Introduction

To meet the global food requirements by 2050, an average annual increase of 44 million tons in food production is necessary (Tester and Langridge 2010). The scarcity of water resources has exacerbated the food shortage situation, and enhancing the drought tolerance (DT) of crops is an effective method to ensure food security (Hu and Xiong 2014). Rice (Oryza sativa L.) not only feeds more than half of the world's population but is also an important model plant in cereals (Xing and Zhang 2010). Drought stress is a main abiotic stress which restricts rice growth and productivity (Singhal et al. 2016). So, enhancing drought tolerance in rice is of great importance.

Enhancing DT in crops is a challenging process, given its complexity, which involves various physiological and molecular responses influenced by multiple alleles with minor effects (Blum 2011; Fukao and Xiong 2013). To comprehend the genetic basis of DT in rice, researchers have utilized quantitative trait locus (QTL) mapping with recombinant inbred lines (RILs) populations (Yue et al. 2006). Yield-associated traits and visible scores for plant performance during or after water deficit conditions are commonly used to assess DT, however; a limited number of QTLs were repetitively identified in various populations or different years or environments. Drought stress significantly reduces grain yield (GY), and identifying trustworthy loci related to DT using GY is challenging, considering that grain yield is influenced by many genes with minor effects and numerous uncontrolled environmental factors in the field (Guo et al. 2018a).

QTL mapping has been widely employed as an influential statistical approach to detect genomic regions related to important traits for breeding (Wang et al. 2019). Numerous QTL-based studies have been conducted on different populations for several DT-associated traits and GY components such as water use efficiency (Zhou et al. 2013), carbon isotope discrimination (Takai et al. 2006), canopy temperature (Prince et al. 2015), flag leaf size (Yue et al. 2008), heading date and delay in flowering (Trijatmiko et al. 2014), drought response index (Kim et al. 2017), leaf drying (Michael Gomez et al. 2010), grain number per panicle (Baisakh et al. 2020), biomass yield (Dixit et al. 2015) and plant height, number of tillers per plant, leaf rolling, leaf drying, harvest index, spikelet fertility, and relative water content (Barik et al. 2019).

Different positions of a QTL in various mapping populations result in a immense confidence interval and an unreliable QTL position. Additionally, several factors such as differences in mapping population size, sampling errors, marker density, experimental replicates and QTL mapping models can further complicate the situation (Darvasi and Soller 1997; Darvasi et al. 1993). Different methods have been applied till now to validate the QTL results, like QTL mapping utilizing first-generation populations, and confirmed in advanced-generation breeding populations of the same cross (Gelli et al. 2017). Furthermore, QTL validation is accomplished using the candidate gene method or positional cloning, followed by incorporating functional and genetic data within the breeding process (de Dorlodot et al. 2007). However, this is a challenging process that requires high-density linkage maps, extensive genomic resources and logical informatics data (de Dorlodot et al. 2007).

Meta-analysis is a statistical method that combines consensus loci from various QTL studies for multiple traitsinto a single dataset to determine the most probable position and confidence interval (CI) of QTL regions (Loni et al. 2023; Bilgrami et al. 2023). This approach has been applied to identify genomic consensus regions over various QTL studies considering their effects and constancy across different genetic backgrounds and environments. Moreover, it can enhanceand validate QTL positions on a consensus map through mathematical models. The identified consistent QTL for a set of QTLs with a CI of 95% via meta-analysis is called meta-QTL (MQTL) (Swamy et al. 2011). Another important advantage of meta-analysis of QTLs is its ability to decrease the CI of the MQTLs compared to QTLs. MQTL analysis determines the most constant QTLs irrespective of the population’s genetic background and field trial conditions, and it efficiently decreases the CI for identifying candidate genes and developing markers (Bilgrami et al. 2023; Khahani et al. 2021). Therefore, QTL-based meta-analysis provides more precise and stronger results. In addition, meta-analysis of QTLs provides a perception into the genetic analysis of complicated traits like drought, salt and heat response. Meta-analysis of QTLs has been applied to accurately evaluate many agronomical traits in different crops (Khowaja et al. 2009).

Several studies have been conducted for the meta-analysis of QTLs controlling grain yield components and DT-related traits under water deficit conditions in rice. Some MQTLs have been reported for plant height througha meta-analysis of published QTLs up to2009 (Khowaja et al. 2009). Swamy et al. (2011) projected 53 grain yield QTLs, reported in 15 studies under drought stress, on a consensus map and conducted a meta-analysis that resulted in the identification of 14 MQTLs on 7 chromosomes. They showed that a grain yield MQTL under drought coincided with at least one of the MQTLs discovered for root and leaf morphology in previous studies (Swamy et al. 2011). In another study, MQTL-analysis for grain yield and yield component under drought stress resulted in identification of a GY MQTL in a region close to the semi dwarf gene (sd1) locus on chromosome 1 which co-localized with QTLs for leaf rolling and osmotic adjustment (OA) (Trijatmiko et al. 2014). In addition, a QTL for percent seed set and grains per panicle under drought stress was discovered on chromosome 8 in the region where a QTL for OA was reported in previous studies (Trijatmiko et al. 2014). Yang et al. (2018a, b) reported some MQTLs for heading date through a meta-analysis of published QTLs till 2018 (Yang et al. 2018a). Khahani et al. (2021) conducted a meta-analysis of 536 QTLs related to yield and yield-associated traits such as yield (YLD), grain weight (GW), heading date (HD), plant height (PH) and tiller number (TN) plus root-architecture related traits under drought stress conditions. They identified 61 stable MQTLs across different genetic backgrounds and environments (Bilgrami et al. 2023; Khahani et al. 2021). Abdirad et al. (2022) combined root tip transcriptome sequencing and meta-analysis of QTLs to find the main genes engaged in drought stress response in rice (Abdirad et al. 2022). In addition, for complex quantitative traits, GWAS is widely employed to identify significant effects of genomic loci. It has been indicated that combining meta-analysis of QTLs and GWAS data can lead to dissecting important genomic regions and the genetic foundation of important quantitative traits (Bilgrami et al. 2023; Bilgrami et al. 2020; Daryani et al. 2022). Furthermore, an integrated approach of meta-QTL analysis was used to identify the genomic regions and candidate genes related to drought tolerance and yield-related traits in foxtail millet (Loni et al. 2023).

In the current study, a meta-analysis of 1,087 QTLs controlling DT and yield-related traits under drought stress conditions in rice was performed. The QTLs were gathered from QTL mapping studies conducted under drought stress in rice from 2000 to 2021. After mapping the 1,087 QTLs onto the consensus genetic map, the density of QTLs, described as the “QTL-overview index”, was computed for the considered interval of 0.5 cM on each chromosome to detect genomic regions significantly associated with yield and yield-related traits under drought stress conditions. Overlap between MQTLs identified using meta-analysis and SNPs identified using the GWAS technique for yield and yield-related traits under drought stress conditions was examined to select candidate genes. In our study, in addition to conducting a distinct MQTL analysis for each trait, an inclusive MQTL analysis was performed on all the investigated traits to identify and introduce hotspots for breeding programs. The consensus genomic regions identified by meta-analysis of QTLs were subsequently confirmed by the GWAS studies. The genes located within the MQTLs were found and categorized based on their function. Furthermore, the rice drought-responsive genes were identified by the RNA-seq and microarray datasets analysis, and the MQTL regions related to yield and yield-related traits were searched to identify the drought-responsive genes. Conclusively, the integration of QTLs, GWAS, and transcriptome data has facilitated the detection of the promising MQTLs and candidate genes. These findings would be utilized in MQTL-assisted breeding to improve yield potential under drought stress in rice.

Materials and Methods

Compilation of QTLs Linked to Drought Tolerance and Yield Associated Traits

All the publications reporting QTLs associated with DT and yield-related traits under drought treatment in rice from 2000 to 2021 were reviewed. Totally, 1,388 original QTLs belonging to 21 different traits from 134 bi-parental rice populations extracted from 76 studies, of which 1,087 QTLs were utilized for the meta-analysis. Out of this dataset, 1,087 QTLs (Table 1) were considered for the meta-analysis, which had the necessary information such as phenotypic variance, population size, etc. Moreover, the QTLs with a large confidence interval and small phenotypic variance were removed. Table 1 provides information on the parents used in the populations, the type and size of the populations, the markers used for genotyping (including AFLP, SSR, SNP, and RFLP), and the number of primary QTLs. The original QTLs were classified into 10 trait categories, including biomass yield (BY), canopy temperature (CT), drought response index (DRI), flag leaf size (FLZ), grain yield (GY), heading/days to flowering (HD), harvest index (HI), plant height (PH), panicle number (PN), and spikelet fertility (SF) (Additional file 4: Table S1).

Table 1 Brief of the QTL mapping studies used in meta-analysis of the QTLs for yield components and drought tolerance-associated traits in rice
Full size table

Consensus Genetic Map

The most comprehensive genetic map, developed by Wu et al., (2016) was used as a reference map for the meta-analysis of QTLs. This map integrated different types of markers such as SSR, RFLP and AFLP, from six rice saturated maps, and contained 6,970 markers spanning 1,823.1 cM with a genetic distance between markers ranging from 0.19 to 0.5 cM on all the chromosomes (Wu et al. 2016) (Additional file 5: Table S2). In order to incorporate those initial QTLs with SNP markers (Table 1) into the reference map, we employed our previous method (Daryani et al. 2022) in which the genomic positions of SNP markers on the rice genome were detected and the closest markers based on the physical position were utilized to project them on the reference map.

Projection of QTLs into the Consensus Map

To project the QTLs on the reference map, the LOD (Logarithm of the odds) score, the phenotypic variation explained by the QTL (R2), the closest or flanking markers, and the position of the QTL-linked markers were used. A simple scaling rule was used to project the QTLs based on the consensus map, which involved the interval of the markers flanking the original QTL and the relevant interval on the chromosome. The projection was done using a Gaussian mixture model-based algorithm to estimate the new CI of a QTL on the consensus map. In the research, BioMercator v4.2 (http://moulon.inra.fr/) was used for meta-analysis of QTLs. The formula CI = 530/(N × R2) was used to calculate the 95% CI for QTLs obtained from backcross (BC) and F2 populations, where N is the population size and R2 is the proportion of phenotypic variance described by a QTL (Darvasi and Soller 1997). For QTLs obtained from doubled haploid (DH) and recombinant inbred (RI) lines, the formulas CI = 287/(N × R2) (Visscher and Goddard, 2004) and CI = 163/(N × R2) (Guo et al., 2006) were used to compute the 95% CI, respectively.

Meta-QTL Analysis and QTL-Overview Index

Integrated QTLs on the consensus map were used to conduct MQTL analysis using BioMercator V4.2 (Arcade et al. 2004). Two types meta-analysis of QTLs were conducted; 1: Distinct MQTL analysis: individual trait-based analysis for 10 traits separately (Additional file 4: Table S1), 2: Inclusive MQTL analysis: a comprehensive analysis using all the original QTLs associated with drought tolerance and yield-related traits under drought stress. Two different methods were applied for MQTL analysis based on the number of primary QTLs. When the number of the primary QTLs was fewer than 10 for a chromosome, the method suggested by Goffinet and Gerber was used (Goffinet and Gerber 2000). Based on this approach, the model with the minimum AIC value was chosen for integrating QTLs and identifying MQTL positions. When the number of primary QTLs for an individual chromosome was at least 10, method recommended byby Veyrieras et al. was utilized (Veyrieras et al. 2007). This method used five criteria (AIC, AICc, AIC3, BIC, and AWE) for selecting the number of potential MQTLs on a chromosome. The model with the minimum value of three criteria out of five was selected as the best model, and the 95% CIs and MQTL positions were defined based on the chosen model. QTLs integration was done such that the peak position of the primary QTLs lay within the confidence interval of the MQTLs. QTLs with a membership probability greater than 60% for an MQTL were assigned to the same MQTL (Chardon et al. 2004). The “QTL-overview index” method was used to estimate the probability of QTL for 0.5-cM-long segment of the reference map (Daryani et al. 2022; Chardon et al. 2004).

Identification of the Genes Located Within the MQTL Regions

The flanking markers were mapped onto the Oryza sativa Japonica group (IRGSP-1.0) reference genome (Kawahara et al. 2013) to determine their physical position. Finally, BioMart tool on the Ensemblplants website (https://plants.ensembl.org/biomart/martview/) was employed to find the genes placed within the MQTL regions (Additional file 6: Table S3 and Additional file 7: S4).

Dataset Collection and Gene Expression Analysis

Differentially expressed genes (DEGs) under drought stress in rice were obtained from related microarray (11 published articles) and RNA-seq (13 published articles) data (Additional file 8: Table S5) available at https://www.ncbi.nlm.nih.gov. The genes having − 1 ≥ log2 fold change ≥ 1 and p-value ≤ 0.05 were identified as DEGs. The drought responsive genes placed within the MQTL regions were identified using Venn diagram (Venn diagram was drawn using a tool on this website: http://www.interactivenn.net/ (Heberle et al. 2015)) (Additional file 9: Table S6, Additional file 1: Fig. S1).

Comparison of the MQTLs with DT-Associated Genome-Wide Association Studies (GWAS)

Reviewing DT-related GWAS studies (Bhandari et al. 2020; Courtois et al. 2013; Guo et al. 2018b; Kadam et al. 2018; Liang et al. 2016; Ma et al. 2016a; Pantaliao et al. 2016; Pariasca-Tanaka et al. 2020; Sandhu et al. 2019; Su et al. 2021; To et al. 2019; Zhang et al. 2021) was done to collect the reported SNP peak positions and discover the overlaps between their positions with MQTLs (Additional file 10: Table S7). Using the physical positions on the Oryza sativa Japonica group (IRGSP-1.0) reference genome, the genes placed within the SNP peak positions (± 25 kb) were identified.

Graphical Representation

To visually represent the data, the Circos software (Krzywinski et al. 2009) were utilized to create a comprehensive graphical summary of the original QTLs, MQTLs, and QTL-overview statistics on all 12 rice chromosomes. Furthermore, additional graphs were generated using the ggplot2 R package.

Results

Collection of Original QTLs Associated with DT and Yield-Related Traits Under Drought Stress in Rice

A sum of 1,087 QTLs related to DT and yield-associated traits under drought stress in rice were collected which were from 134 different rice populations, including F2 (2 populations), BC (11 populations), RILs (30 populations), DH (13 populations), BAC (1 population), BIL (21 populations), BSA (4 populations), F2:3 (3 populations), F3 (1 population) and NILs (1 populations), with population size ranged from 50 (Baghyalakshmi et al. 2016) to 485 (Swamy BP et al. 2013) (Additional file 11: Table S8, Table 1). Additional file 11: Table S8 presents the complete information about the original QTLs including the QTL name, trait controlled by the QTL, related chromosome, LOD score, phenotypic variance described by the QTL, left and right flanking markers, interval between left and right markers (cM), parents of the mapping populations, type of the population, population size, number of used markers, location where the experiment was conducted and the reference.

Analyzing the distribution pattern of these 1,087 QTLs across the rice chromosomes revealed interesting insights. chromosome 1 hosted the highest number of QTLs (176), followed by chromosomes 3 (155), 6 (128), 2 (122), 4 (104), 8 (98), 9 (67), 5 (53), 12 (50), 7 (46), 11 (44), and 10 (44), respectively (Fig. 1 and Table 2). The highest number of the initial QTLs related to grain yield (GY) (304), followed by biomass yield (BY) (151), heading/days to flowering (HD) (128), plant height (PH) (119), drought response index (DRI) (109), panicle number (PN) (102), spikelet fertility (SF) (71), harvest index (HI) (48), flag leaf size (FLZ) (38) and canopy temperature (CT) (21), respectively (Table 2). These QTLs exhibited 95% CI spanning from 0.43 to 50.40 cM, with an average of 12.36 cM. Notably, nearly 30% and 85% of the original QTLs had CIs of less than 10 cM and 20 cM, respectively (Fig. 2a). In terms of the phenotypic variance described by the investigated QTLs, there was considerable variability, ranging from 1.1% to 85% (Fig. 2b). Each trait's original QTLs were ranked based on the proportion of phenotypic variance they accounted for (Fig. 2c). Remarkably, 52.7% of the original QTLs (585 out of 1,087) described more than 10% of the phenotypic variance, while 47.2% of them (524 out of 1,087) had a PVE of less than 10% (Fig. 2b). Specifically, out of the 304 QTLs linked to GY, a total of 173 QTLs surpassed the 10% threshold for PVE (Fig. 2c).

Fig. 1
figure 1

Dispersion pattern of the original QTLs related to yield components and drought tolerance associated traits under drought stress on the 12 rice chromosomes in terms of number

Full size image
Table 2 Distribution of the original QTLs used for the MQTL analysis on 12 rice chromosomes
Full size table
Fig. 2
figure 2

Brief of the 1,087 original QTLs associated with yield components and DT-associated traits utilized under drought stress for MQTL analysis. a The frequency dispersion of the original QTLs density according to diverse levels of the 95% confidence interval. b Original QTLs distribution based on the amount of the phenotypic variance explained by them, c The percentage of original QTLs with different phenotypic variance explained (R2 ≤ 5, 5 < R2 < 10 and R2 ≥ 15) for each of the yield components and DT-associated traits, d The dispersion of original QTLs and MQTLs on rice chromosomes. BY: biomass yield, CT: canopy temperature, DRI: drought response index, FLZ: flag leaf size, GY: grain yield, HD: heading/days to flowering, HI: harvest index, PH: plant height, PN: panicle number, SF: spikelet fertility

Full size image

Identification of MQTLs Controlling Yield Components and DT-Related Traits by Meta-Analysis

The collected original QTLs were categorized into 10 major traits and subjected to MQTL analysis to detect MQTLs controlling each trait associated with DT and yield components under drought stress in rice. The meta-analysis resulted in the identification of 213 MQTLs with at least eight MQTLs on each chromosome (Fig. 2d; Table 2 and 3). The identified MQTLs included 50 MQTLs for GY, 29 MQTLs for BY, 27 MQTLs for HD, 25 MQTLs for PN, 23 MQTLs for PH, 24 MQTLs for DRI, 13 MQTLs for SF, 10 MQTLs for FLZ and HI and two MQTLs for CT. MQTL_GY1.4 was recognized as the most stable MQTL with the greatest number of original QTLs (22) (Fig. 3; Table 3). In addition, the results indicated that 29 MQTLs out of the detected 213 MQTLs were involved in the genetic control of more than one trait (Additional file 12: Table 9). These MQTLs were distributed across the rice chromosomes, with the number of MQTLs per chromosome varying, ranging from 29 MQTLs on chromosome 1 to eight MQTLs on chromosome 11 (Fig. 2d; Tables 2 and 3). The results indicated that all the chromosomes of rice were engaged in controlling yield components and DT-related traits. The number of initial QTLs grouped in a MQTL varied from two to 22 QTLs (Table 3). The proportion of phenotypic variance explained by these MQTLs displayed a range from 3.05% to 70.1%, with an average of 12.76% (Table 3). When it comes to the 95% CIs for these identified MQTLs, they spanned from a minimal 0.12 cM for the RM482-RM213 interval on chromosome 2 to a maximal 19.7 cM for the E2801S-HSP70A interval on chromosome 5 (Table 3).Notably, a narrower confidence interval was observed for each MQTL when compared to the average CI of the original QTLs clustered within that specific MQTL. Specifically, the confidence interval of 17 MQTLs was diminished to less than 1 cM, effectively reducing the length of these MQTLs by approximately 15.66 times when compared to the average CI of the original QTLs. The mean phenotypic variance for these 17 MQTL regions was calculated at 20.3. It's worth noting that the flanking markers of these aforementioned MQTLs present as promising candidates for deploying molecular breeding and marker-assisted selection strategies aimed at bolstering drought tolerance in rice (Table 3).

Table 3 Summary of the identified trait specific MQTLs. The initial QTLs for yield components and drought tolerance associated traits under drought stress in rice were categorized into ten major traits* and subjected to meta-QTL analysis independently
Full size table
Fig. 3
figure 3

Concentric circles indicate different features drawn in Circos (Krzywinski et al. 2009). a Genetic positions (cM) of rice chromosomes represented by bars. b Molecular markers density on rice chromosomes shown on a scale from white to orange to indicate the lowest to highest density. c Distribution of QTLs across the twelve rice chromosomes. d Genetic positions of MQTLs for each distinct trait with 95% CIs. e Heatmap illustrating the QTL-overview index, which is estimated by the frequency of QTLs for yield components and DT-associated traits on each 0.5-cM segment of the rice consensus map. f Genetic positions of inclusive MQTLs with 95% CIs. g Proportion of phenotypic variance explained (R2) by each QTL

Full size image

Identification of DEGs Involved in Drought Stress Response of Rice

RNA-seq and microarray datasets were used to identifydrought-responsive genes in the rice (Additional file 8: Table S5). In total, 1.4814 and 23.722 genes were found to be drought responsive by RNA-seq and Microarray analysis (Additional file 13: Tables S10 and Additional file 14: S11 and Additional file 1: Fig. S1). On the other hand, MQTL regions identified to control DT-associated traits and yield components were investigated to discover the genes placed in those regions. Comparing the DEGs discovered by RNA-seq and microarray analysis with genes located within all 213 identified MQTL regions using Venn diagrams revealed the presence of 6,877 common genes. Furthermore, 375 genes were shared among the DEGs identified by RNA-seq and microarray analysis and genes within the 17 MQTLs with CI < 1 cM (Additional file 1: Fig. S1). The mentioned genes are recognized as differentially expressed candidate genes due to being both drought responsive and locating in MQTL regions.

Estimating QTL-Overview Index for DT-Associated QTLs in Rice and Validating MQTLs Using GWAS Studies

To strongly associate genomic regions with yield components and DT associated traits in rice, the QTL-overview index was estimated. The QTL-overview index, representing QTL density, was calculated for every 0.5 cM segment on each chromosome (Additional file 2: Fig. S2). The findings revealed that 213 overview index peaks exceeded the genome-wide mean value (Chardon et al. 2004), showing the presence of real QTLs controlling yield components and DT-associated traits in rice (Additional file 2: Fig. S2). In addition, 113 peaks (out of the 213 peaks) were higher than the high-value threshold (Chardon et al. 2004) and were considered as QTL hotspots (Additional file 2: Fig. S2).

Significant concurrences were observed between the MQTLs identified through a comprehensive analysis and the SNPs uncovered via the GWAS approach concerning traits associated with DT within the rice genome. Notably, among the total of 213 identified MQTLs, 63 were found to be colocated with 130 peak SNP positions, which were detected through the GWAS approach for traits associated with DT in rice (Additional file 10: Table 7). Totally, 765 rice genes were identified in the SNP peak positions (± 25 kb) overlapped with MQTLs. Furthermore, 61 SNP peak positions (out of the 130 SNP peak positions) overlapped with high-overview-index MQTLs (QTL hotspots). In this study, we identified novel candidate genes among those situated within the significant SNP peaks and regions characterized by high QTL-overview indices associated with yield components and DT traits. Specifically, we discovered nineteen genes within SNP peak positions that coincided with QTL hotspots, thereby designating them as novel candidate genes (Fig. 3, Additional file 10: Table S7).

Conclusively, we integrated the results achieved by the meta-analysis of QTLs, GWAS studies, and the transcriptome data analysis leading to the discovery of 231 candidate genes (Additional file 3: Fig. S3, Additional file 15: Table S12), which might play key roles in rice DT and yield-associated traits under drought stress. In addition, 9 genes were common between GWAS studies results for DT and yield-related traits, DEGs discovered by RNA-seq and microarray analysis and the genes placed in the 17 MQTLs having CI < 1 cM (Additional file 3: Fig. S3, Additional file 15: Table S12). Following functional analysis, the identified candidate genes hold promise for applications in genetic engineering efforts targeting improvements in yield potential, stability, and performance under water deficit conditions, specifically focusing on DT and yield-related traits.

Inclusive Meta-Analysis of Traits, QTL-Overview Index and Identification of Breeding MQTLs

Totally, 96 MQTLs were detected by inclusive meta-analysis of all the 1,087 original QTLs related to yield components and DT-associated traits in rice. The number of the detected MQTLs on each chromosome varied from 13 MQTL on chromosome 3 to 4 MQTLs on chromosome 9 (Fig. 3f; Table 4). The CI of the identified MQTLs was between 0.01 and 9 cM, having a mean of 2.33 cM, which was 4.66 times narrower than the CI average of the initial QTLs. Only 8 MQTLs (out of the 96 identified MQTLs) had a confidence interval > 5 cM. The CI values of 52 and 25 MQTLs were less than 2 and 1 cM, respectively (Table 4, Fig. 3). The number of the initial QTLs clustered in a MQTL ranged from 2 to 59 QTLs (Table 4). Forty MQTLs included more than 10 initial QTLs. For 71 MQTLs, the PVE mean of the initial QTLs was higher than 10 (Table 4, Fig. 3). There were 59 common MQTLs between the obtained results by inclusive meta-analysis for all the traits and meta-analysis for each trait (Fig. 3).

Table 4 Summary of the identified general MQTLs for yield components and drought tolerance associated traits under drought stress in rice
Full size table

After mapping the 1,087 QTLs on the consensus genetic map, the QTL density, described as the "QTL-overview index", was calculated for a 0.5 cM distance on each chromosome to find genomic regions significantly related to the studied traits (Additional file 2: Fig. S2 a). Ninety-six overview index peaks were attained, which were exceeded 0.0043 as the average of the genome-wide statistic and indicated the “Real QTLs” affecting all yield traits in rice. Based on the Additional file 2: Fig. S2 a, out of the 96 peaks considered as “Real QTLs”, 49 peaks overpassed 0.0219 as the high-value threshold and thus can be considered as “QTL hotspots”.

According to the criteria presented by Löffler et al. 2009, an MQTL with a narrow confidence interval, a high number of the original QTLs and a high PVE value for the original QTLs are considered suitable for marker-assisted breeding (Löffler et al. 2009). Based on the achieved results, 13 MQTLs with CI < 1 cM, sum of the original QTLs > 10 and mean PVE value of the original QTLs > 10, identified using the inclusive MQTL analysis of all the investigated traits, can be considered as “Breeding MQTLs” (Table 5, Fig. 3). Interestingly, these thirteen MQTL regions are located in the QTL hotspots.

Table 5 The MQTLs with more than 10 initial QTLs, CI < 1 cM and a PVE average of the original QTLs > 10, which were considered as “Breeding MQTLs”
Full size table

Candidate Genes with Inclusive/Distinct Analysis of Traits

Combining the results of MQTL analysis for yield and DT-related traits (both inclusive and distinct analysis of traits), GWAS studies, and transcriptome data resulted in identification of 82 candidate genes (Fig. 4 and Additional file 16: Table S13). Among these candidate genes, two genes including Os02g0700700 (OsBRXL2) and Os04g0431200 (OsDRF1) are placed both on the SNP peak positions and in the high-overview-index MQTLs i.e. MQTL_BY2.2 and MQTL_GY4.2, respectively (Additional file 10: Table S7).

Fig. 4
figure 4

Venn diagram showing the genes placed in the detected MQTLs through inclusive MQTL analysis on all the traits (orange) and distinct MQTL analysis for each individual trait (green), genes placed in SNP peak positions based on GWAS studies (pink) for yield and DT-associated traits under drought stress conditions, and the drought responsive genes based on RNA-seq (blue) and microarray (yellow) data

Full size image

Discussion

Candidate MQTLs/Genes for Yield Maintenance Under Drought with Potential Application in Breeding Programs

One effective statistical approach for accurately detecting QTLs that control yield components and DT-associated traits in the genome could involve conducting a meta-analysis on a large number of independent QTLs associated with these traits. In this particular study, a meta-analysis was performed on 1,087 QTLs obtained from 76 different studies and 134 distinct rice populations, all related to yield components and DT-associated traits (Additional file 11: Table S8). This analysis led to the identification of 213 MQTLs (Table 3). The utilization of MQTL analysis resulted in a significant reduction in the CI, enabling the identification of a more precise set of candidate genes potentially involved in controlling the investigated traits. The average CI of MQTLs was 4.68 cM, representing a 2.74-fold decrease compared to the average CI of the original QTLs, which was 12.86 cM (Table 3). Notably, 62.4% of the MQTLs had a CI of less than 5 cM, while 17 MQTLs displayed a CI of less than 1 cM (Table 3). Apart from having a narrow CI, an MQTL selected for breeding should possess a high number of original QTLs and a high PVE value. In this study, 17 MQTLs (including MQTL_PH1.2, MQTL_GY6.4, MQTL_GY8.4, MQTL_DRI1.1, MQTL_SF4.2, MQTL_HD6.2, MQTL_GY8.5, MQTL_PH1.4, MQTL_BY6.1, MQTL_GY6.2, MQTL_PN4.4, MQTL_GY3.3, MQTL_HI3.2, MQTL_GY2.4, MQTL_GY2.1, MQTL_HD3.5, and MQTL_GY1.4) fulfilled these criteria, with more than 10 original QTLs, an average PVE of 14.2, and a mean CI of 1.98 cM. Consequently, they can be regarded as potential MQTLs for future breeding programs aimed at enhancing yield and drought tolerance in rice.

To identify candidate genes involved in controlling yield components and DT-related traits in rice, the genes located within the MQTL regions were compared with differentially expressed genes (DEGs) identified through RNA-seq and microarray data analysis (Additional file 9: Table S6). The genes responsible for regulating various investigated traits are discussed below. Furthermore, following the criteria defined by Löffler et al. in 2009, MQTLs exhibiting a high number of original QTLs, a high PVE value, and a narrow CI were identified as "Breeding MQTLs". Additionally, QTL-overview peaks that overlapped with SNP peaks reported in GWAS studies were considered, and the genes situated within these regions were recognized as potential candidate genes involved in controlling yield components and DT-related traits (Fig. 5, Table 3, and Additional file 10: Table S7).

Fig. 5
figure 5

The Circos diagram illustrates the positioning of MQTLs and GWAS-based SNPs associated with yield and drought tolerance (DT)-related traits under water deficit conditions on the physical map of Oryza sativa japonica (Nipponbare). a The twelve rice chromosomes are arranged in a clockwise direction. b The gene density on rice chromosomes shown in yellow to red scale for the lowest to the highest density in 500 kb windows. c The locations of identified MQTLs through comprehensive MQTL analysis on the physical map of each chromosome. d Position of the identified MQTLs for each distinct trait on the physical map of each chromosome. e Showing the genes locating inside both the SNP peak positions and the QTL-overview peaks for yield and DT-associated traits under drought stress conditions as novel candidate genes (Supplementary Table S7). f Points having various colors symbolize significant SNPs discovered through GWAS studies for yield and DT-associated traits under water deficit conditions. g The heatmap representing the Indels density in white to dark blue color scale for the lowest to the highest density. h The heatmap representing the structural variants (SV) density in white to dark red color scale for the lowest to the highest density. Physical positions of all markers (including those used in the flanking markers each MQT and map markers) were determined using the genome assembly IRGSP-1.0

Full size image

MQTLs and Candidate Genes for Grain Yield

Drought stress during the reproductive stage of rice leads to a significant reduction in grain yield (Palanog et al. 2014). According to the obtained results, 50 MQTLs were identified for the GY trait (Table 3). Chromosomes 1, 2, 3, 6 and 8 had the highest number of GY-associated MTQLs with 6 MTQLs per chromosome. The most stable GY-associated MQTLs were MQTL_GY1.4, MQTL_GY3.4, MQTL_GY2.1, MQTL_GY2.4, MQTL_GY3.3, MQTL_GY6.2, MQTL_GY8.5, MQTL_GY6.4, MQTL_GY8 having the uppermost number of original QTLs from 22, 17, 14, 13, 12, 11, 11, 10 and 10 different studies, respectively. The confidence interval of 8 MQTLs including MQTL_GY2.6, MQTL_GY1.6, MQTL_GY6.5, MQTL_GY3.6, MQTL_GY8.6, MQTL_GY3.3, and MQTL_GY1.4 was decreased to less than 1 cM (Table 3), indicating the potential of the mentioned MQTLs to be used in breeding programs to enhance GY under drought stress in rice.

The genes located within each MQTL interval are listed in Additional file 6: Table S3. One of these genes, OsTF1L, located within the MQTL_GY8.3 interval, is engaged in controlling GY in rice under drought stress treatment. Transgenic plants overexpressing OsTF1L indicated more tolerance to drought stress in comparison with wild type plants at the reproductive stage. In addition, transgenic plants produced higher GY compared to wild type plants under drought stress. Encouraging lignin biosynthesis and stomatal closure by the HD-Zip transcription factor of OsTF1L is the reason for improved drought tolerance in transgenic plants (Bang et al. 2019).

OsCBL8 gene, detected in MQTL_GY2.2 interval, is engaged in rice response to abiotic stresses. Up- and down-regulation of OsCBL8 in sense (salt tolerant) and anti-sense (drought tolerant) transgenic lines resulted in a remarkable reduction in both the number of filled grains per panicle and the seed setting rate in rice (GU et al. 2010).

OsGRF4 gene, located in MQTL_GY2.5, encodes a transcriptional regulator and is regulated by OsmiR396c. It has been reported that the module of OsmiR396c-OsGRF4-OsGIF1 is involved in determining GY and size in rice. It was demonstrated that a 2 bp substitution mutation in OsGRF4 disturbs its regulation by OsmiR396c, leading to enhanced GY through increasing the weight and size of grains. In addition, there is direct interaction between OsGRF4 and OsGIF1 and it has been reported that overexpression of OsGIF1 enhanced grain size (Li et al. 2016). Furthermore, increased expression of OsGRF4 caused by mutation, activates brassinosteroid responses, promoting grain development (Hu et al. 2015).

OsLHT1, discovered in the MQTL_GY8.1 interval, encodes Lysine-Histidine-type Transporter 1, which is involved in translocating amino acids from vegetative organs to reproductive ones, determining GY. High GY and N use efficiency are determined by the suitable allocation of nitrogen from source leaves to grains. Panicle length, the grain number per panicle and total grain weight were reduced in knockout mutants of OsLHT1 (Guo et al. 2020). SAPK2, located within MQTL_GY7.4 interval, encodesfor a serine/threonine-protein kinase and is able to increase GY via regulating nitrogen use efficiency under drought stress in the reproductive stage. In addition, contents of nitrogen, phosphorus, and potassium in rice grain are remarkably influenced by SAPK2 (Lou et al. 2020). RL9 (SLL1, AH2, OsADD1), found in MQTL_GY9.2 interval, codes for a MYB domain protein that is involved in the development of hull and grain. RL9 influences on GY, grain size and quality. Rice plants overexpressing SLL1 had longer lateral roots, indicating the potential of SLL1 gene to be used for improving root architecture in rice (Ren et al. 2019; Shelley et al. 2013). RAG2, located in MQTL_GY7.1 interval, codes for a 14–16 kDa α-amylase/trypsin inhibitor. Overexpression of RAG2 led to improved GY and grain quality in rice (Zhou et al. 2017). PLANT ARCHITECTURE AND YIELD 1 (PAY1), located in MQTL_GY8.4 interval, plays a role in enhancing plant architecture and GY in rice. This gene can be used for establishing perfect plant architecture and breeding rice varieties for high yield (Zhao et al. 2015). OsNPF7.2, located in MQTL_GY2.5 interval, which codes for a nitrate transporter, plays a positive role in regulating number of tillers and GY in rice (Wang et al. 2018). OsbHLH107 is located in MQTL_GY2.6 interval, and its homologs play important roles in regulating grain size development and can be used for improving GY in rice (Yang et al. 2018b). OsABCG18, detected in MQTL_GY8.2 interval, codes for an ABC transporter and is involvedin controlling cytokinins transport into shoots and GY in Rice. Enhanced cytokinins in the shoot and increased GY were obtained by overexpression of OsABCG18 (Zhao et al. 2019).

Among the 50 MQTLs identified for the grain yield trait, a total of 22 MQTLs overlapped with 49 SNP peak positions associated with yield-related traits based on GWAS studies (Fig. 5, Additional file 10: Table S7). The genes located both in QTL-overview and SNP peaks, including OsSPO11-5, OsDRF1, FKF1, Os03g0197175, Os03g0197200, Os03g0305000, Os03g0305050, Os03g0232800 and Os08g0170200 are considered as candidate genes for GY under drought stress.

MQTLs and Candidate Genes for Heading Date

Heading date, which is regulated by numerous environmental signals and endogenous cues, plays a crucial role in crop reproduction, yield, and regional adaptability (Wei et al. 2020). Yield and drought tolerance are highly correlated with HD (Xu et al. 2018). We identified 27 MQTLs for HD under drought conditions, with a maximum of 7 MQTLs on chromosome 3, and 4 MQTLs on each of the chromosomes 1 and 6 (Table 3). The highest number of original QTLs (18 QTLs) was observed for MQTL_HD3.5 on chromosome 3 (Table 3). In five MQTLs including MQTL_HD3.5, MQTL_HD3.7, MQTL_HD3.3, MQTL_HD3.1 and MQTL_HD12.2, the CI was decreased to less than 1 cM (Table 3). Out of the 27 identified MQTLs for the HD trait, 4 MQTLs overlapped with 6 SNP peak positions reported by GWAS studies and one MQTL (MQTL_HD3.1) was recognized as QTL-overview peaks overlapped with SNP peaks reported by GWAS studies (Fig. 5, Additional file 10: Table S7).

Some of the genes located in HD-related MQTLs are discussed here. OsMFT, located in the MQTL_HD6.4 interval, increased drought tolerance in rice by interacting with OsbZIP66 and OsMYB26, recognized as main drought-related transcription factors, and regulating their binding to drought-responsive genes (Chen et al. 2021). OsMFT1 suppressed Ehd1, FZP, and SEPALLATA-like genes, resulting in delayed heading date and enhanced spikelets per panicle in rice (Song et al. 2018). Ehd2 gene, located in the MQTL10-1 interval, is involved in adjusting flowering time in rice (Brambilla and Fornara 2013). Ehd2 gene was also detected on a HD-related MQTL on chromosome 10 under drought stress conditions in rice by Khahani et al. 2021 (Khahani et al. 2021).

MQTLs and Candidate Genes for Plant Height

For the PH trait under drought stress in rice, 23 MQTLs were obtained. Chromosome 1 with 7 MQTLs had the most number of MQTLs per chromosome, followed by chromosomes 6 and 8 with 3 MQTLs per each chromosome. The most stable MQTLs for PH were MQTL_PH1.4, MQTL_PH1.2 and MQTL_PH1.7. MQTL_PH1.4 and MQTL_PH1.2 were considered as the most stable MQTLs for PH, because of having the most number of original QTLs with 10 and 11 original QTLs, respectively. MQTL_PH1.7 was identified as one of the most stable MQTLs due to having CI < 1 cM (Table 3).

Some of the genes located in PH-related MQTLs are discussed here. Auxin is involved in regulating plant height (Ma et al. 2016b). Amidase is engaged in the metabolic pathway of indole acetic acid (IAA). The crucial role of IAA phytohormone in cell division, differentiation, elongation, root development and plant height regulation has been reported (Petersson et al. 2009). In a prior investigation, specific genes associated with the auxin metabolic pathway, namely OsYUCCA1, OsYUCCA8, WOX6, and OsRR2, were identified within the genetic interval linked to MQTLs that are correlated with root system architecture in rice. (Daryani et al. 2022). In the current research, the same MQTLs and genes were detected for PH trait. This means that the common MQTLs and genes are involved in controlling both traits of root system architecture and PH in rice. In this study, YUCCA1, YUCCA6, YUCCA4, OsIAA20, OsIAA21, OsIAA2, IAA6 and OsRR33 were detected in the intervals of MQTL_PH1.3, MQTL_PH7.1, MQTL_PH1.1, MQTL_PH6.1, MQTL_PH6.2, MQTL_PH1.1, MQTL_PH1.4 and MQTL_PH8.2, respectively.

Another detected gene is OsFTL1, which locates in MQTL_PH1.1 interval. Pleiotropic effects of OsFTL1 on the total number of secondary rachides, grains number per panicle, plant height and flag leaf length have been reported (Wang et al. 2020). OsbZIP49, detected in MQTL_PH6.3 interval, codes for a transcription factor that is engaged in controlling tiller angle and plant architecture via IAA synthetase. Overexpression of OsbZIP49 in rice led to tiller-spreading phenotype, reduced plant height and internode length (Ding et al. 2021).

Out of the 23 identified MQTLs for PH trait, seven MQTLs overlapped with 14 SNP peak positions found by GWAS studies. Two novel candidate genes for PH trait under drought stress in rice, including Os01g0884500 and Os01g0885300 were located both in QTL-overview and SNP peaks (Fig. 5, Additional file 10: Table S7).

MQTLs and Candidate Genes for Biomass Yield

Enhancing biomass in rice is a key breeding objective, yet it poses challenges due to the intricacies involved and labor-intensive nature of the trait phenotyping (Matsubara et al. 2016). Twenty-nine MQTLs were detected for BY. Chromosomes 1, 3, and 4 each contain four MQTLs, while chromosomes 2 and 6 each harbor three (Table 3). The most stable MQTLs detected for BY under drought stress included MQTL_BY6.1, MQTL_BY3.3, MQTL_BY4.2 and MQTL_BY4.4, having the highest number of original QTLs and two MQTLs of MQTL_BY1.2 and MQTL_BY2.3, having CI of less than 1 cM (Table 3).

OsOFP6, detected within the MQTL_BY2.2 interval, is involved in regulating growth and development, and responses to drought and cold stresses in rice. RNAi-mediated knockdown of OsOFP6 led to semi-dwarf plants with changed grain shape and shorter lateral roots. In addition, slower water loss and less accumulation of H2O2 were observed in OsOFP6 overexpressing plants in comparison with RNAi plants under drought stress, indicating the role of OsOFP6 in both drought avoidance and drought tolerance in rice. As well, a thicker secondary cell wall with enhanced lignin content was noticed in OsOFP6 overexpressing plants in rice (Ma et al. 2017; Sun et al. 2020). OsHk6 (OsCKT1), located within the MQTL_BY2.2 interval, serves as a cytokinin receptor and plays a role in regulating various biological processes such as secondary metabolism, sucrose and starch metabolism, chlorophyll synthesis and photosynthesis (Ding et al. 2017). In addition, green pigmentation and shoot induction were promoted in rice calli by ectopic expression of OsHk6 (Choi et al. 2012). Two genes encoding for glutamine synthetases, OsGS1;1 and OsGS1, both located within MQTL_BY2.2 interval, are probably engaged in drought tolerance in rice based on the previous studies. OsGS1;1 plays a key role in normal growth and grain filling under water deficit conditions. Increased physiological tolerance and agronomic performance were obtained in rice plants co-overexpressing OsGS1;1 and OsGS2 isoforms under adverse abiotic stress conditions (James et al. 2018; Tabuchi et al. 2005). OsBRI1 (D61, OsBRKq1), detected within the MQTL_BY1.3 interval, is engaged in several growth and developmental processes like internode elongation, lamina joint bending and skotomorphogenesis. Altogether, the kinase activity of OsBRI1 is crucial for brassinosteroids to regulate normal plant growth and development in rice (Zhao et al. 2013). OsBRKq1 has the potential to be utilized for increasing yield through enhancing grain size in rice. Additionally, OsBRKq1 was found on chromosome1 through a QTL mapping analysis for 1000 kernel weight, kernel length, and kernel width conducted on SNDH113 populations in which grain sizes were diversely distributed (Park et al. 2021). Expression of rice OsARGOS (detected in MQTL_BY4.3) in Arabidopsis resulted in enhancedcell division and expansion and increased organ size. Considering the role of OsARGOS in organ enlargement, it has the potential to be used for biomass enhancement through genetic engineering (Wang et al. 2009).

Ten MQTLs (Out of the 29 detected MQTLs for BY trait) overlapped with 17 SNP peak positions discovered by GWAS studies (Fig. 5, Additional file 10: Table S7). Three candidate genes including OsBRXL2, DEK1 (ADL1) and Os06g0164300 were identified for PH trait under drought in rice (Fig. 5, Additional file 10: Table S7) that were located in SNP peak positions overlapped with QTL-overview peaks MQTLs.

MQTLs and Candidate Genes for Canopy Temperature

Canopy temperature under stress conditions is a reliable predictor of GY performance (Melandri et al. 2020). We identified two MQTLs for CT consisting of one MQTL on chromosome 2 and another on chromosome 3 (Table 3).

MQTLs and Candidate Genes for Drought Response Index

Twenty-four MQTLs were identified for DRI using the meta-analysis of QTLs. The highest count number of DRI-associated MQTLs per chromosome was 3 and related to chromosomes 2, 3, and 5 (Table 3). MQTL_DRI1.1 was recognized as the most stable DRI-related MQTL having the most number of original QTLs from 10 independent studies. GF14c, located in MQTL_DRI8.2, codes for a 14-3-3 protein. 14-3-3 proteins play main roles in regulating primary metabolism and transducing cellular signals (Ho et al. 2013). It has been demonstrated that overexpression of GF14c resulted in increased drought tolerance in transgenic seedlings of rice (Ho et al. 2013). Another gene that was found in MQTL_DRI2.2 interval is OsDi19-4 (Dehydration-induced 19 homolog 4). The OsDi19 codes for proteins that are engaged in response to abiotic stresses. Wang et al. 2014 indicated that overexpression of OsDi19-4 led to increased drought tolerance in rice through increasing ROS-scavenging activity (Wang et al. 2014). In addition, the expression of some ABA-responsive genes was changed in rice plants overexpressing OsDi19-4 which resulted in strong ABA-hypersensitive phenotypes (Wang et al. 2016). OsDIS1, detected in MQTL_DRI3.2, encodes a SINA-type E3 ligase. OsDIS1 plays a negative role in drought stress tolerance via regulating transcription of several stress-associated genes and probably via regulating its interacting protein OsNek6 at posttranslational level in rice (Ning et al. 2011). OsETOL1, located in the MQTL_DRI3.1 interval, encodes a homolog of ETHYLENE OVERPRODUCER. Two allelic mutants of OsETOL1 indicated enhanced tolerance to drought stress at panicle development stage (Du et al. 2014). OsGRAS23, detected in MQTL_DRI4.3 interval, codes for a stress-responsive GRAS transcription factor. OsGRAS23 plays a positive role in regulating drought tolerance in rice by inducing several stress responsive genes (Xu et al. 2015). OsMT1a (a type 1 metallothionein), found in MQTL_DRI12.2 interval, is engaged in zinc homeostasis and drought tolerance in rice. Yang et al. 2009 indicated that overexpression of OsMT1a resulted in increased drought tolerance in rice via taking part in ROS scavenging pathway directly and also through regulating the expression of zinc finger transcription factors (Yang et al. 2009). OsTF1L, detected in MQTL_DRI8.1, encodes a homeodomain-leucine zipper transcription factor. OsTF1L plays a key role in regulating drought tolerance mechanisms in rice. Up-regulation of drought-inducible genes and the genes involved in stomatal movement and lignin biosynthesis was observed in plants overexpressing OsTF1L. Under drought stress, rice plants overexpressing OsTF1L showed enhanced effective photosynthesis, reduced water loss rate and increased drought tolerance at the vegetative stage. Furthermore, enhanced drought tolerance together with increased GY was observed in the OsTF1L overexpressing plants than in non-transgenic plants at the reproductive stage (Bang et al. 2019).

Out of the 24 identified MQTLs for DRI trait, four MQTLs overlapped with 19 SNP peak positions discovered through GWAS studies (Fig. 5, Additional file 10: Table S7). Os03g0288800 and Os03g0286500 were identified as two novel candidate genes for DRI in rice that was detected in SNP peak positions overlapped with QTL-overview peaks.

MQTLs and Candidate Genes for Flag Leaf Size

Meta-analysis of QTLs resulted in the identification of 10 MQTLs for FLZ in rice. The most number of MQTLs per chromosome was three MQTLs on chromosome3, followed by two MQTLs on each of chromosomes 2 and 4 (Table 3). MQTL_FLZ1.1 was identified as the most stable MQTL for FLZ having the most number of original QTLs from four independent studies.

Within the MQTL_FLZ1.1 interval, OsFBK1 (ORYZA SATIVA F-BOX KELCH 1) was detected, which encodes an E3 ligase subunit. It has been demonstrated that EP3 and OsFBK1, both are functional orthologues of Arabidopsis F-box protein HAWAIIAN SKIRT, influence on plant architecture, organ size, number and size of floral organ, floral morphology, pollen viability, grain size, and weight and affect transcript accumulation of microRNA pathway genes and their targets (Borah and Khurana 2018). The most important mechanism engaged in environmental responses and developmental processes in plants is mitogen-activated protein kinase (MAPK) cascade. OsMAPK2, located in MQTL_FLZ3.1 interval, encodes Oryza sativa MAP kinase 2 gene. OsMAPK2 may be involved in the stress-signaling pathway and panicle development in rice. OsMAPK2 plays a role in plant tolerance to various biotic/abiotic stresses based on previous studies. Hur and Kim (2014) indicated that overexpression of OsMAPK2 affected root development and led to increased tolerance to phosphate deficiency in rice and Arabidopsis (Hur and Kim 2014). The other discovered gene is a cytokinin receptor, called OsHk6 (OsCKT1), detected in the MQTL_BY2.2 interval. OsHk6 is involved in cytokinin regulation of biological processes like secondary metabolism, sucrose and starch metabolism, chlorophyll synthesis and photosynthesis (Choi et al. 2012).

MQTLs and Candidate Genes for Harvest Index

Meta-analysis of QTLs led to the identification of 10 MQTLs for HI. The highest number of MQTLs per chromosome were three MQTLs on chromosome 6, followed by 2 MQTLs on each of the chromosomes 1, 2, and 3 (Table 3). MQTL_HI3.2 having the most number of original QTLs from 12 independent studies and MQTL_HI2.2 having CI < 1 cM was recognized as the most stable MQTLs for the HI under drought stress (Table 3). Out of the 10 detected MQTLs for HI trait, one MQTL overlapped with SNP peak positions discovered in GWAS studies (Fig. 5, Additional file 10: Table S7).

OsNAC6, OsRPK1, OsZFP, OsCOI1a, OsPP15, OsKASI, OsETOL1 and OsMSRMK2 detected in the intervals of MQTL_HI1.2, MQTL_HI1.1, MQTL_HI1.2, MQTL_HI1.2, MQTL_HI1.2, MQTL_HI6.2, MQTL_HI3.2 and MQTL_HI3.2, respectively, were recognized as potential candidate genes for HI under drought stress conditions.

The transcription factor of OsNAC6 up-regulates the expression of the genes involved in several drought tolerance pathways such as genes engaged in membrane modification, nicotianamine biosynthesis, glutathione translocation, 3′-phophoadenosine 5′-phosphosulfate accumulation and glycosylation. Altogether, molecular drought tolerance mechanisms are arranged by OsNAC6, indicating its potential to be used for developing high-yielding crops under drought stress conditions (Lee et al. 2017). The OsRPK1 gene, encoding a Ca2+-independent Ser/Thr kinase, was induced by auxin, ABA, cold and drought stresses. Knockdown of OsRPK1 led to enhanced growth, plant height and tiller number in transgenic rice plants. Furthermore, polar auxin transport and development of root are negatively regulated by OsRPK1 in rice (Zou et al. 2014). OsZFP encodes a C2HC-type zinc finger protein that plays a role in regulating the development of lateral roots through IAA pathways (Cui et al. 2017). The F-box protein OsCOI1 is involved in drought tolerance in rice through participating in the signaling module of OsbHLH148-OsJAZ-OsCOI1 (Seo et al. 2011). OsPP15 (OsPP2C09), encoding a clade A type 2C protein phosphatase, had a positive effect on plant growth but negatively regulated drought tolerance via ABA signaling. On the other hand, OsPP2C09 interacts with DREB TFs and activates DRE-containing promoters. So, drought response regulon is positively regulated by OsPP2C09, leading to the activation of an ABA-independent signaling pathway. Altogether, OsPP2C09 is involved in both ABA-dependent and independent abiotic stress signaling pathways as a bifunctional regulator (Chen et al. 2014; Min et al. 2021). OsKASI encodes β-ketoacyl-[acyl carrier protein] synthase I. OsKASI deficiency led to decreased fertility and a considerable change in the composition and contents of fatty acids in roots and seeds. It was demonstrated that the involvement of OsKASI in fatty acid synthesis is of great importance for rice root development (Ding et al. 2015). OsETOL1 codes for a homolog of ETHYLENE OVERPRODUCER. Two allelic mutants of OsETOL1 indicated enhanced tolerance to drought stress during panicle development stage (Du et al. 2014). Diverse biotic/abiotic stresses resulted in changes in the expression of OsMSRMK2, indicating its role in defense/stress response pathways of rice (Agrawal et al. 2002).

MQTLs and Candidate Genes for Panicle Number

We identified 25 MQTLs for the PN trait using MQTL analysis. The highest number of MQTLs per chromosome was four MQTLs on each of the chromosomes 2 and 4 (Table 3). MQTL_PN4.4 was recognized as the most stable MQTL for PN having the highest number of original QTLs from 11 independent studies.

hbd2, IAA6, OsAHL1, OsC3H10, OsCNX, OsMOGS, OsTSD2, SAPK2, OsCKX9, SRS1/DEP2, OsLHT1 and SAPK2 located in the intervals of MQTL_PN2.3, MQTL_PN1.2, MQTL_PN8.2, MQTL_PN1.2, MQTL_PN4.2, MQTL_PN1.3, MQTL_PN2.4, MQTL_PN7.1, MQTL_PN5.2, MQTL_PN7.1, MQTL_PN8.2 and MQTL_PN7.1, respectively, were detected as potential candidate genes for PN under drought stress conditions.

OsCKX9, located in MQTL_PN5.2 interval, encodes cytokinin oxidase 9. Significant enhancements in tiller number and reduction in plant height and panicle size were observed in both OsCKX9 mutants and OsCKX9-overexpressing plants, proposing that OsCKX9 homeostasis is of great importance for regulating shoot architecture in rice (Duan et al. 2019). SRS1/DEP2 (The Small and Round Seed1/Dense and Erect Panicle2) is engaged in regulating seed size and panicle length in rice (Abe et al. 2010).

OsLHT1 (MQTL_PN8.2) plays key roles in the translocation of amino acids from vegetative to reproductive organs for GY and quality of nutrition and functionality. The amino acid transporter of OsLHT1 exhibits a broad substrate specificity and a tendency for neutral and acidic amino acids, and disturbance of OsLHT1 function noticeably repressed rice growth and fertility. Loss-of-function of OsLHT1 in two oslht1 mutants, produced through CRISPR/Cas9 genome-editing technology, led to inhibition of root and shoot growth and significant reduction of grain yield in rice (Guo et al. 2020). SAPK2 codes for a Serine/threonine-protein kinase and contributes to rice yield by controlling nitrogen metabolic processes under water deficit conditions in the reproductive stage (Lou et al. 2020).

Out of the 25 identified MQTLs for PN trait, 14 MQTLs overlapped with 23 SNP peak positions discovered by GWAS studies (Fig. 5, Additional file 10: Table S7). Three novel candidate genes for PN trait under drought stress conditions in rice include Os01g0915350 and Os02g0752200, Os02g0752250 (Fig. 5, Additional file 10: Table S7) that were detected in SNP peak positions overlapped with QTL-overview peaks MQTLs.

MQTLs and Candidate Genes for Spikelet Fertility

We identified 13 MQTLs for SF trait using MQTL analysis. The highest number of MQTLs per chromosome was two MQTLs on each of the chromosomes 4, 5, 6, 8 and 9 (Table 3). MQTL_SF4.2 having the most number of original QTLs from 10 independent studies and MQTL_SF2.1 having CI < 1.17 cM were recognized as the most stable MQTLs for SF trait under drought stress. Out of the 13 identified MQTLs for SF trait, one MQTL overlapped with one SNP peak position detected by GWAS studies (Fig. 5, Additional file 10: Table S7).

Gibberellic acid (GA) plays an important role in development of floral organs and GA signaling has a key function in spikelet fertility. OsGID1, detected in MQTL_SF5.1, acts as soluble GA receptor and binds directly to the biologically active GA. Then, OsGID1 interacts with SLR1, a DELLA protein that supress GA signalling. This results in degradation of SLR1 and consequently, permitting GA signaling pathway (Kwon and Paek 2016; Ueguchi-Tanaka et al. 2005). Moreover, OsMFT1, detected in MQTL_SF6.2, play an important role in GA biosynthesis and ABA signaling (Lu et al. 2023) and has a main regulatory function under drought stress in rice (Chen et al. 2021). It is also reported that panicle branching and spikelets per panicle in rice is enhanced by OsMFT1 through suppressing a subfamily of MADS-box genes and SEPALLATA-like genes, respectively (Song et al. 2018).

Breeding MQTLs

Thirteen MQTLs were identified as "Breeding MQTLs" meeting the following criteria: having more than 10 initial QTLs, a confidence interval (CI) of less than 1 cM, and an average proportion of phenotypic variance explained (PVE) by the original QTLs exceeding 10. These MQTLs are specifically identified as MQTL2-8, MQTL1-12, MQTL8-9, MQTL9-4, MQTL1-5, MQTL12-4, MQTL3-5, MQTL6-3, MQTL1-6, MQTL12-2, MQTL1-11, MQTL3-6, and MQTL8-4 (Table 5, Fig. 3).

Conclusion

Meta-analysis of yield and drought tolerance associated traits under drought stress conditions led to the discovery of 213 MQTLs, among which 17 MQTLs had a genetic distance of less than 1 cM and accounted for an average phenotypic variance of 20.29%. Notably, 63 MQTLs (out of 213 MQTLs) coincided with SNP positions identified by GWAS for yield components and DT-related traits under drought stress in rice. Moreover, 19 genes precisely situated at the SNP peak positions and QTL-overview peaks were nominated as candidate genes for subsequent functional analysis. These genes were involved in GY (OsSPO11-5, OsDRF1, FKF1, Os03g0197175, Os03g0197200, Os03g0305000, Os03g0305050, Os03g0232800 and Os08g0170200), plant height (Os01g0884500 and Os01g0885300), Biomass yield (OsBRXL2, DEK1 (ADL1) and Os06g0164300), drought response index (Os03g0288800 and Os03g0286500) and panicle number (Os01g0915350, Os02g0752200, and Os02g0752250) under drought stress. On the other hand, the inclusive meta-analysis of QTLs for all the yield-associated traits together led to identification of 13 MQTLs having suitable features to be used as "breeding MQTLs". Finally, integrating the results of MQTL-analysis for yield and DT-associated traits (distinct and combined analysis of traits), GWAS studies, and transcriptome data, resulted in finding 82 candidate genes involved in DT and yield maintenance under drought stress. The promising candidate genes and breeding MQTLs discovered in the current research are valuable sources for genetic engineering and molecular breeding for drought tolerance in rice.