Common wheat cultivar Hanxuan 10 with remarkable tolerance to drought stress was used for genomic sequence isolation of TaSnRK2.3 and gene structure analysis. Twelve accessions of various wheat species, including three A genome accessions (Triticum urartu) (UR204, UR206, and UR207), three S genome accessions (Aegilops speltoides, the putative B genome donor) (Y2003, Y2033, and Y2017), three D genome accessions (Ae. tauschii) (Y125, Y225, and AE38), and three AB genome accessions (T. dicoccoide) (DS1, PS5 and PS9) were selected for target fragment isolation and genomic origin identification.

Thirty-two accessions/genotypes with wide variation screened by SSR markers, were initially chosen to re-sequence for polymorphism analysis. Three hexaploid wheat germplasm populations were selected for different research purposes. Population 1 (262 accessions/genotypes) was firstly employed for association analysis. The accessions were mainly released in the Northern winter wheat and Yellow and Huai River valley facultative wheat zones (Zhang et al., 2015). Population 2 (157 landraces/genotypes) and Population 3 (348 modern cultivars/genotypes) were used to determine temporal haplotypes and analyze geographic distribution aiming to functionally validate TaSnRK2.3-1A/1B markers. Population 2 was mainly from the Chinese wheat mini-core collection representing more than 70% of the genetic diversity of the total Chinese germplasm collection; Population 3 came from the Chinese wheat core collection (Hao et al., 2008; Hao et al., 2011). The two populations (2 and 3) including genotypes from all the 10 Chinese wheat production zones were selected from 23, 705 accessions released or collected in China (Zhang et al., 2002; Dong et al., 2003; Hao et al., 2008).

Population 1 was planted in 10 environments (year × site × water regime combinations) including Changping (116°13′E, 40°13′N), Beijing, in 2010 and 2012, and Shunyi (116°56′E, 40°23′N), Beijing, in 2010, 2011, and 2012. The field experiments were grown under well-watered (WW) and drought-stressed (DS) regimes. The WW plots were irrigated with 750 m³ ha^-1 (75 mm) at each of pre-overwintering, booting, flowering, and grain filling stages (total 300 mm applied as irrigation), while DS plots were rain-fed. The rainfall during the growing seasons were 131 mm in 2010, 180 mm in 2011, and 158 mm in 2012. Measured agronomic traits included 1,000-grain weight (TGW), plant height (PH), peduncle length (PLE), length of penultimate node (LPN), spike length, number of spikes per plant, total number of spikelets per spike, number of sterile spikelets per spike, and grain number per spike.

Stem water-soluble carbohydrates (SWSC) are an important carbon source for grain filling in wheat. They are mainly composed of fructans, sucrose, glucose, and fructose, with the main reserve as fructans at the late stage of WSC accumulation (Ruuska et al., 2006). We obtained SWSC data of Population 1 under WW and DS conditions. SWSC were measured by near-infrared reflectance spectroscopy (MAP multi-purpose FT-NIR analyzer) as previously described (Wang et al., 2011). Five main stems were cut 1 cm above the soil surface at the flowering, mid-grain filling (14 days after flowering), and maturity stages. Leaf blades were removed from samples, and stem samples were cut into two parts, namely, peduncle, and the lower internodes except for peduncle. The WSC was determined for peduncle, lower internode and total stem, using different near-infrared reflectance spectroscopy (NIRS) regression models, which were developed for quantitative determination of WSC using modeling samples of 150 DH (Hanxuan 10 × Lumai 14) lines (Wang et al., 2011).

For each accession, genomic DNA was extracted from young leaves by cetyltrimethylammonium bromide (CTAB) method (Stewart and Via, 1993). Based on known cDNA information, a primer pair GF/R (Table 1) was obtained to amplify the genome sequence of TaSnRK2.3. TransStart Fast Pfu DNA polymerase was used in PCR amplification (TransGen, Inc). PCR products were extracted and cloned into pEASY-Blunt vector, 24 clones for each sample were randomly selected for sequencing by DNA Analyzer 3730XL. To get the whole sequence of TaSnRK2.3, both M13 and three overlapping primers (Table 1) were used for sequence walking. So the sequence of each clone was obtained by assembling five overlapping sequences with the SeqMan program in DNAStar. The genomic origin of each of the sequences were confirmed by comparing them with that from diploid and tetraploid species based on clustalW analysis with the MegAlign program.

A total of four functional markers (2.3AM1, 2.3AM2, 2.3BM1, 2.3BM2) were developed based on four selected polymorphism sites, in order of 1898 bp (C/T) and 2905 bp (A/G) of A genome; 2153 bp (C/T) and 2638 bp (C/G) of B genome. These markers were designed with a specific mismatch in the primer to introduce a restriction enzyme recognition site using an available program dCAPS Finder 2.0¹. Basically, the four systems of PCR and digestion were similar. Genotyping was performed by two rounds of PCR. Firstly, genome specific primer pairs were used to amplify fragments from chromosome 1A/1B in all accessions. The second round of PCR was performed as follows: the first round of PCR product was diluted 50 times, then taking 1 μl as template for the second round of PCR. The annealing temperatures and extension time were set depending on primer pairs and expected PCR product lengths. The PCR products were resolved by electrophoresis in 4% agarose gel after digestion with corresponding restriction enzymes. The primers were listed in Table 1.

TASSEL 2.1 was used to identify significant associations between haplotypes and agronomic traits for Population 1. The general linear model (GLM) was performed using population structure Q matrix, which listed the estimated membership coefficients for each individual in each cluster. Associations were considered significant at P < 0.05. Different effects of haplotype on traits were analyzed by one-way ANOVA using SPSS 16.0 software, and followed by the least significant difference (LSD) method at P < 0.05 (even 0.01).

Based on TaSnRK2.3 cDNA, three genomic sequences of TaSnRK2.3 were isolated from Hanxuan 10 using the primer pair GF/R flanking the open reading frame. The sequencing results indicated that the three types of sequences obtained were consistent with the previous results (Tian et al., 2013). By comparing different TaSnRK2.3 genomic sequences among the ploidy levels of different wheat species, the genomic origin was identified (Figure 1). Since TaSnRK2.3 was located on chromosome 1A, 1B, and 1D (Tian et al., 2013), we named the three copies TaSnRK2.3-1A, TaSnRK2.3-1B, and TaSnRK2.3-1D, respectively.

Multialignment (ClustalW) assays showed that TaSnRK2.3 consisted of nine exons and eight introns (Figures 2, 3). The fragment sizes of TaSnRK2.3-1A, TaSnRK2.3-1B, and TaSnRK2.3-1D were 2,999, 2,862, and 2,817 bp, and their homologies ranged from 92.5 to 94.4%.

As shown in Figure 2, for TaSnRK2.3-1A, polymorphic sites were only identified in non-exon regions in 32 accessions, and the four SNPs formed three haplotypes (Hap-1A-1, Hap-1A-2, and Hap-1A-3). Ten nucleotide variations (eight SNPs and two 1-bp InDels) were detected in the target fragment of TaSnRK2.3-1B in the 32 common wheat accessions (Figure 3), only SNP at 928 bp (C/T) led to an amino acid change (CCG→Pro, TCG→Ser). The 10 SNPs formed three haplotypes, i.e., Hap-1B-1, Hap-1B-2, and Hap-1B-3.

To distinguish the haplotypes of TaSnRK2.3-1A, two functional markers (2.3AM1 and 2.3AM2) were developed based on the chosen SNP sites. One was a HhaI cutting site at SNP-1898C, un-cut at SNP-1898T; the other was an XbaI cutting site at SNP-2905A, un-cut at SNP-2905G (Figure 2). In the same way, markers 2.3BM1 and 2.3BM2 for TaSnRK2.3-1B were successfully developed with restriction enzymes N1aIII and AvaII (Figure 3).

For TaSnRK2.3-1A, Hap-1A-2 was a major haplotype accounting for 71.5% frequency in the natural population, followed by Hap-1A-1 with a frequency of 22.3%, and Hap-1A-3 with the lowest percentage (6.2%). Association analysis showed that the three haplotypes were significantly associated with PH, PLE, LPN, and TGW (Table 2). Hap-1A-1 had the lowest PH, PLE, and LPN among the three haplotypes in almost all the environments, with significant difference (P < 0.01) between Hap-1A-1 and the others. The TGW of Hap-1A-1/2 was higher than Hap-1A-3 in 10 environments, and the differences were significant in seven environments (P < 0.01 or 0.05) (Figure 4). Therefore, Hap-1A-1 could be a superior allele for increasing TGW and reducing PH.

For TaSnRK2.3-1B, Hap-1B-1 and Hap-1B-2 were two major haplotypes, accounting for 46.1 and 53.5% frequency. Hap-1B-3 was a rare haplotype only presented in one accession, thus it was not included in subsequent statistical analysis. Significant associations were identified between TaSnRK2.3-1B haplotypes and agronomic traits, including PH, PLE, and LPN in all 10 environments, and TGW in five environments (Table 3). Hap-1B-1 was significantly associated with lower PH, PLE, and LPN, and higher TGW (Figure 5). Therefore, Hap-1B-1 might be a favorable haplotype in terms of PH and TGW.

For TaSnRK2.3-1B, significant associations were detected between the two haplotypes and SWSC, including WSC1-2-DS, WSC1-3-DS, WSC2-1-DS, WSC2-2-DS/WW, and WSC2-3-DS/WW (Table 4). Hap-1B-1 had higher SWSC than Hap-1B-2 at flowering and mid-grain filling stages with significant difference (P < 0.01) (Figure 6). So Hap-1B-1 might be a potential haplotype to increase SWSC.

The Chinese wheat production area is classified into 10 main agro-ecological zones based on cultivar ecotypes, growing season, and cultivar response to temperature and photoperiod (Zhang et al., 2002). Among landraces, selection pressure on haplotypes in the different zones was not as strong as expected, and the frequencies of the favored haplotype Hap-1A-1 was generally low, none was identified in zone V, VII, and IX (Figure 7A). Similar trends were also observed for TaSnRK2.3-1B (Figure 8A).

AS shown in Figures 7, 8, from Chinese landraces to modern cultivars, the frequencies of the two superior haplotypes Hap-1A-1 and Hap-1B-1 increased across almost all zones except in VI (40% to 33%) and VIII (36% to 33%) of Hap-1B-1. Hap-1B-1 was the most favorable haplotype in major wheat production regions I (58%), II (49%), III (52%), and IV (83%) compared to the other six regions in Population 3. The results indicated the two haplotypes had suffered strong positive selection in Chinese wheat breeding programs.

Three hundred and forty-eight Chinese modern cultivars were divided into six subgroups according to 10-year release intervals to evaluate changes in haplotype frequencies over time. As a whole, the proportions of the two favored haplotypes, Hap-1A-1 and Hap-1B-1, increased gradually with PH reduction and TGW enhancement in Chinese modern cultivars since pre-1950 (Figure 9), suggesting that these favored haplotypes experienced positive selection in the process of wheat breeding. Furthermore, Hap-1B-1 (up to 75.9%) was positively selected by wheat breeders, while Hap-1A-1 (only 55.9%) still has considerable potential for further utilization.