Dissecting pleiotropic functions of the wheat Green Revolution gene Rht-B1b in plant morphogenesis and yield formation Free

Plant height has significant effects on lodging resistance and yield potentials in crops. The introduction of the reduced height (semi-dwarf) genes Rht-B1b and Rht-D1b in the 1960s was a crucial component of the Green Revolution (GR) in wheat (Triticum aestivum L.) (Hedden, 2003). Rht-B1b and Rht-D1 are homoeoloci on chromosome arms 4BS and 4DS, respectively, and encode truncated DELLA proteins homologous to the gibberellin-insensitive (GAI) protein in Arabidopsis (Chandler et al., 2002; Ikeda et al., 2001; Peng et al., 1997, 1999). DELLA proteins have been identified as key integrators of multiple regulatory pathways that modulate plant growth and development in model plants, such as Arabidopsis and rice (Davière and Achard, 2016; Nolan et al., 2020; Tong and Chu, 2018; Xu et al., 2014). DELLA proteins directly bind with GID1, BZR1, SPY, EIN3, JAZ1, ARF7 and D14 in the gibberellin (GA), brassinosteroid (BR), cytokinin (CTK), ethylene (ETH), jasmonate (JA), auxin (IAA) and strigolactone (SL) signaling pathways, respectively (Nolan et al., 2020; Tong and Chu, 2018), and also interact with GRF4 and NGR5, key factors for yield and nitrogen-use efficiency in rice, shaping a module to boost a sustainable GR (Li et al., 2018c; Wu et al., 2020). The GA signaling pathway has been studied extensively to explore the function of DELLA in plant growth and development. GID2 and SLY1 are F-box proteins, components of a ubiquitin E3 ligase complex in Arabidopsis and rice, respectively. The binding of GA with its receptor GID1 promotes the recruitment of DELLA to the SCF^SLY1/GID2 E3 ligase complex for polyubiquitylation and ubiquitinated DELLA is then degraded by the 26S proteasome (Murase et al., 2008; Shimada et al., 2008; Sun, 2011; Wu et al., 2011). However, the truncated proteins encoded by Rht-B1b and Rht-D1b cannot be degraded owing to lack of the DELLA domain that is responsible for the binding with GID1, but they can still repress the promotion effect of GA on plant height probably through binding to GA response factors, such as homologs of phytohormone-interacting factors (PIFs) and BZR1 (de Lucas et al., 2008; Feng et al., 2008; Pearce et al., 2011; Peng et al., 1999; Wu et al., 2011; Bai et al., 2012).

It is well known that GR genes promote yield stability by reducing plant height and concomitant reduced lodging, but how they improve yield potential remains elusive. Linkage mapping and association analyses showed that GR genes confer pleiotropic effects on yield component traits, including spike number per unit area (SN), grain number per spike (GNS) and thousand grain weight (TGW) (reviewed by Cao et al., 2020). Many near-isogenic lines (NILs) have been generated to investigate further the role of GR genes in improving yield (Allan, 1989; Börner et al., 1993; Borrell et al., 1991; Flintham et al., 1997; Keyes and Sorrells, 1989; Kowalczyk et al., 1997; Lanning et al., 2012; Miralles and Slafer, 1995; Velu et al., 2017; Wu et al., 2020). However, both genetic analyses and NIL comparisons produced inconsistent and even opposite results regarding the impacts of GR genes on yield component traits. It is almost a consensus that GR genes reduce TGW (Allan, 1989; Beharav et al., 1991; Börner et al., 1993; Cabral et al., 2018; Flintham et al., 1997; Guan et al., 2018; Keyes and Sorrells, 1989; Kowalczyk et al., 1997; Miralles and Slafer, 1995; Schulthess et al., 2017; Tian et al., 2017; Velu et al., 2017; Xu et al., 2019), but no difference in TGW caused by GR genes was detected in some genetic studies (Achilli et al., 2022; Li et al., 2018a). The majority of reports showed that GR genes increased SN or tiller number (Allan, 1989; Borrell et al., 1991; Keyes and Sorrells, 1989; Wu et al., 2020), although negative or minor effects on SN were observed in several studies (Guan et al., 2018; Li et al., 2018a; Tian et al., 2017). The effects of GR genes on GNS are highly controversial; some reports indicated enhanced GNS (Allan, 1989; Beharav et al., 1991; Börner et al., 1993; Borrell et al., 1991; Flintham et al., 1997; Kuchel et al., 2007; Lanning et al., 2012; Liu et al., 2014; Miralles and Slafer, 1995; Schulthess et al., 2017; Velu et al., 2017), whereas others showed that GR genes could reduce GNS (Keyes and Sorrells, 1989; Kowalczyk et al., 1997; Wu et al., 2020; Zhang et al., 2013).

Transgenic technology is one of the most widely used and reliable tools to investigate gene function. The function of GR genes in regulating plant height was initially confirmed by transformation in rice (Peng et al., 1999; Wu et al., 2011) and later confirmed in wheat (Li et al., 2012; Van De Velde et al., 2021). However, their specific roles in yield formation were not dissected. There is also no molecular evidence to address the reason why GR genes improve yield potential. Here, we pinpointed the function of GR genes in modulating plant architecture and yield traits using transgenic overexpression and knockout assays. The developmental causes for the resulting phenotypic changes were also explored by microscopy and dynamic investigations. We further dissected the underlying molecular regulatory networks through multi-omics analyses.

Rht-B1b and Rht-D1b have equivalent genetic effects on agronomic traits (Eshed and Lippman, 2019; Lanning et al., 2012; Zhang et al., 2013). To clarify the function of GR genes, we chose Rht-B1b to create overexpression (Rht-B1b-OE) and knockout (Rht-B1b-KO) lines. Three Rht-B1b-OE lines and two Rht-B1b-KO lines with frame-shift mutation were used for subsequent analyses (Fig. S1). As expected, plant height was significantly reduced in Rht-B1b-OE lines and increased in Rht-B1b-KO lines compared with transgenic null lines (TNLs) (Fig. 1A,C). Stem internode length was also significantly reduced in Rht-B1b-OE lines and increased in Rht-B1b-KO lines (Fig. 1D). The change in uppermost internodes (peduncles) caused by Rht-B1b was the greatest among all stem internodes (Fig. 1B,D). Flag leaves were significantly smaller in Rht-B1b-OE lines than those in Rht-B1b-KO lines and TNLs, but there was no significant difference in flag leaf size between TNLs and Rht-B1b-KO lines (Fig. 1B,E,F). Rht-B1b-OE lines generated more tillers per plant than the TNLs, and Rht-B1b-KO lines had fewer tillers (Fig. 1A,G). Accordingly, overexpression of Rht-B1b increased earbearing tiller number per plant, i.e. spike number per plant (Fig. 1H). Rht-B1b overexpression also significantly increased tiller angle (Fig. 1A,I). Spikes in Rht-B1b-KO lines were obviously more slender than those in Rht-B1b-OE lines and TNLs (Fig. 1B). Although there was little difference in spike length between Rht-B1b-OE lines and TNLs, spike length was significantly increased in Rht-B1b-KO lines compared with TNLs (Fig. 1J). Overall, Rht-B1b had clear effects on plant height, leaf size, tiller number, tiller angle and spike length, indicating its role in modulating plant architecture.

To ascertain how GR genes improve yield potential, we investigated the effects of Rht-B1b on yield-contributing traits SN, GNS and TGW. Rht-B1b-OE lines had significantly increased total tiller number per m² and SN (herein referring to spike number per m²) compared with TNLs, and both traits were reduced in Rht-B1b-KO lines (Fig. 2A,B). Nevertheless, the earbearing tillering rate was reduced in Rht-B1b-OE lines and enhanced in Rht-B1b-KO lines relative to that in TNLs (Fig. 2C). Thus, Rht-B1b improved SN by enhancing tillering capacity. GNS was significantly increased in Rht-B1b-OE lines and decreased in Rht-B1b-KO lines compared with TNLs (Fig. 2D). However, Rht-B1b-KO lines had significantly higher spikelet number per spike than TNLs, and Rht-B1b overexpression led to a very slight decrease in floret and spikelet number per spike, but this was not statistically significant (Fig. 2E,F). Obviously, Rht-B1b increased GNS by enhancing floret fertility. TGW was lower in Rht-B1b-OE lines and higher in Rht-B1b-KO lines than in TNLs (Fig. 2H). Considering that grain size is a major determinant of grain weight, we investigated grain length (GL) and grain width (GW) (Fig. 2G). GL was significantly increased in Rht-B1b-KO lines compared with TNLs (Fig. 2I). Likewise, grain width was less in Rht-B1b-OE lines and greater in Rht-B1b-KO lines (Fig. 2J). Thus, we conclude that Rht-B1b reduces TGW largely by modulation of grain size. Most importantly, grain yield was significantly enhanced in Rht-B1b-OE lines and reduced in Rht-B1b-KO lines compared with TNLs (Fig. 2K), indicating that Rht-B1b has an overall positive effect on yield-contributing traits. Enhanced SN and GNS were apparently sufficient to compensate for the reduction in grain weight.

Histochemical staining showed that cell length and width in the internodes were significantly reduced in the Rht-B1b-OE lines and increased in the Rht-B1b-KO lines compared with TNLs, suggesting that Rht-B1b functions as a repressor of cell growth and causes a concomitant reduction in plant height (Fig. 3A). Considering that Rht-B1b reduced grain size, we also investigated the cell size of the grain pericarp. Electron microscopy showed that Rht-B1b-OE and Rht-B1b-KO lines were not different from TNLs in this respect (Fig. 3B). As expected, Rht-B1b had little effect on cell number per area in the grain pericarp (Fig. 3B). Rht-B1b is a negative regulator of grain size (Fig. 2G,I,J). Further investigation showed that the enhanced expression of Rht-B1b reduced the area of grain pericarp and caused a concomitant decrease in cell number (Fig. 3B). Rht-B1b reduces plant height by diminishing cell size in the internodes, and reduces grain size by decreasing cell number in the pericarp, suggesting that it modulates plant height and grain size through different pathways.

Given the large effects of Rht-B1b on tiller number, we investigated tiller initiation and outgrowth at the seedling stage. Rht-B1b-OE lines had more tiller buds than the TNLs, and the Rht-B1b-KO lines had fewer (Fig. 3C). Conversely, enhanced expression of Rht-B1b repressed tiller outgrowth, accounting for a lower rate of earbearing tillers in Rht-B1b-OE lines (Fig. 2C). We also investigated developing spikes and observed that there was little difference in spikelet number among the Rht-B1b-OE and Rht-B1b-KO lines and TNLs at the early stage of spike development (Fig. 3D). However, there were smaller spikelets at the top of the spikes in Rht-B1b-OE lines owing to retarded growth at the late stage of spike development, explaining why Rht-B1b reduced spikelet and floret number per spike (Fig. 3D). The Rht-B1b-OE lines had fewer underdeveloped sterile florets than the TNLs and the Rht-B1b-KO lines (Fig. 3E). We also compared pollen fertility in the normal florets and observed little difference in pollen staining among the three groups (Fig. 3F). Thus, Rht-B1b enhances floret fertility by improving floral development.

Transcriptome analysis, which detects differentially expressed genes (DEGs) in key spatial-temporal windows, is an effective way to uncover regulatory pathways or networks underpinning genes of interest (Wang et al., 2009). To mine genes downstream of Rht-B1b, we compared transcriptome profiles between the Rht-B1b-OE lines and TNLs. Considering that plant height and grain yield are the traits most affected by the GR genes, we focused on dissection of the regulatory mechanisms controlling stem elongation, tillering, spike development and grain weight using RNA sequencing (RNA-Seq) assays.

As the jointing and heading stages (from Feekes 6 to 10.5) are the most determinant spatiotemporal windows for plant height, the transcriptome of stems at Feekes 7 (second internode visible) and uppermost internodes at Feekes 10.1 (spike visible) were used to mine key genes downstream of Rht-B1b. In total, 36/30 and 599/411 DEGs were upregulated or downregulated by Rht-B1b overexpression at Feekes 7 and Feekes 10.1, respectively (Tables S1 and S2). Very few common DEGs were detected between these two groups of RNA-Seq data, indicating strong spatiotemporal specificity of the regulatory mechanism underlying Rht-B1b modulation of stem development. Several DEGs were homologs of TCP14, OsMADS57, GID1c, OsPIL1, BRI1, OsYABBY4, HOX12, OsDWARF and OsDWARF10, each of which have been reported to regulate plant height (Fig. 4A; Tables S1 and S2). TCP14 in Arabidopsis increases internode length by promoting cell division in response to auxin (Ferrero et al., 2021; Kieffer et al., 2011) and its homologs were downregulated by Rht-B1b at Feekes 10.1 (Fig. 4A; Table S2). OsMADS57 promotes plant height by repressing GA 2-oxidase genes (Chu et al., 2019). Rht-B1b overexpression downregulated a MADS57-like gene at Feekes 10.1 (Fig. 4A; Table S2). HOX12 in rice negatively regulates plant height (Gao et al., 2016) and one of its homologs is upregulated by Rht-B1b (Fig. 4A; Table S2). Three GA 2-oxidase (GA2ox) genes were upregulated by Rht-B1b (Fig. 4A; Table S2). The GA-GID1 module induces DELLA degradation in model plants (Murase et al., 2008; Shimada et al., 2008). A homolog of GID1c was differentially expressed between Rht-B1b-OE lines and TNLs (Fig. 4A; Table S2). These results indicate that Rht-B1b could modulate plant height through a feedback regulatory loop of GA homeostasis in stem internodes. In addition, Rht-B1b overexpression downregulated a homolog of OsPIL1, a key regulator of internode elongation (Fig. 4A; Table S2) (Todaka et al., 2012). BRI1 functions as a receptor of BR to promote plant height (Montoya et al., 2002), and its homologs were downregulated by Rht-B1b overexpression (Fig. 4A; Table S2). OsYABBY4 is a negative regulator of plant height (Yang et al., 2016), but one of its homologs was downregulated by Rht-B1b overexpression at Feekes 7 (Fig. 4A; Table S2). Likewise, DWARF10 and OsDWARF are positive regulators of plant height (Arite et al., 2007; Hong et al., 2002), but their homologs were upregulated by Rht-B1b overexpression (Fig. 4A; Table S2). The contrasting expression patterns of such genes showed that they had been subjected to functional differentiation between wheat and rice. In addition to GA, other hormone-related DEGs included 9-cis-epoxycarotenoid dioxygenase (NCED) genes for ABA synthesis (Tan et al., 1997) and abscisic acid 8'-hydroxylase (CYP707A) genes for ABA degradation (Yang and Choi, 2006), ABA receptor PYL4, PP2C and ABF genes for ABA signaling (Choi et al., 2000; Hao et al., 2011), ethylene receptor OsERF3 (Wuriyanghan et al., 2009) and ACC oxidase genes (Qin et al., 2007) for ETH, SAUR genes (Zhou et al., 2013) for auxin, and OsCKX11-like genes (Zhang et al., 2021) for cytokinin (Fig. 4A; Table S2). A group of DEGs possibly involved in cell division and growth was also identified, including genes encoding CDC48c and expansins (Fig. 4A; Table S2) (Choi et al., 2003; Rancour et al., 2002). Based on the putative functions of DEGs, we propose a simplified regulatory model underpinning Rht-B1b reducing plant height (Fig. 4B).

A total of 661/214 DEGs in tillers at the tillering stage (Feekes 3) were upregulated or downregulated by Rht-B1b overexpression (Table S3). Some homologs of OsHOX1, OsHOX6, OsHOX12, OsHOX14, TB1, D2 and D14 were differentially expressed between Rht-B1b-OE lines and TNLs. OsHOX6 overexpression enhances tiller number in rice (Rahmawati et al., 2019) and its homologs were upregulated by Rht-B1b overexpression (Fig. 5A; Table S3). Enhanced expression of OsHOX14 results in a dwarf phenotype (Shao et al., 2018) and OsHOX1 has been identified as a positive regulator of tiller angle by affecting shoot gravitropism (Hu et al., 2020). Rht-B1b overexpression led to upregulation of the homologs of OsHOX1 and OsHOX14, suggesting that these genes might mediate the function of Rht-B1b in repression of tiller outgrowth and increase in tiller inclination (Fig. 5A; Table S3). As such, HOX family genes as downstream targets of Rht-B1b play key roles in wheat tillering. Although TaTB1 genes in chromosomes 4A, 4B and 4D repress tillering (Dixon et al., 2018) we detected no differential expression between Rht-B1b-OE lines and TNLs. By contrast, another three homologs of TB1 on homoeologous group 5 chromosomes, were upregulated by Rht-B1b overexpression, suggesting that these three TB1-like genes could be involved in regulation of tillering (Fig. 5A; Table S3). D14 in rice inhibits tillering and participates in the conversion of strigalactone to the bioactive form (Arite et al., 2009), but the wheat homologs of D14 were upregulated by Rht-B1b overexpression (Fig. 5A; Table S3). The mutant d2, which has a mutation in a BR biosynthetic enzyme gene encoding cytochrome P450, has been shown to reduce tiller angle in rice (Dong et al., 2016). A homolog of D2 was upregulated by Rht-B1b overexpression (Fig. 5A; Table S3). Rht-B1b overexpression upregulated some genes encoding GA 2-oxidases (GA2ox), GA 13-oxidases (GA13ox) and GA receptor GID1, showing that it could trigger self-regulation of GA metabolism and signal transduction in tillers (Fig. 5A; Table S3). We also detected other hormone-related DEGs, such as the genes encoding IAA15, IAA18 (Jain et al., 2006) and the ethylene-responsive transcription factor AIL5 (Nole-Wilson et al., 2005) (Fig. 5A; Table S3). Based on the putative key DEGs related to tillering, we present a diagram to illustrate the regulatory network of Rht-B1b responsible for modulation of tiller number and angle (Fig. 5B).

Transcriptome assays to uncover the molecular mechanism of DELLA in regulating spike development, especially floret fertility, were conducted using young spikes at the booting stage (Feekes 10.0, when spikes can be seen in the swollen section of the leaf sheath below flag leaves). A total of 229/150 DEGs were upregulated or downregulated in Rht-B1b-OE lines compared with TNLs (Table S4). Notably, a few DEGs were homologs of genes involved in floral development, such as CBP1, HEC1, MSP1 and TB1. CBP1 is a regulator of transcription initiation in central cell-mediated pollen tube guidance in Arabidopsis (Li et al., 2015). HEC1 is required for female reproductive tract development and fertility (Gremski et al., 2007). MSP1 is required to restrict the number of cells entering into male and female sporogenesis and initiates anther wall formation in rice (Nonomura et al., 2003). TB1 is involved in apical dominance and the formation of female inflorescences (Dixon et al., 2018; Studer et al., 2017). Some homologs of CBP1, HEC1, MSP1 and TB1 were upregulated by Rht-B1b overexpression (Fig. 6A, Table S4). A d10 mutant has been shown to exhibit enhanced branches and apical dominance in rice (Arite et al., 2007) and one of its homologs was downregulated by Rht-B1b overexpression (Fig. 6A). Rice HOX12 negatively regulates panicle exsertion (Gao et al., 2016). A homolog of HOX12 upregulated in Rht-B1b-OE lines likely represses spike elongation. GA signaling plays an important role in spikelet fertility in rice (Kwon and Paek, 2016). Rht-B1b overexpression led to upregulation of two GA 2-oxidase genes, suggesting that it could drive a feedback inhibition for GA homeostasis in young spikes (Fig. 6A; Table S4). In addition to GA-related DEGs, several DEGs encoded homologs of proteins involved in other hormone pathways, including ZmAO-2, ARF13, IAA18, SAUR32, SAUR36, SAUR72, ETR4, ACC oxidase 1, OsEIN2, ERF013, OsCKX3, OsABA8ox3 and AtAIB (Fig. 6A; Table S4). We propose a molecular regulatory model underlying Rht-B1b function in modulating spike growth and development (Fig. 6B).

Considering that grain filling started at 7 days after flowering (DAF) and peaked at 20 DAF, transcriptome analysis of developing grains at these stages were performed to dissect the molecular mechanism underlying the Rht-B1b-driven reduction in grain size and weight (Li et al., 2019; Liu et al., 2016). Totals of 1446/2047 and 965/657 DEGs were upregulated or downregulated by Rht-B1b overexpression in grains at 7 and 20 DAF, respectively (Tables S5 and S6). Remarkably, homologs of a few genes related to grain size and/or weight, such as BG1, OsSPL13/GLW7, ABI5, GW6, OsDWARF and OsDWF4 were differentially expressed between Rht-B1b-OE lines and TNLs (Fig. 7A; Tables S5 and S6). Among them, BG1, OsDWARF, OsSPL13 and OsDWF4 function as promoting factors of grain size in rice and some of their wheat homologs were downregulated by Rht-B1b overexpression in grains at 7 DAF (Fig. 7A; Table S5) (Li et al., 2018b; Liu et al., 2015; Mori et al., 2002; Si et al., 2016). GW6 functions as a positive regulator of grain size and weight in rice, and its homologs were repressed by Rht-B1b overexpression in grains at both 7 and 20 DAF (Shi et al., 2020). ABI5 has been reported to act as a negative effector of seed size in Arabidopsis and its homologs were upregulated by Rht-B1b overexpression in grains at both 7 and 20 DAF (Fig. 7A; Tables S5 and S6) (Cheng et al., 2014). DELLA, encoded by Rht-B1b, is well known as a repressor of the GA pathway. As expected, we detected numerous DEGs involving the GA response (Fig. 7A; Tables S5 and S6). A few homologs of the genes related to GA metabolism, such as GA 20-oxidase (GA20ox) and GA 3-oxidase (GA3ox) were differentially expressed between Rht-B1b-OE lines and TNLs (Fig. 7A; Tables S5 and S6) (Luo et al., 2006). DEGs related to ABA and ETH metabolism were also identified (Fig. 7A; Tables S5 and S6).

As shown in the microscopic investigation, Rht-B1b overexpression led to reduced grain size by decreasing cell number in the grain pericarp. Maize CKX2 negatively regulates cell division or number by catalyzing oxidative degradation of cytokinins (Bilyeu et al., 2001) and one of its homologs was upregulated by Rht-B1b overexpression in grains at 7 DAF (Fig. 7A; Table S5). By contrast, IPT genes promote the biosynthesis of trans-zeatin, a member of the cytokinin family (Miyawaki et al., 2006), and one plant-specific IPT5 gene was downregulated in grains at 7 DAF (Fig. 7A; Table S5). Some DEGs encoded cyclin-dependent kinases (CDK: CDKB1; 1 and CDKB2; 2) and Cyclin (CycB1; 1, CycB1; 5 and CycH1; 1) that participate cell proliferation (Fig. 7A; Tables S5 and S6) (Fabian et al., 2000). Many DEGs related to carbohydrate metabolism were also identified in developing grains (Fig. 7A; Tables S5 and S6). IDD14a represses starch synthesis through activation of QQS, a negative regulator of starch accumulation in Arabidopsis (Seo et al., 2011), but its homolog was downregulated by Rht-B1b overexpression (Fig. 7A; Table S6). Based on the above information, we present a molecular regulatory model underlying Rht-B1b modulation of grain size and weight (Fig. 7B).

Grain metabolites were analyzed by liquid chromatography with tandem mass spectrometry (LC-MS/MS) to decipher the underlying mechanism of Rht-B1b function in modulation of grain development. As in RNA-Seq assays, developing grains at 7 and 20 DAF were used for analysis. Only two metabolites, adenosine 5′-monophosphate (AMP) and guanosine-5′-monophosphate (GMP), had significantly different abundance between Rht-B1b-OE lines and TNLs in grains at 7 DAF, and were not involved in energy metabolism (Table S7). Trans-zeatin induces cell division and is formed from AMP and dimethylallyl pyrophosphate (Kakimoto, 2003). Some homologs of CKX1 and IPT5, key regulators of cytokinin metabolism, were detected in the transcriptome assays (Fig. 7A; Table S5). In the metabolome of grains at 20 DAF, 239/124 metabolites were upregulated or downregulated, respectively, in Rht-B1b-OE lines compared with TNLs (Table S8). Rht-B1b-OE lines had lower AMP abundance than did TNLs, showing that AMP was also a target metabolite of Rht-B1b in grains at 20 DAF (Table S8). Six DEGs involved in cytokinin metabolism or signal transduction were detected in the transcriptome assays (Fig. 7A; Table S6). GA homeostasis is tightly regulated by activation of GA 20-oxidases and GA 3-oxidases, and deactivation of GA 2-oxidases (Bao et al., 2020). GA7 is a bioactive GA isoform (Tudzynski et al., 2003; Tian et al., 2022) and its content was significantly increased in Rht-B1b-OE lines compared with the TNLs (Table S8), consistent with the Rht-B1b-upregulated GA20-oxidase and GA3-oxidase genes in grains according to RNA-Seq assays (Fig. 7A; Table S6). Starch is the major component contributing to wheat grain yield, accounting for approximately 70% of seed dry weight. ADP-glucose pyrophosphorylase (AGP) uses glucose 1-phosphate and ATP as substrates to produce ADP-glucose, which is the glucose donor for starch synthesis (Smidansky et al., 2002). Rht-B1b-OE lines had lower ADP-D-glucose (ADP-Glc) content than TNLs (Table S8). A few DEGs involving starch metabolism were also identified by transcriptome analyses (Fig. 7A; Table S6). Collectively, AMP, GA7 and ADP-Glc are likely key metabolites in regulating grain size and weight.

The introduction of Rht-B1b and Rht-D1b caused considerable increases in yield stability and potential and led to the wheat Green Revolution. However, the underlying mechanisms of GR genes in improving yield remained unknown. Here, we undertook a comprehensive investigation of the effects of Rht-B1b on agronomic traits and specified its function in modulating plant architecture and yield component traits. Rht-B1b significantly increases the number of effective tillers (fertile spikes) through improved tiller initiation. Several putative key genes for tillering capability function downstream of Rht-B1b (Fig. 5). We found that Rht-B1b overexpression increased GNS by enhancing floret fertility. A recent study showed that DELLA promotes floret fertility by interacting with OsMS188 in rice (Jin et al., 2022). We identified a group of DEGs involved in floret fertility in developing spikes between Rht-B1b-OE lines and TNLs by transcriptome assays. In terms of grain weight, Rht-B1b overexpression significantly decreased TGW, which was adverse to yield improvement. Thus, Rht-B1b enhances yield potential by increasing SN and GNS.

Although Rht-B1b promotes tiller bud initiation, it represses tiller outgrowth and consequently reduces ear-bearing tiller rate. Many studies have shown that nitrogen concentration has significant effects on tiller bud growth but does not influence tiller bud initiation (Huang et al., 2021; Lin et al., 2005). Appropriate increases in nitrogen fertilizer are necessary for the wheat varieties carrying GR genes to enhance earbearing tillers. We observed that Rht-B1b reduced grain size but had little effect on cell size in the pericarp, suggesting that Rht-B1b caused a decreased cell number and a concomitant reduction in grain size. By contrast, Rht-B1b decreased plant height by reducing cell size in stem internodes. Transcriptome analyses showed a large difference in the regulatory pathways underpinning the effect of Rht-B1b on plant height and grain size (Figs 4 and 7). Thus, Rht-B1b modulates plant height and grain size or weight through different pathways, providing useful cues to overcome the adverse effect of Rht-B1b on grain size and weight by modifying the downstream targets.

We probed the molecular mechanisms underlying pleiotropic effects of Rht-B1b on agronomic traits. Rht-B1b regulated many homologs of key genes that have been functionally validated to control target traits (Figs 4–7). Numerous genes modulating different traits downstream of Rht-B1b were also identified. Most of the previously identified key genes have been not reported to work downstream of DELLA-encoding genes, so the resultant transcriptome information is also very helpful to broaden the regulatory networks underlying plant architecture and yield formation in other plants. Homologs of OsHOX12 were regulated by Rht-B1b in tillers, stem internodes and young spikes, suggesting that they acted as integrators to orchestrate the growth and development of these organs (Figs 4 and 5) (Hu et al., 2020; Shao et al., 2018). Homologs of OsD14, a key regulator of plant architecture and yield (Arite et al., 2009; Zhou et al., 2013), were differentially expressed between Rht-B1b-OE lines and TNLs in tillers and developing grains (Figs 5 and 7). TaD14-4D, reported to be involved in strigolactone signaling, is associated with yield-contributing traits in wheat (Liu et al., 2021). OsDWARF, encoding a brassinosteroid biosynthetic enzyme, promotes elongation of leaves and stems in rice (Hong et al., 2002) and its homologs were differentially expressed between Rht-B1b-OE lines and TNLs in stem internodes and developing grains (Figs 4 and 7). GR genes encode truncated DELLA proteins, which still function in GA signal transduction but avoid GA-induced degradation because of an inability to interact with the GA receptor GID1 (Murase et al., 2008; Shimada et al., 2008). We observed several DEGs between Rht-B1b-OE lines and TNLs encoding GA-responsive proteins and GA metabolism enzymes in all investigated organs (stem internodes, tillers, young spikes and developing grains). These results suggest that Rht-B1b-modulated spatiotemporal GA homeostasis and signal transduction is a major cause of the resulting phenotypic variation. The above DEGs downstream of Rht-B1b were detected in multiple tissues, so they probably function in both plant morphogenesis and yield formation. As such, the DEGs homologous to OsHOX12, OsD14 and OsDWARF as well as the genes involved in GA metabolism and GA signal transduction are likely more important regulators of plant architecture and yield component traits than the others. Overall, the findings provide informative gene resources for synergistic improvement of wheat yield component traits.

Fielder, a wheat cultivar with Rht-B1b (a dwarfing allele at the RHT-B1) and Rht-D1a (wild-type allele at RHT-D1), was used as a transgenic recipient. The complete coding sequence (CDS) of Rht-B1b (GenBank: MG681100.1) was cloned into the entry vector pDONR207 and then transferred into the destination vector pUbiGW by the Gateway cloning method (Xie et al., 2021). The resultant construct was transformed into immature embryos to generate overexpression lines of Rht-B1b by the Agrobacterium tumefaciens (strain EHA105)-mediated method (Ishida et al., 2015). The Rht-B1b-OE lines were identified by genomic PCR. The expression levels of target genes in Rht-B1b-OE lines were also validated by qPCR assays using the primers and the procedure reported by Xu et al. (2019). CRISPR/SpCas9 was used to create knockout lines of Rht-B1b. The single guide RNA (sgRNA) (PAM-guide sequence 5′-GGAGCCGTTCATGCTGCAG-3′) was designed to target conserved regions of Rht-B1b and cloned into the vector pWMBX110-SpCas9 (Liu et al., 2020). Genetic transformation was performed by infecting immature embryos of Fielder with Agrobacterium tumefaciens strain EHA105 carrying the destination constructs. The resultant mutant lines were identified by genomic and sequencing. Considering that Rht-A1 and Rht-D1 have highly similar sequences to Rht-B1, their corresponding regions with the sgRNA-targeting site of Rht-B1 were also isolated and sequenced as above to determine whether there were off-target events (Fig. S1A). Primers for vector construction and transgenic plant identification are listed in Table S9.

Rht-B1b overexpression and knockout lines were grown in the field at Changping district in Beijing (40°2′N, 116°2′W) during the 2021 crop season and Jinan in Shandong province during the 2022 crop season. They were sown in 1.5-m rows with three replications in February and were harvested in June.

Plant height was measured from the ground to spike (awns excluded) at the grain fill stage. Tiller and spike number per plant, SN (spike number per m²), spike length (awns excluded), floret and spikelet number per spike, and flag leaf size were also investigated. Tiller number per plant was counted at the heading stage. Tiller angle was defined following the method described by Zhao et al. (2020). GNS was deduced from measures on 30 spikes. TGW, grain length and grain width were scored with a Wanshen SC-G seed detector (Hangzhou Wanshen Detection Technology Co.). Analysis of variance (ANOVA) was performed using SAS 9.2 software.

Considering that the uppermost internodes is the most important determinant for the difference in plant height between among Rht-B1b-OE and Rht-B1b-OE lines and TNLs (Fig. 1D), we used the internodes to image their cells in order to uncover the micromorphological cause. The uppermost internodes of the Rht-B1b-OE and Rht-B1b-KO lines and TNLs at the heading stage were fixed in FAA solution (70% alcohol, 5% acetic acid and 0.02% formaldehyde), embedded in paraffin, longitudinally sectioned and stained with Saffron-solid Green dye (G1053, Servicebio). The stem internode cells were imaged with a digital Panoramic MIDI scanner (3DHISTECH), and cell length and width were scored using ImageJ (https://imagej.nih.gov/ij/index.html). The area of grain coats (harvested at 20 days after flowering) was scored with a Wanshen SC-G seed detector (Hangzhou Wanshen Detection Technology Co.). Developing florets were photographed using a stereo microscope M205C (Leica Microsystems). Pollen grains were stained using 1% iodine-potassium iodide (I2-KI) solution and imaged using an SMZ800N Nikon stereomicroscope. Fresh grains of the Rht-B1b-OE and Rht-B1b-KO lines and TNLs at 14 DAF were harvested and the grain pericarps were imaged with an S-3400 scanning electron microscope (Hitachi). Cell length was measured in a minimum of 100 cells per sample using ImageJ. Three biological replicates per sample were used for each experiment.

The Rht-B1b-OE lines and TNLs were used for RNA-Seq experiments. Total RNA was extracted from tiller buds at tillering, stem internodes at Feekes 7, uppermost internodes at Feekes 10.1, spikes at Feekes 10.0, and grains at 7 and 20 DAF. The mRNA was purified from total RNA using poly-T oligo-attached magnetic beads (S1419S, NEB). RNA extraction and library preparation were conducted by Novogene (http://www.novogene.com/).

Raw reads were first filtered through Trimmomatic 0.38 to get clean reads. Reference genome and gene model annotation files were downloaded from the EnsemblPlants database (https://plants.ensembl.org/Triticum_aestivum/Info/Index). Clean reads were aligned to the reference genome using HISAT2 (v2.0.5) and read counts were calculated using FeatureCounts v1.5.0. Differential expression analyses of contrasting lines were performed using the DESeq2 R package (1.16.1). Genes were assigned as DEGs with an adjusted P-value <0.05 and |log₂(FoldChange)|≥1. Heatmaps were produced to display expression patterns of DEGs using TBtools (Chen et al., 2020).

Grains of the Rht-B1b-OE lines and TNLs were collected at 7 and 20 DAF for metabolome analyses. Three grains of each line were bulked for metabolite extraction. Three biological replicates per sample were prepared. The grains were frozen in liquid nitrogen and ground to fine powder. The homogenate was resuspended with prechilled 500 μl 80% methanol and 0.1% formic acid by vortexing. Samples were incubated on ice for 5 min and then centrifuged at 15,000 rpm (20,000 g) for 10 min at 4°C. The supernatants were diluted to a final concentration containing 53% methanol using ultrapure water. The samples were subsequently transferred to fresh tubes and centrifuged at 15,000 g for 20 min at 4°C. Finally, filtrates were injected into the Q Exactive HF-X LC-MS/MS system (Thermo Fisher Scientific). Grains from 7 and 20 DAF plants were used for central carbon metabolism and non-targeted metabolomics assays, respectively (Dunn et al., 2011; Want et al., 2010). Data files generated by the LC-MS/MS system were processed using the SCIEX OS (https://sciex.com/) to integrate and correct the peaks. The KEGG database was used to determine the most important biochemical metabolic and signal transduction pathways (Kanehisa and Goto, 2000).

We are grateful to Professor Robert McIntosh, Plant Breeding Institute, University of Sydney, for revising this manuscript.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201601.reviewer-comments.pdf