A Single Amino Acid Substitution in STKc_GSK3 Kinase Conferring Semispherical Grains and Its Implications for the Origin of Triticum sphaerococcum[OPEN]

Map-based cloning of a gene underlying grain shape in wheat suggests that modest genetic changes induce dramatic phenotypic variations associated with a new wheat subspecies during evolution. Six subspecies of hexaploid wheat (Triticum aestivum) have been identified, but the origin of Indian dwarf wheat (Triticum sphaerococcum), the only subspecies with round grains, is currently unknown. Here, we isolated the grain-shape gene Tasg-D1 in T. sphaerococcum via positional cloning. Tasg-D1 encodes a Ser/Thr protein kinase glycogen synthase kinase3 (STKc_GSK3) that negatively regulates brassinosteroid signaling. Expression of TaSG-D1 and the mutant form Tasg-D1 in Arabidopsis (Arabidopsis thaliana) suggested that a single amino acid substitution in the Thr-283-Arg-284-Glu-285-Glu-286 domain of TaSG-D1 enhances protein stability in response to brassinosteroids, likely leading to formation of round grains in wheat. This gain-of-function mutation has pleiotropic effects on plant architecture and exhibits incomplete dominance. Haplotype analysis of 898 wheat accessions indicated that the origin of T. sphaerococcum in ancient India involved at least two independent mutations of TaSG-D1. Our results demonstrate that modest genetic changes in a single gene can induce dramatic phenotypic changes.

To date, six subspecies of hexaploid wheat (AABBDD) have been reported: Triticum aestivum, Triticum spelta, Triticum vavilovi, Triticum macha, Triticum compactum, and Triticum sphaerococcum. Triticum zhukovskyi was excluded from the list since it has a different genome constitution (AABBGG). T. sphaerococcum, also known as Indian dwarf wheat, is a landrace endemic to India and Pakistan (Ellerton, 1939;Hosono, 1954). Archaeological evidence supports the contribution of this wheat to ancient civilization in the Indus valley dating back to 9000 to 10,000 years ago (Josekutty, 2008;Mori et al., 2013). T. sphaerococcum exhibits semispherical grains, a semidwarf stature, short rigid culms, straight flag leaves, and compact and dense spikes (Percival, 1921;Sears, 1947). In addition, the grain has a shallow crease and high protein content compared to other wheat subspecies, making it a valuable genetic resource in wheat breeding programs (Singh, 1946). Thus, T. sphaerococcum was cultivated in the Indian subcontinent for thousands of years before the Green Revolution.
The spike-related morphological traits of hexaploid wheat, including seed threshability, spike compactness, and grain shape, are strongly influenced by three major genes: Q (free-threshing seeds; Mac Key, 1954;Faris et al., 2005), C (compact ears; Rao, 1972), and s (sphaerococcum grains; Rao, 1977;Salina et al., 2000). However, the genetic and biological basis of round grain development in T. sphaerococcum wheat is largely unknown. Miczynski (1930) crossed T. aestivum wheat with T. sphaerococcum and observed complete dominance of the T. aestivum trait over the T. sphaerococcum trait, indicating that the inheritance of the T. sphaerococcum trait is likely controlled by a single pleiotropic gene. Sears (1947) located the corresponding gene (denoted as s) on chromosome 3D by crossing T. sphaerococcum with nullisomics of wheat cv Chinese Spring and concluded that the T. sphaerococcum gene is hemizygous ineffective, (E) The Tasg-D1 gene was fine-mapped to a 1.01-Mb region between markers 3DS-68 and 3DS-94 using recombinants. The number of recombinants between the molecular marker and Tasg-D1 is indicated. (F) Analysis of genomic architecture and grain shape analysis for each recombinant type. The number of recombinants used for phenotypic analysis was indicated on the right. The black and white rectangles represent the homozygous ND4332 and HS2 regions, respectively. in contrast to the observation that the phenotypes segregate at a ratio of ;1(SS):2(Ss):1(ss; Ellerton, 1939). Subsequent studies confirmed this observation and mapped the s locus close to the centromere region of chromosome 3D (Rao, 1977;Koba and Tsunewaki, 1978;Singh, 1987). The sphaerococcum trait is not only restricted to the responsible gene in the D genome but also can be attributed to its homoeologs on the A and B genomes Johnson, 1963, 1966). Indeed, (G) Predicted high-confidence genes in the mapping region according to the International Wheat Genome Sequencing Consortium wheat genome. Colored arrows indicate the orientation and order of the annotated genes.
(H) Gene structure and sequence analysis of Tasg-D1 between HS2 and ND4332. The SNP information is indicated in red. (A) Gross morphology of ND4332 and its three independent EMS mutants. Bar 5 10 cm. (B) and (C) Spikelet (B) and grain morphology (C) of ND4332 and the mutants; bar 5 5 mm and 5 mm, respectively. Twenty grains were aligned for photograph. (D) to (I) ANOVA of plant height (D), spike length (E), spikelet density (F), thousand-grain weight (G), grain length (H), and grain length/width ratio (I) between ND4332 and the mutants. Ten plants were selected for trait analysis. (J) Gross morphology of Fielder (control) and OE-Tasg-D1 and OE-TaSG-D1 transgenic plants. Bar 5 10 cm. (K) and (L) Spikelet (K) and grain morphology (L) of Fielder (control), OE-Tasg-D1, and OE-TaSG-D1; bar 5 5 mm and 5 mm, respectively. Twenty grains were aligned for photograph. (M) to (R) ANOVA of plant height (M), spike length (N), spikelet density (O), thousand-grain weight (P), grain length (Q), and grain length/width ratio (R) among Fielder, OE-Tasg-D1, and OE-TaSG-D1. Eight plants were selected for data analysis. Values are means, with bars showing the SD. ** and * indicate significant differences at the 1 and 5% levels, respectively. (A) and (B) Comparison of root length (A) and dry mass (B) between NIL-Tasg-D1 and NIL-TaSG-D1 with different concentrations of BL at the seedling stage. NIL-Tasg-D1 and NIL-TaSG-D1 are recombinants with 90% background similarity, identified from the fine-mapping population. The BL concentrations were 0, 0.001, 0.01, and 0.1 mM, respectively. Three independent biological replicates were used (five plants per treatment).
(C) Expression analysis of BR biosynthesis-related and signaling-related genes in Fielder (control), OE-Tasg-D1, and OE-TaSG-D1. TaD11, TaBrd2, TaD2, and TaDwarf4 are BR biosynthesis-related genes. TaBRI1, TaBZR1, TaTUD1, TaRAVL1, and TaDLT are BR signaling-related genes. Values are means, with bars showing the SD. (D) and (E) Scanning electron microscopy of grain pericarp in HS2 and ND4332 (D) and statistical analysis (E). Bar 5 100 mm. A minimum of 100 cells per sample were measured using the instrument's software. Each parent was performed with three biological replicates. ANOVA method was conducted for statistical analysis. Values are means, with bars showing the SD. Asterisks indicate significance determined by ANOVA: *P < 0.05 and **P < 0.01. Salina et al. (2000) mapped the T. sphaerococcum genes s1, s2, and s3 on chromosomes 3D, 3B, and 3A, respectively.
Recent studies have highlighted the importance of brassinosteroids (BRs) in controlling grain shape through BR biosynthetic/ homeostatic or signaling pathways Jiang et al., 2013;Che et al., 2015;Zhu et al., 2015;Tian et al., 2019). BRdeficient mutants often exhibit smaller and less elongated seeds than the wild-type plants due to reduced cell length Che et al., 2015). The glycogen synthase kinase 3-like kinase (GSK) BRASSINOSTEROID INSENSITIVE2 (BIN2), a critical repressor of BR signaling, plays a crucial role in regulating grain shape. For example, loss of function of OsGSK3, a rice (Oryza sativa) ortholog of BIN2, alters the phosphorylation status of BRASSINAZOLE RESISTANT1 and induces its nuclear localization, leading to enhanced BR sensitivity and ultimately resulting in increased seed length (Gao et al., 2019). In addition, GSK2, another ortholog of BIN2, interacts with and represses the transcriptional activation activity of GROWTH REGULATION FACTOR4 in rice (Che et al., 2015), leading to reduced BR responses and small seeds (Che et al., 2015;Hu et al., 2015). However, the role of BR responses in determining grain shape in T. sphaerococcum remains unknown. Identifying the mechanism underlying this trait would elucidate the evolution of this subspecies.
Here, by performing positional cloning based on the semispherical seed phenotype, we isolated the subspecies-forming gene underlying grain shape in T. sphaerococcum. This gene, Tasg-D1, encodes a Ser/Thr protein kinase glycogen synthase kinase 3 (STKc_GSK3), the wheat ortholog of BIN2. In T. sphaerococcum, a single amino acid substitution of STKc_GSK3 enhances protein stability in response to BR, leading to round grain formation. Evolutionary analysis provided evidence that the origin of T. sphaerococcum wheat involved at least two independent mutations of TaSG-D1.

Map-Based Cloning of Tasg-D1
To explore the genetic basis of semispherical grain trait and its origin in T. sphaerococcum, we generated an F2 segregating population using wheat lines HeSheng 2 (HS2, common wheat) and Nongda 4332 (ND4332, derived from a cross between T. aestivum and T. sphaerococcum; Figure 1A), which exhibit significant differences in grain shape ( Figure 1B; Supplemental Data Set 1). The corresponding segregation ratio fits a Mendelian model of 3:1 (208 long grains/62 semispherical grains; x 2 < x 2 0.05, 1 5 3. 84), indicating that the semispherical grain trait is controlled by a single nuclear gene, which is consistent with previous findings (Sears, 1947;Salina et al., 2000).
The gene that determines semispherical grain (hereafter referred to as Tasg-D1 according to the official nomenclature rules in wheat; https://wheat.pw.usda.gov/GG2/Triticum/wgc/2013/; Ta, Triticum aestivum; sg, semispherical grain; D1, the first sg gene identified in genome D) was mapped between markers Xgwm341 and Xgdm72 on the short arm of chromosome 3D near the centromeric region ( Figure 1C). To perform fine mapping of Tasg-D1, we generated an F7 recombinant inbred line (RIL) population of 247 lines derived from a cross between HS2 and ND4332 and confirmed that the locus is located between markers 3DS-68 and 3DS-44 ( Figure 1D; Supplemental Data Set 2). We self-pollinated a residual heterozygous line (RHL) from the F7 RIL population to produce a segregating population and screened for recombinants. Using recombinant-derived progeny, we narrowed the candidate region to a 1.01-Mb region between markers 3DS-68 and 3DS-94 ( Figures 1E and 1F; Supplemental Data Set 2).
This region contains 13 predicted high-confidence genes, as annotated based on the Ensembl Plants database (http://plants. ensembl.org/index.html; Figure 1G; Supplemental Data Sets 3 and 4; Appels et al., 2018). Resequencing of the 13 predicted highconfidence genes in the 1.01-Mb region revealed only one singlenucleotide polymorphism (SNP) in the coding sequences (CDSs). This SNP (A/G) is located in the ninth exon of TraesCS3D01G137200 between ND4332 (Tasg-D1) and HS2 (TaSG-D1; Figure 1H). Although 253 SNPs and 32 insertions/ deletions were detected in the intergenic regions, we focused on the SNP in the coding region, which led to an amino acid substitution from Lys (286K) to Glu (286E; Figure 1H).
TraesCS3D01G137200 is predicted to encode an STKc_GSK3 kinase containing a Thr-283-Arg-284-Glu-285-Glu-286 (TREE) domain ( Figure 1H). TraesCS3D01G137200 is an ortholog of OsGSK1 in rice (Supplemental Figure 1A; Supplemental Data Set 3; Koh et al., 2007). Interestingly, the orthologous OsGSK1 gene in rice is involved in regulating stress responses (Koh et al., 2007), but not grain shape, indicating that the roles of these orthologs have diverged. By contrast, OsGSK2, a homolog of OsGSK1, helps regulate seed development (Tong et al., 2012). Phylogenetic analysis revealed that the TREE domain is highly conserved among different species, including monocots and dicots (Supplemental Figure 1B).
By using the publicly available expression data for wheat (eFP browser; http://bar.utoronto.ca/efp_wheat/cgi-bin/efpWeb.cgi), we predicted the expression profiles of TaSG-D1 in different tissues and found that it was highly expressed in shoot meristem, root, spike, grain, shoot axis, and ovary tissue. To verify these predictions, we performed reverse transcription quantitative PCR (RT-qPCR) analysis of TaSG-D1 in 13 tissue/time points, including spikes and grains at different developmental stages (since we were focusing on seed shape formation). As expected, the expression patterns of the candidate gene were consistent with the publicly available data, with high levels of expression in roots, immature spikes, and seeds (Supplemental Figure 2A).
TaSG1 has three homoeologs since wheat is a typical hexaploid containing three genomes. We investigated their expression patterns according to the publicly available data and found that the three homoeologs shared similar expression patterns (Supplemental Figure 2B).

Genetic Analysis of Tasg-D1
Previous studies have revealed a discrepancy in the inheritance pattern of the sphaerococcum gene, which either has a hemizygous-ineffective recessive effect or an incompletely dominant effect (Sears, 1947;Salina et al., 2000). Since we determined that Tasg-D1 is a gain-of-function allele in wheat line ND4332, we evaluated its genetic effects on grain shape and other traits using a segregating population (F8-RHL-II; see Methods). The grain length showed a 1:2:1 segregation ratio in the examined populations (x 2 5 0.46 < x 2 0.05, 2 5 5.99; Supplemental Figure 3), supporting the notion that Tasg-D1 shows incomplete dominance. Moreover, the phenotypes of heterozygous RHL-TaSG-D1/Tasg-D1 individuals were significantly different from those of homozygous RHL-TaSG-D1 and RHL-Tasg-D1 individuals, including grain length, plant height, spike length, spikelet density, and thousand-grain weight (Supplemental Data Set 5), suggesting that the effects of Tasg-D1 on plant architecture associated with the s locus are indeed pleiotropic.

Functional Analysis of Tasg-D1
The TREE domain likely plays a crucial role in GSK activity in plants (Li and Nam 2002;Koh et al., 2007;Tong et al., 2012). The overexpression of a mutant OsGSK2 allele with a missense mutation in the TREE domain led to the production of small round seeds in rice (Tong et al., 2012). To determine whether the STKc_GSK3 kinase is associated with the s locus responsible for grain shape, we obtained three independent mutants (M2 generation) of TaSG-D1 via ethyl methanesulfonate (EMS) mutagenesis in wheat line ND4332 (Supplemental Figure 4). Homozygous premature termination mutations resulted in plants with longer grains, larger spikes, and greater plant height than the wild-type plants (Figures 2A to 2I).
To investigate the potential functions of the TaSG-D1 homoeologs, we obtained a plant with a mutation in the TREE domain of the TaSG-A1 homoeolog (namely, line IIA236) in the Ningchun 4 (NC4) background by EMS mutagenesis (Supplemental Figure 5A). NC4 contained the same TaSG-A1 allele as HS2, and the NC4 plants exhibited tall stature, large leaf angles, and long grain shape. Interestingly, the IIA236 mutant showed contrasting phenotypic variation, including semispherical grains (Supplemental Figures 5B to 5G). This result is consistent with the finding that the Tasg-D1 gene homoeolog in the A genome contributes to the semispherical grain development Johnson, 1963, 1966;Salina et al., 2000). In addition, BLASTP analysis (match length, >70%; sequence identity, >70%; e value, <1 3 10 25 ) revealed 22 homologs of TaSG-D1 in the wheat genome. Only six of the homologs might be involved in negatively regulating the BR-mediated signaling pathway (according to the annotation in Ensembl Plant), indicating that neo-or subfunctionalization of the gene duplicates occurred independently during evolution, a topic that merits further investigation.
To further confirm our observations, we overexpressed the full CDS of this gene from HS2 (OE-TaSG-D1) and ND4332 (OE-Tasg-D1) driven by the Ubiquitin promoter in wheat cv Fielder. The overexpression of Tasg-D1 appeared to cause more severe phenotypic variation compared to TaSG-D1 ( Figures 2J to 2R). Specifically, the OE-Tasg-D1 lines had reduced grain length, whereas the OE-TaSG-D1 lines had increased grain length compared to the control ( Figures 2K, 2L, and 2Q). In addition, the OE-Tasg-D1 lines exhibited delayed flowering, decreased plant height and spike length, increased spikelet density, decreased thousand-grain weight, and the failure to produce ears in tillers with unelongated internodes compared to control and OE-TaSG-D1 plants ( Figures 2J and 2M to 2P). Together, these findings confirm the notion that the STKc_GSK3 kinase is the s gene and that a single amino acid substitution is sufficient to confer pleiotropic effects on phenotype, including grain shape, between T. sphaerococcum and other wheat subspecies.

Tasg-D1 Functions as a Negative Regulator of the BR Signaling Pathway
The homologs of TaSG-D1 in Arabidopsis and rice are negative regulators of BR signaling (Li and Nam, 2002;Koh et al., 2007;Tong et al., 2012). Therefore, to explore how Tasg-D1 regulates grain shape, we treated the roots of plants of the near-isogenic lines (NILs) NIL-TaSG-D1 and NIL-Tasg-D1 with epi-brassinolide (BL). Root growth was significantly promoted in plants treated with a low BL concentration (0.001 mM, P 5 1.3 3 10 26 ) and inhibited with a high BL concentration (0.1 mM, P 5 9.7 3 10 24 ), compared to untreated NIL-TaSG-D1 plants ( Figure 3A). In addition, the root dry mass significantly decreased with increasing BL concentration ( Figure 3B). By contrast, NIL-Tasg-D1 plants showed less sensitivity to BL application than NIL-TaSG-D1 plants, with only a gradual increase in root length in response to increasing BL concentrations from 0.001 to 0.1 mM. Moreover, the expression of BR biosynthesis and BR signaling-related genes decreased in both Tasg-D1 and TaSG-D1 overexpression plants compared to the wild type, including TaD11, TaBrd2, TaBZR1, TaTUD1, and TaRAVL1 ( Figure 3C). Consistent with the effects of BR on cell growth (Gasperini et al., 2012;Jiang et al., 2013;Che et al., 2015), cell length was significantly lower in ND4332 (Tasg-D1) than in HS2 (TaSG-D1), with a decrease of 12.97% (P 5 0.006; Figures 3D  and 3E), as revealed by scanning electron microscopy of the grain Three types of TREE domains were detected in 898 wheat accessions with varying ploidy levels. I indicates the HS2-type allele; II indicates the ND4332-type allele; and III indicates the third type in which the TREE domain was changed to TGEE. pericarp. These results indicate that Tasg-D1 has a negative effect on cell elongation.
To further investigate the effect of Tasg-D1 at the molecular level, we fused TaSG-D1 and Tasg-D1 with GFP (TaSG-D1-GFP and Tasg-D1-GFP respectively), and introduced them into Arabidopsis to test whether BR affects TaSG-D1 protein abundance. Similar to the results in wheat, Tasg-D1-GFP transgenic Arabidopsis plants showed more severe phenotypes than TaSG-D1-GFP plants (Supplemental Figure 6A). Furthermore, treatment with brassinazole, a specific inhibitor of BR biosynthesis, enhanced the dwarf phenotype of TaSG-D1-GFP plants but had little effect on Tasg-D1-GFP plants (Supplemental Figure 6B). BL treatment resulted in a rapid decrease in TaSG-D1-GFP abundance but no change in Tasg-D1-GFP abundance after 3, 6, 9, and 12 h of treatment (Supplemental Figure 6C). These results indicate that Tasg-D1 is more stable than TaSG-D1 in response to BR. Therefore, our results indicate that a single amino acid substitution in the TREE domain influences the accumulation of TaSG-D1.

The Origin of T. sphaerococcum
To explore the origin of T. sphaerococcum, we analyzed the sequences of the TREE domain of TaSG-D1 in 898 wheat accessions with varying ploidy levels. Of these, 14 T. sphaerococcum accessions (Supplemental Data Set 6) harbored two haplotypes, one identical to that of ND4332 and the other exhibiting an amino acid substitution from R to G in the TREE domain ( Figure 4). Interestingly, all of the examined Ae. tauschii accessions contained the same allele to HS2 (Supplemental Data Set 7). Therefore, we propose that the Tasg-D1 gene or related traits of T. sphaerococcum might have originated from T. aestivum due to a spontaneous mutation in the TREE domain of TaSG-D1 after hexaploidization. The mutations most likely occurred on two separate occasions. All of the other hexaploid wheat accessions with long grain shape contained the allele identical to that in HS2, including all T. spelta accessions (Supplemental Data Sets 8 and 9), indicating that the mutation likely occurred after domestication.

Potential Use of Tasg-D1 for Wheat Yield Improvement
Although Tasg-D1 itself had a negative effect on grain weight, as the NIL with the ND4332 allele (;28.9 g) exhibited significantly lower thousand-grain weight than the NIL containing the HS2 allele (;45.8 g; Supplemental Data Set 5), we evaluated the potential use of Tasg1-D1 combined with other yield-related genes in wheat breeding programs by comprehensively examining yieldrelated traits in various RILs. Compared with the parent HS2, six selected lines had compact plant architecture, with erect leaves,  decreased plant height (;11.2 cm, 12.4%; Figure 5A), reduced grain length/width ratio (;0.5, 23.8%; Figure 5B; Supplemental Figure 7), and more fertile spikelets per spike (;2.1, 11.3%; Figure 5C), but the thousand-grain weight was reduced in these lines by an average of 34.9%. However, the thousand-grain weight of these RILs was 56.1% (16.8 g) larger compared to that of the other parental line, ND4332 (;29.9 g; Figure 5D; Supplemental Figure 7). Collectively, these results indicate that combining Tasg-D1 with other favorable yield-related genes could improve wheat yield potential substantially, even though Tasg-D1 itself has a negative effect on grain weight.
Finally, elliptical grains produced by plants harboring Tasg-D1 could be beneficial for enhancing the flour extraction rate during wheat processing (Evers et al., 1990;Gegas et al., 2010). To test this hypothesis, we performed a flour extraction experiment using ND4332 (Tasg-D1) and HS2 (TaSG-D1) plants. The flour extraction rate was significantly lower in ND4332 (54.8%) versus HS2 (63.4%; Supplemental Figure 8), suggesting that elliptical grains might be one factor affecting the flour extraction rate. Other factors such as seed size and the depth of the crease might also contribute to the flour extraction rate. Thus, we conducted the experiment again using RILs with the same thousand-grain weight (;31.0 g). The flour extraction rate was indeed higher for semispherical grains (62.1%) than for long grains (57.4%; Supplemental Figure 8).
In conclusion, positional cloning and haplotype analysis of Tasg-D1, a gene affecting grain shape in wheat, broadened our view of the origin and evolution of T. sphaerococcum species. Introducing Tasg-D1 into modern wheat cultivars could facilitate the genetic engineering of grain shape and potentially improve the flour extraction rate.

Plant Materials
Line ND4332 was derived from a cross between common wheat and natural Triticum sphaerococcum wheat. HS2 is a common wheat (Triticum aestivum) cultivar with large grains. An F2 population of 270 individuals was generated by self-pollinating the F1 progeny of a cross between HS2 and ND4332 and used for genetic analysis and candidate gene mapping. A population comprising 247 RILs was developed by crossing HS2 and ND4332 and was advanced to the F7 generation by single-seed descent. We identified two residual heterozygous lines in the F7 RIL population, including one line (RHL-I) carrying the heterozygous segment in the genetic region from simple sequence repeat (SSR) markers 3DS-68 to 3DS-94 and the other (RHL-II) carrying the heterozygous segment from SSR markers 3DS-20 to 3DS-94. These two lines were self-pollinated to produce the F8 generations (F8-RHL-I and F8-RHL-II, respectively). The recombinants of F8-RHL-I were self-pollinated to generate a segregation population for fine-mapping of candidate genes. The F8-RHL-II population was used to analyze the effect of Tasg-D1 on grain shape and other agronomic traits.
Seeds from self-pollinated ND4332 plants were used to construct the EMS mutant population. More than 10,000 seeds were soaked in water for 12 h at room temperature and imbibed in a 0.4% (w/v) aqueous solution of EMS (M0880, Sigma-Aldrich) in the dark for 16 h at room temperature. The seeds were washed in water for 12 h and air-dried at room temperature. M1 plants with variant phenotypes were selected and self-pollinated to produce the M2 generation. In addition, IIA236, a sphaerococcum mutant, was identified from the EMS-mutagenized population of the spring wheat cv NC4 kindly provided by Ligeng Ma (Capital Normal University, China) and used for analysis of TaSG-A1.
The Fielder accession (United States spring hexaploid wheat) and Arabidopsis (Arabidopsis thaliana) Columbia (Col-0) ecotype were used for transformation. Transgenic wheat plants were grown in a greenhouse (400W HPS lamps, light intensity of 3000 lux) under 16 h of light at 24°C and 8 h of dark at 16°C. The transgenic Arabidopsis plants were grown in a greenhouse (white light, ;100 mmol m 22 s 21 ) at 22°C/18°C under a 16-hlight/8-h-dark cycle.

Scanning Electron Microscopy
HS2 and ND4332 seeds were cleaned, placed crease down onto aluminum specimen stubs, and sputter coated with gold. The grains were observed and photographed under an S-3400 scanning electron microscope (Hitachi; Wu et al., 2016). Cell length was measured in a minimum of 100 cells per sample using Photoshop 7.0 (Adobe). Each experiment was performed with three biological replicates using seeds from at least three plants per biological replicate. Analysis of variance (ANOVA) for cell size was conducted using SPSS 22.0 software with default parameters. Asterisks indicate significant differences determined by ANOVA (Supplemental File 1): *P < 0.05 and **P < 0.01.

Trait Evaluation
Seeds from the F2, F8-RHL-I, and F8-RHL-II segregating populations, the parent, and M0 seeds of ND4332 were planted in a seasonal growing location in Shangzhuang Experimental Station in Beijing, China, with 20 seeds per row (1.5-m-long rows). The F7 RIL population was grown in Beijing, Shandong, and Hebei using a randomized complete block design with three biological replicates. Each line was hand sown in two-row plots of 1.5-m-long rows with 30 seeds per row. The seeds of each M1 plant were planted in one row in Beijing. The three independent M2 lines and parent (ND4332) were grown in the greenhouse with 10 seeds per plot with three biological replicates. The Fielder and two independent transgenic plants of T1 generation were also grown in the greenhouse with six plants per row and two rows per pot. Agronomic traits, including plant height, spike length, and spikelet density, were examined before harvest. Thousand-grain weight, grain length, and grain width were determined using a cameraassisted phenotyping system (Wanshen Detection Technology). Statistical analysis was performed via one-way ANOVA using SPSS 22.0 software with default parameters. ** and * indicate significant differences at the 1 and 5% levels, respectively. A chi-square (x 2 ) test was used to test whether the distribution of the test statistic for each trait fit a chi-square distribution.

Marker Development
Microsatellite markers were designed and used to construct a genetic map of chromosome 3DS. The chromosome 3DS reference sequence of the Chinese Spring wheat (http://www.wheat genome.org/) and the Aegilops tauschii (accession AL8/78) genome sequence (http://aegilops.wheat. ucdavis.edu/ATGSP/) were used to design new SSR markers as described by Cheng et al. (2015). A parental polymorphism survey and identification of polymorphic SSR markers were conducted using the PCR conditions described by Cheng et al. (2015).
A cleaved amplified polymorphic sequence (CAPS) marker, 3DS-CAPS1, was developed as follows. Based on the sequence of rice gene Os01g0205700 (OsGSK1), the cDNA of the wheat orthologous gene was cloned and sequenced to detect polymorphisms between the two parents. Sequence analysis revealed an SNP in the cDNA sequences between the parents, that is, a MnlI site (CCTC) mutation in ND4332 (TCTC). This SNP was used to develop a chromosome 3DS specific CAPS marker (3DS-CAPS1). The PCR product (10 mL) was digested with restriction enzyme MnlI restriction enzymes for 4 h, and the digestion products were separated in 2% (w/v) agarose gels and visualized under UV light after ethidium bromide staining. As expected, 107-bp PCR products were completely cleaved by MnlI in HS2, but not in ND4332. Detailed information about the polymorphic markers used for genetic map construction is provided in Supplemental Data Set 10.

BL Treatment of Roots
Two individuals with 90% background similarity were selected from the F9 population, which were detected using 210 SSR primer pairs on 21 chromosomes. The two individuals had homozygous segments from markers 3DS-68 to 3DS-94 of ND4332 and HS2; the individuals were named NIL-Tasg-D1 and NIL-TaSG-D1, respectively. The seeds were surface sterilized with 75% (v/v) ethanol for 30 s and 1% (v/v) NaClO for 10 min and germinated for 1 d in Petri dishes. The germinated seeds were incubated at 4°C for 3 d in the dark and exposed to white light for 1 d. Seedlings with uniform growth were sown in boxes containing the appropriate BL (E1641, Sigma-Aldrich) concentration (0, 0.001, 0.01, and 0.1 mM) and grown in a greenhouse for 4 d (16 h of light at 24°C and 8 h of dark at 16°C; 75% humidity). Primary roots were stretched out with forceps and measured with a ruler. The experiment was repeated three times, with at least five plants per treatment. The roots of all five seedlings were removed and placed together in a single kraft-paper bags. Root dry mass was measured after the roots were dried for 3 d at 60°C.

Genome Resequencing
Genomic DNA from HS2 and ND4332 was used to build paired-end sequencing libraries with insert sizes of ;500 bp as described previously (Chai et al., 2018). We performed sequencing with an average 53 coverage of the assembled genome using the Illumina NovaSeq 6000 platform. Highquality reads were aligned to the reference genome of Chinese Spring (RefSeq v1.0; Appels et al., 2018) using the Burrows-Wheeler Aligner 0.7.15 program with default parameters (Li and Durbin, 2009). SNP and insertions/ deletions calling was performed using the HaplotypeCaller module. Illumina reads of all samples were deposited in the Bioproject at the National Center for Biotechnology Information (NCBI; https://www.ncbi.nlm.nih. gov/sra) under accession number PRJNA533588.

Gene Cloning and Sequence Analysis
Homoeolog-specific primer pairs were designed to amplify the genomic and full-length CDSs of the three TaSG1 homoeologs. The PCR assays were performed with high-fidelity Tks Gflex polymerase (TaKaRa) for 35 cycles (94°C for 1 min; cycles of 98°C for 10 s, 58°C for 15 s, and 68°C for 30 s; followed by 68°C for 10 min). The PCR products were confirmed using 1% (w/v) agarose gels and visualized under UV light after ethidium bromide staining. The purified PCR products were cloned into the pEASY-Blunt vector (TaKaRa) and sequenced. DNAMAN software and the online tool Conserved Domain Search (http://www.ncbi.nlm.nih.gov/Structure/cdd/ wrpsb.cgi) were used to analyze the nucleotide and amino acid sequences.

Phylogenetic Analysis
We downloaded the sequences of TaSG-D1 homologous proteins from various species, including both monocots and dicots, by performing a BLASTP (nr) search in the NCBI database (Supplemental File 2). The amino acid sequences were aligned with ClustalW, and a phylogenetic tree was constructed using the neighbor-joining method with MEGA 6.06 software with default parameters (Tamura et al., 2013). The evolutionary distances were calculated using Poisson model. The phylogeny test was computed using bootstrap method with 1000 replications (Supplemental File 3).

RT-qPCR Analysis
For expression analysis, 13 different tissues from HS2 and leaves from transgenic wheat lines and Arabidopsis lines were collected from at least three plants per biological replicate for RNA extraction. Total RNA was extracted from the samples using the standard TRIzol RNA isolation protocol (Invitrogen). The detailed protocols and procedures were as described by Han et al. (2016). RT-qPCR was performed using SYBR Green PCR Master Mix (TaKaRa) with a CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.). The RT-qPCR conditions and analytical methods were the same as those used in Wang et al. (2018). The wheat b-ACTIN gene (Paolacci et al., 2009) was used as an endogenous control. The specific primer pairs used for amplification are listed in Supplemental Data Set 10. Each experiment was performed with three biological replicates.

Vector Construction and Plant Transformation
For wheat, the full-length CDS of TaSG-D1 and Tasg-D1of HS2 and ND4332 were amplified using specific primer pairs and inserted into the pWMB110 vector via homologous recombination technology with a ClonExpress II One Step Cloning Kit (Vazyme). The sequencingconfirmed vectors were transformed into 15-d-old immature wheat embryos via Agrobacterium-mediated transformation.
For Arabidopsis, the 35S:TaSG-D1-GFP and 35S:Tasg-D1-GFP constructs were generated by inserting the full-length CDS of HS2 and ND4332 into a modified pCAMBIAsuper1300 vector containing the green fluorescent protein (GFP) gene. The resulting vectors were introduced into Agrobacterium tumefaciens GV3101 and transformed into the wild-type Arabidopsis plants using the floral dip method (Clough and Bent, 1998). All primers used for PCR amplification are listed in Supplemental Data Set 10.

Chemical Treatment of Arabidopsis Seedlings
As described by Peng et al. (2008), seeds from two independent transgenic Arabidopsis lines of T2 generation were germinated on one-half-strength Murashige and Skoog (MS) medium supplemented with 2 mM brassinazole (SML1406, Sigma-Aldrich), a specific inhibitor of BR biosynthesis, and grown in a greenhouse at 22°C under a 16-h-light/8-h-dark cycle for 2 weeks. The seedlings were transferred to half strength MS liquid medium containing 1 mM BL and incubated for 0, 3, 6, 9, and 12 h prior to immunoblot analysis.

Immunoblot Analysis
Transgenic Arabidopsis seedlings grown on half strength MS medium with or without chemical treatment were collected, transferred to liquid nitrogen, and ground into a fine powder. Total protein was extracted from the samples via incubation at 100°C for 10 min in 53 SDS-PAGE sample buffer (Zhang, et al., 2017), followed by centrifugation at 12,000g for 10 min at room temperature. The protein samples were separated in an 8% (v/v) SDS-PAGE gel for 1 h and 10 min, transferred to a polyvinylidene difluoride STKc_GSK3 Kinase Mutation Affects Grain Shape membrane (Millipore), and blocked with 5% (w/v) fat-free milk in Trisbuffered saline with 0.1% (v/v) Tween-20 (TBST; Guo et al., 2015) for 1 h. The membrane was incubated for 1 h in antibodies [anti-ACTIN antibody [A0480, Sigma] and anti-GFP antibody [ab290, Abcam] diluted 1:5000), washed three times (10 min each) with TBST, incubated with secondary antibody diluted 1:5000 for 1 h, and washed three times (5 min each) with TBST. Specific protein bands were visualized with Immobilon Western Chemiluminescent horseradish peroxidase substrate (http://www. millipore.com). ACTIN served as a loading control.

Flour Extraction Experiment
The flour extraction rate was examined using 300 g of seeds from HS2, ND4332, and two RILs harvested in Beijing in 2017. The thousand-grain weight of HS and ND4332 was 67.4 and 25.0 g, respectively, whereas the two RILs had the same thousand-grain weight (31.0 g). Seed moisture content was measured with a Grain analyzer machine (Infratec 1241, FOSS), and water was added until the moisture content rose to 14%. The seeds were stored at room temperature for 24 h before starting the flour extraction experiment using a flour-milling machine (CD1, Chopin Moulin). Each sample was performed with three biological replicates. Finally, ANOVA was performed to calculate differences in flour extraction rates using SPSS 22.0 software with default parameters. Asterisks indicate significant differences determined by ANOVA: *P < 0.05 and **P < 0.01.