Trehalose-6-phosphate phosphatase-mediated trehalose metabolism shapes sorghum grain domestication for brewing adaptation Authors: Wenzhen Li, Ning Cao, Yanqing Ding, Jianxia Xu, Xu Gao, Bin Cheng, Jichao Zhang, Shangfeng Wang, Ruoruo Wang, Kuiyin Li, Shengjun Li, Liyi Zhang Published: 2026 ABSTRACT AbstractGrain length-to-width ratio (GLWR) is a crucial trait influencing sorghum grain quality, particularly in brewing applications such as Baijiu production. However, the genetic basis of GLWR remains poorly understood. Here, we performed a genome-wide association study (GWAS) on 255 Chinese sorghum accessions and identified a gene, encoding trehalose-6-phosphate phosphatase, named SbTPP1, as a principal modulator of GLWR. Population structure analysis revealed three distinct subgroups, with Chishui subgroup exhibiting the roundest grains. SbTPP1, located on chromosome 7, contains 2 InDels and 5 SNPs in the coding sequence that differentiate long-grain and round-grain alleles, which shows strong selection signatures in Chishui germplasm and higher expression in round-grain accessions. Overexpression of SbTPP1 in transgenic sorghum significantly increased GLWR but reduced thousand grain weight. Haplotype analysis of 737 global sorghum accessions traced the round-grain allele (Hap1 and Hap3) to African origins, while the long-grain allele (Hap2) arose during sorghum’s migration to China. This study establishes SbTPP1 as a candidate gene controlling GLWR underlying grain shape variation and population-specific selection in Chinese sorghum, via the trehalose metabolism pathway. These insights provide molecular insights into sorghum grain shape determination and offer valuable genetic markers for breeding programs aimed at optimizing grain morphology for industrial uses.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00122-026-05279-6. Introduction Sorghum (Sorghum bicolor (L.) Moench, 2n = 2x = 20) ranks as the fifth top cereal crop in the word and is recognized as a potential "star" crop for addressing global food security challenges (https://www.fao.org/faostat/zh/#data). This widely adaptable species is cultivated across tropical, subtropical, and temperate regions, serving diverse purposes including food, fodder, brewing, and bioenergy production (Mace et al. 2013; Smith et al. 2018). In China, sorghum is extensively cultivated from northern to the southern regions, with over 80% of the annual sorghum grain production currently dedicated to distilled liquor (Baijiu) brewing (Chen et al. 2019). It serves as a key raw material for renowned Baijiu brands across different aroma types, including sauce-aroma (e.g., Moutaijiu, Langjiu), strong-aroma (e.g., Luzhoulaojiao, Wuliangye), and light-aroma (e.g., Fenjiu) styles. Notably, distinct brewing processes impose specific requirements on grain properties. For instance, the production of Maotai liquor involves over 30-step, including nine rounds of cooking, eight rounds of fermentation, and seven rounds of distillation. This rigorous protocol demands sorghum grains with exceptional mechanical stirring resistance to maintain structural integrity—a critical trait for ensuring Baijiu quality (Zhao et al. 2024). Grain morphology directly influences this trait: Rounder grains exhibit superior stability during agitation due to uniform stress distribution and reduced surface friction, making them ideal for traditional brewing processes (Zhang et al. 2022). Despite its industrial significance, the genetic basis of grain morphology quantified by the grain length-to-width ratio (GLWR) remains inadequately explored in sorghum. To date, only a limited number of quantitative trait loci (QTL) associated with GLWR and other grain size-related traits have been identified in sorghum, with most distributed on chromosomes 1, 3, 6, and 7 via linkage mapping and GWAS (Guindo et al. 2019; Liu et al. 2019; Tao et al. 2018, 2020, 2021; Zhang et al. 2015). However, fine mapping and functional validation of GLWR-related QTL in sorghum remain largely inadequate. In contrast, due to the substantial impact of grain shape on appearance quality and market value in rice, QTLs and genes controlling rice GLWR have been particularly well studied, with genetic mapping efforts localizing these loci to chromosomes 1, 2, 4, 5, 7, 8, 10, 11, and 12 (Dong et al. 2018; Zhao et al. 2018; Zhang et al. 2020; Dwiningsih et al. 2021; Feng et al. 2021). Several studies have highlighted the role of the GL7/GW7 gene on chromosome 7, which influences grain length, width, and GLWR of rice (Wang et al. 2015a, 2015b). Additionally, genes such as GS3 and GW5 have been shown to be significantly correlations with rice GLWR (Fan et al. 2006; Weng et al. 2008; Mao et al. 2010; Feng et al. 2016). A recent study further revealed that the rice transcription factor OsSPL12 interacts with GW5 to jointly determine rice grain shape (Zhang et al. 2021). Cloned genes regulating rice GLWR are involved in multiple signaling pathways, including G protein signaling, the ubiquitin–proteasome pathway, the mitogen-activated protein kinase (MAPK) signaling pathway, plant hormone pathways, and transcriptional regulation (Li and Li 2016; Si et al. 2016). In contrast, few genes/QTLs associated with GLWR have been reported in sorghum (Boyles et al. 2016; Zou et al. 2020; Takanashi et al. 2021). Trehalose-6-phosphate phosphatase (TPP) is a pivotal regulator in plant biology, catalyzing the dephosphorylation of the signaling molecule trehalose-6-phosphate (T6P) to trehalose. This enzymatic activity modulates T6P levels, thereby governing carbon allocation, growth, stress responses, and sucrose partitioning between source and sink tissues (Figueroa and Lunn 2016; Kerbler et al. 2023). TPP is essential for regulating seed development, with its enzymatic activity controlling key processes such as seed filling and starch biosynthesis. T6P acts as a metabolic signal to coordinate carbon partitioning toward starch synthesis in developing seeds, with TPP activity directly influencing the spatiotemporal distribution of this key regulator. During the grain-filling stage, T6P suppresses the activity of SnRK1 kinase, alleviating growth inhibition and promoting resource allocation to seeds (Du et al. 2022; Wang et al. 2022). Mechanistically, TPP-mediated modulation of T6P levels serves as a key signaling hub to promote auxin biosynthesis, which in turn enhances sucrose uptake capacity in endosperm and cotyledons (Meitzel et al. 2020). Elevated auxin levels enhance sink strength by activating sucrose transporters and triggering the synthesis of storage compounds such as starch. This regulatory axis ensures the efficient translocation of photosynthates (primarily sucrose) from source organs (leaves) to sink organs (seeds), maximizing the conversion of carbon reserves into starch and other storage materials. Consequently, TPP activity fundamentally influences final seed size, weight, and quality (Liu et al. 2024). While TPP genes (e.g., RA3 and TPP4 in maize) contribute to other processes such as inflorescence branching and stress adaptation—including drought tolerance via stomatal regulation and root architecture modulation—their primary function remains focused on seed development (Satoh-Nagasawa et al. 2006; Claeys et al. 2019). Emerging evidence also links trehalose metabolism to panicle architecture, with pleiotropic effects mediated by SPL and GRAS transcription factors influencing grain shape (Yue et al. 2017). In rice, TPP orthologs interact with grain shape determinants like GW5 and GS3 through brassinosteroid and MAPK signaling pathways (Li and Li 2016; Zhang et al. 2021). Despite their well-characterized roles in other cereals, the functions of TPP genes in sorghum remain poorly understood. Notably, their potential involvement in regulating grain development—particularly through trehalose metabolism—has not been systematically explored, representing a critical gap in our understanding of sorghum’s genetic architecture. High-throughput resequencing technology, combined with comprehensive statistical analysis methods, provides a powerful tool for uncovering the molecular genetic mechanisms underlying important traits in sorghum. Genome-wide association study (GWAS) based on genome sequencing technology has been widely used in sorghum to genetically map genes/QTLs associated with traits such as grain weight per panicle, panicle number, and thousand grain weight (TGW) (Boyles et al. 2016; Tao et al. 2019), as well as traits including panicle shape, grain color, plant height, and maturity (Olatoye et al. 2020; Enyew et al. 2022; Kasule et al. 2024). In this study, we performed high-depth whole-genome resequencing of 255 Chinese sorghum accessions to analyze the genetic diversity and population structure of Chinese sorghum and to identify the QTL associated with sorghum grain length-to-width ratio (GLWR). Although part of the germplasm collection overlaps with our previous study, the resequencing data analyzed here were newly generated at a substantially higher sequencing depth, providing a reliable genomic foundation for accurate variant detection and association analysis. Based on this high-resolution dataset, we further characterized a key gene controlling this trait. These results provide a theoretical basis for marker-assisted selection (MAS) in sorghum breeding for Baijiu production. Materials and methods Plant material and phenotype identification The 255 sorghum accessions used in this study were procured from the Crop Germplasm Resources Center at the Chinese Academy of Agricultural Sciences in Beijing, China, and Institute of Upland Crops at the Guizhou Academy of Agricultural Sciences in Guiyang, China. The cultivation was conducted in Guiyang, Guizhou Province, as well as in Lingshui, Hainan Province, and Hangzhou, Zhejiang Province, spanning from 2018 to 2020 (Table S1). Grain length-to-width ratio (GLWR), associated with grain shape, was investigated by a crop grain appearance quality scanner (SC-E, Hangzhou Wanshen Technology Co., Ltd). Genome resequencing We utilized the cetyltrimethylammonium bromide (CTAB) method for the extraction of genomic DNA from 255 samples. Subsequently, the genomic DNA was submitted to Shenzhen Huada Gene Company for genome re-sequencing. Variant detection was carried out following the best practices workflow recommended by the Genome Analysis Toolkit (GATK4, version 3.8.1) as described by McKenna et al. (McKenna et al. 2010). Annotation of single nucleotide polymorphisms (SNPs) and insertions/deletions (indels) was performed using SNPeff6 (version 4.3 T) based on T2T-HYZ reference genome (Ding et al. 2024), following the methodology outlined by Cingolani et al. (Cingolani et al. 2012). Population genetic analysis Genetic population structure was delineated through fastSTRUCTURE v1.0 (Raj et al. 2014). After defining an initial K range (2–12) via cross-validation, ten independent analyses with random seeds were performed on the entire SNP dataset. Q-matrices were subsequently aligned using pong, clustered based on similarity, and matrices from the predominant cluster averaged to yield the final admixture matrix. As described in the previous research, we conducted an assessment of overall and chromosomal r2 values for each sorghum population(Zhang et al. 2022). Subsequently, we performed Linkage Disequilibrium (LD) analysis and inference of haplotype segments containing critical SNP peaks associated with grain length–width ratio using the LDBlockShow software (version: 1.39) developed by Dong et al. (Dong et al. 2021) to identify potential candidate genes. Furthermore, we utilized the VCFtools software (Danecek et al. 2011), employing a 10 kb sliding window with a 10 kb step size, to conduct Tajima’s D analyses on the population. The objective was to pinpoint genomic regions influenced by artificial selection. In all three analyses, regions satisfying the criteria within the top 5% were selected as indicative of signals of selective pressure. Phylogenetic inference utilized the core SNP dataset. A neighbor-joining tree was constructed with PHYLIP software (version 3.6974), complemented by PLINK-derived IBS distance matrices. The resultant phylogeny was rendered graphically using iTOL (v4.0; Letunic & Bork 2019). Principal components analysis, conducted with smartPCA (EIGENSOFT package v6.1.4; Price et al. 2006) on the core SNPs, visualized its first three eigenvectors via the R package. Genome-wide association study The association analysis was performed using the Genome Association and Prediction Integrated Tool (GAPIT version: 3.0) (https://www.maizegenetics.net/gapit), utilizing the Blink, FarmCPU, and MLM statistical methods. The first three principal components (PCs), derived from whole-genome SNPs, were utilized as fixed effects within the mixed model to account for stratification. Random effects were evaluated based on the clustering of kinship relationships integrated into the analysis. The significance threshold for GWAS was determined using the adjusted Bonferroni test, setting the threshold at P = -log10(0.05/2,249,619) = 7.6. Additionally, we conducted Linkage Disequilibrium (LD) analysis and inferred haplotype segments containing peak SNPs related to the recorded traits using LD BlockShow software (version: 1.39) developed by Dong et al. (2021) to identify potential candidate genes. Based on LD decay distance of 62.5 kb on average for the 255 Chinese accessions, QTN was defined as follows: All significant SNPs within a 125 kb were clustered in one interval, where the SNP with largest -log10(P) values was selected as QTN peak position, and its 62.5 kb upstream and downstream were determined as the QTN confidence interval, which was named as follows: QTN + chromosome + physical location (bp), for example, QTN1.16817670 referred to QTN at position of 16,817,670 bp on chromosome 1. Cloning of candidate gene Based on the candidate gene sequences provided by the reference genome HYZ, we designed gene-specific primers: F: 5’-atgacgaagcagggcgtggtgg-3’ and R: 5’-tcaccacttggcggggcgggc-3’. Genomic DNA was extracted from four round-grain sorghum accessions (064, 198, 238, and 250) and five long-grain accessions (029, 031, 052,107, and 129) and used as templates for PCR amplification. The genomic sequences of SbTPP1 from these accessions were obtained through PCR amplification, cloning, and Sanger sequencing. These nine accessions were selected from the resequenced panel based on extreme GLWR phenotypes (round vs. long grains). The PCR mixture consisted of 200 ng of sorghum genomic DNA, 1 μL of each forward and reverse primer (10 mM each), 0.2 μL of ExTaq DNA polymerase, 2 μL of 10 × ExTaq Buffer, 2 μL of dNTPs (2 mM), and 12.8 μL of ddH2O, with a total volume of 20 μL. PCR was performed using the following cycling conditions: 94 °C for 5 min, 35 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 2 min, followed by a final extension step at 72 °C for 5 min. The amplified products were subsequently purified, ligated into vectors, and subjected to sequencing analysis. RNA extraction and gene expression analysis To analyze the expression of SbTPP1, transcriptome sequencing was performed on sorghum accessions exhibiting extreme grain shape phenotypes. Two spherical grain accessions and two elongated grain accessions were selected for comparative expression analysis. Developing grain samples were collected at two key stages, 10 days post-anthesis (S1) and 20 days post-anthesis (S2), with three biological replicates for each sample. Total RNA was extracted, and transcriptome libraries were constructed using the NEBNext® Ultra™ RNA Library Prep Kit, following the manufacturer’s instructions. After quality assessment, paired-end sequencing was conducted on the Illumina NovaSeq 6000 platform using a paired-end sequencing strategy. Raw sequencing reads were processed using Trimmomatic software (Version 0.39)(Bolger et al. 2014) to remove adapters and low-quality sequences. Clean reads were then aligned to the Hongyingzi reference genome using Hisat2 software (Version 2.2.1)(Kim et al. 2019). Gene expression levels were quantified using Cufflinks (Version 2.2.1)(Trapnell et al. 2012) and normalized as fragments per kilobase of transcript per million mapped reads (FPKM). Vector construction and sorghum transformation The open reading frame (ORF) of SbTPP1 was amplified from sorghum varieties 064 (round) seedling cDNA using gene-specific primers SbTPP1-CDS-F (5’-actagggtctcgcaccatgacgaagcagggcgtggtgg-3’) and SbTPP1-CDS-R (5’-actagggtctctcgcctcaccacttggcggggcgggc-3’) using KOD-FX DNA polymerase (TOYOBO, Shanghai, China). The amplified PCR product was digested with BsaⅠ restriction enzymes and ligated into the corresponding vector pEGEOP to construct the overexpression vector pEGEOPubi-N-SbTPP1-GUS. This recombinant plasmid was introduced into sorghum P898012 via Agrobacterium tumefaciens strain EHA105-mediated transformation. Three positive transgenic sorghum lines were, respectively, designated OE-20, OE-25, and OE-31; then, GLWR and thousand grain weight (TGW) of transgenic lines were investigated. Evolutionary analysis of SbTPP1 Based on the amino acid sequence encoded by SbTPP1 (SbiHYZ.07G123900), protein conservation motifs were predicted using the Interpro website (https://www.ebi.ac.uk/interpro/) to analyze the protein sequences encoded by alleles controlling the round- and long-grain traits. Phosphorylation sites of the predicted proteins were analyzed using the online program NetPhos3.1. Homologous sequences in the sorghum genome were searched within the Phytozome database (https://phytozome-next.jgi.doe.gov/, version 13) using the BLAST function. The construction of a systematic evolutionary tree was performed using the neighbor-joining method with Bootstrap analysis (repeated 1000 t) in MEGA7.0 software. Subcellular localization of SbTPP1 protein The open reading frame (ORF) of SbTPP1 was amplified from sorghum varieties 064 and 029 seedling cDNA library using gene-specific primers SbTPP1-GFP-F(ctcaagcttggatccatgacgaagcagggcgtggtggtg)/SbTPP1-GFP-R (gctcaccatactagtccacttggcggggcgggcc). The resulting PCR product was digested and inserted into the same sites in the pHB plasmid (Mao et al. 2005) to generate the recombinant construct 2 × 35S::SbTPP1-CDS-GFP. This plasmid was subsequently transformed into Agrobacterium tumefaciens strain GV3101 using the freeze–thaw method (Weigel et al. 2006). Bacterial cultures were prepared for infiltration by resuspending cells in an induction buffer. Using low-light-grown sorghum seedlings as materials, stem and leaf tissues were harvested and subjected to enzymatic digestion to obtain protoplasts. Subsequently, protoplasts were purified through filtration, centrifugation, and washing with a solution containing 154 mM NaCl, 125 mM CaCl2, 2 mM KH2PO4(W5 solution), and 2 mM MES. Plasmids containing the target gene (for general localization) or both the target gene and marker gene (for co-localization) were introduced into protoplasts using PEG4000 solution, with the reaction conducted in a 24.5 °C water bath. After transformation, the reaction was terminated by adding W5 solution, followed by repeated centrifugation and washing to remove residual reagents. Finally, transformed protoplasts were cultured overnight at 28 °C under low-light conditions. After sample concentration via centrifugation, observation and image acquisition were carried out using a laser confocal microscope. Throughout the process, all operations from post-enzymatic digestion to pre-termination of transformation were performed gently on ice to ensure protoplast viability. Haplotype analysis of SbTPP1 Sequencing data from 737 sorghum germplasm resources, including African and Asian varieties, were used for haplotype analysis. Among these accessions, raw sequencing data for 360 accessions were obtained from previously published studies, while the remaining 377 accessions were sequenced in this study. All sequencing reads were aligned with the reference genome to obtain candidate gene sequences for all materials. The software CandiHap (V1.3.0) was utilized to construct haplotype networks and to investigate the geographical distribution of different haplotypes. Statistical analysis The study data were analyzed using analysis of variance (ANOVA) with SPSS software (IBM Corporation). The least significant difference (LSD) test was applied to compare mean values at a significance threshold of 0.05. Histograms were generated using Origin 8.0 (OriginLab Corporation, Northampton, Massachusetts, USA). Results Population structure in Chinese sorghum To investigate the population structure of Chinese sorghum germplasm, we initially performed whole-genome re-sequencing on 255 Chinese local sorghum varieties collected from 16 sorghum planting provinces across China, along with 6 foreign sorghum varieties (Supplementary Table S1). This resequencing yielded approximately 9.14 Gb of 150 bp paired-end reads, with an average sequencing depth of ~ 46.29 × and an average genome coverage of ~ 94.36%. By aligning the sequences to the ‘T2T-HYZ’ reference genome and identifying single nucleotide polymorphisms (SNPs), a total of 2,249,619 high-quality SNPs were identified across the 255 sorghum accessions. Comprehensive analyses of population structure, linkage disequilibrium (LD), and genetic diversity were conducted using cluster analysis and principal component analysis (PCA). The population structure of the 255 sorghum accessions was inferred using Admixture 1.3.0 software with K values ranging from 2 to 10. The optimal number of clusters (K = 3) was determined using the K-selector (Fig. 1a), resulting in three subgroups: Northern, Southern, and Chishui germplasm. LD analysis showed an average LD decay distance of 120 Kb (r2 = 0.2) across the 255 sorghum accessions (Fig. 1b). Population structure based on 2,249,619 SNP for 255 sorghums. A Genetic structure as inferred with fastSTRUCTURE of K = 2 to 10. B LD decay in different sorghum populations. C Neighbor-joining tree for sorghum accessions. The four subpopulations defined based on structures analysis are indicated by different colors. Blue, green, pink, and orange indicate sorghum form foreign, Chinese Chishui, North, and South China, respectively. D Principal component analysis of Chinese sorghum accessions, showing the first three principal components. Blue, green, pink, and orange indicate sorghum form foreign, Chinese Chishui, North, and South China, respectively Based on SNP-based cluster analysis, the 255 accessions were roughly divided into three subgroups according to their geographic origins (Fig. 1c), which was consistent with the phylogenetic tree analysis. The first subgroup (24 accessions), referred to as "Chishui" germplasm, originated from the Chishui River Basin in Guizhou and Sichuan Provinces (Southern China) and is primarily used for Baijiu-making. The second subgroup (70 accessions) mainly consists of other cultivars from Southern China and foreign germplasm. The third and largest subgroup (156 accessions) predominantly includes accessions from northern China, such as Liaoning, Inner Mongolia, and Hebei Provinces. PCA revealed that the first three principal components (PC1, PC2, and PC3) explained 34.44%, 25.43%, and 12.42% of the genotypic variance, respectively. Sorghum accessions from Northern China, Southern China, and the Chishui region were clearly separated along the PC3 axis (Fig. 1d). Phenotyping of sorghum grain shape In this study, GLWR was evaluated for 255 sorghum varieties across six different environments from 2018 to 2020 (Fig. 2a). The trait exhibited a continuous normal distribution. The greatest variation in GLWR was observed in the Guiyang (Guizhou Province) in 2018, while the smallest variation was noted in Guiyang in 2020. Within a single environment, GLWR values varied from 0.99 to 1.77, with mean values between 1.27 and 1.37, and standard deviation from 0.10 to 0.11. Correlation analysis demonstrated that the Pearson correlation coefficients for GLWR between different environments were highly significant (Fig. 2b), indicating strong genetic control of GLWR across the six environments and minimal influence of environmental factors on this trait. Frequency distribution and correlation analysis of grain length-to-width ratio (GLWR) in environments of Guiyang in 2018 (18GY), 2019 (19GY), 2020 (20GY), as well as Hangzhou in 2018 (18HZ), 2019 (19HZ), Lingshui in 2020 (LS) Selective sweeps analysis across populations Comparison analysis of GLWR among three Chinese sorghum populations revealed that Chishui sorghum had the roundest seeds (Fig. 3a, b), suggesting domestication-related selection. To investigate this, we integrated selective sweep analysis with GWAS (Fig. 3c). Selective sweep was first assessed using Tajima’s D, with the top 5% of the empirical distribution defined as candidate regions under selection. Additionally, Fst was also used to evaluate genetic differentiation between populations, with regions showing significant deviations in both Tajima’s D and Fst considered as candidate regions for selective sweep (Supplementary Table S2 and S3). This combined analysis identified a total of 843, 1410, and 1616 putative selected genes across the Southern, Northern, and Chishui populations, respectively. Pairwise population comparisons revealed 37 shared candidate genes between the Southern and Northern populations, 106 between the Southern and Chishui populations, and 83 between the Northern and Chishui populations. Notably, only three genes(SbiHYZ.03G086700, SbiHYZ.03G086800, and SbiHYZ.08G167400) were found to be shared across all three populations, indicating that the selection signatures are predominantly population-specific (Supplementary Table S4). Candidate gene for GLWR based on GWAS analysis. A Phenotypic difference between round-grain (GZ250) and long-grain (GZ009) accessions. Scale bars, 1.5 cm. B Comparisons of grain aspect ratio among South, North, and Chishui populations. The Chishui sorghum exhibits significantly the roundest seeds among three Chinese populations (***P < 0.0001, Student’s T test). C Selective sweeps region for grain aspect ratio based on Tajima’s D values among the South population, North population, and Chishui population. D Manhattan plots for grain aspect ratio using phenotype data from the 255-accession panel in Guiyang in 2018. E Local Manhattan plot (upper) of grain aspect ratio and linkage disequilibrium (LD) heat map (lower). F Expression pattern of total candidate gene in GZ009, GZ127 (long grain) and GZ198, GZ254 (round grain) in various stages using transcriptome sequencing: S1, 10 days post-anthesis; S2, 20 days post-anthesis Identification of candidate genes regulating GLWR To further refine candidate loci associated with grain shape, GWAS was conducted using 2,249,619 high-quality SNPs across six environments with three complementary models (BLINK, FarmCPU, and MLM). In total, 27 quantitative trait nucleotides (QTNs) were consistently detected (Supplementary Table S4). Among these, QTN7.57454142 on chromosome 7 was repeatedly identified in four environments and exhibited pronounced selection signals in the Chishui population, as evidenced by both Tajima’s D and population differentiation (Fst) analyses (Fig. 3d). This locus resides within a high-LD block containing a single candidate gene (SbiHYZ.07G123900) (Fig. 3d). Transcriptome analysis showed that the expression level of this gene was consistently significantly higher in round-grain materials than in long-grain materials across the two growth stages (Fig. 3e). Functional annotation and homology analysis showed that this gene encodes trehalose 6-phosphate phosphates and is homologous to rice gene LOC_Os08g31630 (OsTPP6) and the maize gene Zm00001d032298 (Kretzschmar et al. 2015; Zhang et al. 2017b; Wang et al., 2021). We designated this candidate gene as SbTPP1. To explore whether natural variation in SbTPP1 is associated with grain shape diversity, we further analyzed the gene structure and sequence variation of SbTPP1 in sorghum accessions with contrasting grain phenotypes. Cloning and sequence alignment of SbTPP1 from round-grain (sorghum varieties GZ064, 198, 238, and 250) and long-grain (sorghum varieties 029, 031, 052,107, and 129) accessions revealed that the full-length gene (starting from ATG) was 1941 and 1948 bp, respectively, both consisting seven exons and six introns (Fig. S1). The coding sequence (CDS) was 1131 and 1128 bp, respectively (Fig. S2). Comparison of SbTPP1 sequences between accessions with extreme phenotypic identified five SNP sites and two insertions/deletions (INDELs), all located in the first exon. Among these, three SNPs at positions + 92 bp, + 329 bp, and + 353 bp (G/T, G/T, and G/C) were synonymous mutations. A SNP at + 105 bp (A/G) resulted in an amino acid change from isoleucine (I) to valine (V) in long-grain accessions, while a G/T difference at + 270 bp led to a change from alanine (A) to serine (S).The first INDEL (CCGCCG, + 142 ~ + 147 bp) caused a deletion of two alanine (A) residues in the SbTPP1 protein of long-grain accessions, and the second INDEL (GGC, + 183 ~ 185 bp) resulted in the insertion of a glycine (G) in long-grained SbTPP1 (Fig. S3). These differences result in a one-amino-acid reduction in the SbTPP1 protein of long-grained accessions compared to round-grained accessions. Functional validation of SbTPP1 in grain shape regulation To functionally validate the role of SbTPP1 in grain development, we generated SbTPP1-overexpressing (SbTPP1-OE) transgenic lines in the sorghum inbred line P898012. Phenotypic evaluation revealed that overexpression of SbTPP1 significantly altered grain morphology, with OE lines producing narrower and more elongated grains compared to the wild-type (Fig. 4a). qRT-PCR analysis confirmed successful overexpression of SbTPP1 in the transgenic lines (Fig. 4b). Quantitative measurements further demonstrated that both grain length and grain width were significantly affected in SbTPP1-OE lines compared with WT(Fig. 4c, d). Notably, the OE-020 line exhibited the largest changes in both grain length and width, which was consistent with its most pronounced alteration in GLWR. GLWR was significantly increased in all SbTPP1-OE lines (P < 0.05, Student’s *t* test)(Fig. 4e), while the TGW was markedly reduced (Fig. 4f). These results demonstrate that SbTPP1 negatively regulates grain size while promoting slender grain shape in sorghum. Functional characterization of SbTPP1 in regulating grain morphology and grain shape in sorghum. A Representative image of mature grains from wild-type (WT) and SbTPP1-overexpressing (SbTPP1-OE) lines (OE-020, OE-025, and OE-031). B Relative expression levels of SbTPP1 in WT and SbTPP1-OE lines determined by qRT-PCR. C Grain length of WT and SbTPP1-OE lines. D Grain width of WT and SbTPP1-OE lines. E Grain length-to-width ratio (GLWR) of WT and SbTPP1-OE lines. F Thousand grain weight (TGW) of WT and SbTPP1-OE lines. G Phylogenetic analysis of trehalose-6-phosphate phosphatase (TPP) family proteins from sorghum (Sorghum bicolor), maize (Zea mays), rice (Oryza sativa), and Arabidopsis thaliana. SbTPP1 is highlighted in red. H Subcellular localization of SbTPP1 in sorghum protoplasts. GFP fluorescence, chloroplast autofluorescence, and merged images are shown. Note: Data in (B–F) represent mean ± SD from three independent biological replicates, with at least 10 mature grains measured per genotype for grain-related traits. Different lowercase letters indicate statistically significant differences at P < 0.05 (Student’s t test) To gain further insights into the evolutionary conservation and potential functional context of SbTPP1, we analyzed the TPP gene family in sorghum and examined the subcellular localization of SbTPP1. Analysis of the sorghum TPP gene family identified a total of 11 TPPs. Phylogenetic clustering of these sorghum TPPs with orthologs from maize (Zea mays) and rice (Oryza sativa) demonstrated that SbTPP1 (identified in our study) and the maize gene Zm00001d032298 are tightly clustered in a single clade and display the smallest genetic distance among the analyzed orthologs (Fig. 4e). To investigate the subcellular localization of SbTPP1, we constructed 35S::SbTPP1(L)-GFP and 35S::SbTPP1(R)-GFP vectors (CDS fused to GFP) and transiently expressed them in sorghum protoplasts. Confocal microscopy showed that while empty 35S::GFP vector exhibited diffuse distribution, both SbTPP1-fusion proteins colocalized with chloroplast autofluorescence (Fig. 4f). Haplotype analysis of SbTPP1 To investigate the distribution and origin of round-grained and long-grained sorghum, we analyzed the full-length coding sequence of SbTTP1 in a global collection of 737 sorghum accessions, including 163 African, 130 Asian, and 444 Chinese accessions(Table S5). The gene sequence variations observed in these accessions were consistent with those identified in the subset of 255 accessions, indicating that the detected variations were not rare and were representative of the broader population. We identified 5 SNPs and 2 INDELs in the coding sequence of SbTPP1 across all accessions, leading to the formation of three core haplotypes (Fig. 5a), while their overall frequencies are displayed in Fig. 5b. Among them, Hap1 was highly enriched across all groups and regions, with the highest frequency (588 occurrences, 80.54%), followed by Hap2 (127 occurrences, 17.40%) and Hap3 (8 occurrences, 1.10%). To assess the phenotypic effects of different haplotypes, GLWR was compared among accessions carrying different SbTPP1 haplotypes. Accessions carrying Hap2 had significantly higher GLWR values compared those carrying Hap1 and Hap3 (Fig. 5c). Haplotype analysis of SbTPP1 gene, phenotypic association, and phenotype distribution for three haplotypes. A Gene structure of HYZ showing exon–intron organization and untranslated regions (UTR), and schematic representation of major haplotypes based on sequence polymorphisms within the gene. Green boxes indicate exons, black lines indicate introns, dark gray boxes indicate the 5’-UTR, and light gray boxes indicate the 3’-UTR. Colored boxes denote nucleotide substitutions and insertion/deletion polymorphisms distinguishing Hap1, Hap2, and Hap3. The number of accessions and frequencies for each haplotype is shown on the right. B Phylogenetic relationships among the three HYZ haplotypes based on sequence variation. C Comparison of grain length-to-width ratio (GLWR) among accessions carrying different HYZ haplotypes. Boxes represent the interquartile range, center lines indicate medians, and whiskers indicate minimum and maximum values. Significant differences between haplotypes are indicated (**P < 0.01). D Geographic distribution of HYZ haplotypes across different regions, including Africa, Asia, Northern China (China_N), Southern China (China_S), and Chishui Analysis of three haplotypes distribution across geographical regions revealed distinct patterns (Fig. 5d). Hap1 was dominant globally, present in sorghum accessions from diverse regions. Notably, all accessions collected from the Chishui region belonged to Hap1. Hap2 was primarily found in sorghum varieties originating from Asia, including both Southern and Northern Chinese sorghum varieties. In contrast, Hap3 was exclusively identified in accessions of Africa origin, albeit at a low frequency. Integrating these findings with the historical migration of sorghum suggests that spherical grain morphology is likely the ancestral state form of SbTPP1, while the elongated grain shape observed in Hap2 appears to have evolved during the geographic expansion of sorghum, pointing to SbTPP1 as a candidate gene involved in grain shape diversification during domestication. Discussion Divergent domestication trajectories driven by brewing selection In China, sorghum has been cultivated since ancient times and has been widely used for liquor production since the Ming Dynasty (Zhang et al. 2022). Resequencing of 255 Chinese sorghum accessions revealed three genetically distinct subpopulations, among which the Chishui subgroup, originating from Guizhou–Sichuan Chishui River Basin, exhibits unique genetic divergence driven by adaptation to liquor-making demands. Notably, this subgroup displays rounder grains (lower GLWR), a critical trait for mechanical resistance during the 30-step brewing process of Maotaijiu. The phenotype of Chishui sorghum contrasts sharply with that of Northern and Southern Chinese varieties and reflects local, use-oriented selection associated with brewing demands, in a crop for which over 80% of China’s current sorghum production is dedicated to Baijiu manufacturing(Chen et al. 2019). Importantly, integration of these regional patterns with the global haplotype distribution of SbTPP1 provides an evolutionary framework for understanding this specialization. Our haplotype analysis supports rounder grain morphology (low GLWR) as the ancestral state of sorghum originating in African, whereas elongated grain types likely emerged during sorghum dispersal and diversification in Asia. The predominance of spherical grains in Chishui sorghum therefore does not represent an independent origin of this trait, rather than as the preferential retention and enrichment of an ancestral grain shape state driven by strong brewing-oriented selection. During sorghum diversification in China, multiple grain shape variants were generated and selected for different end-use purposes; in this context, the Chishui region represents a distinctive case in which brewing practices favored the maintenance of low GLWR grains due to their superior mechanical stability during repeated steaming, stacking, and fermentation. Population genomic analyses further support this specialized domestication. Tajima’s D analysis identified 1,616 genes under selection in Chishui sorghum, including the chromosome 7 locus QTN7.57454142 for GLWR in GWAS. This locus resides within a high-LD block, suggesting strong artificial selection for grain roundness. The Chishui-specific selective sweep at QTN7.57454142 highlights brewing-driven adaptation, as this grain shape regulator shows no comparable selection in non-brewing varieties. Functionally, rounder grains in Chishui sorghum reduce surface friction during agitation, preserving structural integrity—a key trait for quality of sauce-flavor Baijiu production(Sajid Mushtaq et al. 2021). In contrast, sorghum varieties bred primarily for grain yield often prioritize increased TGW as a key breeding target (D’Andrea 2008; Newsom et al. 2011). Together, these patterns illustrate that sorghum domestication in China has followed divergent trajectories, with brewing-oriented selection favoring the retention of ancestral grain morphology in Chishui, whereas other regions and end uses promoted distinct grain shape and yield-related phenotypes. Such end-use-driven domestication patterns are consistent with broader models of crop evolution in which local cultural practices impose strong, trait-specific selection that can override or redirect generalized domestication trends (Dillon et al. 2007; Fuller et al. 2014; Zhang et al. 2018; Wu et al. 2022). SbTPP1: a domestication-related regulator of grain shape In this study, GWAS identified 27 QTNs associated with GLWR. Among these, QTN7.57454142 on chromosome 7 was consistently detected across all environments and overlapped with a previously reported QTL (58.8–61.93 Mb) for GLWR (Liu et al. 2019). High-density SNP mapping (2.2 million markers) precisely localized this signal to SbTPP1, which encodes an enzyme in the trehalose pathway. TPP plays a crucial role in seed development by regulating meristem determinacy, starch accumulation, stress response, and hormonal signaling. Knockout of maize ZmTPS9 increases seed starch content and weight by elevating T6P levels, which activates starch biosynthesis (Hu et al. 2021). Overexpression of rice OsTPP1 in maize improves yield under drought conditions by reducing T6P leves, thereby enhancing sucrose availability for seed development (Nuccio et al. 2015). In Arabidopsis, AtTPPI regulates stomatal aperture via ABA signaling, influencing water use efficiency and carbon supply for seed development (Lin et al. 2020). In peas, the absence of T6P leads to downregulation of TAR2 expression, a reduction in auxin synthesis, which in turn inhibits the activity of starch synthase AGP and results in a wrinkled seed phenotype; conversely, increased T6P levels promote embryo cell expansion and starch granule development by activating auxin signal (Meitzel et al. 2020). In maize, the TPP genes RAMOSA3 (RA3) and TPP4 maintain determinacy of axillary meristems, preventing abnormal branching and ensuring proper seed number (Satoh-Nagasawa et al. 2006; Claeys et al. 2019). Maize ZmTPS9 knockout increases seed starch content and weight by elevating T6P, activating starch biosynthesis (Hu et al. 2021). Overexpression of rice OsTPP1 in maize improves yield under drought by reducing T6P, enhancing sucrose availability for seed development (Nuccio et al. 2015). In wheat, TaTPP-7A acts as a key domestication- and breeding-related gene that regulates grain filling and yield by influencing sucrose decomposition, transport, and utilization through the T6P-SnRK1 signaling pathway and sugar-ABA interaction feedback loop (Liu et al. 2024). Notably, overexpression of TPP genes in cereals enhances yield under stress but may inadvertently alter grain dimensions—a trade-off that requires genetic dissection (Kerbler et al. 2023). Although our results clearly demonstrate that SbTPP1 modulates GLWR in sorghum, the precise downstream molecular cascade linking trehalose-6-phosphate (T6P) homeostasis to grain shape determination remains to be fully elucidated. Based on previous studies in cereals and model plants, SbTPP1-mediated regulation of T6P levels may influence grain morphology through multiple, potentially interconnected pathways, including SnRK1-dependent carbon allocation, auxin-mediated control of cell expansion, and the regulation of sink strength during early grain development. T6P has been shown to act as a central signal reflecting sucrose availability, thereby coordinating cellular growth and carbon partitioning in developing seeds. In this context, altered T6P levels resulting from SbTPP1 activity could affect the balance between longitudinal and radial cell expansion in the grain, ultimately shaping grain length and width. Dissecting the direct downstream targets of SbTPP1 and the spatiotemporal dynamics of T6P signaling during grain development will require further genetic and molecular analyses, which are beyond the scope of the present study but represent an important direction for future research. Our results support that SbTPP1 has been a target of selection during sorghum population differentiation and improvement, contributing to grain shape diversification through GLWR regulation and linking trehalose metabolism with brewing-associated grain shape. Haplotype analysis of SbTPP1 across 737 global sorghum accessions revealed distinct selection patterns. Hap3, which is associated with smaller GLWR, can be traced back to Africa—the center of sorghum domesticated—supporting the notion that the small GLWR (round grain) trait originated from the ancestral gene pool of sorghum. In contrast, Hap2 is exclusively found in Chinese sorghum varieties (excluding those from the Chishui region), suggesting that Hap2 represents a more recent variant selected during local production adaptation to production needs. Implications for sorghum breeding: balancing grain shape and yield TPPs have presently emerged as promising tools in crop breeding, offering significant potential to enhance both yield and quality by regulating the trehalose-6-phosphate (T6P) metabolic pathway. Although the present study primarily focuses on grain shape-related traits, such as GLWR and TGW, which are closely associated with grain mechanical stability during traditional Baijiu brewing, comprehensive evaluation of additional brewing-relevant phenotypes—including grain hardness, and starch content—will be essential to fully assess the industrial utility of SbTPP1. Such evaluations typically require specialized processing conditions and large-scale fermentation assays and therefore represent an important direction for future studies. Evidence from other cereal crops underscores the broad agronomic potential of TPP genes. In wheat, the functional marker TaTPP6AL1-CAPS, derived from TaTPP-6AL1, distinguishes alleles associated with increased TGW, enabling marker-assisted selection for high-starch genotypes (Zhang et al. 2017a). In maize, manipulating TPP/TPS levels profoundly impacts yield and composition. Overexpression of rice OsTPP1 driven by the OsMads6 promoter in maize reduces T6P levels, enhances sucrose availability, and improves yield by 9–123% under diverse water conditions (Nuccio et al. 2015). Conversely, knockout of ZmTPS9 elevates kernel starch content and weight by activating starch biosynthesis (Hu et al. 2021). In rice, OsTPP1 enhances chilling tolerance during seed development, ensuring grain maturation under stress (Zhang et al. 2017b), while OsTPP6 regulates grain size through modulation of the brassinosteroid signaling pathway (Kretzschmar et al. 2015). Crucially, the function of TPP genes exhibits significant phylogenetic conservation among cereals (e.g., maize RA3 and its sorghum orthologs), indicating potential for transferring desirable traits across cereal species (Claeys et al. 2019). Central to this mechanism is T6P-mediated balancing of source–sink carbon allocation during grain development (Satoh-Nagasawa et al. 2006). Leveraging this T6P signaling mechanism provides a powerful strategy to enhance grain-filling efficiency, shed light on evolutionary mechanisms (Kretzschmar et al. 2015; Liu et al. 2024), establish a template for broad-adaptability breeding, and open new avenues for evolutionary breeding approaches. In this study, SbTPP1 overexpression demonstrates a clear trade-off: While it improves GLWR, it reduces TGW. This necessitates that breeders carefully balance these traits to optimize crop performance. Thus, modulating the trehalose pathway through TPP presents a promising approach for enhancing crop yield and quality. The pleiotropic effects of SbTPP1 necessitate strategic breeding approaches. The identified nonsynonymous SNPs and InDels offer potential markers for marker-assisted selection. Pyramiding rounder alleles into elite backgrounds could enhance processing resilience without compromising yield for brewing, as Chishui accessions already exhibit optimal GLWR. On the other hand, fine-tuning SbTPP1 expression (e.g., via promoters specific to early grain development) may mitigate the TGW penalty observed in overexpression lines to improve grain yield. In conclusion, by integrating population genomics, GWAS, functional validation, and evolutionary analysis, this study establishes SbTPP1 as a major domestication gene that shaped sorghum grain morphology in response to brewing selection. The Tre6P pathway emerges as a versatile toolkit for cereal improvement, offering molecular levers to reconcile industrial quality requirements with agronomic performance. These findings not only advance our understanding of sorghum domestication but also provide a framework for studying grain shape evolution in other crops. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PNG 1299 KB) Supplementary file2 (PNG 826 KB) Supplementary file3 (PNG 586 KB) Supplementary file4 (XLSX 529 KB) Author contribution WL and NC drafted the original manuscript. WL and NS, WL, YD, JX, XG, BC, JZ, SW and RW conducted experiments, analyzed and interpreted the data. KL, SL and LZ designed and supervised the research, obtained funding and revised the manuscript. All authors have read and approved the final version of the manuscript. Funding National Natural Science Foundation of China, 32460493, Kuiyin Li, The Outstanding Team of Guizhou Academy of Agricultural Sciences (202606), Liyi Zhang, QKH Basic-ZK [2025]299, Wenzhen Li, Guizhou Academy of Agricultural Sciences, [2025]02, Wenzhen Li, Guizhou provincial key laboratory of biotechnology breeding for special minor cereals, QKHPT[2025]026, Liyi Zhang. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wenzhen Li and Ning Cao have contributed equally to this work. Contributor Information Shengjun Li, Email: sjli@sdau.edu.cn. Liyi Zhang, Email: lyzhang1997@hotmail.com. References Associated Data Supplementary Materials Supplementary file1 (PNG 1299 KB) Supplementary file2 (PNG 826 KB) Supplementary file3 (PNG 586 KB) Supplementary file4 (XLSX 529 KB)