TMS5-mediated R-loop dynamics regulate thermosensitive fertility conversion in rice

Abstract

Hybrid rice is vital for global food security, with the thermosensitive genic male sterility (TGMS) locus tms5 used in over 95% of the two-line breeding system. Despite its widespread application, the regulatory mechanism underlying tms5-mediated TGMS remains unclear. This study demonstrates that under high-temperature conditions, weakened tapetum function and delayed microspore mother cell development lead to pollen defects and male sterility, whereas under low-temperature conditions, tapetal function is restored and microspore mother cell development proceeds normally, allowing fertility comparable to wild-type Zhonghua11 (ZH11). We further reveal that the knockout of TMS5, which encodes an RNase enzyme, leads to significant accumulation of R-loops under high-temperature conditions, compromising genomic stability and causing pollen abortion. This R-loop accumulation correlates with fertility conversion, as R-loop levels are elevated in tms5 at high temperatures when plants show male sterility, but return to ZH11-comparable levels at low temperatures when fertility is restored. Overexpression of the R-loop resolving protein Local Lesions 1 (OsLS1) partially restores tms5 fertility, further supporting a direct role of R-loop dynamics in tms5 fertility. Our findings elucidate the cellular and molecular mechanisms of tms5-mediated TGMS and provide a basis for improving two-line hybrid rice breeding.

This study reveals temperature-dependent accumulation of RNA-DNA hybrids that compromise genome stability and contribute to fertility conversion in tms5, the most widely deployed thermosensitive genic male sterility (TGMS) locus in rice.

Introduction

Rice (Oryza sativa L.) is one of the world’s most important crops and has been extensively studied for improving its yield and quality^1–4. Hybrid rice breeding, primarily through the three-line and two-line hybrid systems, has significantly enhanced productivity. The three-line system, which relies on cytoplasmic male sterility and specific restorer lines, is constrained by limited restorer germplasm diversity. In contrast, the two-line hybrid system utilizes photo-/thermo-sensitive genic male sterile (P/TGMS) lines, which exhibit male sterility under restrictive conditions (e.g., high temperature or long-day photoperiod) but recover fertility under permissive conditions (e.g., low temperature or short-day conditions). This thermosensitive trait is controlled by recessive nuclear genes, which allow breeders to utilize a broader range of germplasm as restorer lines, significantly enhancing the genetic diversity for hybrid development^5–7. By 2019, two-line hybrids made up 44% of China’s hybrid rice cultivation, with the tms5 allele serving as the predominant genetic determinant in over 95% of commercial P/TGMS lines⁸. A comprehensive understanding of tms5-mediated fertility conversion is critical for its application in the two-line hybrid system.

The TGMS line AnS-1, governed by the TMS5 locus⁹, was first identified in 1987. Under high temperatures, AnS-1 displays shrunken, white anthers with aborted pollen. In contrast, pollen fertility is largely restored under low-temperature conditions¹⁰. In 2014, Zhou et al. reported that the TGMS trait in the tms5 donor lines AnS-1 and ZhuS1 is caused by a C-to-A transition at nucleotide 71 of the TMS5 coding sequence, which introduces a premature stop codon and produces a truncated RNase Z^S1 protein predicted to be loss-of-function¹¹. RNase Z^S1 cleaves the 3′ trailer of precursor tRNAs during tRNA maturation, indicating that the widely used TGMS tms5 alleles are loss-of-function truncations^11–14. Recent studies have revealed that TMS5 possesses mRNA endonucleolytic cleavage activity and tRNA cyclophosphatase function^11,15–17. Functional studies of tms5 suppressors demonstrated that either reduced protein ubiquitination levels or overexpression of specific tRNAs can partially restore fertility^8,18,19, suggesting that efficient protein accumulation is a key determinant of fertility restoration in tms5. Despite these advances in understanding the biochemical functions of TMS5, the cytological basis of sterility in tms5 remains ambiguous. During anther development, both microspore mother cells (MMCs) and tapetum play critical roles in pollen development²⁰. Although cytological defects have been observed in both MMCs and the tapetal layer of tms5 mutants^8,11,16,21, the cellular mechanisms underlying TMS5 function in these cell types remain to be elucidated.

R-loops are stable RNA-DNA hybrid structures generated by nascent RNA invasion of double-stranded DNA, serving essential functions in transcription termination, gene expression and DNA repair²². However, their excessive accumulation can lead to genomic instability^23,24. Although well studied in yeast and mammals, the biological functions of R-loops in plants remain poorly understood^25–27. In this study, we demonstrate that TMS5 (RNase Z^S1) is required for the proper development of both the tapetum and MMCs, and that loss of function disrupts pollen development. Furthermore, we reveal that loss of RNase function leads to excessive accumulation of R-loops, which triggers genomic instability and conversion of thermosensitive male fertility in rice.

Results

TMS5 is expressed in all anther cell layers

Previous studies have shown that the mRNA expression of TMS5 was relatively enriched in MMCs¹¹. To examine TMS5’s expression at the protein level, we generated a genomic TMS5 construct fused to a short peptide linker and inserted it into the pCAMBIA1300:VENUS vector (Fig. 1a). In the transgenic lines, fluorescence of TMS5-VENUS accumulated in the cytoplasm of early stages (stage 6-8b) tapetal cells and MMCs. TMS5-VENUS signals were also detected in the cytoplasm of the epidermis, endothecium, and middle layer cells, but were absent in late-stage (stage 9-10) microspores (Fig. 1b). Subcellular localization assays in tobacco further confirmed that TMS5 localizes to the cytoplasm (Supplementary Fig. 1a), consistent with previous findings in rice protoplasts¹¹. Notably, DAPI-counterstained immunofluorescence showed predominantly cytoplasmic TMS5-VENUS, with an additional nuclear-associated signal in reproductive cells at specific stages (Supplementary Fig. 1b, c). We also generated a VENUS-tagged construct of TRZ1, the Arabidopsis homolog of TMS5 (Fig. 1c; Supplementary Fig. 2). TRZ1-VENUS also showed cytoplasmic localization across all four layers of anther, which is similar to that of TMS5 (Fig. 1d).

Fig. 1. TMS5 is broadly expressed in all cell layers of the anther.

a Schematic diagram of the pOsTMS5::OsTMS5-VENUS construct. b Confocal images showing OsTMS5-VENUS fluorescence in rice anthers from developmental stages 6 to 10. VENUS signal (530 nm) is shown in green; chlorophyll autofluorescence (>560 nm) is shown in magenta. Bar = 25 μm. c Schematic diagram of the pAtTRZ1::AtTRZ1-VENUS construct. Data were obtained from three independent experiments, with three independent transgenic lines. d Confocal images of AtTRZ1-VENUS fluorescence in Arabidopsis anthers from stages 5–10. Data were obtained from three independent experiments, with three independent transgenic lines. VENUS signal (530 nm) is shown in green, chlorophyll autofluorescence (>560 nm) is shown in magenta. Bar = 25 μm.

Tapetal regulatory pathway is weakened in tms5 under high temperatures

Previous studies reported that tms5 defects primarily occur in microsporocytes¹¹, yet subsequent evidence has indicated that tapetal cells are also affected in tms5 under high-temperature conditions^16,21. To clarify the cellular basis of male sterility in tms5, we performed semi-thin sections of anthers from tms5 (a CRISPR/Cas9-generated loss-of-function tms5 allele in the ZH11 background) (Supplementary Fig. 3) and ZH11 under both high and low temperatures. During early anther development (stages 6-7), MMCs and tapetal cells in tms5 appeared normal, with only occasional MMC degeneration. However, at stage 8a (anaphase I), initial cytological defects emerged, including partial vacuolization of both dyads and the tapetum. These abnormalities became more pronounced after meiosis II, with swelling of tapetal cells and increasingly disorganized locule contents (Fig. 2a). By stage 9, tapetal vacuolation was severe, and most microspores were degraded, although a small number of surviving microspores persisted in some locules (Supplementary Fig. 4a). These residual microspores ultimately degenerated by stage 10, resulting in complete male sterility (Fig. 2a). As the innermost somatic layer of the anther, the tapetum plays a central role in providing nutrients, enzymes and sporopollenin precursors for pollen wall formation²⁰ (Fig. 2b). To further investigate tapetal malfunction, we performed transcriptome profiling of anthers from wild-type ZH11 and tms5. Gene Ontology (GO) term enrichment analysis of differentially expressed genes (DEGs) showed that transcriptional regulatory processes were significantly affected in tms5 under high-temperature conditions (Supplementary Fig. 5). In rice, the transcriptional cascade UDT1-OsTDF1-TDR-OsMS188-PTC1 regulates tapetal development for pollen formation^28,29. We therefore examined the expression of these transcription factors, RT-qPCR results showed that UDT1, the upstream regulator of this pathway, maintained stable expression in tms5. However, the downstream transcription factors were downregulated, with the most substantial reductions observed in the terminal regulator PTC1, indicating a collapse of the tapetal regulatory cascade in tms5 under high temperatures. OsACOS12, OsPKS2, OsPKS1, TKPR1, DPW, CYP703A3, CYP704B2 and ABCG15 are responsible for sporopollenin precursors synthesis and transportation. Their expressions were also significantly downregulated. Under low-temperature conditions, the fertility of tms5 was restored. The expressions of both transcriptional regulators and sporopollenin-related genes were fully recovered, which was close to or higher than those in the wild-type ZH11 (Fig. 2c, d). These findings suggest that tapetal dysfunction in tms5 is associated with disrupted transcriptional regulation cascades, which likely contributes to the male sterility observed in tms5.

Fig. 2. The regulatory pathway governing tapetum function is weakened in tms5 under high temperatures.

a Semi-thin cross-sections of anthers from ZH11 and tms5 at developmental stages 6–10 under high and low-temperature conditions. Data were obtained from three independent experiments, with three independent lines. Tapetum is outlined with magenta dashed lines. The red dashed boxes highlight representative regions with obvious tapetal abnormalities in tms5 under HT. Bars = 20 μm. b Molecular pathways by which the tapetum contributes to pollen formation. Expression profiles of genes involved in tapetum development (UDT1, OsTDF1, TDR, OsMS188, PTC2 and PTC1) and pollen wall formation (OsACOS12, OsPKS2, OsPKS1, TKPR1, DPW, CYP703A3, CYP704B2 and OsABCG15) in tms5 under high (c) and low temperatures (d). Expression levels of most genes are restored under low temperatures. Data are presented as mean values ± SD. The error bars represent the SDs from three independent biological repeats. Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are 0.3101, 0.0003, 0.0189, 0.0091, 0.0071, 0.0016, 0.0006, 0.0014, 0.0002, <0.0001, 0.0002, 0.0001, 0.0120 and 0.00073, respectively). MMC microspore mother cell, dMMC defective microspore mother cell, ML middle layer, T tapetum, Dy dyad, dDy defective dyad, Tds tetrads, Msp microspore, dMsp defective microspore. Source data are provided as a Source Data file.

Microspore mother cell development is delayed in tms5 under high temperatures

Abnormal development of MMCs was observed in tms5 under high-temperature conditions. As meiosis is a critical step in the transition from MMCs to microspores, its disruption has been widely reported to cause MMC defects and ultimately lead to pollen abortion^30,31. To evaluate meiotic progression, we performed DAPI staining on anthers of tms5 and wild-type ZH11 at meiotic stages under high temperatures. Although occasional chromosomal abnormalities (e.g., micronuclei; ~2% of cells) were observed in tms5, no significant defects occurred during prophase I (Fig. 3a; Supplementary Fig. 4b). We subsequently performed detailed meiotic analysis and categorized floral buds into three groups: small (1.5–2.5 mm), medium (2.6–3.5 mm), and large (3.6–4.5 mm), and quantified their distribution across meiotic phases. In the small group, ZH11 MMCs predominantly progressed beyond the pre-leptotene stage, with only 5% remaining at pre-leptotene, while tms5 MMCs were delayed, with 36% at pre-leptotene and 50% at leptotene. In the medium group, 34% of ZH11 MMCs had progressed to pachytene or later stages, whereas tms5 MMCs were mostly stalled at earlier stages, with no MMCs reaching pachytene. In the large group, 42% of ZH11 MMCs had reached diplonema or later stages, while the majority of tms5 MMCs remained at zygotene (48%) or pachytene (16%), and only 11% progressed to later stages. These data demonstrate a clear developmental lag in tms5 MMCs under high temperatures across all floral bud size groups (Fig. 3b). In contrast, under low-temperature conditions, the timing of MMC meiotic progression in tms5 was comparable to that of ZH11 (Fig. 3c). These results indicate that tms5-mediated male sterility involves the developmental delay of MMCs.

Fig. 3. Microsporocyte development is delayed in tms5 under high temperatures.

a DAPI-stained meiotic chromosome configurations in microsporocytes of ZH11 and tms5 anthers at various floral bud lengths under high temperatures. Bar = 10 μm. Quantification of meiotic stage distribution in ZH11 and tms5 microsporocytes across spikelets of different lengths under high temperatures (b) and low temperatures (c). Developmental stages were classified based on cytological features and bud size. A total of n > 100 nuclei per group were scored and used for the statistical analyses. Source data are provided as a Source Data file.

Expression of TMS5 in either microsporocytes or the tapetum partially restores fertility in tms5

Phenotypic analysis indicated that tms5 disrupts both tapetal function and microsporocyte development, revealing its critical role in anther development. To further understand the impact of MMC and tapetum development on the fertility of tms5, we generated transgenic tms5 lines expressing TMS5 under the control of the microsporocyte-specific OsMSP (MEIOTIC SERINE PROTEINASE-LIKE PROTEIN) promoter or the tapetum-specific OsTDF1 (DEFECTIVE IN TAPETAL DEVELOPMENT AND FUNCTION 1) promoter^28,32(Fig. 4a). Under high-temperature conditions, both transgenic lines exhibited partial fertility restoration, as revealed by iodine-staining and seed-setting assays (Fig. 4b, c; Supplementary Fig. 6a). Specifically, the pMSP::OsTMS5 lines restored pollen I₂/KI staining rate to 18–66% and seed-setting rate to 25–60%. Similarly, the pTDF1::OsTMS5 lines exhibited about 35–65% pollen I₂/KI staining rate and 25–40% seed-setting rate (Fig. 4d, e). RT-qPCR confirmed that the expression of TMS5 in both transgenic backgrounds was higher than that in wild-type ZH11 (Fig. 4f, g). Moreover, expression of key transcription factors and genes associated with tapetum development and sporopollenin biosynthesis was partially restored in both lines (Supplementary Fig. 6b). These findings indicate that TMS5 is functionally required in both MMCs and tapetal cells, and that defects in either cell type contribute to the thermosensitive male sterility observed in tms5.

Fig. 4. Expression of TMS5 in either microspore mother cells or tapetum partially restores fertility in tms5.

a Schematic diagrams of pMSP::OsTMS5-NOS and pTDF1::OsTMS5-NOS constructs, driven by microsporocyte-specific (OsMSP) and tapetum-specific (OsTDF1) promoters, respectively. b Spikelets of tms5, pMSP::OsTMS5-NOS and pTDF1::OsTMS5-NOS plants under high temperatures. Bar = 2 mm. c Representative panicles of tms5 and pMSP::OsTMS5-NOS, pTDF1::OsTMS5-NOS transgenic lines. Bar = 2 cm. Quantification of I₂/KI staining rate (d) and seed setting rate (e) in tms5, pMSP::OsTMS5-NOS and pTDF1::OsTMS5-NOS transgenic lines. Data were obtained from three independent experiments, with three independent transgenic lines. Data are presented as mean values ± SD. The error bars represent the SDs from three independent biological repeats. Relative expression level of pMSP::OsTMS5-NOS (f) and pTDF1::OsTMS5-NOS lines (g). Data were obtained from three independent experiments, with three independent transgenic lines. Data are presented as mean values ± SD. The error bars represent the SDs from three independent biological repeats. Source data are provided as a Source Data file.

Accumulation of R-loops is associated with tms5 sterility

R-loops are RNA-DNA hybrid structures that regulate transcription, gene expression, and DNA repair. It was reported that their excessive accumulation can lead to genomic instability^23,24. While TMS5 is known for its role in tRNA 3’-end processing, previous studies also suggest broader RNase activity, including cleavage of UbL40 mRNAs^11,13,15,33. Given the localization of TMS5 protein during cell division (Supplementary Fig. 1b), we hypothesized that loss of TMS5 function might lead to abnormal accumulation of uncleaved RNAs, which could interfere with DNA and form R-loops, ultimately resulting in genomic instability and high-temperature-induced sterility. To test this hypothesis, we used S9.6 antibodies, which specifically recognize R-loops³⁴, to detect R-loop levels in tms5 and wild-type ZH11. R-loop signals were significantly elevated in tms5 under high temperatures, but no significant difference was observed under low temperatures (Fig. 5a, b; Supplementary Fig. 7). To assess whether this R-loop accumulation was associated with genomic instability, we performed comet assay to quantify DNA damage. At high temperatures, tms5 exhibited more extensive DNA damage than ZH11, with an increased proportion of nuclei exhibiting >30% tail DNA (Fig. 5c, d). This elevated DNA damage was largely restored at low temperatures (Fig. 5d). Furthermore, we performed DNA double-strand break (DSB) labeling by immunostaining with a γH2AX-specific antibody and found that the elevated R-loop signals in tms5 were accompanied by a marked increase in γH2AX signals (Fig. 5e, f). Kendall’s Tau-b rank correlation analysis between R-loop and γH2AX signals further demonstrated a stronger association between R-loop accumulation and γH2AX signals in tms5 than in ZH11 (Fig. 5g). These findings reveal that R-loop accumulation is associated with tms5 sterility, suggesting that R-loops may play a role in tms5 fertility conversion.

Fig. 5. Accumulation of R-loops is associated with tms5 sterility.

a S9.6 immunofluorescence staining of R-loops in anther nuclei from ZH11 and tms5 under high- and low-temperature conditions. R-loop signals (green, 520 nm) were detected with the S9.6 antibody; nuclei were counterstained with DAPI (blue). Bar = 1 μm. b Quantification of nuclear S9.6 fluorescence intensity shown in (a). Images were converted to 8-bit format using ImageJ, followed by threshold adjustment via the Threshold function and subsequent signal measurement within the selected regions. Each dot represents one nucleus; data were collected from n ≥ 30 nuclei per condition. S9.6 fluorescence intensities of ZH11 (HT): Min: 569, Max: 15996 Median: 3934, Box: 1915 (25th)–5681 (75th), Whiskers: Min to Max, n = 35; S9.6 fluorescence intensities of tms5 (HT): Min: 218, Max: 31514, Median: 11840, Box: 6910 (25th)–19487 (75th), Whiskers: Min to Max, n = 33. S9.6 fluorescence intensities of ZH11 (LT): Min: 1299, Max: 23430 Median: 3560, Box: 1960 (25th)–7761 (75th), Whiskers: Min to Max, n = 34; S9.6 fluorescence intensities of tms5 (LT): Min: 16, Max: 24296, Median: 2434, Box: 939 (25th)–5062 (75th), Whiskers: Min to Max, n = 47. Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are <0.0001 and >0.01). c Representative comet assay images showing four DNA damage classes, defined by the percentage of DNA present in the comet tail: 0–5%, 5–10%, 10–30% and >30%. d Frequency distribution of nuclei exhibiting different levels of DNA damage in ZH11 and tms5 under high- and low-temperature conditions. Data are representatives of three independent experiments, with ≥30 nuclei per condition. Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are shown in the figure). e Representative confocal images of anther nuclei co-immunostained with S9.6 (R-loops; green) and γH2AX (DNA damage; red). Nuclei were counterstained with DAPI (blue). Bar = 1 μm. f Quantification of nuclear S9.6 and γH2AX fluorescence intensities (background-subtracted) corresponding to (e) Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, p-values are <0.0001, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001. Each dot represents one nucleus. For each biological replicate, n ≥ 30 nuclei were quantified. S9.6 fluorescence intensities of ZH11: Min: 9008, Max: 221511 Median: 97714, Box: 60434 (25th)–133445 (75th), Whiskers: Min to Max; S9.6 fluorescence intensities of tms5: Min: 27235, Max: 695788, Median: 325995, Box: 240579 (25th)–387161 (75th), Whiskers: Min to Max. γH2AX fluorescence intensities of ZH11: Min: 3293, Max: 613540, Median: 49412, Box: 25206 (25th)–111227 (75th), Whiskers: Min to Max; γH2AX fluorescence intensities of tms5: Min: 108553, Max: 1064875, Median: 390593, Box: 295662 (25th)–665880 (75th), Whiskers: Min to Max. g Analysis of Kendall’s Tau-b rank correlation values between R-loop and γH2AX signals in (f) (p-value is <0.0001). Each dot represents one nucleus; data were collected from n ≥ 30 nuclei per condition. Kendall’s Tau-b correlation coefficients were calculated for each biological replicate and compared between genotypes. Kendall’s Tau-b correlation coefficients of ZH11: Min: 0.2852, Max: 0.9095, Median: 0.6560, Box: 0.5503 (25th)–0.7879 (75th), Whiskers: Min to Max; Kendall’s Tau-b correlation coefficients of tms5: Min: 0.4791, Max: 0.9431, Median: 0.8474, Box: 0.7466 (25th)–0.8790 (75th), Whiskers: Min to Max. Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Expression of an R-loop-resolving enzyme partially rescues the fertility of tms5

In rice, OsLS1 (Local Lesions 1) encodes a subunit of the RNase H2 complex, which has been implicated in R-loop clearance³⁵. Although RNase H enzymes degrade the RNA strand of R-loops, the direct enzymatic activity of OsLS1 in R-loop resolution has not been previously validated. We expressed and purified OsLS1 protein, and the in vitro assay showed that OsLS1 has the ability to resolve R-loops (Supplementary Fig. 8). To test whether excessive R-loop accumulation contributes to the male sterility of tms5, we introduced OsLS1 into tms5 under the control of microsporocyte-specific OsMSP promoter. The transgenic lines pMSP::OsLS1/tms5 plants showed partially restored fertility under high temperatures (Fig. 6a; Supplementary Fig. 9). In the same growth condition, the pMSP::OsLS1/tms5 plants exhibited a marked increase in pollen I₂/KI staining rate (64%, compared with 16% in tms5) and a significantly improved seed-setting rate (23%, compared with 4% in tms5) (Fig. 6b–d). Further immunofluorescence assays showed a significant reduction in nuclear S9.6 signals in the pMSP::OsLS1/tms5 transgenic lines compared with the tms5 mutant under high-temperature conditions (Fig. 6e, f). Consistently, comet assays indicated lower DNA damage in nuclei from the transgenic lines than that in the tms5 (Fig. 6g). These results indicate that reducing R-loop accumulation can alleviate the sterility phenotype in tms5, providing functional evidence that RNA-DNA hybrid clearance plays a key role in maintaining male fertility under restrictive conditions.

Fig. 6. R-loop clearance partially rescues the fertility of tms5.

a Representative panicles of tms5 and pMSP::OsLS1-NOS transgenic lines. Bar = 2 cm. b I₂/KI staining of pollen from tms5 and pMSP::OsLS1-NOS transgenic lines. Bar = 200 μm. Quantification of I₂/KI staining rate (c) and seed setting rate (d) in tms5 and pMSP::OsLS1-NOS transgenic lines. Data are presented as mean ± SD from three biologically independent replicates. Student’s two-tailed t test was used for statistical analysis of the difference between the wild type and mutants, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are shown in the figure). e S9.6 immunofluorescence staining of R-loops in anther nuclei from tms5 and pMSP::OsLS1-NOS transgenic lines. R-loop signals (green, 520 nm) were detected with the S9.6 antibody, and nuclei were counterstained with DAPI (blue). Bar = 1 μm. f Quantification of nuclear R-loop fluorescence intensity shown in (e). Each dot represents one nucleus. For each biological replicate, n ≥ 30 nuclei were quantified, and replicate means were used for statistical comparison. Student’s two-tailed t test was used for statistical analysis of the difference between the tms5 and transgenic lines; ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are shown in the figure). g Frequency distribution of nuclei exhibiting different levels of DNA damage in tms5 and pMSP::OsLS1-NOS transgenic lines. Data were obtained from three independent experiments, with ≥30 nuclei per condition. Student’s two-tailed t test was used for statistical analysis of the difference between the tms5 and transgenic lines, ns, no significant difference, *p < 0.05, **p < 0.01, ***p < 0.001 (p-values are shown in the figure). Source data are provided as a Source Data file.

Discussion

Weakened tapetal function and delayed microspore mother cell development lead to male sterility of tms5

Proper anther development is essential for male gametogenesis in plants, where both the tapetum and microsporocytes play key roles in pollen formation. The tapetum provides nutrients and synthesizes materials such as sporopollenin precursors for pollen development and wall formation. Simultaneously, microsporocytes undergo meiosis to produce haploid microspores, initiating the formation of male gametophyte^20,36. Defects in either cell type typically result in pollen abortion and male sterility. Although TMS5 transcripts have been previously detected in MMCs¹¹, our data demonstrate that TMS5 protein is expressed in both tapetal cells and microsporocytes (Fig. 1; Fig. 4), consistent with the defects in the tapetum and MMCs of tms5. Under high- temperature conditions, the tms5 mutant exhibited tapetal vacuolization accompanied by weakened function (Fig. 2), resembling the phenotypes observed in ostms15, which causes initial tapetal developmental defects³⁷. Meanwhile, MMCs exhibited pronounced meiotic progression delays (Fig. 3), phenotypes that were previously undercharacterized. Using tissue-specific complementation, we further demonstrate that TMS5 activity in either the tapetum or MMCs is sufficient to partially restore fertility, establishing that TMS5 functions in both layers are critical for fertility regulation. In addition to these two key layers, TMS5 is also detected in other somatic cell layers of the anther, including epidermis, endothecium, and middle layer, though the functional significance of these tissues remains to be elucidated.

R-loop accumulation is associated with genome instability and male sterility in tms5

R-loops are known to regulate transcription termination and gene expression, but their aberrant accumulation can compromise genomic stability^23,24. In rice, mutation of OsLS1, a subunit of the RNase H2 complex, results in R-loop accumulation³⁵. We confirmed that OsLS1 has R-loop resolving activity by in vitro enzymatic assays (Supplementary Fig. 8), and overexpression of OsLS1 in tms5 anthers partially restored fertility under high temperatures (Fig. 6), suggesting that R-loop accumulation acts as a key contributor to male sterility in tms5. Previous studies indicate that TMS5 loss impairs tRNA maturation and proteostasis, and that overexpression of specific aminoacyl-tRNA synthetases can partially rescue pollen fertility under high temperatures¹⁷. Our OsLS1 overexpression data support a complementary mechanism in which reducing R-loop-linked DNA damage also improves tms5 fertility under high temperatures. In addition, tms5 exhibits elevated nuclear S9.6 signals together with increased γH2AX and an increased nucleus-by-nucleus co-occurrence between R-loop and γH2AX signals (Fig. 5e–g), indicating DNA damage-associated genome instability during anther development. This genome instability is accompanied by elevated DSB-associated DNA damage signaling. In animal cells, DSBs and their signaling can trigger transcriptional silencing in cis to DNA breaks^38–40. Therefore, we propose that R-loop-driven genome instability and the associated DNA damage response may indirectly impact the expression of key tapetal regulators and compromise anther development. This model is supported by prior studies showing that R-loop homeostasis factors are required to maintain appropriate gene expression programs^41,42. Previous studies showed that tms5-mediated sterility under high temperatures is associated with increased ubiquitination of unstable proteins and reduced mature tRNAs, ultimately impairing translational efficiency^8,18,19. Together with our findings, male sterility in tms5 may result from a convergence of molecular disruptions, including enhanced ubiquitination of unstable proteins, reduced levels of mature tRNAs, and R-loop-associated DNA damage signaling and genome instability. Interestingly, similar phenotypes are observed in animals, where deletion of RNase H1, a key enzyme for R-loop resolution, causes accumulated R-loops and spermatogenic defects⁴³. These findings indicate the conserved role of R-loop homeostasis in safeguarding genome stability and reproductive development across eukaryotes.

The mechanisms of fertility restoration in tms5

Fertility restoration in tms5 under low temperatures was accompanied by a marked reduction in R-loop accumulation to wild-type ZH11 levels (Fig. 5), implying that R-loop dynamics may underlie the thermosensitive fertility conversion in this mutant. Previous reports suggest that fertility restoration in P/TGMS lines under low temperature is associated with slow development⁴⁴. Meanwhile, reactive oxygen species (ROS) show a temperature-dependent pattern in P/TGMS lines, accumulating at high temperatures but declining at low temperatures, and this shift has been linked to low-temperature-induced fertility restoration in P/TGMS lines^45,46. In tms5, ROS levels are efficiently cleared under low temperatures, supporting a role for ROS homeostasis in fertility recovery⁴⁵. Further, we propose that the fertility restoration of tms5 under low-temperature conditions may be associated with low-temperature-induced slow development in plants. Under low-temperature conditions, reduced transcriptional activity likely leads to a lower global RNA load. This physiological slowdown may provide an extended temporal window that enables more efficient clearance of R-loops by endogenous RNase machinery. As a result, reduced R-loop accumulation may relieve transcriptional repression and help reestablish protein homeostasis, providing a molecular basis for fertility recovery in tms5.

Together, these findings reveal a thermosensitive fertility restoration mechanism in which R-loop dynamics regulate male fertility in rice. Under high-temperature conditions, TMS5 is essential for clearing excessive RNA to maintain efficient expression of key developmental genes, thereby ensuring proper anther development and pollen formation. However, loss of TMS5 impairs RNA degradation, resulting in excessive R-loop accumulation that is accompanied by DNA damage-associated genome instability, which may contribute to defects in tapetal and MMC development, ultimately causing pollen abortion. By contrast, under low temperatures, slower development helps maintain R-loop homeostasis in the absence of TMS5, supporting normal pollen development and fertility recovery (Fig. 7). Our study not only elucidates a unique regulatory pathway underlying tms5-mediated P/TGMS, but also underscores the broader significance of R-loops in plant development and temperature-sensitive fertility conversion. These insights provide a conceptual foundation for future strategies to exploit RNA-mediated genome regulation in crop breeding.

Fig. 7. Proposed model illustrating the role of R-loops in tms5-mediated TGMS in rice.

Under high temperatures, TMS5 functions as an RNase that degrades specific structured RNAs to maintain RNA homeostasis and prevent excessive R-loop accumulation. In tms5, loss of RNase activity leads to aberrant R-loop accumulation, which is associated with genomic instability and DNA-damage signaling, and may indirectly perturb transcriptional programs of anther developmental genes, thereby contributing to tapetal and MMC developmental defects and ultimately causing pollen abortion. By contrast, under low temperatures, reduced transcriptional activity and slower development limit R-loop accumulation in tms5, allowing partial restoration of fertility. Figure was created in BioRender https://BioRender.com/adk6j1u.

Methods

Plant materials and growth conditions

The rice lines used in this study included the wild-type ZH11 (Oryza sativa ssp. japonica) and the tms5 mutant, kindly provided by Prof. Hai Zhou (South China Agricultural University). The tms5 mutant was generated in the ZH11 background using CRISPR/Cas9-mediated genome editing and confirmed to be transgene-free after segregation. The edited allele carries 1618-bp deletion spanning from 761 bp upstream of the ATG start codon to 857 bp within the third exon, effectively disrupting gene function (Supplementary Fig. 3). The rice lines were cultivated in paddy fields in Shanghai. All transgenic lines were grown at the Shanghai Normal University Botanical Garden. Rice transformation was carried out by Wuhan BioRun Biosciences Co., Ltd. (China).

Phylogenetic analysis of RNase Z homologs

Amino acid sequences of RNase Z family proteins from Arabidopsis thaliana and Oryza sativa were retrieved from the TAIR (https://www.arabidopsis.org/) and RAP-DB (https://rapdb.dna.affrc.go.jp/) databases, respectively. A condensed phylogenetic tree was constructed with MEGA11 software, using the Nearest-Neighbor-Intercharge method with the following parameters: Jones-Taylor-Thornton model, Uniform rates, Use all sites, and 500-replicates Bootstrap. The homolog proteins are named according to TAIR (Arabidopsis thaliana) and RAP-DB (Oryza sativa).

Generation of transgenic plants

For protein expression analysis, the genomic sequence of OsTMS5 including a 2438-bp native promoter was amplified, fused with a short peptide linker (GGGGSGGGGS), and inserted into the pCAMBIA1300:VENUS vector (CAMBIA, Australia) for transformation into wild-type ZH11 plants. A second construct containing a 2409 bp OsTMS5 native promoter was amplified, fused to a short peptide linker (GGGGSGGGGS) and inserted into the pCAMBIA1300:VENUS vector for transformation into Arabidopsis thaliana Col. To generate cell-type-specific expression constructs, a 2346-bp OsMSP promoter or a 2114-bp OsTDF1 promoter was amplified from ZH11 genomic DNA and fused with the OsTMS5 genomic DNA sequence, then inserted into the pCAMBIA1300 vector. For R-loop clearance experiments, the OsLS1 genomic sequence was fused to the OsMSP promoter and cloned into pCAMBIA1300 vector. All constructs were sequence verified and introduced into rice via Agrobacterium-mediated transformation. Genotypes of transformants were confirmed by PCR, and primer sequences are listed in Supplementary Data Set S1. At least three independent lines were used for phenotypic and fluorescence analyses.

Phenotypic analysis

Plant and panicle images were captured using a Nikon D7000 digital camera (Nikon, Japan). Anther morphology was observed with an Olympus SZX10 dissecting microscope (Olympus, Japan), and I₂/KI stained pollen grains were imaged using an Olympus BX51 fluorescence microscope (Olympus, Japan). I₂/KI staining reflects starch accumulation and is therefore presented as the percentage of I₂/KI-stainable (starch-positive) pollen grains at the time of sampling, rather than as a direct measure of pollen fertility. For histological analysis, semi-thin sections of anthers were prepared and imaged using a Leica microtome system (Leica, Wetzlar, Germany). Protein expression was examined using a laser confocal microscope FV3000 (Olympus, Japan).

Gene expression analysis

Total RNA was extracted from inflorescences of ZH11 and tms5 plants grown under different temperature conditions using TRIzol (Invitrogen, Waltham, USA). Reverse transcription was performed using the TransScript Fly First-Strand cDNA Synthesis SuperMix (TransGen Biotech, Beijing, China). RNA-seq was performed using three biological replicates. RT-qPCR assays were performed with Taq Pro Universal SYBR qPCR Master Mix (Vazyme Biotech Co., Ltd., Nanjing, China) on an ABI 7300 system (Life Technologies, Carlsbad, USA), using OsACTIN as the internal reference. Each experiment was conducted with three biological replicates.

Comet assay

Anthers at meiotic stages were collected, and MMCs were released by mechanical disruption. Cells were embedded in low-melting-point agarose for slide preparation. After solidification, slides were immersed in lysis and alkaline buffers as previously described⁴⁷. Electrophoresis was conducted for 10 min, followed by DAPI staining. DNA damage was visualized under a laser confocal microscope FV3000 (Olympus, Japan). The images were analyzed using ImageJ to determine the percentage of DNA in each comet tail (Tail DNA%: the percentage of DNA content in the tail compared to the total DNA content in the comet). Comets were categorized into four groups based on the amount of Tail DNA: 0–5% (0–5% DNA in the tail), 5–10% (5–10% DNA in the tail), 10–30% (10–30% DNA in the tail), >30% (more than 30% DNA in the tail). At least 30 comet images were analyzed per slide, with three biological replicates per condition.

Meiotic chromosome preparation and staging

Young panicles were fixed in Carnoy’s solution and stored at −20 °C until use. Microsporocytes were released from anthers and spread onto microscope slides. The slides were treated with 50% (v/v) acetic acid, air-dried, and briefly frozen in liquid nitrogen for 20 s. Samples were then dehydrated with absolute ethanol and air-dried at room temperature. DAPI staining solution was applied for observation, and meiotic stages were examined under an Olympus BX51 fluorescence microscope (Olympus, Japan).

Immunofluorescence

Rice buds at the microsporocyte stage were collected and fixed in pre-chilled methanol for 30 min, followed by two washes in deionized water (5 min per wash). Anthers were placed on lysine-coated slides, gently squashed to release MMCs, and air-dried at room temperature. For improved permeability, slides were flash-frozen in liquid nitrogen for 20 s and air-dried again. Membranes were permeabilized in Wash Buffer I (1× PBS with 1% Triton X-100) for 40 min at room temperature, followed by blocking with 3% BSA in PBST (1×PBS with 0.1% Tween 20) for 40 min at room temperature. Primary antibodies (S9.6, Abcam, ab234957) were diluted 1:150 (v/v) in PBST solution, applied to the slides, and incubated in humid chamber at 4 °C for 16 h. After three 15-min washes in PBST, fluorochrome-coupled secondary antibodies (Goat anti-mouse IgG H&L (Alexa Fluor® 488), Abcam, ab150113) were applied (1: 200 in PBST) and incubated for 1 h at 37 °C. Nuclei were counterstained with DAPI. Images were captured using a laser confocal microscope FV3000 (Olympus, Japan). Images were converted to 8-bit format using ImageJ, followed by threshold adjustment via the Threshold function and subsequent signal measurement within the selected regions.

Dot blot assay

Genomic DNA was extracted from rice floral buds using the CTAB method and quantified with a NanoDrop™ One spectrophotometer. DNA samples from wild-type ZH11 and tms5 mutant plants were diluted to 50 ng/μl, and 2 μl aliquots were spotted onto methanol-activated PVDF membranes. Membranes were air-dried, rinsed briefly with PBST (1× PBS containing 0.1% Tween 20) for 2 min, and blocked with 3% BSA in PBST for 1 h at room temperature. Membranes were then incubated with the primary antibody (S9.6, Abcam, ab234957) diluted (1:1000 dilution in PBST) for 1 h at room temperature, followed by three washes with 3% BSA in PBST (5 min each). Afterward, membranes were incubated with the secondary antibody (Goat Anti-Mouse IgG, HRP Conjugated, Cwbio, CW0102S) (1:5000 dilution in PBST) for 1 h at room temperature. R-loop signals were detected using the Tanon-5500 chemiluminescent imaging system (Tanon, Shanghai, China). “IntDen” (defined as integrated density) on the Y-axis was quantified in ImageJ by applying the invert function, subtracting the background, and subsequently measuring the corresponding regions of interest.

In vitro enzymatic assay

Equal amounts of tms5 genomic DNA were incubated with GST, RNase H (Thermo Fisher Scientific, EN0201)(positive control) or GST-OsLS1(test sample) in 20 μl reaction volumes. GST alone was used as the negative control. Reactions were carried out at room temperature for 30 min, followed by DNA purification using phenol:chloroform:isoamyl alcohol (25:24:1, v/v/v) extraction. R-loop levels were subsequently analyzed by dot blot assay as described above.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Z.-N.Y. and N.W. conceived and supervised the study. B.-S.Z. and M.-Y.R. performed protein expression and analysis. B.-S.Z. and A.-A.C. conducted cytological observations and gene expression analysis. Y.-M.Y. and B.-S.Z. prepared and analyzed meiotic chromosomes. B.-S.Z. carried out the dot blot assay and analysis. B.-S.Z., J.-S.X. and X.C. performed the immunofluorescence experiments and analysis. B.-S.Z., J.Z. and A.-A.C. assessed fertility phenotypes. Z.-N.Y., N.W. and B.-S.Z. wrote the manuscript. All authors reviewed and approved the manuscript.

Peer review

Peer review information

Nature Communications thanks Jungnam Cho, Anna Dobritsa, Qianwen Sun and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within this paper and its Supplementary Information files. All the biological materials generated in this work are available from the authors upon request. The RNA-seq raw data are provided at Figshare (10.6084/m9.figshare.31698265)⁴⁸. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71491-z.