Breeding and study of two new photoperiod- and thermo-sensitive genic male sterile lines of polyploid rice (Oryza sativa L.)

Male sterile lines play an important role in the utilization of heterosis. To explore and exploit the heterosis of polyploid hybrid rice, two photoperiod- and thermo-sensitive genic male sterile lines of polyploid rice, PS006 and PS012, were bred via chromosome doubling, complex hybridization and self-breeding. The characteristics of these two lines, including the agronomic traits, growth, development, fertility transformation and combining ability, were investigated. Both lines had good agronomic characteristics and flowering habits, a high outcrossing rate, obvious fertility alterations and good combining abilities. Their hybrids showed strong heterosis and great potential for increasing rice productivity and quality. The new polyploid rice photoperiod- and thermo-sensitive genic male sterile lines will provide material for further research into polyploidy and hybrid vigour in rice and promote the exploitation of polyploid hybrid rice.

Stigma exsertion is an important trait that contributes to seed production in hybrid rice. Both PS006 and PS012 had good exsertion rates and similar single stigma exsertion rates. However, the double stigma exsertion rate of PS006 was higher than that of PS012 (Fig. 2b), thus indicating the PS006 was more conducive to hybrid seed production.
Characteristics of fertility alteration of PS006 and PS012. In both Hainan and Wuhan, PS006 and PS012 showed fertility alteration characteristics, and the stage sowing experiment showed similar results ( Fig. 2c and d and Table 2). When planted in Wuhan, the two male sterile lines had white and small anthers at the heading stage and displayed complete male sterility before September under a photoperiod of 12.5-14.0 h and a temperature of 25-34 °C. In mid-September, which presented a photoperiod of 10.9-12.3 h and a temperature of 21-30 °C, pollen fertility began to recover. During this period, the lines produced few seeds by self-pollination. After October, the lines reverted to infertility. The recovery periods of PS006 and PS012 differed and presented differences among years. When planted in Hainan, a region in southern China that is warmer and suitable for rice growing in winter, PS006 and PS012 were fertile before April (a photoperiod of 11.0-12.5 h and a temperature of 21-28 °C), and both had good seed-setting rates. After April (a photoperiod of 11.0-12.5 h and a temperature of 24-29 °C), the pollen gradually became sterile. The accurate recovery period also showed differences among years. These phenotypes were observed consistently from 2008-2016 in both locations, suggesting that male sterility may be controlled by both temperature and photoperiod.
The male sterile lines PS006 and PS012 also presented obvious characteristics of fertility transformation. In the different stages of fertility transformation, significant differences were observed in the anthers and pollen. In the sterile stage, the anthers were white and small ( Fig. 3a and   prismatic under the microscope, and most of the mature pollen grains (>90%) were typical abortive pollen and did not result in pollination ( Fig. 3b and f). These phenotypes indicated that pollen abortion of PS006 and PS012 primarily occurred at the microspore stage, suggesting that the two polyploid rice PTGMS lines were the sporophyte male sterile. In the fertile stage, PS006 and PS012 had normal plump yellow anthers and normal pollen grains ( Fig. 3a-h).
Fertility performance of PS006 and PS012 in a phytotron. PS006 and PS012 were grown in a phytotron at temperatures of 23, 24 and 28 °C and with two illumination times of 11.5 and 13.5 h. Temperature and illumination time had different effects on the fertility of PS006 and PS012 (Table 3). Significant correlations were observed between temperature and fertility for PS006 (F-value 163.777, P < 0.01) and PS012 (F-value 180.589, P < 0.01). The correlation coefficients between illumination time (F-value 20.833, P < 0.01) or photothermal interaction (F-value 22.642, P < 0.01) and fertility of PS012 were significant; however, none of these correlations were significant for PS006. The analysis illustrated that temperature was the main factor affecting fertility of PS006. Both temperature and illumination time affected fertility of PS012, although temperature was the major factor.

Sensitive stage, duration and sterile critical point of temperature (CPT) of fertility alteration.
According to the methods of Mou 27 , the correlation between the daily mean temperatures at 0-21 d before heading and pollen fertility of PS006 and PS012 were analysed ( Table 4). The results showed that specific stages of young panicle development in the two lines were sensitive to temperature. The sensitive stages were 6-15 d before heading, and the sensitive duration was 10 d.
Regression analyses between the average daily mean temperature in the sensitive stages indicated that the CPT of fertility alteration were 23.6 °C and 24.4 °C for PS006 and PS012, respectively (Table 5). This implied that when the daily mean temperature during the sensitive stage was below 23.5 °C, both lines were fertile, and when it was above 24.5 °C, the lines were sterile. As a result, the lines could be used for hybrid seed production or reproduction according to the temperatures in the environments in which they are grown.
Outcrossing characteristics of PS006 and PS012. The polyploid rice PTGMS lines PS006 and PS012 were crossed with high seed-setting polyploid rice restorer lines. Both PS006 and PS012 had good outcrossing rates (Table 6). However, differences were observed between the restorer lines. Generally, PS006 had a higher outcrossing rate than PS012. Pollen fertility rate (%) Self-seeds rate (%)
PS012 had the same sequence as PA64S and NK58S, which contained the same substitution of C-to-G (789 site) compared with NK58. The sequence of PS006 was the same as HD9802S, which contained a four-point mutation at position 524 (G-to-A), 866 (T-to-G), 1115 (A-to-G) and 1231 (T-to-G) and a single-base deletion at the 561 site (Fig. 4c). A sequence analysis of the tms5 gene in PS006, PS012, HD9802S, PA64S, AnnongS-1 (NnS-1) and Annong (NnN) revealed that PS006 had the same sequence as HD9802S and NnS-1, which contained the mutated nucleotide (C-to-A) at position 71 compared with NnN. However, PS012 and PA64S had the same substitution of T-to-G at position 70. The results indicated that the two polyploid rice PTGMS lines had different PTGMS gene backgrounds, with PS006 carrying the TGMS gene tms5, which was transferred from HD9802S, and PS012 carrying the pms3 gene, which was transferred from PA64S.    These results together suggested that PS006 and PS012 were typical two-line male sterile lines and exhibited sterility under high-temperature and long-day conditions and fertility under low-temperature and short-day conditions. These lines presented good plant types, good flowering habits, and high stigma exsertion and outcrossing rates; therefore, they appear to represent suitable materials for polyploid hybrid rice research and exploitation.

Combining ability analysis of the main characteristics of the hybrid combinations. Eight crosses
were generated according to a partial-diallel cross using the two polyploid rice PTGMS lines and four polyploid rice restorer lines. The results of a variance analysis of the eight main characteristics showed significant differences between the combinations (Table 7), which indicated genetic differences among the genotypes for these traits and genetic differences among the combinations.
Further analysis of the general combining ability (GCA) of PS006 and PS012 was performed based on the aforementioned analysis. The GCA in the same parent differed among the eight traits ( Table 8). The panicle number per plant (PP), filled grain number per panicle (FGP), total grain number per panicle (TGP), seed-setting rate (SR) and grain weight per plant (YP) of PS006 mainly showed additive effects; however, the plant height (PH), spikelet length (PL) and 1000-grain weight (GW) had negative GCA effects. The GCA values showed that PS006 was a good male parent for increasing FGP, TGP and SR and could increase the production of the F 1 hybrids. In contrast, the GCA effects of the PH, PL and GW for PS012 were significantly positive, whereas the PP, FGP, TGP, SR and YP were negative. The hybrids of PS012 may show tall plants, long spikelets and heavy grains. Thus, high-yield characteristics of large spikelets and heavy grains could be obtained using PS012.
To further investigate the yield level of the hybrids, we analysed the specific combining ability (SCA) of the polyploid hybrid rice combinations. The SCA values showed differences among hybrids for the same combination of different traits or the same trait with different combinations (Table 9). Additionally, the SCA effect of different combinations from the same male parent also differed, and the results indicated diverse gene interactions in polyploid hybrid rice. Thus, the SCA values can be used to guide the breeding of high-yield polyploid hybrid rice combinations.

Heterosis and utilization potential analysis of polyploid hybrids.
To investigate the heterosis and potential application value of polyploid rice hybrids, we compared the main agronomic traits among polyploid hybrids, diploid hybrids and conventional rice cultivars. Compared with the parents, the polyploid hybrids had stronger growth and tillering ability ( Fig. 5a and b), and they also had larger spikelets, more grains per spike and heavier grains (Tables 1 and 10). These results suggest that heterosis occurred in the F 1 generation of the polyploid hybrid rice, which was consistent with the results of the GCA and SCA analyses. The application value of polyploid rice depends on whether it presents advantages when compared with diploid rice. Compared with the diploid hybrid rice Liangyou 287 (Early Hybrid Rice, bred by professor Zhou, Hubei University, China), the polyploid hybrid rice line XH216 had sturdier stems, which could help reduce lodging (Table 10 and Fig. 5c). In addition, studies of the yield traits showed obvious differences in the GW, with the weight of XH216 (43.69 g) nearly twice that of Liangyou 287 (24.56 g). The TGP and SR of XH216 were lower; however, the GW per plant (54.70 g) was higher than that of Liangyou 287 (37.11 g) and Yangdao 6 (33.26 g). The results indicated that, compared with their parents, the F 1 hybrids of polyploid hybrid rice showed heterosis and that, compared with diploid rice, the F 1 hybrids demonstrated the potential for higher rice yields.

Discussion
The discovery of male sterile lines plays a crucial role in the utilization of rice heterosis 1,8,28 . To explore and exploit rice heterosis at the polyploid level, we first established a breeding technology for obtaining polyploid rice PTGMS lines based on chromosome doubling, complex hybridization and self-breeding. Certain crucial parameters are required for the success of such technology. First, the parent materials of the PTGMS lines and PMeS lines are required to provide the PTGMS and PMeS genes, particularly the PMeS gene, which can promote the SRs of the polyploid rice PTGMS lines in their fertile stage. Second, the chromosome doubling frequency affects the progress of breeding. Usually, the vitality of the callus, the concentration of colchicine and the time of colchicine treatment are key factors for success. Using this technology, two polyploid rice PTGMS lines were successfully bred: PS006 and PS012. Previous research has shown that PS006 and PS012 are tetraploid indica rice lines, and they present unique agronomic characteristics that are useful for rice breeding, such as strong stems, large panicles and stigmas, and long oval grains, which conform to the typical features of polyploidy. Flowering habit studies have revealed that both of these lines have good panicle uniformity, concentrated flowering periods, and good stigma exsertion rates, which would be more conducive to hybrid seed production 29 . In addition, we noticed that PS006 and PS012 had fertility alteration characteristics. Under high-temperature (above 23.6 °C for PS006 and 24.4 °C for PS012) and long-day conditions, the lines are male sterile. However, under low-temperature (below 23.6 °C for PS006, and below 24.4 °C for PS012) and short-day conditions, these lines convert to male fertile; thus, they can self-pollinate. We inferred that the fertility of these lines was mainly induced by temperature and photoperiod. In two-line hybrid breeding, the cultivation of sterile lines with low critical sterility-inducing temperature (CSIT) is a key requirement for ensuring the purity of hybrid seeds 30 , and the CSIT shows that PS006 (below 23.6 °C) would be safer than PS012 (below 24.4 °C) in two-line polyploid hybrid rice breeding. We hypothesize that the differences in CSIT are related to the different PTGMS gene backgrounds. In our study, the PTGMS genes of  Table 9. The specific combining ability values of polyploid hybrid rice combinations. PS006 were transferred from HD9802S, which presents a low CSIT (<23.5 °C) 31 . HD9802S and PS006 carried the same tms5 mutation. This mutation leads to the TGMS trait through a loss of RNase Z S1 function, which is responsible for processing Ub L40 mRNAs and controlling thermo-sensitive genic male sterility in rice 32 . However, the PTGMS genes of PS012 were transferred from PA64S, which is a TGMS rice line that was developed by transferring PTGMS genes from NK58S. The TGMS trait is conferred by p/tms12-1 (pms3), which encodes a unique noncoding RNA that produces a 21-nucleotide small RNA 9,10 . However, the sterility gene from NK58S, in which the PGMS trait is determined by pms1, pms2 and pms3 10,33-35 via a single genetic background, usually has a high CSIT 14,30 . The thermo-photoperiod sensitivity characteristics of PS012 are consistent with this conclusion.
Recently, Zhou et al. 36 suggested that TGMS lines with higher CSITs could be crossed with lower CSIT lines to select new TGMS lines with lower CSITs. Thus, the polyploid rice PTGMS lines with lower CSITs from PS006 and PS012 populations are selected according to this suggestion. In the present study, we also investigated the combining ability for the main agronomic traits in the hybrids generated by PS006 and PS012. We found that PS006 presented additive effects for PP, FGP, TGP, SR and total YP but had negative effects for PH, PL and GW. However, PS012 showed the opposite effects. The analysis of the combining ability indicated that high-yield polyploid hybrid rice combinations could be bred using suitable parents. Studies of the yield and yield-related traits among polyploid parents and hybrids, diploid hybrids and conventional rice have confirmed these findings. Compared with the parents, the polyploid F 1 hybrids had high parent heterosis for the tillers per plant, TGP, SR and grain yield per plant but not for the grain length and width. These results are consistent with other studies, which also found high parent heterosis for filled grains per panicle, SR and yield but negative high parent heterosis for grain length and width in polyploid rice [14][15][16]21 . Researchers have inferred that a complex genetic mechanism controls the heterosis in polyploid rice, and many genes related to fertility and heterosis in autotetraploid rice have been found [37][38][39] . The complex regulatory mechanisms might soon be revealed 14,40 . Our work suggests that PS006 and PS012 represent suitable material for further studies of polyploidy and hybrid vigour in rice.
In summary, the findings reported in this study provide new germplasm data for rice research and insights for studies of the two-line hybrid system of polyploid rice. Breeding procedures. The breeding process of PS006 and PS012 included parental selection, cross and composite cross, chromosome doubling, polyploid identification, fecundity identification and self-crossing to produce stable lines. PS006 is used as an example (Fig. 6).

Methods
(1) Parental selection. In the breeding of PS006, the parent materials were HD9802S-2x, HN2026-2x, HD9802S-4x and HN164-4x. (2) Cross. During flowering, HD9802S-2x was used as the female parent in crosses with HN2026-2x. (3) Chromosome doubling. The method used was modified according to the reports of Liu 41 and Li 42 . The F 1 (2x) of HD9802S and HN2026 was planted in the field. In the stage of panicle differentiation, young panicles of F 1 were cultured on an N6 solid medium to induce callus. Approximately 3 weeks later, the vigorous calli were transferred to the liquid medium with 500-750 mg l −1 colchicine for chromosome doubling. After a recovery culture process, the calli treated with colchicine were placed in a differentiation medium to form buds. The shoots were induced to produce roots on a 1/2 MS medium. Plantlets were transferred to the field and allowed to grow into whole rice plants. (4) Polyploid identification. In the breeding process, the synthetic autotetraploid rice HD9802S-4x, PA64S-4x and F 1 -4x and the polyploid rice PTGMS line PS006 were identified via morphological observations and the root tip chromosome number. (5) Backcross and sterile plant selection. The identified tetraploid F 1 (♀) was hybridized with HD9802S-4x (♂). Then, sterile plants were selected from the BC 1 F 1 in Wuhan (under high-temperature and long-day conditions) and then transported to Hainan to identify their fertility under low-temperature and short-day conditions and allowed to produce seeds (BC 1 F 2 ). (6) Composite cross and sterile plant selection. BC 1 F 2 (♀) was crossed with HN164-4x (♂) to obtain RC 1 F 1 . These RC 1 F 2 plants were planted in Wuhan. Then, sterile plants were selected from the RC 1 F 2 plants in Wuhan and transported to Hainan to identify their fertility. (7) Self-crossing for line stability. The plants selected from the PS006 lines were self-crossed and formed stable lines, and their final line numbers matched the original numbers at initial selection (BCF 2 ), i.e., PS006.
Chromosome identification and morphological observations. Plants from the synthetic autotetraploid rice HD9802S-4x, PA64S-4x and F 1-4x and the polyploid rice PTGMS lines PS006 and PS012 were examined by counting the chromosome numbers in their root tips according to the methods of Li 42 . The observations and photographic recordings were performed using an Olympus BX51 microscope (made in Japan). The key morphological traits PH, spikelet number, PL, grain length and width, awn length, shattering trait, seed colour and seed set were investigated. The recording methods and standards were set according to the protocols of Gai 43 . , where Xii is the indica homozygous genotype II, Xjj is the japonica homozygous genotype JJ, Xij is the indica-japonica heterozygous genotype IJ, and N is the number of InDel molecular markers.

PTGMS gene background investigation.
To investigate the PTGMS genes of PS006 and PS012, the sequences of pms3 (1236-bp) and tms5 (a 1060-bp fragment) were amplified via PCR. Genomic DNA from the leaves of plants was extracted using the sodium dodecyl sulphate method 45 . PCR amplifications were performed using two specific primers for the pms3 gene (primer 1, F: 5′-ggcatgtgtcttagggttttta-3′, R: 5′-accatgcctcccactcctatat-3′; and primer 2, F: 5′-aagcagagacatagatgagcaaca-3′, R: 5′-agcctatgtttcttctgccttg-3′), and one specific primer for the tms5 gene (F: 5′-tggccaaacagctgctacttca-3′, R: 5′-atggcgtggtaggtcttgaagg-3′) surrounding the designed target sites. The PCR products were purified and then directly sequenced (by the Tsingke biological technology company, Beijing, China). The sequences were analysed using BLAST (http://www.ncbi.nlm.nih.gov/BLAST/). Fertility alteration research. The characteristics of fertility alteration of the PS006 and PS012 lines were experimentally studied by sowing rice by stage, breeding in a phytotron and observing the pollen fertility under natural conditions in both Wuhan and Hainan. Pollen fertility was determined as the percentage of pollen grains stained with 1% I 2 -KI and observed under an optical microscope.
Combining ability analysis. The polyploid rice PTGMS lines PS006 and PS012 were crossed with four high seed-setting polyploid rice restorer lines in Wuhan. Then, all parent lines and their F 1 progeny were planted in Hainan. When the materials were mature, five representative plants of each type were randomly selected, and their main agronomic characteristics were investigated as follows: tiller number, PH, PL, grain number and GW. Then, the GCA and SCA values of the polyploid hybrid rice combinations were analysed using SPSS and Microsoft Excel.
Polyploid hybrids and diploid rice comparison. The main morphological traits of the polyploid hybrids and diploid hybrid and conventional rice cultivars were compared.