Sorghum (*Sorghum bicolor*) is a crucial cereal crop globally, providing essential nutrients. However, increasing demands due to population growth and climate change necessitate enhancing both the quantity and quality of sorghum grain production. Understanding the molecular mechanisms driving sorghum seed development is crucial for achieving this goal. Sorghum seeds are complex systems comprising genetically distinct tissues: a diploid embryo, a triploid endosperm, and diploid maternal tissues. Following double fertilization, the zygote forms the embryo, and the central cell develops into the endosperm, a nutrient-rich storage tissue vital for embryo growth and germination. Sorghum seed development typically takes 40–45 days, divided into three stages: an early stage (before 6 dpa), a middle stage (6–24 dpa), and a late stage (25–35 dpa). Starch and protein metabolism are key processes during seed development, involving various enzymes. Kafirins, the predominant seed storage proteins in sorghum, are crucial for grain quality. While transcriptomic analyses have been conducted in other plant species to understand seed development, a comprehensive investigation of sorghum seed development, integrating transcriptomic and metabolomic profiles, is lacking. This study addresses this gap by conducting an in-depth transcriptomic and metabolomic analysis of developing sorghum seeds, differentiating between embryo, endosperm, and whole seed tissues throughout seed development to identify key genes and metabolites regulating mature sorghum seed chemistry profiles.

Previous studies have investigated seed development in various plant species such as *Arabidopsis thaliana*, *Oryza sativa*, *Zea mays*, and soybean, using transcriptomic analyses. These studies have identified genes and transcription factors (TFs) associated with the biosynthesis of storage compounds like starch, protein, and oil. For example, RNA-Seq in maize identified genes associated with amylose and amylopectin biosynthesis, while transcriptome analyses of developing soybean seeds revealed hub genes implicated in oil and protein accumulation. Integrated metabolomic and transcriptomic analyses of rice seeds have also provided insights into anthocyanin modification. However, a comprehensive understanding of gene expression dynamics and the interplay between transcriptome and metabolome during sorghum endosperm and embryo development remains limited. A recent study analyzed the developing sorghum seed transcriptome from 5 to 25 dpa but did not differentiate between embryo and endosperm. This study aims to fill this knowledge gap by providing a more detailed and tissue-specific analysis of gene expression and metabolite accumulation during sorghum seed development.

The study used the sorghum inbred line 'BTx623', grown under field conditions. Developing seeds were collected at various timepoints (1–25 dpa) for morphological, kafirin, metabolomic, and transcriptomic analyses. Morphological changes were observed using scanning electron microscopy (SEM) to monitor starch granule development and kafirin accumulation. Kafirins (storage proteins) were quantified using HPLC to measure kafirin 1 and kafirin 2 levels. Untargeted metabolomic profiling using LC-MS identified and quantified metabolites. Transcriptome profiling involved RNA extraction from the embryo, endosperm, and whole seed at various developmental stages, followed by RNA sequencing (RNA-Seq). Data analysis included PCA, hierarchical clustering, co-expression network analysis using the STRING database and Mfuzz for soft clustering, and GO enrichment analysis using KEGG and ShinyGO. Quantitative real-time PCR (qRT-PCR) validated gene expression levels. Tissue-specific (TS) genes were identified by comparing the expression levels with previously published non-seed sorghum RNA-seq datasets using a TS scoring algorithm. Statistical analyses included Pearson correlation, Fisher's exact test, and false discovery rate (FDR) adjustments.

Morphological analysis showed starch granules appearing at 5 dpa, and kafirin accumulation increased significantly between 10 and 15 dpa. Metabolomic analysis identified 2073 metabolites across five timepoints, with significant changes in metabolite profiles throughout development. Pathway analysis revealed starch biosynthesis initiating at 5 dpa, transitioning to protein biosynthesis after 15 dpa. Transcriptome profiling identified 21,971 expressed genes, with higher expression in early whole seeds and endosperms compared to later stages. PCA separated samples based on tissue type, and hierarchical clustering grouped samples based on developmental phases. K-means clustering revealed co-expression clusters enriched in different pathways across developmental stages. In the endosperm, co-expression network analysis identified 361 hub genes associated with starch synthesis, mainly involved in the TCA cycle, ribosome biogenesis, oxidative phosphorylation, and starch metabolism. Similarly, 207 hub genes were identified in relation to kafirin biosynthesis, with functions in lipid metabolism, amino acid metabolism, and MAPK signaling. Analysis of tissue-specific genes identified 499 genes uniquely expressed in either the early whole seed, embryo, or endosperm, highlighting tissue-specific functions and potential regulatory roles.

The findings provide a comprehensive view of the spatiotemporal dynamics of gene expression and metabolite accumulation during sorghum seed development. The identified hub genes and metabolic pathways represent potential targets for genetic improvement to enhance grain quality. The shift from starch to protein biosynthesis after 15 dpa suggests a potential trade-off between these two major nutritional components, offering insights into optimizing carbon partitioning. The tissue-specific gene expression profiles highlight the distinct regulatory programs driving the development of each tissue. The study's findings contribute significantly to our understanding of sorghum seed development, offering a foundation for molecular breeding strategies to improve grain quality and yield.

This study provides a detailed transcriptomic and metabolomic analysis of sorghum seed development, identifying key genes and pathways involved in starch and protein biosynthesis. The identified hub genes and tissue-specific genes offer targets for molecular breeding to enhance sorghum grain quality. Future research should investigate the functional roles of the identified genes and TFs, focusing on allelic variation across diverse sorghum germplasm to elucidate how genetic diversity impacts carbon partitioning and seed traits under various environmental conditions.

The study focused on a single sorghum inbred line ('BTx623'), limiting the generalizability of the findings to other varieties. Environmental factors were not extensively controlled, potentially influencing the observed results. Further research with a broader range of sorghum genotypes and under controlled environmental conditions is needed to confirm the generality of these findings.