BOTRYOID POLLEN 1 regulates ROS-triggered PCD and pollen wall development by controlling UDP-sugar homeostasis in rice

Abstract

Uridine diphosphate (UDP)-sugars are important metabolites involved in the biosynthesis of polysaccharides and may be important signaling molecules. UDP-glucose 4-epimerase (UGE) catalyzes the interconversion between UDP-Glc and UDP-Gal, whose biological function in rice (Oryza sativa) fertility is poorly understood. Here, we identify and characterize the botryoid pollen 1 (bp1) mutant and show that BP1 encodes a UGE that regulates UDP-sugar homeostasis, thereby controlling the development of rice anthers. The loss of BP1 function led to massive accumulation of UDP-Glc and imbalance of other UDP-sugars. We determined that the higher levels of UDP-Glc and its derivatives in bp1 may induce the expression of NADPH oxidase genes, resulting in a premature accumulation of reactive oxygen species (ROS), thereby advancing programmed cell death (PCD) of anther walls but delaying the end of tapetal degradation. The accumulation of UDP-Glc as metabolites resulted in an abnormal degradation of callose, producing an adhesive microspore. Furthermore, the UDP-sugar metabolism pathway is not only involved in the formation of intine but also in the formation of the initial framework for extine. Our results reveal how UDP-sugars regulate anther development and provide new clues for cellular ROS accumulation and PCD triggered by UDP-Glc as a signaling molecule.

UDP-Glc and its derivatives induce the expression of NADPH oxidase genes, inducing ROS-triggered PCD as signaling molecules, but also regulate pollen wall development as metabolites.

Introduction

Sugars, such as glucose and galactose, are essential nutrients involved in energy metabolism, are structural components of the cell, and act as important physiological signaling molecules in both prokaryotes and eukaryotes (Kelly et al. 2012). In plants, monosaccharides such as glucose, galactose, and glucuronide are the major sugar components of the cell wall (Rottmann et al. 2018), which requires enzymatic conversion from sugars to UDP-sugars prior to incorporation into the cell wall via polymerization (Geserick and Tenhaken 2013). There are two major biosynthetic pathways for UDP-sugars: de novo biosynthesis and the recycling pathway (Seifert 2004). In the de novo pathway, uridine diphosphate glucose (UDP-Glc) is synthesized from glucose-1-phosphate as a substrate and UDP-glucose-pyrophosphorylase (UGPase) as catalytic enzyme. Other UDP-sugars are derived from UDP-glucose, which is also the glucose donor for the biosynthesis of many types of oligosaccharides and polysaccharides, as well as an important signal molecule in animal gene regulatory networks (Geserick and Tenhaken 2013). The recycling pathway is catalyzed by sugar-specific kinases and UDP-sugar-pyrophosphorylase and is an important route for the recovery of monosaccharides hydrolyzed from polymers such as cell walls (Bar-Peled and O'Neill 2011).

Sugar metabolism occupies an important position throughout pollen development in rice (O. sativa L.). Callose is a β-1,3-glucan polysaccharide complex synthesized from UDP-glucose and plays an important role in meiosis and the formation of the pollen wall. Indeed, dysfunction of callose metabolism-related genes leads to abnormal anther development (Nishikawa et al. 2005; Chen et al. 2007; Wan et al. 2011; Shi et al. 2015; Zhang et al. 2018; Wang et al. 2020). In the late stage of pollen development, the timely accumulation of starch in pollen is essential for the normal development of pollen grains (Lee et al. 2016). The pollen wall is an important barrier that protects pollen development from environmental interference and is the location for intercellular interactions during pollination (Jia et al. 2015). Abnormal sugar metabolism in the pollen wall may lead to defects in intine deposition and culminate in gametophytic male sterility (Moon et al. 2013).

UDP-glucose 4-epimerase (UGE) is an enzyme that catalyzes the interconversion of UDP-Glc and uridine diphosphate galactose (UDP-Gal). The Arabidopsis (Arabidopsis thaliana) genome encodes five UGEs that all exhibit catalytic capacity in vitro (Barber et al. 2006). The rice genome encodes four UGEs. The expression of OsUGE1 was shown to be induced by low-nitrogen levels; moreover, the leaves of OsUGE1-overexpressing plants accumulate more glucose and galactose in hemicellulose under low-nitrogen conditions (Guevara et al. 2014). Recently, OsUGE2 was cloned and characterized (Liu et al. 2020; Zhang et al. 2020, 2021). In Osuge2 mutants, the shortage of arabinogalactan proteins (AGPs) caused by abnormal glycosylation resulted in the disordered deposition of cellulose. In addition, monogalactose diacylglycerol (MGDG), the main component of the chloroplast membrane, accumulates to much lower levels in Osuge2 mutants, resulting in the collapse of the chloroplast membrane and eventually leading to a brittle stem and zebra leaf phenotypes. The contents of uronic acids, neutral noncellulosic monosaccharides, and cellulose were altered as well in this mutant.

Here, we show that a UGE enzyme plays a vital role in several stages of male reproductive development in rice. Our results also suggest that UDP-Glc, a substrate for UGE, is a potential signaling molecule that regulates anther development in rice. Our findings expand the knowledge of the genetic network underlying the regulation of male fertility in rice and provide a new germplasm resource for hybrid rice breeding.

Results

The bp1 mutant has botryoid pollen and is caused by a mutation in a single gene

We isolated a completely sterile mutant from the transgenic progeny of the rice cultivar Taichung 65 (T65) (Fig. 1, A and B). The anthers of this mutant were smaller and pale yellow in color, with no pollen grain, as observed by optical microscopy following I₂-KI staining, compared to T65 (Fig. 1, C–H). However, we detected abnormal pollen grains when investigating the cytological characteristics of mutant anthers in transverse sections. We used scanning electron microscopy (SEM) to reveal the anther microstructure in T65 and bp1 mutants and observed only slight differences on their outward surfaces (Fig. 1, I and M). After removing the anther wall, we noticed botryoid pollen—resembling a cluster of grapes—that adhered together in the mutant (Fig. 1N). The pollen grains clumped together and were difficult to separate, with a crumpled and rough surface (Fig. 1O). Based on pollen appearance, we named this mutant botryoid pollen 1 (bp1).

Figure 1.

Phenotypic observation and scanning electron microscopy (SEM) analysis of T65 and botryoid pollen 1 (bp1). A) Plant morphology, B) panicle, C–E) spikelets, F–H) pollen stainability. I–P) SEM analysis of anthers and pollen grains in T65 (I–L) and in bp1(M–P). Scale bars, 20 cm (A), 5 cm (B), 1 mm (C–E), 100 μm (F–H), 100 μm (I, M), 20 μm (J, N), 2 μm (K, L, O, P). Het, bp1/BP1 plant.

Notably, homozygous bp1 mutant plants were completely sterile, while bp1-1/BP1 plants had a seed setting rate of 63.9% ± 0.16% (SD) when allowed to self. We also measured pollen fertility of 80 flowers from bp1-1/BP1 plants, and pollen fertility ranged from 0% to 98.0% (Supplemental Fig. S1, A–C). The fertility of different florets of heterozygous plants ranged from 0% to 98% and were mainly distributed at both ends (>70% and <30%), and there were very few florets with a fertility of 50%. Genetic analysis revealed that the segregation ratio of full fertility/lower fertility/fully sterile plants in the F₂ progeny from a bp1-1/BP1 plant is close to 1:2:1 (188:373:164, X₂ = 2.20, n = 2). These results indicate that bp1 is an incompletely dominant mutation in a single gene.

Abnormal degradation of the anther wall in bp1 mutants

To define the abortion period and characteristics of bp1-1 pollen, we investigated anther development at multiple stages in semithin sections (Fig. 2). Compared to T65, the bp1-1 tapetum started to swell and the middle cells were thicker at Stage 8 (S8), but the pollen mother cells in bp1-1 were able to undergo meiosis. At S9, the bp1-1 tapetum continued to thicken, with a clearly visible middle layer and the formation of irregular young microspores that adhered together. At S10, bp1-1 tapetum did not disappear as in T65 but became thicker, pollen had a sparse and irregular cytoplasm, the middle layer was still visible, and there were no large vesicles in the center of microspores. At S11, the bp1-1 tapetum degraded, leaving numerous cell fragments; the epidermis, endothecium, and middle layer were clearly visible and neatly arranged, the middle layer was thicker, and the microspores appeared dry and irregular. At S12, bp1-1 microspores were completely crumpled and aborted, but the middle and inner layers still existed and were thick. These results suggest that BP1 regulates the degradation of rice anther walls.

Figure 2.

Semi-thin cross-section analysis of anthers from T65 and bp1. E, epidermis; En, endothecium; ML, middle layer; Sp, sporogenous cell; T, tapetum layer; MMC, microspore mother cell; MC, meiotic cell; Tds, tetrads; Msp, microspore parietal cell; BP, binuclear pollen; MP, mature pollen; DP, defective pollen. Scale bars, 50 μm.

We used transmission electron microscopy (TEM) to observe the features of anther ultrastructure in greater detail (Fig. 3; Supplemental Fig. S2). At S7, the bp1-1 tapetum appeared vacuolated, and the cell borders were not clearly defined. From S7 to S9, the vacuolation of bp1-1 tapetum increased, with sparse cytoplasmic texture. At S10, the bp1-1 tapetum thickened; from S11 to S12, we observed many Ubisch bodies and a residual densely textured tapetal layer in bp1-1. However, the T65 tapetum completely degenerated into cell fragments, and we observed few Ubisch bodies between the outer wall of microspores and the anther wall, which were loosely connected. These results suggest that the loss of BP1 function leads to abnormal degradation of the tapetum and middle layers.

Figure 3.

Ultrastructure observations of pollen wall in T65 and bp1 from Stage 8 to 12 using transmission electron microscopy (TEM). Tds, tetrads; E, epidermis; En, endothecium; T, tapetum layer; Tn, tapetal nucleus; V, vacuole; Msp, microspore; Pl, proplastid; Mt, mitochondrion; Bp, binuclear pollen; DP, defective pollen; Ub, ubisch body; Gl, Golgiosome; Tc, tectum; Ne, nexine; Ba, bacula; In, Intine; Cy, cytoplasm. The arrows point to the microspore wall. Scale bars, 5 μm (A17–A18); 2 μm (A9–A12); 1 μm (A1–A4, A13–A16); 500 nm (A5–A8; A19–A20).

To investigate the mechanism behind the abnormal degradation of the tapetum and the middle layer in bp1-1, we performed a terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay to monitor the extent of programmed cell death (PCD) in the tapetal and middle layer. In T65, we detected apoptotic signals in tapetum at S8–S10 (Fig. 4A), with the strongest apoptotic signal at S9, which was consistent with the results from transverse sections. In bp1-1 tapetum, we detected strong PCD signals from S5 to S11 (Fig. 4A), often with two independent apoptotic signals in some tapetal cells, suggestive of dividing nuclei. Meanwhile, we observed strong PCD signals in the other anther wall of bp1 mutants from S5 to S11. There were still weak PCD signals from S12 to S13 in the tapetal and middle layer of bp1-1. These results indicate that BP1 is an important homeostat for PCD of the tapetal and the middle layers to accurately regulate the initiation and termination of PCD. Indeed, the loss of BP1 function led to an earlier initiation and a delayed termination of PCD in the tapetal and the middle layers. In addition, we detected PCD signals in the epidermis and endothecium at S5–S7 and S10 in bp1-1, indicating that BP1 plays a key role in PCD of the anther wall.

Figure 4.

PCD and ROS accumulation in anthers from T65 and bp1. A) PCD observation in T65 and bp1 anthers from Stages 4 to 10. The nuclei were stained by propidium iodide (PI) and visualized under confocal laser scanning microscopy. Scale bars, 50 μm. B–D) ROS analysis in anthers of T65 and bp1. B) NBT staining of superoxide anion production in anthers from Stages 4 to 11. Scale bars, 0.5 mm. C) NBT staining of anther sections from Stages 5 to 10. Scale bars, 50 μm. 2P, secondary parietal cell layer; E, epidermis; En, endothecium; ML, middle layer; T, tapetum; Sp, sporogenous cell; MMC, microspore mother cell; MC, meiocyte; Tds, tetrads; Msp, microspore; BP, binuclear pollen; MP, mature pollen; DP, defective pollen; Dy, dyad cell. D) ROS (H₂O₂) contents in anthers by spectrophotometric analysis; *P < 0.05; data are means ± SD, n = 3.

Abnormal formation of the pollen wall in bp1 mutants

We investigated the defects of pollen development in bp1 mutants by TEM (Fig. 3). In agreement with the ultrathin sections above, the primordium walls of bp1-1 microspores were scattered and loose at S8. At S9, the microspores adhered together, and the outer layers of the microspores thickened and formed three rough density lines and irregularly spaced dark granules in bp1-1, which were different from those of T65, with three clear electron-dense lines and regularly spaced dark granules. At S10, the outer wall of microspores had irregularly thickened, nexine and sexine were hard to distinguish, and the outer wall of the microspores adhered together in bp1-1. At S11, the intine began to form in wild type; however, in bp1-1, the outer wall of microspores further thickened, no intine formed, and the microspores were crumpled and adhered together. At S12, we visualized intine on the inner side of nexine in T65, but did not observe any intine in bp1-1. The exines of bp1-1 pollen, which normally take on an I-shape in T65, appeared to grow together in one solid mass. These results suggest that the irregular development of the pollen wall in bp1-1 leads to adhered pollen walls and the formation of pollen that resembles bunches of grapes.

Cloning of BP1

Since we isolated the bp1 mutant from a transgenic line, we first tested whether the phenotype was caused by the T-DNA insertion. Accordingly, we looked for linkage between the T-DNA and the sterility phenotype. Of 102 progenies from the T₀ generation, 24 individuals were completely sterile; importantly, 100 individuals out of the 102 plants contained a T-DNA based on hygromycin resistance, indicating the presence of multiple T-DNA insertions in the bp1 background. Next, we used thermal asymmetric interlaced PCR (TAIL-PCR) to identify the insertion sites (Liu et al. 1995), but this approach was not fruitful, possibly due to the multiple insertions affecting the efficiency of PCR amplification. We thus used map-based cloning by crossing bp1 to the indica-type rice cultivar Huanghuazhan (HHZ) and narrowed down the location of the bp1 mutation between molecular markers 528897 and 529989 within a 1.1-Mb interval on chromosome 5 (Fig. 5A). By coincidence, we had established that the progenies of F₂ line 45 show a segregation ratio for the presence/absence of the T-DNA close to 3:1; the T-DNA was also tightly linked to the sterility phenotype, suggesting that this line contains a single T-DNA responsible for male sterility. TAIL-PCR revealed one T-DNA insertion in the 5′ untranslated region (5′ UTR) of Os05g0595100 (Fig. 5B). Complementation tests with a construct encompassing the genomic region of Os05g0595100 from T65 restored pollen fertility of bp1 mutants (Fig. 5C). Conversely, we generated knockout lines for Os05g0595100 using clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated nuclease 9 (Cas9)-mediated gene editing (Ma et al. 2015; Zhou et al. 2016) and determined that these mutants recapitulate the bp1 phenotype (Fig. 5, D and E). These results demonstrate that Os05g0595100 is the gene whose loss of function results in male sterility in bp1.

Figure 5.

The cloning and functional identification of BP1. A) Primary mapping of BP1 using a segregation F₂ population derived from a cross between bp1 and Huanghuazhan (HHZ). B) Determination of the T-DNA insertion site by TAIL-PCR. C) Functional complementation analysis of bp1-1. The full-length BP1 gene from wild type was transformed into bp1-1 (T-DNA insertion mutant). The resulting transgenic plants are labeled BP1FC. D) Morphology of T₀ transgenic plants. E) Genomic structure of T0 transgenic plants. Scale bars, 20 cm (C1, D1), 1 mm (C2–4, D2–4), 100 μm (C5–7, D5–7).

As with the original bp1 mutant allele (bp1-1 thereafter), we investigated the setting rate of progeny from heterozygous plants for the gene-edited allele bp1-2. The segregation ratio of full fertility/lower fertility/fully sterile plants in the F₂ generation of bp1-2/BP1 heterozygous plants was 1:2:1 (14:43:17, X₂ = 2.19, n = 2), and the average setting rate of bp1-2/BP1 plants was 65.5% ± 26%, which was consistent with the results of bp1-1/BP1 plants above (Supplemental Fig. S1D). Further, we inspected semithin sections of anthers from bp1-2/BP1 plants at different stages (Supplemental Fig. S3). No obvious abnormalities were observed in these bp1-2/BP1 plants before S7. Importantly, starting at S7, locules on different florets or the four locules of the same anther in bp1-2/BP1 plants showed inconsistent development. At S9, two newly separated microspores showed inconsistent staining, with one microspore entering the S9 stage and the other apparently remaining in the S8 stage. At S10, the degradation of tapetum was slow in bp1-2/BP1 anthers. At S11, some of the microspores in bp1-2/BP1 plants appeared thin, at a time when the microspores in wild type would form a crescent shape. At S12, some microspores were not filled with starch grains, and some microspores were very thin, and even crumpled. We also investigated the development of pollen wall in bp1-2/BP1 plants in ultrathin sections (Supplemental Fig. S4). In microspores with normal fertility, we observed the formation of intine from S11 that became distinct at S12. By contrast, in abortive microspores, we observed no intine. These results indicate that the pollen abortion characteristics of bp1/BP1 plants are similar to those of bp1 homozygous plants, although asynchronous development of pollen mother cells and microspores and its severity was much lower, suggesting a dosage effect of BP1.

BP1 phylogeny and expression analysis of BP1

BP1 encodes UDP-glucose 4-epimerase 1 (UGE1), containing 355 amino acids. To assess the conservation of BP1, we identified 16 UGEs from Arabidopsis, rice, barley (Hordeum vulgare), pea (Pisum sativum), turnip (Brassica rapa), fission yeast (Schizosaccharomyces pombe), and human (Homo sapiens) by BLAST at the National Center for Biotechnology Information using the full-length protein sequence of BP1 as query. The phylogeny tree using the above 16 UGE proteins identified three clades. BP1 clustered with UGE1, UGE2, and UGE4 from monocots and with UGE2, UGE4, and UGE5 from dicots. BP1 showed the highest percentage identity with barley UGE1, with 89.3%. Notably, UGE1 from dicot species formed a separate clade with UGE3 from dicots and monocots. UGEs from fission yeast and human were more distant (Supplemental Fig. S5).

Since BP1 regulated pollen development of rice, we employed reverse transcription quantitative PCR (RT-qPCR) to analyze the expression pattern of BP1 in spikelets at different developmental stages. BP1 was expressed throughout anther development and increased gradually until S9, followed by a dramatic decline at S11 onward in T65 (Fig. 6A). BP1 was expressed to much lower levels in bp1-1 relative to T65 from S4 to S10, which was consistent with the position of the T-DNA responsible for the male sterility of this mutant (Supplemental Fig. S6). In situ hybridization in T65 anthers showed that BP1 transcripts accumulate to high levels in the four layers of the anther wall and pollen mother cells from S4 to S7, in the tapetum and tetrad at S8, and in tapetum and microspore from S9 to S10 as well (Fig. 6B). These results were consistent with the defects observed for PCD of the four layers of the anther wall and the formation of the pollen wall in bp1-1, suggesting that BP1 plays an important role in all stages of anther development. We also investigated the subcellular localization of BP1 by generating a construct encoding a fusion protein between BP1 and green fluorescent protein (GFP). We transiently transfected rice protoplasts with the resulting construct (pYL322-BP1-GFP), using the empty vector pYL322-GFP as a control. We detected green fluorescence in the cytoplasm and the nucleus for free GFP and BP1-GFP, indicating that BP1 is widely distributed throughout the cell (Fig. 6C). To independently confirm the subcellular localization of BP1, we performed a subcellular fractionation assay followed by immunoblot analysis, separating total proteins extracted from the leaves of transgenic BP1-Flag seedlings into cytosolic and nuclear fractions. We detected a signal for BP1 in the nuclear and cytosolic fractions, with no clear contamination between cytoplasmic and nuclear proteins, as determined by the pattern seen for the cytosolic marker RNAse Z^S1 and the nuclear marker histone H4 (Fig. 6D). This result further corroborated that BP1 localizes in the cytoplasm and nucleus.

Figure 6.

Analysis of BP1 expression. A) RT-qPCR analysis of BP1 transcript levels in T65 anthers from Stages 1 to 14 (S1–S14); Data are means ± SD, n = 3. B) In situ hybridization of BP1 transcripts in T65 anthers from S4 to S11. 2P, secondary parietal cell layer; E, epidermis; En, endothecium; ML, middle layer; T, tapetum; Sp, sporogenous cell; MMC, microspore mother cell; MC, meiocyte; Tds, tetrads; Msp, microspore; BP, binuclear pollen. Scale bars, 50 μm. C) Subcellular localization of BP1-GFP in rice protoplasts. The nucleus was stained with DAPI. Scale bars, 20 μm. D) Subcellular localization of BP1 examined by immunoblot analysis of subcellular fractions. The nuclear and cytoplasmic fractions were extracted from leaves of transgenic BP1pro:BP1-Flag plants, with RNase Z^S1 and histone H4 used as markers for the cytoplasm and nucleus respectively. CP, cytoplasmic proteins; NP, nuclear proteins; Total, Total protein.

BP1 catalyzes the reversible conversion of UDP-glucose and UDP-galactose

As BP1 is predicted to be a UGE, we tested the ability of recombinant BP1 to catalyze the reversible conversion between UDP-glucose and UDP-galactose in vitro. Indeed, we detected both UDP-glucose and UDP-galactose in the enzymatic reactions when adding only UDP-glucose (Fig. 7A2) or only UDP-galactose (Fig. 7A3) as substrate, in the presence of recombinant BP1 for 10 min. We concluded that BP1 can reversibly convert UDP-glucose to UDP-galactose. The K_m value of BP1 was 7.58 mM with UDP-glucose as the substrate, and 90.1 mM with UDP-galactose as the substrate, indicating that BP1 has a greater affinity for UDP-glucose than UDP-galactose (Fig. 7A4–5). Moreover, mutated BP1 with amino acid transitions in functional sites showed little enzymatic activity (Supplemental Fig. S7).

Figure 7.

Enzymatic activity of BP1 in vitro and sugar contents of anthers in vivo. A) Analysis of BP1 enzyme activity in vitro by HPLC. (A1), control reaction consisting of only BP1; (A2), reaction with the addition of 1.5 mM UDP-Glc as substrate; (A3), reaction with the addition of 1.5 mM UDP-Gal as substrate. (A4), Enzymatic kinetic parameters calculated by the double reciprocal method with UDP-Glc as the substrate (R² = 0.998). (A5), Enzymatic kinetic parameters calculated by the double reciprocal method with UDP-Gal as the substrate (R² = 0.998). B) Sugar levels from anthers of T65 and bp1-1 at Stage 9. (B1), Quantitative analysis of free nucleotide sugar by LC–MS/MS. (B2), Quantitative analysis of free sugar by UHPLC-HRMS. Data are means ± SD, n = 6. *, P < 0.05; **, P < 0.01; ***, P < 0.001, two-tailed Student's t-test.

Since BP1 can catalyze the reversible conversion of UDP-glucose and UDP-galactose in vitro, we wondered about the effect of the loss of UGE activity on the content of nucleotide sugars in vivo. We thus measured the contents of free nucleotide sugars from S9 anthers in T65 and bp1-1 by liquid chromatography tandem mass spectrometry (LC-MS/MS; Fig. 7B1). Compared to T65, the contents of UDP-Glc, UDP-Gal, UDP-GalA, UDP-GlcA, UDP-GlcNAc, UDP-Arap, GDP-Man, GDP-Fuc, and ADP-Glc were significantly higher in bp1-1, with the highest content for UDP-Glc in bp1-1 and the largest fold increase, 20.4-fold higher than T65. Similarly, the content for UDP-Gal in bp1-1 was 2.8-fold higher relative to T65. The content for UDP-GlcA in bp1-1 was 2.6-fold higher than T65 levels, and the content of UDP-GlcNAc was 1.8-fold higher compared to T65. Only the content for UDP-Araf was significantly lower in bp1-1 compared to T65, while UDP-Rha and UDP-Xyl showed no significant difference in their abundance between T65 and bp1-1. These results were consistent with the in vitro BP1 activity assay, suggesting that the loss of BP1 function might affect the conversion of UDP-Glc to UDP-Gal and lead to an increase in free UDP-Glc.

We also investigated possible differences in the metabolism of other monosaccharides in bp1-1 and T65 anthers by measuring the contents of glucose, galactose, arabinose, xylose, apiose, fucose, and rhamnose by ultraperformance liquid chromatography–high-resolution mass spectrometry (UHPLC–HRMS) (Fig. 7B2). We determined that the contents of galactose and arabinose in bp1-1 are significantly lower compared to T65, with a 36.3% decrease in arabinose, while the contents of UDP-Gal and UDP-Arab (presented as the isomers UDP-Araf and UDP-Arap) were higher in bp1-1, indicating that monosaccharide metabolism in anthers is abnormal in the mutant.

UDP-Glc affects ROS accumulation and PCD in anthers

UDP-Glc and its derivatives were previously advanced to act as signaling molecules, whose elevated content might induce the production of reactive oxygen species (ROS), which are important signals for PCD initiation (Xie et al. 2014; Yu et al. 2017). Given the earlier onset of PCD in bp1-1 tapetum and the higher levels of UDP-Glc in the mutant, we wished to look at ROS accumulation. To this end, we stained T65 and bp1-1 anthers with Nitrotetrazolium Blue chloride (NBT; Fig. 4B). As the anthers developed, we did not observe staining in T65 anthers between S4 and S6, but we noticed a light blue color from S7 onward, indicative of the initiation of ROS production at this stage. As the anthers developed, ROS continued to accumulate, reaching the darkest staining at S9 and gradually becoming lighter after S9, which was consistent with the tapetal PCD. In bp1-1, we detected staining in anthers as early as S4, with very intense staining from S4 to S6, becoming lighter at S7, and gradually increasing again from S8 to S10. These fluctuations in ROS levels in bp1-1 were consistent with the PCD signals from the TUNEL assay. Observations of paraffin sections confirmed the staining of tapetum and microspores in bp1-1 (Fig. 4C). We also measured hydrogen peroxide levels in T65 and bp1-1 anthers (Fig. 4D) using a hydrogen peroxide assay kit (Beyotime, Shanghai, China). We observed a greater accumulation of hydrogen peroxide at S5, S6, and S10 in bp1-1 relative to T65, suggesting that the accumulation of UDP-Glc induces abnormal ROS accumulation, which in turn affected the normal progress of PCD in the tapetum.

UDP-sugar homeostasis affects callose degradation and pollen wall formation

Callose is a β-1,3-glucan polysaccharide complex synthesized from UDP-glucose and catalyzed by β-1,3-glucan synthase. Callose is widely found in angiosperms and is deposited at different plant developmental stages (Nishikawa et al. 2005). During anther development in rice, callose biosynthesis and deposition around the pollen mother cell begin at S6. At S9, callose is degraded and microspores are released from the tetrads (Li and Zhang 2010). An abnormal timing of callose biosynthesis and degradation will lead to abnormal pollen development and its eventual abortion (Chen et al. 2007; Wan et al. 2011; Shi et al. 2015; Wang et al. 2020). As bp1-1 accumulated more UDP-glucose, we stained anther cross sections from T65 and bp1-1 plants with aniline blue to detect callose from S7 to S9 (Fig. 8A). We detected a strong signal around the pollen mother cells in T65 and bp1-1 at S7. The area of fluorescence became smaller in bp1-1 at S8. When reaching S9, callose was almost undetectable around T65 microspores, whereas a hazy fluorescence signal remained around bp1-1 microspores, which was similar to the phenotype of the Osg1 mutant with loss of β-1,3-glucanase1 function (Wan et al. 2011). These results suggest that the delayed degradation of callose in bp1-1 may prevent the release of microspores, leading to adhered pollen walls.

Figure 8.

Callose observation and immunofluorescence analysis of sugar components in anthers. A) Callose observation in T65 and bp1-1 from S7 to S9. MC, meiocyte; Tds, tetrads; Msp, microspore; T, tapetal cell; S7–S8, meiosis stage; S9, Mononuclear microspore stage; Scale bars, 25 μm. B) Analysis of cell wall polysaccharide contents in rice anthers by immunofluorescence. JIM13 binds to Glca-β (1, 3)-Gala-α (1, 2) Rha of AGPs; LM5 recognizes (1, 4)-β-D-galactan side chains of pectic rhamnogalacturonan I; CCRC-M7 recognizes an arabinosylated (1, 6)-β-D-galactan of arabinogalactan II; CCRC-M1 binds to a-L-fucosyl (1, 2)-β-D-galactosyl side chains of xyloglucan; LM10 is an antibody against (1, 4)-β-D-xylan; and LM2 recognizes a β-D-glucuronosyl residue on AGP. E, epidermis; T, tapetum; Msp, microspore. Scale bars, 50 μm.

UDP-sugars are important donors for the biosynthesis of the cell wall (Verbancic et al. 2018). Since the inspection of thin sections indicated an abnormal development of the pollen wall in bp1-1, we explored the possible changes of sugar contents in the pollen wall by immunohistochemistry with specific monoclonal antibodies against six types of sugars. As shown in Fig. 8B, we detected weak signals in T65 but strong signals in bp1-1 for the antibodies JIM13 (recognizing the trisaccharide GlcA-β (1-3)-GalA-α (1-2) Rha on AGPs) (Yates et al. 1996), LM5 (recognizing (1→4)-β-D-galactan side chains of pectic rhamnogalacturonan I) (Jones et al. 1997), LM10 (recognizing (1→4)-β-D-Xylan), and CCRC-M1 (recognizing L-fucosyl (1→2)-β-D-galactosyl side chains of xyloglucan; Puhlmann et al. 1994), especially in the microspore wall and the anther outer layer. We observed signal for CCRC-M7 and LM2, recognizing (1,6)-β-D-galactose of arabinogalactan II (Puhlmann et al. 1994) and a β-D-glucuronosyl residue (Yates et al. 1996) on AGPs, respectively, in the microspore wall and the anther outer layer in T65, but only weakly in bp1-1. Although the JIM13, CCRC-M7, and LM2 antibodies all recognized AGPs in plant cell walls, the JIM13 signal was very strong in bp1-1, while the signals of CCRC-M7 and LM2 were more obvious in T65, indicating that the biosynthesis of the AGP epitopes recognized by JIM13, CCRC-M7, and LM2 change in their abundance, probably due to disordered sugar metabolism in bp1-1. AGPs, containing protein backbones and arabinogalactan and galactan side chains, may function as signaling molecules in the information exchange between tapetal cells and microspores (Kawaguchi and Shibuya 2000). These results indicate that UDP-sugars are components of the pollen wall and that bp1-1 perturbs the incorporation of UDP-sugars in the pollen wall.

BP1 affects the expression of genes involved in ROS generation and clearance

The proper accumulation of ROS is necessary for the initiation of tapetal PCD. NADPH oxidases from the RESPIRATORY BURST OXIDASE HOMOLOG (RBOH) family play a key role in ROS production in plants (Luo et al. 2013; Huang et al. 2016). To identify the cause for the elevated ROS levels in bp1 mutants, we examined the expression levels of RBOH family members in bp1-1. We established that OsRBOHb, OsRBOHc, OsRBOHd, OsRBOHe, and OsRBOHh (Doyle et al. 2010) are upregulated at S4–S6 in bp1-1 anthers (Supplemental Fig. S8, A–I). These results suggested that the elevated expression of RBOH genes leads to the early accumulation of ROS in bp1-1 anthers, which may act as a signaling molecule to regulate PCD in tapetum. We also characterized the expression levels of DITERPENE CYCLASE 1 (OsDTC1) (Yi et al. 2016) and METALLOTHIONEIN 2b (OsMT2b) (Ren and Zhao 2009) in bp1-1, as they are dynamic ROS homeostasis-related genes. OsDTC1 and OsMT2b expression levels were lower in bp1-1 relative to T65 at different stages of anther development; in particular, OsMT2b was significantly downregulated from S5 to S9, suggesting that the excessive accumulation of ROS in bp1-1 might also be associated with the decreased levels of the ROS scavenger-encoding gene OsMT2b. However, DEGENERATED PANICLE AND PARTIAL STERILITY 1 (DPS1) (Zafar et al. 2020), encoding a protein that interacts with mitochondrial thioredoxin, another ROS scavenger, was expressed at higher levels in bp1-1 compared to T65 in the early stage of anther development, suggesting that DPS1 expression is induced due to excessive ROS accumulation in bp1-1. MADS3 is a key transcriptional regulator involved in modulating ROS levels through OsMT-1-4b. Notably, MADS3 was significantly upregulated in bp1-1 at S9, but downregulated at S10 probably due to negative feedback (Supplemental Fig. S8). ARGAUNOTE 2 (OsAGO2) is another important gene involved in the dynamic balance regulation of ROS in anthers and PCD development, which regulates pollen development (Zheng et al. 2019). OsAGO2 expression was upregulated during the pollen mother cell stage (S4–S6) and significantly downregulated after S9, suggesting that the expression of ROS homeostasis-regulated genes is affected in bp1-1.

BP1 affects the expression of genes involved in anther, callose, and pollen wall development

To further elucidate the function of BP1 in anther development, we analyzed the expression levels of key regulatory genes in anthers for tapetal development at S4–S10 in T65 and bp1-1 plants by RT-qPCR (Supplemental Fig. S9). The expression levels of OsEAT1 (ETERNAL TAPETUM 1) (Niu et al. 2013), OsPTC1 (PERSISTENT TAPETAL CELL 1) (Li et al. 2011), OsAP25 (ASPARTIC PROTEASE 25) (Niu et al. 2013), OsAP37 (Niu et al. 2013), OsTGA10 (TGA MOTIF-BINDING PROTEIN 10) (Yang et al. 2016), GAMYB (GIBBERELLIC ACID-INDUCED MYB) (Liu et al. 2010), PTC2 (Uzair et al. 2020), and OsMYB80/OsMYB103 (Pan et al. 2020) were significantly lower in bp1-1 than in T65 at the late stages of anther development, while the expression level of TIP3 (TDR INTERACTING PROTEIN 3) (Yang et al. 2019b) was significantly downregulated at S5–S9 in bp1-1. Likewise, the expression level of OsTDL1A (TAPETAL DETERMINANT1-LIKE 1A) (Zhao et al. 2008) was significantly lower at S4 in bp1-1 compared to T65. Importantly, loss-of-function mutations in the above genes all display delayed tapetal degradation. This observation suggest that the downregulation of these genes functioning at late stages of anther development may lead to a delayed end of degradation and swelling of tapetum in bp1-1.

Notably, OsUGP1 (UDP-GLUCOSE PYROPHOSPHORYLASE 1), OsUGP2, and Osg1 expression levels were significantly altered in the bp1-1 mutant in different directions and at different stages relative to T65 (Supplemental Fig. S10, A–C). Their encoded proteins regulate callose biosynthesis and degradation. The downregulation of Osg1 in the late stage of anther development might account for the delayed degradation of callose and the adhered microspores.

The expression levels of CYP704B2 (CYTOCHROME P450 704B2) (Li et al. 2010), OsUGP1 (Chen et al. 2007), WDA1 (WAX-DEFICIENT ANTHER 1) (Jung et al. 2006), CAP1 (COLLAPSED ABNORMAL POLLEN 1) (Ueda et al. 2013), CYP703A3 (Yang et al. 2014), OsABCG15 (ATP-BINDING CASSETTE G15) (Wu et al. 2014), and DPW (DEFECTIVE POLLEN WALL) (Shi et al. 2011), related genes that are involved in pollen wall formation, were significantly lower in bp1-1 compared to T65 at the late stages of anther development (Supplemental Fig. S10, A–I). Their encoded proteins are mainly involved in fatty acid metabolism, biosynthesis, and transport of sporopollenin precursors in extine, which mainly functions at the late stage of pollen development. The significant downregulation of these genes was consistent with the abnormal pollen wall development in bp1-1.

Discussion

UGE regulates homeostasis of UDP-sugars and pollen sterility

Hybrid rice has greatly contributed to solving the food shortage problem in developing countries. Male-sterile mutants are important germplasm resources in hybrid breeding (Zhou et al. 2012, 2014; Chen et al. 2020). In this study, we isolated the male-sterile bp1 mutants with botryoid and adhered pollen from the progeny of a transgenic rice line. Unlike the previously reported sterile genes cloned from rice, which affect a certain developmental stage or regulate the formation of a given structure, we established here that BP1 encoding a UGE functions throughout pollen development by controlling UDP-sugar homeostasis, ROS accumulation, PCD of the anther wall, callose degradation, and the development of the pollen wall. In Arabidopsis, UGE3 is specialized for pollen development, likely due to its high expression in anther (Barber et al. 2006). The mutation of OsUGE2 in the brittle culm and zebra leaf 1 (bz1, also named fragile culm 24 [cf24]) caused changes in the contents of UDP-Glc and UDP-Gal, but did not cause pollen abortion, as the expression of this gene is quite low in anthers (Liu et al. 2020; Zhang et al. 2020), suggesting that UGEs from different species share the same biochemical function regardless of the extent of sequence similarity, with their biological function mainly depending on the expression pattern of their encoding genes.

Homozygous bp1 plants were completely male-sterile, whereas bp1/BP1 plants showed a pollen fertility ranging from 0% to 98% mainly distributed at both ends (>70% and <30%), and there were very few florets with fertility around 50%. Heterozygous mutant individuals in sporophytic genes are generally fully fertile, while heterozygous mutant individuals in gametophytic genes are generally semi-sterile with a fertility rate about 50% (Chen and Liu 2014; Zou et al. 2017). This notion is thus inconsistent with our observation of the fertility of bp1/BP1 individuals, suggesting that BP1 is not a gametophyte fertility gene. Sporophytic genes are usually expressed in the tapetum or the microspore mother cell, while gametophytic genes are specifically expressed in microspores. In situ hybridization showed that BP1 is highly expressed in the tapetum and pollen mother cells but was also expressed in the microspores at S9, indicating that BP1 mainly functions in the sporophyte. The accumulated BP1 in pollen mother cells may be unequally transmitted to the microspores and therefore may affect the development of certain microspores in bp1/BP1 plants. In this study, we observed that ROS accumulation is very high at S4, thus prior to PCD of the tapetum starting at S5, also suggesting that BP1 is a sporophytic gene. Therefore, we consider BP1 as a sporophytic gene.

Although previous studies had demonstrated that UGEs catalyze the interconversion of UDP-Glc to UDP-Gal in vitro, a change in the contents of free UDP-sugars in vivo has not been reported in plants following the loss of UGE function (Rösti et al. 2007; Kim et al. 2009; Guevara et al. 2014; Liu et al. 2020; Zhang et al. 2020). In this study, we determined that the contents of several UDP-sugars increase in bp1-1 relative to T65, with UDP-Glc experiencing the greatest increase of 20.4-fold. The de novo pathway and the recycling pathway are the two routes for UDP-sugar biosynthesis (Bar-Peled and O'Neill 2011). The biosynthesis of UDP-Glc derivatives through the de novo pathway was blocked in bp1-1, resulting in increased UDP-Glc contents. However, this rise in UDP-Glc levels may have accelerated the biosynthesis of UDP-sugars through the recycling pathway. Indeed, we measured lower contents for galactose and arabinose acting as donors in the recycling pathway in bp1-1. Furthermore, the Imbalance of UDP-sugars might also affect the biosynthesis of free UDP-sugars into polysaccharide complexes, yielding the higher levels of multiple UDP-sugars seen in bp1-1.

The imbalance of UDP-sugar contents in bp1-1 led to defects in anthers and pollen at various developmental stages starting at S4. UDP-Glc is an important metabolite, but may also act as a signaling molecule (Janse van Rensburg and Van den Ende 2017). As a metabolite, UDP-Glc may affect the degradation of callose, the enrichment of pollen starch, and the formation of the pollen wall. As a potential signaling molecule, UDP-Glc may induce the production of ROS and thus affect tapetal PCD. Since BP1 functions at multiple developmental stages resulting in anther defects and, thus, complete male sterility, BP1 may be a potentially useful sterility gene in third-generation hybrid rice using a recessive nuclear male sterile line (Chang et al. 2016; Liao et al. 2021), which could overcome the breeding disadvantages of first- and second-generation hybrid rice technologies using cytoplasmic and photo-thermosensitive genic male sterile lines, respectively (Zhou et al. 2012, 2014, 2016; Chen et al. 2020).

Disorder of UDP-glucose metabolism leads to abnormal PCD in the anther wall

In animals, UDP-Glc has been suggested to be a signaling molecule that stimulates multiple signaling pathways, such as mitogen-activated protein kinase signaling, by interacting with G-protein-coupled receptors (Janse van Rensburg and Van den Ende 2017). The status of UDP-Glc as a signaling molecule is uncertain in plants, although several studies have demonstrated that UDP-Glc and its derivatives can induce cellular PCD. We reported previously that overexpression of OsHXK1 (HEXOKINASE 1), whose encoded protein catalyzes the biosynthesis of Glc-6-phosphate using glucose, leads to ROS accumulation and early PCD in the tapetum (Zheng et al. 2019). We propose that the OsHXK1 overexpression phenotype may be caused by an increase in UDP-Glc resulting from an accumulation of Glc-6-phosphate.

Tapetal development and PCD strongly correlate with the changes in ROS levels in anthers (Hu et al. 2011; Luo et al. 2013; Zafar et al. 2020). In T65 anthers, ROS levels remained relatively low before S7, during which tapetum develops to maturity. From S7 onward, ROS contents started to accumulate, subsequently initiating tapetal PCD. By S10, when tapetal PCD typically terminates, ROS contents returned to a low level. By contrast, bp1-1 anthers showed ROS accumulation starting at S4, with tapetal PCD initiating at S5.

We speculate that the accumulated UDP-Glc in bp1-1 may act as a signaling molecule that interacts with an unknown receptor to activate MAP kinase signaling, similar to its proposed role in animals (Janse van Rensburg and Van den Ende 2017). MAP kinase cascades have been reported to be activated when ROS accumulate in plants (Lee et al. 2020). The analysis of expression from marker genes (RBOH, DTC1, OsAGO2, OsMADS3; Hu et al. 2011; Yi et al. 2016; Zheng et al. 2019) involved in ROS accumulation and anther development also showed a significant change, indicating that the mutation of BP1 did affect the expression of ROS homeostasis-related genes in bp1-1, which in turn caused an imbalance in ROS levels, affecting the development of anthers. The RBOH genes encoding NADPH oxidases that are responsible for much of intracellular ROS biosynthesis were upregulated in bp1-1 at an early stage of anther development. In the bp1-1 mutant, the accumulation of UDP-Glc may activate the MAP kinase signaling pathway, thus promoting the expression of RBOH genes. These results suggest that UDP-Glc in plants may act as a signaling molecule, regulating the triggering of premature PCD.

However, the swollen tapetum and PCD signal persisted until S13 in bp1-1, indicating a delayed end of apoptosis. This phenotype may be caused by lower expression of genes (TDR INTERACTING PROTEIN 2 [TIP2], TIP3, EAT1, TGA10, APETALA2/ETHYLENE RESPONSE FACTOR 37 [OsAP37], OsMYB80/OsMYB103) associated with tapetal PCD at a later stage of anther development in bp1-1, compared to the wild type (Zhao et al. 2008; Niu et al. 2013; Ko et al. 2014; Yang et al. 2019b; Pan et al. 2020).

Abnormal degradation of callose leads to adhesion of pollen wall

Callose is a β-1,3-glucan complex synthesized by callose synthase located on the plasma membrane of pollen mother cells using UDP-Glc as its building block and serves as a temporary cell wall that separates microspore mother cells (Somashekar et al. 2022). When meiosis is completed, callose is degraded and microspores are released. The proper formation and degradation of callose is therefore critical to the proper development of the pollen wall. OsGSL5 (GLUCAN SYNTHASE-LIKE 5), OsUGP1, and Osg1 are important genes regulating callose biosynthesis and its timely degradation (Chen et al. 2007; Wan et al. 2011; Shi et al. 2015). For instance, the UDP-glucose pyrophosphorylase OsUGP1 catalyzes the reversible conversion of Glc-1-phosphate and UTP to UDP-Glc and pyrophosphate, with UDP-Glc being essential for callose deposition during pollen mother cell meiosis. OsGSL5 is a callose synthase critical for pollen development, as OsGSL5 knockdown lines have a defective callose wall. Osg1 is a glucanase and plays a vital role in timely callose degradation. Importantly, all three proteins are involved in UDP-sugar metabolism, indicating that abnormal UDP-Glc metabolism directly affects the biosynthesis and degradation of callose, which in turn affects rice fertility.

In bp1-1, UDP-Glc content increased significantly at S9 and callose degradation was delayed, resulting in an abnormal development of the primary cell wall that likely accounted for the adhered microspores and their difficulty in separating from one another. Our results suggest that BP1 may play a regulatory role in the biosynthesis and degradation of callose via its interconversion of UDP-Gal and UDP-Glc, the callose building block. Moreover, the expression pattern of other callose-related genes was affected in the bp1-1 mutant: OsUGP1 expression decreased in bp1-1 from S7 to S9, likely a result of product inhibition. Excessive UDP-Glc accumulation in bp1-1 also lowered the expression of Osg1, possibly explaining the delayed callose degradation seen in the mutant, resulting in the impaired release of microspores and formation of the pollen wall (Fig. 9).

Figure 9.

A working model illustrating how BP1 regulates male fertility in rice. BP1 catalyzes the interconversion between UDP-Glc and UDP-Gal. The loss of BP1 function leads to massive accumulation of UDP-Glc. The imbalance of UDP-sugars not only induces the expression of RBOH family genes, which results in ROS accumulation and thereby eliciting the early PCD of anther walls but also causes a slow degradation of callose and deficiency of intine and extine, leading to an adhesive microspore and pollen abortion. CW, callose wall; PM, plasma membrane; Pe, primexine; Pb, probacula; Ne, nexine; Ba, baculum; Ty, tryphine; In, intine.

Sugars define the framework for pollen wall formation

Despite variation in pollen wall morphology between species, its fundamental structure consists mainly of three layers: exine, intine, and tryphine. Exine is commonly composed of two layers: outer sexine and inner nexine, forming an I-shaped structure (Ariizumi and Toriyama 2011). The main component of exine is sporopollenin, a complex macromolecule made up of very-long-chain fatty acids, aromatic compounds, and phenyl propionic acid compounds. Intine is composed of hydrolytic enzymes, hydrophobic proteins, cellulose, hemicellulose, and pectic polymers (Ariizumi and Toriyama 2011). Various types of defects in the pollen wall result in pollen sterility (Fu et al. 2014; Ko et al. 2014; Yi et al. 2016; Yang et al. 2019a). We also observed severe defects in the pollen wall of the bp1-1 mutant. Ultrathin sections showed that intine in bp1-1 is missing, and exine is abnormal from the primary wall stage onward; the pollen wall of individual grains adhered to one another, leaving a disordered structure instead of the typical complex multiple-layer outer surface of pollen (Fig. 9).

Previous studies have demonstrated that sugars are associated with the formation of intine, as well as the biosynthesis of exine, by affecting the formation and degradation of callose (Chen et al. 2007; Mu et al. 2009; Wan et al. 2011; Shi et al. 2015; Wang et al. 2020). Sugars and their derivatives are vital structural components of the plant cell wall (Barnes and Anderson 2018; Verbancic et al. 2018). However, the sugar components of exine have rarely been explored in rice, although we hypothesized that sugar metabolism would also be important for exine formation (Sumiyoshi et al. 2015; Lee et al. 2016). In this study, immunofluorescence assays with antibodies recognizing specific sugars revealed the presence of sugars in exine. In addition, the polysaccharide contents of exine changed noticeably in bp1-1 compared to T65, with the contents of pectic rhamnogalacturonan I, xylan, and xyloglucan increasing, similar to the results in a previous publication looking at cell wall composition (Rösti et al. 2007). We showed that the loss of BP1 function caused abnormal metabolism of UDP-sugars and monosaccharides, which may explain the above results. For example, the anther of bp1-1 mutant was labeled by CCRC-M1 specific for fucosylated xyloglucan, whereas T65 was not. The analysis of free nucleotide sugars from anther of T65 and bp1-1 at S9 (Fig. 7) showed that the content of UDP-Glc in bp1-1 anthers increased significantly, accounting for the increased content of GDP-Fuc, which is an important donor for the biosynthesis of fucosylated xyloglucan and accumulates to significantly higher levels than the wild type, leading to an increased content of fucosylated xyloglucan in the cell wall (Kleczkowski et al. 2010; Verbancic et al. 2018).

Observations of ultrathin sections indicated that exine in bp1-1 is irregular and thick with no clear separation between nexine and sexine, compared to the normal complex I-shaped pollen wall in T65. At S8, we observed a layer of fine electron-dense lines on the primexine of microspores in T65, while we detected only a thin and scattered layer in bp1-1. Primexine is a microfibrillar matrix that acts as a fine template for the deposition and subsequent polymerization of sporopollenin precursors (Ariizumi and Toriyama 2011). At S9, a thin tectum had already formed on the T65 microspore wall, as well as the typical I-shaped structure. By contrast, the microspore wall of bp1-1 retained a loose granular structure (Fig. 9). We therefore propose that sugars may act as a framework structure for the formation of the pollen wall. In the bp1-1 mutant, the disordered metabolism of UDP-sugars or monosaccharides disrupted this framework, thus preventing the normal deposition of sporopollenin, resulting in the abnormal structure of the entire pollen wall. The above results indicate that BP1 may play an important role in the early stages of pollen wall formation. Moreover, UDP-sugars may participate in the formation of the initial framework for pollen wall development as metabolites.

Materials and methods

Plant materials

All rice varieties were planted either in the field of South China Agricultural University (23° N, 113° E) or in phytotrons. The growing conditions were 20–35°C most of the time and 50% to 90% relative humidity (RH) in the field. The day length was 11–13 h, the average temperature was 25-28°C, and the RH was 45–70% in phytotrons. The bp1 (bp1-1, T-DNA insertion allele) mutant is derived from a transgenic plant generated in the Taichung65 (T65, O. sativa L. ssp. japonica) background. F₂ mapping populations were constructed from a cross between the bp1-1 mutant and Huanghuazhan (HHZ).

Characterization of the phenotype

The images of mature rice plants were captured with a Canon 550D digital camera. Anthers were analyzed with an Olympus SZX10 dissecting microscope. The pollen grains were treated with 1% (w/v) I₂-KI staining solution before being photographed with a Revolve FL microscope (Echo, San Diego, CA, USA). Anther sections were prepared according to previous publications (Zhou et al. 2014; Chen et al. 2020) and observed by transmission electron microscopy (TEM). The semi-thin sections of anthers from 102 flowers of bp1-2/BP1 plants were observed, the number of observed flowers at different developmental stages ranged from 7 to 23. The micro-structure of anthers was imaged by SEM as previously described (Chen et al. 2020). First, the material was fixed with 4% (w/v) paraformaldehyde and 0.25% (w/v) glutaraldehyde in phosphate-buffered saline (PBS) (0.1 M, pH 7.0) overnight at 4°C, rinsed with 0.1 M PBS (pH 7.2) three times, followed by fixation in 1% (w/v) osmic acid for 2 h. Subsequently, the material was washed with PBS three times before being dehydrated through a gradient ethanol series. For SEM observation, the dehydrated material was dried with supercritical carbon dioxide and sputtered with palladium using IXRF SYSTEMS (MSP-2S) under a current of 20 µA for 20 s, then captured by SEM (JEOL, Japan) with an accelerated voltage of 30 kV. For semi-thin sections and TEM, the dehydrated material was infiltrated with 100% acetone twice, then immersed in acetone: Eponate 12^TM resin (3:1, 1:1, and 1:3, v/v) for 12 h each, and finally placed in pure resin for 12 h twice, followed by a 24-h incubation in a 65°C thermostat. Slices with 1–2 µm thickness were prepared using a semi-thin microtome (Lycra, Germany) or an ultrathin microtome then were observed using a Revolve FL microscope or TEM.

Molecular cloning of BP1

Primary mapping of BP1 was carried out from a segregating F₂ population derived from a cross between bp1-1 and HHZ, which was grown in the field in South China Agricultural University, Guangdong province in China. Details of the markers used for mapping are listed in Supplemental Data Set 1. The candidate genes harboring a T-DNA insertion were identified by thermal asymmetric interlaced PCR (TAIL-PCR) (Liu et al. 1995).

Vector construction and rice transformation

For functional complementation, a 5,484-bp genomic DNA fragment for BP1 was amplified from Zhonghua11 (ZH11, O. sativa L. ssp. geng/japonica) genomic DNA using primer pairs BP1FC F/R and cloned into the vector pCAMBIA3301 (Zhou et al. 2014). For knockout of BP1, the BP1Cas vector containing two targets for BP1 was constructed as described previously (Ma et al. 2015; Zhou et al. 2016). To generate the 35S:BP1-GFP vectors, the full-length BP1 cDNA was amplified and cloned in-frame and upstream of the GFP sequence in the pYL322 vector under the control of the cauliflower mosaic virus (CaMV) 35S promoter. To generate the BP1pro:BP1-Flag vectors, the full-length BP1 gDNA (start codon at ATG upstream 2 kb and coding sequence region of BP1) was amplified and cloned in-frame and upstream of the Flag sequence in pC1300. All primers are shown in Supplemental Data Set 1.

Phylogenetic analysis

To conduct the phylogenetic analysis, the full-length protein sequences of BP1 and 15 UGE family proteins were obtained by a BLAST_P search at NCBI (https://www.ncbi.nlm.nih.gov/). The sequence alignment was performed using CLUSTAL W. Based on the resulting alignment, a neighbor-joining tree was reconstructed using MEGA 10.0 (Bootstrap replicates = 1,000).

Expression analyses

Total RNA from anthers or spikelets at different developmental stages was isolated with TRIzol (Invitrogen, USA) reagent. First-strand cDNA was synthesized using HiScript II Q RT SuperMix for qPCR kit (Vazyme, Nanjing, China), Quantitative PCR was conducted using a 2хRealStar Green Fast Mixture kit (GenStar, Beijing, China) as previously described (Zhou et al. 2014). Each experiment was performed with three replicates, using ACTIN as the control gene. The value of 2^–△Ct was considered to be the relative expression level of the target gene. The primers used for qPCR are listed in Supplemental Data Set 1.

Subcellular localization assay

For transient expression, protoplasts were isolated from the leaf sheaths of 2-week-old rice plants; the subcellular localization assay was performed as previously described (Yang et al. 2013). In brief, for protoplast transfection, 3 to 5 μg plasmid carrying BP1-GFP, 200 μL protoplasts and 200 μL polyethylene glycol (PEG) solution (40% [w/v] PEG4000, 0.3 M mannitol and 0.1 M CaCl₂) were mixed gently and incubated for 15 min at room temperature. The cells were then washed with 1 mL cold W5 solution (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, 2 mM MES, pH 5.7) and resuspended in 1 mL WI solution (4 mM MES, 0.5 M mannitol, 20 mM KCl, pH 5.7). The cells were incubated at 28°C overnight after transfection. GFP signals were observed under a LSM510 Meta confocal microscope (Carl Zeiss, Germany) with excitation and emission wavelengths being 488 and 507 nm, respectively. To confirm the results of the subcellular localization, cytosolic and cytoplasmic proteins were obtained from total protein extracted from leaves of transgenic plants harboring the transgene BP1pro:BP1-Flag, and the localization of BP1 was examined by immunoblot analysis, using RNase Z^S1 and histone H4 (Cat No: 16047-1-AP, diluted by a factor of 5,000 before use) as subcellular markers for the cytoplasm and nucleus, respectively (Zhou et al. 2014).

Paraffin sectioning

Paraffin sectioning was performed according to our previous publication (Wang et al. 2022). The lodicules were fixed in 4% (w/v) paraformaldehyde and 0.25% (w/v) glutaraldehyde in 0.1 M PBS (pH 7.2) and stored in a refrigerator at 4°C overnight. The lodicules were dehydrated through a gradient ethanol series and treated with dimethylbenzene for 1 h three times, followed by immersion in melted paraffin and xylene (1:1, v/v) at 60 °C in an oven overnight. The solution was exchanged with fresh paraffin twice a day for 3 days. Then the lodicules were embedded in melted paraffin and cooled down; slices with thickness of 5 to 6 mm were cut using a Leica RM2235 paraffin slicer (Lycra, Germany).

RNA in situ hybridization

Specific regions of the BP1 coding sequence were amplified using the corresponding primers (Supplemental Data Set 1) and transcribed in vitro as probes using a DIG RNA Labeling kit (Roche, 11277073910, Switzerland). The prepared paraffin sections were incubated in hybridization buffer containing the probes (100 μL per slide) on slides with coverslips at 45°C overnight. Immunological detection of the hybridized probes was performed using a DIG Nucleic Acid Detection kit (Roche, 11175041910, Switzerland) according to the manufacturer's protocol.

Immunocytochemical analysis

Immunocytochemical analysis was conducted as previously reported with minor modifications (Rösti et al. 2007). The paraffin sections were treated with xylene for 15 min twice, and rehydrated with different concentrations of ethanol (100%, 95%, 85%, 75%, 50%, 30%, all v/v) and 1 × PBS for 5 min each time. The slices were immersed in 0.01 M, pH 7.2 sodium citrate buffer for 10 to 15 min at 95°C to restore the antigen and then blocked with 3% (w/v) BSA for 1 to 2 h, and subsequently washed with PBS for 5 min. The slices were respectively incubated with 100 to 200 μL 3% (w/v) BSA containing plant immunofluorescence antibodies (JIM13 (Cat.No.JIM13), LM5 (Cat.No.LM5), CCRC-M7 (lot:1605), CCRC-M1 (lot:1605), LM10 (Cat.No.LM10), LM2 (Cat.No.LM2), PlantProbes; All antibodies should be diluted by a factor of 10 before use.) for 1.5 h at 37°C and then rinsed with 1 × PBS for 10 min three times. Next, the slices were immersed in 100 to 200 μL 3% (w/v) BSA containing corresponding secondary antibody fuzed to GFP for 1 h at 37°C and washed with 1 × PBS for 10 min three times. Finally, the slices were mounted in 10% (v/v) glycerol and captured using a laser scanning confocal microscope. The negative control was performed as described above but without the addition of primary antibodies. The CCRC-M1 and CCRC-M7 were purchased from Cedarlane (USA), the secondary antibodies (Cat.NO.BA1108, China, diluted by a factor of 64 before use) for CCRC-M1, and CCRC-M7 were from mouse and purchased from BOSTER (China), JIM13, LM5, LM2, and LM10 were purchased from PlantProbes (UK) and the secondary antibodies (Cat.NO.3020-02, USA, diluted by a factor of 100 before use) for JIM13, LM5, LM2, and LM10 were from rat and purchased from SouthernBiotech (USA).

Callose observation

For callose wall observation, the dehydrated paraffin sections were stained with 0.1% (w/v) aniline blue, mounted in 10% (v/v) glycerol and then photographed using an Olympus BX51 microscope.

TUNEL assay

Anther samples were collected at different developmental stages. The TUNEL assay was performed as described (Li et al. 2006) using a TUNEL kit (DeadEnd Fluorometric TUNEL system; Promega, LOT0000466391) according to the manufacturer's instructions. The samples were analyzed using a confocal laser-scanning microscope (LSM510, Zeiss). The overlays of fluorescein signal and propidium iodide signal were taken as TUNEL-positive signals. All photographs were taken using the same settings.

ROS levels analysis

The contents for superoxide anion radical and hydrogen peroxide in the various stages of anthers were assayed by NBT staining (Zheng et al. 2019) and a Hydrogen Peroxide Assay Kit, respectively. For NBT staining, in brief, anthers at various developmental stages were immersed in 10 mM potassium-citrate buffer (pH 6.0) and vacuum-infiltrated for 5 to 10 min, then exchanged with 0.5 mM NBT solution and incubated for 2 to 3 h at 25 °C. The stained anthers were washed with 75% (v/v) ethanol, then imaged directly, or made into paraffin sections and mounted before being observed under a Leica DNRXA dissecting microscope. The NBT staining experiment was performed twice, using about 30 flowers (180 anthers) for every developmental stage each time. For detection of hydrogen peroxide, anthers at various developmental stages were collected, then the H₂O₂ concentration was measured using a Hydrogen Peroxide Assay Kit (Beyotime, Shanghai, China) according to the manufacturer's manual. NBT staining and analysis of H₂O₂ experiments were conducted using two different batches of materials.

UGE activity assay

The prediction of BP1 activity was performed by searching functional sites using SMART (Simple Modular Architecture Research Tool, http://smart.embl-heidelberg.de). The mutant BP1 (mBP1) was prepared by introducing three base transitions in the BP1 coding sequence (C to T at nucleotide 398; AA to TT at nucleotides 559 to 560; and AA to TT at nucleotides 619 to 620 from the ATG start codon). The full-length BP1 and mBP1 cDNA were individually inserted into the bacterial expression vector pET23d (Novagen) and introduced into Escherichia coli BL21 (DE3). The production of recombinant BP1 or mBP1 is induced by the addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (final concentration). The reaction was started by adding 1 μg of recombinant protein in 50 mM Tris-HCl (pH 7.6), 1% (w/v) BSA, 1 mM DDT, 1 mM EDTA, and different concentrations of UDP-Gal or UDP-Glu, reacted for 10 min at 37°C and terminated by incubation for 10 min at 100°C; the substrates and products were separated and detected by high performance liquid chromatograph (HPLC).

Nucleotide sugar measurements

The nucleotide sugar contents from anthers at S9 were measured according to a published study (Ebert et al. 2018). First, the samples were collected, ground into a fine powder in liquid nitrogen, immersed with chloroform/methanol (3:7, v/v), extracted with ice-cold water, frozen by liquid nitrogen and lyophilized until dry. The samples were then treated with the Solid Phase Extraction (SPE) method and measured by liquid chromatography/tandem mass spectrometry (LC-MS/MS).

Monosaccharide content analysis

Anthers at S9 were placed into Eppendorf 1.5 mL tubes with 0.5 mL water and vortexed for 30 s. The samples were ground for 4 min and sonicated for 5 min with the frequency of 40 KHz under an ice-water bath, and the procedure was repeated three times. The supernatant was collected by centrifugation at 13,523 rcf and at 4 °C for 15 min and concentrated with nitrogen flow. One hundred micorliter methanol solution containing 1,2-dihydro-5-methyl-2-phenyl-3h-pyrazol-3-one (10 mg/L) and 30 μL NaOH solution (0.1 M) were added into the concentrated sample. The mixture was incubated at 70°C for 30 min and cooled down to room temperature. Then, 30 μL HCl solution (0.1 M) was added before drying. The samples were redissolved with water, extracted with chloroform to remove excessive derivatization reagents, and filtered with a 0.22-μm filter before being measured by Ultra Performance Liquid Chromatography (ACQUITY UPLC, Waters).

Accession numbers

Sequence data from this article can be accessed in the GenBank database under the following accession numbers: Osg1 (LOC107275949), OsRBOHb (LOC4326027), OsRBOHc (LOC4339397), OsRBOHd (LOC4339045), OsRBOHe (LOC4327453), OsRBOHh (LOC4352439), OsDTC1 (LOC4329724), OsMT2b (LOC4337596), OsMT-1-4b (LOC4352563), DPS1 (LOC4338701), OsAGO2 (LOC4336991), OsEAT1 (LOC4336865), OsPTC1 (LOC4347213), OsAP25 (LOC4331874), OsAP37 (LOC4335981), GAMYB (LOC4327362), PTC2 (LOC4331173), OsTDL1A (LOC4352199), CYP704B2 (LOC4331756), OsUGP1 (LOC4347800), WDA1 (LOC3974662), CAP1 (LOC4328266), DPW (LOC3974662), OsUGP2 (LOC4328091), OsABCG15 (LOC4341486), OsHXK1 (LOC4343113), RBOH (LOC4352439), DTC1 (LOC9271129), OsMADS3 (LOC9271572), TIP2 (LOC4325164), TIP3 (LOC4333910), EAT1 (LOC4336865), TGA10 (LOC830575), OsAP37 (LOC4335981), OsMYB80 (LOC9272494), OsMYB103 (LOC9272494), OsGSL5 (LOC4340315), UGE1 in O. sativa (LOC4339812), UGE1 in A. thaliana (LOC837834), UGE1 in S. pombe (LOC2540938), UGE1 in B. rapa (LOC103836145), UGE1 in P. sativum (LOC12107955), UGE1 in H. vulgare (LOC732681), UGE2 in A. thaliana (LOC828492), UGE2 in H. vulgare (LOC32700), UGE3 in A. thaliana (LOC842622), UGE3 in H. vulgare (LOC732680), UGE3 in O. sativa (LOC4347638), UGE4 in A. thaliana (LOC822483), UGE4 in O. sativa (LOC4346723), UGE5 in A. thaliana (LOC826696), CF24/BZ1 (Osuge2) in O. sativa (LOC4345415), UGE in H. sapiens (2582).

Acknowledgments

We thank Chuan-he Liu (Instrumental Analysis and Research Center of South China Agricultural University) for transmission electron microscopy. We thank Lu-yi Pan (Instrumental Analysis and Research Center of South China Agricultural University) for HPLC analysis. This work was supported by the open competition program of top ten critical priorities of Agricultural Science and Technology Innovation for the 14th Five-Year Plan of Guangdong Province (2022SDZG05), the Guangdong Basic and Applied Basic Research Foundation (2019B030302006, 2022B1515120036), the National Natural Science Foundation of China (31921004, 32172017, 32172097), the Natural Science Foundation of Anhui Province (2008085MC98) and Open Funding of Rice Genetics and Breeding of Anhui Province Key Laboratory (SDKF-2020-01, SDKF-2021-02), the Double First-class Discipline Promotion Project (2021B10564001).

Author contributions

A.M.W., J.L.H., Z.L.L., C.X.Z., Y.Y.X., Z.Z.C., and H.Z. conceived the experiments; H.Q.C., S.Q.Z., R.Q.L., and G.Q.P. performed most of the experiments; C.R. analyzed the content of UDP-sugars; Z.Z.C. performed the molecular cloning of BP1; W.P.C. and Y.Y.X. contributed mutant materials; F.S.S. and J.L.N. planted the materials. M.L.L. and L.Y.Z. analyzed data and/or contributed materials; Y.P.X. and J.L.H. performed the semi-thin and ultrathin observations of bp1-2/BP1 sections; H.Q.C. and H.Z. wrote the paper; H.Z. designed the study. All authors commented on the manuscript.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Fig. S1. Pollen fertility rate and seed setting rate of the wild types T65 and ZH11, the bp1 mutants, and bp1-1/BP1.

Supplemental Fig. S2. Ultrastructure observations of anthers in T65 and bp1-1 from Stages 4 to 11 using TEM.

Supplemental Fig. S3. Semi-thin cross-section analysis of anthers in ZH11 and bp1-2/BP1 heterozygous plants.

Supplemental Fig. S4. Ultrastructure observations of anthers in bp1-2/BP1 heterozygous plants using TEM.

Supplemental Fig. S5. Phylogeny of BP1 homologs in various plant species.

Supplemental Fig. S6. RT-qPCR analysis of Os05g0595100 transcript levels in T65 and bp1-1 plants.

Supplemental Fig. S7. Enzymatic activity of mBP1 in vitro.

Supplemental Fig. S8. Expression pattern analysis of RBOH genes and ROS homeostasis-related genes during various anther development stages in T65 and bp1-1 plants.

Supplemental Fig. S9. RT-qPCR analysis of several tapetum development-related genes from stage 4 to stage 10 in T65 and bp1-1 plants.

Supplemental Fig. S10. RT-qPCR analysis of pollen wall development-related genes from Stage 4 to 10 in T65 and bp1-1 plants.

Supplemental Data Set 1. Primers used for mapping, cloning and expression analysis in this study.

Supplemental Data Set 2. Detailed statistical analysis in this study.

Supplemental File 1. Sequence alignment (with gaps) used for the phylogenetic analysis shown in Supplemental Fig. S5.

Supplemental File 2. Machine-readable tree file shown in Supplemental Fig. S5.