A natural allele of OsMS1 responds to temperature changes and confers thermosensitive genic male sterility

Changes in ambient temperature influence crop fertility and production. Understanding of how crops sense and respond to temperature is thus crucial for sustainable agriculture. The thermosensitive genic male-sterile (TGMS) lines are widely used for hybrid rice breeding and also provide a good system to investigate the mechanisms underlying temperature sensing and responses in crops. Here, we show that OsMS1 is a histone binding protein, and its natural allele OsMS1wenmin1 confers thermosensitive male sterility in rice. OsMS1 is primarily localized in nuclei, while OsMS1wenmin1 is localized in nuclei and cytoplasm. Temperature regulates the abundances of OsMS1 and OsMS1wenmin1 proteins. The high temperature causes more reduction of OsMS1wenmin1 than OsMS1 in nuclei. OsMS1 associates with the transcription factor TDR to regulate expression of downstream genes in a temperature-dependent manner. Thus, our findings uncover a thermosensitive mechanism that could be useful for hybrid crop breeding. Thermosensitive genic male-sterile (TGMS) lines are widely used for two-line hybrid rice breeding, but the mechanism of TGMS has not been fully elucidated. Here, the authors show that natural allele of the OsMS1 gene, encoding a histone binding PHD finger protein, responds to temperature change and confers TGMS in rice.

A s one of the most important cereal crops, rice (Oryza sativa L.) is a staple food for nearly half of the world's population 1 . Hybrid rice leads to a 10-20% grain yield increase compared with parental inbred lines 2-4 , thereby contributing greatly to food security. The three-line system (the cytoplasmic male-sterile line [CMS], the maintainer line, and the restorer line) and two-line system (the thermosensitive genic male-sterile [TGMS] line/photoperiod-sensitive genic male-sterile [PGMS] line and the restorer line) are widely used for hybrid rice breeding 5,6 . TGMS lines are male-sterile at restrictive (high) temperatures but male-fertile at permissive (low) temperatures. Therefore, the male-sterile line in two-line system also serves as the maintainer line, resulting in a dramatically decreased cost in seed production. Considering that most cultivars can restore the fertility of TGMS lines 5,7,8 , two-line system utilizes broader genetic resources to generate hybrids than three-line system in hybrid rice breeding. Recently, TGMS lines have made tremendous contributions to heterosis breeding for increasing rice yield in China 2,5 .
As the male gametophyte development has been identified as the most heat-vulnerable stage in flowering plants 9 , high temperature often causes male sterility. TGMS lines therefore provide a good system for studying the tissue-specific thermosensitive mechanisms. A few factors conferring TGMS phenotype have been identified in rice, including RNase Z S1 , long non-coding RNA, and receptor-like kinases, but temperature does not directly alter the properties (e.g., activity or protein abundance) of these factors 5,7,8 .
Male reproductive development is a dynamic fine-tuning regulatory process involving the formation of male meiocytes, gametophytes, and gametes 10 . The male reproductive organ, the anther, forms male gametophytes via meiosis and mitosis within the flower, and releases mature pollen grain 11 . Tapetum, the innermost anther wall layer, plays essential roles in supporting gametogenesis, and undergoes programmed cell death (PCD)-triggered degradation after the meiosis of microspore mother cells 10 . Thus, abnormal development and/or premature or delayed degradation of the tapetum results in male sterility. Several conserved receptor-like protein kinases and transcriptional regulators have been identified to be associated with tapetal cell development in rice 10 . Rice MULTIPLE SPOROCYTE 1 (MSP1) and MICROSPORELESS 2 (MIL2) encode receptor-like protein kinases and work coordinately to promote the correct progression of anther development 12,13 . Rice MYB transcription factor encoded by GAMYB plays an important role in floral organ development and pollen development 14,15 . Rice UDT1 (UNDEVELOPED TAPETUM 1) 16 , TDR (TAPETUM DEGENERATION RETARDATION) 17 , EAT1 (ETERNAL TAPETUM 1) 18 , and TIP2/bHLH142 (TDR INTER-ACTING PROTEIN 2) 19,20 , encoding basic helix-loop-helix (bHLH) transcription factors, play key roles in maintaining tapetum development in rice. TIP2/bHLH142 acts downstream of UDT1 and GAMYB but upstream of TDR and EAT1 in pollen development [18][19][20] . PTC1 (PERSISTENT TAPETAL CELL1)/OsMS1 (MALE STERILITY 1) encodes a transcriptional activator that regulates tapetal PCD and pollen exine formation in rice [21][22][23] . However, how PTC1/OsMS1 is involved in the regulation of transcriptional activation remains unclear.
Here, we report that OsMS1 is a histone binding protein, and its natural allele OsMS1 wenmin1 from two indica TGMS lines (Tian1S and HengnongS-1) confers thermosensitive genic male sterility in rice. The abundances of OsMS1 and OsMS1 wenmin1 are temperature dependent. OsMS1 associates with the transcription factor TDR in nuclei to activate its downstream gene expression in a temperature-dependent manner. These findings define an important mechanism of thermosensitive genic male sterility control and open a way for hybrid crop breeding.

Results
The wenmin1 allele confers thermosensitive male sterility in both indica and japonica backgrounds. To dissect the temperature sensing mechanism, we collected multiple temperaturesensitive male-sterile lines used for two-line hybrid rice breeding. Among them, the indica TGMS line Tian1S showed male sterility under high temperature (restrictive temperature), but exhibited normal male fertility under low temperature (permissive temperature), irrespective of the day length ( Supplementary Fig. 1a, b). Our genetic analysis revealed that the thermosensitive malesterile phenotype was controlled by a single recessive locus ( Supplementary Fig. 1c, Supplementary Table 2), hereafter named this locus wenmin1 (wenmin: temperature-sensitive in Chinese).
To explore whether the wenmin1 allele could also confer the thermosensitive male sterility in japonica varieties, we backcrossed Tian1S with a wide compatibility japonica cultivar ZhongJing1 (ZJ1) and generated the nearly isogenic line ZJ1-wenmin1, which contains the wenmin1 allele in ZJ1 background ( Supplementary  Fig. 2a, b). ZJ1-wenmin1 exhibited normal vegetative growth ( Supplementary Fig. 2b), but was temperature-sensitive male sterile, irrespective of the day length ( Supplementary Fig. 2c). To investigate the temperature range of wenmin1 in detail, we examined the pollen fertility of Tian1S and ZJ1-wenmin1 at day average temperatures (DAT) 22°C, 23°C, 25°C, 27°C, and 29°C under the fixed photoperiods. Pollen fertility of Tian1S and ZJ1-wenmin1 was normal at 22°C, whereas they exhibited a gradual abortion with increasing temperatures (Fig. 1a, b). Tian1S and ZJ1-wenmin1 showed no mature pollen grains at 27°C and 25°C, respectively. Together, these results indicate that the wenmin1 allele confers the temperature-sensitive male sterility in both indica and japonica cultivars at restrictive temperature and that the male sterility-fertility transition temperature of the wenmin1 allele depends on genetic backgrounds.
We further examined anther development to characterize the cellular abnormalities of ZJ1-wenmin1. At restrictive temperature, the early development events of anthers and pollen grains of ZJ1-wenmin1 were similar to those of ZJ1, whereas ZJ1-wenmin1 tapetum appeared abnormally enlarged, and microspores became degenerated with few or no mature pollen grains compared with normal tapetal layer degradation and mature pollen grains formation of ZJ1 ( Supplementary Fig. 2d).
Identification of the wenmin1 mutation. To identify the wen-min1 mutation, we conducted a map-based cloning by crossing Tian1S with ZJ1. Fine mapping of the wenmin1 mutation using 2102 F 2 individuals narrowed the candidate region to a 31.6 kb interval between markers RM24453 and A09-13 (Fig. 1c). Interestingly, this mapping region contains three genes, and one of them (LOC_Os09g27620) is the PTC1/OsMS1 gene 21,23 . Notably, mutations in several PTC1/OsMS1 homologs including Arabidopsis thaliana MS1 (AtMS1) [24][25][26][27] , barley MALE STERILITY 1 (HvMS1) 28 , and maize MALE STERILITY 7 (ZmMS7) 29 resulted in the male sterility, although these mutations did not cause the temperaturesensitive male-sterile phenotype. Thus, LOC_Os09g27620 was the most promising candidate gene. To identify the causal mutation, we firstly investigated the expression of LOC_Os09g27620 in ZJ1 and ZJ1-wenmin1. We found that LOC_Os09g27620 is mainly expressed in stages 8 and 9 of anther development ( Supplementary Fig. 3a), consistent with the public data ( Supplementary Fig. 3b, c) 30 . Similarly, OsMS1-GUS fusion proteins in gOsMS1:GUS transgenic plants were also detected in stages 8 and 9 of anther development (Fig. 1d). We observed that the transcript levels of LOC_Os09g27620 were similar in both ZJ1 and ZJ1-wenmin1 grown at restrictive temperature ( Supplementary Fig. 3a), indicating that the mutation in ZJ1-wenmin1 does not affect the expression of LOC_Os09g27620.
We then sequenced the LOC_Os09g27620 gene and found a nucleotide transition (T902C) in the second exon of LOC_Os09g27620 in Tian1S and ZJ1-wenmin1 compared with ZJ1 and indica cultivars 93-11 and Pei'Ai 64 (PA64) (Fig. 1c), which resulted in an amino acid change (Leu 301 to Pro, L301P) (Fig. 1e). These results suggest that T902C mutation in LOC_Os09g27620 might be responsible for the temperature-sensitive male-sterile phenotype of Tian1S and ZJ1-wenmin1. To confirm this, we conducted a genomic complementation assay. A genomic fragment of LOC_Os09g27620 (gOsMS1) from 93-11 was transformed into ZJ1-wenmin1. Transgenic plants (gOsMS1:ZJ1-wenmin1) showed normal fertility under restrictive temperature ( Fig. 1f and Supplementary Fig. 4). Thus, these results demonstrate that the T902C mutation in LOC_Os09g27620/OsMS1 is responsible for the temperature-sensitive male sterile phenotype in Tian1S and ZJ1-wenmin1, and that the wenmin1 allele is a weak allele. Therefore, the wenmin1 allele was referred to as OsMS1 wenmin1 thereafter.
To explore whether other TGMS lines could contain the OsMS1 wenmin1 mutation, we sequenced the OsMS1 gene in multiple TGMS lines. The indica thermosensitive male sterile line HengnongS-1 (HNS-1) was bred from a cross progeny between wild rice and indica cultivars, and has been widely used in twoline hybrid rice breeding since the 1990s in China (China Rice Data Center, http://www.ricedata.cn). Interestingly, genomic DNA sequencing showed that HNS-1, but not other TGMS lines, harbors the same mutation as Tian1S and ZJ1-wenmin1 (Fig. 1c). Notably, the tms9-1 was also mapped to the OsMS1 wenmin1 region using an F 2 population from a cross between HNS-1 and Minghui63 (MH63) 31 . The OsMS1 was identified as the candidate gene for TMS9-1, although genomic complementation test has not been conducted 31 . Consistent with this, our allelic tests revealed that OsMS1 wenmin1 was allelic to tms9-1, but not to tms5 (Supplementary Table 5). We then analyzed the sequences of 4,726 rice genomes (http:// ricevarmap.ncpgr.cn/v2) and found that none of these rice varieties contained the OsMS1 wenmin1 mutation, indicating that OsMS1 wenmin1 is a rare allele. OsMS1 and its homolog genes are shared in monocot and dicot species (Supplementary Figs. 5 and 6), and the Leu 301 of OsMS1 is evolutionarily conserved (Supplementary Fig. 7), suggesting a promising common target to engineer TGMS lines in the plant kingdom.
OsMS1 is a histone binding protein. OsMS1 encodes a 683amino acid protein that contains a putative nuclear localization motif (NLS) at the N-terminus, a LXXLL motif (where L is leucine and X is any amino acid) in the middle region 32 , and a plant homeodomain (PHD) in the C-terminus ( Fig. 1e and Supplementary Fig. 5). In animals, the PHD fingers were proposed to play crucial roles in transcriptional regulation and were known to recognize a subset of post-translational histone modifications, such as histone H3 trimethylated at Lys4 (H3K4me3), unmethylated tail of histone H3 (H3K4me0), histone H3 trimethylated at Lys9 (H3K9me3), and histone H3 or H4 acetylated at various lysine residues (H3/H4Kac) [33][34][35] . However, little is known about the function of tissue-specific expressed PHD proteins in plants 36 . Interestingly, sequence alignment revealed that the PHD domain of OsMS1 is highly conserved in OsMS1 and its homologs (Supplementary Fig. 8a, b), but with a unique amino acid composition ( Supplementary Fig. 8b, c), possessing more acidic amino acid that is different from the known PHD domains of human BPTF and PHF2 that specifically recognize H3K4me2/3, of human AIRE and BHC80 that bind H3K4me0, of human DPF3b that recognizes H3K14ac ( Supplementary Fig. 8b, c). We therefore asked whether OsMS1 could bind histone tails through its PHD finger. To answer this, we tested several biotin-labeled histone H2A, H2B, H3, and H4 peptides and modified histone peptides bearing methylations and acetylations on different residues in peptide pull-down assays. Unexpectedly, FLAG-tagged full-length OsMS1 preferentially binds to H4 (lane 9), H4 acetylated at lysines 5, 8, 12, 16 (lanes 10-13), and H3 (lane 5) (Fig. 2a, upper panel). Meanwhile, OsMS1 wenmin1 also interacted with H4, acetylated H4, and H3 ( Supplementary  Fig. 9a), indicating that the mutation in OsMS1 wenmin1 does not affect its histone-binding ability. A glutathione S-transferase (GST) tagged C-terminal region of OsMS1 containing the PHD domain (GST-OsMS1C; amino acids 626-683) retains the histone-binding specificity of the full-length OsMS1 (Fig. 2a, middle panel), indicating that the PHD domain-bearing C-terminal region of OsMS1 is sufficient for binding these histone peptides. We further mutated the conserved C642, C645, H650, and C653, which are important for PHD domain structure, into Ala ( Supplementary Fig. 8b). We found that the mutations (C642A, C645A, H650A, and C653A) within the PHD domain abolished the interactions between OsMS1 and histone peptides ( Supplementary Fig. 9b), indicating that the PHD domain structure is crucial for the specific interaction. We confirmed the interaction between full-length OsMS1 and H4 in planta using a luciferase complementation imaging (LCI) assay and a bimolecular fluorescence complementation (BiFC) assay in N. benthamiana leaves (Fig. 2b, c). The association of OsMS1 with H4 in rice anthers is further demonstrated by the semi-in-vivo pulldown and in vivo co-immunoprecipitation assays (Fig. 2d, e). Together, these results indicate that OsMS1 is a histone binding protein.
The L301P mutation in OsMS1 wenmin1 affects its nucleus localization. Given the putative nuclear localization motif in OsMS1 ( Fig. 1e and Supplementary Fig. 5) and the fact that the LXXLL motif was proposed to play a role in determining nuclear localization of Dax-1 (dosage-sensitive sex reversal-adrenal hypoplasia congenita critical region on the X chromosome, gene 1) 37,38 , we asked whether the L301P mutation in the conserved LXXLL motif of OsMS1 could influence its subcellular localization. We observed that the fluorescent signal of green fluorescent protein (GFP)-OsMS1 fusion protein was exclusively found in the nuclei in transgenic rice ( Supplementary Fig. 10a), yeast ( Fig. 3b and Supplementary Fig. 10b, upper panels), N. benthamiana (Supplementary Fig. 10c, upper panels), Arabidopsis ( Supplementary Fig. 10d, upper panels), and rice protoplast ( Supplementary Fig. 10e, upper panels), whereas GFP-OsMS1 wenmin1 is distributed in both the cytoplasm and nuclei (Fig. 3b, Supplementary Fig. 10b, lower panels, Supplementary Fig. 10c, middle and lower panels, Supplementary Fig. 10d, lower panels, Supplementary Fig. 10e, lower panels). To validate these results, we then investigated the localization of OsMS1 and OsMS1 wenmin1 using stable rice transgenic plants expressing OsMS1-FLAG and OsMS1 wenmin1 -FLAG under native OsMS1 promoter. Consistent with the results in yeast, N. benthamiana, and Arabidopsis, we found that OsMS1 was exclusively detected in the nuclei, while OsMS1 wenmin1 was detected in both the cytoplasm and nuclei ( Supplementary Fig. 10f). Together, these results indicate that the residue at position 301 (L301) is important for the nuclear localization of OsMS1, and also suggest that the L301P mutation affects the function of OsMS1, at least in part, by influencing its subcellular localization.
Temperature regulates the abundances of OsMS1 and OsMS1 wenmin1 . Besides the histone binding domain, LXXLL motif also functions in transcriptional regulation in mammals 32,39 . Notably, the L301P substitution in OsMS1 wenmin1 happens in the extremely conserved LXXLL motif ( Fig. 1e and Supplementary Fig. 7). We therefore tested whether the L301P mutation in OsMS1 influences its transcriptional activation activity in yeast cell. Using BD-OsMS1 and BD-GFP-OsMS1 as baits and AD as prey, we found that OsMS1 and GFP-OsMS1 had the transcriptional activation activity in yeast cells (Fig. 3a). Furthermore, the PHD domain-bearing C-terminal region, but not the N-terminus, is necessary for the transcriptional activation of OsMS1 ( Supplementary Fig. 11). Interestingly, we observed that yeast cells with BD-OsMS1 wenmin1 grew slowly at 22°C, but had no obvious growth at 30°C (Fig. 3a and Supplementary  Fig. 12a), indicating a temperature-dependent transcriptional activation activity of OsMS1 wenmin1 in yeast cells. These results suggest that the L310P mutation in OsMS1 wenmin1 could suppress the transcriptional activation activity and/or decrease its protein abundance in a temperature-dependent manner. Consistent with this hypothesis, we observed that OsMS1 and OsMS1 wenmin1 protein abundance is sensitive to ambient temperatures in yeast cells (Fig. 3b). Compared with OsMS1, high temperature caused less OsMS1 wenmin1 accumulation in the nuclei ( Fig. 3b and Supplementary Fig. 12b). Thus, it is possible that the L301P mutation in OsMS1 wenmin1 reduces its protein abundance in the nuclei, thereby decreasing the transcriptional activation activity in yeast cells, which is consistent with the fact that GFP-OsMS1 was detected in nuclei, while GFP-OsMS1 wenmin1 was distributed in both nuclei and cytoplasm ( Fig. 3b and Supplementary Fig. 10b). However, we could not fully exclude the possibility that the L301P mutation may directly repress the transcriptional activation activity in yeast cells.
We next asked whether the abundances of OsMS1 and OsMS1 wenmin1 proteins are temperature-dependent in planta.
We examined GFP-OsMS1 and GFP-OsMS1 wenmin1 protein levels using a transient expression assay in N. benthamiana. GFP signals of GFP-OsMS1 and GFP-OsMS1 wenmin1 were readily detectable at 22°C, but became less abundant at 30°C (Fig. 3c, d), although their transcript levels were similar ( Supplementary  Fig. 13a). Notably, GFP signals of GFP-OsMS1 wenmin1 in nuclei were obviously weaker than that of GFP-OsMS1 in nuclei (Fig. 3c, d and Supplementary Fig. 10c). We then quantified protein levels of OsMS1 and OsMS1 wenmin1 in response to ambient temperature changes using a dual-luciferase system (Fig. 3e). Surprisingly, we found that total luciferase activity of OsMS1-LUC and OsMS1 wenmin1 -LUC at 22°C was similar (Fig. 3f). Considering that the mutation in OsMS1 wenmin1 affects the distribution of OsMS1 wenmin1 in cytoplasm and nuclei, the luciferase activity of OsMS1 wenmin1 -LUC in nuclei should be lower than that of OsMS1-LUC in the nuclei at 22°C. The total luciferase activity of OsMS1 wenmin1 -LUC was significantly lower than that of OsMS1-LUC at 30°C (Fig. 3f). Importantly, the luciferase activity of both OsMS1-LUC and OsMS1 wenmin1 -LUC was decreased at 30°C compared with that at 22°C (Fig. 3f), although OsMS1-LUC and OsMS1 wenmin1 -LUC transcripts were not significantly affected by temperatures (Fig. 3g). We further examined OsMS1 and OsMS1 wenmin1 proteins using pOsMS1:OsMS1-GUS and pOsM-S1:OsMS1 wenmin1 -GUS transgenic rice plants grown at permissive or restrictive temperatures. Consistently, we found that GUS signals in both pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS transgenic lines were detectable at 22°C, but became less apparent at 30°C (Fig. 3h and Supplementary Fig. 14a). Notably, GUS signals in pOsMS1:OsMS1 wenmin1 -GUS transgenic lines were barely detected at 30°C (Fig. 3h, Supplementary Fig. 14a), although transcript levels of OsMS1 and OsMS1 wenmin1 in these transgenic lines were similar (Supplementary Fig. 13b-d). We then quantified protein levels of OsMS1-GUS and OsMS1 wenmin1 -GUS using a GUS antibody. Consistently, we found that the abundances of OsMS1-GUS and OsMS1 wenmin1 -GUS proteins were obviously lower at 30°C than those at 22°C, and that the abundance of OsMS1 wenmin1 -GUS was barely detected at 30°C (Fig. 3i, j and Supplementary Fig. 14b). To further examine the tissue expression pattern, we conducted transverse section analysis using stage 9 anthers of pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS transgenic plants grown at 22°C. Consistent with the previous results 21 , we found that OsMS1-GUS was expressed in tapetum ( Supplementary Fig. 14c). Notably, L301P mutation in OsMS1 did not affects its tapetum expression at 22°C (Supplementary Fig. 14c). Taken together, these results indicate that ambient temperature affects the abundances of OsMS1 and OsMS1 wenmin1 , and OsMS1 wenmin1 is more sensitive to temperature changes than OsMS1.
The interactions of OsMS1 and OsMS1 wenmin1 with TDR are temperature-dependent. Considering that OsMS1 has the transcriptional activation activity (Fig. 3a) and lacks a predicted DNA-binding domain ( Fig. 1e and Supplementary Fig. 5), it is possible that OsMS1 could interact with transcription factors to regulate gene expression. To identify the potential OsMS1interacting proteins, we used the N-terminal fragment of OsMS1 (OsMS1 1-625 ) as the bait to screen a yeast cDNA library derived In vitro histone pull-down assay with the indicated proteins against biotinylated peptides. Proteins were pulled down by strepavidin beads and detected by western blots with antibody against FLAG or GST. b OsMS1 associates with H4 in a LCI assay. The indicated constructs were transiently expressed in N. benthamiana at 22°C for 48 h. Luciferase activity is depicted with false color from low (blue) to high (red). Nluc, N-terminal half of firefly luciferase; Cluc, C-terminal half of firefly luciferase. c OsMS1 associates with H4 in a BiFC assay. Assays were done as in (b). nYFP, N-terminal portion of YFP; cYFP, C-terminal portion of YFP. Black arrows indicate nuclei. Scale bars, 50 μm. d OsMS1 associates with H4 in a semi-in vivo pull-down assay. GST-OsMS1 was incubated with nuclear proteins extracts from stage 9 anthers of rice young panicles. Proteins were pulled down by Glutathione sepharose beads and detected by western blots with antibody against GST or H4. e OsMS1 associates with H4 in a co-immunoprecipitation assay. Proteins extracts from stage 9 anthers of transgenic 35S:GFP-OsMS1 rice was immunoprecipitated with a GFP-Trap ® _A agarose beads and detected by western blots with antibody against GFP or H4. 35S:GFP transgenic rice was used as a negative control. Experiments were repeated more than three times with similar results. Source data are provided as a Source data file.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29648-z ARTICLE NATURE COMMUNICATIONS | (2022) 13:2055 | https://doi.org/10.1038/s41467-022-29648-z | www.nature.com/naturecommunications from rice inflorescence, because the full-length OsMS1 had a selfactivating activity. Interestingly, three independent clones corresponding to TDR were identified in this screen. We further confirmed the interaction of OsMS1 1-625 with the full-length TDR in yeast cells (Fig. 4a). We then observed the interaction between full-length OsMS1 and TDR in planta using a LCI assay in N. benthamiana leaves (Fig. 4b, c and Supplementary Fig. 15a, b). Furthermore, bimolecular fluorescence complementation (BiFC) assays in N. benthamiana leaves revealed that OsMS1 associates with TDR in the nuclei (Fig. 4d), where OsMS1 and TDR were colocalized ( Supplementary Fig. 16). Consistently, coimmunoprecipitation assays showed that OsMS1 associates with TDR in transgenic rice ( Supplementary Fig. 15c). The pull-down assay using in vitro translated GST-tagged OsMS1 and FLAGtagged TDR further demonstrated a direct interaction between OsMS1 and TDR (Fig. 4e). Supporting this, OsMS1/PTC1 and TDR were spatio-temporally co-expressed (Supplementary Considering that OsMS1 wenmin1 contributes to temperaturesensitive male sterility, we next asked whether OsMS1 and OsMS1 wenmin1 could interact with TDR in a temperaturedependent manner. Yeast cells cotransfacted AD with BD-OsMS1 but not BD-OsMS1 wenmin1 can grow in SD-4 plates (Supplementary Fig. 11). Using BD-OsMS1 wenmin1 as bait and AD-TDR as prey, yeast growth was observed on SD-4 plates incubated at 22°C (Fig. 4a), albeit at a lower level. However, we did not observe yeast growth on SD-4 plates incubated at 26°C or 30°C (Fig. 4a). These results indicate that the temperature influences the abundance of OsMS1 wenmin1 , thereby resulting in the reduced interaction between OsMS1 wenmin1 and TDR. Similarly, the temperaturedependent interaction between OsMS1 L304L305AA with mutations in LXXLL motif and TDR was also observed ( Supplementary  Fig. 18). We further used LCI assays to examine the interactions of The GFP-OsMS1 and GFP-OsMS1 wenmin1 protein levels were quantitated and normalized to the mCherry level. P values indicate the significant differences relative to GFP-OsMS1 at 22°C. e Dual-luciferase system. LUC, firefly luciferase. Ren-LUC, Renilla luciferase. f Relative luciferase activity of OsMS1-LUC and OsMS1 wenmin1 -LUC in N. benthamiana plants during the low temperature to high-temperature transition. g OsMS1 and OsMS1 wenmin1 mRNA levels as in (f). h-j Protein levels of OsMS1-GUS and OsMS1 wenmin1 -GUS are temperature-dependent in rice. pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS transgenic plants were grown at 22°C or 30°C in growth chamber. Protein levels of OsMS1-GUS and OsMS1 wenmin1 -GUS were detected by GUS-staining of transgenic plant spikelets at stage 9 (h). OsMS1-GUS (i, left) and OsMS1 wenmin1 -GUS (j, left) proteins were detected by ani-GUS antibody. Quantifications of OsMS1-GUS (i, right) and OsMS1 wenmin1 -GUS (j, right) protein levels were relative to ACTIN. Scale bars, 5 μm (b), 50 μm (c), 1 mm (h).
In d, f, g, i (right) and j (right). Data are presented as means ± s.e.m. (n = 6 biological replicates for d, n = 3 biological replicates for f, g, i, and j). Two-tailed unpaired t-test was used for statistical analysis. n.s., not significant. Source data are provided as a Source data file.
OsMS1 and OsMS1 wenmin1 with TDR in planta. We found that the abundance of the OsMS1 wenmin1 -TDR complex was temperaturedependent, and was also significantly lower than that of the OsMS1-TDR complex at both low and high temperatures (Fig. 4b, c and Supplementary Fig. 19a). Notably, we also observed that the interaction between OsMS1 and TDR was gradually decreased with increasing temperatures (Fig. 4b, c and Supplementary Fig. 19a), although the temperature did not obviously affect the interaction between OsMS1 1-625 and TDR in yeast cells (Fig. 4a). Furthermore, BiFC assays confirmed the temperature-dependent interactions of OsMS1 and OsMS1 wenmin1 with TDR in the nuclei ( Fig. 4d and Supplementary Fig. 19b). Considering that high temperature reduced the levels of OsMS1 and OsMS1 wenmin1 proteins, and that the abundance of TDR proteins is not affected by the temperatures (Supplementary Fig. 20), it is possible that the temperaturedependent interactions of OsMS1 and OsMS1 wenmin1 with TDR are due to the temperature-dependent abundances of OsMS1 and OsMS1 wenmin1 in the nuclei. Supporting this idea, in vitro pulldown assay showed that OsMS1 and OsMS1 wenmin1 had similar ability to bind TDR (Fig. 4e).
The OsMS1-TDR complex activates EAT1 expression. To identify the downstream factors of the OsMS1-TDR complex in regulating male fertility, we examined the expression levels of several genes involved in male fertility regulation at different stages of anther development 10 . We found that expression levels of EAT1, CP1, and C6 in ZJ1-wenmin1 were significantly reduced in comparison to those in ZJ1 under restrictive temperature ( Supplementary Fig. 21). Importantly, expression of EAT1, CP1, and C6 was also downregulated in tdr mutant 17,19 , supporting that OsMS1 and TDR share common downstream targets. TDR has been known to bind E-box elements (CANNTG) to activate gene expression 17 . Although EAT1 has been proposed to act downstream of TDR to regulate tapetal development, and mutation in TDR resulted in the decreased expression of EAT1 18 , it is unclear whether TDR could directly bind the promoter of EAT1. To test this, we examined the promoter sequence of EAT1 and found putative TDR binding elements (CANNTG) (Fig. 5a). We subsequently performed electrophoretic mobility shift assays (EMSAs) to test whether TDR could bind the E-boxes in the EAT1 promoter. We expressed MBP-conjugated OsMS1 (MBP-OsMS1), TDR (MBP-TDR), or MBP in Escherichia coli. We found that MBP-TDR but not MBP-OsMS1 alone can bind directly to the promoter of EAT1 in vitro (Fig. 5b). Moreover, excess unlabeled probe can compete with the corresponding labeled probe and eliminate the observed shift (Fig. 5b), indicating the specificity of the interaction. These data support a GST-OsMS1   direct interaction between TDR and the promoter of EAT1 via E-box elements. We further investigated the effects of OsMS1 and OsMS1 wenmin1 on the binding of TDR to the E-box elements in the EAT1 promoter. We found that OsMS1 enhances TDR's binding to the EAT1 promoter (Fig. 5c). Notably, OsMS1 wenmin1 exhibited similar but a bit weaker ability than OsMS1 to enhance TDR's binding to the EAT1 promoter (Fig. 5c). Given the fact that OsMS1 and TDR are tapetum specific expressed genes (Supplementary Figs. 3) 17,21 , we performed chromatin immunoprecipitation (ChIP)-qPCR assays in rice leaf protoplasts to investigate the effect of OsMS1 on the binding of TDR to the EAT1 promoter. We transiently expressed MYC-OsMS1 and MYC-TDR in rice protoplasts and detected the association of OsMS1 and TDR with the EAT1 promoter. ChIP-qPCR results showed that MYC-TDR, but not MYC-OsMS1 alone associated with the EAT1 promoter ( Fig. 5d and Supplementary Fig. 22a). When MYC-OsMS1 was co-expressed with GFP-TDR, MYC-OsMS1 associated with the EAT1 -1347

Wildtype probe
Mutant probe OsMS1 enhances the binding of TDR to the EAT1 promoter. a Schematic representation of partial structure of the EAT1 locus. The conserved E-box elements (CANNTG) are shown in black boxes. The mutated sequences are shown in red. P1 and P2 represent DNA fragments used for ChIP-qPCR analysis. P1 fragment contains two E-box elements, while P2 fragment contains no E-box element. b EMSA assay reveals that TDR but not OsMS1 directly binds to the EAT1 promoter in vitro. The wild-type and mutated probes were shown in (a). MBP was used as a negative control. 10-, 20-,100-fold excess unlabeled probes were used as competitors. c EMSA assay showing the effects of OsMS1 and OsMS1 wenmin1 on the binding of TDR to the E-box elements in the EAT1 promoter in vitro. Biotin-labeled probes were incubated with a constant amount of MBP-TDR protein with increasing amounts of MBP-OsMS1, MBP-OsMS1 wenmin1 or MBP. MBP was used as a negative control. d, e ChIP-qPCR assays showing the effects of OsMS1 and OsMS1 wenmin1 on the binding of TDR to the EAT1 promoter in rice leaf protoplasts (n = 3 biological repeats). Chromatins were incubated with anti-MYC antibody, and precipitated by protein A + G magnetic beads. ChIP-qPCR results were quantified by normalization of the MYC-IP signal with the corresponding input signal (IP/input), and different letters represent significant differences (P < 0.01; Duncan's multiple range test). f Temperature-dependent enrichment of OsMS1 and OsMS1 wenmin1 at the EAT1 promoter by ChIP assays in rice stage 9 anthers. pOsMS1:OsMS1-FLAG and pOsMS1:OsMS1 wenmin1 -FLAG transgenic lines were grown at 22°C or 30°C in growth chamber. Chromatins were incubated with anti-FLAG antibody, and precipitated by protein A + G magnetic beads. ChIP-qPCR results were quantified by normalization of the FLAG-IP signal with the corresponding input signal (IP/input). P values indicate the significant differences relative to pOsMS1:OsMS1-FLAG at 22°C. g mRNA levels of OsMS1-FLAG and OsMS1 wenmin1 -FLAG in (f). h qPCR analysis of EAT1 expressions in ZJ1 and ZJ1-wenmin1 stage 9 anthers. ZJ1 and ZJ1-wenmin1 were grown at 22°C or 30°C in growth chamber. For d-h, values are means ± s.e.m. (n = 3 biological repeats). For f-h, two-tailed unpaired t-test was used for statistical analysis. n.s., not significant. Experiments in b and c were repeated three times with similar results. Source data are provided as a Source data file.
promoter ( Supplementary Fig. 22c, d), indicating that the association of OsMS1 with the EAT1 promoter depends on the TDR. In addition, we observed that OsMS1 enhanced the association of TDR with the EAT1 promoter, and OsMS1 wenmin1 exhibited similar but a bit weaker ability than OsMS1 to enhance the association of TDR with the EAT1 promoter in rice protoplasts ( Fig. 5e and Supplementary Fig. 22b).
Finally, we performed ChIP-qPCR using pOsMS1:OsMS1-FLAG and pOsMS1:OsMS1 wenmin1 -FLAG transgenic rice plants to test if OsMS1 and OsMS1 wenmin1 were enriched at the promoter region of EAT1 at different temperatures. We found that both OsMS1 and OsMS1 wenmin1 were enriched more at the EAT1 promoter at 22°C than 30°C (Fig. 5f, g). Notably, OsMS1 wenmin1 was barely enriched at the EAT1 promoter region at 30°C compared with OsMS1 ( Fig. 5f, g). We further examined the expression of EAT1 in ZJ1 and ZJ1-wenmin1 at different temperatures. We found that the expression of EAT1 in ZJ1 and ZJ1-wenmin1 were comparable at 22°C, while expression of EAT1 was significantly decreased in ZJ1-wenmin1 at 30°C compared with that in ZJ1 (Fig. 5h), indicating that OsMS1 wenmin1 influences the expression of EAT1 in a temperature-dependent manner.

Discussion
Male sterile lines are widely used for crop hybrid breeding, contributing greatly to food security. In rice, CMS lines and TGMS/PGMS lines are extensively utilized to explore hybrid vigor in three-line system and two-line system, respectively, which dramatically increase rice grain yield. As male reproduction is temperature-vulnerable in flowering plants 9 , rice TGMS lines give a good system to understand how plants sense and respond to changing ambient temperature. In this study, we discovered that the natural allele OsMS1 wenmin1 confers temperature-sensitive male sterility in rice, and OsMS1 is a PHDfinger protein that interacts with histones. Consistent with its role, OsMS1 is localized in nuclei. By contrast, OsMS1 wenmin1 is localized in both cytoplasm and nuclei, indicating that the mutation in OsMS1 wenmin1 influences its subcellular localization. We further found that temperatures regulate the abundances of OsMS1 and OsMS1 wenmin1 , and OsMS1 wenmin1 is more sensitive to temperature changes than OsMS1. At permissive temperatures, the wild-type and the OsMS1 wenmin1 allele possess suitable levels of OsMS1 and OsMS1 wenmin1 proteins to keep normal pollen development, respectively, although the level of OsMS1 wenmin1 in the OsMS1 wenmin1 allele is lower in nuclei than that of OsMS1 in the wild type. Both OsMS1 and OsMS1 wenmin1 interact with the transcription factor TDR to activate the expression of downstream genes, thereby generating fertile pollens. At restrictive temperatures, the abundance of OsMS1 proteins was decreased, but there are still enough OsMS1 proteins in nuclei to interact with TDR to activate the expression of downstream genes, therefore producing fertile pollens. In contrast, the high temperatures much strongly reduce the OsMS1 wenmin1 protein levels, thus there are no enough OsMS1 wenmin1 proteins in nuclei to interact with TDR, resulting in a dramatic decrease of downstream gene expression and forming sterile pollens. Thus, our findings uncover an important thermosensitive mechanism by which OsMS1 interacts with the transcription factor to control male fertility in a temperature-dependent manner (Supplementary Fig. 23).
As sessile organisms, plant cells must be capable of adapting to various environmental signals, especially for ambient temperature [40][41][42][43][44][45][46] . In plants, several proteins have been described to respond to ambient temperature 40,41,45,46 . Factors that sense and respond to temperature have been proposed to exist in anywhere in the cell 47 , and different tissues may also possess distinct mechanisms to detect temperature changes 43,44,47 . We found that temperature controls the abundance of OsMS1, which is specifically expressed in anthers. The natural allele OsMS1 wenmin1 causes temperature-dependent male sterility. Thus, our results indicate that temperature-dependent abundance of OsMS1 in nuclei is crucial for tissue-specific response to temperature changes.
The LXXLL motif was originally reported to be a signature sequence that mediates the interactions of different proteins with nuclear receptors, and was proposed to be a defining feature of a new family of nuclear proteins in human 48 . In animals, it was reported that LXXLL motifs influence nuclear localizations by interacting with nuclear proteins 32,38 . For example, the LXXLL motif in the N-terminal repeat region of Dax-1 plays a crucial role in its nuclear localization through interacting with the nuclear protein Ad4BP/SF-1 38 . Interestingly, the L301P mutation in the LXXLL motif of OsMS1 wenmin1 affects its subcellular localization. It will be interesting to investigate whether OsMS1 could interact with nuclear proteins to determine its nuclear localization and why the L301P mutation affects the nuclear localization of OsMS1 wenmin1 in the future.
LXXLL motifs are present in many transcription factors and cofactors, and involved in activating or repressing transcription 32 . The L301P mutation in the LXXLL motif of OsMS1 wenmin1 decreases its transcriptional activity (Fig. 3a, Supplementary  Figs. 11 and 12a). Considering that the L301P mutation caused the reduced abundance of OsMS1 wenmin1 in a temperaturedependent manner, it is plausible that the changes in the subcellular localization and abundance of OsMS1 wenmin1 cause the decreased transcriptional activity. LXXLL motif was reported to mediate interactions between E proteins and p300/CBP histone acetyltransferases (HATs) 39 . Notably, the L301P substitution in OsMS1 wenmin1 happens in the extremely conserved LXXLL motif ( Fig. 1e and Supplementary Fig. 7). Considering that OsMS1 interacts with histones ( Fig. 2a and Supplementary Fig. 9), we then conducted a mononuclease (MNase) digestion-qPCR assay to test whether OsMS1 could affect the chromatin status of the EAT1 promoter. As shown in Supplementary Fig. 24a, b, we found that the OsMS1 wenmin1 allele slightly influences the chromatin status of the EAT1 promoter. It seems plausible that OsMS1 might interact with histone modifiers (e.g., histone acetyltransferases) to influence the chromatin status and gene expression. In addition, a previous study showed that OsMS1 interacts with TIP2/bHLH142 in Y2H and BiFC assays 23 , suggesting a possibility that OsMS1 may interact with multiple transcription factors to regulate gene expression.
Hengliangyouyihao bred from the indica HNS-1 with the OsMS1 wenmin1 allele and MH63 was once the most widely planted two-line hybrid rice in 1990s in China (http://www.ricedata.cn/). However, the relatively higher male sterility-fertility transition temperature of HNS-1 restricted its further application 31 . Rice breeders prefer to use TGMS lines that had 22-24°C transition temperature in practical two-line hybrid rice breeding 7,49,50 , because higher transition temperature usually causes selfpollination of TGMS lines during hybrid rice seed production under fluctuating environmental temperature conditions. We observed the transition temperatures of the OsMS1 wenmin1 allele depend on genetic backgrounds. The japonica ZJ1-wenmin1 confers temperature-sensitive genic male sterility with significantly decreased sterility-fertility transition temperature at 23°C (Fig. 1a), indicating that it has great potential as a better TGMS line for hybrid rice breeding. Considering that Leu 301 of OsMS1 is evolutionarily conserved in plant species, it opens a way to create TGMS lines by generating the L301P mutation in various rice cultivars and other crops using genome editing technology.

Methods
Plant materials and growth conditions. The rice parents, transgenic materials, and NILs used in this study were listed in Supplementary Table 1 For transverse section analysis of OsMS1-GUS and OsMS1 wenmin1 -GUS in rice anther, pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS transgenic plants were grown at 22°C or 30°C in growth chamber.
Arabidopsis thaliana ecotype Col-0 was used as wild-type control. Arabidopsis plants were grown on soil in a growth chamber at 22°C ± 2°C under long-day photoperiod (16 h light/8 h dark) with light intensity of 120 μmol photons m −2 s −1 .
To grow Arabidopsis seedlings on MS medium, seeds were surface-sterilized for 10 min in 15% commercial bleach, washed thoroughly with sterile water, and plated on half-strength MS medium with 1% sucrose and 0.8% agar. Plants were stratified at 4°C for 2 days in the dark and then transferred to a growth chamber at the indicated times at 22°C ± 2°C under long-day photoperiod (16 h light/8 h dark) with light intensity of 120 μmol photons m −2 s −1 .
N. benthamiana plants were grown on soil in a growth chamber at 22°C ± 2°C under long-day photoperiod (16 h light/8 h dark) with light intensity of 120 μmol photons m −2 s −1 . Five-to six-week-old N. benthamiana plants were used for transient expression assays.
Yeast strains and growth conditions. Saccharomyces cerevisiae strain AH109 was cultured in yeast extract peptone dextrose (YPD) media at the indicated temperatures. The presence of transgenes was confirmed by growth on SD/-Trp/-Leu plates (Clontech, Cat. No: 630317). The transformed yeasts were suspended in 0.9% NaCl to OD 600 = 1.0. Then, 3 μL of suspended yeast was spotted onto the indicated selection medium.
Bacterial strains and growth conditions. The Agrobacterium strain GV3101 was incubated overnight and resuspended in the activation buffer (10 mM MES, pH 5.6, 10 mM MgCl 2 , and 150 µM acetosyringone). The desired Agrobacterium combinations were mixed and incubated at room temperature for at least 2 h by gentle rocking with the final concentration of each construct at OD 600 = 1. The activated Agrobacterium strains were then infiltrated into fully expanded N. benthamiana leaves through a 1 mL needleless syringe.
E. coli BL21(DE3) transformed with the indicated constructs were grown to a density OD 600 of 0.4-0.5 on LB supplemented with 100 µg/mL of ampicillin or 50 µg/mL Kanamycin. Protein expression was then induced by addition adding 0.1 mM IPTG and incubated 3 h at 37°C with agitation.
Generation of ZJ1-wenmin1. A NIL for OsMS1 wenmin1 (ZJ1-wenmin1) was generated by backcrossing Tian1S with a wide compatibility japonica variety ZJ1 as the recurrent parent. The ZJ1-wenmin1 was selected from the BC 6 F 2 population based on pollen fertility and marker-assisted selection.  Supplementary  Table 3.
Pollen fertility analysis. For pollen fertility analysis in Fig. 1a, 100 anthers of 20 mature spikelets from four plants in heading date were sampled just before their blooming. All anthers were taken out and stained with 1% potassium iodide-iodine solution (I 2 -KI). Pollen grains were released using a needle and debris was removed. More than 200 pollen grains were photographed in randomly picked microscopic fields under a Leica DM2500 microscope (Leica Microsystems) using a ×20 magnification objective to evaluate pollen fertility. Darkly stained and round pollen grains were defined as fertile, whereas unstained, lightly or incompletely stained, irregularly-shaped grains were defined as sterile. For detecting the pollen fertility of Tian1S at permissive temperature (22°C) and restrictive temperature (30°C) under photoperiod conditions of short day (10 h light and 14 h dark) and long day (14 h light and 10 h dark) in Supplementary Fig. 1b, temperature profile set of the growth chamber was shown in Supplementary Table 4. For examining pollen fertility in Fig. 1a, ZJ1, Tian1S, and ZJ1-wenmin1 were planted at the indicated different day average temperature (DAT) under photoperiod conditions of 12 h light and 12 h dark in growth chamber. For statistical analysis of Fig. 1a in Fig. 1b, ten repeats of pollen fertility experiments were conducted for ZJ1, Tian1S, and ZJ1-wenmin1, each with ten anthers being examined. For examining pollen fertility in Fig. 1f, ZJ1, ZJ1-wenmin1, and gOsMS1:ZJ1-wenmin1 lines (#1 and #2) were planted at natural high temperature (~30°C, DAT).
Map-based cloning of wenmin1. Tian1S is a thermosensitive male sterile indica variety. To identify the causal wenmin1 gene, we conducted a map-based cloning by crossing Tian1S with a wide compatibility japonica variety ZJ1. The wenmin1 was first delimited to a genomic interval between the SSR markers RM6235 and RM3600 on chromosome 9, and was then narrowed down to the markers RM24453 and A09-13, within a region covered by one BAC (AP005308). Three genes are located in this mapping region. Schematic structure of LOC_Os09g27620 showed that LOC_Os09g27620 has three exons and two introns. DNA sequencing showed that T902C mutation in LOC_Os09g27620 is responsible for the temperature-sensitive male sterility phenotype of Tian1S and ZJ1-wenmin1.
Notably, the OsMS1 wenmin1 mutation creates an Hpa II restriction site in the OsMS1 coding sequence. To distinguish OsMS1 and OsMS1 wenmin1 , genomic DNA was isolated from rice seedlings by CTAB method. The resulting DNA was used in a 20 μL PCR reaction with the primers A09-37F and A09-37R. The PCR products were digested with Hpa II and then separated using 2% agarose gel electrophoresis. The Hpa II digested PCR products of OsMS1 and OsMS1 wenmin1 were 168 bp and 87 bp, respectively.
Allelism and functional complementation test. Hengnong S-1 (HNS-1), bred by the Hengyang Academy of Agricultural Sciences, was a widely used thermosensitive male sterile indica cultivar in the two-line hybrid breeding system in China (http://www.ricedata.cn/) 52 . To determine whether OsMS1 wenmin1 in Tian1S was allelic to the causal mutation in HNS-1, we performed an allelic noncomplementation assay in the F 1 generation of a cross between Tian1S and HNS-1. To examine whether OsMS1 wenmin1 is allelic to tms5, we crossed Tian1S with Annong S-1 and Zhu1S, both of which are indica and carry tms5 5 . We also crossed Tian1S with 93-11 as a control. As shown in Supplementary  (19°54′N, 110°33′E), Hainan Province when panicle development started in the summer. Male fertility was assayed as the seed setting rate, and plants with a seed setting rate lower than 10% were considered sterile.
For the functional complementation test, a fragment of 5.1 kb in length (designated gOsMS1), containing the entire OsMS1-coding region, along with 5′ and 3′ flanking regions was amplified from 93-11 genomic DNA using primers gOsMS1-F and gOsMS1-R, and cloned into the binary vector pMDC99. All constructs were sequencing-verified at TSINGKE Biological Technology (Beijing, China), and were individually transformed into Agrobacterium tumefaciens strain GV3101. gOsMS1 was used to transform the japonica TGMS line ZJ1-wenmin1. We chose ZJ1-wenmin1 for transformation because the indica variety Tian1S was difficult to transform. The presence of transgenes was determined by genomic PCR using oligonucleotide primers specific to the transgene constructs. We analyzed at least 10 independent lines.
Phylogenetic analysis of OsMS1 and its homologs. The protein sequence of OsMS1 was used to identify its homologs via BLAST search. A condensed phylogenetic tree was constructed with MEGA software (version 5.2) 55  Gene expression pattern analysis. Tissue-specific expression patterns of OsMS1 and TDR were obtained from Botany Array Resource (http://bar.utoronto.ca/ efprice/cgi-bin/efpWeb.cgi) 56 . The signal threshold was set to 1000 for all data extraction. The expression specificity of OsMS1 and TDR were also obtained from Rice Expression Profile Database (RiceXPro) (https://ricexpro.dna.affrc.go.jp/). Plasmid construction and plant transformation. All constructs used in this study were generated using an EZfusion kit (Genera, Cat. No: GR6086). To generate the plasmid 35S:GFP-OsMS1, the full-length open reading frame (ORF) of OsMS1 was amplified with the primers GFP-OsMS1-F and GFP-OsMS1-R from young panicles of indica variety 93-11 and cloned into pMDC43 vector. To generate the plasmid 35S:GFP-OsMS1 wenmin1 , the ORF of OsMS1 wenmin1 was amplified with the primers GFP-OsMS1-F and GFP-OsMS1-R from young panicles of Tian1S and cloned into the pMDC43 vector. To generate the plasmid 35S:MYC-TDR, the ORF of TDR was amplified with the primers MYC-TDR-F and MYC-TDR-R from young panicles of ZH11 and cloned into the pCAMBIA1300-221-MYC vector. To generate the plasmid gOsMS1-GUS, the genomic fragment of OsMS1 was amplified with the primers gOsMS1-F and gOsMS1-R from young panicles of 93-11 and cloned into the pMDC164 vector. To generate the plasmid pOsMS1:OsMS1-GUS and pOsM-S1:OsMS1 wenmin1 -GUS, the full-length ORF of OsMS1 and OsMS1 wenmin1 was amplified and fused in frame with pOsMS1 and then cloned into the pMDC164 vector. 35S:GFP-OsMS1, 35S:GFP-OsMS1 wenmin1 , 35S:MYC-TDR, pOsMS1:OsMS1-GUS, and pOsMS1:OsMS1 wenmin1 -GUS were introduced into ZH11 via Agrobacterium-mediated transformation 57 . 35S:GFP-OsMS1 and 35S:GFP-OsMS1 wenmin1 were also introduced into wild-type Arabidopsis Col-0 via Agrobacterium-mediated transformation using standard protocol 58 . We analyzed at least 10 independent lines for each construct. Primers are summarized in Supplementary Data 1.  Table 6.
Histone pull-down assay. To generate the plasmid FLAG-OsMS1, the ORF of OsMS1 was amplified with the primers FLAG-OsMS1-F and FLAG-OsMS1-R and cloned into the pET-nT vector. To generate the plasmid FLAG-OsMS1 wenmin1 , the ORF of OsMS1 wenmin1 was amplified with the primers FLAG-OsMS1-F and FLAG-OsMS1-R and cloned into the pET-nT vector. To generate the plasmid GST-OsMS1C, the ORF of OsMS1C was amplified with the primers GST-OsMS1C-F and GST-OsMS1C-R and cloned into the pGEX4T-1 vector. To generate the plasmid FLAG-OsMS1C with the mutations C642A, C645A, H650A, C653A, the synthesized mutated OsMS1C was amplified with the primers FLAG-OsMS1C-F and FLAG-OsMS1C-R and cloned into the pET-nT vector. To generate the plasmid FLAG-GST, the ORF of GST was amplified with the primers FLAG-GST-F and FLAG-GST-R and cloned into the pET-nT vector. To generate the plasmid FLAG-GFP, the ORF of GFP was amplified with the primers FLAG-GFP-F and FLAG-GFP-R and cloned into the pET-nT vector. The peptide pull-down assays were performed in a three hundred microliters system 59  Semi-in vivo glutathione S-transferase pull-down assay. To generate the plasmid GST-OsMS1, the ORF of OsMS1 was amplified with the primers GST-OsMS1-F and GST-OsMS1-R and cloned into the pGEX4T-1 vector. Primers are summarized in Supplementary Data 1. To examine whether OsMS1 interacts with H4 in Fig. 2d, we conducted a semi-in vivo glutathione S-transferase (GST) pull-down assay using nuclear proteins of stage 9 rice anthers. In brief, purified GST-OsMS1 was incubated with nuclear proteins extracts from stage 9 anthers of rice young panicles. Proteins were pulled down by Glutathione Sepharose 4B (GE Healthcare, Cat. No: 45-000-139) and detected by western blots using anti-H4 antibody (Active motif, Cat. No: 61521, Clone: MABI 0400, dilution 1:5000) or anti-GST antibody (Abmart, Cat. No: M20007, dilution 1:5000). GST was used as a negative control.
Transcription activation assay. Transcription activation assay was performed using Matchmaker GAL4 Two-Hybrid System 3 (Clontech, USA) following the manufacturer's instructions. The full-length and a series of truncations or mutations of OsMS1 were amplified and fused with the DNA-binding domain (BD) in vector pGBKT7 (Clontech, USA). All the resulting constructs were sequencing-verified and cotransformed with pGADT7 (Clontech, USA) into Saccharomyces cerevisiae strain AH109. The presence of transgenes was confirmed by growth on SD/-Trp/-Leu plates (Clontech, Cat. No: 630317). The transformed yeasts were suspended in 0.9% NaCl at OD 600 = 1.0. Then, 3 μL of suspended yeast was spread into SD/-Trp/-Leu/-His (Clontech, Cat. No: 630319) + 2 mM 3-aminotriazole (Sigma-Aldrich, Cat. No: A8056) or SD/-Trp/-Leu/-His/-Ade (Clontech, Cat. No: 630323) plates. Yeast growth was observed after incubation at the indicated temperatures. For detecting the transcriptional activation activity in Fig. 3a, yeast cells were spotted on SD-2 or SD-3 plus 2 mM 3-aminotriazole (3-AT) plates and grown at low (22°C) or high (30°C) temperatures for 4d. For quantifying the transactivation activities in Fig. 3a, we conducted yeast liquid assays. Equal numbers of diploid yeast cells were cultured in the SD-2 or SD-3 liquid medium at 22°C (OD 600 = 0.05) and 30°C (OD 600 = 0.01) for the indicated time. Relative yeast growth was shown as (OD 600 in SD-3)/(OD 600 in SD-2).
Yeast two-hybrid assay. Yeast two-hybrids were performed using Matchmaker GAL4 Two-Hybrid System 3 following the manufacturer's instructions. As the fulllength OsMS1 autoactivated the reporter gene when fused with BD, we used the N-terminal fragment of OsMS1 (referred to as OsMS1 1-625 , amino acids 1-625) as bait to screen a cDNA library derived from rice inflorescence.
To check for the interaction between OsMS1 1-625 , OsMS1 wenmin1 , OsMS1 L304L305AA , and TDR, the coding sequence of OsMS1 wenmin1 , OsMS1 L304L305AA , and OsMS1 1-625 were fused with the BD in pGBKT7 (Clontech, Cat. No: 630443), and the coding sequence of TDR was fused with the DNA-activation domain (AD) in pGADT7 (Clontech, Cat. No: 630442). Constructs for testing interactions were cotransformed into yeast strain AH109. The presence of transgenes was confirmed by growth on SD/-Trp/-Leu plates (Clontech, Cat. No: 630317). The transformed yeasts were suspended in 0.9% NaCl to OD = 1.0. Then, 3 μL of suspended yeast was spotted onto a stringent selection medium SD/-Trp/-Leu/-His/-Ade (Clontech, Cat. No: 630323). Interactions were observed after incubation at the indicated temperatures. For detecting the temperature-dependent association of OsMS1 wenmin1 and TDR in Fig. 4a, yeast cells (OD 600 = 1) were spotted on the SD-4 plates and grown at 22°C, 26°C, and 30°C for 96 h.
To quantify the temperature-dependent association of OsMS1 wenmin1 and OsMS1 L304L305AA with TDR in yeast in Supplementary Fig. 18a, equal amounts of yeast diploid cells were cultured in 10 mL liquid SD-4 (Clontech, Cat. No: 630322) medium at 22°C, 26°C, and 30°C for about 96 h until yeast cells coexpressing OsMS1 1-625 -BD and TDR-AD were grown to an OD 600 of~2.0. Yeast growth were shown as OD 600 and normalized to yeast cells coexpressing OsMS1 1-625 -BD and TDR-AD.
Luciferase complementation imaging assay. For confirming OsMS1-interacting proteins, coding sequences of OsMS1 were cloned into vector pCAMBIA1300-35S-nLUC-RBS 60 using primers OsMS1-Nluc-F and OsMS1-Nluc-R. The full-length ORF of TDR was cloned into the vector pCAMBIA1300-35S-cLUC-RBS using primers Cluc-TDR-F and Cluc-TDR-R. All the constructs were introduced into Agrobacterium tumefaciens strain GV3101. Primers are summarized in Supplementary Data 1. The LCI assay was carried out in N. benthamiana leaves 61,62 . In brief, Agrobacterium strains GV3101 with individual expression constructs were incubated overnight and resuspended in the activation buffer (10 mM MES, pH 5.6, 10 mM MgCl 2 and 150 µM acetosyringone). The desired Agrobacterium combinations were mixed and incubated at room temperature for at least 2 h by gentle rocking with the final concentration of each construct at OD 600 = 1. The activated Agrobacterium strains were then infiltrated into fully expanded N. benthamiana leaves through a 1 mL needleless syringe. After infiltration, plants were incubated at 22°C for 44 h and then transferred at the indicated temperatures for 4 h until LUC activity was measured. For examining the interactions in Fig. 2b, the indicated Nluc and Cluc construct combinations were transiently expressed in N. benthamiana at 22°C for 2 d. For examining the temperature-dependent associations of OsMS1 and OsMS1 wenmin1 with TDR in LCI assays in Fig. 4b, N. benthamiana plants infiltrated with the Agrobacteria combinations were incubated at 22°C for 44 h and then transferred to 30°C or not for another 4 h.
To detect luciferase activity, 1 mM D-luciferin solution (Promega, Cat. No: E1602) was sprayed onto the leaves of N. benthamiana, followed by incubation in darkness for 5-10 min. A low-light cooled CCD imaging apparatus (NightOWL II LB983 with Indigo software) was used to capture the image. An exposure time of 5 min was used to take images.
To quantify protein-protein interaction intensity, leaf discs infiltrated with the desired Agrobacterium combinations were taken and incubated with 50 μL water containing 1 mM D-luciferin in a Costar ® 96-Well White Flat Bottom Polystyrene High Bind Microplate (Corning, Cat. No: 7400) for 5-10 min. The luminescence activity was measured using the GLOMAX 96 microplate luminometer (Promega, USA). The protein-protein interaction intensity was shown as relative luminescence unit (RLU). To avoid the bias of transfection efficiency, we used at least twelve individual leaves of six tobacco plants for each combination.
The Agrobacterium bacteria were activated and then infiltrated into N. benthamiana plants as LCI assay. After infiltration, plants were incubated at 22°C for 44 h and then transferred to the indicated temperatures. The firefly and Renilla luciferase activity were detected using the Dual-Luciferase® Reporter Assay System (Promega, Cat. No: E1960) in the GLOMAX 96 microplate luminometer (Promega, USA) following the manufacturer's instructions. The firefly luciferase activity was normalized to the Renilla luciferase activity to control for transfection efficiency. The relative firefly luciferase activity reflects the protein abundance of OsMS1, OsMS1 wenmin1 , or TDR at the indicated temperatures. Three independent replicates were performed for each condition.
For the subcellular localization of OsMS1 and OsMS1 wenmin1 in Arabidopsis, root tip cells and cotyledons of 4-day-old 35S:GFP-OsMS1 and 35S:GFP-OsMS1 wenmin1 transgenic Arabidopsis plants were visualized using a laser scanning confocal microscope (Nikon confocal microscope A1R, Japan).
For the subcellular localization of OsMS1 and OsMS1 wenmin1 in N. benthamiana, 35S:GFP-OsMS1 + 35S:mCherry or 35S:GFP-OsMS1 wenmin1 + 35S:mCherry were transiently co-expressed in fully expanded leaves of 6-week-old N. benthamiana as described in LCI assay. The GFP and RFP fluorescence signals were visualized using a laser scanning confocal microscope (Zeiss LSM710 META, Germany) after 48-72 h incubation at 22°C in the dark.
For the subcellular localization of OsMS1 and OsMS1 wenmin1 in rice protoplast, we used the japonica variety ZH11 to prepare and transform rice protoplasts. The plasmids (10 μg per construct) were introduced by PEG-mediated transfection 63 . The GFP and DAPI signals were visualized using a laser scanning confocal microscope (Zeiss LSM710 META, Germany) after 18 h incubation at 28°C in the dark.
To examine whether GFP-OsMS1 and GFP-OsMS1 wenmin1 protein levels are temperature-dependent in Fig. 3c, N. benthamiana expressing 35S:GFP-OsMS1 or 35S:GFP-OsMS1 wenmin1 were grown at 22°C for 44 h then transferred to 30°C or not for another 4 h. The GFP signal in leaves was visualized under a laser scanning confocal microscope (Zeiss LSM710 META, Germany). The excitation wavelength used for DAPI, GFP/YFP, and mCherry is 405 nm, 488 nm, and 543 nm, respectively. For quantitative analysis of Fig. 3c in Fig. 3d, the GFP-OsMS1 and GFP-OsMS1 wenmin1 protein levels were quantified by comparing the GFP fluorescence intensity of the cell expressing 35S:GFP:OsMS1 or 35S:GFP:OsMS1 wenmin1 with that of RFP fluorescence intensity of 35S:mCherry in the same cell. Fluorescence intensity was determined using ImageJ software. All the images in a single experiment were captured under the same setting parameters. Zen Black 2009 (Zeiss) was used to export pictures taken by Zeiss LSM710 META.
Bimolecular fluorescence complementation assays. To generate the plasmid 35S:OsMS1-cYFP and 35S:OsMS1 wenmin1 -cYFP the ORF of OsMS1 and OsMS1 wenmin1 was amplified with the primers OsMS1-YC-F and OsMS1-YC-R and cloned into the pSPYCE(M) vector 64 . To generate the plasmid 35S:H4-nYFP, the ORF of H4 was amplified with the primers H4-YN-F and H4-YN-R and cloned into the pSPYNE173 vector 64 . To generate the plasmid 35S:TDR-nYFP, the ORF of TDR was amplified with the primers TDR-YN-F and TDR-YN-R and cloned into the pSPYNE173 vector 64 . Primers are summarized in Supplementary Data 1. Co-transfection of the indicated constructs (for example, those encoding OsMS1-cYFP and TDR-nYFP) into N. benthamiana leaf epidermal cells was performed as LCI assay. YFP fluorescence was visualized and photographed using a confocal scanning microscope (Zeiss LSM710 META, Germany). The excitation wavelength used for YFP and mCherry is 488 nm and 543 nm, respectively. Each BiFC assay was repeated at least three times. For detecting the interactions of OsMS1 with H4 in Fig. 2c, the indicated nYFP and cYFP construct combinations were transiently expressed in N. benthamiana at 22°C for 48 h. For detecting the temperature-dependent interactions of OsMS1 and OsMS1 wenmin1 with TDR in Fig. 4d, the co-infiltrated tobacco plants were incubated at 22°C for 44 h and then transferred to 30°C for another 4 h or not.
In vitro pull-down assay. To produce GST-tagged OsMS1 and OsMS1 wenmin1 proteins, the full-length coding sequences of OsMS1 and OsMS1 wenmin1 were amplified and cloned into the GST vector pGEX-4T-1. To produce FLAG-tagged TDR protein (referred to as FLAG-TDR), the coding sequence of TDR was amplified and cloned into the vector pET-nT. Primers are summarized in Supplementary Data 1.
Rice protoplast preparation. We planted ZH11 in growth chamber for two weeks. Approximately 50-200 leaves at the three-leaf stage were treated with macerozyme and cellulase to prepare protoplast. Briefly, the leaves were cutted into~0.5 mm fragments, and then incubated in an enzyme solution (0.75% Macerozyme R-10, 1.5% Cellulose RS, 0.6 M mannitol, 10 mM MES at pH 5.7, 10 mM CaCl 2 , and 0.1% BSA) for 4-5 h in the dark with gentle rotation. After digestion, the same volume of W5 solution (154 mM NaCl, 5 mM KCl, 125 mM CaCl 2 , and 2 mM MES at pH 5.7) was added, and followed by vigorous shaking for 10-30 s. Protoplasts were collected by filtering through 40 μm nylon mesh with three to five washes using W5 solution. The pellets were collected by centrifugation at 150 × g for 5 min. After washing twice with W5 solution, the pellets were then resuspended in MMG solution (0.4 M mannitol, 4 mM MES at pH 5.7, and 15 mM MgCl 2 ) at a concentration of 2× 10 6 cells mL −1 for subcellular localization or 2 × 10 7 cells mL −1 for ChIP-qPCR. Protoplasts were further transiently transformed with the indicated constructs by PEG-mediated transfection according to a published protocol 63 . Briefly, the indicated DNA plasmids (10 μg) were mixed together with 300 μL protoplasts and the freshly prepared 330 μL polyethylene glycol (PEG) solution (0.4 M mannitol, 40% w/v PEG4000, 0.1 M CaCl 2 ), and then incubated at room temperature for 20 min. After serial dilutions with W5 solution, the transfected protoplasts were resuspended in W5 solution and cultured overnight at 28°C in the dark. To examine the subcellular localization of OsMS1 and OsMS1 wenmin1 in Supplementary Fig. 10f, nuclear and cytoplasmic fractions were prepared from stage 9 anthers of pOsMS1:OsMS1-FLAG and pOsMS1:OsMS1 wenmin1 -FLAG transgenic rice. About 0.5 g anthers were ground into powder in liquid nitrogen and homogenized in 5 mL Honda buffer (20 mM HEPES-KOH at pH 7.4, 1.25% ficoll, 2.5% Dextran T40, 0.44 M sucrose, 10 mM MgCl 2 , 0.5% (v/v) Triton X-100, 5 mM DTT, 1 mM PMSF, Roche protease inhibitor cocktail), and then filtered through 62 μm nylon mesh. Triton X-100 was added to a final concentration of 0.5%, and the mixture was incubated on ice for 15 min. The solution was centrifuged at 1500 × g for 10 min, and the supernatant was used to detect the cytoplasmic protein. The remaining pellets were washed with Honda buffer containing 0.1% (v/v) Triton X-100 and then used to extract the nucleus protein.
Proteins were detected by western blots with antibody against FLAG (Abmart, Cat. No: M20008, dilution, 1:5000), H4 (Active Motif, Cat. No: 61521, dilution, 1:5000) or BIP (Abcam, Cat. No: ab108615, dilution, 1:5000). H4 and BIP were used as a marker for nuclei and cytoplasm, respectively. GUS histochemical staining. Transgenic rice plants were planted in the open field until their inflorescence length reached~0.5 cm, an indicator of anther development stage 3 7 , and then transferred to the growth chamber under the indicated day average temperature (DAT). For GUS-staining, panicles were submerged in buffer containing 50 mM NaPO 4 buffer, 0.4 mM each K 3 Fe(CN) 6 /K 4 Fe(CN) 6 , 0.1% (v/v) Triton X-100, and 1 mM X-Gluc and stained at 37°C until blue color became visible. Then the samples were incubated with 75% ethanol until chlorophyll was removed. The GUS-staining panicles were taken photos using a microscope. Notably, pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS panicles were treated under the same setting parameters. For detecting endogenous OsMS1-GUS proteins in Fig. 1d, pOsMS1:OsMS1-GUS transgenic plants were grown at planted in the open field and then transplanted to 22°C DAT in growth chamber. For detecting endogenous OsMS1-GUS or OsMS1 wenmin1 -GUS proteins in Fig. 3h and Supplementary Fig. 14a, pOsMS1:OsMS1-GUS and pOsMS1:OsMS1 wenmin1 -GUS transgenic plants were grown in the open field and then transplanted to 22°C or 30°C DAT in growth chamber.
Electrophoretic mobility shift assay. To generate the plasmid MBP-OsMS1 and MBP-OsMS1 wenmin1 , the ORF of OsMS1 and OsMS1 wenmin1 was amplified with the primers MBP-OsMS1-F and MBP-OsMS1-R and cloned into the pMAL-c2 vector. To generate the plasmid MBP-TDR, the ORF of TDR was amplified with the primers MBP-TDR-F and MBP-TDR-R and cloned into the pMAL-c2 vector. Primers are summarized in Supplementary Data 1. MBP-OsMS1 and MBP-TDR proteins were expressed and purified as described in in vitro pull-down assay. The oligonucleotide sequences containing E-box sequences in the rice EAT1 promoter were found by manual search in the promoter sequence. Double-stranded biotinlabeled DNA probes were generated by annealing sense and antisense oligonucleotides. Annealing reactions were performed at 95°C for 10 min with 0.5°C decrease per minute until 25°C in a PCR thermocycler.
EMSA was conducted using Chemiluminescent Nucleic Acid Detection Module Kit (Thermo Scientific). EMSA reactions were performed in a 20 μL system containing 10 mM Tris-Hcl pH 7.5, 2.5% glycerol, 1 mM DTT, 50 mM KCl, 5 mM MgCl 2 , 50 ng/μL Poly (dI·dC), and 0.05% NP-40. The binding assays of the indicated proteins to the EAT1 promoter were performed by incubating 10 fmol Biotin-dsDNA probe with 1 μg MBP-TDR or MBP-OsMS1 proteins (1 μL) at room temperature for 30 min. To test whether TDR binds to the EAT1 promoter differently with OsMS1 or OsMS1 wenmin1 , biotin-labeled probes were incubated with a constant amount of MBP-TDR protein with increasing amounts of MBP-OsMS1, MBP-OsMS1 wenmin1 or MBP. Protein-DNA complexes were resolved on native 9% (w/v) polyacrylamide gels at 0.5 × TBE at 100 V for 90 min at 4°C. Electrophoresed protein-DNA complexes were transferred to a Hybond N + nylon membrane (Amersham, GE Healthcare Life Sciences) at 100 V for 60 min using 0.5 × TBE at 4°C. The transferred protein-DNA complexes were further crosslinked to the membrane for 10 min with the membrane facing a transilluminator equipped with 312 nm bulbs. Detection of Biotin-labeled DNA by chemiluminescence was performed following the instructions of the manufacturer.
Chromatin immunoprecipitation and quantitative real-time PCR analysis. The proteins expressed in protoplasts were used for the ChIP assay. In brief, we used ZH11 to prepare rice protoplasts as described in rice protoplast preparation section 63 . The indicated plasmids with final concentration at 5 μg per 100 μL were introduced by PEG-mediated transfection. ChIP experiments were performed according to the published protocol 65 . Briefly, approximately 10 mL protoplasts were cross-linked with 1% (v/v) formaldehyde in ice for 10 min. The crosslink was stopped with the addition of 0.125 M glycine for 5 min. After shearing chromatin into about 500 bp fragments via sonication, anti-MYC antibody (Abcom, Cat. No: ab32, dilution, 1:200) was used to immunoprecipitate genomic DNA fragments. Immunoprecipitated DNA was extracted with phenol/chloroform and precipitated with ethanol. The ChIP experiments were quantified by qPCR in triplicate with appropriate primers (Supplementary Data 1). Empty MYC vector was used as the control. The relative enrichment of promoter DNA was calculated by the % input method with qPCR analysis.
In Fig. 5f, young panicles of pOsMS1:OsMS1-FLAG and pOsMS1:OsMS1 wenmin1 -FLAG transgenic plants grown at 22°C or 30°C in growth chamber were used. Approximately 2 g young panicles with stage 9 anthers of the transgenic lines were ground into powder in liquid nitrogen and cross-linked using extraction buffer with 1% (v/v) formaldehyde in ice for 10 min. The crosslink was stopped with the addition of 0.125 M glycine for 5 min. The chromatin was isolated and the immunoprecipitation assay was conducted as described above. Anti-FLAG antibody (Abcom, Cat. No: ab 205606, dilution, 1:200) was used to immunoprecipitate genomic DNA fragments. The ChIP experiments were quantified by qPCR in triplicate with appropriate primers (Supplementary Data 1). The relative enrichment of promoter DNA was calculated by the % input method with qPCR analysis.
Open chromatin detection by MNase digestion. Open chromatin was detected using a published protocol 66 . In brief,~0.5 g young panicles with stage 9 anthers of ZJ1 and ZJ1-wenmin1 were ground into power in liquid nitrogen and suspended in 10 mL of cold nuclei isolation buffer (1 M hexylene glycol, 20 mM PIPES-KOH at pH 7.6, 10 mM MgCl 2 , 1 mM EGTA, 15 mM NaCl, 0.5 mM spermidine, 0.15 mM spermine, 0.5% Triton X-100, 10 mM β-mercaptoethanol and 1× Roche protease inhibitor cocktail) with gentle rotation for 15 min. The suspension was filtered through 30 nm CellTrics, and the elute was centrifuged for 10 min at 1500 × g at 4°C. The pellets were resuspended as crude nuclei extracts with digestion buffer (40 mM Tris-HCl at pH 7.9, 0.3 M Suc, 10 mM MgSO 4 , 1 mM CaCl 2 , and 1× Roche protease inhibitor cocktail). Equal aliquot of crude nuclei extracts was subject to digestion with increasing MNase amounts at 30°C for 20 min and DNA was extracted using phenol:chloroform treatment followed by ethanol precipitation and used for qPCR analysis.
Statistics and reproducibility. All data are shown as the mean ± s.e.m. unless indicated otherwise. Statistical analysis was performed using GraphPad Prism 7 software (GraphPad Software, Inc. San Diego, CA, USA). All details on statistics have been indicated in figure legends. For all quantitative measurements, a normal distribution was assumed. For comparisons between two groups, the two-tailed unpaired Student's t test was applied. No method was undertaken to calculate statistical power and corresponding sample size. Instead, sample size was determined empirically on the basis of experience with similar assays and from sample NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-29648-z ARTICLE NATURE COMMUNICATIONS | (2022) 13:2055 | https://doi.org/10.1038/s41467-022-29648-z | www.nature.com/naturecommunications sizes generally used by other investigators 5,61 . Images were processed and analyzed with Adobe Photoshop and ImageJ.