ADP-glucose pyrophosphorylase genes, associated with kernel weight, underwent selection during wheat domestication and breeding.

ADP-glucose pyrophosphorylase, comprising two small subunits and two large subunits, is considered a key enzyme in the endosperm starch synthesis pathway in wheat (Triticum aestivum L.). Two genes, TaAGP-S1-7A and TaAGP-L-1B, were investigated in this study. Haplotypes of these genes were associated with thousand kernel weight (TKW) in different populations. Mean TKWs of favoured haplotypes were significantly higher than those of nonfavoured ones. Two molecular markers developed to distinguish these haplotypes could be used in molecular breeding. Frequencies of favoured haplotypes were dramatically increased in cultivars released in China after the 1940s. These favoured haplotypes were also positively selected in six major wheat production regions globally. Selection of AGP-S1 and AGP-L-1B in wheat mainly occurred during and after hexaploidization. Strong additive effects of the favoured haplotypes of with other genes for starch synthesis were also detected in different populations.

Introduction

Increasing the yield of wheat (Triticum aestivum) will become a greater challenge in the future. Although average yields in China increased by 2.12% in the last 10 years (http://faostat3.fao.org), the annual rate of increase is gradually decreasing. Yield potential can be increased in three ways: higher spike number per unit area (SN), higher grain number per spike (GN) and higher thousand kernel weight (TKW). Crop yields at the physiological level are determined mainly by source and sink strengths (Rossi et al., 2015). In cereals such as wheat and barley (Hordeum vulgare), sink capacity has proved more important than source accumulation (Serrago et al., 2013); therefore, improvement in the expression of key enzymes involved in starch biosynthesis in the endosperm may be an effective way to increase yield in wheat (Bahaji et al., 2014; Hou et al., 2014).

ADP‐glucose pyrophosphorylase (AGPase) is a rate‐limiting enzyme in the synthesis of starch in plant endosperm (Dickinson and Preiss, 1969; Müller‐Röber et al., 1992; Tsai and Nelson, 1966), that is the catalysis of glucose‐1‐Pi to ADP‐glucose. AGPase in most plants is a tetrameric protein complex comprising two large subunits (LSUs) and two small subunits (SSUs; Copeland and Preiss, 1981; Okita et al., 1990). The LSUs mainly have regulatory roles and the SSUs have catalytic roles, and recent studies indicate that LSUs also contribute substrate binding (Cakir et al., 2015; Hwang et al., 2008; Kavakli et al., 2001). AGPases are activated by 3‐phosphoglyceric acid and inhibited by inorganic phosphate (Boehlein et al., 2013; Figueroa et al., 2013; Heldt et al., 1977; Neuhaus et al., 1989). In nonphotosynthetic cells, AGPases have different localizations in dicots and monocots (Comparot‐Moss and Denyer, 2009). AGPases are mainly active in the plastids of dicots, whereas in monocots they are active in both the cytosol and plastids, with greater activity occurring in the cytosol. The proportion of cytosolic activity in the endosperm may be as high as 85% in barley (Thorbjørnsen et al., 1996), 95% in maize (Zea mays) (Denyer et al., 1996) and 94% in wheat (Burton et al., 2002).

AGP mutation and overexpression are possible ways to increase yield. Heterologous expression of Escherichia coli glgC16 in rice (Oryza sativa) increased seed weight by 7%–24% (Nagai et al., 2009; Sakulsingharoj et al., 2004), in maize by 13%–25% (Wang et al., 2007) and tuber yield in potato (Solanum tuberosum) by 35% (Stark et al., 1992). The Shrunken2 (Sh2) and Brittle2 (Bt2) loci in maize endosperm separately encode the large and small subunits (Hannah and Nelson, 1976). Expression of the Sh2r6hs allele with enhanced heat stability and reduced inorganic phosphate inhibition increased the yield of wheat by 38% (Smidansky et al., 2002) and of rice by 23% (Smidansky et al., 2003). Transgenic maize lines simultaneously overexpressing the Bt2 and Sh2 alleles increased starch content by 74% and seed weight by 15% compared to wild‐type (WT) lines (Li et al., 2011). Overexpression of LSU in wheat lines enhanced AGPase activity and increased TKW by 5.2%–9.1% compared to WT (Kang et al., 2013).

Most crops undergo significant genetic changes during domestication and breeding (Shi and Lai, 2015). Previous studies indicated that SSUs were more conserved than LSUs, yet both AGPase subunits are sensitive to enzyme activity that leads to amino acid changes (Georgelis et al., 2007) and have undergone selection in maize during domestication and breeding (Corbi et al., 2011). In this study, genes TaAGP‐S1‐7A and TaAGP‐L‐1B were mainly investigated. Four haplotypes of each gene were detected and were associated with TKWs when assayed in near‐isogenic lines (NILs), a mini core collection (MCC) and modern cultivars (MC). Based on their effects, these haplotypes were artificially classified into subgroups Hap‐I and Hap‐II and cleaved amplified polymorphic site (CAPS) markers were developed to identify them. There has been strong selection for the two favoured haplotypes over the past seventy years of wheat breeding in China. These favoured haplotypes also occur at high frequencies in cultivars released in other geographic regions during the last century.

Results

Cloning by homology and chromosome locations of TaAGP‐S and TaAGP‐L

Full‐length homologous cDNA and genomic DNA sequences of TaAGP‐S1 (GenBank: EF405961, cytosolic), TaAGP‐S2 (AY727927, plastidic) and TaAGP‐L (DQ839506, cytosolic) were cloned from Chinese Spring (CS). TaAGP‐S1‐A and TaAGP‐L‐D were located on chromosomes 7AL (scaffold53517, C‐7AL1‐0.39) and 1DL (scaffold2135, bin203), respectively, by sequence alignments to the Triticum urartu and Aegilops tauschii genome sequences. TaAGP‐S1, TaAGP‐S2 and TaAGP‐L were located on homoeologous chromosome groups 7, 5 and 1, respectively, and verified by PCR amplification of DNA from CS and its nulli‐tetrasomic lines with genome‐specific primers (Table S1).

Sequence lengths of coding regions of TaAGP‐S1‐7A, ‐7B and ‐7D were 7531, 6245 and 6072 bp, respectively; each had nine exons and eight introns. TaAGP‐S1‐7A has an alternative first exon in the first intron that encodes plastid SSU in leaves (Comparot‐Moss and Denyer, 2009), and also has a 1.1‐kb insertion in the first intron compared to ‐7B and ‐7D. This inserted region had a GC content above 70% and was also present with high homology in AGP‐S1 in T. urartu.

Sequence lengths of the coding regions of TaAGP‐S2‐5A, ‐5B and ‐5D were 3732, 3774 and 3771 bp, respectively; each had nine exons and eight introns. Sequence lengths of coding regions of TaAGP‐L‐1A, ‐1B and ‐1D were 3332, 3351 and 3340 bp, respectively; each had 15 exons and 14 introns. Although the three genes within each homoeologous group had homologies above 92%, the small and large subunit genes have very low homology at both the cDNA (52%–54%) and DNA (40%) levels (Figure S1).

Genetic haplotypes and their effects on TKW in a MCC and NILs

We sequenced TaAGP‐S1, TaAGP‐S2 and TaAGP‐L in 245 accessions from the MCC, and polymorphic sites were detected only in TaAGP‐S1‐7A and TaAGP‐L‐1B (Figure 1).

Figure 1

Haplotypes of Ta and Ta. (a) Coding regions of Ta. ▼SNP at position 5090. (b) Coding and 2‐kb upstream regions, and polymorphic sites of Ta. ▼SNP at position ‐122. Numbers indicate deletion size (bp). Vertical thin lines indicate polymorphic site differences between haplotypes.

A SNP at position 5090 in the third exon of TaAGP‐S1‐7A (Figure 1a) led to an amino acid change (GCA→TCA, Ala→Ser). Dendrogram analysis showed that this site was Ala in most species and the amino acid sequences around it were highly homologous (Sarma et al., 2014). In potato, this site was A196 and next to E197 and K198 in AGP‐L (Jin et al., 2005). E197 and K198 are essential for binding substrate ADP‐glucose. Therefore, the change from Ala to Ser might result in lower enzyme activity associated with an allosteric change (Figure 2). Interestingly, this site was also Ser in plastid TaAGP‐S2‐5A, ‐5B and ‐5D. In a previous study, the activities of cytosolic SSUs in wheat endosperm reached 94%, and activities of plastid SSUs accounted for only 6% of the total (Burton et al., 2002). Thus, this SNP might cause reduced enzyme activity.

Figure 2

Key polymorphic sites on protein sequences of TaAGP‐S1‐7A‐Hap‐I and ‐Hap‐II. (a) Partial amino acid (aa) sequence alignments of TaAGP‐S1, TaAGP‐S2, ZmAGP‐S (maize), OsAGP‐S (rice) and StAGP‐S (potato). Key site differences between TaAGP‐S1‐7A‐Hap‐I and ‐Hap‐II are in the rectangle. (b) Secondary protein structure of StAGP‐S (30 aa) predicted by the SWISS‐MODEL (Arnold et al., 2006; Guex and Peitsch, 1997; Schwede et al., 2003).

Another eight SNPs and two insertions/deletions (InDels) were identified among the 245 sequenced accessions, forming four haplotypes, named ‐Hap‐1, ‐Hap‐2, ‐Hap‐3 and ‐Hap‐4 (Figure 1a). The numbers of accessions in the ‐Hap‐2 and ‐Hap‐4 groups were quite small, at four and four in landraces compared to nine and zero in modern cultivars. Only the SNP (G/T) at position 5090 in the third exon caused an amino acid difference between ‐Hap‐1/2 and ‐Hap‐3/4; other polymorphic sites were all in introns. Significant differences in TKW were detected between ‐Hap‐1 and ‐Hap‐3 in both landraces and modern cultivars of the MCC (Table 1; Figure S2a). There were no significant differences in SN, GN and TKW between ‐Hap‐1 and ‐Hap‐2 in 2014 plot tests (Table S2; Figure S3a). The relative expression level of ‐Hap‐1 was not significantly different from that of ‐Hap‐2 at 15 days postanthesis (DPA) (Table S3; Figure S3b), suggesting that ‐Hap‐1 and ‐Hap‐2 had similar effects on TKW.

Table 1

Mean TKWs of TaAGP‐S1‐7A and TaAGP‐L‐1B haplotypes and their combinations in the MCC

	Landraces in MCC				Modern cultivars in MCC
	Number of accessions	2002	2005	2006	Number of accessions	2002	2005	2006
TaAGP‐S1‐7A
Hap‐1	28	34.53 ± 1.20a	34.17 ± 1.54a	35.66 ± 1.41a	60	40.55 ± 0.58a(A)	40.23 ± 0.70a(A)	42.52 ± 0.72a(A)
Hap‐2	4	31.56 ± 1.06ab	28.40 ± 0.97ab	32.51 ± 1.41ab	9	37.81 ± 1.76ab(AB)	35.28 ± 2.70ab(AB)	38.88 ± 1.92ab(AB)
Hap‐3	120	31.58 ± 0.38b	29.94 ± 0.54b	32.41 ± 0.44b	19	35.71 ± 1.09b(B)	34.24 ± 1.27b(B)	36.33 ± 1.27b(B)
Hap‐4	4	32.83 ± 0.77ab	30.86 ± 0.98ab	34.88 ± 0.29ab
Hap‐I	32	34.16 ± 1.07a	33.35 ± 1.38a(A)	35.27 ± 1.26a	69	40.19 ± 0.56a(A)	39.70 ± 0.71a(A)	42.05 ± 0.68a(A)
Hap‐II	124	31.62 ± 0.36b	29.97 ± 0.53b(B)	32.49 ± 0.43b	19	35.71 ± 1.09b(B)	34.24 ± 1.27b(B)	36.33 ± 1.27b(B)
TaAGP‐L‐1B
Hap‐1	23	34.16 ± 0.98a	34.49 ± 1.16a(AB)	36.09 ± 1.16a(A)	22	41.19 ± 1.08a	39.93 ± 1.28a	43.07 ± 1.06a
Hap‐2	9	35.77 ± 2.50a	37.86 ± 2.92a(B)	36.30 ± 2.79ab(AB)	22	39.53 ± 0.86ab	38.80 ± 1.10a	41.22 ± 1.00ab
Hap‐3	30	32.09 ± 0.56ab	30.18 ± 0.83b(AC)	33.01 ± 0.67ab(AB)	30	38.50 ± 0.90ab	38.79 ± 1.19a	40.22 ± 1.23ab
Hap‐4	65	30.98 ± 0.49b	28.82 ± 0.59b(C)	31.58 ± 0.57b(B)	13	37.23 ± 1.56b	36.00 ± 2.09a	37.88 ± 1.93b
Hap‐I	62	32.97 ± 0.51a(A)	31.90 ± 0.73a(A)	34.10 ± 0.60a(A)	74	39.60 ± 0.56a	39.14 ± 0.68a	41.37 ± 0.67a
Hap‐II	65	30.98 ± 0.49b(B)	28.82 ± 0.59b(B)	31.58 ± 0.57b(B)	13	37.23 ± 1.56a	36.00 ± 2.09a	37.88 ± 1.93b
SI+LI	29	34.48 ± 1.14a(A)	34.07 ± 1.47a(A)	35.64 ± 1.33a(A)	61	40.32 ± 0.59a(A)	39.94 ± 0.73a(A)	42.25 ± 0.71a(A)
SI+LII	3	31.05 ± 2.48a(A)	27.30 ± 1.76a(A)	31.62 ± 3.87a(A)	8	39.20 ± 1.85ab(AB)	37.73 ± 2.79ab(AB)	40.52 ± 2.46ab(AB)
SII+LI	62	32.26 ± 0.52a(A)	30.94 ± 0.82a(A)	33.39 ± 0.63a(A)	13	36.22 ± 1.16b(B)	34.56 ± 1.23b(B)	37.24 ± 1.42b(B)
SII+LII	62	30.98 ± 0.50b(B)	28.90 ± 0.62b(B)	31.58 ± 0.58b(B)	6	34.60 ± 2.45b(B)	33.60 ± 3.14b(AB)	34.36 ± 2.63b(B)

SI, SII: TaAGP‐S1‐7A‐Hap‐I (‐Hap‐1 and ‐2), ‐Hap‐II (‐Hap‐3 and ‐4). LI, LII: TaAGP‐L‐1B‐Hap‐I (‐Hap‐1, ‐2 and ‐3), ‐Hap‐II (‐Hap‐4).

Different capital and small letters indicate significant differences between haplotypes at P < 0.01 and P < 0.05, respectively.

Nine SNPs in the coding region, and 13 SNPs and five InDels in the 5′ UTR and promoter regions of TaAGP‐L‐1B formed four haplotypes, ‐Hap‐1, ‐Hap‐2, ‐Hap‐3 and ‐Hap‐4 (Figure 1b). No amino acid change was detected within any of the four haplotypes. Significant differences in TKW were detected between ‐Hap‐1 and ‐Hap‐4 in landraces and modern cultivars (Table 1; Figure S2b), whereas no significant differences were detected between ‐Hap‐1, ‐Hap‐2 and ‐Hap‐3. SN, GN, and TKW data for ‐Hap‐1 and ‐Hap‐3 in 2014 plot tests were not significantly different (Table S2; Figure S4a). The relative expression level of ‐Hap‐1 was not significantly different from that of ‐Hap‐3 at 20 DPA (Table S3; Figure S4b). TKW of ‐Hap‐1 was not significantly different from ‐Hap‐2, but was significantly higher than ‐Hap‐4 by 6.0 and 6.8 g in 2011 and 2012 plot tests (Table S2; Figure S4c, d).

In the case of TaAGP‐S1‐7A, ‐Hap‐1 and ‐Hap‐2 were combined as ‐Hap‐I, and ‐Hap‐3 and ‐Hap‐4 were combined as ‐Hap‐II due to their similar effects. TaAGP‐L‐1B‐Hap‐1, ‐Hap‐2 and ‐Hap‐3 were classified as ‐Hap‐I, and ‐Hap‐4 became ‐Hap‐II. The TKW of ‐Hap‐I was significantly higher than that of ‐Hap‐II in landraces and modern cultivars (Table 1; Figure S2c, d). The frequency of TaAGP‐S1‐7A‐Hap‐I was 20.5% in landraces compared to 78.4% in modern cultivars, and the frequency of TaAGP‐L‐1B‐Hap‐I was 48.8% in landraces compared to 85.1% in modern cultivars.

The distinct haplotype effect of TaAGP‐L‐1B was caused by variation in one SNP in the promoter region

There were three SNPs between ‐Hap‐I and ‐Hap‐II, at positions −122 (T–C), 150 (C–T) and 283 (G–C) in TaAGP‐L‐1B. The SNP at position 150 involved a synonymous mutation (AAC‐AAT) in the first exon, position 283 was in the first intron, and position −122 was located in an E2F‐DP‐binding motif (WTTSSCSS). Rose et al. (2015) reported that 47 wheat cultivars genotyped at TaAGP‐L‐1B consisted of five haplotypes, and H1 to H4 in their study could be regarded as ‐Hap‐I in the present research. The SNP at position 182 (−122 in this study) was not mentioned because this SNP was present only at H5 in their study, and only in one cultivar (Chinese Spring).

To detect whether the SNP at position −122 caused a difference, rice seeds transformed with ‐Hap‐I and ‐Hap‐II promoter‐GUS were employed to measure promoter‐driven activity. Gus expression was mainly located in the aleurone layer at 21 DPA (Figure 3a, b), as also reported by Thorneycroft et al. (2003). Gus activity in developing seeds transformed with ‐Hap‐I was significantly higher than in those transformed with ‐Hap‐II (Table S4; Figure 3c). This suggested that the SNP difference at position −122 caused the phenotypic variation.

Figure 3

GUS staining and expression driven by the promoters of Ta, ‐Hap‐2, ‐Hap‐3 and ‐Hap‐4 in developing seeds at 21 DPA. (a) GUS staining of whole seeds in wild type (WT) and four haplotypes. (b) GUS staining of half‐seeds in WT and four haplotypes. (c) activities in seeds of WT and four haplotypes.

Haplotype effects associated with TKW and their selection in the MC

Two cleaved amplified polymorphic site (CAPS) markers based on restriction enzyme digestion were developed to distinguish each haplotype. Position 5090 of TaAGP‐S1‐7A‐Hap‐II (T) fell within the restriction site CTNAG that was cleaved by restriction endonuclease DdeI (Figure 4a) although about 3.0 kb of sequence around the SNP was highly conserved. A 3.3‐kb genome‐specific fragment was amplified by primers S1‐7A‐M1F and S1‐7A‐M1R in the first step, and a 581‐bp fragment was amplified by primers S1‐7A‐M2F and S1‐7A‐M2R using the PCR products as template following a 25× dilution. The 581‐bp TaAGP‐S1‐7A‐Hap‐II fragment was digested into 333‐ and 248‐bp subcomponents by DdeI, whereas the ‐Hap‐I fragment was not cleaved. Position 150 of TaAGP‐L‐1B‐Hap‐II (T) fell within the restriction site GCAATGNN that was cleaved by BSrDI (Figure 4b). The 460‐bp genome‐specific fragment amplified by primers L‐1B‐MF and L‐1B‐MR was cleaved into 359‐ and 101‐bp subfragments by BSrDI, whereas ‐Hap‐I was not cleaved.

Figure 4

CAPS markers discriminating Ta and Ta haplotypes. PCR products were restrictively digested by DdeI (a) and (b).

When 348 MC accessions were assayed by the two CAPS markers, the mean TKW of TaAGP‐S1‐7A‐Hap‐I was significantly higher than that of ‐Hap‐II by 5.7 g in 2002, 4.9 g in 2006 and 4.7 g in 2010 (Table 2; Figure 5a). The TKW of TaAGP‐L‐1B‐Hap‐I was significantly higher than ‐Hap‐II by 4.2 g in 2002, 3.5 g in 2006 and by 3.7 g in 2010 (Table 2; Figure 5b). Each ‐Hap‐I was defined as a favoured haplotype relative to the corresponding ‐Hap‐II. The proportion of favoured haplotypes gradually rose during six decades of breeding, reaching 97% in the 1990s (Figure 5c, d).

Table 2

Mean TKWs of TaAGP‐S1‐7A and TaAGP‐L‐1B haplotypes and their combinations in the MC

	Number of accessions	2002	2006	2010
TaAGP‐S1‐7A
Hap‐1	248	43.16 ± 0.40a(A)	40.69 ± 0.39a(A)	40.82 ± 0.39a(A)
Hap‐2	23	42.05 ± 1.15ab(AB)	37.39 ± 1.52b(AB)	39.71 ± 1.04ab(AB)
Hap‐3	58	37.32 ± 0.89b(B)	35.50 ± 0.96b(B)	36.01 ± 0.79b(B)
Hap‐I	271	43.07 ± 0.38a(A)	40.41 ± 0.38a(A)	40.72 ± 0.36a(A)
Hap‐II	58	37.32 ± 0.89b(B)	35.50 ± 0.96b(B)	36.01 ± 0.79b(B)
TaAGP‐L‐1B
Hap‐1	130	40.74 ± 0.52a(A)	43.08 ± 0.53a(A)	40.84 ± 0.56a(A)
Hap‐2	79	39.91 ± 0.86a(A)	42.53 ± 0.73a(A)	40.77 ± 0.66a(A)
Hap‐3	92	39.04 ± 0.75ab(AB)	41.77 ± 0.78a(AB)	39.62 ± 0.69a(AB)
Hap‐4	44	36.55 ± 0.95b(B)	38.33 ± 1.07b(B)	36.78 ± 0.94b(B)
Hap‐I	301	42.53 ± 0.38a(A)	40.01 ± 0.39a(A)	40.44 ± 0.36a(A)
Hap‐II	44	38.33 ± 1.07b(B)	36.55 ± 0.95b(B)	36.78 ± 0.94b(B)
SI+LI	245	43.41 ± 0.39a(A)	40.67 ± 0.41a(A)	41.04 ± 0.39a(A)
SI+LII	24	39.73 ± 1.31ab(AB)	37.70 ± 0.99b(AB)	38.20 ± 1.07ab(AB)
SII+LI	41	37.82 ± 1.02b(B)	35.76 ± 1.13b(B)	36.34 ± 0.96b(B)
SII+LII	16	36.36 ± 1.86b(B)	34.71 ± 2.01b(B)	35.44 ± 1.53b(B)

SI, SII: TaAGP‐S1‐7A‐Hap‐I (‐Hap‐1 and ‐2), ‐Hap‐II (‐Hap‐3 and ‐4). LI, LII: TaAGP‐L‐1B‐Hap‐I (‐Hap‐1, ‐2 and ‐3), ‐Hap‐II (‐Hap‐4).

Different capital and small letters indicate significant differences between haplotypes at P < 0.01 and P < 0.05, respectively.

Figure 5

Mean TKWs of two genes in the MC. (a) TKWs of Ta and ‐Hap‐. (b) TKWs of Ta and ‐Hap‐. (c) Ta and ‐Hap‐ frequencies in cultivars released in China from the 1940s to 2010. (d) Frequencies of Ta and ‐Hap‐. **P < 0.01, ***P < 0.001.

Haplotype combinations and selection within the MCC and MC

To determine whether there were additive effects of favoured haplotypes at the two loci, we combined haplotypes and tested their effects on TKW in the two collections. TaAGP‐S1‐7A‐Hap‐I and ‐Hap‐II were assigned as S‐I and S‐II, and TaAGP‐L‐1B‐Hap‐I and ‐Hap‐II were named as L‐I and L‐II, respectively. The mean TKW of the S‐I + L‐I cohort was significantly higher than that of S‐II + L‐II in both populations, and the proportion of accessions with the favoured combination also gradually rose over time (Table 1; Figure S5). TKW differences among the four haplotype combinations clearly indicated the existence of a genetically additive effect of the favoured haplotypes.

Geographic distributions of haplotypes and additive effects with other important starch synthesis genes in world cultivar populations

Frequencies of favoured haplotypes at TaAGP‐S1‐7A and TaAGP‐L‐1B were assessed by CAPS markers in 384 European, 447 North American, 53 CIMMYT, 82 Russian and 51 Australian cultivars (Figure S6a,b). These six regions account for about 60% of global wheat production (http://faostat3.fao.org). Favoured haplotypes at TaAGP‐S1‐7A and TaAGP‐L‐1B reached very high frequencies (from 82% to 100%) in all five groups, indicating they had undergone overall positive selection in wheat breeding.

We previously analysed genes encoding sucrose synthase I and II in the starch synthesis pathway, and determined their haplotypes and global distributions in cultivars released during a century of wheat breeding (Hou et al., 2014). Haplotype effects on TKW between favoured and nonfavoured haplotypes at six loci (TaAGP‐S1‐7A, TaSus1‐7B, TaAGP‐L‐1B, TaSus2‐2A, TaSus1‐7A and TaSus2‐2B) were compared in the MC (Figure S6c). The largest difference was found at TaAGP‐S1‐7A. The frequencies of favoured haplotypes reached very high levels at all loci in cultivars released in Europe, North America and China (Figure S6d).

As expected, cultivars with more favoured haplotypes had higher mean TKWs in the MC and MCC (Table 3; Figure 6a, b). The frequencies of cultivars with four favoured haplotypes at TaAGP‐S1‐7A, TaSus1‐7B, TaAGP‐L‐1B and TaSus2‐2A increased during six decades of Chinese, European and North American wheat improvement (Figure 6c, d), with Chinese cultivars rising fastest.

Table 3

Frequency of favoured haplotypes (FH) and mean TKWs in the MCC and MC

Population	Number of FH	Number of accessions	2002	2005	2006	Population	Number of FH	Number of accessions	2002	2006	2010
Landraces in the MCC	0	9	30.65 ± 1.22ab	27.43 ± 1.23a(A)	31.61 ± 1.50ab	MC	0	6	31.18 ± 3.45a(AB)	31.60 ± 2.03a(A)	31.32 ± 1.99a(A)
	1	65	31.06 ± 0.50a	29.35 ± 0.67a(A)	31.68 ± 0.58a		1	21	34.45 ± 1.78a(B)	31.74 ± 1.55a(A)	33.08 ± 1.14a(AB)
	2	52	32.62 ± 0.55ab	31.09 ± 0.86a(AB)	33.90 ± 0.67ab		2	46	39.15 ± 0.80a(BC)	36.30 ± 0.75b(AB)	37.93 ± 0.78ab(AB)
	3	17	33.09 ± 1.35ab	31.03 ± 1.54ab(AB)	33.72 ± 1.61ab		3	114	42.54 ± 0.57b(AC)	39.75 ± 0.62c(BC)	40.11 ± 0.64bc(C)
	4	14	35.26 ± 1.86b	36.97 ± 2.33b(B)	36.45 ± 2.10b		4	157	43.92 ± 0.51b(A)	41.45 ± 0.50c(C)	41.67 ± 0.45c(C)
Modern cultivars in the MCC	1	8	34.12 ± 1.98a(A)	33.62 ± 2.57a	34.33 ± 2.17a(A)
	2	10	36.55 ± 1.30ab(AB)	34.54 ± 1.96a	37.26 ± 1.56ab(AB)
	3	24	40.50 ± 0.99b(B)	39.79 ± 1.30a	42.07 ± 1.15b(B)
	4	46	40.03 ± 0.67b(B)	39.57 ± 0.82a	42.06 ± 0.83b(B)

Different capital and small letters indicate significant differences between haplotypes at P < 0.01 and P < 0.05, respectively.

Figure 6

TKW effects and frequency changes of cultivars with different numbers of favoured haplotypes (FH) at Ta, TaSus1‐7B, Ta and TaSus2‐2A in different populations. (a) TKW differences in cultivars with different FH in the MCC. (b) TKW differences in the MC accessions. (c) Frequency changes among accessions with different FH in the MC over six decades. (d) Frequency changes among accessions with four favoured haplotypes in Chinese cultivars (CC), European cultivars (EC) and North American cultivars (NAC).

Relationship between ancestral species and hexaploid accessions

Fifteen diploid and 36 tetraploid wheat accessions were genotyped at AGP‐S1. Distinct differentiation occurred between T. urartu and T. boeoticum (T. monococcum). Polyploid wheats were closer to T. urartu, supporting the currently held opinion that T. urartu is the A genome donor to polyploid wheats (Feldman and Levy, 2015). When neglecting polymorphic sites that were less than 10% in frequency, most tetraploid accessions clustered into three haplotype groups (Figure S7). Hap‐1 and Hap‐2 had the same polymorphic sites as TaAGP‐S1‐7A‐Hap‐1 and ‐Hap‐2, but Hap‐3 was detected only in tetraploid wheat, indicating that it was lost during hexaploidization of common wheat.

Nucleotide diversity (π) dramatically declined with polyploidization (Figure 7a). The πD/πT was only 1.43, but πT/πH was 9.11 (Table 4). Pairwise differences between populations (F ST) showed that the F ST between cultivated and wild tetraploid wheats was only −0.036, but this value reached 0.436 between hexaploid landraces in the MCC and cultivated tetraploid wheat. This value was 0.154 when modern cultivars and landraces in the MCC were compared (Figure 7c). These results indicated that selection on AGP‐S1‐7A mainly occurred during and after hexaploidization.

Figure 7

Nucleotide polymorphisms (π) and genetic differentiation () between pairs of populations at and . (a) π values in diploid, tetraploid and hexaploid accessions at ; the line under the horizontal axis represents the coding region. (b) π values in diploid, tetraploid and hexaploid accessions at . (c) in DS, DM, LA and MC at , and blue colour gradient represents changes values from dark (1.0) to light (0.0). (b) in DS, DM, LA and MC at . DS, Triticum dicoccoides; DM, other tetraploid accessions used in the study; LA, landraces in the MCC; MC, Chinese modern cultivars in the MCC.

Table 4

Diversity and Tajima tests at the AGP‐S1‐7A and AGP‐L‐1B loci

	π	θ	Tajima's D	P
AGP‐S1‐7A
Diploid accessions	0.00234	0.00344	−1.72416	<0.05
Tetraploid accessions	0.00164	0.00338	−1.90879	<0.05
Hexaploid accessions	0.00018	0.00019	−0.11657	>0.10
AGP‐L‐1B
Aegilops speltoides	0.03122	0.02525	0.96839	>0.10
Tetraploid accessions	0.00817	0.00643	1.01117	>0.10
Hexaploid accessions	0.00113	0.00042	3.61558	<0.001

Diversity at AGP‐L‐1B was higher than that at AGP‐S1‐7A (Table 4). It also declined dramatically during polyploidization (πD/πT = 3.82, πT/πH = 7.23; Figure 7b). F ST between the cultivated and wild tetraploid wheat was only 0.001, yet this value reached 0.906 between hexaploid landraces and cultivated tetraploid wheat. However, this value was only 0.099 between modern cultivars and common wheat landraces (Figure 7d). These data indicated that strong selection on AGP‐L‐1B had occurred during hexaploidization, domestication and breeding of common wheat. Tajima's D index also supports this opinion (P < 0.001).

Discussion

Haplotype differences at TaAGP‐S1‐7A and TaAGP‐L‐1B are associated with variation in TKW

AGPases synthesize the starch substrate ADP‐glucose. AGPases are ancestral plastidial enzymes expressed in photosynthetic tissues, but are highly expressed in the developing endosperm of grasses due to duplication of the AGPase multigene family (Comparot‐Moss and Denyer, 2009). In this study, we focused on genes TaAGP‐S1‐7A and TaAGP‐L‐1B expressed in developing wheat endosperm. Four haplotypes of each gene were detected in modern cultivars. Some haplotypes at each locus showed similar expression patterns and no variation in TKW in the MCC and NILs (Figures S2, S3, S4). These haplotypes were pooled into single subgroups based on their similar phenotypic effects, and polymorphic sites were reduced to one at TaAGP‐S1‐7A and three at TaAGP‐L‐1B. These sites might be causative SNPs affecting TKW.

The SNP at position −122 of TaAGP‐L‐1B is associated with variation at the transcript level

The SNP at position −122 of TaAGP‐L‐1B‐Hap‐II is within an E2F‐DP‐binding motif (WTTSSCSS) and therefore may change TaAGP‐L‐1B into an E2F target gene. This motif is not present in the 1‐kb sequence upstream of ‐Hap‐I. E2F transcription factors can be inhibitors or activators of E2F target genes in Arabidopsis thaliana (Mariconti et al., 2002; Vandepoele et al., 2002). Proteins encoded by plant E2F target genes are involved in cell cycle regulation, DNA replication and repair, chromatin dynamics and many other functions (Vandepoele et al., 2005). In this study, seeds transformed with TaAGP‐L‐1B‐Hap‐II promoter‐GUS had lower GUS expression (Figure 3), suggesting this nonfavoured haplotype generated a lower transcript level due to the E2F motif.

The TaAGP‐S1‐7A and TaAGP‐L‐1B loci underwent strong selection in both domestication and breeding of common wheat

The extreme difference between the πD/πT and πT/πH ratios indicated that much stronger selection occurred in hexaploid than in tetraploid wheats. Of course, the bottleneck effect at hexaploidization of wheat cannot be ignored as formation of common wheat was a very rare event (Feldman and Levy, 2015). Evidence for selection on the two AGPase genes in breeding was strong. Firstly, the frequency of TaAGP‐S1‐7A‐Hap‐I was about 20% in landraces, but nearly 80% in modern cultivars in the MCC, and the frequency of TaAGP‐L‐1B‐Hap‐I was less than 50% in landraces, but 85% in recent modern cultivars. Secondly, the frequencies of cultivars carrying favoured haplotypes also steadily increased among varieties released after the 1940s (Figure 5).

These favoured haplotypes were also selected in five other wheat production regions (Figure S6). Compared with yield‐related genes TaTEF and TaCWI (Jiang et al., 2015; Zheng et al., 2014) TaAGP had higher favoured haplotype frequencies and selection intensities. All these data indicated that TaAGP were critical genes selected for larger grain size during wheat domestication and breeding.

Additive effects of major genes influencing grain weight in wheat

The TKWs of accessions combining the two favoured haplotypes were significantly higher than those of accessions carrying the nonfavoured haplotypes in both the MCC and MC (Figure S5), suggesting that genetically additive effects favoured selection at TaAGP‐S1 and TaAGP‐L. TaSus1 and TaSus2, another pair of key genes for enzymes involved in the starch synthesis pathway, showed a similar trend (Hou et al., 2014; Jiang et al., 2011). The four favoured haplotype cohorts showed highest TKW and positive selection in Chinese, European and North American cultivars (Figure 6). All of these examples indicate additive effects of major genes influencing TKW. Pyramiding these favoured haplotypes at multifunctional loci should be beneficial for future yield improvement in wheat.

Conclusion

TaAGP‐S1 and TaAGP‐L are important genes for enzymes involved in starch synthesis in developing endosperm. Haplotypes at TaAGP‐S1‐7A and TaAGP‐L‐1B are associated with TKW, and mean TKWs of favoured haplotypes are significantly higher than those of nonfavoured ones in different populations. Favoured haplotypes underwent strong positive selection during wheat domestication and breeding due to their additive genetic effects on TKW.

Experimental procedures

Plant materials

Two wheat populations were used: 245 accessions from the Chinese wheat Mini Core Collection (MCC) (Table S5) and 348 modern cultivars (MC) (Table S6) from the Chinese Wheat Core Collection (Hao et al., 2011). The MCC and MC were planted at the CAAS Luoyang Experiment Station in Henan province in 2002, 2005 and 2006, and CAAS Shunyi Experiment Station in Beijing in 2010.

Wheat NILs derived from an F5 population from cross Youzimai/Zhou 18*3//Handan 6172 were used for estimating SN, GN, and TKW and transcript level differences between TaAGP‐S1‐7A‐Hap‐1 and ‐Hap‐2. Another set of NILs from an F4 population from cross Yangmai 158/Zhou 18*3//Handan 6172 were used for measuring the same yield components and transcript level differences between TaAGP‐L‐1B‐Hap‐1 and ‐Hap‐3 in 2014. Two NILs derived from a BC3F5 population from Isengrain/5*Yanzhan 4110 and a BC3F6 population of Jianmai/6*Zhou 18 were used for detecting TKW differences between TaAGP‐L‐1B‐Hap‐1 and ‐Hap‐2, and TaAGP‐L‐1B‐Hap‐1 and ‐Hap‐4, respectively. The data for all NILs are listed in Table S2.

Three hundred and eighty‐four European, 429 North American, 53 International Maize and Wheat Improvement Center (CIMMYT), 82 Russian and 51 Australian cultivars were surveyed to investigate the global distribution of favoured haplotypes of the TaAGP‐S1‐7A and TaAGP‐L‐1B genes and frequency changes of the favoured haplotypes during the last 100 years. Release dates and origins of all materials are provided in Hou et al. (2014).

Thirty‐seven accessions of diploid wheat and 36 accessions of tetraploid wheat were sequenced for evolutionary studies of TaAGP‐S1‐7A and TaAGP‐L‐1B (Table S7).

Haplotype analysis

Primers were designed by Premier 5.0 (http://www.premierbiosoft.com/) and synthesized by BGI Tech (Shenzhen, China). PCR was conducted in reaction volumes of 15 mL containing 50–100 ng DNA, 7.5 mL buffer, 2.4 mL of 2.5 mm deoxynucleotide triphosphates, 0.5 mL of 10 mm forward and reverse primers, and 0.15 mL of LA Taq polymerase (TaKaRa Biotechnology). PCR was carried out in a Veriti 96‐Well Thermal Cycler (Applied Biosystems) with the following steps: denaturing at 95 °C for 5 min, followed by 32 cycles of 95 °C for 30 s, annealing at 55–63°C for 1 min, 72 °C for extension (1 kb/min) and a final extension of 72 °C for 10 min. PCR products were purified by purification kit DP‐1502 (TIANGEN), then cloned by the pGEM‐T Easy Cloning Vector (TIANGEN), and transformed into Escherichia coli cells by the heat shock method. Plasmids were extracted with a DP‐1002 Kit (TIANGEN). DNA was sequenced by an ABI 3730XI DNA Analyser (Applied Biosystems). Sequence alignments and SNP identification were made using DNASTAR (http://www.dnastar.com/). The digestion sites of restriction endonucleases were detected by Primer Premier 5.0. Two restriction endonucleases used in the study were produced by New England Biolabs (Beijing).

Transcript analysis

RNA was extracted from developing grains of two near‐isogenic wheat lines (described in Plant materials) at 5, 10, 15, 20, 25 days postanthesis (DPA) using an RNAprep pure plant kit DP432 (TIANGEN) and reverse transcribed with an M‐MLV reverse transcriptase kit (Invitrogen, Shanghai, China). Relative real‐time PCR was carried out by Rotor‐Gene Q (Qiagen). Primers are listed in Table S1, and data are provided in Table S3.

Promoter‐driven GUS expression in transgenic rice

About 1.4‐kb sequences upstream of the ATG start codon of two haplotypes at TaAGP‐L‐1B were amplified by primers L‐1B‐PEF and L‐1B‐PER (Table S1), and inserted into the pCAMBIA1391Z vector via SalI and EcoRI digestion. The construct was mobilized into Agrobacterium tumefaciens EHA105 and then transformed into rice (Oryza sativa L. ssp. japonica) cv. Kitaake as described by Hiei et al. (1994). Wild‐type and transformed rice seedlings were planted at the CAAS Langfang Experiment Station in Hebei province in 2016. Each haplotype was represented by three positive transgenic lines, and 10 plants of each line were tested for GUS activity. GUS activities in developing seeds at 21 DPA were measured following the protocol of Hänsch et al. (1995). For each sample, 200 mg of fresh seeds was ground into powder in liquid nitrogen, and then, 1 mL of extraction solution (0.02885 m Na2HPO4, 0.02115 m NaH2PO4, 0.01 m EDTA, 0.1% SDS, 0.1% β‐mercaptoethanol, and 0.1% Triton × 100) was added, and centrifuged at 12 000  for 10 min at 4 °C. Protein concentration in the supernatant was measured following the Bradford (1976) method. The supernatant was assayed for GUS activity with 4‐MUG (4‐methylumbelliferyl‐β‐D‐glucuronide) substrate as described by Hänsch et al. (1995). Fluorescence intensity levels were measured on a Tristar LB941 fluorescence spectrophotometer (Berthold) with 4‐MU (4‐methylumbelliferon) as the calibration control. GUS staining was performed following the protocol of Kosugi et al. (1990). The images of stained samples were captured using a Discovery V20 stereomicroscope (Carl Zeiss, Germany). All data are listed in Table S1.

Statistical analysis

Data processing was carried out using Excel 2010 and SPSS Statistics 17.0 (http://www-01.ibm.com/software/analytics/spss/).

Evolutionary studies

Diversity analysis and Tajima D tests of the two genes were carried out by DnaSP 5.10 (http://www.ub.es/dnasp). F ST tests at AGP‐S1‐7A were carried out by Arlequin 3.5.1.2 (http://cmpg.unibe.ch/software/arlequin3).

Conflict of interest

The authors declare that they have no conflict of interests.

Figure S1 Alignments of TaAGP‐S1‐7A, ‐7B, ‐7D, TaAGP‐S2‐5A, ‐5B, ‐5D, TaAGP‐L‐1A, ‐1B and ‐1D based on cDNA (a) and genomic DNA (b) sequences.

Figure S2 Mean TKWs of two genes in the MCC.

Figure S3 Differences between TaAGP‐S1‐7A‐Hap‐1 and ‐Hap‐2 in F5 NILs derived from Youzimai/Zhou 18*3//Handan 6172.

Figure S4 Differences between TaAGP‐L‐1B haplotypes.

Figure S5 Haplotype combinations in two populations.

Figure S6 Global distribution of haplotypes.

Figure S7 Dendrogram based on AGP‐S1‐7A sequences among diploid, tetraploid and hexaploid wheat accessions. Major polymorphic sites in tetraploid accessions are shown in the table below.

Click here for additional data file.

Table S1 Primers used in this study.

Table S2 Near‐isogenic lines derived from four combinations.

Table S3 Haplotype transcript measurements at five stages of developing seeds.

Table S4 GUS activity measurements in transgenic and wild type rice.

Table S5 Accessions of hexaploid wheats in the Mini Core Collection.

Table S6 Accessions of modern cultivars.

Table S7 Progenitor accessions used in this study.

Click here for additional data file.