Unraveling the Genetic Architecture of Photoperiod Sensitivity in Myanmar Rice Landraces Through Integrated GWAS and Transcriptome Analysis

Abstract

Photoperiod sensitivity (PS) is the major determinant of flowering time in rice and has played a critical role in adaptation across diverse ecotypes. To dissect the genetic and molecular architecture of PS in MYR landraces, we combined GWAS with transcriptomic profiling on 236 diverse accessions. Thirteen major QTL underlying heading date were mapped on chromosomes 1, 2, 3, 6, 7, and 8, consisting of the previously reported flowering genes (OsHd1, OsFTIP9) and a number of novel loci specific to Myanmar germplasm. Comparative RNA-seq analysis, using a photoperiod-sensitive (V10) and a photoperiod-insensitive (V3) indica genotype to the japonica cultivar ‘H479B’ as reference, showed distinct transcriptional reprogramming in response to short-day conditions, with higher-expression plasticity occurring in V10. By integrating GWAS signals with differential expressions, we narrowed our candidate gene set of two high-confidence regulators: Os06g0275000, encoding a zinc finger transcription factor, and Os07g0606600 (NF-YB10). Both genes were highly expressed in a stage-specific manner and further confirmed by qRT-PCR. Our results suggest a complex genetic regulatory network attracting conserved photoperiod pathways with unique novel allelic variant populations in Myanmar landraces. These candidate genes will be potential targets for precision breeding to optimize flowering time and enhance adaptation both in response to climate change and photoperiodic changes.

1. Introduction

Rice (Oryza sativa L.) is a vital stable crop for global food security, sustaining over half of the world’s population. The extensive adaptability is largely regulated by photoperiod sensitivity (PS), a key mechanism that facilitates the alignment of reproductive development with seasonal day length to optimize growth and yield [1,2]. As a facultative short-day (SHD) plant, rice accelerates flowering in short-day conditions and delays it under long-day (LOD) environments, facilitating adaptation across various latitudes and climatic zones [2,3].

Heading date (HD) and flowering time (FTI) are agronomically important traits that are considerably affected by PS [4,5]. The photoperiod-sensitive phase occurs in the leaves 2–3 weeks before panicle initiation, during which photoreceptors perceive day length signals and regulate the expression of floral transition genes [5,6]. Flowering regulation in rice is also influenced by temperature: high temperatures can reduce PS, inducing earlier flowering even under long days, while low temperatures delay the floral transition [6,7]. Most rice varieties have a critical day length requirement of 12–13.5 h, with facultative SHD types being able to flower under both short and long-day conditions [2,3]. Generally, day length is determined by the location’s geographical latitude [8]. To facilitate rice breeding, PS and HD conditions with special short day (7–9 h of daily light) and high temperatures (28–30 °C of daily mean temperature) will enable us to cope with genotypic differences in PS and HD in smart breeding chamber [9,10] to advance the transfer of breeding materials rapidly and achieve quick generation times (at least 3–4 generations annually). In Myanmar, rice landraces cultivated across a wide latitudinal range exhibit significant genetic diversity [11,12]. In PS traits, highland varieties typically show late flowering under long days, and delta landraces exhibit early flowering to align with the monsoon [13,14]. However, the molecular mechanisms and genetic factors that control PS in Myanmar rice remain poorly characterized, limiting their effective use in breeding programs.

Previous molecular studies of rice in Myanmar have mostly employed limited marker systems, including SSR and DArT markers, to investigate genetic diversity and population structure [15,16]. PhyC haplotypes (PHYCHapA/C) have been recently proposed to be candidate regulators of HD adaptation derived from genome-wide studies, reflecting the importance of functional genomics in elucidating PS processes [17]. However, a full understanding of the gene and regulatory networks that control PS remains insufficient. PS in rice is influenced by a complex genetic control architecture, including approximately 600 quantitative trait loci (QTLs), as well as many important regulatory genes, such as Hd1, Ghd7, Ehd1, Hd3a, and RFT1, which serve as the major parts of photoperiodic and florigen signaling pathways [18,19,20].

Genome-wide association studies (GWASs) have a high-resolution power of identifying loci associated with complex traits using natural genetic variation across multiple populations [21,22]. At the same time, transcriptome analysis (RNA-seq) allows for a dynamic understanding of the gene expression networks in response to environmental stressors, including changes in photoperiod [23]. Integrating GWASs with transcriptomics enhances the discovery of functional genes, regulatory pathways, and gene networks related to complex quantitative traits. Integrative investigations have successfully identified genetic determinants of heading and flowering time, stress tolerance, and yield-related traits in rice, maize, and other crops [24,25].

This study employed a combined GWAS and RNA-seq methodology to clarify the genetic structure of PS in Myanmar rice landraces. Our objectives were to (1) identify genetic loci linked to PS through high-density genotyping and rigorous statistical models, (2) characterize transcriptomic responses under regulated SHD conditions to pinpoint differentially expressed genes (DEGs) and potential regulators, and (3) offer molecular insights to enhance breeding for better adaptability to varying photoperiodic and climatic conditions. This work not only enhances understanding of PS mechanisms in rice but also underscores the genetic and ecological importance of Myanmar landraces as a resource for global rice improvement.

2. Results

2.1. Population Structure and Genetic Diversity of Myanmar Rice Landraces

To elucidate the genetic architecture of Myanmar rice landraces, we performed genome-wide single-nucleotide polymorphism (SNP) analysis on 236 accessions. After applying quality filters, including minor allele frequency (MAF) ≥ 0.05 and missing data ≤ 20%, a total of 2,647,384 high-confidence SNPs were retained and distributed across all 12 chromosomes, providing comprehensive genome coverage for diversity and structure analyses. Chromosome-wide SNP density exhibited notable heterogeneity; chromosome 8 displayed the highest density with extensive contiguous regions of elevated SNP counts, whereas chromosome 3 showed the lowest density, characterized by broad regions of variation (Figure 1A).

Principal component analysis (PCA) and neighbor-joining (NJ) phylogenetic clustering revealed distinct genetic groupings aligned with geographic origins. PCA was conducted to investigate the genetic structure of 236 Myanmar rice landraces. The first three principal components explained 75% of the total genetic variance, with PC1 explaining 45%, PC2 20%, and PC3 10% of the variation. To enhance visualization and interpretation, the PCA results were presented as two two-dimensional plots: PC1 vs. PC2 and PC1 vs. PC3 (Figure 1C,D). In both projections, accessions from different agroecological regions exhibited separation primarily along PC1, which captured the largest proportion of genetic variation. Accessions from the Central Dry Zone (CDZ) and Eastern Mountain Region (EMR) tended to occupy distinct positions relative to other regions. In contrast, accessions from the Southern Plain and Delta (SPD) and Western Hills Region (WHR) were more broadly distributed, indicating greater genetic variability within these regions. The NJ tree corroborated these findings, delineating major clades corresponding to the geographic groups, CDZ (yellow), EMR (green), and the Northern Mountain Region (NMR, blue), with SPD (purple) and WHR (yellow) accessions exhibiting longer branches and internal diversity suggestive of admixture events (Figure 1B and Figure S1). Collectively, these analyses indicate that the geographic origin is associated with population structure among Myanmar rice landraces.

2.2. Subspecies Composition of Myanmar Landraces

To determine subspecies composition, we conducted a genome-wide similarity analysis using SNP data aligned to the reference genomes of O. sativa subspecies ‘Nipponbare’ (japonica) and ‘9311’ (indica). The indica-japonica proportions of each accession were estimated based on 100 kb non-overlapping windows. The majority of landraces exhibited predominant similarity to the indica reference, confirming an overall indica background within the panel. However, a small proportion of accessions exhibited notable japonica components or admixed ancestry, reflecting the long history of germplasm exchange and local adaptation in Myanmar (Figure 2D,E). These patterns provide insight into the genetic diversity of Myanmar rice and offer a genetic basis for GWASs and transcriptomic investigations.

2.3. Phenotypic Variation in Heading Date

HD, a major indicator of PS and flowering regulation, was assessed across three contrasting environments: Yezin, Myanmar (MM), Yuanyang, China (YY), and Xundian, China (XD). The 236 landraces displayed wide phenotypic variation across all sites, confirming substantial genetic diversity and genotype-by-environment interaction (Figures S2 and S3A–C, Table S1).

In MM, HD ranged from 50 to 110 days, with a near-normal distribution centered around 81.5 days (±11.7), reflecting a balanced mix of early and late heading genotypes (Figures S3A and S4A,B). In XD, the distribution was broader, mostly spanning 60 to 160 days, with a mean of 115.3 days (±27.5), which was skewed towards later FTIs (Figure S3B), except for a few non-heading ones (Figure S5). In YY, the HD ranged from 60 to 160 days, with a mean of 104.3 days (±29.4). Most accessions flowered between 80 and 120 days (Figure S3C), except for a few non-heading accessions (Figure S4C,D). These regional differences not only underscore the strong environmental influence on HD but also reveal that Myanmar landraces harbor diverse adaptive alleles responsive to local photoperiod and temperature regimes. The phenotypic diversity observed provides a valuable foundation for dissecting the genetic basis of HD through integrative GWAS and transcriptomic approaches.

According to the heading performance of 236 Myanmar rice landraces across the three experimental regions, five representative photoperiod-insensitive and five photoperiod-sensitive indica varieties were selected for detailed evaluation. The days to heading were measured for these ten indica accessions, photoperiod-insensitive (V1–V5) and photoperiod-sensitive (V6–V10), along with two japonicas control varieties (CK1 and CK2), under SHD (7–9 h light) and control (12–13 h light) conditions (Figure S6).

After applying SHD treatment for 30 days from the 3–4 leaf stage, the photoperiod-insensitive group (V1–V5) headed between 67 and 92 days under SHD and between 82 and 113 days under control conditions, indicating only a slight acceleration of 7–21 days under SHD. In contrast, the photoperiod-sensitive group (V6–V10) displayed a more pronounced response, heading between 51 and 71 days under SHD compared to between 91 and 99 days under control, corresponding to a reduction of 24–44 days. Notably, V10 displayed the most pronounced photoperiodic response, flowering 44 days earlier under SHD (51 vs. 95 days). The japonica control varieties (CK1 and CK2) showed minimal differences between treatments (e.g., CK1: 44 vs. 45 days), confirming their role as photoperiod-insensitive controls (Figure S6).

Based on these results, V3 (photoperiod-insensitive), V10 (photoperiod-sensitive), and CK1 (japonica-insensitive) were selected for transcriptomic analysis as representative genotypes exhibiting contrasting responses to photoperiod (Figure 2A–C). These findings clearly distinguish the two indica groups, with SHD treatment markedly accelerating flowering in the sensitive accessions, while the insensitive varieties and japonica controls remained largely unaffected.

2.4. GWAS Identifies QTLs and Candidate Genes Regulating HD in Rice

To elucidate the genetic architecture underlying variation in heading date (HD) in rice, GWAS was performed using the mixed linear model (MLM) based on phenotypic data from 236 rice accessions evaluated across three environments: MM, XD, and YY. Manhattan plots (Figure 3A–C) illustrated genomic regions showing significant marker-trait associations, with QTLs defined as genomic intervals containing at least two SNPs exceeding the genome-wide significance threshold of −log₁₀(P) ≥ 5. This stringent cut-off was applied to reduce false positives while enabling detection of loci contributing to HD variation.

Across all environments, thirteen HD-related QTLs were identified on multiple chromosomes, indicating a polygenic basis of FTI regulation (

Table 1. Thirteen GWAS regions associated with photoperiod sensitivity in three regions.

QTL ID.	Region	Chr	No. of Significant SNPs	Position Range (bp) in 9311	Position Range (bp) in Nipponbare	Lead SNP	p-Value	Previous QTLs/Genes	References
qPS2-1	MM	2	2	27,619,720–28,019,720	27,118,006–27,494,913	27,819,720	7.18 × 10−6	OsIDD4	[30]
qPS6-1	MM	6	1	9,354,323–9,754,323	9,180,932–9,566,967	9,554,323	1.15 × 10−6	OsHd1	[26]
qPS6-2	MM	6	1	10,129,236–10,529,236	9,938,311–10,380,263	10,329,236	1.59 × 10−6	OsNF-YB9	[27]
qPS6-3	MM	6	1	10,573,332–10,973,332	10,479,281–10,927,028	10,773,332	2.01 × 10−6
qPS6-4	MM	6	3	11,012,834–11,412,834	11,002,119–11,332,676	11,212,834	2.62 × 10−6	OsBBX19 (DTH2) Similar to Hd1(indica)	[28]
qPS6-5	MM	6	2	21,700,399–22,100,399	21,677,543–22,082,490	21,900,399	2.73 × 10−6
qPS7	MM	7	5	24,337,992–24,737,992	24,356,961–26,263,836	24,537,992	1.6 × 10−6
qPS8-1	MM	8	1	21,215,772–21,615,772	19,139,609–19,469,347	21,415,772	1.5 × 10−6
qPS8-2	MM	8	3	27,862,151–28,262,151	25,426,651–25,804,334	28,062,151	1.55 × 10−6
qPS1	XD	1	3	24,502,894–24,902,894	22,776,946–23,123,998	24,702,894	3.3 × 10−6	OsFTIP9	[29]
qPS8-3	XD	8	41	21,525,166–21,925,166	19,427,423–20,007,889	21,725,166	2.42 × 10−6
qPS2-2	YY	2	1	28,864,389–29,264,389	28,390,170–28,784,587	29,064,389	2.66 × 10−6
qPS3	YY	3	1	36,836,537–37,236,537	33,522,519–33,854,736	37,036,537	4.68 × 10−6

). In the MM environment (Figure 3A), eight QTLs were detected on chromosomes 6, 7, and 8, including qPS6-1 to qPS6-5. These regions overlapped with loci harboring genes previously reported to be involved in flowering regulation, such as OsHd1 [26], OsNF-YB9 [27], and OsBBX19/DTH2 [28]. Additional loci (qPS7, qPS8-1, qPS8-2) further support the involvement of multiple genomic regions in photoperiod-related HD variation within the Myanmar landraces. In the XD environment (Figure 3B), significant associations were observed on chromosome 1 (qPS1), co-localizing with OsFTIP9 [29], and on chromosome 8 (qPS8-3), where a cluster of 41 significant SNPs suggested a region with a strong association signal specific to this environment. In the YY environment (Figure 3C), distinct loci on chromosomes 2 and 3 (qPS2, qPS3) were identified, with lead SNPs exceeding the significance threshold and mapping to genomic intervals not previously reported in major flowering time studies, indicating potential population or environment-specific effects.

Within these thirteen QTL regions, a total of 906 annotated genes were identified, with gene counts per locus ranging from 26 to 286. Several genes located within these regions have reported or putative roles related to flowering regulation or developmental timing. For instance, Os01g0587300 within qPS1 encodes an FTI-interacting protein (OsFTIP1-like), which has been implicated in florigen transport and photoperiodic flowering regulation.

In qPS2-1 and qPS2-2, we have candidate genes, such as a C2H2 zinc finger transcription factor (ZF-TF) (Os02g0672100) and an R2R3-type MYB protein (Os02g0695200), both of which belong to gene families previously associated with photoperiodic flowering pathways. Additionally, Os02g0698800, encoding a WRKY transcription factor, suggests a link between stress responses and flowering regulation. Within qPS3, Os03g0808600 encodes a calcium-dependent protein kinase, which may participate in signaling pathways integrating developmental and environmental cues.

Notably, qPS6-1 harbored Os06g0275000, a ZF-TF homologous to Hd1, while qPS6-4 contained Os06g0298200, encoding a BBX-domain TF related to Hd1. These genes are consistent with known components of photoperiodic flowering regulation. The qPS7 region included multiple TFs, such as AP1/FUL-like MADS-box genes (Os07g0605200) and NF-Y family members (Os07g0608200, Os07g0606600), which have been reported to influence FTI in rice. In qPS8-1, Os08g0407200 encodes an auxin transporter, associated with floral meristem development, while Os08g0513700 (SPL15) in qPS8-2 is known to function in the vegetative-to-reproductive phase transition.

2.5. Haplotype Analysis of Key SNPs Within GWAS Peaks Associated with HD in Rice

To investigate allelic variation underlying HD, haplotype analyses were performed at major GWAS peaks, focusing on SNPs within regions of strong linkage disequilibrium (LD; r² = 0.6–1.0). On chromosome 1 (24.55–24.80 Mb), a significant association peak encompassed several SNPs surpassing the −log₁₀(P) ≥ 5 threshold (Figure 4A). Haplotype analysis within this LD block identified two predominant haplotypes. Accessions carrying HapB (n = 75) exhibited significantly delayed heading compared to HapA (n = 161; p < 0.01) (Figure 4B,C). This region contains Os01g0587300, which encodes an OsFTIP1-like MCTP protein involved in florigen transport from leaves to the shoot apex. This suggests that natural variation in OsFTIP-mediated transport efficiency may contribute to PS divergence in Myanmar landraces.

On chromosome 6 (11.20–11.40 Mb), a prominent LD block included multiple SNPs near the GWAS significance threshold (Figure 4D). Two major haplotypes were identified; accessions with HapA (n = 140) displayed significantly later HD than HapB (n = 96; p < 0.0001) (Figure 4E,F). This region harbors Os06g0298200, encoding a BBX TF homologous to Hd1, a well-characterized photoperiodic floral initiation regulator.

A strong association was also observed on chromosome 7 (24.55–24.75 Mb), where multiple SNPs within a high-LD block defined two main haplotypes (Figure 4G). Accessions with HapB (n = 40) flowered significantly later than HapA (n = 196; p < 0.0001) (Figure 4H,I). Candidate genes in this region include Os07g0606600, encoding NF-YB10, a CCAAT-box TF involved in HD regulation, and Os07g0598200, a ZGT-like gene associated with circadian clock regulation.

On chromosome 8 (21.60–21.85 Mb), a significant peak encompassed SNPs defining two major haplotypes (Figure 4J). Accessions with HapB (n = 100) exhibited significantly later heading than HapA (n = 36; p < 0.0001) (Figure 4K,L). This region contains Os08g0407200, encoding an auxin transporter involved in floral meristem initiation and the transition to reproductive development.

Haplotype analyses identified allelic variation at key loci associated with HD regulation, including those involved in florigen transport, circadian rhythms, and hormone signaling. These results highlight the genetic diversity present in Myanmar rice landraces and provide a foundation for future functional and transcriptomic studies.

2.6. Transcriptomic Profiling and Functional Enrichment Reveal PS Mechanisms in Rice

2.6.1. Principal Component Analysis (PCA) and Sample Correlation

To investigate molecular mechanisms underlying PS, RNA-seq transcriptome profiling was conducted on three rice genotypes with contrasting photoperiodic responses: V3 (photoperiod-insensitive, indica), V10 (photoperiod-sensitive, indica), and CK (‘H479B’) (photoperiod-insensitive, japonica). Samples were collected at two critical time points: (i) the earliest heading stage, representing rapid-responding genotypes (early heading group), and (ii) the latest heading stage, representing delayed-heading genotypes (late heading group).

At the early heading time point, PCA revealed that PC1 and PC2 accounted for 78.6% of total variance (PC1: 46.8%; PC2: 31.9%) (Figure 5A). PC1 separated genotypes by genetic background, while PC2 distinguished photoperiod-sensitive V10 from -insensitive genotypes (V3 and CK). Biological replicates clustered tightly, and correlation analysis confirmed high reproducibility (Figure 5B), validating the dataset for downstream differential expression analysis.

At the late heading stage, PCA explained 75.4% of variance (PC1: 47.0%; PC2: 28.4%) (Figure 5C). Genotypes were again separated along PC1, while PC2 captured transcriptomic features associated with delayed heading in V10. Replicates clustered by genotype and treatment, with strong intra-group correlation (Pearson’s r > 0.9; Figure 5D), confirming the dataset’s reliability. These analyses indicate clear transcriptomic differentiation between photoperiod-sensitive and -insensitive genotypes, providing a basis for identifying stage-specific photoperiod-responsive genes.

2.6.2. DEGs at Earliest HD

To characterize transcriptional dynamics at floral initiation, RNA sequencing was performed at the earliest heading stage following SHD treatment across V3, V10, and CK. Differential expression analysis identified 4764 DEGs in V3 (2456 upregulated and 2308 downregulated) (Table S3), 5421 DEGs in V10 (2533 upregulated and 2888 downregulated) (Table S4), and 4765 DEGs in CK (2350 upregulated and 2415 downregulated) (Table S5) (|log₂FC| ≥ 1, adjusted p < 0.05) (Figure 6A).

Venn diagram analysis revealed 995 upregulated DEGs shared across all genotypes, suggesting conserved floral-induction mechanisms (Figure 6B). Genotype-specific responses were observed, with 800 DEGs unique to CK, 619 to V3, and 823 to V10, and 501 were shared between V3 and V10. Similarly, downregulated genes showed similar patterns; 991 were shared across genotypes, with 956, 580, and 761 uniquely downregulated in V10, V3, and CK, respectively (Figure 6C). V10 exhibited the highest level of transcriptional plasticity, consistent with its photoperiod-sensitive phenotype.

2.6.3. GO Enrichment Analysis at the Earliest HD

GO enrichment analysis categorized DEGs into biological process (BP), cellular component (CC), and molecular function (MF) terms. In V3, enriched BP terms included positive regulation of SHD photoperiodism and floral transition (GO:0048576), response to abscisic acid (GO:0009737), and cold response (GO:0009409). CC terms highlighted chloroplast (GO:0009507) and plastid compartments, while MF terms encompassed ribosomal structural constituents (GO:0003735) and amino acid transporter activity (Figure 6D, Table S9-1). In V10, BP terms emphasized translation and photoperiod-regulated floral pathways, including cytoplasmic translation (GO:0002181) and positive regulation of SHD photoperiodism (GO:0048576). CC enrichment highlighted ribosome and chloroplast stroma; MF annotations included rRNA and mRNA binding (Figure 6E, Table S9-2). In CK, enrichment was similar to V10, with BP terms centered on translation, rRNA processing, and ribosome biogenesis; CC terms emphasized chloroplast and ribosomal components; MF terms included ribosomal structural constituents and RNA-binding activities (Figure 6F, Table S9-3).

These patterns indicate that while all genotypes activate translational machinery, V3 is characterized by the enrichment of stress and hormone pathways, V10 is characterized by enhanced pathways associated with rapid protein synthesis, and CK emphasizes ribosomal and RNA-related regulation.

2.6.4. KEGG Pathway Enrichment at the Earliest HD

KEGG pathway enrichment analysis identified both conserved and genotype-specific regulatory pathways modulating photoperiodic floral responses. In the V3 genotype, DEGs were significantly enriched in pathways related to ribosome biogenesis (ko03008), photosynthesis (ko00195), and plant circadian rhythm (ko04712). Notably, hormone biosynthesis pathways, such as brassinosteroid biosynthesis (ko00905), and core metabolic processes, like glycolysis and gluconeogenesis (ko00010), were also prominent, indicating coordinated regulation of metabolism and signaling networks critical for FLT (Figure S7A, Table S10-1).

In contrast, the V10 genotype exhibited enrichment in pathways associated with fundamental cellular processes, including ribosome function (ko03010), RNA polymerase activity (ko03020), and amino acid metabolism, specifically glycine, serine, and threonine metabolism (ko00260). Additional enrichment in circadian rhythm (ko04712) and non-homologous end joining (ko03450) suggests an integrated role of transcriptional regulation and DNA repair mechanisms in the photoperiodic response (Figure S7B, Table S10-2).

In the CK genotype’s pathway, enrichment was mainly associated with translational and organellar functions, such as ribosomal activity, RNA degradation (ko03018), and chloroplast-related processes. Unique enrichments included purine metabolism (ko00230), pantothenate and CoA biosynthesis (ko00770), and glyoxylate and dicarboxylate metabolism (ko00630), indicating distinct energy management and nucleotide biosynthesis pathways during photoperiod adaptation (Figure S7C, Table S10-3).

Collectively, GO and KEGG analyses reveal a core set of photoperiod-responsive pathways centered on ribosome function, circadian regulation, and hormonal signaling. Nonetheless, each genotype employs distinct regulatory strategies: V3 emphasizes metabolism and hormone pathways, V10 enhances biosynthetic and DNA repair processes, and CK prioritizes translational control and organellar restructuring. These molecular insights underpin genotype-specific sensitivity to photoperiod during the early stage of floral induction.

2.6.5. DEGs at Latest HD

At the terminal flowering stage under SHD conditions, transcriptome profiling uncovered substantial, genotype-specific transcriptional reprogramming among V3, V10, and CK (Figure 7A). Differential expression analysis identified 3935 DEGs in V3 (1994 upregulated; 1941 downregulated), 2943 in V10 (1307 upregulated; 1636 downregulated), and 1877 DEGs in CK (1060 upregulated; 817 downregulated) (Tables S6–S8). The higher DEG count in V3 indicates a more dynamic transcriptional response compared to the more moderate reprogramming of V10 and CK.

Venn diagram analysis revealed both conserved and genotype-specific expression patterns (Figure 7B,C). Among upregulated genes, 237 were shared by all three genotypes, representing a core regulatory module associated with late-stage FLT. Conversely, V3 exhibited the highest number of unique upregulated DEGs (1190), followed by V10 (643) and CK (553), indicating divergent regulatory pathways. Pairwise overlap showed 362 shared genes between V3 and V10, 205 between CK and V3, and 65 between CK and V10 (Figure 7B). Downregulated DGEs followed a similar trend, with 266 conserved across all genotypes and 1200, 910, and 262 unique to V3, V10, and CK, respectively (Figure 7C). This distinct expression profile highlights V3’s prolonged activation of flowering and metabolism-related genes compared with the more stress-associated transcriptional responses of V10 and CK.

2.6.6. GO Enrichment Analysis at the Latest HD

GO enrichment analysis at the latest heading/flowering stage revealed an association of chloroplast-related processes and stress responses in all three genotypes. In V3, cellular component (CC) terms were dominated by chloroplast (GO:0009507), chloroplast stroma (GO:0009570), and thylakoid membrane (GO:0009535), while biological process (BP) terms included chloroplast organization (GO:0009658), photosynthesis (GO:0015979), and response to abscisic acid (GO:0009737) (Figure 7D, Table S11-1). This suggests that V3 maintains an active plastid and hormonal regulatory program even at late stages.

In V10, enriched CC terms also highlighted chloroplasts and light-harvesting complexes (GO:0009507; GO:0009517), while key BP terms included photosynthesis, light harvesting (GO:0009765), chloroplast organization (GO:0009658), response to hypoxia (GO:0071456), and protein refolding (GO:0042026) (Figure 7E, Table S11-2), indicating a dual strategy of sustaining photosynthetic activity while buffering stress.

In CK, GO enrichment highlighted plastid- and stress-related functions, with CC terms such as chloroplast, chloroplast stroma, and thylakoid membrane. BP categories were enriched for protein folding (GO:0006457), response to water deprivation (GO:0009414), and protein refolding (GO:0042026), with molecular function (MF) terms including unfolded protein binding (GO:0051082) and protein folding chaperone activity (GO:0044183) (Figure 7F, Table S11-3), consistent with a focus on maintaining protein stability and stress mitigation rather than extensive metabolic activation.

Together, these results show that all three genotypes rely on chloroplast regulation under SHD, but V3 emphasizes metabolic and hormone pathways, whereas V10 and CK shift toward protein stability and stress protection.

2.6.7. KEGG Pathway Enrichment at the Latest HD

KEGG pathway enrichment further revealed genotype-specific responses during late-stage flowering. In V3, DEGs were enriched in photosynthesis (ko00195), metabolic pathways (ko01100), biosynthesis of secondary metabolites (ko01110), and transitional energy pathways, such as glycolysis/gluconeogenesis (ko00010) and starch and sucrose metabolism (ko00500) (Figure S8A, Table S12-1). This broad enrichment indicates sustained metabolic activation and photoperiod monitoring.

In V10, enriched pathways emphasized biosynthetic and adaptive regulation, including the biosynthesis of secondary metabolites (ko01110), photosynthesis–antenna proteins (ko00196), and the circadian rhythm–plant (ko04712) pathway (Figure S8B, Table S12-2). Additional enrichment in alpha-linolenic acid metabolism (ko00592), carotenoid biosynthesis (ko00906), and flavonoid biosynthesis (ko00941) suggested that V10 relies on specialized metabolite pathways to balance reproduction and stress adaptation.

In CK, late-stage DEGs were predominantly enriched in porphyrin metabolism (ko00860), circadian rhythm plant, and photosynthesis (ko00195). CK-specific pathways included pentose phosphate pathway (ko00030), protein processing in the endoplasmic reticulum (ko04141), and glyoxylate and dicarboxylate metabolism (ko00630), indicating a strategy centered on energy production and protein quality control rather than extensive biosynthetic activation (Figure S8C, Table S12-3).

In summary, transcriptome profiling at the latest flowering time point revealed both conserved and genotype-specific gene expression responses to photoperiod. All three genotypes engaged core regulatory pathways involving the circadian rhythm, chloroplast activity, and stress responses. However, V3 maintained a broad, metabolically active program indicative of prolonged floral activation, whereas V10 and CK prioritized stress buffering, protein regulation, and energy management, reflecting subspecies- and genotype-specific adaptation to SHD conditions.

2.7. Candidate Gene Identification Through Integrated GWAS and RNA-seq Analyses in Rice

To elucidate the genetic basis of PS and HD adaptation in rice, we employed an integrative approach combining the GWAS with transcriptomic profiling via RNA-seq. This strategy identified 221 DEGs within thirteen GWAS-defined QTL regions: qPS2-1 (24 DEGs), qPS6-1 (8), qPS6-2 (5), qPS6-3 (9), qPS6-4 (8), qPS6-5 (7), qPS7 (79), qPS8-1 (6), qPS8-2 (11), qPS1 (11), qPS8-3 (11), qPS2-2 (23), and qPS3 (20) (Table S13).

Although some selected genes exhibited fold changes (log₂FC) of less than 2, they were chosen based on their functional importance in PS and HD regulation. Genes were prioritized for their roles in photoperiodic flowering pathways, transcriptional regulation, and stage- or genotype-specific expression, regardless of fold change magnitude, as well as their mapping to QTLs associated with PS and HD. Even modest fold changes can be physiologically relevant, particularly when genes play key roles in flowering regulation.

Within the qPS2-1 locus, we identified three genes that exhibited strong, stage-specific expression responses in the photoperiod-sensitive V10 genotype: Os02g0672100, encoding a C2H2-type ZF-TF, was significantly upregulated during the late developmental stage (log₂FC = 1.18); Os02g0671000, a homolog of APK1B protein kinase, exhibited strong early induction (log₂FC = 3.26), and Os02g0671100, an F-box domain-containing protein, was upregulated early (log₂FC = 2.36). These results indicate that qPS2-1 harbors multiple regulatory components that function in different temporal windows, suggesting coordinated control of heading under photoperiod sensitivity.

In the qPS6-1 region, Os06g0275000, a ZF-TF associated with HD regulation, exhibited marked upregulation at the late stage (log₂FC = 2.29). Similarly, within qPS6-4, Os06g0298200, a BBX domain TF with high homology to Hd1, displayed sustained upregulation at both early (log₂FC = 1.61) and late (log₂FC = 1.20) stages, highlighting its role in linking photoperiod sensitivity with spikelet development.

The qPS7, Os07g0613200, encoding a RING-type ZF-TF, was downregulated early (log₂FC = −6.78), while Os07g0604000, a 6-phosphogluconolactonase-like gene, was consistently upregulated across stages (log₂FC = 2.22 and 2.37). Additionally, Os07g0666000, an NF-Y CCAAT box binding factor, was significantly repressed at the late stage (log₂FC = −2.01). The contrasting expression patterns of TFs and metabolic genes within qPS7 suggest complex regulatory and metabolic interactions during photoperiodic flowering.

In qPS8-2, Os08g0513700 (OsSPL15), a member of the squamosa promoter-binding-like (SPL) TF, was induced at the early stage (log₂FC = 1.13), aligning with its known role in flowering induction. Additionally, Os02g0696900, a G2-like MYB TF within qPS2-2, exhibited early upregulation (log₂FC = 1.82), highlighting its potential involvement in developmental phase transitions associated with photoperiod adaptation. These findings demonstrate that both SPL and MYB-type TFs are key drivers of developmental plasticity under short-day conditions.

Collectively, this integrative approach identified key regulatory genes with stage- or genotype-specific expression changes within GWAS loci. The identified transcription factors (C2H2, BBX, SPL, NF-Y, MYB) and regulatory proteins (F-box, protein kinase, metabolic enzymes) represent important targets for further functional validation and breeding efforts to optimize heading date adaptation.

2.8. Validation of Candidate Genes via qRT-PCR

To confirm the RNA-seq results, we performed qRT-PCR analyses on eight candidate genes across photoperiod-sensitive (V10) and insensitive (V3) rice genotypes under control and SHD conditions (Figure 8).

The expression profiles obtained by qRT-PCR closely mirrored those from RNA-seq, demonstrating strong reproducibility and affirming the reliability of the transcriptomic data. Notably, Os01g0588500, Os06g0275000, and Os07g0603800 were consistently upregulated in V10 under SHD treatment, supporting their roles as positive regulators of photoperiod-induced flowering. Os07g0626700 showed the highest expression in V10 following treatment, indicating that it may function as a key activator within the short-day response pathway. Conversely, Os08g0414600 and Os07g0606600 exhibited elevated expression in V3, particularly after treatment, suggesting genotype-specific regulatory divergence between V3 and V10. Interestingly, Os07g0605400 peaked under control conditions in V10 but increased in V3 after treatment. Overall, the similarity between qRT-PCR and RNA-seq data validates the reliability of the integrative approach and supports the identification of these candidate genes as key regulators in photoperiod flowering pathways.

3. Discussion

3.1. GWAS Reveals Conserved and Novel Genetic Loci for PS

Our GWAS study identified 13 putative QTLs associated with natural variations in PS and HD, highlighting a genetic architecture that incorporates both conserved and novel loci from the diverse panel of Myanmar rice landraces. The finding of these QTLs, using multi-model and multi-locus approaches and linkage disequilibrium (LD) decay distances comparable to those in previous studies of cultivated rice [31], demonstrates the effectiveness of association mapping in this germplasm. Several of the QTLs co-located with known flowering time genes, providing robust biological validation. For example, qPS6-1 to qPS6-5 cover OsHd1, a master regulator of photoperiod flowering [26], and other regions involve the core integrators OsNF-YB9 and OsBBX19/DTH2 [27,28]. These findings are consistent with earlier studies in indica and japonica populations, where QTLs on chromosomes 6, 7, and 8 have been linked to major loci influencing photoperiod response and florigen signaling [7,18,32]. These genomic regions, therefore, represent conserved hubs for flowering-time regulation.

In addition to the well-established loci, we identified new QTLs (e.g., qPS7, qPS8-1, and qPS8-2), which have received less attention in previous reports. These strong association signals are likely driven by rare allelic variation, reflecting local adaptation, and suggest that Myanmar rice represents an important source of genetic diversity. In total, 906 annotated genes within these QTL regions were identified, including TFs (Zn finger-C2HC type; MYB-related protein); WRKY DNA-binding proteins (BBX), florigen transporters, and genes involved in hormone signaling. This wide range of genetic variation supports the model that circadian clocks, photoperiod, and developmental regulators collectively control heading date regulation [33,34]. The canonical and noncanonical QTLs detected across three environments in this study provide new insights into heading date regulation, contributing to the identification of candidate genes for functional validation and potential application in breeding strategies.

3.2. Transcriptional Dynamics Uncover a Two-Phase Model of Flowering Regulation

Transcriptomic profiling across early and late heading stages under SHD conditions resulted in highly dynamic stage-specific responses, transitioning from a conserved ‘initiation’ phase to specialized adaptive strategies that are genotype dependent. A core conserved transcriptional response, consisting of 1986 DEGs, was detected in the early heading phase, shared by all genotypes, and enriched for key processes such as photoperiodic flowering, ribosome biogenesis, and hormone-signaling pathways [34,35]. This early activation of genes, including well-known circadian clock regulators (LUX, ELF3), flowering activators (Hd3a and Ehd1), and repressors (Ghd7, PRR37) [36,37,38,39], establishes a molecular basis for the rice floral transition.

As development progressed until the late heading stage, this conserved transcriptional landscape diverged substantially to establish distinct, genotype-specific profiles. The total number of DEGs decreased compared to earlier time points, indicating transcriptional stabilization; however, genotypic contrasts were more pronounced. The photoperiod-sensitive genotype V10 exhibited large-scale reprogramming of stress response pathways, including those related to secondary metabolite biosynthesis and protein folding machinery [40], suggesting a stress-adaptive strategy to cope with developmental and environmental challenges while maintaining floral integrity. In contrast, the less sensitive genotype V3 persistently enriched processes related to chloroplast organization, photosynthesis, and carbohydrate metabolism [41], underscoring a continued, energy-centric approach to support reproductive development. This genotype-specific specialization was supported by KEGG pathway analysis, which revealed enriched pathways for secondary metabolite and lipid metabolism in V10 and primary metabolic and circadian rhythm pathways in V3 [42].

Altogether, these results indicate a distinct two-stage control model of heading date in Myanmar rice landraces. This study reveals a two-phase transcriptional pattern underlying photoperiodic flowering in Myanmar rice landraces. An early, conserved induction of circadian- and florigen-related pathways supports robust flowering initiation across genotypes [32,43]. The early conserved transcriptional response ensures robust flowering initiation, whereas the later genotype-specific response enables fine-tuning of local adaptation under distinct photoperiodic conditions. Together, these patterns demonstrate the transcriptional plasticity of Myanmar germplasm and its role in local adaptation.

3.3. Candidate Gene Identification Through Integrated GWASs and Transcriptomics

Integrating the GWAS with transcriptomic profiling provided a useful strategy for dissecting the genetic basis of PS and HD variation in Myanmar rice landraces. Combining both datasets, 221 DEGs were mapped to thirteen GWAS-identified QTLs related to flowering time index (FTI), allowing refinement of the candidate gene pool to key stage- and genotype-specific regulators. In the qPS2-1 region, three genes showed distinct temporal expression in the photoperiod-sensitive genotype V10. Os02g0672100, encoding a C2H2-type zinc finger transcription factor [30], was upregulated at the late heading stage, indicating a role in floral transition under short-day conditions. Os02g0670100, homologous to APK1B kinase, was induced early, implying involvement in light-responsive signaling. Similarly, Os02g0671100, an F-box domain protein, was upregulated early, consistent with its role in proteolysis during development transitions [44].

In the qPS6 region, Os06g0275000 (a ZF TF; [26]) was upregulated at the late stage, while Os06g0298200, a BBX domain TF closely related to Hd1, showed sustained expression across both stages [28]. These findings reinforce the regulatory significance of BBX family genes in photoperiodic flowering through circadian modulation and FT activation [26,28]. At qPS7, several candidate genes displayed contrasting expression patterns. The RING-type ZF TF Os07g0613200 was strongly repressed early, whereas Os07g0604000, a 6-phosphogluconolactonase-like enzyme, was consistently induced. Os07g0606600 (NF-YB10) was downregulated at the late stage, aligning with its known regulatory role in FTI through transcriptional complexes [45,46]. Additionally, Os08g0513700 (OsSPL15) was induced early, supporting its function in photoperiod-driven floral initiation [47]. This integrated approach effectively highlighted candidate genes involved in photo perception, signal transduction, and transcriptional regulation, key processes controlling HD [48,49]. From these, three genes (Os06g0275000, Os08g0414600, and Os07g0606600) were validated by qRT-PCR (Figure 8). Os06g0275000 was more highly expressed in the photoperiod-sensitive genotype V10, supporting its role in photoperiod-induced flowering, while Os08g0414600 and Os07g0606600 were more highly expressed in V3, indicating alternative regulatory roles.

Based on their significant GWAS association, differential expression, and potential functions, Os06g0275000 and Os07g0606600 were prioritized for CRISPR/Cas9 functional validation in our ongoing study. Os06g0275000, a ZF-TF, likely promotes heading and flowering under short days [26], whereas Os07g0606600 (NF-YB10) participates in circadian and photomorphogenic regulation [45]. Functional knockout analysis will clarify their causal roles and support precision breeding of Myanmar rice for enhanced photoperiod adaptability.

3.4. Implications for Breeding and Cultivar Development

The molecular resources identified in this study, including new QTLs and the high-confidence candidate genes, offer a robust genetic framework for precision breeding of climate-resilient rice. This study is one of the first attempts to integrate information on Myanmar germplasm and provides valuable gene sources related to photoperiod sensitivity, which can be directly utilized by breeders. The molecular targets identified, such as Os06g0275000 and Os07g0606600 [32,50], offer potential for advanced breeding strategies, including marker-assisted selection for key QTLs, as well as CRISPR/Cas9-based gene editing [48,51]. Additionally, the genotype-specific transcriptional regulatory mechanisms uncovered in this study can be leveraged to select germplasm with expected adaptive responses to specific agroecological conditions. Furthermore, to ensure high-yielding and stable cultivars, functional validation of the candidate genes identified, coupled with multi-location trials, is crucial. These efforts will help to optimize yield stability while contributing to sustainable food security under changing climate conditions [52].

3.5. Comparison with Previous Studies

The genetic architecture of PS and HD in rice is shaped by both conserved major-effect genes and population-specific loci, enabling adaptation to diverse environments. Key regulators, such as Hd1, Ehd1, and Ghd7, as well as members of the BBX and SPL families, have been identified as central to PS and HD regulation in Asian rice germplasm [7,18]. Integrative GWASs and transcriptomic studies in indica and japonica populations have frequently detected QTLs on chromosomes 6, 7, and 8, which overlap with major loci influencing circadian clock regulation, florigen expression, and developmental timing [7,18,32]. Recent meta-analyses further highlight the coexistence of conserved regulatory hubs and locally adapted loci [32,51].

For Myanmar rice landraces, studies have identified novel loci that expand beyond globally recognized QTLs. A GWAS of Myanmar indica accessions revealed a major locus on chromosome 3 containing the PhyC gene, influencing HD and adaptation across Myanmar’s agroecological zones [17]. Although not detected in our study, our GWAS–RNA-seq analysis identified novel loci on chromosomes 2, 6, 7, and 8 that show strong transcriptional evidence under short-day conditions, which have not been reported globally. Both previous and current studies implicate circadian rhythm regulation, light perception, and phytohormone signaling as critical to FTI regulation. Studies have also highlighted the roles of BBX transcription factors (e.g., Hd1, OsBBX4) and SPL family members (e.g., OsSPL15) in photoperiodic flowering [7]. Our findings extend this understanding by demonstrating that Myanmar landraces exhibit a unique two-phase regulatory model, beginning with early, conserved activation of circadian and florigen pathways, followed by genotype-specific transcriptional divergence during late flowering. This model, which has not been emphasized in previous studies, highlights the distinctive contribution of Myanmar germplasm to photoperiodic flowering regulation.

4. Materials and Methods

4.1. Plant Materials and Field Experimental Design

A comprehensive panel comprising 236 landrace accessions of Oryza sativa L. was assembled, representing the seven principal agroecological zones of Myanmar: the Central Dry Zone (CDZ), Delta Region (DR), Eastern Coastal Region (ECR), Eastern Mountainous Region (EMR), Northern Hilly Region (NHR), Western Coastal Region (WCR), and Western Mountainous Region (WMR). Seeds were obtained from the Myanmar Department of the Agricultural Research Seed Bank (Figure S1A,B).

To evaluate PS across diverse environmental contexts, the accessions were cultivated at three geographically and photoperiodically distinct sites: Yezin, Myanmar (19.82° N, 96.26° E), representing SHD conditions (~11.03 h daylight); Yuanyang, China (23.16° N, 102.74° E), representing mid-long-day (LOD) conditions (~11.22 h daylight); and Xundian, China (25.56° N, 103.26° E), representing long-day (LOD) conditions (~12.45 h daylight). Day length (time between sunrise and sunset) was calculated for the experimental growing period in 2021 (Myanmar) and 2023 (China) (from sowing to heading dates) using the solar declination/hour angle method, and the results were verified using the U.S. Naval Observatory’s Duration of Daylight/Darkness Table for One Year (https://aa.usno.navy.mil/data/Dur_OneYear), accessed on 28 January 2024 (Table S15).

Field trials employed a randomized complete block design with plots measuring 60 × 100 cm (three rows with ten plants per row). Standard agronomic practices were uniformly applied across all sites. In Yezin, sowing occurred on 11 January 2021, with transplanting on 3 February 2021. In Yuanyang, sowing was conducted on 20 January 2023 and transplanting on 20 February 2023, while in Xundian, sowing took place on 23 March 2023 and transplanting on 25 April 2023. During the experimental periods, maximum temperatures ranged from 32.4 °C to 37.8 °C in Yezin, 26.3 °C to 34.2 °C in Yuanyang, and 25.1 °C to 32.8 °C in Xundian (greenhouse), reflecting distinct thermal regimes (Figure S2A,B). Temperature data were obtained from the NASA Prediction Of Worldwide Energy Resources (POWER) database (https://power.larc.nasa.gov/data-access-viewer/), for the experimental growing periods in 2021 (Myanmar) and 2023 (China) (Table S16), accessed on 28 January 2024. Heading date was recorded as the number of days from sowing to the emergence of the first panicle above the flag leaf sheath in at least 50% of the plants within an accession.

4.2. DNA Extraction and Whole-Genome Sequencing

Genomic DNA was extracted from young leaf tissue of each accession using a modified cetyltrimethylammonium bromide (CTAB) protocol optimized for high-molecular-weight DNA suitable for sequencing. DNA quality and concentration were checked by spectrophotometry and electrophoresis. Sequencing libraries were prepared with the MGIEasy Universal DNA Library Prep Kit (MGI Tech Co. Ltd., Shenzhen, Guangdong Province, China) following the manufacturer’s instructions (fragmentation, end repair, adapter ligation, PCR enrichment). Library size distributions were assessed using an Agilent Bioanalyzer. Paired-end sequencing (150 bp, PE150) was performed on a DNBSEQ-T7 platform (Huazhi Biotechnology Co., Ltd., Changsha, Hunan province, China), targeting an average genome coverage of ~15× per accession.

Raw reads were processed using the Sentieon DNAseq pipeline (version 202112.06). Reads were aligned to the Oryza sativa reference genome (IRGSP-1.0) using BWA-MEM; PCR duplicates were marked, and local realignment was applied. Variant calling was performed according to GATK best practices [53], resulting in a raw SNP set that was subsequently filtered using VCFtools v0.1.17. Filtering retained bi-allelic SNPs with minor allele frequency (MAF) ≥ 0.05 and missing genotype rate ≤ 20%, yielding 2,647,384 high-confidence SNPs for downstream analyses.

4.3. Indica–Japonica Subspecies Classification

Subspecies composition (O. sativa ssp. indica vs. japonica) was estimated using genome-wide similarity to two reference genomes (‘9311’ for indica and ‘Nipponbare’ for japonica). Genomic similarity was calculated in non-overlapping 100 kb windows by comparing each accession’s genotype to reference genotypes; a subspecies index was defined as the proportion of windows with higher similarity to each reference. The subspecies index provides a continuous measure of indica/japonica composition for each accession (Figure 2D–F).

4.4. Population Structure, Kinship, and Linkage Disequilibrium

Population structure was assessed using PCA based on the filtered SNP set; the top five principal components were used as covariates in association models to account for population stratification. Kinship matrices were computed using the VanRaden method implemented in the rMVP package [54]. Linkage disequilibrium (LD) decay was estimated genome-wide to inform QTL window size. It was visualized using pairwise r² values computed in R with the SNPRelate package. LD heatmaps for candidate regions were generated using the LD heatmap package in R [55].

4.5. GWAS Analysis

GWAS analyses were implemented with the rMVP package in R [54]. Genotypic data were formatted in HapMap format, and phenotypes were stored as tab-delimited files with additive SNP coding. Three complementary models were applied to enhance detection robustness: (1) the General Linear Model (GLM) as a baseline, (2) the Mixed Linear Model (MLM) incorporating the VanRaden kinship matrix to account for relatedness, and (3) FarmCPU (Fixed and Random Model Circulating Probability Unification), which iteratively estimates fixed and random effects. Among these, the MLM and FarmCPU controlled false positives most effectively, and the MLM was selected for the final presentation of the results, given its balanced performance. The top five principal components were included as covariates in all models. Visualization included Manhattan and Q–Q plots (−log10 p). A uniform genome-wide significance threshold of p < 1 × 10⁻⁵ was applied across models in accordance with prior rice GWAS practice [31]. GWAS scripts and parameter settings are available at (Xiaolei-lab, China) https://github.com/xiaolei-lab/rMVP.

4.6. QTL Definition, Haplotype Analysis, and Candidate Gene Annotation

Considering the typical LD decay in cultivated rice (~100–200 kb) [31], QTL regions were defined as ±100 kb (200 kb intervals) centered on lead SNPs exceeding the significance threshold (p < 1 × 10⁻⁵). Significant SNPs located within the same LD block were grouped into single QTLs, with the most important marker designated as the lead SNP. Pairwise LD among SNPs was calculated and visualized using the LD heatmap package [55] in R, focusing on regions harboring GWAS peaks to elucidate haplotype structures.

Haplotype analysis was conducted using SNPs and insertion/deletion variants located within coding sequences and up to 2 kb upstream promoter regions of candidate genes. Variants with missing or heterozygous calls were filtered out to ensure high-quality genotypic data. Accessions were classified according to distinct haplotypes, and phenotypic differences among haplotype groups were evaluated by t-tests based on trait mean comparisons, enabling the assessment of allelic effects on phenotypic variation.

Candidate genes within or adjacent to significant SNP regions were functionally annotated using multiple reference databases. Genomic and protein sequences from the 9311 assembly and Nipponbare reference genome were retrieved from SNP-Seek database (International Rice Research Institute (IRRI), Los Baños, Laguna, Philippines) (https://snpseek.irri.org) and cross-validated using the Rice Genome Informatics database (Huazhong Agricultural University, Wuhan, Hubei, China) (https://riceome.hzau.edu.cn/), NCBI (National Library of Medicine, Bethesda, MD, USA) (https://www.ncbi.nlm.nih.gov/), and RAP-DB (National Agriculture and Food Research Organization (NARO), Tsukuba, Ibaraki, Japan) (https://rapdb.dna.affrc.go.jp/) databases. Sequence similarity searches were performed using BLASTp (BLAST+ suite, NCBI, Bethesda, MD, USA) to identify homologous proteins and functional domains. Gene function prediction integrated InterPro annotations and literature-based evidence [56,57], allowing robust inference of gene roles associated with photoperiod response and adaptation in rice.

4.7. Transcriptome Sampling, Library Preparation, and Sequencing

To elucidate the molecular mechanisms underpinning photoperiodic response in rice, transcriptomic profiling was conducted on selected indica varieties (O. sativa ssp. indica) exhibiting contrasting PS. From the three regions, two groups of accessions were selected based on days to heading: five varieties from a photoperiod-sensitive group [V6 (Mm3), V7 (Mm40), V8 (Mm60), V9 (Mm70), V10 (Mm88)] and five from an insensitive group [V1 (Mm111), V2 (Mm164), V3 (Mm192) V4 (Mm205), V5 (Mm214)], along with two japonica (O. sativa ssp. japonica) control varieties (CK1 ‘H479B’ CK2 ‘Hwayeongbyeo’) for RNA sequence analysis. Plants were cultivated under controlled conditions (with natural light and timed darkness) at a greenhouse in Kunming, China (25.08° N, 102.45° E, 1950masl). During the experimental periods, maximum temperatures ranged from 25 to 35 °C in the greenhouse.

At the 3–4 leaf stage, plants were subjected to either a control group under natural photoperiods (12–13 h light/day) or an SHD treatment (7–8 h light/day) for 4–5 weeks [4,5,58], with three biological replicates per genotype treatment combination. Specifically, two genotypes were chosen based on their treatment response: V3 (‘Nga Kywe Dume’, Mm192), a photoperiod-insensitive indica-clined genotype from the Western Mountainous Region (WMR), and V10 (‘In Ma Ye Baw’, Mm88), a photoperiod-sensitive indica genotype from the Eastern Coastal Region (ECR) (Table S1, Figure 2A–C). ‘H479B’ (japonica genotype) served as a reference control. Leaf tissues were sampled at two developmental stages corresponding to the earliest and latest heading dates observed, flash-frozen in liquid nitrogen, and stored at −80 °C for subsequent RNA extraction. A total of 36 samples were collected (3 genotypes × 2 treatments × 3 replicates × 2 time points). Total RNA was isolated from the 36 samples, and mRNA was purified using magnetic bead-based enrichment. Sequencing libraries were prepared with the VAHTS Universal V6 RNA-seq Library Prep Kit (Illumina^®), fragmented, and reverse-transcribed following the manufacturer’s protocol. Libraries were sequenced on the Illumina NovaSeq X Plus platform to generate 150 bp paired-end reads. Raw sequencing data underwent quality control and adapter trimming using SOAPnuke.

4.8. RNA-seq Processing and Differential Expression Analysis

Raw reads were filtered and trimmed using SOAPnuke to remove adapters and low-quality reads. Clean reads were aligned to the Oryza sativa IRGSP-1.0 reference genome using HISAT2 [59]. Transcript assembly and quantification were performed with StringTie v2.1.2 [60]. Read counts were imported into DESeq2 [61] for normalization and differential expression analysis. DEGs were identified with adjusted p ≤ 0.05 (Benjamini–Hochberg FDR) and |log₂ fold change| ≥ 1.

Functional enrichment of DEGs was assessed by GO and KEGG pathway analyses. GO annotations were retrieved from the GO resource, the Gene Ontology Consortium (http://www.geneontology.org/), and KEGG pathway mapping was performed using KEGG resources (Kanehisa Laboratories, Kyoto University, Kyoto, Japan) (https://www.kegg.jp/). Enrichment significance was evaluated using the hypergeometric test with Bonferroni correction for multiple testing; adjusted p ≤ 0.05 was considered significant.

4.9. qRT-PCR Validation

Eight candidate genes selected from the integration of the GWAS and RNA-seq results (|log₂FC| > 1) were validated by quantitative real-time PCR (qRT-PCR). Total RNA was extracted from leaf tissue using the RNA Extraction Kit (Tsingke Biotechnology Co., Ltd, Beijing, China) and treated with DNase I to remove genomic DNA. First-strand cDNA synthesis was performed from 2 μg total RNA using SynScript^® III RT SuperMix for qPCR (Tsingke Biotechnology Co., Ltd, Beijing, China). qRT-PCR was conducted with ArtiCanCEO SYBR qPCR Mix (Tsingke Biotechnology Co., Ltd, Beijing, China) on an ABI QuantStudio StepOnePlus system (Thermo Fisher Scientific Inc., Waltham, MA, USA). Reaction and cycling conditions followed the manufacturer’s recommendations. Relative expression was calculated using the 2^−ΔΔCT method with OsACT1 (β-actin) as the reference gene. All reactions were run in triplicate. Primer sequences are provided in Table S14.

4.10. Statistical Analyses and Software

Phenotypic summaries (means and standard errors) were calculated in Microsoft Excel 2016. Visualization (bar graphs, LD heatmaps, Manhattan plots) was produced using GraphPad Prism v10 and R (RStudio v4.2.2) with packages including LDheatmap [55] and rMVP [54]. GWAS and population genetics analyses used rMVP and PLINK as indicated above. RNA-seq processing used SOAPnuke, HISAT2, StringTie, and DESeq2 (versions listed above). All software and package versions are described in the corresponding method subsections.

5. Conclusions

This study demonstrates the effectiveness of integrating GWASs with time-course transcriptomics to dissect the genetic basis of PS in Myanmar rice landraces. Thirteen significant QTLs and over 200 differentially expressed genes with stage- and genotype-specific patterns were identified, highlighting both conserved and population-specific flowering regulators. Two high-confidence candidate genes, Os06g0275000 (ZF-TF) and Os07g0606600 (NF-YB10), were prioritized for their robust association signals and dynamic expression, providing promising targets for functional validation. These findings advance understandings of photoperiod adaptation and offer valuable resources for developing climate-resilient rice cultivars with optimized flowering time, contributing to both regional breeding efforts and the broader rice research community.