HvPHO1 Controls a Glucose 1‐Phosphate–SnRK1 Signalling Module That Coordinates Starch Accumulation in Barley Endosperm Authors: Yulong Li, Yujia Chen, Jing Liu, Ping Yang, Guangqi Gao, Carlos Guzmán, Qiang Xu, Yazhou Zhang, Yanlin Liu, Pengfei Qi, Mei Deng, Shoufen Dai, Jian Ma, Guoyue Chen, Jirui Wang, Youliang Zheng, Yuming Wei, Qiantao Jiang Published: 2026 ABSTRACT ABSTRACTDeveloping barley grains must balance the use of imported sucrose between soluble sugars and storage starch, but how this balance is controlled remains unclear. Here we show that loss of the plastidial α‐glucan phosphorylase HvPHO1 causes starch and soluble sugars to accumulate concurrently in barley grains. Developing pho1 grains showed sustained increases in glucose 1‐phosphate (G1P) and adenosine diphosphate glucose (ADP‐glucose), and exogenous G1P also promoted starch accumulation, supporting a stimulatory effect of elevated G1P on starch synthesis. G1P directly bound sucrose non‐fermenting 1‐related protein kinase 1 (HvSnRK1), suppressing its kinase activity and its phosphorylation of sucrose synthase 2 (HvSUSY2); the metabolic phenotypes of snrk1 and susy2 mutants supported a role for this phosphorylation module in sucrose utilization and soluble‐sugar homeostasis. HvPHO1 also interacted with the starch‐granule initiation factor starch synthase 4 (HvSS4). Loss of HvPHO1 increased the abundance of B‐type starch granules and shifted their size distribution, indicating a parallel role in storage‐starch organization. pho1 lines showed greater grain dry weight and test weight. Thus, loss of HvPHO1 initiates G1P‐mediated metabolic feedback that reconfigures the balance between sucrose utilization and starch accumulation and is associated with enhanced grain carbon deposition and grain weight. These findings reveal G1P‐dependent feedback linking starch‐precursor status to SnRK1‐mediated carbon allocation in barley endosperm. Introduction Sucrose imported from source tissues provides the main carbon input for developing cereal grains and must be partitioned between soluble carbohydrate pools and storage starch (Emes et al. 2003; Tetlow and Emes 2017). The amount and composition of starch formed during this process are major determinants of grain yield and end‐use quality (Fox and Bettenhausen 2023; De Schepper and Courtin 2024). In barley, starch is the major reserve compound of the mature grain and is produced from sucrose‐derived carbon during endosperm development. Its accumulation requires coordinated sucrose cleavage, sugar‐phosphate interconversion, and ADP‐glucose production, but how these steps are controlled remains unclear. In cereal endosperm, a substantial proportion of ADP‐glucose can be synthesized outside the amyloplast and then imported into plastids for starch synthesis, making starch accumulation sensitive to upstream carbon allocation (Beckles et al. 2001; Denyer et al. 1996; Rösti et al. 2006; Patron et al. 2004; Bowsher et al. 2007). Studies in cereals show that carbon flows dynamically between soluble sugars and starch precursors during grain development, highlighting the importance of regulatory points that control this allocation (Zhong et al. 2024; Jiang et al. 2024). These findings suggest that barley starch accumulation is not determined only by the capacity of starch biosynthetic enzymes (Tetlow et al. 2004; Hennen‐Bierwagen et al. 2009), but also by upstream control of sucrose‐derived carbon allocation. Glucose 1‐phosphate (G1P) is a key intermediate linking sucrose‐derived sugar phosphate metabolism to ADP‐glucose production for starch synthesis. In developing cereal endosperm, carbon imported as sucrose is converted through hexose phosphates and nucleotide sugars before entering the starch biosynthetic pathway (Zeeman et al. 2010; Smith and Zeeman 2020). G1P can be converted to ADP‐glucose by ADP‐glucose pyrophosphorylase, providing the immediate glucosyl donor used by starch synthases (Emes et al. 2003; Tetlow and Emes 2017). G1P is also used by plastidial α‐glucan phosphorylase PHO1 [also referred to as PHS1 in cereal literature (Kamble et al. 2023)], which participates in glucan metabolism in developing barley endosperm (Cuesta‐Seijo et al. 2017; Shoaib et al. 2021; Seung et al. 2025). This places G1P at a point where precursor supply and plastidial G1P utilization intersect, and PHO1 function is critical to maintain G1P homeostasis rather than allow it to accumulate excessively. Under a simple biosynthetic model, loss of a G1P‐using phosphorylase would be expected to limit starch production. An alternative possibility is that PHO1 helps maintain G1P balance, preventing this starch precursor from gaining a regulatory role in developing endosperm. Distinguishing between these possibilities is important for understanding how barley grains control sucrose‐derived carbon allocation into storage starch. SnRK1 is a conserved protein kinase that links sugar and energy status with plant growth and carbon use (Baena‐González et al. 2007; Peixoto and Baena‐González 2022; Nukarinen et al. 2016). In plants, inhibition of SnRK1 by trehalose 6‐phosphate (Zhang et al. 2009; Nunes et al. 2013; Blanford et al. 2024) has provided a clear example of how a sugar phosphate can act as a signal of carbon availability (Lunn et al. 2014; Figueroa and Lunn 2016; Baena‐González and Lunn 2020; Martínez‐Barajas et al. 2011; Nuccio et al. 2015). Whether a starch precursor can also signal through SnRK1 is less clear. G1P is a strong candidate for such a role because it is both a precursor for ADP‐glucose production and a substrate used by PHO1. If G1P accumulates when plastidial utilization is impaired, it could provide a metabolic cue that connects starch precursor status with sink carbon allocation. Sucrose synthase provides a plausible output for this signal because it links imported sucrose to UDP‐glucose, fructose and downstream sugar phosphate pools (Jiang et al. 2011; Sturm and Tang 1999; Koch 2004; Ruan 2014; Stein and Granot 2019). Its activity and stability can also be regulated by phosphorylation, raising the possibility that a SnRK1–SUSY2 branch transmits changes in G1P status to sucrose utilization during grain filling (Hardin et al. 2003; Hu et al. 2022; Yang et al. 2024). In addition to its potential role in signalling through G1P, PHO1 may also contribute to starch granule formation. Plastidial phosphorylases have been shown to participate in the initiation of B‐type starch granules in cereal endosperm (Kamble et al. 2023). Although PHO1 has been shown to interact with granule initiation factors in other cereals such as wheat, its role in barley granule architecture remains to be determined (Kamble et al. 2023; Hawkins et al. 2021). These findings suggest that PHO1 might influence starch structure in parallel with its role in controlling carbon allocation, providing a secondary layer of regulation that shapes starch granule architecture during grain development. Here, we used CRISPR–Cas9‐edited pho1 material to examine how HvPHO1 affects carbohydrate accumulation, starch‐precursor metabolism, and starch‐granule organization in barley endosperm. We first characterized the metabolic effects of HvPHO1 loss. We then used biochemical and genetic approaches to test whether G1P regulates HvSnRK1 and HvSnRK1‐dependent phosphorylation of HvSUSY2. Finally, we analysed starch‐granule architecture and tested the association between HvPHO1 and HvSS4. Result Loss of HvPHO1 Increases Grain Carbohydrate Deposition and Grain Filling To define the role of HvPHO1 in barley grain filling and yield formation, we generated three homozygous T3 single plant derived pho1 segregant lines (L2‐4, L2‐6 and L2‐9), all carrying the same CRISPR/Cas9‐induced 5‐bp deletion. This shared frameshift was predicted to truncate HvPHO1 before the catalytic phosphorylase domain; no sequence variants were detected at the candidate off‐target sites that were examined (Figure S1). HvPHO1 transcripts were most abundant early in grain development, and the HvPHO1 fusion protein localized to plastids (Figures S2 and S3). Plant height, spike length, productive spike number and grain‐shape parameters did not differ significantly between the pho1 lines and the wild type (Figure 1a,b and Figure S4). By contrast, thousand‐grain weight and test weight were significantly higher in pho1 (Figure 1c,d). Loss of HvPHO1 therefore increased grain weight without detectably altering major architectural traits. Loss of HvPHO1 increases grain weight and carbohydrate deposition. (a) Representative field and greenhouse phenotypes of Golden Promise (GP), the non‐edited control (NE) and pho1 lines. (b) Representative mature‐grain morphology. (c, d) Thousand‐grain weight (c) and grain test weight (d) of GP and pho1 lines. (e) Periodic acid–Schiff (PAS) staining of developing caryopsis sections, showing polysaccharide deposition in the central endosperm. SG, starch granule. (f–i) Total starch (f), sucrose (g), glucose (h) and fructose (i) contents in mature grains. (j–m) Activities of ADP‐glucose pyrophosphorylase (AGPase; j), granule‐bound starch synthase (GBSS; k), soluble starch synthase (SSS; l) and starch‐branching enzyme (SBE; m) in developing grains. (n) Starch‐granule‐bound proteins extracted from purified mature‐grain starch. (o) Immunoblot analysis of major starch‐biosynthetic proteins, with actin as the loading control. Data are mean ± s.d.; ns, not significant; *p < 0.05; **p < 0.01. Scale bars are indicated in the panels. PAS staining revealed a stronger polysaccharide signal in the central endosperm of pho1 grains (Figure 1e and Figure S5). Quantitative analysis showed that mature grains of all three pho1 lines contained significantly more total starch than the wild type, representing an increase of approximately 5–6 percentage points on a dry‐weight basis (Figure 1f,g and Figure S6). Glucose, fructose and sucrose were also significantly elevated in mature pho1 grains (Figure 1h,i and Figure S6). The additional starch therefore did not arise at the expense of soluble sugars; the major soluble sugars increased in parallel, indicating that loss of HvPHO1 altered the overall carbohydrate‐deposition state of the grain. To determine whether altered HvPHO1 dosage generated a reciprocal metabolic response, we examined stable transgenic lines expressing full‐length HvPHO1 from its native promoter (Figure S7). Despite a substantial increase in HvPHO1 transcript abundance, total starch, glucose, fructose and sucrose in mature grain did not differ significantly from the wild type (Figure S7). Thus, within the native‐promoter developmental context, increased HvPHO1 transcript abundance was insufficient to alter the major carbohydrate pools of mature grain; loss and increased expression of HvPHO1 did not produce a symmetric dosage response. We next asked whether the greater starch accumulation in pho1 resulted from broad enhancement of the canonical starch‐synthetic machinery. Activities of major starch‐metabolic enzymes were measured, starch‐granule‐bound proteins were profiled, and the abundance of selected starch‐biosynthetic proteins was examined by immunoblotting. AGPase, GBSS and SBE activities were not broadly increased in pho1 (Figure 1j–m), and neither the major band composition nor the overall abundance of granule‐bound proteins showed a systematic increase (Figure 1n and Figure S8). Immunoblot analysis of independent 15‐DAP endosperm samples showed reduced abundance of AGPS1, GBSS1, SS2a and SBE2b in pho1, whereas AGPL1 and SBE2a showed no consistent reduction (Figure 1o). Thus, the higher starch content of pho1 occurred despite reduced abundance of several starch‐biosynthetic proteins and was not explained by global activation of the starch‐synthetic machinery. This pattern is consistent with altered upstream carbon‐metabolite and starch‐precursor pools. Loss of PHO1 Expands G1P and ADP‐Glucose Pools in Developing Grains Because the increase in pho1 starch was not accompanied by broad enhancement of starch‐synthetic enzymes or granule‐bound proteins, we examined upstream precursor metabolism. Major sugar phosphates and nucleotide sugars were quantified across grain filling. Relative to the wild type, G1P remained elevated in pho1 at multiple developmental stages and was accompanied by increased ADP‐glucose, whereas UDP‐glucose was generally lower. G6P increased only at selected stages and to a lesser and less persistent extent than G1P (Figure 2a–d and Figure S9). As the direct glucosyl donor for starch synthesis, elevated ADP‐glucose was consistent with greater starch accumulation in pho1. The sustained increase in G1P was the most pronounced upstream metabolic feature of HvPHO1 loss. Loss of HvPHO1 expands glucose 1‐phosphate and ADP‐glucose pools and alters sucrose‐related metabolism. (a–d) Contents of glucose 1‐phosphate (G1P), adenosine diphosphate glucose (ADP‐glucose), uridine diphosphate glucose (UDP‐glucose) and glucose 6‐phosphate (G6P) in GP and pho1 grains. (e) G1P‐dependent glucan‐extension activity of purified HvPHO1. (f) G1P consumption and inorganic phosphate (Pi) release in complete reactions containing HvPHO1, G1P and a glucan acceptor and in the indicated controls. (g) Gene metabolite association network and correlation matrix linking starch‐biosynthetic genes with soluble sugars, sugar phosphates and nucleotide sugars. (h) Integrated representation of sucrose unloading, cytosolic sugar metabolism and amyloplast starch biosynthesis based on transcriptomic and metabolomic differences between GP and Hvpho1 grains. Heat blocks indicate relative transcript or metabolite abundance in each genotype. Data are mean ± s.d.; **p < 0.01. To test whether G1P accumulation in pho1 reflected restricted HvPHO1‐dependent utilization, we examined the G1P‐dependent activity of purified HvPHO1. In reactions containing G1P and a glucan acceptor, HvPHO1 catalysed glucan extension in the synthetic direction, and the reaction rate increased with G1P concentration (Figure 2e). The complete reactions consumed more G1P and released more Pi than the corresponding controls (Figure 2f). HvPHO1 can therefore use G1P for glucan extension in vitro. Together with persistent G1P accumulation in pho1 grain, these data support the conclusion that loss of HvPHO1 removes one route of G1P utilization and thereby contributes to the expanded G1P pool. The increases in G1P and ADP‐glucose identified a larger steady‐state starch‐precursor pool associated with greater starch accumulation, but these concentration measurements do not by themselves establish increased carbon flux into starch. Lower UDP‐glucose coincided with higher sucrose in pho1, suggesting that sucrose cleavage was altered. To identify regulators that could connect G1P accumulation to this sucrose‐metabolic shift, we integrated transcriptome–metabolome associations with phosphoproteomic screening. HvSnRK1 and HvSUSY2 emerged as candidates associated with variation in G1P and sugar metabolism. The phosphoproteomic screen further detected reduced HvSUSY2 phosphorylation in pho1 grain, which also had lower total SUS activity (Figure 2g,h and Figures S10–S14). We therefore selected HvSnRK1 and HvSUSY2 for direct testing of whether G1P participates in HvSnRK1–HvSUSY2‐dependent control of sucrose metabolism. G1P Directly Binds SnRK1 and Inhibits Its Kinase Activity HvSnRK1 emerged as a candidate link between G1P accumulation and altered sugar metabolism. We first tested whether G1P acted directly on HvSnRK1 by measuring binding and kinase inhibition. Isothermal titration calorimetry indicated direct binding of G1P to purified HvSnRK1, with an apparent dissociation constant (Kd) of approximately 12.18 μM under the assay conditions (Figure 3a). In vitro kinase assays further showed concentration‐dependent inhibition of HvSnRK1 by G1P, with a half‐maximal inhibitory concentration (IC₅₀) of 2.62 mM (Figure 3b). The ITC‐derived Kd describes the binding interaction, whereas the IC₅₀ reflects the G1P concentration required for kinase inhibition and should not be interpreted as a binding‐affinity parameter. Together, these results indicate that G1P binds HvSnRK1 and inhibits its kinase activity. G1P binds HvSnRK1 and suppresses its kinase activity. (a) Isothermal titration calorimetry analysis of G1P binding to purified HvSnRK1. (b) Concentration‐dependent inhibition of HvSnRK1 kinase activity by G1P. (c) Phos‐tag immunoblot analysis of HvSnRK1 phosphorylation in GP and pho1 grain protein extracts after addition of the indicated G1P concentrations; conventional immunoblotting shows total HvSnRK1 abundance. (d) Predicted G1P‐binding region in HvSnRK1, with candidate contact residues shown. (e) Energy‐component analysis of the simulated HvSnRK1–G1P complex. (f) Residue level binding‐energy decomposition. (g) Gibbs free‐energy landscape of the HvSnRK1–G1P complex during molecular dynamics simulation. (h) Residue root‐mean‐square fluctuation profile of HvSnRK1 during simulation. (i) Iodine staining of barley leaves treated with buffer control, glucose, Pi, glucose plus Pi or 5 mM G6P and G1P. (j) Starch content after the indicated treatments. (k) HvSnRK1 transcript abundance during exogenous G1P treatment. Data are mean ± s.d.; *p < 0.05; **p < 0.01. Phosphatase‐validated Phos‐tag analysis showed that addition of G1P to wild‐type grain protein extracts progressively weakened the phosphorylated HvSnRK1 band as G1P concentration increased. In pho1 grain extracts, the phosphorylated HvSnRK1 signal became undetectable after G1P addition, while total HvSnRK1 abundance also declined with increasing G1P (Figure 3c). These extract‐based Phos‐tag data are interpreted as physiological evidence that HvSnRK1 phosphorylation status is altered in the pho1 background, rather than as independent evidence for direct G1P binding; direct binding and inhibition are supported by the purified‐protein assays. To investigate the structural basis of G1P recognition, we performed molecular docking and molecular‐dynamics simulations. The calculations predicted a candidate G1P‐binding region in HvSnRK1, with R61, V161 and Q483 making comparatively large predicted contributions to complex stability (Figure 3d–f). During the 100‐ns simulations, G1P remained within the predicted region and maintained stable contacts with the candidate residues (Figure 3g,h and Figure S15). Alanine substitution of candidate residues attenuated G1P‐mediated kinase inhibition to different degrees, with the strongest effects observed for R61A, V161A and Q483A (Figure S16). Because these variants retained basal kinase activity, the data implicate R61, V161 and Q483 in the functional response of HvSnRK1 to G1P. We next tested whether elevated G1P was sufficient to alter carbohydrate accumulation in plant tissue. The cut bases of dark‐pretreated detached leaves were immersed in treatment solution so that G1P was taken up through the transpiration stream rather than applied to the leaf surface. Exogenous G1P increased intracellular G1P and starch accumulation in barley leaves, with the strongest response at 5 mM G1P (Figure 3i,j and Figure S17). In the parallel feeding series, exogenous G6P also promoted starch accumulation relative to the control, but the effect was weaker than that of G1P. Neither equimolar glucose nor Pi, alone or in combination, reproduced the G1P response, arguing against simple provision of glucose or phosphate as its cause. HvSnRK1 transcript abundance progressively declined with treatment time (Figure 3k), showing that elevated G1P was also accompanied by a change in HvSnRK1 expression. These feeding data show that elevated G1P promotes starch accumulation, but they do not establish that this response is specific to G1P because G1P and G6P are metabolically interconvertible. G1P Weakens SnRK1‐Mediated Phosphorylation of SUSY2 Having established that G1P binds and inhibits HvSnRK1, we asked how this effect reaches sucrose metabolism. HvSUSY2 was prioritized as a candidate downstream component because it associated with G1P and sucrose‐metabolic variation, and its phosphorylation signal was reduced in pho1 grain. HvSUSY2 catalyses the reversible reaction between sucrose and UDP; in the cleavage direction it produces UDP‐glucose and fructose and provides a major entry route for sucrose‐derived carbon into grain metabolism (Jiang et al. 2011; Sturm and Tang 1999; Koch 2004; Ruan 2014). Yeast two‐hybrid assays detected an interaction between HvSnRK1 and HvSUSY2. This interaction was independently supported in plant cells by co‐immunoprecipitation and bimolecular fluorescence complementation (Figure 4a–c). G1P suppresses HvSnRK1‐dependent phosphorylation of HvSUSY2, with S814 evaluated as a candidate site. (a) Yeast two‐hybrid assay of the HvSnRK1–HvSUSY2 interaction. AD–Large T plus BD–p53 and AD–Large T plus BD–Lamin were used as positive and negative controls, respectively. (b) Co‐immunoprecipitation of HvSUSY2–Myc with HvSnRK1–FLAG in Nicotiana benthamiana leaves. (c) Bimolecular fluorescence complementation analysis of the HvSnRK1–HvSUSY2 interaction in tobacco epidermal cells. (d) Phos‐tag analysis of HvSUSY2 phosphorylation by HvSnRK1 in vitro. (e) Concentration‐dependent effects of G1P on HvSnRK1 autophosphorylation and HvSnRK1‐dependent phosphorylation of HvSUSY2. (f) Kinase assays using kinase‐inactive HvSnRK1K49M and phospho‐null HvSUSY2S814A variants. (g–j) SUS activity and sucrose, glucose and fructose contents in barley mesophyll protoplasts coexpressing HvSnRK1 with HvSUSY2WT, HvSUSY2S814A or phosphomimetic HvSUSY2S814D, with or without 5 mM G1P. Data are mean ± s.d.; ns, not significant; **p < 0.01. Scale bars are indicated in the panels. We next tested whether HvSUSY2 was a phosphorylation substrate of HvSnRK1. In ATP‐containing reactions, HvSnRK1 induced a distinct slow‐migrating HvSUSY2 band on Phos‐tag gels, identifying HvSUSY2 as an in vitro substrate (Figure 4d). Addition of G1P reduced both HvSnRK1 autophosphorylation and HvSnRK1‐dependent HvSUSY2 phosphorylation in a concentration‐dependent manner (Figure 4e), showing that G1P‐mediated kinase inhibition was transmitted to the downstream substrate. Guided by the candidate site detected in the phosphoproteomic screen, we generated the phospho‐null variant HvSUSY2S814A. Relative to wild‐type HvSUSY2, S814A strongly reduced the HvSnRK1‐induced Phos‐tag mobility shift (Figure 4f), supporting S814 as a candidate site contributing to HvSnRK1‐dependent phosphorylation of HvSUSY2. To determine the functional consequence of this site, HvSnRK1 was co‐expressed in barley mesophyll protoplasts with HvSUSY2WT, HvSUSY2S814A or the phosphomimetic HvSUSY2S814D, and SUS activity and soluble‐sugar contents were measured with or without 5 mM G1P. G1P reduced SUS activity and increased sucrose accumulation in protoplasts expressing HvSUSY2WT. S814A likewise lowered SUS activity and promoted sucrose accumulation, whereas S814D retained higher SUS activity under G1P treatment and substantially attenuated sucrose accumulation. Glucose and fructose changed in parallel, indicating that substitutions at candidate S814 affected broader soluble sugar homeostasis (Figure 4g–j). These candidate site substitution results support a role for S814 in the intracellular sucrose cleavage function of HvSUSY2, and the S814D substitution partially bypassed G1P‐dependent inhibition of HvSnRK1. We then assessed whether G1P altered the HvSnRK1–HvSUSY2 association. Treatment with 5 mM G1P reduced HvSnRK1 fluorescence in the nucleus and cytoplasm and HvSUSY2 fluorescence in the cytoplasm, without clearly changing their principal subcellular distributions (Figure 5a,b). Under the same conditions, their cytoplasmic co‐localization signal declined. Pull‐down assays recovered less HvSUSY2 with HvSnRK1 after G1P treatment, and the split‐luciferase signal was likewise reduced (Figure 5c–e). Thus, beyond suppressing HvSnRK1 catalytic activity, G1P may destabilize the HvSnRK1–HvSUSY2 complex. Consistent with these molecular and cellular data, total SUS activity was lower in pho1 grain, together with reduced UDP‐glucose and increased sucrose. These data suggest that HvSnRK1 supports sucrose cleavage through phosphorylation of HvSUSY2, with S814 examined here as a candidate regulatory site, whereas G1P accumulation lowers HvSUSY2 phosphorylation by inhibiting HvSnRK1. The partial resistance of the S814D substitution to G1P further supports S814 as a candidate regulatory node through which G1P–HvSnRK1 may regulate sucrose metabolism. G1P weakens the intracellular association between HvSnRK1 and HvSUSY2. (a) Confocal images of HvSnRK1–RFP and HvSUSY2–GFP in barley mesophyll protoplasts under control conditions or after treatment with 5 mM G1P. (b) Quantification and line‐scan profiles of nuclear and cytoplasmic HvSnRK1 fluorescence, cytoplasmic HvSUSY2 fluorescence and cytoplasmic co‐localization. (c) In vitro pull down analysis of the HvSnRK1–HvSUSY2 association in the absence or presence of 3 or 5 mM G1P. (d) Split luciferase complementation assay of the HvSnRK1–HvSUSY2 interaction in tobacco leaves with or without G1P. (e) Quantification of the split‐luciferase signal. Data are mean ± s.d.; **p < 0.01. Scale bars are indicated in the panels. The HvSnRK1–HvSUSY2 Module Coordinates Grain Sugar Accumulation With Starch‐Precursor Formation We next asked whether the G1P–HvSnRK1–HvSUSY2 relationship coordinates soluble‐sugar accumulation with starch formation in grain by analysing snrk1 and susy2 mutants (Figures S18 and S19). Mature snrk1 grains accumulated significantly more total starch, sucrose, glucose, and fructose, partially reproducing the concurrent increase in starch and soluble sugars observed in pho1. By contrast, the major soluble sugars increased in susy2 grain, whereas total starch remained indistinguishable from the wild type (Figure 6a–d). Soluble‐sugar accumulation was therefore not sufficient to increase starch; the starch phenotype caused by reduced SnRK1 function also involved altered upstream precursor metabolism. HvSnRK1 and HvSUSY2 differentially regulate grain sugar accumulation and starch‐precursor supply. (a–d) Total starch (a), sucrose (b), glucose (c) and fructose (d) contents in mature grains of HTX, snrk1 and susy2. (e–h) G6P (e), G1P (f), UDP‐glucose (g) and ADP‐glucose (h) contents in the same genotypes. (i) Relative expression of genes involved in sucrose metabolism, sugar transport, hexose phosphorylation and starch precursor supply. (j, k) Total SUS (j) and AGPase (k) activities. Data are mean ± s.d.; ns, not significant; **p < 0.01. To examine this distinction, we quantified sugar phosphates and nucleotide sugars. G6P, G1P, UDP‐glucose and ADP‐glucose all increased in snrk1 grain, and the coordinated increases in G1P and ADP‐glucose matched the increase in total starch. In susy2, G6P increased, whereas UDP‐glucose, G1P and ADP‐glucose all decreased (Figure 6e–h). Loss of SUSY2 therefore caused soluble sugars to accumulate while restricting UDP‐glucose and starch‐related precursors, uncoupling sugar accumulation from starch formation. Expression and enzyme‐activity analyses revealed distinct metabolic adjustments in the two mutants. Although SUSY2 expression was lower in snrk1, several genes involved in sucrose synthesis and cleavage, sugar transport and hexose phosphorylation were upregulated, including SPS1, SPS2, CWINV1, INV1, SUT1 and HXK1. Total SUS activity remained at the wild‐type level in snrk1, consistent with a compensatory metabolic adjustment by other sucrose‐metabolic components, whereas AGPase activity increased substantially in parallel with ADP‐glucose and starch (Figure 6i–k). Because gene expression does not directly measure metabolic flux, the induction of invertase‐ and hexokinase‐related genes is interpreted as a plausible compensatory route rather than evidence that carbon flux through this route increased. In susy2, by contrast, induction of SUSY1, SUSY3 and several other sugar‐metabolic genes did not restore total SUS or AGPase activity, both of which remained significantly lower. Transcriptional compensation was therefore insufficient to restore overall sucrose‐synthase activity or to maintain UDP‐glucose, G1P and ADP‐glucose accumulation. Reduced SnRK1 function thus caused soluble sugars, G6P, G1P, ADP‐glucose and starch to rise together, whereas loss of SUSY2 separated soluble‐sugar accumulation from starch‐precursor formation. Reduced SUSY2 activity alone was therefore insufficient to reproduce the starch phenotype. These genetic results support a broader SnRK1‐dependent metabolic reprogramming in addition to the HvSUSY2 branch, with the latter contributing primarily to sucrose utilization and soluble‐sugar homeostasis. HvPHO1 Contributes to B‐Type Granule Architecture and Starch Physicochemical Properties Having established that loss of HvPHO1 altered carbohydrate deposition and sugar‐metabolic regulation, we examined starch‐granule morphology and size distribution in the pho1 endosperm. Scanning electron microscopy revealed altered granule morphology, with the most pronounced differences among B‐type granules (Figure 7a). Coulter‐counter analysis further revealed changes in A‐ and B‐type granule fractions, including an increased contribution and mean diameter of B‐type granules (Figure 7b–g). Loss of HvPHO1 therefore shifted the size composition of the endosperm B‐type granule population. HvPHO1 associates with HvSS4 and regulates B‐type starch granules architecture. (a) Scanning electron micrographs of endosperm starch granules from GP and pho1 grains. (b, c) Starch‐granule diameter distributions calculated by volume (b) and particle number (c). (d–g) B‐type granule volume fraction (d), mean B‐type granule diameter (e), A‐type granule volume fraction (f) and mean A‐type granule diameter (g) in GP and pho1 lines. (h–k) Yeast two‐hybrid, split‐luciferase complementation, co‐immunoprecipitation and bimolecular fluorescence complementation analyses of the HvPHO1–HvSS4 interaction. (l) Scanning electron micrographs of starch granules from GP and ss4 grains. (m, n) Amylopectin chain‐length distribution (m) and difference plot relative to GP (n). (o) ¹H NMR spectra of purified starch. (p) X‐ray diffraction (XRD) profiles of purified starch. Data are mean ± s.d.; **p < 0.01. Scale bars are indicated in the panels. To determine whether HvPHO1 associated with known granule‐initiation components (Kamble et al. 2023; Hawkins et al. 2021; Szydlowski et al. 2009; Roldán et al. 2007; Crumpton‐Taylor et al. 2013; Bürgy et al. 2021), we tested its interaction with candidate proteins implicated in B‐type granule formation (Kamble et al. 2023; Hawkins et al. 2021). Among those examined, HvPHO1 interacted with HvSS4 (Figure 7h–k and Figure S20), supporting a role for HvSS4 in the B‐type granule phenotype associated with HvPHO1. HvSS4 fusion proteins localized to plastids, and ss4 grains showed granule‐morphology defects resembling those of pho1 (Figure 7l). These results support a functional association between HvPHO1 and HvSS4 during endosperm starch‐granule formation. Because granule morphology and size distribution were altered in pho1, we assessed amylopectin fine structure and starch physicochemical properties. Amylopectin chain‐length distributions were unchanged, indicating that loss of HvPHO1 did not measurably alter glucan chain‐length composition (Figure 7m,n). Proton nuclear magnetic resonance and X‐ray diffraction revealed changes in branching characteristics and crystalline order (Figure 7o,p), whereas differential scanning calorimetry detected changes in gelatinization parameters (Figure S21). HvPHO1 therefore influences starch beyond the amount deposited, contributing to granule population structure and higher‐order physicochemical organization despite an unchanged amylopectin chain‐length distribution. HvPHO1 Is Highly Conserved, Whereas Coding‐Haplotype Diversity Is Reduced in Cultivated Barley To assess sequence conservation and coding variation of HvPHO1 across barley germplasm, we analyzed HvPHO1 protein sequences, tissue‐expression profiles and coding‐sequence haplotypes in distinct populations (Mascher et al. 2017; Monat et al. 2019). A protein phylogeny constructed from 997 aligned amino‐acid positions did not separate wild barley, landraces and improved cultivars into discrete groups; instead, accessions were interspersed on short branches (Figure S22). HvPHO1 protein divergence was therefore limited across the germplasm examined. HvPHO1 transcript abundance was highest in caryopses, remained relatively high in inflorescences and aerial tissues, and was lower in embryos and roots (Figure S23a), consistent with a prominent role in grain starch metabolism. Across 1681 barley accessions, we identified 131 single‐nucleotide polymorphisms in the HvPHO1 coding region and used them to define 82 coding‐sequence haplotypes (Mascher et al. 2017; Monat et al. 2019). Seventy haplotypes occurred in wild barley but only 18 in cultivated barley, and haplotype‐diversity indices confirmed substantially lower coding variation in the cultivated population (Figure S23a,b). Hap01 was shared by wild and cultivated barley; Hap02 differed from Hap01 by a single synonymous Leu424 variant and was enriched among cultivars. Together, Hap01 and Hap02 accounted for 91.94% of cultivated accessions but only 29.55% of wild accessions (Figure S23a,b), demonstrating a marked concentration of HvPHO1 coding haplotypes in cultivated germplasm. Rare missense variants occurred mainly in wild barley or population‐restricted accessions, indicating that wild germplasm retains allelic resources for resolving functional variation at HvPHO1. Discussion This study identifies PHO1 as a plastidial control point that links G1P use with sink carbon partitioning in barley endosperm. The central finding is that loss of PHO1 increases grain starch accumulation even though the expression or activity of several starch biosynthetic components is reduced. Instead of simply activating starch synthesis, PHO1 loss changes the balance of soluble sugars, G1P, ADP‐glucose and starch. This shows that PHO1 helps set the metabolic state of developing grains. This is important because G1P is usually viewed as a starch precursor, but our data show that it can also act as a regulatory metabolite when its plastidial use is restricted. In pho1 grains, G1P accumulation was associated with direct inhibition of SnRK1, weaker SnRK1‐dependent phosphorylation of SUSY2 and altered sucrose‐derived carbon partitioning. Thus, PHO1 appears to prevent a starch precursor from reaching a level at which it becomes a signal that feeds back on carbon use. In parallel, PHO1 also works with SS4 to shape B‐type starch granule architecture. These two branches define PHO1 as a link between precursor use, sugar signalling, and starch structure, rather than only as an enzyme in starch polymer metabolism. Under a simple model of starch metabolism, loss of a G1P‐using phosphorylase might be expected to reduce starch formation. Plastidial phosphorylases have long been linked to glucan metabolism and G1P use in starch‐related pathways (Smith and Zeeman 2020; Cuesta‐Seijo et al. 2017; Shoaib et al. 2021; Seung et al. 2025). The opposite response in pho1 grains shows that PHO1 has a broader role than supplying glucan synthesis directly. The increase in mature‐grain starch was modest but consistent, approximately 5–6 percentage points on a dry‐weight basis, and occurred even though the expression, abundance or activity of several starch‐biosynthetic components was unchanged or reduced. Loss of PHO1 removes one route of G1P utilization, making expansion of the G1P pool the most direct precursor‐level consequence of the mutation. ADP‐glucose also increased, providing a larger steady‐state starch‐precursor pool. However, ADP‐glucose concentration is a pool‐size measurement rather than a flux measurement and could reflect increased production, reduced utilization, or both. The modest change in AGPase activity in pho1 is therefore unlikely to explain the starch phenotype by itself. Consistent with a networked hexose‐phosphate response, exogenous G6P also promoted starch accumulation, although less strongly than G1P; because G1P and G6P are reversibly interconverted (Periappuram et al. 2000), part of the physiological response to G1P feeding may involve the broader hexose‐phosphate network. Together, these observations support altered precursor homeostasis upstream of starch polymer synthesis rather than simple activation of the canonical starch‐biosynthetic machinery. The G1P–SnRK1 relationship identified here should be placed within the established sugar‐phosphate regulation of SnRK1. Nunes et al. (2013) previously showed that G1P inhibits SnRK1 activity in plant extracts and that G1P and trehalose 6‐phosphate (T6P) can act cooperatively. The novelty of the present study therefore does not lie in the observation that G1P can inhibit SnRK1, but in demonstrating a direct interaction with a cereal SnRK1 protein and linking that regulation to endosperm carbon allocation. The ITC‐derived binding parameter and the millimolar concentration required for kinase inhibition describe different properties and should not be conflated. Comparison with T6P manipulation in wheat further indicates that the two sugar phosphates can produce overlapping but non‐identical metabolic outputs. Increasing T6P availability has been reported to promote a broad source–sink response and activation of sucrose‐to‐starch metabolism (Griffiths et al. 2026), whereas the higher starch content of pho1 is not accompanied by general upregulation of the canonical starch‐biosynthetic machinery. T6P is closely associated with sucrose availability, whereas G1P occupies a more proximal position in starch‐precursor metabolism and accumulates directly when PHO1‐dependent G1P utilization is disrupted. Because sucrose is elevated in pho1, a contribution of T6P to SnRK1 inhibition is biologically plausible, particularly given the cooperative inhibition reported previously (Nunes et al. 2013). However, T6P was not measured in this study, so its concentration and contribution relative to G1P cannot be resolved. The present data establish that G1P can directly bind and inhibit HvSnRK1, but they do not exclude an additional contribution from T6P in vivo. Reduced SnRK1 activity provides a second, coupled component of the starch phenotype. In snrk1 grain, soluble sugars, G6P, G1P, ADP‐glucose and starch all increased, and AGPase activity was substantially elevated. This shows that reduced SnRK1 function is sufficient to produce a broader metabolic state in which hexose phosphates and starch‐precursor pools increase together with starch deposition. Several invertase‐ and hexokinase‐related genes, including CWINV1, INV1 and HXK1, were induced in snrk1, consistent with a compensatory adjustment in sucrose cleavage and hexose phosphorylation; because expression is not a direct flux measurement, this remains a plausible route rather than demonstrated carbon flux. The susy2 phenotype separates this adjustment from the mechanism responsible for increased starch: although soluble sugars accumulated, G1P and ADP‐glucose decreased and total starch did not increase. Reduced SUSY2 activity alone is therefore insufficient to reproduce the starch phenotype of pho1 or snrk1. We consequently interpret increased starch as the combined outcome of precursor‐pool expansion and broader SnRK1‐dependent metabolic reprogramming, whereas the HvSnRK1–HvSUSY2 branch primarily contributes to sucrose utilization and soluble‐sugar homeostasis (Baena‐González et al. 2007; Peixoto and Baena‐González 2022; Nukarinen et al. 2016; Ruan 2014; Hu et al. 2022; Yang et al. 2024). PHO1 also affects starch structure through a branch that is separate from the G1P–SnRK1–SUSY2 pathway. The changes in pho1 starch granule morphology, amylopectin fine structure, crystallinity and thermal properties show that PHO1 loss alters more than carbon partitioning. The interaction between PHO1 and SS4, together with the similar granule phenotype of ss4, suggests that PHO1 works with SS4 to shape B‐type starch granule architecture in barley endosperm. This result is consistent with previous work linking plastidial phosphorylases and SS4‐related factors to granule initiation in cereal endosperm (Kamble et al. 2023; Hawkins et al. 2021; Szydlowski et al. 2009; Roldán et al. 2007; Crumpton‐Taylor et al. 2013; Bürgy et al. 2021). The repeated immunoblot analysis provides an additional biochemical context: AGPS1, GBSS1, SS2a and SBE2b were reduced in pho1, whereas AGPL1 and SBE2a were comparatively maintained. Because GBSS1, SS2a and SBE2b participate in glucan synthesis, branching and the organization of starch‐biosynthetic enzyme complexes, these reproducible abundance changes could plausibly contribute to the altered B‐type granule distribution and physicochemical properties, but the present data do not establish direct causality for any individual protein (Tetlow et al. 2004; Hennen‐Bierwagen et al. 2009; Kamble et al. 2023). It also explains why PHO1 cannot be considered only as a regulator of starch amount. Instead, PHO1 appears to connect two levels of starch biology: precursor use and granule architecture. This dual role may be relevant for barley quality, because starch amount, granule size distribution and starch physicochemical properties all affect grain end use (Fox and Bettenhausen 2023; De Schepper and Courtin 2024; Ayala et al. 2015). However, PHO1 should not yet be treated as a simple breeding target. The high conservation of HvPHO1 across barley germplasm is consistent with functional constraint rather than with a simple role as a negative regulator of starch accumulation. Loss of PHO1 increased starch only modestly in this study while also changing soluble sugars, precursor pools and starch structure. The haplotype analysis broadens the genetic context of PHO1, but its value lies mainly in guiding future functional tests. These natural haplotypes have not been experimentally linked to HvPHO1 activity, G1P abundance, starch content or granule architecture and should not be interpreted as functionally associated variants at present. Overall, our findings define PHO1 as a plastidial link between metabolic state and starch structure in barley endosperm. One branch acts through G1P and SnRK1 to tune sucrose‐derived carbon partitioning, whereas a second branch involves SS4 and shapes B‐type starch granule architecture. This dual role explains how a plastidial phosphorylase can affect both starch amount and starch organization, and broadens the role of G1P from a biosynthetic precursor to a signal of sink carbon regulation during grain filling (Figure S24). Materials and Methods Plant Materials and Growth Conditions The barley cultivar Golden Promise was used as the transformation recipient for clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR‐associated protein 9 (Cas9) editing (Harwood 2014). Following established barley transformation and genome‐editing procedures, CRISPR/Cas9 constructs targeting HvPHO1 or HvSS4, together with a full‐length HvPHO1 overexpression construct, were introduced into immature embryos of Golden Promise (Harwood 2014; Jiang et al. 2023). Regenerated plants were genotyped by PCR and sequencing to identify pho1 and ss4 edited lines and stable HvPHO1‐overexpression lines. For HvPHO1, the three homozygous T3 lines analyzed (L2‐4, L2‐6 and L2‐9) were single plant derived segregant lines carrying the same 5‐bp deletion recovered from T1 heterozygous plants. All transgenic materials used for phenotypic, biochemical and molecular analyses were homozygous T3 lines. Lines carrying mutations in HvSnRK1 or HvSUSY2 were identified by Targeting Induced Local Lesions IN Genomes (TILLING) screening of the HTX ethyl methanesulfonate (EMS)‐mutagenized population and are referred to as snrk1 and susy2, respectively (Jiang et al. 2022). Both mutants were backcrossed twice to the HTX wild‐type parent and then selfed once before homozygous lines were selected for analysis. Each mutant was compared with the corresponding wild‐type control. Plants were grown under greenhouse and field conditions. Greenhouse plants were cultivated in soil under a long‐day photoperiod (16 h light/8 h dark). Field experiments were conducted in Sichuan, China (30.5° N, 103.6° E), using a randomized complete‐block design with at least three biological replicates. Each plot comprised 20 rows of fifteen plants, with 25 cm between rows and 10 cm between plants. Spikes used for developing‐grain analyses were tagged at anthesis, and grains were collected at the indicated days after pollination (DAP). Developing grains used for RNA, protein, enzyme activity, and metabolite analyses were frozen in liquid nitrogen and stored at −80°C. Mature grains were air‐dried to constant weight before analyses of agronomic traits, grain morphology, grain composition, and starch physicochemical properties. Agronomic Traits and Grain‐Filling Dynamics Major agronomic traits were recorded at maturity. For each biological replicate, 50 plants were randomly selected to determine tiller number and productive spike number per plant. Fifty mature main‐stem spikes were randomly sampled per biological replicate to determine spikelet number per spike. Thousand‐grain weight was measured after mature grains had been dried to constant weight, with three technical measurements per biological replicate. Grain‐filling dynamics were assessed on tagged spikes. At each sampling point, 100 developing grains were collected at random per biological replicate and weighed to determine fresh mass. Samples were then heated at 105°C for 30 min and dried at 65°C to constant weight before dry mass was determined. Grain Morphology and Histological Analysis Mature‐grain morphology was quantified by image analysis. One hundred mature grains were randomly selected from each biological replicate, scanned with an Epson EU‐88 scanner and analyzed for grain length and width using WinSEEDLE. Representative grains were imaged with a Leica stereomicroscope. Dissected endosperm tissue was fixed in formalin–acetic acid–alcohol (FAA) for 24 h, dehydrated through an ethanol series, cleared in xylene, embedded in paraffin and sectioned at 10 μm. Sections were deparaffinized, rehydrated and stained with periodic acid–Schiff (PAS) reagent to visualize starch‐rich regions. Grain Composition Mature grains were milled before compositional analysis. Total starch, amylose and resistant starch were quantified using Megazyme kits K‐TSTA, K‐AMYL and K‐RSTAR, respectively, according to the manufacturer's instructions. Protein content was measured by the Kjeldahl method using a Kjeltec 8400 analyzer (FOSS, Denmark). Reagent blanks, calibration curves and dilution‐linearity controls were included in each assay. Quantification of Soluble Sugars and Sugar‐Metabolic Intermediates Soluble sugars were quantified by ethanol extraction followed by ion chromatography. Ground material (10–100 mg) was extracted in 80% ethanol at 50°C for 2 h. Extracts were analyzed using a Thermo ICS 5000+ ion‐chromatography system equipped with a CarboPac PA1 column and an electrochemical detector. Separations were performed at 30°C using water and 100 mM NaOH as the mobile phases. G1P, glucose 6‐phosphate (G6P), ADP‐glucose and uridine diphosphate glucose (UDP‐glucose) were quantified using commercial assay kits from Coibo (CB10838 and CB11761) and Geruisi (G0825W48 and G0881W48). Assays were performed according to the manufacturers' instructions; absorbance was recorded with a Varioskan LUX microplate reader, and concentrations were calculated from external calibration curves. Activities of Starch‐Metabolic Enzymes Grains collected at 15 DAP were assayed using Solarbio kits according to the manufacturer's instructions for the activities of ADP‐glucose pyrophosphorylase (AGPase), granule‐bound starch synthase (GBSS), soluble starch synthase (SSS), starch‐branching enzyme (SBE) and starch‐debranching enzyme (DBE). RNA Extraction and Reverse‐Transcription Quantitative PCR Total RNA was extracted using the BIOFIT Plant RNA kit and treated to remove genomic DNA. Complementary DNA was synthesized with HiScript III RT SuperMix (Vazyme). Reverse‐transcription quantitative PCR (RT–qPCR) was performed with ChamQ Universal SYBR qPCR Master Mix (Vazyme) on a QX real‐time PCR system. HvActin and HvGAPDH were used as reference genes, and relative transcript abundance was calculated by the 2−ΔΔCt method (Livak and Schmittgen 2001). Primer sequences and amplification information are provided in Table S3. Starch Granule Bound Protein Analysis Starch‐granule‐bound proteins were extracted from purified mature‐grain starch using a modified established protocol (Yamamori and Endo 1996). Purified granules were washed sequentially with sodium dodecyl sulphate (SDS)‐containing buffer, water and acetone, and granule‐bound proteins were released by boiling in sample buffer. Extracted proteins were resolved by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and visualized by silver staining. Total Protein Extraction and Immunoblotting Total protein was extracted from three independent biological samples from 15‐DAP endosperm in potassium phosphate buffer containing ethylenediaminetetraacetic acid (EDTA), glycerol, protease inhibitors and dithiothreitol (DTT). Protein concentration was determined by the bicinchoninic acid (BCA) assay. Equal amounts of protein were separated by SDS–PAGE, transferred to polyvinylidene difluoride (PVDF) membranes and immunoblotted with antibodies against starch‐biosynthetic proteins, including Bt2, Sh2, SBE2a, SBE2b, SS2a and GBSS1 (PhytoAB, Shanghai, China). HvPHO1 Catalysed G1P‐Dependent Glucan Extension and Kinetic Analysis The catalytic activity of purified HvPHO1 was measured using a modified plant α‐glucan phosphorylase assay. HvPHO1 was incubated with G1P in HEPES buffer containing soluble starch or maltohexaose as a glucan acceptor, together with DTT and bovine serum albumin (BSA). Reactions were conducted at 30°C and terminated by heating at 95°C. Because HvPHO1 uses G1P to extend glucans in the synthetic direction and releases inorganic phosphate (Pi), Pi release was used as the activity readout. Pi was quantified by the malachite green–molybdate method against a KH2PO4 standard curve. Residual G1P after reaction termination was independently quantified with a phosphoglucomutase/glucose‐6‐phosphate dehydrogenase‐coupled assay. G1P was first converted to G6P by phosphoglucomutase; G6P was then oxidized to 6‐phosphogluconate by glucose‐6‐phosphate dehydrogenase, with concomitant reduction of NADP+ to NADPH. Residual G1P was calculated from NADPH formation monitored at 340 nm. For apparent G1P kinetics, G1P was varied from 0 to 40 mM while the glucan‐acceptor concentration was held constant. Reaction time and enzyme amount were optimized to maintain linear product formation and less than 10% substrate consumption. Initial rates were fitted to the Michaelis–Menten equation; if rates declined at high G1P concentrations, a substrate‐inhibition model was used. Reactions lacking HvPHO1, G1P or glucan acceptor, or containing heat‐inactivated HvPHO1, served as negative controls. Subcellular Localization For localization in tobacco leaf cells, full‐length coding sequences without stop codons were cloned into pCAMBIA1302‐GFP to generate in‐frame green fluorescent protein (GFP) fusions. Sequence‐verified constructs were introduced into Agrobacterium tumefaciens GV3101. Agrobacteria carrying each target protein–GFP construct were mixed in equal volumes with agrobacteria carrying an appropriate red fluorescent marker and infiltrated with a needleless syringe into the abaxial surface of fully expanded leaves of 4‐week‐old Nicotiana benthamiana. Empty GFP vector served as a negative control. Plants were kept in darkness for 12–24 h after infiltration and then returned to 16 h light/8 h dark. Epidermal cells were imaged 48 h after infiltration. Localization was independently assessed in barley cells by cloning full‐length coding sequences without stop codons into pJIT163‐GFP or pRTVc‐RFP to generate in‐frame GFP or red fluorescent protein (RFP) fusions. Sequence‐verified constructs were transiently introduced into barley mesophyll protoplasts by polyethylene glycol (PEG)‐mediated transfection. Target‐protein fusions were co‐transfected with the corresponding fluorescent marker, with empty vectors as controls. Protoplasts were incubated for 16 h under dim light or in darkness before imaging. Fluorescence was captured with a Leica STELLARIS STED confocal laser‐scanning microscope (Leica Microsystems, Germany). Each experiment included at least three independent biological replicates; at least 10 independent fields of view or 20 expressing cells were examined per replicate, and consistent localization patterns were obtained. Yeast Two‐Hybrid Assay Bait coding sequences were cloned into pGBKT7 and prey coding sequences into pGADT7. Plasmid pairs were co‐transformed into Y2HGold yeast cells, and transformants were selected on synthetic defined medium lacking leucine and tryptophan (SD/−Leu/−Trp). Positive colonies were serially diluted and spotted onto SD/−Leu/−Trp and SD/−Ade/−His/−Leu/−Trp media to assess protein interactions. Empty‐vector pairs, bait plus empty prey, empty bait plus prey, and a known interacting pair were used as negative and positive controls, as appropriate. All bait constructs were tested for autoactivation and toxicity. Co‐Immunoprecipitation Target proteins were cloned into transient‐expression vectors carrying GFP, Myc or FLAG tags and introduced into A. tumefaciens GV3101. Bacteria harbouring different constructs were collected by centrifugation and resuspended in infiltration buffer (10 mM MES, pH 5.6–5.8, 10 mM MgCl2 and 100–150 μM acetosyringone), adjusted to the same optical density at 600 nm (OD600), mixed in equal volumes and infiltrated into leaves of 4‐week‐old N. benthamiana. Leaves were harvested after 48–72 h, ground in liquid nitrogen and extracted in protein lysis buffer. Clarified extracts were incubated for 2–4 h at 4°C with anti‐GFP, anti‐FLAG or anti‐Myc magnetic beads. Beads were washed four to five times with lysis or wash buffer, and immunoprecipitated proteins were eluted in SDS sample buffer. Input and immunoprecipitated fractions were immunoblotted with anti‐GFP, anti‐Myc or anti‐FLAG antibodies. Empty‐tag proteins, singly expressed samples and IgG‐ or bead‐only reactions served as negative controls. Bimolecular Fluorescence Complementation Target coding sequences were cloned into vectors encoding the N‐ or C‐terminal half of yellow fluorescent protein (nYFP or cYFP). Sequence‐verified constructs were introduced into A. tumefaciens GV3101, adjusted to equal OD600 and infiltrated into N. benthamiana leaves. Yellow fluorescent protein (YFP) fluorescence was examined by confocal laser‐scanning microscopy 48 h after infiltration. YFP was excited at 514 or 488 nm and emission was collected from 520 to 560 nm. Empty nYFP or cYFP combinations, singly expressed constructs and known non‐interacting protein pairs were used as negative controls; a known interacting pair served as a positive control. Exogenous G1P Treatment Exogenous G1P treatment was adapted from detached‐leaf sugar‐feeding procedures. Uniform barley seedlings were used as intact plants or to obtain fully expanded detached leaves. To reduce endogenous starch before treatment, plant material was preincubated in darkness. The cut bases of detached leaves were immersed in G1P‐containing solution so that G1P was taken up through the transpiration stream; for intact seedlings, the roots were placed in treatment solution. G1P was not sprayed or applied to the leaf surface. Samples were then incubated under controlled illumination. G1P concentration and treatment duration were selected from preliminary experiments. All solutions were prepared in pH‐matched buffer, adjusted to the same pH, and, when necessary, adjusted for osmotic potential and ionic strength with mannitol or the corresponding salt solution. To distinguish the effects of intact G1P from those of its glucose moiety or phosphate group, equimolar glucose, Pi, glucose plus Pi and G6P, pH‐matched buffer and untreated controls were included. After treatment, a subset of leaves was boiled in 80% ethanol to remove pigments and stained with iodine/potassium iodide (I2/KI) to visualize starch. Parallel samples were immediately frozen in liquid nitrogen and stored at −80°C for measurement of total starch, intracellular G1P, HvSnRK1 transcript abundance, and expression of related SnRK1 target genes. Intracellular G1P was measured to verify that exogenous treatment increased cellular G1P. Each treatment included at least three independent biological replicates. Molecular Docking Molecular docking of G1P to HvSnRK1 was used to predict a potential binding mode. The three‐dimensional structure of G1P was obtained from PubChem, energy‐minimized and assigned the predominant protonation state before docking. The predominant protonation state and net charge of G1P at pH 7.4 were predicted with small‐molecule preparation software and manually inspected. An experimentally determined homologous structure was used when available; otherwise, a homology model was generated for HvSnRK1. Where no experimental barley structure was available, an AlphaFold model was used and evaluated using predicted local distance difference test (pLDDT) and predicted aligned error (PAE) scores; low‐confidence disordered regions were removed (Jumper et al. 2021). Docking was restricted to the HvSnRK1 catalytic domain or the conserved kinase domain homologous to Arabidopsis KIN10/SnRK1. Before docking, water molecules were removed, hydrogens and Gasteiger charges were added, and the structure was energy‐minimized. Docking was performed with AutoDock Vina (Trott and Olson 2010). Guided by the structural mechanism of T6P–KIN10 (Baena‐González and Lunn 2020)/SnRK1 recognition and by sequence and structural alignment of HvSnRK1 with KIN10, the docking grid was centred on a candidate sugar‐phosphate‐binding region. For semi‐flexible docking, residues surrounding the candidate pocket that could form hydrogen bonds or electrostatic contacts with G1P were treated as flexible, while the remainder of the receptor was held rigid. Representative poses were selected by binding score, conformational clustering, hydrogen‐bond and salt‐bridge formation, and conservation of candidate contact residues. Docking poses were visualized in PyMOL (Schrödinger, LLC 2021) and with interaction‐analysis software. Molecular‐Dynamics Simulation A representative docked HvSnRK1–G1P pose was used as the starting structure for molecular dynamics simulations in GROMACS (Abraham et al. 2015). The protein was described with the AMBER99SB‐ILDN force field, and G1P parameters were generated with the General Amber Force Field 2 (GAFF2) and converted to a GROMACS‐compatible topology using AmberTools/ACPYPE. The protonation state and net charge of G1P were set for pH 7.4 and verified before topology generation. The complex was placed in a three‐dimensional periodic water box, solvated with the TIP3P water model, neutralized with Na+ and Cl−, and adjusted to 0.15 M NaCl. The system was energy‐minimized by steepest descent and then equilibrated sequentially under constant particle number, volume and temperature (NVT) and constant particle number, pressure and temperature (NPT), with restraints on protein and ligand heavy atoms. Unrestrained production simulations were run for 100 ns with a 2‐fs integration step. Long‐range electrostatics were treated with particle mesh Ewald, and bonds involving hydrogen were constrained with LINCS. At least three independent simulations were performed for each complex. Trajectories were analyzed for protein‐backbone root‐mean‐square deviation (RMSD), residue root‐mean‐square fluctuation (RMSF), radius of gyration, ligand RMSD, the number of protein–ligand hydrogen bonds, distances between G1P and key residues, and conformational changes in the binding pocket. Representative conformations were extracted by clustering and visualized in PyMOL. G1P‐Binding Assay Direct binding between HvSnRK1 and G1P was tested using purified recombinant HvSnRK1–maltose‐binding protein (MBP). Purified proteins were exchanged into binding buffer before analysis. HvSnRK1–MBP and G1P were analyzed by isothermal titration calorimetry (ITC) in the same buffer, with MBP alone as a negative control. Binding parameters were fitted in NanoAnalyze using an independent‐site model. Where indicated, excess free G1P or structurally related sugar phosphates were included for competition or specificity controls. Kinase‐Activity Assay Kinase activity was quantified with the ADP‐Glo Kinase Assay kit (Promega). After the in vitro kinase reaction, ADP‐Glo reagent was added according to the manufacturer's instructions to terminate the reaction and deplete residual ATP. Kinase Detection Reagent was then added to convert ADP to ATP, which was quantified by a luciferase‐based luminescence reaction. ATP/ADP calibration curves were used in every experiment to convert luminescence into ADP production. Reactions containing G1P but no kinase, G1P but no ATP, or buffer alone were used to subtract background and to test whether G1P interfered with the ADP‐Glo detection chemistry. In Vitro Pull‐Down Assay Target proteins were cloned into expression vectors carrying MBP or His tags. Constructs were introduced into Escherichia coli BL21(DE3), and fusion proteins were induced with isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) and purified with the corresponding affinity matrix. Equal amounts of immobilized bait and purified prey were rotated in pull‐down binding buffer for 2–4 h at 4°C. Beads or resin were washed extensively, and bound proteins were eluted in SDS sample buffer and denatured by boiling. Input and pull‐down fractions were separated by SDS–PAGE and immunoblotted with the corresponding antibodies. Empty‐tag proteins, bead‐ or resin‐only reactions and unrelated protein pairs served as negative controls. Split‐Luciferase Complementation Assay Target proteins were fused to the N‐ or C‐terminal fragment of luciferase (nLUC or cLUC) and introduced into A. tumefaciens GV3101. Agrobacteria carrying nLUC and cLUC constructs were mixed at equal OD600 and infiltrated into N. benthamiana leaves. At 48–72 h after infiltration, 1 mM D‐luciferin was uniformly sprayed onto or infiltrated into the leaves. After 3–5 min in darkness, luminescence was captured with a chemiluminescence imaging system. Empty nLUC/cLUC pairs, target proteins paired with empty fragments and unrelated protein combinations were used as negative controls. In Vivo Phosphorylation Analysis The phosphorylation state of HvSnRK1 in plant cells was examined in vivo. Sequence‐verified HvSnRK1–GFP and mutant constructs were transiently introduced into pho1 leaves by particle bombardment. Gold or tungsten particles were coated with plasmid DNA using a standard biolistic procedure, and bombarded leaves were allowed to recover under dim or normal light before harvest. Wild‐type leaves, empty GFP vector, a kinase‐inactive HvSnRK1 variant and non‐transformed leaves served as controls. Samples were ground in liquid nitrogen and extracted in ice‐cold buffer containing protease and phosphatase inhibitors. Clarified extracts were incubated with GFP‐Trap beads or anti‐GFP magnetic beads at 4°C with gentle rotation to enrich HvSnRK1–GFP. After washing, beads were divided into two fractions: one was eluted directly in SDS–PAGE sample buffer, and the other was treated with λ protein phosphatase. Before phosphatase treatment, immunoprecipitates were washed thoroughly and exchanged into a compatible buffer lacking EDTA, NaF, Na₃VO₄ and other phosphatase inhibitors. Dephosphorylation was performed according to the manufacturer's instructions. Samples were resolved by conventional SDS–PAGE and Phos‐tag SDS–PAGE. Phos‐tag gels were treated with EDTA before transfer to improve protein recovery. Following transfer to PVDF membranes, mobility shifts of HvSnRK1–GFP were detected with anti‐GFP antibody. Conventional SDS–PAGE was used to assess total HvSnRK1–GFP, whereas Phos‐tag SDS–PAGE resolved phosphorylation‐dependent mobility forms. A slow‐migrating band that disappeared or was markedly reduced after λ‐phosphatase treatment was considered phosphorylation dependent. In Vitro Phosphorylation The effects of G1P on HvSnRK1 autophosphorylation and phosphorylation of its substrate HvSUSY2 were assessed in vitro. Recombinant HvSnRK1–MBP, kinase‐inactive HvSnRK1–MBP, HvSUSY2–His and a phospho‐null HvSUSY2 variant were expressed in E. coli and purified. Protein purity was assessed by SDS–PAGE, and proteins were exchanged into kinase‐reaction buffer before use. Standard kinase reactions contained HEPES, MgCl2, DTT, ATP, purified HvSnRK1–MBP and HvSUSY2–His. To test G1P effects, HvSnRK1–MBP was pre‐incubated with a concentration series of G1P before ATP was added to initiate the reaction. Reactions were incubated at 30°C and terminated with SDS–PAGE sample buffer or ADP‐Glo termination reagent. G1P concentrations were selected from preliminary experiments; equimolar glucose, Pi, glucose plus Pi and buffer controls distinguished effects of intact G1P from those of the sugar or phosphate moiety. HvSnRK1 autophosphorylation was measured in reactions containing HvSnRK1–MBP alone, whereas HvSUSY2 phosphorylation was measured in reactions containing both HvSnRK1–MBP and HvSUSY2–His. Products were separated by Phos‐tag or conventional SDS–PAGE and immunoblotted with anti‐MBP and anti‐His antibodies, respectively. Kinase‐inactive HvSnRK1, the phospho‐null HvSUSY2 variant, reactions lacking ATP, kinase or substrate, and tag‐only proteins served as negative controls. Starch‐Granule Size Distribution and Morphology Starch granules were purified from mature grains by milling, filtration and Percoll density‐gradient centrifugation. Size distributions were measured with a Multisizer 4e Coulter counter (Beckman Coulter) using Isoton II electrolyte. The proportions of A‐ and B‐type granules were estimated by fitting the data to a bimodal distribution. Purified granules were mounted on coverslips, air‐dried, sputter‐coated with gold and imaged with a Gemini SEM 300 scanning electron microscope. Starch Structure and Physicochemical Properties Starch crystallinity was analyzed by X‐ray diffraction over a 2θ range of 4°–60°. Relative branching characteristics were measured by proton nuclear magnetic resonance (1H NMR) after dissolution in D₆‐dimethyl sulfoxide. Amylopectin chain‐length distributions were measured by ion chromatography on a Thermo ICS 5000+ system equipped with a CarboPac PA200 column after isoamylase debranching. Thermal properties were determined with a Discovery DSC 25 differential scanning calorimeter by heating samples from 40°C to 100°C at 10°C min−1. HvPHO1 Haplotype Analysis Coding‐region haplotypes of HvPHO1 were analysed as described by Gao et al. Sequence variants were inspected in Sequencher v4.8, and haplotype networks were generated in PopART. Raw data and detailed statistical outputs are provided in the Source Data file. Statistical Analysis Immunoblot band intensities were quantified in ImageJ/Fiji (Schindelin et al. 2012; Schneider et al. 2012). Two‐group comparisons were performed with two‐sided Student's t‐tests when statistical assumptions were met. Data from multiple treatments, concentration series, or time courses were analyzed by one‐ or two‐way analysis of variance, followed by an appropriate post hoc test for the experimental design. Sample sizes, statistical tests, p values and definitions of central tendency and dispersion are specified in the figure legends and Source Data. Unless stated otherwise, analyses and plots were generated in R (R Core Team 2024). Author Contributions Q.J., Y.W., and Y.Z. designed the experiments; Y.L., Y.C., J.L., G.G., and C.G. performed the experiments; Q.X., Y.Z., Y.L., P.Q., M.D., and P.Y. analyzed the data; S.D., J.M., G.C., and J.W. wrote the manuscript. All authors read and approved the manuscript. Funding This work was supported by the Sichuan Science and Technology Program, China (grant nos. 2024YFNH0023 and 2025YFHZ0113) and the Open Foundation of Key Laboratory of Wuliangye‐flavour Liquor Solid‐state Fermentation, China National Light Industry (grant no. 2023JJ011). Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Generation and molecular characterization of the HvPHO1 knockout mutant. Figure S2: Spatiotemporal expression of HvPHO1 during barley grain development. Figure S3: Subcellular localization of HvPHO1 in tobacco epidermal cells and barley protoplasts. Figure S4: Agronomic traits and mature‐grain morphology of GP and pho1 plants. Figure S5: PAS staining of developing GP and pho1 grains. Figure S6: Additional grain‐composition and soluble‐carbohydrate profiles of GP and pho1. Figure S7: Native‐promoter expression of full‐length HvPHO1 does not alter major mature grain carbohydrate pools. Figure S8: Expression and abundance of starch‐biosynthetic components in GP and pho1 grains. Figure S9: Developmental dynamics of sugar phosphates and nucleotide sugars in GP and pho1 grains. Figure S10: Transcriptome quality assessment and functional enrichment of differentially expressed genes in GP and pho1 grains. Figure S11: Metabolome quality assessment, chemical classification and pathway enrichment in GP and pho1 grains. Figure S12: Proteome quality assessment and functional enrichment of differentially abundant proteins in GP and pho1 grains. Figure S13: Phosphoproteomic profiling links HvPHO1 deficiency with altered sucrose synthase phosphorylation and activity. Figure S14: WGCNA‐based integrative analysis reveals coordinated transcriptome and metabolome modules in WT and pho1 grains. Figure S15: Molecular‐dynamics simulation supports the stability of the HvSnRK1–G1P complex. Figure S16: Functional analysis of predicted G1P‐contact residues in HvSnRK1. Figure S17: Dose‐ and time‐dependent effects of exogenous G1P on barley leaf carbohydrate status. Figure S18: Developmental expression dynamics of HvSnRK1 and HvSUSY2 during barley grain filling. Figure S19: Molecular characterization and grain composition of snrk1 and susy2 mutants. Figure 20. Developmental expression pattern and subcellular localization of HvSS4. Figure S21: Differential scanning calorimetry of starch from GP and pho1 grains. Figure S22: Phylogenetic analysis of HvPHO1 across barley germplasm. Figure S23: Tissue expression and coding‐haplotype variation of HvPHO1 in diverse barley germplasm. Figure S24: Model for HvPHO1‐dependent metabolic feedback and carbon partitioning in barley grain. Table S1: Phosphoproteomic analysis of PHO1‐mediated phosphorylation cascade between wild‐type and Hvpho1 grains at 15 DAP. Table S2: Nucleotide sequences of the HvPHO1 haplotypes. Table S3: Primers used in the study.