Genomic regions associated with flag leaf and panicle architecture in rice (Oryza sativa L.) Authors: Gurjeet Singh, Subroto Das Jyoti, Priyanka Uppalanchi, Rushika Chepuri, Sejuti Mondal, Chersty L Harper, Punniakotti Elumalai, Ken Mix, Nicole Wagner, Darlene L Sanchez, Stanley Omar PB Samonte, Shyamal K Talukder Published: 2024 ABSTRACT AbstractBackgroundFlag leaf (FL) and panicle architecture (PA) are critical for increasing rice grain yield as well as production. A genome-wide association study (GWAS) can better understand the genetic pathways behind complex traits like FL and PA.ResultsIn this study, 208 diverse rice germplasms were grown in the field at the Texas A&M AgriLife Research Center at Beaumont, TX, during 2022 and 2023 following Augmented Randomized Complete Block Design. After heading, eight different flag leaf and panicle architecture (FLPA) related traits were measured. GWAS analyses were performed to identify potential genomic regions associated with FLPA traits. A total of 97 quantitative trait loci (QTLs) (48 in 2022 and 49 in 2023) were distributed across all 12 chromosomes. GWAS revealed four QTLs (qSBPP4-2, qFLW6-2, qGNPP9, and qGWPP2-3) with phenotypic variation ranging from 11.7 to 22.3%. Two genetic loci were identified as multi-trait QTLs, i.e., S04_32100268 (qFLL4-1 and qFLA4-1) and S04_11552936 (qFLW4 and qFLA4-2) during 2022 and 2023, respectively. Additionally, these loci were further utilized to analyze candidate genes, and 65 genes were predicted in the 100-kb genomic region upstream and downstream. In silico expression analysis revealed 15 genes were expressed during the reproductive stage. These genes were associated with protein kinase, heat shock transcription factor family, sugar transporter conserved site and transcription factor bHLH95- like basic helix-loop-helix domain protein, as well as those that regulate the FLPA-related traits. Os04g0631100 was identified as a potential candidate gene that is highly expressed during the endosperm development stage, and it is associated with an important sugar transporter protein that will be helpful in grain improvement.ConclusionGWAS results revealed four major and two multi-trait QTLs. Expanding their candidate genes, and expression analysis provide the genetic information for molecular improvement of the FLPA-related trait in rice breeding programs.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12864-024-11037-z.Keywords: Rice, Flag leaf, Panicle architecture, GWAS, Candidate gene Background Over half of the world’s population relies on rice as their staple food [1], and it is a key source of calories and nutritional quality in the majority of nations with limited resources [2, 3]. Since the global population is increasing, rice production needs to be expanded to ensure food security [4] under low-input agriculture [5]. While rice is cultivated in many regions worldwide, climate change makes sustainable intensification of rice yield difficult [6]. In this context, developing strategies to improve yield contributing traits by dissecting complex traits through genome-wide association study (GWAS) is pivotal for enhancing genetic gain in rice [7]. Rice grain yield is determined from the number of panicles, grain number per panicle (GNPP), and grain weight per panicle (GWPP) [8]. These three major components are also affected by panicle length (PL), number of primary branches per panicle (PBPP), and number of secondary branches per panicle (SBPP). Leaf shape influences yield potential by altering light interception and photosynthetic rates [9]. The flag leaf (FL) is the source of more than half of the carbohydrates accumulated during rice grain formation [10]. Flag leaf and panicle architecture (FLPA) related traits have major implications for increasing rice grain yield and production [11–14]. Hence, pre-breeding for FL and PA is important for understanding the genetic and molecular basis of rice yield. Improving flag leaf length (FLL), flag leaf width (FLW), flag leaf area (FLA), and PA will be crucial in increasing the yield potential of rice cultivars to meet the increasing demand [15]. Several quantitative trait loci (QTLs)/genes regulate the FLPA-related traits in rice. Over the years, researchers have identified multiple genes controlling panicle-related traits in rice. For example, DST (DROUGHT AND SALT TOLERANCE) regulates PL, PBPP, and SBPP [16, 17], OsGRF4 positively regulates grain shape and PL [18], OsLG1 controls PA [19], lt1 (low tiller 1) affects panicle branching [20], OsSPL14 increases grain productivity [21], and IPA1 (Ideal Plant Architecture 1) modulates GNPP and tiller number [22]. In addition, these genes have been linked to several pathways, such as the ubiquitin-proteasome pathway, G-protein signaling pathway, MAPK signaling pathway, and the cytokinin signaling pathway [23]. Protein kinase and the protein heat shock transcription factor family play an important role during the panicle development stage [24, 25]. The sugar transport protein family and bHLH (basic helix-loop-helix) protein family regulate the FL traits and help in different plant development stages [26–29]. Protein kinase and sugar transport proteins play significant role in the development of FLPA related traits, which ultimately helps to improve the rice yield. Although significant progress has been achieved in identifying regulatory genes, identifying QTLs followed by validating and developing molecular markers are more feasible for crop improvement programs due to the quantitative nature of these traits. Interestingly, numerous QTLs that control PA in rice have been identified. For example, qPN1controls multiple panicle traits including PL, GNPP [30], qPL6 on chromosome 6 covering 25 kb region controls PL [31], qOPW11controls panicle weight [32], qTSN12.1 and qTSN12.2 increase total number of spikelet by increasing the branching in a panicle [33], qSBN7 controls SBPP [34], and qSPP2.2 regulates spikelet per panicle [35]. Likewise, several studies identified QTLs controlling FL morphology in rice. Some examples of such QTLs include qFLW7.2 for FLW [31], qFLL9 for FLL [36], qFL1 for overall flag leaf size [11], qFLL6.2 for FLL [37], qFLW4 for FLW and qFLAG5 for flag leaf angle [38]. Recently, GWAS has been more convenient for studying FLPA-related traits from diverse germplasm accession than traditional bi-parental mapping. GWAS offers a more convenient and potentially more powerful approach to studying genetic traits by utilizing existing naturally occurring genetic variation within a diverse set of accessions, rather than requiring the extensive time to development of mapping populations. Furthermore, sequencing technology is improving while its cost is decreasing substantially. Consequently, GWAS has been used frequently in candidate gene mining and marker identification for complex traits [39]. Fortunately, many studies have conducted GWAS on FLPA-related traits. For instance, GWAS conducted on diverse rice accessions identified 153 QTLs governing PA-related traits [14], 29 QTLs controlling panicle morphological traits [40], multiple marker-trait associations for panicle morphology and grain structures [41], four QTLs controlling PL [42], multiple QTLs controlling FL size through SNP and haplotype-based GWAS [43], and 64 QTLs identified for FL-related traits [12]. The results of these studies help to develop molecular markers for marker-assisted selection, which will speed up the selection process. For example, the findings of a GWAS was used to develop markers for PA-related traits [44]. Identified potential candidate genes were used for expression analysis by RiceXPro to better understand the gene function and see the expression level under specific tissues of plants [45]. Multiple results have been reported using RiceXPro to identify the potential candidate genes at various growth stages for impotent traits, such as agronomic, root-related, and reproductive [46–48]. Besides, genomic variants identified from GWAS results can be used as a starting point in genome editing programs to develop desirable changes in the rice genome. Due to the complex inheritance nature of these FLPA-related traits, the environment plays a significant role in the phenotypic variation of these traits. Therefore, it is crucial to study these important traits in the target environment to understand the genotype-environment interaction better. To date, all the GWAS studies that were conducted had focused either on FL or PA related traits. In this context, the current study was done to perform a combined GWAS on FLPA traits to identify suitable markers associated with these important traits for rice improvement. In this present research, we studied eight FLPA-related traits, namely FLL, FLW, FLA, PL, PBPP, SBPP, GWPP, and GNPP, in 208 diverse genotypes over two years in the field to perform GWAS. We aimed to identify marker-trait associations and candidate genes that govern FLPA-related traits in rice. The results of these investigations will help design markers for rice molecular breeding programs. Additionally, the research will shed light on the possible genomic regions that can be modified through genome engineering techniques for desirable changes in the future. Similarly, potential germplasm will be identified and integrated into the breeding program for target trait improvement. Materials and methods Plant materials and experiment design A total of 208 (including four check varieties) diverse rice accessions, including indica and japonica cultivars, landraces, and inbred lines [49], were phenotyped at the Texas A&M AgriLife Research Center at Beaumont, TX, USA. The germplasms were grown following Augumented Randomized Complete Block Design in six blocks, with four checks (C1 = Antonio, C2 = Cheniere, C3 = Cocodrie, and C4 = Presidio) as replication in each block, in 2022 and 2023. During both years, each germplasm line was direct-seeded into three rows of 2.43 m each, using a six-row planter. The row-to-row distance of the planting was 0.28 m. A maintained and recommended package of cultural practices was followed to raise a healthy crop. Phenotypic data collection and analysis The data were recorded for flag leaf and panicle architecture (FLPA)-related traits using five randomly selected plants from the three rows of each germplasm line. The FLPA traits included flag leaf length in cm (FLL), flag leaf width in cm (FLW), flag leaf area in cm2 (FLA), panicle length in cm (PL), primary branches per panicle (PBPP), secondary branches per panicle (SBPP), grain weight per panicle in gram (GWPP), and grain number per panicle (GNPP). All phenotypic data were analyzed in RStudio using the “Agricolae” (Team 2013) and “augmented RCBD” packages [50]. A statistical normality test was performed using the Shapiro-Wilk test in RStudio with the ‘dplyr’ package. Pearson’s correlation coefficient analysis was carried out using the “Corrplot” package built in RStudio [51]. Using the replicated data of four checks, which corresponded to all the measured traits, adjusted means were calculated by “augmented RCBD” [50], which were then used for GWAS analysis. Molecular marker data and population structure The marker information of the population was earlier reported [49, 52]. In brief, a total of 1,075,302 sequence-based SNP markers were used at the start of this study. The raw marker data were filtered to eliminate SNPs with > 50% missing data and minimum allele frequency (MAF) < 5%, reducing the raw marker data to 1,075,302 SNPs, and imputation was carried out using BEAGLE V4.0 [53]. The selected markers were imputed and filtered using TASSEL 5.2.61 [54], where SNPs having low-quality clustering and MAF < 5% were removed, and 854,832 SNPs were finally selected for further association analysis. All marker data are available on Dryad (10.5061/dryad.4qrfj6qbs). STRUCTURE v2.3.4 software was used to categorize sub-populations to observe the population structure in the rice panel used in this study based on the admixture model with correlated allele frequency [55]. The factors that may cause the false trait-SNP associations, i.e., population structure (Q) and genetic relatedness (K), were controlled using principal component analysis (PCA) and kinship matrix, respectively. Previous research using the same population as this experiment has calculated the PCA kinship matrix [56] and linkage disequilibrium (LD) decay [49, 52, 57]. Of the entire genetic variation, 56.7% has been described by the first four main principal components (PC). The first PC separated the indica from the japonica rice, explaining 38.9% of the variation. The second PC, contributing to 12.0% of the variation, further classified the japonicas into temperate and tropical subpopulations. There were admixed rice accessions between these three main subgroups [49, 52, 57]. To assess the required resolution for association mapping, LD decay, defined as the distance at which the mean r2 decreased to half its highest value, was estimated using TASSEL 5.2.61 [54]. The average genome-wide LD decay of the population has been calculated to be approximately 150,000 bp [49, 52, 57]. Genome-wide association analyses, candidate gene identification, and expression analysis The adjusted mean of both years was used to perform GWAS analysis. The small impact of genetic loci in complex polygenic attributes can be captured using multi-locus approaches, which have recently gained popularity and feasibility. So, we used these multiple methods, taking advantage of the algorithmic relevance of distinct models and supporting their findings one after the other. The R package mrMLM v5.0 (https://cran.r-project.org/package=mrMLM) was used with default parameters to perform GWAS analysis using the multi-locus approach [58]. There are six multi-locus models in the package, including mrMLM [59, 60], FASTmrMLM [61], FASTmrEMMA [62], pKWmEB [63], pLARmEB [64] and ISIS EMBLASSO [65]. When estimating QTLs, the FASTmrEMMA model outperforms mrMLM in speed, statistical power, and accuracy [61]. In comparison to mrMLM and FASTmrEMMA, ISIS EM-BLASSO has the best robustness and accuracy for detecting significant associations [65]. The pLARmEB model integrates least angle regression with empirical Bayes, allowing for a more accurate estimate of significantly linked QTLs in polygenic backgrounds [64]. QTLs with LOD scores greater than 3.00 were considered important (highlighted in pink on the Manhattan plot). Furthermore, QTLs found in at least two models were considered to be reliable. QTLs (marker sequences) strongly linked with FLPA (flag leaf and panicle architecture) traits were searched against the O. sativa japonica genome assembly IRGSP-1.0 Online Web Resource Ensembl Plants (https://plants.ensembl.org/Oryza_sativa/Info/Index/Tools/Blast) using BLASTn with standard parameters to locate potential candidate genes. Furthermore, we investigated the physical positions of the SNP marker’s upstream and downstream 100 kb regions and identified genes inside each locus. Further identified candidate genes were used for in silico expression analysis by RiceXPro (Rice Expression Profile Database https://ricexpro.dna.affrc.go.jp/) during the entire growth period and reproductive stage of the crop [45]. We used different crop stages for expression analysis, including spatio-temporal gene expression of various tissues/organs throughout the entire growth in the field, gene expression profile during reproductive organ development, grain gene expression profile at the early developmental stage, and embryo and endosperm gene expression profile at ripening stage [66, 67]. Results Phenotypic variation for FLPA All traits showed normal phenotypic distribution in 2022 and 2023 (Fig. 1). The Shapiro-Wilk test was used to assess statistical normality, revealing that the p-values for FLW and PBPP were greater than 0.05, while the p-values for other traits (FLL, FLA, PL, SBPP, GWPP, and GNPP) were almost equal to 0.05. Year-wise analysis of variance (ANOVA) showed significant variation in all the studied traits except PBPP and GWPP in 2023 (Table 1). Considering year as environment and as a factor, genotypes were significantly different for all traits except PBPP, and environmental influence was significant for all traits (Table 1). All traits exhibited non-significant differences in the pooled ANOVA at the block level; likewise, the genotype x environment interaction indicated non-significant differences except for PL and SBPP. Descriptive statistical analysis was performed for FLPA-related traits, and a wide range of variability was observed during 2022 and 2023 (Table 2). The average of FLL was 27.56 and 28.58 cm, FLW was 1.46 and 1.50 cm, and FLA was 30.57 and 32.62 cm2 in 2022 and 2023, respectively. For panicle-architectural traits, the mean PL was 20.24 and 20.78 cm, PBPP was 9.77 and 10.34, and SBPP was 3.18 and 2.45. GWPP was 1.88 and 1.83 g, and GNPP was 118.71 and 125.51 in 2022 and 2023, respectively. Among all the studied traits, the flag leaf traits exhibited lower coefficients of variation (CV) than the panicle traits. The lowest was observed for FLW (6.13 cm in 2022 and 10.57 cm in 2023), while the highest CV was observed for SPBB (36.74 in 2022 and 31.31 in 2023). Skewness showed significant differences for all the traits except FLW and PBPP, whereas kurtosis also showed significant differences for FLL, FLA, and PL during both years (Table 2). Pearson’s correlation coefficients were carried out for independent years and the mean data of both years for all traits (Fig. 2). FLA had a positive and significant correlation with FLL and FLW in 2022 (r = 0.85 and r = 0.73), 2023 (r = 0.85 and r = 0.78), and the average of both years are r = 0.86 and r = 0.76, respectively. GNPP showed significant correlations with GWPP of r = 0.62, r = 0.50, and r = 0.56 in 2022, 2023, and with the average data of both years, respectively. Phenotypic data distribution of diverse rice accessions: flag leaf length in 2022 (a), and during 2023 (b); flag leaf width during 2022 (c), and during 2023 (d); flag leaf area during 2022 (e), and during 2023 (f); panicle length during 2022 (g), and during 2023 (h); primary branch per panicle during 2022 (i), and during 2023 (j); secondary branch per panicle during 2022 (k), and during 2023 (l); grain weight per panicle during 2022 (m), and during 2023 (n); grain number per panicle during 2022 (o), and during 2023 (p) Table 1Analyses of variance for flag leaf and panicle architecture-related traits of diverse rice germplasmSource of variationdfYearMean SquaresFLLFLWFLAPLPBPPSBPPGWPPGNPPANOVA based on single year Block (ignoring Genotypes)5202223.910.02*35.4127.29**4.438.80**0.58*180.57202313.480.0334.727.56*1.781.470.27473.08Treatment (eliminating Blocks)207202234.24**0.05**79.56**14.35*6.055.78**0.75**1433.54202333.13*0.07*98.27**6.87**3.481.68*0.401264.02Treatment: Checks3202225.210.0253.94141.68**100.83**8.15**1.05**1949.72202364.86*0.0148.2914.41**3.802.81*0.43ns1363.11Treatment: Genotype and Genotype vs. Check204202234.37**0.05**79.93**12.484.665.74**0.75**1425.95202332.66*0.07*99.00**6.76 **3.471.67*0.401262.56Residuals1520229.290.0119.945.964.531.440.19970.42202315.250.0329.931.944.560.610.201177.33Pooled ANOVA based on both years Genotype211-55.74***0.09684***149.1***14.18***5.94.2***0.8751***2251***Block4-12.50.0203340.54.811.712.130.1721288Environment (two years)1-121.97**0.13411*457.5***40.19**58.32**66.33***1.607**3756*Genotype x Environment211-10.660.0190425.47.4*3.523.32***0.2793465Residuals28-15.70.01803334.075.241.110.1737705df Degree of freedom, FLL Flag leaf length (cm), FLW Flag leaf width (cm), FLA Flag leaf area (cm2), PL Panicle length (cm), PBBP Primary branch per panicle (number), SBPP Secondary branch per panicle (number), GWPP Grain weight per panicle (g), and GNPP Grain number per panicle (number)Level of significant: *P <= 0.05; **P <= 0.01, ***P <= 0.001 Analyses of variance for flag leaf and panicle architecture-related traits of diverse rice germplasm df Degree of freedom, FLL Flag leaf length (cm), FLW Flag leaf width (cm), FLA Flag leaf area (cm2), PL Panicle length (cm), PBBP Primary branch per panicle (number), SBPP Secondary branch per panicle (number), GWPP Grain weight per panicle (g), and GNPP Grain number per panicle (number) Level of significant: *P <= 0.05; **P <= 0.01, ***P <= 0.001 Table 2Descriptive statistics for flag leaf and panicle architecture in diverse rice germplasmTraitsYearMeanMinMaxSESDSkewnessKurtosisCVFLL202227.5612.3450.150.426.080.76**4.21**11.16202328.5815.8551.820.426.130.76**4.42**13.78FLW20221.460.802.100.020.240.082.936.1320231.500.862.370.020.26−0.063.0310.57FLA202230.578.8962.970.659.380.58**3.81*14.73202332.6211.7369.630.7210.350.65**3.82*16.93PL202220.2411.0334.560.253.580.50**4.28**12.14202320.7813.0631.610.192.770.54**3.98*6.72PBPP20229.773.8916.060.172.510.102.6821.72202310.345.3215.160.142.010.152.5120.32SBPP20223.180.0011.440.172.400.66**2.9336.7420232.450.007.050.101.440.64**3.0131.31GWPP20221.880.014.470.060.830.60**3.3622.3120231.830.393.980.040.640.50**3.5124.37GNPP2022118.7144.43238.432.5737.010.50**3.3825.52023125.5154.31240.392.5136.200.46**2.9426.83FLL Flag leaf length (cm), FLW Flag leaf width (cm), FLA Flag leaf area (cm2), PL Panicle length (cm), PBBP Primary branch per panicle (number), SBPP Secondary branch per panicle (number), GWPP Grain weight per panicle (g), and GNPP Grain number per panicle (number). SE Standard error, SD Standard deviation, and CV Coefficient of variation. Level of significant: *P < = 0.05; **P < = 0.01 Descriptive statistics for flag leaf and panicle architecture in diverse rice germplasm FLL Flag leaf length (cm), FLW Flag leaf width (cm), FLA Flag leaf area (cm2), PL Panicle length (cm), PBBP Primary branch per panicle (number), SBPP Secondary branch per panicle (number), GWPP Grain weight per panicle (g), and GNPP Grain number per panicle (number). SE Standard error, SD Standard deviation, and CV Coefficient of variation. Level of significant: *P < = 0.05; **P < = 0.01 Correlation plot of diverse rice germplasm accessions for flag leaf and panicle architecture related traits: year 2022 (a), year 2023 (b), and pooled data of both years (c) GWAS analysis for FLPA The GWAS analysis was carried out utilizing 854,832 SNP markers using six multi-locus models independently. Adjusted mean data of each year (2022 and 2023) were used for GWAS analysis for all FLPA-related traits. Based on pooled ANOVA, the years had a significant impact on all traits; hence, GWAS was performed independently for 2022 and 2023 data (Table 1). A total of 97 significant QTLs (Table S1; Fig. 3) were identified for all the studied traits having LOD scores ≥ 3.0 in ≥ 2 multi-locus models. Among them, 48 and 49 QTLs were identified during 2022 and 2023, and 25 and 72 QTLs were identified for FL and panicle-related traits, respectively (Fig. 3). For all eight traits in both years, significant QTLs were shown in Manhattan and quantile-quantile (Q-Q) plots (Fig. S1-S8). PA-related traits were associated with higher number of QTLs than FL traits, with the highest number found in PL (19), followed by SBPP (17). The lowest number of QTLs was found in FLL (8) and FLA (8). The highest number of QTLs were found on chromosome 4 (16 QTLs), followed by 2 (12 QTLs) (Figs. 3 and 4). Four potential QTLs were identified with phenotypic variation (R2) ranging from 10.32 to 22.30% (Table 3). Among them, qGNPP9 (chromosome 9) was identified in 2022, and qSBPP4-2 (chromosome 4), qFLW6-2 (chromosome 6), and qGWPP2-3 (chromosome 2) were identified in 2023. Summary of 97 marker trait associations (MTAs) with flag leaf and panicle architecture: Distribution of MTAs with different traits across the environments (a), and Distribution of the MTAs on all 12 chromosomes (b) Distribution of quantitative trait loci (QTLs) identified for FLPA-related traits on the rice chromosomes during 2022 and 2023. Each QTL is depicted as a horizontal bar, with the QTL name on the right side and its physical position (bp) on the left side. The QTLs highlighted in red color represent multi-trait QTLs. The light blue color represents the overlapped QTLs with the previous study (for all previous QTL references, see supplementary Table S1) Table 3Potential QTLs identified for FLPA-related traits from diverse rice accessionsS. No.QTL nameTrait nameYearChromosomeMarker IDMarker position (bp)LOD scoreR2(%)MethodQTLs identified with major phenotypic effect1.qSBPP4-2SBPP20234S04_1709072917,090,7293.9722.301, 2, 3, 52.qFLW6-2FLW20236S06_1186855411,868,5543.9116.401, 53.qGNPP9GNPP20229S09_2055130420,551,3045.0013.053, 44.qGWPP2-3GWPP20232S02_1221491412,214,9145.6911.721, 2, 3, 4, 5Multi-trait QTLs (Co-localized QTLs)5.qFLW4FLW20234S04_1155293611,552,9364.865.062, 56.qFLA4-2FLA20234S04_1155293611,552,9364.897.192, 57.qFLL4-1FLL20224S04_3210026832,100,2684.485.521, 2, 58.qFLA4-1FLA20224S04_3210026832,100,2684.9310.322, 5Model: 1 = FASTmrMLM, 2 = pLARmEB, 3 = mrMLM, 4 = pKWmEB, and 5 = ISIS EM-BLASSOFLL Flag leaf length, FLW Flag leaf width, FLA Flag leaf area, SBPP Secondary branch per panicle, GWPP Grain weight per panicle, and GNPP Grain number per panicle. Potential QTLs identified for FLPA-related traits from diverse rice accessions Model: 1 = FASTmrMLM, 2 = pLARmEB, 3 = mrMLM, 4 = pKWmEB, and 5 = ISIS EM-BLASSO FLL Flag leaf length, FLW Flag leaf width, FLA Flag leaf area, SBPP Secondary branch per panicle, GWPP Grain weight per panicle, and GNPP Grain number per panicle. Co-localized QTLs for multiple traits Among all the identified QTLs, two were found as multitrait-QTLs (Table 3). In 2023, the S04_11552936 genomic region on chromosome 4 was linked to two QTLs, qFLW4 for FLW and qFLA4-2 for FLA. In 2022, the S04_32100268 genetic locus located on chromosome 4 was linked with two QTLs, namely qFLL4-1 for FLL and qFLA4-1 for FLA. Based on specific genomic regions (within 0.5 Mb of a locus region), six genetic loci were associated with the multiple traits (Fig. 5). On chromosome 2, the genomic region from 13.35 to 13.52 Mb was associated with two QTLs: qGNPP2-3 (for GNPP in 2022) and qSBPP2-2 (for SBPP in 2023). Similarly, the 0.12–0.17 Mb genomic region on chromosome 4 was associated with qFLL4-2 (for FLL in 2023) and qPL4-1 (for PL in 2022), while the 32.10–32.14 Mb region was linked to qFLA4-1 (for FLA in 2022) and qGNPP4-2 (for GNPP in 2023). Chromosome 6, 11.06–11.26 Mb region contains two QTLs: qPL6 (identified in 2022 for PL) and qFLA6 (identified in 2023 for FLA). QTLs qPBPP8 (2023 for PBPP) and qSBPP8 (2022 for SBPP) was located in the 18.50–18.62 Mb interval on chromosome 8. The genomic interval of 21.78–21.22 Mb on chromosome 11 is associated with two QTLs namely qFLL11 and qPBPP11-2 for FLL in 2022 and PBPP in 2023, respectively. Potential QTLs selected for candidate gene analysis and manhattan plots showing significant SNP-trait associations single trait (a-d) and co-localized QTLs with two trait (e and f on same genetic loci while g and h on the same genetic loci): qSBPP4-2 QTL identified for SBPP during 2023 (a), qFLW6-2 QTL identified for FLW during 2023 (b), qGNPP9 QTL identified for GNPP during 2022 (c), qGWPP2-3 QTL identified for GWPP during 2023 (d), qFLA4-1 QTL identified for FLA during 2022 (e), qFLL4-1 QTL identified for FLL during 2022 (f), qFLW4 QTL identified for FLW during 2023 (g), and qFLA4-2 QTL identified for FLA during 2023 (h) Identification of candidate genes and expression analysis For candidate gene analysis, among the 97 QTLs, four genetic loci (S04_17090729, S06_11868554, S09_20551304, and S02_12214914) were selected based high phenotypic variation explained (≥ 10), and two genetic loci (S04_11552936 and S04_32100268) were selected based on potential pleotropism (Table 3). Based on the calculated LD decay (150 kb) of the population [49, 52], we took 100 kb upstream and downstream of each marker position as the target region for candidate gene identification. A total of 65 putative candidate genes were found in the target region of the six QTL locations (Table S2), and among them, three genes were associated with the protein kinase domain. Three QTLs associated with PA, such as qSBPP4-2 on chromosome 4, qGNPP9 on chromosome 9, and qGWPP2-3 on chromosome 2, revealed 5, 21, and 4 putative candidate genes, respectively (Table S2). For FL architecture, qFLW6-2 QTL on chromosome 6 revealed 10 putative candidate genes in the region. Five and twenty-five potential candidate genes for FL-related traits have been identified by two multi-trait genetic loci, namely S04_11552936 associated with qFLW4 and qFLA4-2 on chromosome 4 and S04_32100268 linked to qFLL4-1 and qFLA4-1 on chromosome 4 (Table S2). The Os04g0356600 gene, linked to protein kinase and its responsibility for SBPP, has been identified by the qSBPP4-2 QTL. The heat shock transcription factor family and zinc finger ring type protein are linked to the Os09g0526600 and Os09g0525400 gene, which has been found in the region of qGNPP9 QTL for GNPP on chromosome 9. The fasciclin-like arabinogalactan protein associated with Os02g0308400 gene was identified in qGWPP2-3 QTL, which is related to GWPP trait. Os04g0273100 and Os04g0276200 genes belong to the transposase protein and the pumilio homology domain protein, respectively. In addition, these genes were found from multi-trait genetic loci (S04_11552936) that are linked to the qFLW4 and qFLA4-2 QTLs. S04_32100268 genetic loci associated with qFLL4-1 and qFLA4-1 for FLL and FLA, respectively, on chromosome 4 was neighbored by two genes, Os04g0631100 and Os04g0631600. These genes are associated with important proteins like sugar transporter conserved site and transcription factor bHLH95-like basic helix-loop-helix domain, respectively, which help with genetic improvement for FL traits. RiceXPro tool was used for expression analysis of 65 identified candidate genes (Fig. S9 and S10) for the crop development and reproductive stages (Fig. 6), and potential candidate genes were identified for FLPA-related traits (Table 4). Based on expression levels, three genes were identified, including Os04g0273100, Os04g0631100, and Os04g0632100, which were associated with transposase, sugar transporter, and protein kinase, which is highly expressed in endosperm and palea development stages (Table 4). Heatmap for potential putative candidate gene expression analysis for FLPA-related traits through RiceXPro: putative candidate gene expression during reproductive organ development stage (a), putative candidate gene expression during early grain developmental stage (b), putative candidate gene expression during embryo as well as endosperm development stage, (c) and putative candidate gene expression throughout the entire crop growth (d). Red represents higher gene expression and green indicates lower gene expression level Table 4List of potential candidate genes associated with FLPA-related traits in rice that are highly expressed during the reproductive stageQTL nameMarker IDChromosomeGene stable IDTranscript stable IDGene start (bp)Gene end (bp)Protein description qSBPP4-2 S04_170907294 Os04g0356600 Os04t0356600-0117,044,34917,046,921Protein kinase domain Os04g0357300 Os04t0357300-0117,063,61717,065,887Similar to helicase-like protein Os04g0359100 Os04t0359100-0117,180,20717,183,428Polyketide synthase, enoylreductase domain qFLW6-2 S06_118685546 Os06g0311000 Os06t0311000-0011,840,61111,844,171Glycosyltransferase 61 qGNPP9 S09_205513049 Os09g0525400 Os09t0525400-0120,528,64820,532,222Zinc finger, RING-type Os09g0526700 Os09t0526700-0220,600,61620,603,003UDP-glucose 4-epimerase Os09g0526600 Os09t0526600-0220,591,25220,595,143Heat shock transcription factor family Os09g0525500 Os09t0525500-0120,534,16520,534,903Bifunctional inhibitor/plant lipid transfer protein/seed storage helical domain Os09g0527700 Os09t0527700-0120,641,46520,643,139AUX/IAA protein qGWPP2-3 S02_122149142 Os02g0308400 Os02t0308400-0212,113,90712,114,966Fasciclin-like arabinogalactan protein, group A Os02g0308800 Os02t0308800-0112,127,91312,129,233FAS1 domainqFLW4 and qFLA4-2S04_115529364 Os04g0273100 Os04t0273100-0111,463,29011,464,387Similar to transposase Os04g0276200 Os04t0276200-0111,650,39411,660,149Pumilio homology domainqFLL4-1 and qFLA4-1S04_321002684 Os04g0629400 Os04t0629400-0132,017,65632,018,383Uncharacterized protein Os04g0629400-like Os04g0630800 Os04t0630800-0132,085,61232,088,272Similar to anthocyanidin reductase Os04g0630900 Os04t0630900-0132,090,88032,093,039Similar to H0105C05.3 protein Os04g0631100 Os04t0631100-0132,139,08432,140,993Sugar transporter, conserved site Os04g0631200 Os04t0631200-0232,145,68132,147,446Glycoside hydrolase family 16 Os04g0632100 Os04t0632100-0132,178,83132,182,153Protein kinase, ATP binding site List of potential candidate genes associated with FLPA-related traits in rice that are highly expressed during the reproductive stage Discussion Understanding the genetic basis for FLPA is a critical step in developing molecular markers to enhance breeding efficiency and rice grain yield. Rice FL is the primary photosynthetic organ during grain filling, and its structural characteristics directly impact light interception, light dispersion, and energy usage. As a result, it is a major target trait for rice breeding programs to improve grain yield [68]. According to reports, FL photosynthesis supplies more than half of the carbohydrates in rice seeds [10]. The FLA is a critical factor in determining photosynthetic capability; it has two primary components, i.e., FLL and FLW, and is regulated by many QTL and their interactions with the environment. Similarly, PA-related traits (PL, PBPP, SBPP, GWPP, and GNPP) are complex quantitative traits regulated by multiple genes and controlled by the environment [14, 15, 69]. Both FL and PA-related traits are important components of improving rice yield. The diversity of genetic variation in rice germplasm accessions is an important and potential source of significant genetic loci. Few reports [14, 70–73] were published earlier on GWAY analysis using FLPA traits, however study was done using new germplasm phenotyped twice in a US environment to explore further and expand knowledge regarding the trait diversity. To further understand the genetic basis of FLPA and its importance in rice breeding, dissecting the QTL regulated by unique genomic locations and those important genomic region experess under various environmental factors is necessary. Our findings will help in the design of molecular breeding techniques for developing new rice cultivars as well as finding germplasms suitable for rice breeding and improvement. Phenotypic variations for targeted traits GWAS analysis results depend on the precise and appropriate phenotyping of the target traits. This study looked at two broad categories: FL and PA-related traits in two distinct environments. During 2022 and 2023, 208 diverse germplasm accessions’ data for all parameters showed normal distributions (Fig. 1), which was appropriate for GWAS analysis. We performed a Shapiro-Wilk test for normality, and it was determined that all study traits almost equal to normal distribution. Previously, the same population was used for a GWAS on key agronomic traits and quality parameters using data from 2018 to 2019 [49, 52]. Significant variability among the genotypes found in this study in both years for most studied traits was supported by previously published research [14, 74]. The mean value of all FLPA-related attributes decreased in 2022 compared to 2023 (Table 2), indicating that genetic and environment-related factors influenced these quantitative traits. The yield of rice grain is considered as an important polygenic trait influenced by many phenotypic traits, including FLPA, along with many environmental factors [75, 76]. FL-related (FLL, FLW, and FLA) traits CV ranged from 6.13 to 14.73 (Table 1), which was similar to the previous study [77]. For the PA (PL, PBPP, SBPP, GWPP, and GNPP) trait, CV ranged 6.72–36.74 (Table 1), which were higher than FL traits, these results were supported by earlier studies [70, 75, 78]. These results indicated that, being a quantitative nature of FLPA traits, FL was less affected by environmental factors compared to PA during both years. Most of the traits under study showed a positive correlation except SBPP (Fig. 2). For FL-realted traits, FLA showed a positive and significant correlation (r = > 0.7) with FLL and FLW in both years (Fig. 3). Panicle archititure traits, GNPP showed a positive correlation (between r = 0.2–0.6) with all traits except SBPP during 2022, 2023, and the mean of both years. These association analysis indicators effectively reflect the FL compared to panicle traits, which are beneficial for improving rice grain yield by GWAS, our results were also supported by earlier studies [72, 73]. Overall, phenotypic data revealed that several significant genetic regions may be responsible for FLPA-related attributes, and the researcher needs to further investigate distinct germplasm accession for GWAS analysis before reaching a final conclusion. Genome-wide association study for FLPA-related traits The speedy advancement of genome sequencing technology has made GWAS an effective tool for QTL analysis and has been utilized widely in rice research for analyzing complex features, including FLPA [12, 71, 73]. GWAS is a groundwork study that identifies genomic regions associated with FLPA-related traits. Using 854,832 polymorphic SNPs [49, 52] for this investigation, we were able to identify 97 QTLs (48 in 2022 and 49 in 2023) (Fig. S1-S8) out of which 24 were found based on > 3 models and were all significantly associated with all studied traits (Table S1). Previously, a GWAS study used a diverse set of rice germplasm accessions to identify FLPA related traits and revealed eight QTLs [79], 15 QTLs [73], and 64 QTLs [12] for FL-related traits, as well as 29 QTLs [40], 153 QTLs [14], 106 QTLs [71], and 49 QTLs [72] for PA-related traits. During 2022 and 2023, the present study found 25 and 72 QTLs for FL and PA-related traits, respectively (Fig. 4 and Table S1). Earlier research focused on FL and PA-related traits individually, but our objective was to combine both traits in a GWAS to discover suitable markers that would aid in developing rice cultivars with enhanced FLPA-related traits. Comparison of identified QTLs with previously published FLPA genes/QTLs Multiple studies explored diverse rice germplasms for QTLs related to FLPA traits. As a result, reviewing published literature to identify overlapped QTLs will validate the importance of specific genomic regions. From this study, 12 QTLs overlapped with previously reported QTLs. Four QTLs related to FLL were reported earlier (Fig. 4 and Table S1). Several of our QTLs overlap with previously reported QTLs/genes, revealing consistent genetic regions influencing these traits. For example, qFLL2-1 coincided with previously identified QTLs TC80 [68] and qFLL2.1 [80]. Another QTL qFLL4-1 is linked to the single sequence repeat (SSR) markers RM1112 [68] and RM_5709 [81], which were previously reported to be significantly associated with FLL. Further, qFLL8 identified from our study was also reported previously [31]. Three QTLs linked with FLW were also documented earlier. For example, qFLW9 identified in this study overlapped with TC136 [68], qFLW10-1 was reported as qFWn10-1 [12], and qFLW10.1 [80] in previous studies. These results will help narrow down the identified QTLs for fine-mapping FL-related traits. Likewise, some QTLs controlling PA were also reported earlier (Fig. 4 and Table S1). For example, qSBPP4-3 QTL controlling SBPP overlapped with qSBN-4 [72]. Also, four QTLs related to PL were documented previously in multiple studies. For example, qPL1 identified in our study overlapped with qPL1-8 [14], qPL1 [82], and qPl1 [13], qPL4-1 overlapped with RM551 [81], qPL7-1 and qPL7-3 overlapped with qPL7-7 [14] and RM_234 [81], respectively. Future research should focus on comparing identified QTLs with overlapping QTLs to refine genetic markers, discover new functional mechanisms, and integrate omics data. This will improve trait understanding, boost breeding efficiency, and expand applications across diverse rice germplasms. Putative candidate genes and expression analysis To further refine the potential identification of QTLs, six specific genetic loci were selected for candidate gene analysis (Table 3; Fig. 5). This selection aims to narrow down the potential candidate genes associated with the QTLs by focusing on these loci, allowing for a more precise investigation into their roles and effects on the traits of interest. With a resolution of 100 kb radius genomic regions at the identified QTL locations, a minimal number of possible genes were identified; a similar approach was also utilized to identify the candidate genes in previously published research [83–85]. Of the 97 QTLs, four were chosen based on phenotypic variation (> 10%) and two for their pleiotropic effects on multiple traits for candidate gene analysis (Table 3). A total of 65 potential putative candidate genes were identified from Ensemble plants (Table S2). We identified 35 candidate genes for FL and 30 for PA traits. These putative QTLs give the foundation information for future studies to identify the FLPA-related genomic region from diverse rice germplasm. Protein kinase plays an essential role during panicle development, especially panicle apical spikelets [24]. In the present study, one of the SBPP trait controlling the QTLs, qSBPP4-2, which is identified as Os04g0356600 gene and is associated with protein kinases. The protein heat shock transcription factor family has the highest levels of gene expression in reproductive organs, where it regulates the embryo and lemma to enhance rice grain yield [25]. Similarly, this protein has been linked to spikelet fertility in rice [86] and anther development in wheat [87]. Zinc-finger ring-type protein expressed during the spikelet development in rice [88]. From qGNPP9 QTL, we identified the Os09g0526600 and Os09g0525400 genes, which are linked to the heat shock transcription factor family and zinc finger ring type protein, respectively. The sugar transport protein family plays an important role in sink tissue (the major site of carbohydrate storage in leaves), which helps with grain filling in rice [26] and different stages of plant development [27, 28]. Our study also found the Os04g0631100 gene for qFLL4-1 and qFLA4-1 QTLs, which are associated with the sugar transporter protein. In rice, the bHLH (basic helix-loop-helix) protein family influences FL shape and angle [29, 79]. For qFLL4-1 and qFLA4-1 QTLs, our study identified the Os04g0631600 gene that corresponds to the same protein family and directly contributes to improving FL traits. Plant molecular biology uses several strategies that are improved by gene expression profiling, especially when characterizing gene function [89]. In the current study, we used the gene expression profile database, i.e., RiceXPro (https://ricexpro.dna.affrc.go.jp/index.html), to find more comprehensive information on the rice transcriptome, encompassing the entire growth cycle as well as the reproductive stage at field experimental conditions [45, 67, 89]. A total of 65 potential putative candidate genes were identified in this study (Table S2; Fig. S9 and S10) throughout the entire crop growth period of crop. Among them, 15 candidate genes (Fig. 6) were highly expressed during the reproductive stage (Table 4). Os04g0631100 gene was highly expressed during the endosperm development stage and was also associated with sugar transporter protein; further, this gene needs to be validated for functional characterization. Many of the previous studies used RiceXPro to find the potential candidate gene for different traits like reproductive traits [48], roots-related traits [47], and important agronomic traits [46]. In silico expression analysis was chosen for validating gene expression during the reproductive stage because it efficiently screens large datasets to identify candidate genes at this stage. All things considered, the QTLs and potential candidate genes identified in this research suggest that several unexplored factors could play a role in determining FLPA-related phenotypes and be helpful in molecular marker development and marker-assisted selection in rice breeding programs. Conclusion Based on independent GWAS analysis from both years, we identified 97 QTLs with significant associations with all studied traits. Among these, 24 QTLs were found significant in at least three multi-locus models and four of which (qSBPP4-2, qFLW6-2, qGNPP9, and qGWPP2-3) showed high phenotypic variance explained ranging from 11.72 to 22.30%, and two genetic loci were linked to multi-trait QTLs. Overall, 12 QTLs were found to be reported in previous studies for FLPA traits, and among them, seven QTLs were associated with FL-related traits, and five QTLs with PA. Two multi-trait QTLs, S04_32100268 (qFLL4-1 and qFLA4-1) and S04_11552936 (qFLW4 and qFLA4-2) were found associated with 65 putative candidate genes. In silico expression analysis by RiceXPro revealed that 15 candidate genes were highly expressed during the reproductive stage. This approach is essential for FLPA traits, as reproductive stage-specific expression significantly affects fertility and contributes to increased grain yield, thus providing important insights for rice breeding strategies. Among 15 candidate genes, five putative candidate genes i.e., Os04g0356600, Os09g0526600, Os09g0525400, Os04g0631100, and Os04g0631600, were associated with protein kinase, heat shock transcription factor family, zinc-finger ring-type, sugar transporter conserved site, and transcription factor bHLH95-like basic helix-loop-helix domain, respectively, regulating different FLPA-realted traits. These genes will be the focus for future research into their possible roles in the rice breeding program. Supplementary Information Supplementary Material 1: Table S1. Summary of 97 QTLs associated with flag leaf and panicle architecture-related traits during 2022 and 2023. Supplementary Material 2: Table S2. Summary of 65 putative candidate genes located in the six potential genetic loci intervals. Supplementary Material 3: Fig. S1. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf length: During 2022 (a), and during 2023 (b). Supplementary Material 4: Fig. S2. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf width: During 2022 (a), and during 2023 (b). Supplementary Material 5: Fig. S3. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf area: During 2022 (a), and during 2023 (b). Supplementary Material 6: Fig. S4. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with panicle length: During 2022 (a), and during 2023 (b). Supplementary Material 7: Fig. S5. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with primary branch per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 8: Fig. S6. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with secondary branch per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 9: Fig. S7. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with grain weight per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 10: Fig. S8. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with grain number per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 11: Fig. S9. All QTLs name should be in italic format likeq SBPP4-2, qFLW6-2, qGNPP9, and GWPP2-3. Supplementary Material 12: Fig. S10. All QTLs name should be in italic format likeq FLW4, qFLA4-2, qFLL4-1 and qFLA4-1. Acknowledgements Texas Rice Research Foundation (TRRF) and Texas A&M AgriLife Research Center at Beaumont. Authors’ contributions GS analysed data and drafted the manuscript. SKT conceptualized and designed the experiment, helped GS in interpreting the results, edited the final manuscript and supervised the study. PU, RC, PE and CLH collected data for the experiment. SDJ helped GS in analyzing data and drafting the manuscript. SM, KM, and NW supervised PU and RC, interpreted results and edited the manuscript. SOPBS and DLS designed, planted, managed the field experiment, and edited the manuscript. All authors read and approved the final manuscript. Funding Funding was provided by the Texas Rice Research Foundation and Texas A&M AgriLife Research. Data availability The datasets generated and/or analyzed during the current study are not publicly available due to sensitivity but are available from the authors on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1: Table S1. Summary of 97 QTLs associated with flag leaf and panicle architecture-related traits during 2022 and 2023. Supplementary Material 2: Table S2. Summary of 65 putative candidate genes located in the six potential genetic loci intervals. Supplementary Material 3: Fig. S1. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf length: During 2022 (a), and during 2023 (b). Supplementary Material 4: Fig. S2. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf width: During 2022 (a), and during 2023 (b). Supplementary Material 5: Fig. S3. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with flag leaf area: During 2022 (a), and during 2023 (b). Supplementary Material 6: Fig. S4. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with panicle length: During 2022 (a), and during 2023 (b). Supplementary Material 7: Fig. S5. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with primary branch per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 8: Fig. S6. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with secondary branch per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 9: Fig. S7. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with grain weight per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 10: Fig. S8. Manhattan and quantile-quantile (Q-Q) plots showing significant marker trait associations with grain number per panicle: During 2022 (a), and during 2023 (b). Supplementary Material 11: Fig. S9. All QTLs name should be in italic format likeq SBPP4-2, qFLW6-2, qGNPP9, and GWPP2-3. Supplementary Material 12: Fig. S10. All QTLs name should be in italic format likeq FLW4, qFLA4-2, qFLL4-1 and qFLA4-1. Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available due to sensitivity but are available from the authors on reasonable request.