Hybrid rice, which yields ~10–20% more grain than conventional rice, has been planted in ~60% (~17 million hectares) of the rice-growing area in China and is also cultivated in many other countries, thus providing a key component of the global food supply1,2,3. Rice is strictly self-pollinating; therefore, hybrid varieties are bred using male-sterile maternal lines that fail to produce viable pollen, thus preventing self-pollination. Hybrid rice breeding uses the well-developed three-line and two-line systems4,5. The three-line system uses a cytoplasmic male-sterile (CMS) line, a restorer line and a CMS maintainer line to produce F1 hybrid seeds and maintain the CMS line6,7,8. Three-line hybrid rice has been grown since the 1970s and is a major type of hybrid rice1,4,9. However, the limited germplasm resources of restorer lines and the genetic diversity between CMS and restorer lines have limited further improvements in three-line breeding10. The two-line breeding system uses thermosensitive genic male sterility (TGMS) lines or photoperiod-sensitive genic male sterility (PGMS) lines as maternal parents to produce hybrid seeds. TGMS and PGMS lines are male-sterile under restrictive conditions (high temperatures for TGMS and long-day for PGMS) but convert to male-fertile under permissive conditions (low temperatures for TGMS and short-day for PGMS); thus, they can self-pollinate under permissive conditions10,11. The TGMS and PGMS traits are controlled by nuclear, recessive genes and most normal rice cultivars can restore male fertility, thus providing broader genetic resources for rice breeding to produce hybrids with strong hybrid vigour10,11,12. Two-line hybrid rice has been planted in China since 1993 and over 3.4 million hectares were grown in 2011 (Supplementary Fig. 1).

AnnongS-1 (AnS-1), the first indica rice (Oryza. sativa ssp. indica) TGMS line, was found in 1987 as a spontaneous mutant13. AnS-1 and its derived TGMS lines have been widely used for two-line hybrid rice breeding, and their TGMS trait is controlled by a single recessive locus, tms5 (ref. 13). A candidate region on the short arm of chromosome 2 has been reported14,15,16,17; however, the identity of the TMS5 gene has not been confirmed and the molecular mechanism by which it produces male sterility remains unclear.

Here we report that tms5 confers the TGMS trait through a loss-of-function mutation in the gene coding for RNase ZS1, which regulates UbL40 mRNA levels during male development. Our results uncover a novel mechanism of RNase ZS1-mediated UbL40 mRNA processing, which controls TGMS in rice and has potential applications in hybrid crop breeding.


AnS-1 and Zhu1S are temperature-sensitive TGMS lines

Zhu1S is one of the indica TGMS lines frequently used for commercial two-line hybrid rice breeding (Supplementary Fig. 1 and Supplementary Table 1). To examine the phenotype of AnS-1 and Zhu1S in detail, growth chambers and Phytotrons were used to grow plants under fixed photoperiod conditions and at different day average temperatures (DAT; See Methods and Supplementary Table 2a,b). AnS-1 and Zhu1S were grown in the field until the panicle length was ~1 cm and then the plants were transferred into growth chambers or Phytotrons for 2–3 weeks, respectively. At the permissive temperature (~23 °C DAT), AnS-1 is male-fertile, similar to the normal male-fertile line AnnongN (AnN, a parental line of AnS-1). At restrictive temperatures, AnS-1 becomes male-sterile, producing abortive pollen grains at ~26 °C DAT, and no pollen grains at 28 °C or higher DAT (Fig. 1a–h and Supplementary Fig. 2a). At 23 °C, AnS-1 developed normally (Fig. 1i–n), whereas at 28 °C few microspore mother cells (MMCs) undergo normal meiosis, and any microspores that are produced eventually disintegrate (Fig. 1o–t). By contrast, the day length had little effect on male sterility of AnS-1 (Supplementary Fig. 2). We observed a similar phenotype in Zhu1S grown at the restrictive temperatures of >23 °C DAT (Supplementary Fig. 3).

Figure 1: Phenotypes and complementation of AnS-1.
figure 1

(ah) Anther morphology and pollen fertility of AnN and AnS-1. Normal anthers (a,b) and pollen (e,f) in AnN and AnS-1 at the permissive temperature (23 °C). Normal anthers (c) and pollen (g) in AnN and abnormal anthers (d) and abortive pollen (h) in AnS-1 grown at the restrictive temperature (28 °C). (it) Comparison of transverse sections of anthers of AnS-1 grown at 23 °C (in) and 28 °C (ot). At 23 °C, the MMCs underwent meiosis to produce microspores (immature pollen grains); at 28 °C, few or no microspores were generated. (i,o) Early MMC stage; (j,p) late MMC stage; (k,q) meiotic leptotene stage; (l,r) meiotic pachytene stage; (m,s) dyad stage; (n,t) young microspore stage. E, epidermis; En, endothecium; ML, middle layer; T, tapetum; MMC, microspore mother cell, Dy, dyad; Msp, microspore. Scale bars, 1 mm (ad), 100 μm (eh) and 20 μm (it). (u) The protein encoded by Os02g0214300 could not be detected in AnS-1 by immunoblot analysis. The constitutively expressed rice Actin protein was used as an internal control. (v) Pollen fertility was restored in the AnS-1 background containing TMS5 genomic DNA (TMS5g-AS) grown at the restrictive temperature (28 °C; above, right). No pollen was produced in anthers of AnS-1 grown under the same conditions (above, left). TMS5 protein was detected in TMS5g-AS line but not in AnS-1 (middle). The Actin protein was used as an internal control.

Cloning the gene associated with TGMS in AnS-1 and Zhu1S

To clone tms5 and the gene conferring the TGMS trait in Zhu1S, we developed and characterized two mapping populations (see Methods, Supplementary Fig. 4a,c), both of which narrowed the target locus to the same region on the short arm of chromosome 2 (Supplementary Fig. 4a,d). Sequencing of the mapped region in the male-sterile lines AnS-1 and Zhu1S revealed a C-to-A transition at position 71 of the gene Os02g0214300 compared with that in AnN, creating a premature stop codon (Supplementary Fig. 4b,e). This change is identical to the premature stop codon mutation mapped to the same gene in the ptgms2-1 locus18. The protein encoded by Os02g0214300 was detected by immunoblotting in AnN, but not in AnS-1 (Fig. 1u). Therefore, we considered Os02g0214300 to be a candidate gene for TMS5.

For genetic complementation to verify that Os02g0214300 indeed corresponds to TMS5, genomic DNA fragments harbouring Os02g0214300 from AnN and the rice cultivar Nipponbare (O. sativa ssp. japonica) were transformed into the tms5 AnS-1 and Zhu1S lines, respectively. Male fertility was restored in transgenic plants of AnS-1 (designated TMS5g-AS) and Zhu1S (designated TMS5g-ZS) at the restrictive temperature (Fig. 1v, Supplementary Fig. 5). Moreover, knockdown of TMS5 by RNA interference (RNAi) in the japonica lines Zhonghua11 (ZH11; designated TMS5i-ZH, Supplementary Fig. 6a–d) and Nipponbare (designated TMS5i-NB, Supplementary Fig. 6e–f) showed similar pollen abortion as in AnS-1 and Zhu1S at restrictive temperatures. These results demonstrate that the loss-of-function mutation of TMS5 confers the TGMS phenotype in AnS-1 and Zhu1S.

tms5 is a major TGMS gene widely used in rice breeding

Dozens of TGMS lines are used in two-line hybrid rice breeding19. To explore whether other TGMS lines also harbour a tms5 mutation, we sequenced the TMS5 locus in 25 other widely used TGMS lines and found that 24 of them carried the same tms5 mutation (Fig. 2a). In 2011, 71 commercial two-line hybrid rice cultivars bred with tms5-containing TGMS lines accounted for at least 71% of all two-line hybrid rice cultivars and 83.8% (~2.9 million hectares) of all the land used to grow two-line hybrid rice in China (Fig. 2b,c and Supplementary Table 3). Accordingly, tms5 serves as the major TGMS genetic resource for two-line hybrid rice breeding.

Figure 2: Application of tms5 in two-line hybrid rice breeding.
figure 2

(a) Left, names of wild-type and TGMS lines. Right, alignment of 75 base-pair DNA sequence of the coding region of TMS5 in wild-type and different TGMS lines. Nucleotide numbers within the TMS5-coding sequence are shown at the top. The mutated nucleotide at position 71 (arrowed) in tms5 lines is shown in red. (b) Number of different two-line hybrid rice cultivars bred using tms5-TGMS lines (tms5-hybrid) compared with other two-line hybrid rice cultivars grown in China in 2011 (left figure). Proportions of the two-line hybrid rice-growing area (B3.4 million hectares) planted with tms5-hybrid cultivars and other hybrid cultivars bred with others and non-sequenced T/PGMS lines in China in 2011 (right figure). (c) Proportions of the two-line hybrid rice-growing area (~3.4 million hectares) planted with tms5-hybrid cultivars and other hybrid cultivars bred with others and non-sequenced T/PGMS lines in China in 2011.

TMS5 encodes a conserved RNase Z protein

TMS5 encodes an evolutionarily conserved ribonuclease Z (RNase Z), which occurs in almost all kingdoms of life20,21,22. Two forms of RNase Z exist in eukaryotes, the short form (RNase ZS) and the long form (RNase ZL; Supplementary Fig. 7), while prokaryotes only have RNase ZS (refs 20, 22). The TMS5 protein, with 302 amino acids, belongs to the short-form group and we refer to it as RNase ZS1. In rice, RNase ZS1 is constitutively expressed in various tissues (Supplementary Fig. 8a). At the protein level, RNase ZS1 accumulated stably at both permissive and restrictive temperatures (Supplementary Fig. 8b). In anther, its mRNA expression was relatively enriched in MMCs and this did not change at low (23 °C) and high (28 °C) temperatures (Supplementary Fig. 8c,d). These results indicate that RNase ZS1 itself is temperature-insensitive.

RNase ZS1 is cytoplasmic and processes tRNAs in vitro

RNase Z proteins process the 3′ ends of tRNAs21,23,24,25. The CCA sequence in 3′ terminal of the tRNA is critical for tRNA maturation and function. RNase Z mainly processes CCA-less pre-tRNAs; however, it can process CCA-containing pre-tRNAs in the nucleus or mitochondria26,27. To investigate the enzymatic activity of RNase ZS1, we tested whether RNase ZS1 cleaved two pre-tRNAs, pre-tRNA9-AspATC (without CCA) and pre-tRNA35-MetCAT (with CCA). We expressed recombinant RNase ZS1 in Escherichia coli (E. coli) and found that it processed the 3′ ends of both pre-tRNAs, with or without CCA, in vitro (Fig. 3a), indicating that RNase ZS1 possesses endonuclease activity. We then examined the in vivo accumulation of mature and precursor tRNA in AnN, AnS-1, ZH11 and TMS5i-ZH but found no obvious difference between wild-type and tms5 plants (Fig. 3b and Supplementary Fig. 9a). The mature tRNA levels were unchanged at permissive or restrictive temperatures when analysed in NIL5 and NIL8 (near-isogenic lines 5 and 8 of Zhu1S in the background of the indica cultivar 93-11) and 93-11 (Supplementary Fig. 9b). These results indicate that RNase ZS1 does not affect tRNA 3′ end processing in vivo.

Figure 3: RNase ZS1 processes tRNA in vitro and localizes in the cytoplasm.
figure 3

(a) Recombinant His-tagged RNase ZS1 processed the 3′ ends of both tRNA9-AspATC-CCA and tRNA35-MetCAT+CCA in vitro. Arrows indicate the processed product bands. (b) RNA blot analysis showed no obvious difference in tRNA precursor accumulation in wild-type (AnN and ZH11), AnS-1 and TMS5i-ZH plants. 5S rRNA/tRNA stained with ethidium bromide was used as a loading control. (c) RNase ZS1-GFP (green) was observed in the cytoplasm (top) but not in the nucleus (middle, stained with DAPI, blue). Scale bars, 10 μm. (d) RNase ZS1 was detected in cytoplasmic protein extracts using immunoblot analysis. Osβ′-COP and OsSRT1 were used as the indicators of cytoplasmic and nuclear protein controls, respectively. CP, cytoplasmic proteins; NP, nuclear proteins.

To explore where RNase ZS1 localizes, we expressed RNase ZS1-GFP (green fluorescent protein) in rice protoplasts and found that it localizes in the cytoplasm but not in the nucleus (Fig. 3c). This observation was further verified by immunoblot analysis of partially purified nuclear and cytoplasmic proteins (Fig. 3d). Therefore, we propose that RNase ZS1 functions in the cytoplasm but does not act in tRNA processing in the nucleus.

UbL40 mRNAs accumulate in tms5 plants at high temperature

To further address whether RNase ZS1 has a role in mRNA metabolism in the cytoplasm, we performed independent whole-genome microarray and RNA-seq analyses, using RNA from young panicles (flower inflorescences) of wild-type lines (AnN, ZH11 and 93-11) and tms5 or RNase ZS1 knockdown lines (AnS-1, TMS5i-ZH, NIL5 and NIL8) grown at permissive and restrictive temperatures. These analyses identified three common mRNAs that accumulate at higher levels in tms5 plants at the restrictive but not the permissive temperature (Fig. 4a,b; Supplementary Fig. 10a and Supplementary Table 4). These mRNAs are from genes of the conserved ubiquitin-60S ribosomal protein L40 family (UbL40), namely UbL401 (Os09g0452700), UbL402 (Os03g0259500) and UbL404 (Os09g0483400; Supplementary Fig. 11). At the permissive temperature, UbL401, UbL402 and UbL404 mRNA levels were similar in wild-type and tms5 plants; at the restrictive temperature, the expression of these three genes, especially UbL401, was dramatically elevated in tms5 but not in wild-type plants (Fig. 4a,b; Supplementary Figs 10a, 12a and 13). Furthermore, UbL401, UbL402 and UbL404 preferentially expressed in stamens and were largely induced by restrictive temperature in tms5 stamens but not in other tissues (Supplementary Fig. 12b–e). Thus, accumulation of UbL401, UbL402 and UbL404 transcripts is temperature-sensitive in the tms5 background, and RNase ZS1-dependent, suggesting that RNase ZS1 might be involved in processing UbL40 mRNAs.

Figure 4: UbL401 and UbL404 mRNAs accumulate in tms5 plants at high temperature.
figure 4

(a,b) Overaccumulation of UbL401 and UbL404 mRNAs in tms5 plants at the restrictive temperature. RNA-seq profile of UbL401 (a) and UbL404 (b) in 93-11, NIL5 and NIL8 (near-isogenic line 5 and line 8 of Zhu1S-derived tms5 in 93-11 background) at 22 and 30 °C. The ubiquitin domain and 60S ribosomal protein L40 domain are shown as blue and orange boxes, respectively. Transcripts of UbL401 (one mRNA isoform) and UbL404 (three mRNA isoforms) are shown with thick boxes representing exons, thin lines representing introns and grey boxes representing untranslated regions (UTRs). Wiggle plots of RNA-seq data represent reads per 10 million (rp10M). (c) In vitro processing of CYP86A1 (left, a negative control), UbL401 (middle) and UbL404 (right) mRNAs by recombinant wild-type RNase ZS1 and enzymatic null (RNase ZS1H81L), respectively. The assay time is marked above the lanes. Arrows indicate bands processed by RNase ZS1.

RNase ZS1 processes mRNAs of UbL40 genes

To address this hypothesis, we transcribed UbL401, UbL402 and UbL404 RNAs in vitro and incubated the RNAs with recombinant RNase ZS1 proteins. All three UbL40 mRNAs were processed into multiple fragments by the wild-type but not the enzymatic null RNase ZS1 proteins in vitro (Fig. 4c, Supplementary Fig. 10b). Furthermore, we employed RNA ligase-mediated 5′ rapid amplification of cDNA ends (RNA ligase-mediated-5′RACE) and mapped multiple degradation sites in the UbL401 mRNA of wild-type plants (ZH11, AnN and 93-11) but not in AnS-1 and NIL5 (Supplementary Fig. 14), suggesting that RNase ZS1 is involved in the processing of UbL401 mRNA in vivo. We hypothesize that RNase ZS1 may recognize the secondary or tertiary structure of the target mRNAs and process them. Thus, we propose that at restrictive temperatures, the loss-of-function mutation of RNase ZS1 in tms5 plants results in unprocessed UbL401, UbL402 and UbL404 RNAs and the excess UbL40 transcripts might lead to abnormal male development in the anther.

UbL40 genes are expressed in MMCs

To further test our hypothesis, we examined the spatial and temporal expression of UbL40 genes by in situ mRNA hybridization. In wild-type anthers, compared with sense probe signals, antisense probes of UbL401 and UbL404 detected stronger signals in MMCs than in other cell layers (Fig. 5a,b). In addition, UbL401, UbL402 and UbL404 expression levels were similar between high (28 °C DAT) and low (23 °C DAT) temperatures in wild type (Fig. 5c and Supplementary Fig. 15). In contrast, all three UbL40 genes tested showed stronger signals in MMCs at high temperature than in the RNase ZS1 mutants at low temperature, indicating the interplay between RNase ZS1 and UbL40 mRNAs in male fertility (Fig. 5d and Supplementary Fig. 15). Among the UbL40 family members, UbL404 may have a major role in the regulation of thermosensitive genic male sterility since the expression level of UbL404 was much higher than that of UbL401 and UbL402, consistent with the RNA-seq data in NIL5 and NIL8, and RT–PCR data in AnS-1 (Figs 4a,b and 5d; Supplementary Figs 10, 12, 13 and 15).

Figure 5: UbL401 and UbL404 mRNAs preferentially express in MMCs.
figure 5

(a,b) In situ hybridization of UbL401 (a) and UbL404 (b) transcripts in anther. (c,d) At indicated temperatures, expression level of UbL404 in MMCs under high (28 °C) and low (23 °C) temperature in AnN (c) and AnS-1 (d). Scale bars, 10 μm (a,b). MMC, microspore mother cells.

Overexpression of UbL401 and 4 leads to male sterility

To test whether the accumulation of UbL401 and UbL404 in tms5 plants could affect male fertility, we generated UbL401- and UbL404-overexpressing transgenic plants in the ZH11 background (designated UbL401OE and UbL404OE). These plants produced higher levels of UbL401 and UbL404 mRNAs than ZH11 and exhibited partial pollen abortion (Supplementary Fig. 16), mimicking the phenotypes of tms5 lines at restrictive temperatures. Furthermore, we created UbL401 or UbL404 knockdown plants in the tms5 background, designated UbL401i and UbL404i. At restrictive temperatures, these plants exhibited reduced levels of UbL401 and UbL404 mRNAs and partially restored male fertility (Supplementary Fig. 17). Thus, we hypothesize that regulation of UbL40 mRNA levels by RNase ZS1 is critical for MMC development at high temperature in rice.


Male sterility has been extensively applied to the breeding of crops for hybrid vigour, and these hybrids have larger stature and higher yields than the parent inbred lines. In rice, CMS lines and TGMS/PGMS lines are widely used in three-line and two-line hybrid rice breeding, respectively. Here we uncovered the molecular mechanism of rice TMS5, which functions in RNase ZS1-mediated UbL40 mRNA regulation during male development (Fig. 6). At the permissive temperature, the level of UbL40 mRNAs remains low in the tms5 mutant plants, allowing the production of normal pollen. At restrictive temperature, however, UbL40 mRNAs are not processed by RNase ZS1 and their high level cause male sterility. The preferential expression of UbL401, UbL402 and UbL404 in MMCs indicates that these mRNAs are crucial for MMC development. We notice that overexpression of UbL401 and UbL404 in wild type leads to male sterility without causing any other obvious defects. We hypothesize that additional factors that are restricted to the MMC might be required to assist UbL40 genes in their functions (Fig. 6). Our results thus uncover a novel function for the RNase ZS1 family, deepening our understanding of the molecular genetic basis of TGMS in rice. These findings have profound implications for hybrid breeding in crops.

Figure 6: A functional model for RNase ZS1 controlling TGMS in rice.
figure 6

In rice anther, UbL40 genes preferentially express in MMCs (in purple). UbL40 mRNAs are induced by high (restrictive) temperature. In wild type, RNase ZS1 processes mRNAs of UbL40 and maintains them at normal levels. In RNase ZS1 defective mutant plants, excessive mRNAs of UbL40 accumulate in anthers and lead to male sterility. MMC, microspore mother cells.

Male reproduction is very sensitive to alterations in environmental conditions. A long non-coding RNA (lncRNA), the long-day-specific male-fertility-associated RNA, causes a PGMS trait and was identified in Nongken58S11. At the same locus in the indica line Pei’ai64S, a non-coding RNA producing small RNAs was shown to confer a P/TGMS trait10. In addition, a temperature-sensitive splicing defect of the transcript encoding UDPase leads to male sterility in rice at high temperature28,29. Our results show that, in tms5 mutants, high temperature induces accumulation of UbL40 mRNAs in MMCs and causes male sterility. It is possible that environmental conditions such as temperature and photoperiod may affect RNA metabolism at the post-transcriptional level. Defects in these processes might be harmful to the cells in anther development, leading to male sterility.

The endonuclease RNase Z exhibits two forms, RNase ZS and RNase ZL, which are evolutionarily conserved in almost all organisms21,22. RNase Z proteins have various RNA substrates, including tRNAs and mRNAs. Different RNase Z proteins might process different substrates and functions in different subcellular compartments21. In prokaryotes, RNase ZS is involved in tRNA 3′ terminal processing and has a role in mRNA decay30. In Arabidopsis, four RNase Z proteins can process tRNA in vitro25: Trz1 and Trz2 (RNase ZL) localize in the nucleus and mitochondria and may process tRNAs; Trz3 and Trz4 (RNase ZS) localize in the chloroplasts and cytoplasm, respectively. Loss-of-function of Trz4 leads to embryo lethality, and the function of Trz3, which localizes in the cytoplasm, is largely unknown25. The rice genome encodes three RNase Z proteins, including one long form and two short forms (RNase ZS1 and RNase ZS2; Supplementary Fig. 7). RNase ZS2 localizes in the chloroplast and is essential to chloroplast development31. In humans, ELAC2 (RNase ZL) localizes in the nucleus and mitochondria32,33. ELAC2 can process the lncRNA MALAT1 and plays a role in the generation of a mammalian herpesvirus microRNA34,35. ELAC1 (RNase ZS) occurs in the cytoplasm but has unknown functions36,37. The enzymatic activity of ELAC2 is over 1,600-fold higher than ELAC1, suggesting that ELAC1 might have new functions other than tRNA processing in the cytoplasm38. Our data provide the first direct experimental evidence demonstrating the biochemical roles of a short form of RNase Z (RNase ZS1) in mRNA metabolism in the cytoplasm of an eukaryote. Our findings could provide clues to the functions of the RNase Z family not only in plants but also in animals.


Plant materials and growth conditions

Rice accessions, including AnnongS-1 (O. sativa ssp. indica, AnS-1, tms5), AnnongN (O. sativa ssp. indica, AnN, WT siblings of AnS-1), Zhonghua11 (O. sativa ssp. japonica, ZH11, WT), tms5-ZH11 (the progeny of AnS-1 after two backcrosses to ZH11) and transgenic lines TMS5g-AS (RNase ZS1 transgene in AnS-1 background), TMS5i-ZH (RNase ZS1 RNAi in ZH11 background), UbL401OE (UbL401 overexpression in ZH11 background), UbL404OE (UbL404 overexpression in ZH11 background), UbL401i (UbL401 RNAi in ZH11 background) and UbL404i (UbL404 RNAi in ZH11 background) were grown in the field under normal conditions at the South China Agricultural University, Guangzhou, or grown in a growth chamber at DATs of 23, 24, 26, 28 and 30 °C (Supplementary Table 2a). The photoperiod was 12 h light and 12 h dark, unless specifically indicated. Zhu1S (O. sativa ssp. indica, tms5), 93-11 (O. sativa ssp. indica, WT), NIL5 and NIL8 (near-isogenic line 5 and 8 of Zhu1S in the 93-11 background), Nipponbare (O. sativa ssp. japonica) and transgenic lines TMS5g-ZS (genetic DNA of RNase ZS1 in Zhu1S background) and TMS5i-NB (RNase ZS1 RNAi in Nipponbare background) were planted in the field in Beijing or in Phytotron growth chambers (Koithtron S-153w, Japan) at DAT of 22, 24, 26, 28 and 30 °C with 13.5 h light and 10.5 h dark at the China National Rice Research Institute, Hangzhou (Supplementary Table 2b). The rice seedlings were moved to growth chambers or the Phytotron at indicated temperature when panicle length reaches ~1 cm (before the stages of early premeiosis) for ~2 weeks until flowering. All samples from AnS-1 and its controls were harvested between 1500 and 1700 hours in the growth chamber and Zhu1S and its controls were harvested between 0800 and 1100 hours in the Phytotron.

Characterization of AnS-1 and Zhu1S phenotypes

Rice florets were photographed with an OLYMPUS DP70 digital camera and an OLYMPUS SZX10 dissecting microscope. Mature pollen grains were stained with 1% I2–KI solution and observed with an OLYMPUS BX51 or Leica DNRXA microscope. Panicles from various stages of development were fixed with a mixture of 4% paraformaldehyde and 1.5% glutaraldehyde in PBS solution at 4 °C. The samples were then fixed again for 1 h with 1% osmic acid. Fixed anthers were washed three times for 10 min each in PBS solution. After washing, the samples were dehydrated in five successive graded ethanol baths (from 50 to 95%) for 10 min each and then washed in 100% propylene oxide twice for 15 min each. The samples were infiltrated with resin/propylene oxide (1:1 mixture) for 2 h and resin/propylene oxide (3:1 mixture) for 3 h, embedded in pure resin overnight, and then placed in an oven at 60 °C for 48 h. After capsule-embedding, blocks were trimmed on an LKB pyramitome. Trimmed blocks were cut to 2-μm semithin sections on an LKB ultramicrotome with glass knives. The semithin sections were stained with 0.2% toluidine blue O (Chroma-Gesellschaft, Germany) and examined with the Leica DNRXA light microscope or OLYMPUS BX51 microscope.

Fine mapping of tms5 and the gene conferring TGMS in Zhu1S

The F2 mapping population was generated from crosses between Xiang125S (derived from AnS-1) and Jingxian 89 (O. sativa ssp. indica) and grown in the field during the summer in Guangzhou (DAT>28 °C), Guangdong province in China. The introgressed mapping population (BC1F2 and BC6F2) was generated according to the programme shown in Supplementary Fig. 4c and planted in the summer at Changsha (DAT>28 °C), Hunan province in China. Details of the markers used for mapping are listed in Supplementary Table 5. The candidate genes were sequenced and compared in wild type (AnN and 93-11) and TGMS (AnS-1 and Zhu1S) using the primers shown in Supplementary Table 6.

Vector construction, plant transformation and analysis

For functional complementation tests, a 5.3-kb DNA fragment containing 2.3 kb of upstream sequence, the entire TMS5-coding sequence, and 0.9 kb of downstream sequence, was amplified from AnN genomic DNA using primers TMS5g-AS F/R (Supplementary Table 6), and then digested with BamHI and HindIII and ligated to the binary vector pCAMBIA1300 for gene transformation. A 5.1-kb DNA fragment containing a 2.2-kb upstream sequence, the entire TMS5 gene and a 1.0-kb downstream region from a BAC library (OSJNBb00118O13, Arizona Genomics Institute), was digested with HindIII and ligated to the binary vector pCAMBIA2300 to get the complementary construction, XF2108. For TMS5i-ZH and TMS5i-NB RNAi (construction of XF2114), two DNA fragments (415 and 360 bp, respectively) were cloned from TMS5 cDNA using the primer pairs TMS5i-ZH F/R and TMS5i-NB F/R (Supplementary Table 6), and then inserted into pYLRNAi39 and pUCC-RNAi40. For UbL401 and UbL404 RNAi, two DNA fragments (242 and 322 bp, respectively) were amplified from UbL401 and UbL404 using two pairs of primers, UbL401i F/R and UbL404i F/R (Supplementary Table 6), and inserted into pYLRNAi. For UbL401 and UbL404 overexpression, two DNA fragments (424 bp) from UbL401 and UbL404 were amplified using two pairs of primers, UbL401OE F/R and UbL404OE F/R, and inserted into the pYLox vector39 driven by the ubiquitin promoter. Transgenic plants were generated via Agrobacterium-mediated transformation41,42. For each vector, 15 transgenic lines were selected and genotyped. Positive transgenic lines were selected for further analysis by expression level of target genes. Transgenic lines with moderate expression levels were next analysed and average phenotypes were photographed. The relevant PCR primers are listed in Supplementary Table 6.

Subcellular localization

The open reading frame of TMS5 was amplified and fused with the N terminus of GFP43 in the pUC18 vector under the control of the CaMV35S promoter. Protoplasts from leaf sheaths of rice plants were isolated as described previously44. Briefly, for protoplast transformation, 10 μl of plasmid carrying TMS5 and GFP fragments, 100 μl of protoplasts and 110 μl of PEG solution (40% PEG4000, 0.3 M mannitol and 0.1 M CaCl2) were mixed gently and incubated for 15 min. After transformation, cells were washed with W5 solution and then resuspended in WI solution (4 mM MES, pH 5.7, 0.5 M mannitol and 20 mM KCl). Cells were incubated ~20–24 h after transformation. GFP signals were observed under the OLYMPUS BX51 fluorescence microscope.

Expression analyses

Total RNA was isolated from rice roots, stems, leaves, panicles, stamens and pistils at different temperatures using TRIzol (Invitrogen, USA) reagent. DNase I-treated total RNA (2.0 μg) was used for reverse transcription using the M-MLV-RT kit (TAKARA, Japan). RT–PCR and real-time quantitative PCR were performed as previously described39,40. Briefly, equal amounts of RT products were used to perform PCR. Quantitative PCR analyses were performed three repeats for each sample using the SsoFast EvaGreen Supermix kit (Bio-Rad, USA) with CFX96 Real-Time PCR Detection System (Bio-Rad). The cDNA levels of target genes were normalized to the internal standard genes OsActin1 and OsEF1-α. RNA blotting was performed as previously described42. The hybridization signals were visualized using a Typhoon 9000 System (GE, USA). Nuclear and cytoplasmic proteins were purified using Plant Total Soluble Nuclear Protein High-Pure Kit (GENMED, USA). Total protein was extracted from rice panicles grown at different temperatures. Immunoblot analysis was performed using Li’s methods42. The complementary sequences corresponding to tRNAs were used as probes labelled with γ-32P-ATP using T4 polynucleotide kinase. The hybridization was performed at 42 °C in a hybridizer. The protein extracts were separated by 12% SDS–polyacrylamide gel electrophoresis and transferred to a Pure Nitrocellulose Blotting Membrane (Pall Corporation). The membrane was successively incubated with the anti-RNase ZS1 antiserum (1:1,000 dilution) and secondary antibody goat anti-rabbit IgG HPR (1:10,000 dilution). Actin (Code: M20009; Abmart, China) was selected as the internal standard protein. OsSRT1 (ref. 45) and Osβ′-COP46 were selected as the nuclear and cytoplasmic protein controls. Anti-TMS5, anti-OsSRT1 and anti-Osβ′-COP were generated by immunizing rabbits. The relevant PCR primers are listed in Supplementary Table 6.

Overproduction and purification of RNase ZS1

The RNase ZS1 and RNase ZS1H81L sequences were amplified (primers see Supplementary Table 6) and cloned into plasmid pET23d (Novagen, Germany). After transformation of the E. coli BL21 (DE3) strain with the resulting plasmid constructs in 2 l of LB culture medium at 37 °C, expression of histidine-tagged RNase ZS1 was induced by the addition of 0.1 mM isopropyl-β-d-thiogalactopyranoside and grown at 20 °C overnight. The culture was harvested, pelleted, resuspended in 30 ml of buffer containing lysozyme and stirred for 3 h at 4 °C. The suspension was treated with an ultrasonic disintegrator (SONICS, USA) at 4 °C and then centrifuged at 15,000 g for 30 min to remove cell debris. The supernatant was loaded onto a Ni2+-NTA column (Qiagen, Germany). RNase ZS1 was purified using the histidine tag according to the manufacturer’s instructions (Novagen) and immediately dialysed. Protein purity was then verified using SDS–PAGE analysis (12%) and estimated at >95%.

Synthesis of tRNA precursors and UbL40 RNAs in vitro

Precursor tRNAs, UbL40 and CYP86A1 were amplified using specific primers (Supplementary Table 6) and cloned into plasmid pEASY-T3 (Transgene, China). Precursor tRNAs, UbL40 and CYP86A1 were transcribed in vitro using T7 RNA polymerase and labelled with [α-32P]-CTP, according to the manufacturer’s instructions (Promega, USA). The template DNA was removed by the addition of DNase I (Fermentas, USA). The labelled RNAs were purified using phenol and ethanol.

RNase ZS1 enzymatic activity assays

RNase ZS1 activity was assayed as previously described24. Briefly, the labelled precursor tRNAs or UbL40 RNA were incubated for 0.5, 1, 2 or 4 h at 37 °C with 200 ng purified RNase ZS1 or enzymatic null RNase ZS1H81L, in which a substitution of histidine by leucine in position 81 in the conserved enzymatic domain of RNase Z proteins causes the loss of enzymatic activity47. The RNase ZS1 cleavage products were purified using the RNA probe purification kit (OMEGA, USA). Equal volumes of reactant solutions and 2 × loading buffer (95% formamide, 20 mM EDTA, 0.05% bromophenol blue and 0.05% xylene cyanol) were mixed. Denaturing gel electrophoresis was performed by 8 or 15% PAGE containing 7 M urea and the signals were visualized with the Typhoon 9000 System (GE).

RNA ligase-mediated 5′ RACE

A 5′ RNA adaptor (5′- CUGACUGCACUCAGAGUACUACAGCCGAC -3′) was ligated to ~2.0 μg of total RNA using T4 RNA ligase 1 (NEB, USA). The ligated mRNAs were then reverse transcribed using oligo(dT)15 primer with AMV reverse transcriptase (TAKARA). Three rounds of 5′ RACE reactions were performed with two nested primers (outer, CX1544: 5′- GCTGATGGCGATGAATGAACACTG -3′; inner, CX1545: 5′- CGCGGATCCGAACACTGCGTTTGCTGGCTTTGATG -3′) and three gene-specific primers (Supplementary Table 6). The PCR products were gel-purified, cloned (pEASY-T1, Transgen, China) and sequenced.

Microarray and RNA-seq analysis

RNA samples used for microarray analysis were prepared from young panicles from stages of MMC to meiosis in wild-type (AnN and ZH11) and tms5 (AnS-1 and TMS5i-ZH) plants grown at permissive and restrictive temperatures. RNA purification and Affymetrix microarray hybridization were performed by the Capital Bio Corporation (, Beijing, China). The microarray data were analysed with the Gene Chip Operating software (GCOS 1.4). The different arrays were normalized using DNA-chip analyzer (dChip). The significantly differentially expressed genes between wild-type and tms5 plants were analysed using the Significant Analysis of Microarray software48. Single-end 86-nucleotide RNA-seq reads were aligned to the TIGR6.1 genome sequences using TopHat49. Uniquely mapped reads were used for subsequent analysis50.

In situ hybridization analyses

Specific regions of RNase ZS1-, UbL401- and UbL404 were amplified using corresponding primers (Supplementary Table 6) and then transcribed in vitro as probes using the DIG RNA labeling kit (Roche, Switzerland). Fresh ZH11 young panicles from different developmental stages were fixed immediately, embedded in paraffin (Sigma-Aldrich, USA) and sectioned to 8-μm thickness. Hybridization and immunological detection were performed according to the previously described method51. Briefly, sections were over night incubated at 45 °C with coverslips in hybridization buffer (40 μl per slide) containing the probes. Immunological detection of the hybridized probes was performed using a DIG nucleic acid detection kit (Roche) according to the manual.

Additional information

How to cite this article: Zhou, H. et al. RNase ZS1 processes UbL40 mRNAs and controls thermosensitive genic male sterility in rice. Nat. Commun. 5:4884 doi: 10.1038/ncomms5884 (2014).

Accession codes: The microarray and RNA sequencing data have been deposited in the Gene Expression Omnibus (GEO) under the accession codes GSE42367 and GSE42314, respectively.