Efficient development of practically usable thermo-photo sensitive genic male sterile lines in wheat through doubled haploid breeding

Background: Two-line hybrid wheat system using thermo-photo sensitive genic male sterility (TPSGMS) is now a dominant and promising approach of wheat heterosis utilization in China. However, during past twenty years only several TPSGMS lines have been capable of practical application in hybrid wheat breeding and production, which reduced the opportunities and efficiency of creating hybrids with strong heterosis. Introducing doubled haploid (DH) breeding could be a helpful strategy to efficiently develop practically usable TPSGMS lines. Results: F 1 s and selected F 2 and F 3 sterile plants from eight crosses made from two commercial TPSGMS lines were used to produce DH lines by using the wheat × maize system. Twenty four elite sterile lines possessing stable sterility, good outcrossing and yield potential, resistance to yellow rust and powdery mildew, and desirable plant height (50-60 cm) were obtained within 4 years through at least one year evaluation. Twenty from twenty four elite lines showed stable sterility in repeated tests of two or three years, will be selected for hybrid breeding. The percentage of elite lines within total tested DH lines produced from filial generations was in the order of F 2 > F 3 > F 1 in this study. Conclusions: Our study shows that DH breeding is more efficient for the selection of traits controlled by recessive gene(s) compared with conventional breeding, especially for the sterility of TPSGMS wheat. Coupling DH techniques with conventional breeding would be an efficient strategy for developing practically usable wheat TPSGMS lines in respect to number and saving time, which is helpful for further improving the efficiency of wheat hybrid breeding. Producing DHs from F 2 generation appeared to be the better choice considering the balance of shortening breeding time and overall breeding efficiency.

, could be helpful for improving development of wheat TPSGMS lines.
This study aimed to evaluate the efficiency of developing wheat TPSGMS lines by using DH technique based on wheat × maize in breeding program with sterile materials derived from different filial generations of F 1 , F 2 and F 3 .

Production of DH lines
During summer sowings of 2014-2016, we produced DHs with four F 1 s and sterile plants selected from F 1 and F 2 segregating populations by wheat ´ maize system (Table 2). A total of 920 DH lines were obtained from all eight crosses (Table 3). In Dec. 2016, a serious frost damaged some doubling treated plants that were heading, resulting in partial failure to obtain DH seeds. Variance analysis showed that there were significant difference in embryo rate (P=0.00) and haploid seedling rate (P=0.00) among different combinations, which indicated that embryo rate and haploid seedling rate were more susceptible to genotypic influence. The averages of embryo rate, seedling rate of embryos and chromosome doubling rate of seedlings were 36.76%, 62.65% and 86.42% respectively, exhibiting good efficiency in DH production as showed in our previous study [22].
Temperate climate at Kunming especially from May to October allows planting of spring and vernalized winter wheat materials throughout the year under natural conditions here ( Fig. 1 and Supplemental Figure 1), which facilitates mass production of wheat DHs by wheat ´ maize crosses because plenty of fresh pollens are available from naturally and repeatedly planted maize plants from late April to early November [22][23][24][25][26].

Selection of candidate DH sterile lines
All 920 DH lines produced from F 1 , F 2 and F 3 generations were separately tested for sterility during 2016-2018. Among them, 295 DH lines showed normal seed set were excluded from further testing.
When sterile lines were sown on Oct. 15 (1 st sowing) and Nov. 20 (2 nd sowing), their sensitive periods (causing fertility alteration) would be the dates from middle to late February and from late March to early April, respectively. Consequently, during sensitive periods, the 1 st sown sterile lines would have low temperature and short-day to fully exhibit sterility, while the 2 nd sown lines would have relatively higher temperature and longer day length that cause the early heading spikes sterile and the late heading tillers partially fertile to produce a few seeds for propagation ( Fig. 1 and Supplemental Figure   1).
According to our experiences, TPSGMS lines that exhibit 100% sterility in the 2 nd sowing date are usually stable in sterility but difficult in propagation, which make them not suitable for practical application. In southwest of China wheat is normally sown from middle Oct. to early Nov., thus a TPSGMS line with seed setting rate <5% in both sowing dates (from Oct. 15 to Nov. 20) would be able to meet the demand for safe production of qualified hybrid seeds.

Sterility determination of DH lines derived from F 1 generation
Ten elite DH sterile lines derived from F 1 were repeatedly tested during 2016-2018. In 2016/2017, the seed setting rates of all lines were 0 in the 1 st sowing date, and ranged from 2.98% to 4.87% in the 2 nd sowing date (Table 5). In further test of ten sowing dates in 2017/2018, the seed setting rates of the ten elite lines were < 1% from the 1 st to the 3 rd sowings (Oct.22-Nov. 5), < 5% till the 5 th sowing (Nov. 19), and ≥ 50% in the 10 th sowing (Dec. 24), suggesting sowings before Nov. 5 and Nov. 19 were separately the optimum and suitable times for hybrid seed production while after Dec. 24 the suitable time for propagation of these sterile lines ( Fig. 3 and Supplemental Figure 3).
During 2016-2018, ten elite TPSGMS lines derived from F 1 generation showed highly or 100% sterility in three years when sown from Oct. 15 to Nov. 5, although the average temperatures varied from 12℃ to 15℃ during their sensitive periods from the second half Feb. to the first half Mar. (Fig.1 and Supplemental Figure1), will be utilized in hybrid breeding later.

Sterility evaluation of sterile lines derived from F 2 generation
Ten selected lines in 2016/2017 derived from F 2 were tested again in winter sowing of 2017/2018. The seed setting rates of all lines were 0 in the 1 st sowing, and ranged from 1.99% to 4.04% in the 2 nd sowing ( Table 6). All elite lines showed stably in sterility during two-year cycles, and will be planted in ten sowing dates for further determination in sterility, suitable times for hybrid seed production and propagation. Meanwhile, preliminary test-crosses will be conducted with these lines.

Evaluation of out-crossing potential for elite TPSGMS lines
In winter sowing of 2018/2019, the out-crossing potential of 20 elite TPSGMS lines derived from F 1 and F 2 generations were evaluated. The out-crossing rates of 20 lines ranged from 70.46 % to 93.90% with average of 82.87%. There were 13 lines including 8 derived from F 2 generation with out-crossing rate > 80%, 4 lines between 75% and 80% and 3 lines between 70% and 75% (Table 7). All 20 lines showed high out-crossing potential even only one round of selection was done after DH production, because doubled haploids had 'genetically fixed' the trait, which confirmed our previous results [27].
However, more lines derived from F 2 generation appeared to have better out-crossing ability (>80%) compared with that from F 1 generation, suggesting one more cycle of selection before DH production would help to further concentrate the target trait. In fact, the results of out-crossing rates here were obtained by pollination with nearly unlimited pollen supply, and need to be further assessed in practical hybrid seed production.

Breeding efficiency of different generations
According to seed setting rates < 5% in both sowing dates, 41 DH sterile lines including 13, 15, and 13 lines separately derived from F 1 , F 2 and F 3 generations were selected, with breeding efficiency (percentage of selected DH lines in total DH lines tested) of 4.14%, 7.35% and 12.15% in F 1, F 2 and F 3 , respectively. When out-crossing ability, resistance to diseases and other desired traits were further considered, 24 elite lines were left and the overall breeding efficiency in F 1, F 2 and F 3 were 3.18%, 4.90% and 3.74%, respectively. U-test analyses indicated that there were significant differences (P<0.01) in breeding efficiency of producing DHs from F 1 , F 2 and F 3 generations (Table 4).
The trend of breeding efficiency for a single trait (sterility) was in the order of F 3 > F 2 > F 1, while for comprehensive traits was F 2 > F 3 > F 1.

Discussion
Utilization of male sterility is the basis of commercial application of hybrid wheat. It's common for all genetically controlled sterility systems that the more sterile lines capable of commercial utilization are used for test-crossing, the more opportunities of creating hybrids with super heterosis would present. Though the TPSGMS based two-line hybrid system was established as early as 1990S [7,28] but less than ten TPSGMS lines capable for commercial usage have been developed and utilized up to now in north and south wheat zones of China. Pedigree method is commonly used in developing TPSGMS lines [29,30], however, there were some difficulties hindering the breeding efficiency. The sterility of TPSGMS line is controlled by recessive nuclear major genes plus minor genes[7, [31][32][33][34], causing a very low proportion of highly sterile plants in segregating populations, especially in F 2 s derived from crosses between sterile lines and normal fertile lines. When other important traits like outcrossing ability, plant height, yield potential, disease resistance, etc. are considered together, the breeding efficiency would become extremely low. Theoretically, the probability of homozygous recessive individuals in F 2 population is 1/4 n , while that in DH population produced from F 1 would be 1/2 n , suggesting DH breeding is more efficient for selection of traits controlled by recessive genes, especially for the sterility of TPSGMS wheat. Also, crossing between semi-sterile materials and sterile lines further increased the proportion of highly sterile plants in segregating populations of this study, which is similar in effectiveness to backcrossing with sterile lines [29].
Meanwhile, few effective molecular markers are currently usable for marker assisted selection in sterility [35]. Consequently, it would take long time to develop a genetically stable TPSGMS line because the expression of sterility in TPSGMS lines needs restricted temperature and light conditions which could be found only once a year [30]. In our previous breeding program, only two TPSGMS lines K78S and K456S capable of commercial usage were developed by pedigree methods from 1996 to 2010, while in this study we developed 24 elite TPSGMS lines with complete homozygosity and other desired traits within 4 years by introducing DH techniques.
Another issue we address is to identify the better generation for producing DHs. Most breeders prefer to produce DHs from F 1 generation to shorten the breeding cycle, but it may limit the opportunity for recombination [15]. Therefore, producing DHs with selected individuals from F 2 generation of single crosses or F 1 generation of pyramiding crosses seems better than that from F 1 generation of single crosses [36]. Similarly, Snape and Simpson (1981) inclined to produce DHs from F 2 generation in barley by comparing the gain in genetic variation for 6 agronomic traits with DH lines derived from F 1 , F 2 , F 3 and intermated F 2 (S3) generations [37]. In contrast, Iyamabo and Hayes (1995) did not found more favorable genotypes in DH lines produced from F 2 generation than that from F 1 generation in barley, therefore, they preferred to use F 1 generation for producing DHs [38]. In present study, the overall breeding efficiency of producing DHs from filial generations was in the order of F 2 >F 3 >F 1 , indication F 2 generation is better for producing DHs in breeding efficiency, which confirmed the majority results above. However it still needs to be further investigated by comparing the breeding efficiency of producing DHs with F 1 and selected plants of F 2 and F 3 derived from the same cross.
Producing DHs from F 1 generation had less breeding efficiency because only one round of recombination occurred and no selection was done. As a result, a high frequency of agronomically undesirable lines were produced [37], which was confirmed in this study as most fertile lines discarded came from F 1 generation. However, using F 1 generation for DH breeding has the edge in saving time, it could be useful for crosses with better predictability and coping with urgent needs for developing varieties with resistance to diseases such as yellow rust for its frequently varying pathogenic races.

Conclusion
In this study we developed at least 20 practical TPSGMS lines of wheat that showed stable sterility in two or three years tests, high outcrossing potential and other desirable traits within 4 years, which verified that DH breeding is more efficient for the selection of traits controlled by recessive gene(s) compared with conventional breeding, especially for the sterility of TPSGMS wheat. Introducing DH technique is an efficient strategy in developing TPSGMS lines of wheat, both in number and saving time. Generally, producing DHs from F 2 generation appears to be the better choice with balance of breeding efficiency and shortening of breeding cycle. However, this result should be further investigated by using diverse genetic materials of different filial generations derived from the same combinations. More practically usable TPSGMS lines would further improve the efficiency of wheat hybrid breeding by increasing the opportunity of developing hybrids with high heterosis.

Plant materials
Two TPSGMS lines and five semi-sterile advanced lines of wheat were used in the study (Table 1). A maize variety "Baitiannuo" was used as pollen donor in DH production. All wheat and maize materials were bred by Institute of Food Crops, Yunnan Academy of Agricultural Sciences, Kunming, China.

Crossing and DH production
Wheat materials were late sown in January 2014 to make semi-sterile materials fertile for crossing with sterile lines K78S and K456S at Kunming, Yunnan province, China (25°02'N, 102°42'E, altitude 1960 m), where spring wheat could be planted and harvested throughout the year. For DH production, maize sowing (in April) began two months before wheat sowing (in June) to synchronize their flowering dates, and maize was sown in three dates with interval of 14 days. The crossings and handling of subsequent generations are summarized in Table 2. Before producing doubled haploids, pedigree selection was adopted to select sterile plants from segregating population of F 1 and F 2 generations according to performances in sterility, out-crossing potential including glume opening and stigma exertion [27], plant height (50-60 cm), resistance to yellow rust and powdery mildew, tillering ability and the yield potential. Seeds of sterile plants were harvested from regenerated tillers by cutting all spikes of sterile plants followed by intensive irrigation and fertilization.

Method of producing DHs
For DH production, an improved protocol [22,23]  20, respectively. Lines with high sterility were kept for further evaluation. At least 10 spikes per line in each sowing were randomly bagged before flowering to measure the seed setting rate and outcrossing potential [27]. Other important traits such as disease resistances and yield potential were