The genomic regions and candidate genes associated with drought tolerance and yield-related traits in foxtail millet: an integrative meta-analysis approach

Drought stress is one of the most significant limiting factors limiting crop productions. Foxtail millet (Setaria italica) is among the most drought-tolerant crop plants, with a high degree of collinearity with other staple cereals. The present study used a meta-analysis approach to identify genomic regions and candidate genes associated with drought tolerance and yield-related traits in foxtail millet. A meta-analysis employing all 448 collected original quantitative trait loci (QTL) identified 41 meta-QTL (MQTL) on the nine foxtail millet chromosomes. The confidence interval (CI) of the identified MQTL was determined to be 0.31–14.47 cM (5.23 cM average), which was 3.5 times narrower than the mean CI of the original QTL. Based on the available RNA-seq and microarray data, 1631 differentially expressed genes (DEGs) were detected in 41 MQTL. Furthermore, through synteny analysis, 8, 4, and 2 ortho-MQTL were recognized within co-linear regions of foxtail millet with rice (Oryza sativa), barley (Hordeum vulgare), and maize (Zea mays), respectively. To detect the most significant genome regions involved in the genetic control of drought tolerance and yield maintenance in foxtail millet, 10 MQTL with physical intervals of less than 1 Mb and seven hotspot regions with a high QTL-overview index were identified. Several candidate genes involved in foxtail millet sensing and signaling, transcription regulation, ROS inhibition, and adaptation to abiotic stress were detected by seeking drought-responsive genes in MQTL regions with a CI < 1 Mb. We hope that the achieved results would aid in developing new high-yielding drought-tolerant genotypes.


Introduction
Plants often encounter diverse environmental conditions and numerous biotic and abiotic stresses throughout their lifespan. Drought is regarded as the most severe of all abiotic stresses, critically threatening global crop growth and sustainable agriculture (Mittler 2006). Drought stress can adversely affect plant growth quantity, quality, and yield stability, resulting in substantial economic losses (Salehi-Lisar and Bakhshayeshan-Agdam 2016). Thus, improving crop production while simultaneously decreasing agricultural water input is necessary to ensure the security of our global food supply and protect our declining freshwater resources (Feldman et al. 2018;Qie et al. 2014), and developing new crop varieties that can produce more grain while consuming less water will be crucial .
Foxtail millet is equipped with nitrogen use efficiency and several morphological and physiological adaptations, including compact and deep root systems, smaller leaf area, epidermal cell arrangements, and thick cell walls, which are reported to lead to robust tolerance to various abiotic stresses, primarily drought, heat, and salinity (Diao et al. 2014;Lata et al. 2013). The majority of cultivated foxtail millet is utilized as animal feed or bird seed. However, it is also commonly cultivated for human consumption in Asia (Bhat et al. 2019). Many essential traits of agriculture, including abiotic stress tolerance, yield, nutritional value, and biomass, are quantitatively inherited. A large number of genes, environmental variations, and genotype-by-environment interactions (GEIs) affect these complex traits.
Only a few studies have been reported on quantitative trait loci (QTL) of foxtail millet. Several QTL associated with biomass, plant height, panicle length, panicle weight, and panicle diameter, have been identified for drought tolerance in foxtail millet (Fang et al. 2016;Wang et al. 2014;Zhang et al. 2017). However, these studies were conducted using diverse populations and independent experiments, and it was reported that QTL might not be directly comparable. On the other hand, a single QTL may correspond to multiple candidate genes; it has been debated whether QTL locations specified for a particular trait in one population correlate with those detected in other populations. Thus, metaanalyses can generate stronger conclusions than individual studies (Rifkin 2012).
In plants, the concept of meta-analysis of QTL has recently been considered for a variety of plant traits, such as drought, salt stress, and waterlogging tolerance in barley Zhang et al. 2017), root system architecture and salt tolerance in rice (Khahani et al. 2019;Daryani et al. 2022), grain yield and anthesis silking interval under drought stress in maize (Semagn et al. 2013), and drought tolerance related traits in maize (Almeida et al. 2013).
The consistent QTL identified by the meta-analysis technique for a collection of QTL at a 95% confidence interval (CI) is referred to as a meta-QTL (MQTL). The MQTL with the lowest CI and the strongest influence on a trait are advantageous to plant breeders (Swamy et al. 2011). The conclusive goals of MQTL analysis are to obtain a better appraisal of a QTL position and, thus, to specify a colocation with the candidate genes (Goffinet and Gerber 2000). These studies provide significant information for genetic enhancement and gene detection.
Numerous genes have been identified in foxtail millet to date, including the LEA gene SiLEA14, which plays a crucial role in abiotic stress tolerance. A study demonstrated that the overexpression of SiLEA14 improved the tolerance of foxtail millet to abiotic stresses . The foxtail millet overexpressing SiARDP induced the expression of drought and salinity-relevant genes and increased abiotic stress tolerance . Furthermore, overexpression of SiMYB3 assisted root development of transgenic Arabidopsis under low-nitrogen conditions (Ge et al. 2019). SiATG8a induced and expressed in the stem under low-phosphorus (LP) stress. The overexpression SiATG8a in wheat (Triticum aestivum) improved tolerance under LP stress and increased grain yield (He et al. 2023).
Only one MQTL study on agronomic traits under normal conditions has been conducted in foxtail millet (Wang et al. 2019). Consequently, information on MQTL related to abiotic stress tolerance in foxtail millet has remained insufficient.
Iran comprises large areas of arid and semi-arid zones, where water availability is the major limiting factor for agriculture (Noori et al. 2021). Climate change is likely to exacerbate water scarcity, but drought-resilient crops such as foxtail millet can contribute to sustainable agriculture. In the present study, an integrative meta-analysis approach was used to discover the genomic regions and candidate genes involved in drought tolerance and yield-related traits of foxtail millet, which could be beneficial for marker-assisted selection in breeding programs. In addition, synteny analysis was used to identify the co-linear chromosome regions among foxtail millet with other well-known cereals such as rice, barley, and maize. The obtained results would provide beneficial information to understand better plants' genetic mechanisms underlying drought tolerance.

Collecting QTL data from bi-parental mapping studies
An inclusive literature review was done looking for QTL related to drought tolerance and yield-associated traits including plant height (PH), water use efficiency (WUE), biomass (BM), root length (RL), tiller number (TN), panicle length (PL), node number (NN), yield (YLD) and stem diameter (SD) in foxtail millet grown under drought or normal conditions. Table 1 provides detailed information on the collected QTL comprising the parents in developing mapping populations, the population size, the markers used, and the number of recognized QTL. In addition, Table S1 outlines the reported genetic position, flanking molecular markers, confidence interval (CI), QTL peak position, LOD score (logarithmic odds ratio), and proportion of phenotypic variance explained (PVE or R 2 ) for each QTL.

Construction of consensus genetic map
A consensus genetic map was made, and a meta-analysis was done employing Biomercator v4.2 (https:// urgi. versa illes. inra. fr/ Tools/ BioMe rcator-V4/). Two well-known genetic maps (Bennetzen et al. 2012;Gupta et al. 2012) were combined to create a consensus map onto which all the QTL could be projected. The consensus map, which contained 1,116 markers (992 single nucleotide polymorphisms (SNPs) and 124 simple sequence repeats (SSR), served as a reference map for the QTL meta-analysis. This map was constructed using common markers from various publications (Table S2).

Meta-analysis of QTL and QTL-overview index
To compute a 95% of CI for the QTL, the CI = 530/(N × R 2 ) formulation for F 2 lines was employed along with CI = 163/ (N × R 2 ) for RILs lines, in which N denotes the size of the population, and R 2 represents the proportion of phenotypic variance explained (PVE) by the QTL. The Akaike Information Criterion (AIC) was applied to select the meta-analysis model on each chromosome. As a result, the meta-analysis model with the lowest AIC value was deemed a suitable model for determining the number of MQTL (Swamy et al. 2011).
The meta-analysis initially determined the best model based on the following model selection criteria: AIC, AICc, AIC3, BIC (Bayesian information criterion), and AWE (average weight of evidence). The optimal QTL model was selected when at least three of the five models had the lowest model selection criteria values (Chardon et al. 2004). A total of 448 independent QTL on nine chromosomes linked to various traits including BM, PH, NN, PL, SD, TN, RWUE, YLD, and RL under normal and drought stress conditions were utilized in this metaanalysis of QTL. Table S3 contains information regarding each MQTL, including genetic position (cM), 95% CI (cM), mean phenotypic variance explained (PVE), flanking markers, the number of initial QTL, and the number of studies. The QTL-overview index statistic was computed to examine the density of QTL associated with drought traits on each chromosome for each 1 cM segment of the genetic reference map (Chardon et al. 2004).

Identification of genes located in the MQTL regions
To detect the MQTL spanning genes, the CI's flanking markers of each MQTL were selected from the consensus genetic map, and their physical locations on respective chromosomes were found. The genome assembly of foxtail millet. was used as the reference genome (Bennetzen et al. 2012). Consequently, the genes were identified using Biomart on the Ensemble Plants website (https:// plants. ensem bl. org/ bioma rt/ martv iew/) (Table S4). Gene ontology (GO) enrichment analysis GO enrichment analysis of the genes located in the MQTL regions was conducted using AgriGO v2.0 (http:// syste msbio logy. cau. edu. cn/ agriG Ov2/). The gene function prediction by the singular enrichment analysis (SEA) tool was summarized into three main groups (biological process, cellular component, and molecular function). GO terms were deemed statistically significant if the p-value was ≤ 0.05.
Collecting drought-responsive genes in foxtail millet using high-throughput sequencing data Differentially expressed genes (DEGs) were collected from six independent RNA-seq and one microarray experiment conducted on foxtail millet in response to drought stress conditions. Genes with a p-value ≤ 0.05 and the cut off of log 2 fold change ≥ 1 or ≤ -1 were considered as DEGs (Table S5). A Venn diagram was used to show the genes, which are both located in the MQTL and also found to be differentially expressed by microarray and RNA-Seq investigations. The common DEGs were mapped onto regulation, metabolism, and transcription pathway overviews using MapMan (version 3.5.1 R2).

Ortho-MQTL mining in the genomes of barley, maize and rice
Three independent studies were used for the ortho-MQTL analysis (Table S6) (Almeida et al. 2013;Khahani et al. 2019Khahani et al. , 2020. In order to find syntenic regions, based on the MQTL physical positions in barley, maize and rice, orthologous genes were obtained at the Ensemble Plants database.

Plant materials and drought stress treatment
Based on our previous works on foxtail millets, two contrasting genotypes were selected including Ilam1, as a drought tolerant and Kerman3 as a susceptible genotype (Hossein et al. 2017;Vaezi et al. 2020). Healthy seeds of these two foxtail millet genotypes were disinfected using 5% sodium hypochlorite for one minute. The seeds were then washed five to six times with autoclaved distilled water to remove sodium hypochlorite completely (Khan et al., 2014). Afterward, the disinfected seeds were cultured on Petri dishes enclosing two layers of filter paper. Petri dishes were kept in a phytotron at a temperature of 20 ± 3 °C, a photoperiod of 16/8 (day/night), and relative humidity of 75%. Finally, the uniformly germinated seeds were transferred to 8 kg pots containing equal amounts of farm soil, sand, and animal manure on August 10, 2019. The experiment was performed in a randomized complete block design (RCBD) with five replications (Feldman et al. 2017). Each pot that contained three plants was considered a replicate. The experimental treatments included normal irrigation (by relative water content (RWC) of 85% and field capacity of 100%) and severe drought stress (RWC of 35 ± 5% and field capacity of 20 ± 5%). Experimental treatments were adjusted using evaporation data from the Class A evaporation pan as well as by measuring leaf relative water content (RWC) and soil moisture percentage. Soil field capacity was measured by random sampling of pots weekly based on the weighting method. Drought stress was applied at the heading stage (about 6 weeks after planting) by complete cessation of irrigation. The amount of RWC was measured 47, 54, 60, and 62 days after planting (Fig.  S1). Sampling was performed 62 days after cultivation while field capacity was around 20 ± 5% and 100% under drought stress and normal conditions, respectively. Sampling was done from three healthy terminal leaves of the plants from Kerman3 genotype under normal conditions (RWC of 86.0%) and drought stress (RWC of 34.61%), Ilam1 genotype under normal conditions (RWC of 86.2%) and stress conditions (RWC of 34.59%). The samples were collected in separate autoclaved bags, frozen in liquid nitrogen, and stored at − 80 °C. Clean and sterile equipment was used in all stages.

RNA isolation and cDNA library synthesis
RNA was extracted from from one to three grams of leaf tissue from each sample was used to extract RNA using the phenol/SDS method (Kansal et al 2008). The integrity of the extracted RNA was checked using agarose gel electrophoresis. The quantity and quality (the 260/280 nm ratio as an indicator of protein contamination and 260/230 nm ratio as an indicator of additional contaminants such as carbohydrates and phenols) of all RNA samples were investigated using Nanodrop (Acosta-Maspons et al. 2019). The cDNA of Leaf samples was synthesized starting from 300 ng of total RNA using random nonamers. According to the manufacturer's instructions, a qScript cDNA Synthesis Kit (Sinaclon, IRAN) was used for cDNA synthesis.
Normalization was performed using EF-1a (Elongation factor 1-alpha) as an internal control gene (Kumar et al. 2013). Gene-specific primers were designed by Oligo 7.0 (ver. 5.0; National Bioscience Inc., Plymouth, USA) and are reported in Supplementary Table S7 and S14. The relative expression level of each gene was calculated compared to the same gene in the susceptible genotype under control conditions using the 2 −ΔΔCt procedure and cycle threshold values (Livak and Schmittgen 2001).

Distribution of initial QTL on nine chromosomes in foxtail millet
A total of 452 QTLs out of the 448 QTLs (99%) from ten published studies on yield in foxtail millet were successfully projected on the reference map (Table S1). The studies included ten diverse experimental crosses with 11 parental lines and 2092 progeny lines and two population types including three F 2 and seven recombinant inbred lines (Table 1), while the population size ranged from 124  to 543 individual genotypes (Wang et al. 2019). The original QTL were dispersed unevenly on all the nine chromosomes of foxtail millet (χ 2 = 439.3), the highest number of QTL were found on chromosome 9 with 90 QTL, followed by chromosomes 5 (n = 80), 7 (n = 58), 1(n = 57) and chromosome 8 had the fewest QTL with 19 QTL (Table S1).
Among the 448 initial QTL, 351 (78.3%) and 97 (21.7%) were found under normal and drought stress conditions, respectively (Fig. 1a). Under normal conditions, the number of QTL varied from 16 QTL on chromosome 8 to 66 QTL on chromosome 5. The distribution of QTL under drought conditions was also diverse on various chromosomes; 25 QTL were located on chromosome 9, and the lowest number of QTL (4 QTL) were located on chromosome 1. The number of QTL per trait ranged from 7 to 118, where TN, PH, and BM exhibited the highest number of initial QTL with 118, 92, and 62 QTL, respectively (Fig. 2). Moreover, the distribution of QTL for each trait varied across environments; TN and PH QTL were mainly evaluated under normal conditions, with 114 and 80 QTL, respectively. The number of initial QTL reported under drought stress conditions ranged from 1 PL to 46 WUE for each trait.
The 95% CI for each QTL differed from 1.45 to 49.05 cM with a mean of 18.37 cM. Approximately 89 (19.8%) of the collected QTL had a CI less than 10 cM, whereas the CI of 196 QTL (43.7%) was between 10 and 20 cM (Fig. 1c). The average PVE for the initial QTL was 7.5%, with a maximum of 61.4% (Fig. 1d). Among the 448 initial QTL, 26 QTL (5.8%) exhibited a PVE greater than 15% (Fig. 1b), and the logarithm of odds ratio (LOD score) ranged from 2.03 to 49.75, with a mean value of 5.8 for combined normal and drought stress in foxtail millet.
Under normal conditions, the average value of 95% CI was 20.65 cM, with approximately 11.68% of the collected original QTL demonstrating a CI lower than 10 cM, and 53.2% a CI lower than 20 cM. The PVE by the single QTL fluctuated from 1.99 to 61.4% with a mean of 6.3%. Among the 351 initial QTL detected under normal conditions, 297 QTL (84.6%) indicated a PVE less than 10%, while 54 QTL (15.3%) explained more than 10% of the phenotypic variance, and the logarithm of odds ratio (LOD score) ranged from 2.03 to 49.75 with a mean value of 5.72 (Table S1). Among the 97 initial QTL found under drought stress, 48 QTL (49.4%) exhibited a CI of less than 10 cM, and 85 QTL (87.6%) exhibited a CI of less than 20 cM. PVE value explained by these initial QTL ranged between 4.96 and 27.83 (average: 10.86%). The range of LOD scores for drought tolerance QTL was 2.03 to 19.52, with a mean of 5.18.

Drought-tolerance associated meta-analysis of QTL
The collected 448 individual QTL were mapped to the consensus genetic map, and 41 meta-QTL (MQTL) regions were identified on the nine chromosomes of foxtail millet (Table S3). The number of MQTL per chromosome extended from seven MQTL on chromosome 2 to two MQTL on chromosome 8 (Table S3).
Under normal conditions, 38 MQTL were found on all the chromosomes (Table S3). MQTL5-1 with 23 original QTL, had the highest number of QTL, followed by MQTL9-2 and MQTL1-4, with 22 initial QTL, and CIs of 2.83, 1.98, and 6.5 cM, respectively (Supplementary Table S3). MQTL5-5, with 14 initial QTL had the highest PVE (17.30%), containing two studies for three traits including PH, BM, and YLD, followed by MQTL5-4 with 11.85% PVE, which can be considered the most effective QTL for the involved traits. Some MQTL for different traits were identified under normal conditions; for instance, MQTL5-1 consisted of four studies for six traits, including BM, PH, TN, NN, PL and SD, and MQTL9-2 consisted of three studies for five traits, including YLD, TN, PH, BM, and NN.

Estimation of QTL-overview index
The QTL-overview index was considered for cM units over all chromosomes (Fig. S2). A total of 33 overview peaks were achieved, of which 20 exceeded the average statistic value for each chromosome and represented "real QTL" regions. Additionally, seven of the 33 peaks exceeded the high-value threshold (measured as 0.0743 for nine chromosomes) and represented "QTL hotspot" regions. These seven overview peaks corresponded to seven of the 41 MQTL (17.03%) and contained 99 initial QTL (22.09%). Two of the seven MQTL on chromosome 2 (28.5%) had a high overlap with overview peaks. Moreover, after mapping the 50 QTL onto the chromosome 7 consensus genetic map, one of the five MQTL (20%) was detected in hotspot overview peaks. A total of six overview peaks were obtained for 68 QTL on chromosome 5; two of the six (27 QTL) MQTL were located in the overview peak interval. Moreover, four overview peaks were identified on chromosome 9, of which two (33 QTL) were identified based on meta-QTL results (Fig. S2).

Distinguishing differentially expressed genes (DEGs) in foxtail millet under drought conditions
The drought-responsive genes were collected from RNA-seq data of six independent studies and one microarray experiment (Table S8). According to the RNA-seq data analysis, 9356 DEGs (6016 up-regulated and 3340 down-regulated) were found under drought stress compared to normal conditions. In addition, the microarray assay suggested 26 downregulated and 74 up-regulated DEGs (Table S5).
A total of 1631 and 314 common genes were identified using the Venn diagram among the DEGs derived from RNA-seq and microarray data and the genes located in the MQTL regions for all the 41 MQTL and the 10 MQTL with CI lower than 1 Mb, respectively ( Fig. S3; Table S9).

Functional annotation of DEGs located in the MQTL regions
Among the 8564 genes located in 41 MQTL, 1631 were differentially expressed candidate genes and 6933 constitutively expressed candidate genes (Table S10, Fig. S4). The Singular Enrichment Analysis (SEA) was used to inspect the gene ontology enrichment using the DEGs list located in 41 identified MQTL regions (1631 DEGs out of 8564 genes) and 10 MQTL regions with an interval of less than 1 Mb (314 DEGs out of 1605 genes).
The results of the GO analysis demonstrated that 1070 of the 1631 genes were functionally annotated using the AgriGO v.2.0 database (Table S11). The number of annotated genes ranged from 0 to 73 (MQTL5-4), and the percentage of significant GO terms varied from 0 to 65.12. Only 19 out of 40 MQTL contained a total of 231 significant GO terms. In the biological process category (BP), GO analysis of the DEGs found in the MQTL regions showed that signaling, metabolic process, biological regulation, cellular process, and response to stimulus were significantly enriched (Fig. S4). Several molecular function categories were significantly enriched, including binding, transcription regulator, structural molecule, electron carrier, catalytic, and transporter activity. Significantly enriched terms of cellular Fig. 2 The number of initial QTL for drought tolerance and yield-related traits (used in meta-analysis of QTL) on each foxtail millet chromosome under normal (N) and drought stress (DS) conditions component (CC) category, were comprised membraneenclosed lumen, organelle, symplast, macromolecular complex and extracellular region (Fig. S4).

Metabolic pathways of drought-regulated genes
To supplement the GO analysis, MapMan analysis was employed to search for the putative functions of the identified drought-response genes in the MQTL regions. The Map-Man file contained the fold-change data and locus identifiers for 1631 drought-responsive genes located in the 41 MQTL (Table S12). The overview of regulatory pathways of the 1631 DEGs indicated the up-regulation of 86 transcription factors (TFs), 27 genes associated with protein degradation, and 48 genes related to protein modification in foxtail millet under drought stress (Fig. S5a). Genes involved in hormone signaling pathways, including IAA, ABA, ethylene, and jasmonic acid, were also enriched (Fig. S5a). Additionally, the regulation overview revealed the identification of 17 genes for calcium regulation, 44 genes for receptor kinase, six genes for G-proteins, one gene for phosphoinositide, and five genes for light signaling pathways (Fig. S5a).
The overview of stress response pathways showed that genes involved in abiotic stress response, signaling, redox tion of initial QTL on nine foxtail millet chromosomes (black lines) and position of MQTL with a 95% confidence intervals (red areas). c QTL-overview index for the QTL on consensus genetic map. d Proportion of phenotypic variance explained for each initial QTL. e Coefficient of reduction in CI from mean original QTL to MQTL state, peroxidases, and glutathione S-transferases were enriched (Fig. S5b). Genes involved in cell wall modification (xyloglucosyl transferase, xyloglucan endotransglycosylase) and cell wall degradation (polygalacturonase) represented another set of important components that were enriched (Fig. S5b). The results of mapping the DEGs to secondary metabolite pathways overview revealed that lignin, terpenoids, flavonoids, and phenylpropanoids pathways were among the enriched pathways (Fig. S5c). The cellular overview pathway indicated that genes encoding abiotic stressrelated miscellaneous enzyme families (misc) and development were up-regulated in foxtail millet under drought stress (Fig. S5d). Although the genes involved in transcription pathways were mapped in 41 MQTL, our results uncovered that most of the TF families belonged to the eight myeloblastosis (MYB), nine Apetala2/Ethylene Responsive Element Binding Proteins (AP2/EREBPs), nine Basic Helix-Loop-Helix (bHLH) genes, seven homeobox genes (HB), six Chromatin remodeling factors, two SET-domains, 11 WRKY domains, three MADS-box genes, 10 Histone genes, 7 C2H2 zinc finger family, five putative DNA-binding, 4 G2-like transcription factor family (GARP) and four basic leucine zipper (bZIP) (Fig. S5e).

Validation of promising candidate genes via qRT-PCR
Expression pattern of the nine drought-regulated promising candidate genes was examined in the two droughtcontrasting genotypes (Ilam1 as a drought-tolerant genotype and Kerman3 as a susceptible genotype) through qRT-PCR to validate the MQTL results (Table S7, Fig. 5, Fig S6). The results demonstrated that eight genes were up-regulated, and one gene was down-regulated in both genotypes, which was consistent with the RNA-seq data. Interestingly, the expression level of the eight genes was significantly higher in the tolerant genotype even in the normal conditions and up-regulation was much more in Ilam1 (log2 fold change ranged between 2.17 and 8.6 in the drought-tolerant genotype, while it was 1.1 -1.6 in the susceptible genotype).

Discussion
Drought stress is one of the major abiotic stresses affecting crop production. Distinguishing the traits, pathways, and gene variants related to drought tolerance is valuable for optimizing crop yields under drought conditions (Qie et al. 2014). Exploring major QTL that exhibit consistent effects across diverse genetic backgrounds and experiments is crucial for the applied and accurate use of the identified consensus genomic regions in conventional plant breeding via marker-assisted selection (MAS). MQTL mapping can reduce adverse impacts of genetic background, population type, and planting environment on QTL and beneficially combine QTL details in several backgrounds. Meta-analysis of QTL has been used to integrate the QTL data to pinpoint regions of the genome that are most frequently involved in the trait variation and to narrow down the confidence interval of the QTL (Yang et al. 2021). The meta-analysis results intensely depend on the quality of the studies identifying QTL, markers density on the genetic map, quality of QTL projection, and confidence intervals of QTL (Goffinet and Gerber 2000).

Advantages and applications of the identified MQTL
Based on integrating a consensus genetic map with 1,116 molecular markers and through meta-analysis, the current study combined 448 initial QTL on nine foxtail millet chromosomes into 41 MQTL regions. The average 95% CI of the recognized MQTL was 5.23 cM, approximately  Table 2 3.5 times narrower than the mean CI of the collected original QTL. The genetic 95% CI of 58.6% and physical CI of 61.0% identified MQTL were narrower than 5 cM and 2 Mb, respectively. The meta-analysis reduced the initial QTL' average CI for drought stress and normal conditions from 9.87 cM and 20.75 cM to 3.38 cM and 5, respectively. Among the 41 MQTL identified in the current research, the physical distance of 10 MQTL (24.39%) was below 1 Mb (Table S3). MQTL with a physical distance of < 1 Mb can be assumed as significant chromosomal regions for identifying of yield-related and drought-tolerance-associated genes in foxtail millet. The QTL-overview index was also calculated, indicating 33 peaks exceeding the statistics' average value. Furthermore, 7 of the 33 peaks exceeded a high-value threshold and represented hotspots for QTL (Fig. S2). The physical distance of 4 MQTL was less than 1 Mb (MQTL2-6, MQTL5-5, MQTL9-1, MQTL9-2), which co-localized with overview peaks suggesting that these MQTL were the most important regions involved in yield-related traits. MQTL2-6 (0.9 Mb, 4.34 cM), MQTL5-5 (0.4 Mb, 3.4 cM), MQTL9-1 (0.65 Mb, 4.1 cM) and MQTL9-2 (0.81 Mb, 5 cM) can be beneficial for fine-mapping studies of QTL. These MQTL involve important traits, including BM, TN, PH, WUS, YLD, NN, and TN (Fig. S2).

Ortho-MQTL with possible cereal-wide application
Recent advances in genome sequencing provide a positive outlook for comparative genomics research on cereals. Previous researches demonstrated coordinated trait evolution in crops with conserved orthologous chromosomal positions The data are shown on a log2 scale as the mean ± SD (n = 3 each) for the gene expression ratio with treatments relative to con- LOC101777378) and the isolation of genes by map-based cloning at orthologous chromosomal positions (Quraishi et al. 2011). There is a high degree of synteny between the genomes of foxtail millet and economically significant cereals like rice, maize, and barley. The comparative meta-QTL analysis provides the opportunity to identify orthologous regions that control important traits, such as drought tolerance in foxtail millet, as demonstrated by this study.
The orthologous regions can be considered candidates for functional validation or, at the very least, a source for identifying SNPs associated with desirable traits for the foxtail-breeding program based on genomic selection. In the present study, ten ortho-MQTL were recognized, representing conserved genomic regions, and therefore, may be recommended for use across cereals ( Table 2). One of the detected ortho-MQTLs, ortho-MQTL9-3, was identified in each of the four crops examined (Fig. 4). OrthoMQTL9-3 is associated with the important traits including BM, PH, TN, WUE, and YLD. Of the 41 identified foxtail millet MQTL, 31 were not located on the syntenic regions of the studied species' chromosomes. These MQTL in foxtail millet, which have no orthologs in other studied species, are expected to provide novel sources for controlling crop yield and drought tolerance.
Ortho-MQTL mining can detect underlying regulatory genes with an evolutionary past and a conserved function. The conservation of ortho-MQTL suggests that they are associated with regulatory elements, each of which influences many genes (Quraishi et al. 2011;Khahani et al. 2020). Foxtail millet MQTL1-5 derived from QTL associated with PH, WUE, and TN are located in a syntenic region on rice containing orthoMQTL1-5. Alternative oxidase 1C (AOX1C), known to be induced by drought and salinity in rice (LI et al. 2013), was identified in the orthoMQTL1-5 genomic region of foxtail millets.
AOX participates in optimizing respiratory metabolism, facilitating the continuous turnover of the tricarboxylic acid cycle, and regulating the cellular energy status in response to changing biological processes or environmental conditions (LI et al. 2013). Dissection of orthoMQTL9-3 in rice led to the detection of OsMYB52. This gene was characterized as a transcription factor and plays a crucial role in plant development, secondary metabolism, hormone signal transduction, disease resistance, and tolerance to abiotic stress (Katiyar et al. 2012;Xiong et al. 2014). The deduced proteins of LOC101762009, LOC101752968, and LOC101752968 contained a highly conserved basic helix-loop-helix (bHLH) domain. They had an orthologous gene in the identified maize MQTL intervals (Almeida et al. 2013), which is located at MQTL9-3. In plants, bHLH TFs participate in a physiological process and regulate plant responses to numerous abiotic stresses (Thirunavukkarasu et al. 2014). The orthologous genes identified in this research could be utilized in cereal breeding programs.

Potential candidate genes with a possible role in drought tolerance of foxtail millet
A total of 1,631 drought-responsive genes were identified in 41 MQTL regions based on the available RNA-seq and microarray data. The MQTL regions with CIs of less than 1 Mb were explored looking for the up-regulated genes Fig. 6 The schematic view of the molecular response to drought stress in Setaria italica. Some drought-responsive candidate genes are demonstrated, located on the MQTL with a CI of less than 1 Mb. ↑: up-regulated. ↓: down-regulated under drought stress, resulting in the detection of 104 potential candidate genes (Table S13) involved in sensing and signaling, transcription regulation, ROS inhibition and adaptation to abiotic stress of foxtail millet. These potential candidate genes were categorized into several GO terms, some of which will be discussed in the following sections (Fig. 6).

Sensing and signaling pathways
Several signal transduction-related genes were among the candidate genes. The wall-associated kinase (WAK) gene family, encodes a protein containing an intracellular Ser-Thr kinase domain and responds to abiotic stress affecting signal transduction between the cell wall and cytoplasm (Zhang et al. 2005). Based on the results of the MapMan analysis, receptor kinases and calcium regulation were among the most significantly enriched gene groups involved in signaling pathways (Fig. S5a). A gene encoding WAK (SiWAK8, LOC101766117) was found in MQTL5-4 and up-regulated in the leaf during drought stress (Table S13, Fig. 6). Furthermore, the gene encoding CSC1-like protein (SiCSC1, LOC101785619), which may act as an osmosensitive calcium-permeable cation channel was also located in MQTL3-4 and up-regulated in response to drought stress (Fig. 5).
The Na + /Ca 2+ Exchanger (NCX) protein family is a member of the Cation/Ca 2+ exchanger superfamily, and the role of this protein family in abiotic stresses is reported. Our results indicated that a gene encoding a mitochondrial proton/calcium exchanger protein (SiNCX, LOC101770330) belonging to the NCX protein family is located at MQTL1-4 and up-regulated under drought stress at the seedling stage (Table S13, Fig. 6) (Shi et al. 2018;Singh et al. 2015). In addition, OsCDPK7 encoding a calcium-dependent protein kinase, was reportedly involved in the signaling pathway during dehydration stress in rice (Saijo et al. 2000). Moreover, SiCDPK3 (LOC101763140) was found in MQTL5-5 and up-regulated under drought treatment. SiCDPK3 might be involved in signal transduction to activate downstream genes during drought stress (Table S13, Fig. 6).
Phytohormones have critical roles in regulating plant growth and abiotic stress responses; consequently, genes involved in hormone signaling pathways, such as cytokinin, auxin, and ethylene components, are involved in abiotic stress tolerance (Egamberdieva et al. 2017;Tang et al. 2017). The MapMan analysis of the drought-responsive genes located on the MQTL positions revealed that several DEGs were involved in hormone signaling pathways, including IAA, ABA, ethylene, and jasmonic acid (JA) (Fig. S5a, b). Cytokinins can be crucial in enhancing crop drought tolerance, including regulating cell division and photosynthetic parameters, such as chlorophyll (Pospíšilová et al. 2000).

Transcription regulation
Transcription factor families such as WRKY, AP2/EREBP, and C2H2 can mediate plant responses to abiotic stresses and plant growth by regulating various stress-inducible genes. Several drought-responsive genes encoding transcription factors were identified in the MQTL regions detected in the current study (Fig. S5a, b).
Prior research demonstrated that C2H2 transcription factors play a significant role in abiotic stress signaling in Arabidopsis and rice (Davletova et al. 2005;Devi et al. 2016). Overexpression of the zinc finger (C2H2-type) 1 3 family (ZAT12) has also been reported to increase drought tolerance in Arabidopsis thaliana (Davletova et al. 2005). In the current research, it was revealed that SiZAT12 (LOC101753014), which belongs to the zinc finger (C 2 H 2 -type) family, is located in the MQTL5-5 region, and was up-regulated by drought stress (Table S13, Fig. 5,  Fig. 6).

ROS inhibition
Due to the altered energy dissipation process in photosynthesis, respiration, and photorespiration, reactive oxygen species (ROS) are produced in various organelles, such as chloroplast, mitochondria, and peroxisomes, during droughtinduced oxidative stress. Conversely, plant antioxidant enzymes (e.g., catalase (CAT) and ascorbate peroxidases (APX), catalyze the dismutation of H 2 O 2 into H 2 O and O 2 , which could play a significant role in quenching oxygen species and establishing homeostasis in plants (Nematpour et al. 2019). Peroxidases and glutathione S-transferases were elevated in foxtail millet under drought stress, as determined by MapMan analysis (Fig. S5b). Catalase activities are also enhanced with increasing H 2 O 2 production. Catalase plays important protecting role in H 2 O 2 removal. Under drought stress, the transgenic tobacco lines had higher CAT activity than the wild type (WT) (Zhou et al. 2012). Lata et al., (2013) reported that accumulation of ascorbate peroxidases helps the foxtail millet to survive under drought stress. It has been reported that overexpression of OsAPX2 increased APX activity and improved drought stress tolerance in rice (Zhang et al. 2013). Based on the achieved results, the gene coding for catalase isozyme 3 (SiCAT3, LOC101775718) and ascorbate peroxidase (SiAPX3, LOC101776821) located in MQTL1-3 and MQTL7-5, respectively, and up-regulated under drought stress.

Photosynthetic response and adaptation
Photosynthesis is a significant process that has sustained plant life on earth and is essential to plant growth and development. A significantly positive correlation was observed between proteins related to carbohydrates and ATP metabolism. The gene coding for fructose-bisphosphate aldolase (SiFBA, LOC101771156) with potential roles in drought response was identified among the DEGs and was located in MQTL2-3. SiPKS_ER (LOC101763588) is involved in carbohydrates and ATP metabolism by encoding the photosynthesis-related protein quinone-oxidoreductase (found in MQTL2-4). Furthermore, SirbcLA (LOC101768770), involved in photosynthesis and codes for RUBISCO (ribulose bisphosphate carboxylase), was located in MQTL3-6 and down-regulated under drought stress conditions (Fig. 6). Reduced gene expression may be due to the need to resist drought stress, the flow of energy shifts from the biosynthesis of macromolecules involved in photosynthesis to the respiratory pathways for energy supply (Amirbakhtiar et al. 2021).

Conclusions
This study employed an integrative meta-analysis survey to detect the most important genome regions and candidate genes involved in the genetic control of drought tolerance and yield maintenance in foxtail millet. The meta-analysis of the initial 448 QTL resulted in the detection of 41 MQTL regions, while 9 and 38 MQTL were explicitly identified under drought stress and normal conditions, respectively.
The expression pattern of nine drought-regulated candidate genes was validated via qRT-PCR in the two drought contrasting genotypes (Ilam1 as a drought-tolerant genotype and Kerman3 as a susceptible genotype). Intriguingly, the expression levels of the eight up-regulated genes (SiZAT12, SiPFK2, SiWRKY24, SiACO, SiCSC1, SiHSP, SiCURT1B, and SiAP2/ERF) at normal conditions and their up-regulation by drought stress were significantly higher in the drought-tolerant genotype (Fig. 5, Fig.S6), indicating that they might be involved in drought tolerance.
Seven hotspot regions with high QTL-overview index values, corresponding to seven MQTL, were identified, with the physical intervals of four MQTL being less than 1 Mb (MQTL2-6, MQTL5-5, MQTL9-1, MQTL9-2). MQTL2-6 and MQTL5-5 overlapped with the MQTL were specifically determined under drought stress, indicating that these MQTL are involved in drought tolerance and yield-related traits (TN, PH, WUS, YLD, NN, TN). These findings facilitate understanding the molecular basis of drought tolerance and yield-related traits in foxtail millet. In addition, they might be useful for developing drought-tolerant cultivars via molecular breeding or genetic engineering.
Author contributions FL and HDR performed the meta-analysis, drafted the manuscript and drew the graphs. FL analyzed the microarray data. The project was conceived, coordinated and supervised by Z-SS, who also revised the manuscript. The final manuscript was checked by AI and BN. All the authors reviewed and approved the final version of the manuscript. Data availability All data generated or analyzed during this study are included in this published article and its supplementary information files.