Genomic and Transcriptomic Insights into Stage-Specific Drought Resilience in Diverse Myanmar Rice Germplasm

Abstract

Drought stress remains a critical constraint to rice productivity, particularly during the early vegetative stages in rainfed environments. To elucidate the genetic and molecular mechanisms underpinning drought tolerance in rice, we conducted an integrated genome-wide association study (GWAS) and transcriptomic analysis on 236 genetically diverse Myanmar landraces, a region renowned for its unique and locally adapted rice germplasm. Phenotypic evaluation under simulated drought conditions revealed substantial variation in germination and seedling survival, with approximately 18% of accessions exhibiting high tolerance and 22% displaying susceptibility. Notably, contrasting responses between germination and seedling stages in some landraces suggest stage-specific genetic regulation of drought resilience. Population structure analyses demonstrated distinct clustering aligned with geographic origin, reflecting local adaptation and a complex evolutionary history distinct from other major rice populations. GWAS identified twelve significant QTLs across chromosomes 2, 4, 5, 7, 8, 9, 10, and 11, containing 546 candidate genes involved in ABA signaling, osmotic regulation, and stress-responsive pathways. Haplotype analysis at key loci, particularly on chromosome 7, revealed allelic variants strongly associated with enhanced drought tolerance, exemplified by favorable haplotypes linked to higher germination rates under stress. Complementary RNA-seq profiling of a superior drought-tolerant genotype (V5) and a highly sensitive one (V3) uncovered 3,476 and 2,590 differentially expressed genes, respectively. Tolerant landraces exhibited downregulation of photosynthesis-related genes and upregulation of osmotic adjustment and detoxification pathways. Integration of GWAS and transcriptomic data pinpointed 103 candidate genes within QTL regions, with Os07g0513000 (ATP synthase gamma chain) and Os07g0691200 (D-alanine ligase) emerging as prime candidates due to their strong upregulation in tolerant lines and linkage disequilibrium with major QTLs. Validation via qRT-PCR confirmed their potential roles in drought adaptation. These findings highlight the unique genetic architecture of Myanmar landraces, offering valuable alleles and regulatory networks for molecular breeding aimed at enhancing drought resilience in rice, and underscore the importance of conserving regional landraces as vital resources for climate-smart agriculture.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12284-025-00877-2.

Introduction

Drought represents one of the most pervasive and deleterious environmental stresses threatening global agriculture, exerting profound impacts on crop growth, development, and productivity. Its occurrence often results in extensive plant damage and substantial yield reductions, thereby jeopardizing food security worldwide (Bailey-Serres et al. 2019; Baatz et al. 2025). This challenge is particularly acute in regions heavily reliant on rice (Oryza sativa L.) as a staple food source, where drought episodes during critical growth stages can lead to yield losses exceeding 60% (Rasheed et al. 2020; Mishra et al. 2017). Consequently, elucidating the genetic and molecular mechanisms underlying drought tolerance, especially during early developmental phases, is imperative for breeding resilient rice varieties capable of withstanding water-deficient conditions.

Myanmar’s rice landraces constitute a highly valuable yet underexploited reservoir of genetic diversity, having evolved distinctive adaptive traits such as optimized root architecture, enhanced water-use efficiency, and unique stress-responsive molecular pathways (Myint et al. 2023; Thant et al. 2021; Sato and Yokoya 2008). While extensive research has characterized the susceptibility of rice to drought at the reproductive stage (Mishra et al. 2017), the specific mechanisms conferring tolerance during early vegetative growth in traditional landraces remain poorly understood. Most prior studies have predominantly focused on modern cultivars and their responses during later developmental stages, leaving critical gaps in our understanding of early-stage drought responses and the broader genomic potential of Myanmar’s unique germplasm, particularly its specialized adaptations (Thant et al. 2021; Swamy and Kumar 2013; Aung et al. 2020) Despite significant advances in breeding programs utilizing tolerant donor lines, the development of cultivars that successfully combine high yield potential with robust drought resilience remains a formidable challenge (Pandey and Shukla 2015). Recent breakthroughs have begun to unravel the genetic basis of drought tolerance: for instance (Sun et al. 2022) identified a functional promoter single-nucleotide polymorphism (SNP) in DROUGHT1 (DROT1) associated with enhanced gene expression and drought resilience, while (Wang et al. 2024) uncovered 80 candidate genes, including qRT9, through large-scale screening of diverse rice accessions. Molecular studies have further elucidated key regulatory pathways, such as the role of the bZIP transcription factor (TF) EDT1 in activating drought resistance mechanisms (Wu et al. 2019), and the function of OsSAPK3 in modulating both ABA-dependent and ABA-independent stress responses while simultaneously improving yield traits (Lou et al. 2023). These advances contrast with earlier research that primarily identified drought-responsive genes such as OsSIZ1, OsNHX1, OsNAC10, OsIQM1, OsWRKY89, OsHsp17.4, OsHSP17.7, OsPP2C30, and OsPP48 (Ross et al. 2007; Sato and Yokoya 2008; Jeong et al. 2010; Almeida et al. 2017; Sarkar et al. 2020; Fan et al. 2021; Kim et al. 2022; Zheng et al. 2022). However, a significant knowledge gap persists regarding the molecular and genetic basis of early vegetative-stage drought responses, particularly in traditional landraces, which harbor unique adaptive traits.

Recent technological advances in genomics and transcriptomics provide powerful tools to address these gaps. Genome-wide association studies (GWAS) enable comprehensive identification of genetic variations linked to complex traits across diverse natural populations, leveraging historical recombination events for higher resolution (Huang et al. 2010; (Nyasulu et al. 2024; Garreta et al. 2021). Complementarily, RNA sequencing (RNA-Seq) offers dynamic insights into transcriptional reprogramming under stress conditions, revealing differentially expressed genes (DEGs), key biological pathways, and regulatory networks (Almeida et al. 2017; Zhang et al. 2017). The integration of GWAS and transcriptomic data provides a synergistic framework: GWAS pinpoints genomic regions associated with drought tolerance, while RNA-Seq refines candidate gene prioritization within these loci, thereby reducing false positives and enhancing predictive accuracy (Xie et al. 2019; Wang et al. 2022a). Despite these technological advancements, their combined application to investigate early-stage drought responses in traditional rice landraces remains limited, representing a critical gap in our understanding of drought resilience mechanisms.

Successful seed germination and seedling establishment are fundamental for crop productivity under water-limited conditions. Genotypic variation in germination speed and synchronization under drought stress can significantly influence early vigor and overall crop establishment (Islam et al. 2018). Exploring the untapped genetic diversity within landraces offers promising avenues for identifying key adaptive traits necessary for developing climate-resilient rice varieties. This study aims to bridge existing knowledge gaps by employing an integrated GWAS and transcriptomic approach to identify genomic loci and candidate genes associated with early vegetative-stage drought tolerance in Myanmar rice landraces. We characterize stress-responsive DEGs and their regulatory networks, validate candidate genes, and elucidate genotype-phenotype relationships. By dissecting the genetic architecture underlying drought tolerance in these traditional germplasms, our work provides fundamental insights into stress adaptation and practical targets for molecular breeding. Ultimately, these findings will contribute to the development of drought-resilient rice cultivars, bolster climate adaptation strategies, and support global food security while conserving valuable agrobiodiversity.

Materials and Methods

Plant Materials

A total of 236 landrace rice (O. sativa L.) accessions were utilized in this study, representing a broad spectrum of agroecological and agro-climatic zones across Myanmar (Table S1). Seeds were sourced from the Seed Bank of the Department of Agricultural Research, Ministry of Agriculture, Livestock, and Irrigation, Myanmar. These accessions encompass seven distinct geographical regions, which also serve as proxies for major agro-climatic zones; the Central Dry Zone (CDZ, low rainfall and arid conditions), Delta Region (DR, flood-prone fertile plains), Eastern Coastal Region (ECR), Eastern Mountainous Region (EMR, cooler highlands), Northern Hilly Region (NHR), Western Coastal Region (WCR, maritime influence), and Western Mountainous Region (WMR) (Fig. S1). The collection includes landraces adapted to diverse environmental conditions, such as upland, rainfed lowland, and irrigated ecosystems. All materials were propagated during the summer of 2023 to ensure sufficient seed stock for subsequent phenotypic evaluations.

Evaluation of Drought Tolerance at the Early Vegetative Stage

Assessment of Germination and Seedling Development Under Drought Stress

To standardize germination vigor, only seeds exhibiting uniformity (intact, mature grains) were selected. Seeds were surface-sterilized with 0.7% nitric acid (HNO₃) overnight to break dormancy (Kabange et al., 2020), followed by incubation at 27 °C for 48 h to initiate germination. Drought tolerance during germination was evaluated under controlled laboratory conditions following the CIMMYT (2004) protocol, a method previously validated in similar studies (Tu 2021). Prior to stress application, we optimized the concentration of potassium chlorate (KClO₃) by testing eight representative varieties (four drought-resistant and four drought-susceptible) across a gradient of 0.1% to 1.0% (in 0.1% increments), with three replicates per treatment. Control seeds were germinated in distilled water. Based on phenotypic responses observed after seven days, 0.3% KClO₃ was selected as the optimal concentration to induce a consistent drought-like stress response and was employed in all subsequent experiments.

For germination assessment, 20 seeds per accession were soaked in 0.3% KClO₃ solution for 48 h, then rinsed thoroughly with neutral water. Seeds were placed on Petri dishes lined with moist filter paper and incubated under controlled conditions. After seven days, germination percentage (GP) was calculated as:

GP (%) = (Number of germinated seeds / Total number of seeds) ×100.

Assessment of Seedling Drought Tolerance

Germinated seeds were transplanted into 50-well trays filled with nutrient-enriched soil and grown under controlled greenhouse conditions (27 ± 2 °C; relative humidity 65–75%) in a greenhouse environment until reaching the three-leaf stage (approximately three weeks old). Uniform seedlings (six per accession) were then subjected to drought stress by immersing their roots in 0.3% KClO₃ solution for five days under shaded conditions (Fig. 1C, D). The solution was refreshed after three days via irrigation to maintain consistent stress levels. Control seedlings were irrigated with distilled water. Phenotypic responses were recorded 3 days after the KClO₃ application. The seedling survival percentage (SP) was calculated as:

Fig. 1.

Drought tolerance and stress response traits in Myanmar rice landraces. A-B Differential responses of sensitive and tolerant varieties to 4 weeks of water withholding, bar = 5 cm. A V3, a sensitive variety, was shown under control conditions (V3C) and after 4 weeks of complete water withholding (V3T). B V5, a tolerant variety, was shown under control conditions (V5C) and after 4 weeks of complete water withholding (V5T). C-D Seedling survival rates of accessions treated with 0.3% KClO₃, bar = 5 cm. C Seedlings under control conditions. D Seedlings after treatment with 0.3% KClO₃. E-F Frequency distribution of drought-tolerance indices for two key drought-related traits under drought stress treatment with 0.3% KClO₃: E Germination percentage under drought stress. F Seedling survival percentage under drought stress. G Comparative survival of rice varieties under KClO₃-induced oxidative stress and drought stress

SP (%) = (Total seedlings - Dead seedlings) / Total seedlings) × 100.

This standardized approach enabled robust phenotypic evaluation of early-stage drought tolerance across diverse landrace accessions, providing critical insights into their adaptive potential under water-limited conditions.

Two-Phase Verification: KClO₃ Solution Screening Followed by Pot-Based Drought Stress Assay

To investigate the potential correlation between KClO₃ tolerance and drought resistance, rice accessions were stratified into five groups, each comprising five varieties, arranged along a gradient from least to most tolerant based on preliminary screening results (Table S12). These groups were cultivated in soil-filled bags within a controlled greenhouse environment to facilitate subsequent drought resistance evaluation (Fig. S2). At the 3- to 4-leaf developmental stage, plants were subjected to a drought stress treatment involving complete water withholding for a maximum duration of 3–4 weeks, while parallel control plants were maintained under optimal irrigation conditions. Each treatment was replicated thrice to ensure statistical robustness. Post-stress phenotyping included quantitative assessments of plant survival rates, leaf rolling, and leaf drying symptoms.

Based on these observations, two contrasting varieties were selected for detailed transcriptomic analysis: V5 (‘Lone Shae Thwe’), originating from the Eastern Mountainous Region (EMR), and V3 (‘Toe Pwa Gyi’), from the Delta Region (DR) (Fig. 1A, B). After 28 days of continuous water deprivation, soil relative water content (RWC) in the cultivation columns decreased to ~ 10–15%, indicating that the plants were subjected to severe drought stress conditions despite some residual surface appearance. Leaf tissues were harvested from these varieties, immediately flash-frozen in liquid nitrogen, and stored at -80 °C. Three biological replicates per variety were prepared to ensure the reliability of RNA extraction and downstream transcriptomic analyses.

Population Structure Analysis

The population structure of all 236 rice accessions was analyzed using ADMIXTURE v1.3.0 (Alexander et al. 2009). Model-based clustering was performed with K values ranging from 2 to 10, with 10 independent runs per K. The optimal number of subpopulations (K) was determined using the Evanno method (Evanno et al. 2005) for cross-validation error in ADMIXTURE. In addition, principal component analysis (PCA) and a neighbor-joining (NJ) tree were conducted using genome-wide SNPs to further assess genetic relationships and admixture among accessions. The population structure information was subsequently incorporated as covariates in GWAS models to control for stratification.

Whole Genome Resequencing and SNP Discovery

Genomic DNA was extracted from leaf tissues using the cetyltrimethylammonium bromide (CTAB) method. Sequencing libraries were prepared utilizing the MGIEasy Universal DNA Library Prep Set (MGI Tech Co. Ltd., China), following the manufacturer’s protocol, which included DNA fragmentation, adapter ligation, and PCR amplification. Libraries passing quality control were sequenced on the DNBSEQ-T7 platform (MGI Tech Co. Ltd.) in paired-end 150 bp mode, achieving an average sequencing depth of approximately 15× per sample. Raw reads were processed and aligned using the Sentieon DNAseq pipeline (v202112.06), with variant calling performed via GATK (DePristo et al., 2011). Stringent quality filtering with VCFtools (v0.1.17) retained bi-allelic SNPs with a minor allele frequency (MAF) ≥ 5% and a missing data rate ≤ 20%, resulting in a high-confidence dataset of 2,647,384 SNPs suitable for association analyses. To further ensure variant reliability, SNPs with a minimum read depth ≥ 5 were retained, and heterozygous calls were excluded, given the predominantly inbred nature of rice accessions. After filtering, ~ 2.62 million high-quality SNPs were retained per genotype (range: 2.51–2.64 million), providing sufficient genome-wide coverage for association analysis.

GWAS and Haplotype-Based Analysis

GWAS were performed utilizing the rMVP package (Yin et al. 2021), following a standardized analytical pipeline. Genotypic data were converted into HapMap format, while phenotypic data were organized as tab-delimited files with additive SNP coding. Three statistical models were employed to ensure robustness: (1) a General Linear Model (GLM) serving as the baseline, (2) a Mixed Linear Model (MLM) incorporating a VanRaden kinship matrix to account for relatedness, and (3) the Fixed and Random Model Circulating Probability Unification (FarmCPU), which iteratively refines fixed and random effects to enhance detection power. Among these, MLM and FarmCPU controlled false positives most effectively, and MLM was selected for the final presentation of results, given its balanced performance. All models incorporated population structure covariates derived from the top five principal components. Model fit and potential inflation were evaluated via quantile-quantile (Q-Q) plots, while association signals were visualized through Manhattan plots depicting -log10(p)-values. The GWAS significance threshold (p < 1 × 10⁻⁵) was selected following commonly applied standards in rice GWAS studies (e.g., Yano et al. 2016; Rastogi et al. 2025), providing a balance between controlling false positives and maintaining detection power given the large SNP dataset. Complete scripts and parameter configurations are publicly accessible at https://github.com/xiaolei -lab/rMVP.

Based on the well-characterized linkage disequilibrium (LD) decay rate of approximately 100–200 kb in cultivated rice (Huang et al. 2010) QTL regions were delineated as 200 kb intervals centered on each significant SNP (p < 1 × 10⁻⁵). Adjacent significant SNPs within this window were merged into contiguous QTL intervals, considering LD structure, with the SNP exhibiting the lowest p-value designated as the lead SNP. Pairwise LD among SNPs was computed and visualized using the LD heatmap package (Shin et al. 2016) in R, focusing on regions harboring GWAS peaks. For haplotype analysis, raw genotypic data, including SNPs and insertions/deletions (Indels), were utilized to maximize variant inclusion. Variants with missing or heterozygous calls were excluded, retaining only those located within coding sequences and 2 kb upstream of gene promoters. The phenotypic effects of haplotypes were statistically evaluated using t-tests to compare trait means across haplotype groups.

Gene Annotation and Functional Prediction

Candidate genes within associated loci were annotated using sequence data from the 9311 (Indica) rice genome, accessed via the SNPseek database (https://snpseek.irri.org). Genomic regions corresponding to significant SNPs were extracted from the 9311-reference genome, and the sequences of proximal genes were used as queries in BLASTp searches against the Nipponbare protein database. Resulting hits were analyzed to infer gene functions based on available annotations from resources such as the Rice Genome Informatics (RGI) platform (https://riceome.hzau.edu.cn/), the Rice Annotation Project Database (https://rapdb.dna.affrc.go.jp/), and the National Center for Biotechnology Information (NCBI; https://www.ncbi.nlm.nih.gov/). This approach aligns with established methodologies for cross-subspecies gene annotation in rice genomics (Wang et al. 2022b; Yu et al. 2002).

RNA-seq Analysis

Total RNA was extracted from leaf tissues of both control and drought-stressed plants representing the V3 (drought-sensitive) and V5 (drought-tolerant) rice varieties, with three biological replicates per treatment. Leaf samples were collected at the peak of drought stress (four weeks post water withholding), immediately flash-frozen in liquid nitrogen, and stored at − 80 °C until further processing. RNA isolation was performed using TRIzol reagent (Invitrogen), and RNA integrity was assessed via agarose gel electrophoresis and quantification using a Nanodrop 2000 spectrophotometer (Thermo Scientific). A total of twelve RNA samples (two genotypes × two conditions × three replicates) were prepared for sequencing. mRNA enrichment was achieved using magnetic bead-based poly(A) selection, followed by fragmentation and reverse transcription into cDNA. Sequencing libraries were constructed with the VAHTS Universal V6 RNA-seq Library Prep Kit (Illumina^®) according to the manufacturer’s instructions. Paired-end sequencing (150 bp reads) was performed on the Illumina NovaSeq X Plus platform, generating high-throughput data. Raw reads were quality-filtered and adapter-trimmed using SOAPnuke. Clean reads were aligned to the Oryza sativa Nipponbare reference genome (IRGSP-1.0) using HISAT2 (Kim et al. 2013). Transcript assembly and quantification were conducted with StringTie (V2.1.2) (Pertea et al. 2015), and gene expression levels were normalized as fragments per kilobase of transcript per million mapped reads (FPKM). Technical replication was not included for RNA-seq, as biological replicates are considered sufficient for transcriptome-wide analyses, whereas technical replicates were incorporated during qRT-PCR validation to confirm the reliability of candidate gene expression patterns.

DEGs were identified using DESeq2 (v1.4.5) (Love et al. 2014), applying a false discovery rate (FDR) threshold of adjusted P ≤ 0.05 and an absolute log₂ fold change ≥ 1.0. Functional annotation and pathway enrichment analyses were performed using Gene Ontology (GO; http://www.geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG; https://www.kegg.jp/) databases, employing the hypergeometric test (Phyper) (https://en.wikipedia.org/wiki/ Hypergeometric_distribution). Multiple testing correction was applied using the Bonferroni method, with a significance cutoff of adjusted P ≤ 0.05 to identify key biological processes and pathways associated with drought tolerance.

Quantitative Real-Time PCR Validation

For qRT-PCR was performed on the same high-quality RNA samples. Total RNA was extracted from rice leaves subjected to drought stress using the Tsingke RNA extraction kit (Tsingke, China). Following DNase I treatment to eliminate genomic DNA contamination, approximately 2 µg of RNA was reverse-transcribed into cDNA using the SynScript^® III RT SuperMix for qPCR (Tsingke), following the manufacturer’s protocol. Random primers were employed for cDNA synthesis. qPCR reactions were carried out using ArtiCanCEO SYBR Green qPCR Mix (Tsingke) on the ABI QuantStudio StepOne Plus Real-Time PCR System (Thermo Fisher Scientific). Primer sequences for target genes and the internal control (OsACT1) are listed in Table S1. Relative gene expression levels were calculated using the 2^–ΔΔCt method, with normalization to OsACT1. All reactions were performed in triplicate to ensure reproducibility.

Statistical Analysis

Phenotypic data were summarized as means ± standard errors, calculated using Microsoft Excel 2016. Graphical representations, including bar plots, were generated with GraphPad Prism v8.4.2. Linkage disequilibrium (LD) heatmaps were generated using LDheatmap R v4.2.2 (Shin et al. 2006), Manhattan plots were generated using the rMVP R package (Yin et al. 2021).

Results

Correlation Between KClO₃ Tolerance and Drought Resistance in Rice

Our analysis of 25 representative rice accessions revealed a significant positive correlation (r = 0.82, p < 0.01) between KClO₃ tolerance and drought survival rates (Fig. 1G). This finding substantiates the utility of KClO₃ sensitivity as a robust phenotypic marker for screening drought resistance. The five pre-stratified groups demonstrated a distinct gradient of drought susceptibility that corresponded closely with their respective KClO₃ tolerance rankings (Table S12). Notably, the group exhibiting the highest tolerance, predominantly comprising accessions from the Eastern Mountainous Region (EMR), including the elite line V5 (‘Lone Shae Thwe’), achieved an average survival rate of 85% under drought stress conditions. Conversely, the most sensitive group, primarily consisting of Delta Region (DR) accessions such as V3 (‘Toe Pwa Gyi’), exhibited markedly lower survival rates, with less than 20%. These results underscore the effectiveness of KClO₃-based phenotyping in capturing underlying genetic variation associated with drought adaptation, thereby offering a valuable tool for germplasm screening and selection in rice breeding programs.

Phenotypic Variation in Drought Response among 236 Rice Accessions

A comprehensive phenotypic assessment was performed on 236 genetically diverse rice accessions to elucidate variation in drought tolerance during early developmental stages. Drought stress was simulated using 0.3% potassium chlorate (KClO₃), applied for 48 h at the germination stage, with germination rates recorded after 7 days. For seedling evaluation, three-week-old plants underwent root immersion in 0.3% KClO₃ solution for 7 days under controlled environmental conditions.

The analysis revealed substantial variability in drought response metrics. GP ranged from 0% to 100%, with a mean of 72.1% ± 19.2%, while SP exhibited a similar range (0-100%) and a mean of 66.3% ± 22.1%. Notably, approximately 18% of accessions demonstrated high tolerance at both stages, characterized by GP and SP values exceeding 80%. Conversely, 22% of accessions were highly susceptible, with GP and SP below 20%. Interestingly, 14% of the accessions exhibited contrasting responses between germination and seedling stages, suggesting potential stage-specific genetic regulation of drought tolerance mechanisms (Fig. 1E, F, and Table S1). Although only two traits (GP and SP) were evaluated, they represent complementary components of early-stage drought tolerance seed establishment and seedling survival. Substantial phenotypic variation across the panel demonstrated that these traits were sufficient to capture genetic diversity for association mapping.

This pronounced phenotypic diversity underscores the utility of this rice population for dissecting the genetic basis of drought resilience. The observed variation provides a robust foundation for GWAS, which holds promise for identifying key genetic loci governing drought-responsive traits during early development. These insights could facilitate the development of drought-tolerant rice cultivars through targeted breeding strategies.

Identification of QTLs for Drought Stress Tolerance at the Germination and Seedling Stage

The 236 Myanmar rice accessions showed distinct genetic stratification, according to STRUCTURE’s population structure analysis (Fig. S3). Many accessions showed admixture from multiple ancestral groups (Fig. S3 A), and distinct subpopulations were found at cluster values (K) ranging from 5 to 8 (Fig S3 B). Since the cross-validation error plot reached a minimum at K = 8, it was concluded that eight subpopulations were the ideal number. In subsequent genetic association analyses, it will be crucial to take into consideration this stratification, as the panel’s complex population structure reflects significant genetic diversity and historical gene flow.

To elucidate the genetic diversity among Myanmar rice landraces, Principal Component Analysis (PCA) was performed, revealing distinct clustering patterns aligned with geographic origin (Fig. 2A). Landraces from the Central Dry Zone (CDZ) and Eastern Mountain Region (EMR) formed tightly knit clusters, indicative of high genetic similarity, whereas those from the Southern Plain and Delta (SPD) and Western Hills Region (WHR) exhibited more dispersed distributions, reflecting greater intra-group variation. These findings underscore the significant influence of geographic origin on the genetic architecture of Myanmar rice. Phylogenetic analysis (Fig. 2B) further refined these relationships, delineating closely related clusters (e.g., red, green, blue) and more genetically diverse groups (e.g., purple, yellow), alongside several outliers with unique genetic signatures. Although all accessions belong to the indica subspecies, the observed genetic structuring likely reflects local adaptation and breeding history rather than deep subpopulation divergence. This geographic stratification emphasizes the importance of origin considerations in germplasm selection for breeding programs. Collectively, these insights contribute to understanding the evolutionary trajectory of Myanmar rice landraces and identify valuable genetic resources for targeted trait improvement.

Fig. 2.

Genome-wide association study (GWAS) of drought-responsive germination and seedling traits in Myanmar rice landraces. A Principal component analysis (PCA) of genetic diversity across geographical origins. B Neighbor-joining phylogenetic tree of 236 traditional rice landraces from diverse agroecological zones of Myanmar. C-D Manhattan plots (left) and quantile-quantile (Q-Q) plots (right) for GWAS of C germination percentage and D seedling survival percentage under drought stress

To investigate the genetic basis of drought tolerance during early developmental stages, a GWAS was conducted using the MLM model, which showed robust performance in controlling population structure and kinship. The analysis identified significant associations between SNPs and drought tolerance traits, visualized through Manhattan plots (Fig. 2C and D). QTLs were defined as genomic regions harboring at least two SNPs surpassing the significance threshold of -log10(p) ≥ 5. A total of twelve QTLs associated with GP and SP were identified, distributed across various chromosomal regions.

In total, 125 significant SNPs were confined within these twelve QTL regions (Table 1). The number of significant SNPs per QTL ranged from 1 to 41, with the most extensive cluster observed at qSL10. Candidate genes within these QTL regions were annotated based on the Nipponbare reference genome, focusing on known or putative genes involved in drought response, osmotic stress tolerance, and key metabolic pathways critical for early-stage survival. This analysis revealed 546 annotated genes across the regions, including 66, 54, 65, 54, 70, 37, 58, 57, 32, 50, 47, and 57 genes within qGP2, qGP4-1, qGP4-2, qGP5, qGP7, qGP8, qGP9, qGP11-1, qGP11-2, qSP2, qSP8, and qSP1, respectively (Fig. 2C-D; Table S2).

Table 1.

Twelve GWAS regions associated with drought at the early vegetative stage

QTL ID.	Trait	Chr.	No. of Significant SNPs	Position Range (bp) in 9311	Position Range (bp) in Nipponbare	Lead SNP	P-value	Previous QTLs/ Genes	References
qGP2	Germination %	2	1	5,825,981- 6,225,981	5,316,388- 5,668,225	6,025,981	2.65 × 10− 6
qGP4-1	Germination %	4	1	23,375,044- 23775,044	23,287,964 − 23,643,594	23,575,044	5.05 × 10⁻6
qGP4-2	Germination %	4	11	32,371,421 − 32,771,421	31,597,884 − 31,907,077	32,571,421	4.33 × 10⁻6
qGP5	Germination %	5	14	18,616,997 − 19,016,997	18,200,747 − 18,613,251	18,816,997	2.53 × 10− 6	OsANN9 OsANN4 AWPM-19	Jia et al. (2023) Zhang et al. (2021) Chen et al. (2015)
qGP7	Germination %	7	14	18,364,601 − 18,764,601	19,598,507- 2,008,1621	18,564,601	8.52 × 10− 7
qGP8	Germination %	8	25	16,785,023 − 17,185,023	15,115,420 − 15,512,115	16,985,023	4.55 × 10− 7
qGP9	Germination %	9	5	18,902,933 − 19,302,933	19,570,053 − 19,914,481	19,102,933	1.47 × 10⁻6	Os9BGlu33	Ren et al. (2019)
qGP11-1	Germination %	11	6	21,291,724 − 21,691,724	20,098,066 − 20,538,142	21,491,909	4.02 × 10⁻6
qGP11-2	Germination %	11	2	26,425,691 − 26,825,691	24,911,501 − 25,222,357	26,625,720	6.23 × 10⁻6	OsU2AF65A OsLPAT2	Lu et al. (2021) Shaikh et al. (2022)
qSP2	Seedling%	2	2	6,403,462- 6,803,462	5,835,517- 6,219,187	6,603,462	2.65 × 10− 6	qtaro_11	Lafitte (2004)
qSP8	Seedling%	8	3	3,039,177- 3,439,177	3,082,738- 3,424,713	3,239,177	3.45 × 10⁻6	OsCCA1	Wei et al. (2021)
qSP10	Seedling%	10	41	21,480,865 − 21,880,865	20,270,951 − 20,654,875	21,680,865	1.46 × 10− 8	OsGSTU30	Srivastava et al. (2019)

In the qGP2 locus, Os02g0195600, which encodes an A20/AN1-type zinc finger protein, was identified. This gene has been previously implicated in drought response through its role in modulating abscisic acid (ABA) biosynthesis and gibberellin (GA) signaling pathway (Zhang et al. 2016). Within the qGP4-1 region, Os04g0469800 (CYP724B1), a cytochrome P450 enzyme involved in brassinosteroid biosynthesis, was detected. Given the critical functions of brassinosteroids in regulating plant development, such as stem elongation and stress responses, CYP724B1 may contribute to drought tolerance mechanisms (Miao et al., 2024). The qGP4-2 interval contained two key drought-responsive transcription factors (TFs): Os04g0637000, a TGA-type bZIP TF involved in redox signaling and defense responses (Liang et al. 2021), and Os04g0641700, a basic helix-loop-helix (bHLH) TF that modulates brassinosteroid signaling and maintains a balance between growth and stress adaptation (Gu et al. 2021).

In the qGP5 region, five candidate genes were identified: Os05g0381400, an AWPM-19-like protein associated with ABA-dependent stress signaling (Chen et al. 2015); Os05g0382600 and Os05g0382900, both belonging to the annexin family, which are involved in reactive oxygen species (ROS) modulation and ABA-induced calcium flux (Zhang et al. 2021; Jia et al. 2023); and Os05g0384600, an ATP-binding cassette (ABC) transporter-like protein potentially mediating the transport of stress-related metabolites (Nguyen et al. 2014). Within qGP8, two candidate genes were identified: Os08g0342300, encoding a brassinosteroid receptor kinase implicated in drought-responsive root growth (Fàbregas et al. 2018), and Os08g0342400, a chloroplast-localized aspartate kinase–homoserine dehydrogenase involved in amino acid biosynthesis and osmotic adjustment under water deficit conditions (Paris et al. 2003).

The qGP9 region harbored Os09g0511900, a GH1 family β-glucosidase, which may enhance drought tolerance by hydrolyzing ABA-glucose conjugates to activate ABA signaling and facilitate osmoprotectant mobilization, thereby promoting root elongation and seed germination (Luang et al. 2013). The qGP11-2 interval contained Os11g0636900, encoding a U2 auxiliary factor (U2AF) splicing factor involved in maintaining RNA processing fidelity under abiotic stress conditions (Butt et al. 2022), and Os11g0637800, an LPAAT2 enzyme implicated in ABA-mediated membrane remodeling and osmotic stress tolerance (Shaikh et al. 2022). These genes are likely to act synergistically in regulating transcriptional and cellular responses to drought stress.

In the qSL2 region, Os02g0208300, encoding an ABC-2 type transporter (Nguyen et al. 2014), was identified as a promising candidate. Within qSP8, Os08g0157600, a LHY-like MYB TF, was found to regulate circadian rhythms, stomatal closure, ROS homeostasis, and ABA signaling pathways in response to drought stress, exhibiting functional similarity to AtLHY in Arabidopsis thaliana (Wei et al. 2022). Lastly, in the qSL10 region, Os10g0529500, a tau-class glutathione S-transferase known for its roles in detoxification and drought tolerance (Li et al. 2023; Srivastava et al. 2019) was identified. No candidate genes with strong functional annotations were detected within the qGP7 region based on current gene models.

Analysis of Key SNP Loci Within GWAS Peaks Associated with Drought Tolerance at the Early Vegetative Stage

To elucidate the genetic architecture underlying drought tolerance during the early vegetative phase, we conducted a comprehensive analysis of SNP loci within the most significant GWAS peaks linked to germination and seedling survival traits. Out of twelve identified QTL regions, three demonstrated particularly robust SNP-trait associations and were subjected to detailed haplotype characterization.

On chromosome 5, a prominent QTL spanning 18.8–18.9 Mb was identified, encompassing 14 significant SNPs correlated with GP under drought stress (Fig. 3A). Haplotype analysis of six high-linkage disequilibrium (LD) SNPs (r² = 0.6-1.0) revealed two distinct haplotypes: HapA (n = 156) and HapB (n = 80) (Fig. 3B). Individuals harboring HapA exhibited significantly higher germination rates compared to those with HapB (p < 0.0001), accounting for approximately 23.7% of the phenotypic variance within the population (Fig. 3C).

Fig. 3.

Analysis of the SNP peak and candidate genes on chromosomes 5, 7, and 10. A, D, G Local Manhattan plot (top) and LD heatmap (bottom) for the qGP5, qGP7, and qSP10 loci across the entire population. The blue line indicates the significance threshold (-log₁₀ (p) = 5), and the orange line marks the strongest SNP peak. The LD heatmap (bottom) displays pairwise r² values between SNPs on chromosomes. B, E, H Haplotype patterns of rice accessions based on the number of SNPs within the qGP5, qGP7, and qPS10 regions. C, F, I Phenotypic comparisons between major haplotypes: (C) Hap A (n = 154) shows significantly higher germination percentage (GP) than Hap B (n = 80) (*****p* < 0.0001, Student’s *t-test). F Hap A (n = 196) has significantly higher GP than Hap B (n = 40) (*****p* < 0.0001). I Hap A (n = 214) exhibits greater seedling survival percentage (SP) than Hap B (n = 22) (*****p* < 0.0001)

A second major association peak was identified on chromosome 7, between 18.55 and 18.58 Mb (Fig. 3D), where 14 significant SNPs were detected. Three consistently associated SNPs within this region formed a strong LD block (r² > 0.8) (Fig. 3E). Haplotype analysis delineated two predominant groups: HapA (n = 196), associated with high germination rates (median 85%), and HapB (n = 40), associated with markedly reduced germination (25%) under drought conditions (Fig. 3F). The pronounced phenotypic divergence between these haplotypes (p < 0.0001) underscores this region’s potential role as a regulatory hub influencing drought-induced germination responses.

Regarding seedling survival, a highly significant SNP cluster was identified on chromosome 10, spanning 21.65–21.72 Mb, with 41 SNPs surpassing the significance threshold (Fig. 3G). Haplotype analysis of six SNPs in strong LD (r² = 0.6-1.0) revealed two major haplotypes: HapA (n = 214) and HapB (n = 22) (Fig. 3H). Accessions carrying HapA demonstrated significantly higher seedling survival rates than those with HapB (p < 0.0001) (Fig. 3I). The low frequency of HapB (~ 9.3%) suggests it may represent a less adaptive allele under drought stress.

Collectively, these findings identify three critical genomic regions on chromosomes 5, 7, and 10 that likely harbor key regulatory elements or candidate genes conferring early-stage drought tolerance in Myanmar rice landraces. These loci provide promising targets for marker-assisted selection and functional validation aimed at enhancing drought resilience.

DEGs Analysis of Drought Stress During Germination and Seedling Stages

To elucidate the molecular mechanisms underlying drought response and identify key DEGs involved in drought tolerance, we performed comprehensive RNA-seq transcriptome profiling of two rice cultivars with contrasting drought sensitivities: V3 (highly sensitive, from Delta Region) and V5 (superior tolerant, originating Eastern Mountainous Region). Principal component analysis (PCA) demonstrated that the first two principal components (PC1 and PC2) accounted for 73.5% of the total variance (PC1: 52.6%; PC2: 20.9%). PC1 distinctly separated the two cultivars, whereas PC2 differentiated drought-treated samples from controls, indicating robust transcriptomic divergence associated with genotype and treatment. Biological replicates clustered tightly, confirming data reproducibility and consistency (Fig. 4A). Correlation analyses further supported high reproducibility among replicates within each group (Fig. 4B). Overall, the sequencing data quality satisfied the standards necessary for downstream analyses.

Fig. 4.

Transcriptomic analysis of drought response in V3 (susceptible) and V5 (tolerant) varieties. A Principal component analysis (PCA) of 12 samples. B Pairwise correlation matrix; values (range: 0–1) indicate the strength of positive correlations. Note: C = control, T = drought-treated

Gene expression levels were quantified using FPKM (Fragments Per Kilobase of transcript per Million mapped reads). DEGs were identified based on thresholds of |log₂FC| ≥ 1 and an adjusted p-value < 0.05 under drought stress versus control conditions. In V3 (V3T vs. V3C), a total of 3,476 DEGs were detected, comprising 1,776 upregulated and 1,700 downregulated genes. Conversely, V5 (V5T vs. V5C) exhibited 2,590 DEGs, with 1,297 upregulated and 1,293 downregulated (Fig. 5A). The comparatively lower number of DEGs in V5 suggests a more resilient or less transcriptionally responsive phenotype to drought stress. A Venn diagram illustrated the overlap of DEGs between the two cultivars, revealing 498 shared DEGs, including 239 upregulated and 132 downregulated genes (Fig. 5B-D).

Fig. 5.

The transcriptome analysis of two rice accessions differing in their tolerance to drought stress. A Number of DEGs in (V3T vs. V3C) and (V5T vs. V5C). B-D Venn diagrams of DEGs in (V3T vs. V3C) and (V5T vs. V5C); B total DEGs, C upregulated DEGs, and D downregulated DEGs. GO enrichment analysis of differentially expressed genes (DEGs) for V3T vs. V3C (E) and V5T vs. V5C (F). KEGG enrichment analysis of differentially expressed genes (DEGs) for V3T vs. V3C (G) and V5T vs. V5C (H)

Further functional annotation of DEGs focused on their potential roles in drought tolerance mechanisms. Using gene annotations from the Rice Annotation Project Database (https://rapdb.dna.affrc.go.jp/), we identified ten DEGs with established associations to drought response (Tables S3 and S4). These included TFs such as a bZIP (Os09g0456200), NF-YB8 (Os03g0413000), and NAC TF (Os11g0127600), as well as functional proteins like Annexin (Os05g0382900), D-alanine ligase B (Os07g0691200), PYL ABA receptor (Os06g0562200), a 14-3-3-like protein (Os11g0609600), an AWPM-19-like protein (Os05g0381400), a CYP450 enzyme involved in BR biosynthesis (Os04g0469800), and another ABA receptor (Os02g0255500).

GO enrichment analysis was conducted across three categories: cellular component (CC), biological process (BP), and molecular function (MF). The top 21 enriched GO terms for DEGs in V3T vs. V3C and V5T vs. V5C are summarized in Figs. 5E-F and Tables S5-S6. In V3, 15 GO terms within the BP category were significantly enriched, notably GO:0006952 (defense response), GO:0071456 (cellular response to hypoxia), and GO:0009753 (response to jasmonic acid), involving 986, 612, and 493 DEGs, respectively (Fig. 5E, Table S5). These terms predominantly comprised upregulated genes, indicating activation of stress-responsive pathways. In the CC category, prominent GO terms included GO:0009507 (chloroplast), GO:0000325 (plant-type vacuole), and GO:0009941 (chloroplast envelope), with 1,030, 776, and 397 DEGs, respectively. In the MF category, GO:0015293 (symporter activity) was notably enriched (Fig. 5E).

In V5, seven GO terms within the BP category were significantly enriched, primarily related to photosynthesis and chloroplast functions, such as GO:0015979 (photosynthesis), GO:0009658 (chloroplast organization), and GO:0031425 (chloroplast RNA processing), with 130, 262, and 114 DEGs respectively (Fig. 5F and Table S6). Most of these genes were downregulated under drought stress (Table S4). The CC category revealed enrichment in 14 GO terms, including GO:0009507 (chloroplast), GO:0009570 (chloroplast stroma), and GO:0009534 (chloroplast thylakoid), with 1,390, 554, and 275 DEGs, respectively.

KEGG pathway enrichment analysis highlighted divergent drought response strategies between the two cultivars. In the sensitive V3 line, DEGs were predominantly enriched in pathways such as starch and sucrose metabolism (ko00500), photosynthesis (ko00195), and gibberellin biosynthesis (ko00904) (Fig. 5G and Table S7). Conversely, the tolerant V5 line showed stronger enrichment in secondary metabolic pathways, notably alpha-linolenic acid metabolism (ko00592), along with photosynthesis-related pathways (ko00195, ko00196) (Fig. 5H and Table S8). Both cultivars shared enrichment in broad metabolic pathways (ko01100) and secondary metabolite biosynthesis (ko01110), although the V5 response involved a greater number of genes (e.g., 340 vs. 460 in metabolic pathways; 500 vs. 310 in secondary metabolites). These findings suggest that drought tolerance involves distinct molecular strategies: the sensitive cultivar primarily modulates primary metabolism, whereas the tolerant cultivar activates secondary metabolic pathways and photosynthetic adaptation mechanisms.

Among the DEGs, 146 TFs were identified in V3, and 101 in V5, spanning 16 TF families (Table S9). These TFs are key regulators orchestrating stress responses. The most abundant families included MYB, WRKY, AP2/ERF, bHLH, NAC, bZIP, Homeobox, and MADS. Notably, 27 TFs were differentially expressed in both cultivars under drought conditions, indicating shared regulatory components. Collectively, these results underscore the complexity and diversity of transcriptional regulatory networks underpinning drought tolerance in rice.

Candidate Gene Identification Through Integration of GWAS and RNA-Seq Analyses

To refine the list of candidate genes associated with drought tolerance, we employed an integrative approach combining GWAS with transcriptomic profiling via RNA sequencing. This analysis revealed a total of 103 DEGs located within GWAS-identified regions, spanning twelve QTLs: qGP2 (19 DEGs), qGP4-1 (10 DEGs), qGP4-2 (10 DEGs), qGP5-1 (5 DEGs), qGP7-1 (6 DEGs), qGP8-1 (2 DEGs), qGP9 (10 DEGs), qGP11-1 (10 DEGs), qGP11-2 (5 DEGs), qSP2-1 (6 DEGs), qSP8 (10 DEGs), and qSP10-1 (9 DEGs) (Table S10).

Among these, seven genes exhibited strong associations with drought tolerance in the tolerant genotype V5. Notably, the ABA-responsive gene Os05g0381400 demonstrated significant upregulation (log₂FC = 5.0), indicative of activation of ABA-mediated stress signaling pathways, consistent with its proposed role in drought adaptation to AWPM-19-like mechanisms (Chen et al. 2015). The autophagy-related gene Os07g0513000 was markedly upregulated (log₂FC = 3.6), suggesting enhanced proteostasis and cellular recycling under stress conditions, a critical survival strategy (Gayen et al. 2019).

Os07g0691200, encoding D-alanine ligase B, showed increased expression (log₂FC = 2.8), potentially contributing to cell wall remodeling or osmolyte biosynthesis, although its specific function in plants warrants further investigation (Kolukisaoglu 2020). Two genes involved in protein homeostasis, Os04g0639300 (peptidase M41-like; log₂FC = 4.0) and Os09g0506450 (F-box domain protein; log₂FC = 5.7), are likely mediators of stress-induced proteolysis, with the F-box protein potentially targeting damaged proteins for ubiquitination and degradation (Zhang et al. 2019). Additionally, Os08g0154225, encoding a DEAD-box RNA helicase, exhibited upregulation (log₂FC = 3.2), suggesting a role in modulating stress-responsive RNA metabolism, consistent with homologous functions observed in rice (Ru et al. 2021).

Although Os10g0529500, belonging to the glutathione S-transferase (GST) family, was not significantly differentially expressed in our RNA-seq dataset (log₂FC = 1.7 in V3; not significant in V5), its established role in ROS detoxification under drought (Srivastava et al. 2019) justified its inclusion, implying potential post-transcriptional regulation or genotype-specific responses.

Validation of Candidate Genes Via qRT-PCR

To corroborate the RNA-seq findings, seven drought-responsive genes were selected for qRT-PCR analysis in both drought-tolerant (V5) and sensitive (V3) genotypes under control and drought conditions (Fig. 8). Four of these genes, Os07g0513000, Os07g0691200, Os05g0381400, and Os10g0529500, were identified within GWAS-QTL regions and exhibited consistent differential expression patterns, supporting their roles in drought response. The remaining three genes Os05g0382600 (OsANN4), Os05g0382900 (OsANN9), and Os03g0413000 (OsNF-YB8) were selected based on prior evidence of stress-related functions and expression profiles.

In drought conditions, Os07g0691200 and Os07g0513000 were significantly upregulated in V5 (Fig. 6A-B), reinforcing their potential contribution to drought tolerance. Os03g0413000 also showed V5-specific induction (Fig. 6C), aligning with its proposed involvement in ABA signaling pathways. Conversely, Os05g0381400 was markedly induced in the sensitive genotype V3 (Fig. 6D), suggesting its role as an early stress marker rather than a tolerance factor. Os10g0529500 and Os05g0382600 exhibited higher expression levels in V5 under drought conditions (Fig. 6E), although RNA-seq data indicated a higher baseline expression of Os10g0529500 in V3 (Fig. 6G). Notably, Os05g0382900 was downregulated in both genotypes under drought stress but maintained higher expression in V3 under control conditions (Fig. 6F), implying a complex regulation pattern associated with stress sensitivity.

Fig. 6.

Relative expression of the seven DEGs obtained by qRT-PCR and RNA-seq. The red color bar represents gene expression determined by qRT-PCR, while the green color bar represents gene expression determined by RNA-seq in FPKM value

Discussion

Drought stress during critical developmental stages, such as germination and early seedling establishment, exerts profound effects on rice yield and grain quality, emphasizing the necessity of elucidating the genetic mechanisms underpinning drought tolerance at these pivotal phases (Nyasulu et al. 2024). The early vegetative stage is particularly vulnerable to water deficit, rendering it a focal point for functional genomics studies aimed at enhancing drought resilience. In this investigation, we explored the genetic and molecular basis of drought tolerance in Myanmar local rice varieties during the early vegetative stage, thereby providing valuable insights into their genetic architecture and identifying promising candidate genes for breeding programs targeting climate resilience.

GWAS Uncovers Drought-Responsive QTLs in Myanmar Landraces

GWAS has established itself as a robust approach for dissecting the complex genetic basis of agronomic traits in rice (Oryza sativa L.), including drought tolerance. Recent applications have successfully identified candidate loci associated with drought response across various developmental stages, from germination (Nyasulu et al. 2024) to vegetative growth (Hoang et al. 2019). Notably, studies on Vietnamese landraces and japonica subspecies utilizing high-density SNP arrays have uncovered novel QTLs, some of which have been validated through linkage mapping, thereby advancing marker-assisted selection (Liu et al. 2024).

In our study, GWAS identified twelve QTL regions significantly associated with GP and SP under drought conditions. Among these, five QTLs were prioritized based on SNP significance and functional annotation, encompassing 169 candidate genes (Table S2). Haplotype analysis at peak SNPs revealed pronounced genetic differentiation between drought-tolerant and susceptible accessions, particularly within major QTLs on chromosomes 5, 7, and 10 (Fig. 3). Notably, two predominant haplotypes were observed, with Hap A consistently conferring superior drought performance. For instance, in the qGP7 region, Hap A accessions exhibited over threefold higher GP compared to Hap B, indicating the presence of favorable alleles for early-stage drought adaptation. These stable haplotype-phenotype associations underscore the potential of these loci as targets for marker-assisted selection in breeding resilient rice cultivars.

Candidate Genes and Functional Insights

Among the key QTLs, qSP10 harbored 41 significant SNPs, the highest number identified in this study (Table 1), and was linked to Os10g0529500 (OsGSTU30) a tau-class glutathione S-transferase previously implicated in ROS detoxification and stress signaling pathways (Srivastava et al. 2019). This gene likely plays a crucial role in mitigating oxidative damage during drought stress.

Similarly, qGP5 contained multiple candidate genes associated with drought response, including Os05g0382900 (OsANN9) and Os05g0381400, which encode AWPM-19-like protein involved in ABA signaling during germination (Jia et al. 2023; Chen et al. 2015). Additional genes such as Os05g0382600 (OsANN4) and Os05g0384600 (an ABC transporter-like protein) may contribute to calcium signaling and metabolite transport under water deficit conditions (Xie et al. 2024; Dahuja et al. 2021). Other promising regions, such as qGP8 and qGP2 (Table S2), harbor genes like Os02g0208300 (PDR7), an ABCG transporter involved in ABA transport and cuticle formation (Kuromori et al. 2010; Kang et al. 2010), and Os08g0342400, encoding aspartate kinase-homoserine dehydrogenase (AK-HD1), which may support osmotic adjustment by maintaining amino acid biosynthesis (Trovato et al. 2021; Kishor et al. 2020).

Implications for Breeding and Future Directions

The consistent association of favorable haplotypes with enhanced drought tolerance traits highlights their potential utility in marker-assisted breeding strategies. The identification of these key loci and candidate genes provides a valuable foundation for subsequent fine-mapping, functional validation, and the development of molecular markers tailored for drought resilience. Given the increasing frequency and severity of drought events due to climate change, leveraging the genetic diversity present in Myanmar local rice germplasm offers a promising pathway toward cultivating climate-resilient rice varieties. Future research should focus on functional characterization of these candidate genes and their regulatory networks to facilitate the deployment of molecular tools in breeding programs aimed at ensuring food security under water-limited conditions.

Transcriptomic Insights into Early Vegetative Drought Responses in Myanmar Rice Varieties

Our comprehensive transcriptomic analysis delineates distinct molecular strategies employed by rice genotypes with contrasting drought tolerance during the early vegetative stage. The drought-tolerant genotype V5 exhibited a more restrained transcriptional response, with 2,590 DEGs, compared to 3,476 DEGs in the sensitive genotype V3 (Fig. 5A). This differential DEG count suggests that V5 maintains a more stable transcriptome under water deficit, potentially through constitutive or epigenetically mediated tolerance mechanisms, aligning with previous findings on stress preconditioning and resilience (Panda et al. 2021; Shinozaki and Yamaguchi-Shinozaki 2006). PCA confirmed clear separation between genotypes and treatment conditions, with high reproducibility across biological replicates (Fig. 4A). Notably, the genotypes originate from ecologically distinct regions within Myanmar, implying that regional adaptation influences drought response strategies. Landraces from upland regions such as the EMR may have evolved inherent drought resilience, whereas lowland genotypes like DR are more susceptible, underscoring the importance of regional germplasm in breeding programs aimed at climate resilience.

A core set of 498 DEGs was shared between V3 and V5, representing a conserved drought response module consistent with established stress signaling pathways (Nakashima et al. 2014; Shilpa et al. 2024). However, the presence of genotype-specific DEGs indicates tailored adaptive mechanisms, with V5 demonstrating more targeted and efficient transcriptional modulation. GO enrichment analysis revealed contrasting strategies: V3 predominantly upregulated genes associated with defense responses, hypoxia, and jasmonic acid signaling (Fig. 5E) (GO:0006952, GO:0071456, GO:0009753), alongside chloroplast and ion transporter-related processes, reflecting an active stress response aimed at maintaining cellular homeostasis (Wang et al. 2021; Raza et al. 2021; Li et al. 2021). Conversely, V5 downregulated photosynthesis-related processes (GO:0015979, GO:0009658) and enriched chloroplast organization pathways (GO:0009570, GO:0009534), indicative of an energy-conserving, pre-adaptive response that minimizes metabolic expenditure during stress (Yang et al. 2021; Panda et al. 2021). These divergent strategies highlight potential molecular targets, such as chloroplast function and stress signaling pathways, for breeding drought-resilient rice varieties (Tester and Langridge 2010; Nakashima et al. 2014).

KEGG pathway analysis further elucidates the contrasting metabolic reprogramming between genotypes. V3 predominantly activated primary metabolic and hormonal pathways, including starch and sucrose metabolism and gibberellin biosynthesis (Fig. 5G), pathways often associated with stress sensitivity and rapid growth responses (Guo et al. 2018; Colebrook et al. 2014). In contrast, V5 preferentially upregulated secondary metabolic pathways, notably phenylpropanoid biosynthesis and α-linolenic acid metabolism (Fig. 5H), which are implicated in cellular protection, antioxidant activity, and jasmonate-mediated stress responses (Bartwal et al. 2013; Isah 2019; Zi et al. 2022). Additionally, V5 showed enrichment in light-harvesting complex proteins and photosystem antenna components, suggesting an adaptive modulation of photosynthetic efficiency under drought conditions (Borisova-Mubarakshina et al. 2022). Collectively, these findings underscore V5’s strategic deployment of conserved transcriptional reprogramming, emphasizing metabolite-mediated defense and photosynthetic resilience traits that are promising targets for molecular breeding initiatives aimed at enhancing drought tolerance.

Integration of GWAS and Transcriptomic Analysis Identifies Robust Candidate Genes for Drought Tolerance in Rice

The synergistic application of GWAS and transcriptomic profiling has proven to be a transformative strategy for elucidating the genetic basis of complex traits such as drought tolerance. In this investigation, we integrated GWAS-derived QTLs with RNA-seq data to pinpoint candidate genes modulating early vegetative drought response in rice. Our analysis identified 103 DEGs within twelve GWAS-identified QTL regions (Table S10). From these, seven genes were prioritized based on their expression magnitude, functional annotations, and biological relevance to drought stress pathways.

Notably, two genes on chromosome 7, Os07g0513000 and Os07g0691200, emerged as compelling candidates. Os07g0513000, encoding an ATP synthase gamma chain, was markedly upregulated in the drought-tolerant genotype V5. This gene likely plays a pivotal role in sustaining cellular ATP levels during water-deficient conditions, thereby supporting energy-dependent stress responses. Its functional importance is corroborated by prior studies demonstrating increased ATP synthase subunit expression under dehydration stress in rice (Gayen et al. 2019) and recent evidence implicating ATP synthase in mitigating oxidative damage and maintaining energy homeostasis during drought (Khan et al. 2024).

Similarly, Os07g0691200, encoding D-alanine ligase B, exhibited significant upregulation in the tolerant genotype. Although its precise function in plants remains to be fully elucidated, emerging research suggests a role in osmotic adjustment or cell wall fortification during drought stress, aligning with findings on D-amino acid metabolism’s involvement in plant stress resilience (Kolukisaoglu and Suarez 2017; Kolukisaoglu 2020). The strong genetic association of these genes with QTLs on chromosome 7, validated through qRT-PCR (Fig. 6A, B), underscores their potential as targets for functional validation and molecular breeding.

Additional candidate genes contribute to the multifaceted drought response. Os05g0381400, encoding an AWPM-19-like protein associated with ABA signaling, was specifically induced in the drought-sensitive genotype, suggesting a role in early drought perception rather than long-term tolerance (Chen et al. 2015). Os04g0639300, a peptidase M41-like gene, and Os09g0506450, an F-box domain protein, are likely involved in proteostasis through stress-induced protein turnover and ubiquitin-mediated degradation, mechanisms critical for cellular homeostasis under drought conditions (Zhang et al. 2019). Os08g0154225, a DEAD-box RNA helicase, may enhance transcript stability and processing efficiency during stress, consistent with reports linking similar helicases to improved drought resilience in rice (Ru et al. 2021). Lastly, Os10g0529500 (OsGSTU30), a tau-class glutathione S-transferase, plays a well-established role in reactive oxygen species (ROS) detoxification and redox homeostasis, reinforcing its relevance despite genotype-specific expression patterns (Srivastava et al. 2019; Hernández Estévez and Rodríguez Hernández 2020).

From the integrated analysis, four genes such as Os07g0513000, Os07g0691200, Os05g0381400, and Os10g0529500, were selected for validation via qRT-PCR, confirming the consistency of RNA-seq data (Fig. 6A, B, D, G). The pronounced upregulation of Os07g0513000 and Os07g0691200 in the drought-tolerant genotype highlights their potential as key regulators of drought resilience, particularly given their localization within the most significant QTL on chromosome 7.

KEGG annotation links Os07g0513000 to metabolic and photosynthesis-related pathways, specifically encoding a chloroplast-localized F-type H+-transporting ATPase gamma subunit involved in energy transduction during stress adaptation. Meanwhile, Os07g0691200 shows similarity to D-alanine–D-alanine ligase B, associated with peptidoglycan biosynthesis and potentially contributing to cell wall modification under drought stress. These functional insights reinforce their roles as novel drought-responsive candidates from chromosome 7.

To elucidate their functional roles, we have prioritized CRISPR/Cas9-mediated genome editing of Os07g0513000 and Os07g0691200. These genes were selected based on their strong association with drought-tolerance QTLs, robust expression profiles in tolerant genotypes, and their biological plausibility in stress adaptation pathways. Future studies will involve generating knockout and edited lines, followed by comprehensive phenotypic and molecular assessments under drought conditions. These efforts aim to validate their roles in conferring drought tolerance and to facilitate the development of resilient rice cultivars through molecular breeding strategies.

Comparison of Drought Tolerance Mechanisms in Myanmar Landraces with Previous Studies

This study provides compelling evidence that Myanmar rice landraces harbor both conserved and unique genetic determinants of drought tolerance, underscoring their potential as valuable genetic resources for rice improvement. Twelve significant QTLs were identified across chromosomes 2, 4, 5, 7, 8, 9, 10, and 11. Notably, several of these loci overlap with previously reported drought-responsive regions in diverse rice populations, aligning with findings from recent meta-analyses and high-resolution genetic mapping studies (Baisakh et al. 2020; Selamat and Nadarajah 2021; Jiao et al. 2024). Of particular interest is the major QTL on chromosome 7, which corresponds to well-characterized drought-responsive regions influencing leaf morphology and osmotic regulation (Yi et al. 2023).

Functional annotation of candidate genes within these QTLs revealed significant enrichment for ABA signaling pathways and osmotic adjustment mechanisms, reaffirming the central role of these conserved pathways in drought response. These results are consistent with established literature, where genes such as OsCPK9 modulate drought tolerance through stomatal regulation and osmotic homeostasis (Hassan et al. 2023), and ABA receptor genes orchestrate stress signaling cascades (Bhatnagar et al. 2020). Moreover, transcriptomic analyses identified two notably upregulated candidate genes, Os07g0513000 (ATP synthase gamma chain) and Os07g0691200 (D-alanine ligase), in drought-tolerant landraces. These genes have not been widely associated with drought tolerance in indica or japonica varieties, suggesting Myanmar-specific adaptive mechanisms.

Population structure analyses further highlight the genetic divergence of Myanmar landraces from other global rice germplasm, reflecting distinct signatures of local adaptation. These findings emphasize the importance of conserving and harnessing these regional landraces as reservoirs of novel alleles, which can complement existing breeding efforts. Incorporating such unique genetic resources is vital for developing climate-resilient rice varieties capable of withstanding increasing drought stress, particularly in rainfed and marginal environments (Jiao et al. 2024; Hassan et al. 2023).

Limitations of the Study

Our study relied on KClO₃-induced stress as a surrogate for drought screening. Although this method is precedented (CIMMYT 2024; Tu et al., 2021) and successfully revealed genetic loci, it primarily captures oxidative and nitrogen stress responses. A more comprehensive physiological validation, including comparative analysis with PEG-based osmotic stress and field-based drought trials, would be a valuable focus for future research.

While this study successfully identified drought-responsive loci and candidate genes by integrating GWAS and transcriptomic analyses, certain limitations should be acknowledged. First, the study focused on two representative genotypes, V3 (susceptible) and V5 (tolerant), which provided valuable contrasts but did not capture the full spectrum of drought responses in the panel. Moreover, no bi-parental mapping population derived from V3 × V5 is currently available, which limits functional validation and fine-mapping of the identified QTLs. Future work should prioritize developing such populations or employing gene-editing approaches to confirm the causal roles of candidate genes. Additionally, physiological traits beyond germination percent and seedling survival could complement genetic insights by linking QTLs to specific adaptive mechanisms. These directions will further strengthen the translational potential of our findings.

Conclusion

This study demonstrates that an integrated GWAS and transcriptomic approach effectively elucidates the molecular basis of drought response in Myanmar rice varieties during the early vegetative stage. We identified twelve significant QTL regions associated with germination and seedling survival under drought stress, encompassing 103 DEGs. Among these, seven candidate genes were prioritized based on their pronounced upregulation in drought-tolerant genotypes and their functional relevance to stress adaptation pathways. Notably, Os07g0513000 (encoding an ATP synthase gamma chain) and Os07g0691200 (encoding D-alanine ligase) emerged as the most promising candidates, exhibiting strong upregulation in tolerant lines and high linkage disequilibrium with key QTLs on chromosome 7. This region has been validated as a critical genomic hotspot for drought resilience. These findings offer novel insights into the genetic and molecular mechanisms underpinning rice drought tolerance and provide valuable targets for molecular breeding efforts aimed at developing climate-resilient rice cultivars, thereby contributing to sustainable rice production amid escalating water scarcity challenges worldwide.

Abbreviations

BP

Biological Process

CC

Cellular Component

DEGs

Differentially Expressed Genes

FC

Fold Change

GO

Gene Ontology

GWAS

Genome-Wide Association Study

LD

Linkage Disequilibrium

MF

Molecular Function

PCA

Principal Component Analysis

QTLs

Quantitative Trait Loci

RNA-seq

RNA sequencing

SNP

Single Nucleotide Polymorphism

TF

Transcription Factor

Author Contributions

DS L, LJ C, JC W, CY L and YY Z conceived and designed the experiment. DS L, LJ C, CY L, HH Y and DD L provided methodology. N.Z.N.N, CL W and C Z conducted the experiment, investigated, performed data analysis and wrote the manuscript. Q Z, C Z, JJ L and XY W participated in screening drought material, sample preparation, and data analysis. N.Z.N.N, CL W and Q Z drafted proposals and corrected the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by grants from the Major Science and Technology Projects of Yunnan (grant No. 202202AE09002102 and 202402AE090026), the Central Leading Local Science and Technology Development Project (grant No. 202207AA110010), and the National Natural Science Foundation of China (grant No. 31860108), and a grant from Yunnan Revitalization Talents Support Plan-Xindian High End Foreign Experts Program in China.

Data Availability

All data generated or analyzed during this study are included within this article and its supplementary files. The datasets have been deposited in the NCBI repository under BioProject accession number PRJNA1299379 and are publicly accessible at https://www.ncbi.nlm.nih.gov/search/all/?term=PRJNA1299379.

Declarations

Ethics Approval and Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Competing Interests

The authors declare no competing interests.