Contrasting genotypes for drought tolerance were identified based on multi-year evaluation of genotypes for various morphophysiological traits (Maheswari et al., 2016). The drought-tolerant genotype SNJ201126 and susceptible genotype HKI161 were used for development of a mapping population following the single cob method. Initial biparental crossing between tolerant and susceptible genotypes, was conducted during the rainy season of the year 2014. Subsequently, the F₁ generation was self-pollinated for nine generations to develop a mapping population consisting of 264 single-plant progenies.

The RIL populations, consisting of 264 single-plant progenies and their parents, were planted in a single-row plot of 2.5 m, with 60 cm spacing between rows and 25 cm between plants. Separate experiments were conducted under both WW and WD conditions, following a randomized complete block design (RCBD) with three replications. In the experiment under WW conditions, populations were grown under normal growth conditions until maturity, with irrigation whenever required. However, in the experiment under WD conditions, irrigation was provided only up to the vegetative stage, i.e., 45 days after sowing (DAS); this was followed by imposition of a water deficit for a period of 10 days in order to expose plants to drought stress, which coincided with the anthesis-silking interval (ASI). These two sets of experiments with different treatments were repeated for two seasons: specifically, the rainy season of 2018 and the post-rainy season 2018–19. The experiments were carried out with appropriate plant protection measures in place, and in accordance with recommended practices for growing healthy crop. During the 2018 rainy season, the weekly average temperature varied between 17.9°C and 31.3°C, with relative humidity of 57.1%–84.5%; during the post-rainy season, the temperatures recorded fell between 11.4°C and 30.4°C, with relative humidity of 40.9%–83%. The total rainfall recorded was 377 mm and 16 mm during the rainy and post-rainy seasons, respectively ( Figure S1 ).

Various morphophysiological parameters were recorded under both WW and WD conditions: these consisted of Normalized Difference Vegetation Index (NDVI), net CO₂ assimilation rate (A_net), stomatal conductance to water vapor (g_s), transpiration rate (TR), leaf temperature (LT), and anthesis-silking interval (ASI). Yield-contributing traits (i.e., cob height (CH), cob weight (CW), total biomass (TB), and grain yield (GY) were recorded for three representative plants of each genotype. NDVI was measured using a GreenSeeker^® 505 device (Manuel NTech Industries Inc., Ukiah, CA, USA). This device measures the reflected light on the canopy of crops in the 660 nm (red) and 770 nm (near-infrared) bands. The NDVI value for any given point in the image, at a particular phenophase of the crop, is equal to the difference in the intensities of reflected light in the red and infrared range divided by the sum of these intensities. A_net, g_s, LT, and TR were measured using the LI-6400 portable photosynthesis system (LI-COR Instruments, Inc., Lincoln, NE, USA).

The mean phenotyping data for the mapping population (consisting of 264 single-plant progenies), as evaluated under both WW and WD conditions, were used for a cluster analysis, carried out by the average-linkage distance method using SAS^® version 9.3 statistical software (SAS Institute Inc., Cary, NC, USA; Cary, 2011). The cluster analysis using combined mean data on morphophysiological and yield-related traits [i.e., relative water content, canopy temperature depression, quantum yield, Soil Plant Analysis Development (SPAD) chlorophyll meter readings, NDVI, proline content, net CO₂ assimilation rate, g_s, TR, LT, plant height, CH, cob length, cob girth, number of kernel rows per cob, number of kernels per row, CW, grain yield, TB per plant, and harvest index] under WD conditions was used to group the mapping population into diverse groups. Specifically, the population was grouped into eight clusters ( Table S1 ) based on the average distances between all pairs of cluster members between the clusters. The RIL IDs (264 in total), their corresponding cluster IDs, and the corresponding distances are shown in Table S1 . A subset of 79 RILs were selected from these eight clusters in such a way as to fully capture the genetic diversity of the mapping population ( Table S2 ). The frequency distribution of this subset, when compared with the whole population for different traits, was found to represent the phenotypic variation of the population. Along with this subset of 79 RILs, the parents SNJ201126 and HKI161 (in triplicate) were subjected to SNP genotyping at Bionivid Technology Pvt. Ltd., Bengaluru, India. The Illumina NGS workflow for SNP genotyping was employed, as follows. First, the young leaves of 15-day-old seedlings of each genotype were used for DNA isolation using the DNAeasy Plant Mini Kit (Qiagen, Hilden, Germany). Next, the DNA quality and quantity were determined via agarose gel electrophoresis and a NanoDrop spectrophotometer, respectively. For library construction, DNA was fragmented randomly and adapters were ligated to the 5′ and 3′ ends. These fragments were then amplified by PCR and purified from the gel. Clusters were generated by loading the library into a flow cell, where fragments were captured on a lawn of surface-bound oligos complementary to the library adapters. After cluster generation, the templates were sequenced. Sequencing was carried out using Illumina SBS technology, a system that detects single bases as they are incorporated into template strands. As all four reversible terminator-bound dNTPs are present during each sequencing cycle, natural competition minimizes incorporation bias and reduces the raw error rate in comparison to other technologies. This enables highly accurate base-by-base sequencing that virtually eliminates sequence-context-specific errors, even within repetitive sequence regions and homopolymers. Sequencing data were subsequently converted to raw data for analysis. The Illumina sequencer generates raw images utilizing sequencing control software for system control and base calling through an integrated primary analysis software tool called Real-Time Analysis (RTA). The base calls (BCL) binary was converted to FASTQ format using Illumina package bcl2fastq. The total numbers of bases and reads, along with values of GC (%), Q20 (%), and Q30 (%), were calculated for all samples. The raw sequences of genotyping data were deposited in the NCBI database (http://www.ncbi.nlm.nih.gov/sra/PRJNA913688) with accession number PRJNA913688.

The phenotyping data of the subset of 79 RILs, which were generated in two seasons under WW and WD conditions, were used for statistical analysis, as the sequencing data were generated for the same set. Individual and combined (i.e., over seasons and treatments) analyses of variance (ANOVAs) were carried out for morphophysiological traits; Pearson’s simple correlation coefficients were also calculated and heritability estimates were made using SAS version 9.3. For combined analyses, the homogeneity of variance component was determined using Bartlett’s test (Bartlett, 1937). Broad-sense heritability was calculated as per the following formula:

where σ 2 G is the total genotypic variance and σ 2 P is the total phenotypic variance. Frequency distribution histograms for all traits were generated using Matplotlib tools (Hunter, 2007). Matplotlib is a cross-platform data visualization and graphical plotting library for Python. The Pyplot module was used to generate plots, and the Scipy.stats module (Virtanen et al., 2020) was used to compute and draw histograms of the WW and WD data; evenly spaced points over a specified interval on the x-axis were created using numpy.linspace, and the norm of the probability density function is displayed on the plots. The Kolmogorov–Smirnov method (Kolmogorov–Smirnov Test, 2008) was used to test for normality. Descriptive statistics were calculated using SAS, version 9.3.

The population consisted of 79 RILs, which demonstrated wide variation in their morphophysiological and yield-related traits (i.e., NDVI, A_net, g_s, TR, LT, ASI, CH, CW, TB, and GY) under WW and WD conditions ( Table S3 ). All traits were affected by WD. Histograms of the frequency distributions of these traits in the WW and WD conditions are presented in Figure 1 . All traits followed a near-normal distribution, with the exception of stomatal conductance, which is positively skewed in the WW condition. Based on the significance values obtained using the Kolmogorov–Smirnov test, the traits A_net, LT, CH, CW, GY, and TB followed a normal distribution under both under WW and WD conditions, whereas the traits g_s and TR followed a normal distribution under the WD condition ( Table S4 ). This indicates that the selected RILs captured the genetic variability of the entire population to be utilized for QTL identification. The descriptive statistics of the mapping population under WW and WD conditions for morphophysiological and yield-related traits, including coefficients of variation (CV%), skewness, kurtosis, and heritability, are provided in Table S5 . Moderate broad-sense heritability (H²), ranging between 0.45 and 0.71, was observed for all traits, and a wide range of coefficients of variation was also observed among these traits.

A combined ANOVA over trials across seasons (i.e., rainy vs. post-rainy) and conditions (i.e., WW vs. WD) indicated significant interactions among season, treatment type, and genotype for most traits. For total biomass (TB), all interaction effects were significant except seasons × treatments ( Table S6 ). For ASI, the interaction effects for treatments × genotypes and seasons × treatments × genotypes were non-significant. The significant variation by genotype, environment, and their interaction for a number of traits indicated that these traits were influenced by both genetic and environmental factors. A simple correlation coefficient analysis revealed significant positive correlations of NDVI with CH, CW, GY, and TB; A_net with g_s and TR; g_s with TR; CH with TB, GY, and TB; CW with GY and TB; and GY with TB under both WW and WD conditions. In addition, a significant positive correlation of NDVI with LT, and of CH with CW and GY, was observed under WD conditions ( Table S7 ).

The numbers of raw SNPs and polymorphic SNPs between the parents after filtering with MAF< 5%, along with their distribution in various chromosomes, the number of mapped SNPs in the linkage map, and the average marker interval, are presented in Table 1 . The largest number of markers (219) was detected on chr 2, and the smallest (65) on chr 10. In the present study, 27 QTLs were identified; of these, 13 were detected under WW conditions, 12 under WD conditions, and two (qCH1–1 and qCW2–1) under both WW and WD conditions ( Table 2 ; Figure 2 ). Major and minor QTLs were detected only in chromosomes 1, 2, 3, 5, 6, and 7 (out of the 10 maize chromosomes studied under WW and WD conditions), and no QTLs were detected in chromosomes 4, 8, 9, or 10. Under WD conditions, QTLs for CH (qCH1–1 and qCH1–2) and ASI (qASI1–1) and for NDVI, CW, TB, and GY were detected in chromosomes 1 and 2. Similarly, three (qA_net3–1, qg_s3–1, qCH3–1), two (qg_s6–1, qGY6–1) and one (qg_s7–2) QTLs were detected in chromosomes 3, 6, and 7, respectively, under WD conditions ( Figure 2 ). QTLs with PVE% greater than 15% were regarded as major QTLs. Among the 13 QTLs detected under WW conditions, three major QTLs were identified for traits A_net, g_s, and TR, with LOD scores ranging from 2.54 to 6.07 and PVE% ranging from 15.86% to 21.47%. Ten minor QTLs were detected for traits LT, TR, g_s, TB, CH, and ASI, with LOD scores ranging from 2.54 to 3.2 and PVE% ranging from 6.75% to 14.91% ( Table S8 ). Among the 12 QTLs detected under WD conditions, three were major QTLs identified for traits NDVI, g_s, and CH, with LOD scores ranging from 3.49 to 4.45 and PVE% ranging from 15.05% to 18.59%. Nine minor QTLs were detected for traits A_net, NDVI, ASI, CH, g_s, TB, and GY, with LOD scores ranging from 2.52 to 3.37 and PVE% ranging from 9.69% to14.25% ( Table 2 ). Among these, two QTLs were detected under both WW and WD conditions: these were a minor QTL for the CH trait and a major QTL for the CW trait, located on chromosomes 1 and 2, respectively.

For the trait NDVI, one major and one minor QTL were detected on chr 2, with LOD scores of 3.93 and 3.19 and capturing 18.59% and 14.02% PVE, respectively. For the A_net trait, major and minor QTLs were detected on chr 3, with LOD scores of 3.24 and 2.52 and capturing 17.24% and 14.25% PVE, respectively. For the g_s trait, two major QTLs on chr 6 and three minor QTLs, one on chr 3 and two on chr 7, were detected. The LOD scores for these ranged from 2.54 to 4.45, and PVE% ranged from 9.69% to 18.21%. For the TR trait, one major QTL was detected on chr 7 (with a LOD score of 6.07 and PVE% of 21.47%), and two minor QTLs were detected, one on chr 1 (with a LOD score of 2.87 and PVE% of 10.81%) and one on chr 3 (with a LOD score of 2.72 and PVE% of 10.11%).

For the ASI trait, three minor QTLs were detected, two on chr 1 and one on chr 6, with LOD scores of 2.61, 3.02, and 3.2 and PVE% of 13.92%, 14.91%, and 13.56%, respectively. For the CH trait, one major QTL, with a LOD score of 3.49 and PVE% of 15.05%, was detected on chr 1. In addition, three minor QTLs were detected on chromosomes 1, 3, and 8, with LOD scores of 2.52, 2.8, and 2.77 and PVE% of 13.84%, 11.18%, and 14.9%, respectively. For the TB trait, three minor QTLs were detected, two on chr 2 and one on chr 5, with LOD scores of 2.53, 2.58, and 2.55 and PVE% of 13.51%, 11.50%, and 13.04%, respectively. For the GY trait, two minor QTLs were detected, one on chr 2 and one on chr 6, with LOD scores of 2.78 and 2.54 and PVE% of 10.54% and 11.33%, respectively. Finally, for the CW trait, one major QTL was detected on chr 2, with a LOD score of 3.03 and PVE% of 17.91%.

Of the 27 QTLs detected under WW and WD conditions, three pairs of co-localized QTLs were identified. For traits TR and ASI, two QTLs (qTR1–1 and qASI1–1) were co-localized at the marker interval rs128441140–rs128842621 on chr 1, with LOD scores of 2.87 and 2.61 and PVE% of 10.81% and 13.92%, respectively. For traits g_s and GY, two QTLs (qg_s6–2 and qGY-6–1) were co-localized at the marker interval S6_89546385–S6_179562549 on chr 6, with a LOD score of 2.54 and 2.54 and PVE% of 15.86% and 11.77%, respectively. Finally, for the g_s and TR traits, two QTLs (qg_s7–1 and qTR7–1) were co-localized at the marker interval S7_166501967–rs130671858 on chr 7, with LOD scores of 2.73 and 6.07 and PVE% of 11.3% and 21.07%, respectively.

Eleven significant digenic epistatic QTLs for traits A_net, g_s, TR, LT, ASI, CW, and TB were detected ( Table 3 ; Figures S2A, B ). The corresponding LOD scores ranged between 5.0 and 5.91, and PVE% ranged from 16.60% to 37.43%. The negative epistatic values (add-by-add) indicated that there was a stronger epistatic effect of recombinant genotype than of parental genotype. The epistatic effect of QTLs was negative for traits TB, A_net, and LT, but positive for traits g_s, ASI, TR, A_net, and CW. The genomic region on chr 2 between the markers rs129243511 and S2_229825946 showed epistatic interaction for the CW trait on chr 8, located between the markers S8_145945904 and S8_148216407. This interaction contributed 37.43% of the PVE. Two epistatic interactions were identified for A_net. The first involved the region on chr 1 between the markers S1_3798004 and rs128441140, which showed epistatic interaction with the region on chr 3 located between the markers rs129555629 and S3_183844216. This interaction contributed 17.60% of the PVE. For the second, a region on chr 2 between the markers rs72722896 and rs276685886 showed epistatic interaction for A_net with a region on chr 3, located between the markers S3_2276048 and S3_89193637. This interaction contributed 16.71% of the PVE. Finally, the genomic region on chr 2 between the markers rs131971876 and rs129196105 showed epistatic interaction for the TR trait with the region on chr 7 located between the markers S7_131791490 and S7_136261108. This interaction contributed 29.55% of the PVE.

Using the MET (multi-environmental trials) module of ICIM, a total of 53 QTLs were identified for various traits ( Table S9 ). Of these, 21 QTLs were common to those identified by the ICIM-ADD method. One QTL for each of the traits g_s, ASI, and GY and two QTLs for the LT trait were identified using the ICIM-ADD method but were not detected by MET analysis. Conversely, the MET analysis identified 32 QTLs for traits A_net, CH, ASI, TB, GY, NDVI, LT, and CW that had not been detected using the ICIM-ADD method. The proportion of phenotypic variation captured by additive and dominance effects [PVE(A)] ranged from 0.13% to 14.36%, and the proportion captured by additive- and dominance-by-environment effects [PVE(A by E)] for corresponding QTLs ranged from 0% to 7.92% ( Table S9 ). Thus, PVE(A by E) was significantly lower than PVE(A). Most QTLs detected through the MET module of ICIM were non-significant. However, traits A_net, ASI, g_s, and TR were found under MET to make greater contributions to phenotypic variation (with contributions ranging from 5.24% to 21.7%), they were not found to contribute to the significant QTL × E interaction effect ( Table 4 ).

The functional annotation of genes associated with the major and minor QTLs for morphophysiological and yield-related traits and their biological/molecular functions under WD and WW conditions are shown in Tables 5 , S10 , respectively. The QTL regions with important genes imparting stress tolerance for different traits belonged to the categories of signal transduction (GY: Zm00001eb297570, protein serine/threonine phosphatase); transcription factors [NDVI: Zm00001eb118010, G2-like transcription factor 27; ASI: Zm00001eb295810, NAC-type transcription factor (NAC87)]; transporter activity (A_net: Zm00001eb146040, chloride channel protein; g_s: Zm00001eb324180, sugar carrier protein C; TR: Zm00001eb015510, phospholipid-transporting ATPase; CH: Zm00001eb363270, calcium-transporting ATPase); cell wall biosynthesis and its organization (TR: Zm00001eb144960, lipoxygenase; TR: Zm00001eb145080, pectin acetyl esterase); photosynthesis (g_s and TR: Zm00001eb324240, chlorophyll a/b-binding protein, chloroplast); and carbon utilization (CH: Zm00001eb002270, glyceraldehyde phosphate dehydrogenase B1).

Table S11 shows the names of the QTLs, their chromosome locations, left markers, right markers, QTL intervals (cM), the physical location of each QTL region (i.e., start and end), its size, and the number of genes within the QTL region. Of the 14 QTLs identified under WD conditions, two QTLs, qg_s7–2 and qCH1–1, encompassed a single gene, namely the Nicotinamide adenine dinucleotide phosphate-binding Rossmann-fold superfamily protein (Zm00001eb316940) and LIM homeodomain proteins transcription factor 2 (Zm00001eb011970), respectively. Although the physical distances between the markers of the QTLs qg_s6–1, qASI1–1, qCW2–1, and qGY6–1 were large, linkage with the trait could not be ascertained, meaning that these were not used. The physical sizes of the QTLs qNDVI2–1, qNDVI2–2, qA_net3–1, qg_s3–1, qCH1–2, qCH3–1, qTB2–2, and qGY2–1 were below< 5 Mbp; these encompassed 22, 108, 264, 35, 31, 64, 372, and 372 protein-coding genes, respectively. Genes that played a significant role in WD tolerance in these QTL intervals were also identified. The details of QTLs detected in earlier studies in the QTL regions, i.e., qA_net3–1, qg_s3–1, qASI1–1, qCH3–1, and qGY6–1, are listed in Table S12 .

The genetic map in this study consisted of 1,241 SNP markers spread across 10 linkage groups, covering a total genetic distance of 5,471.55 cM, with an average marker density of 4.4 cM per marker. RIL mapping populations are widely used for QTL identification (Li et al., 2013; Yang et al., 2016), and in this study, 27 QTLs were identified for various morphophysiological and yield-related traits under both WW conditions (13 QTLs) and WD conditions (12 QTLs). Notably, for the NDVI trait, both a major and a minor QTL were detected on chr 2, with bin position 2.10 under WD conditions. Previous studies have also reported QTLs for NDVI and plant height in two BC₁F_2:3 backcross populations (LPSpop and DTPpop), with 18 QTLs identified in total (Trachsel et al., 2016). In the DTPpop population, QTLs for NDVI influencing stay-green habit (SEN6) were also detected in bins 8.01 and 2.07. The present study identified several QTLs associated with various morphophysiological and yield-related traits under both WW and WD conditions. For the A _net trait, a major QTL (under WW conditions) and a minor QTL (under WD conditions) were identified on chr 3 (bp 3.05), which also showed the presence of a minor QTL for the TR trait under WW conditions. QTLs for the TR trait were also detected on chrs 1 and 7. Two major QTLs for the g_s trait were identified on chr 6 (under WD conditions) and two minor QTLs, one on chr 3 (under WD conditions) and one on chr 7 (under WW conditions), were also identified. A previous study by Yu et al. (2015) identified 32 QTLs associated with chlorophyll a, chlorophyll b, total chlorophyll content, net CO₂ photosynthetic rate, stomatal conductance, intercellular CO₂ concentration, and TR. Another study reported on the photosynthetic performance of maize grown in drought environments (Zhao and Zhong, 2021). In a previous study conducted by Pelleschi et al. (2006), 19 major QTLs were identified for various physiological traits under drought-stressed and WW regimes. QTLs for the stomatal conductance trait have been found on all chromosomes in maize except for chr 5 (Quarrie et al., 1994; Sanguineti et al., 1999). In the present study, two minor QTLs for the LT trait (under WW conditions) were identified on chrs 6 and 8. Three minor QTLs for the ASI trait were also identified: two on chr 1 (one under WW and the other under WD conditions) and one on chr 6 (under WW conditions). A total of 33 QTLs for the ASI trait under WD stress have been reported in earlier studies, distributed across all chromosomes (Semagn et al., 2013). Additionally, QTLs for ASI under both WW and WD conditions were also identified in another study (Hu et al., 2021). In this study, we identified one major QTL under WD stress and two minor QTLs (one under WW and one under WD conditions) for cob height on chr 1, as well as one minor QTL on each of chrs 3 (under WD conditions) and 8 (under WW conditions). Previous studies have also reported QTLs for cob height on all 10 chromosomes (Beavis et al., 1994; Lubberstedt et al., 1997; Li et al., 2016), with 21 QTLs identified for both PH and CH in three common genomic regions in two biparental populations (Li et al., 2016). However, our study identified major and minor QTLs (qCH1–2 and qCH1–1) at different positions on chr 1 compared with those identified in previous studies. QTLs for cob height have also been reported on chrs 3, 4, 5, 6, and 8 in previous studies, but at different bs (Yang et al., 2008; Zhao et al., 2018; Tadesse et al., 2020). Grain yield is a highly complex trait that is influenced by many genes, each exerting only a small effect (Hallauer and Miranda, 2010). Earlier studies have reported on a QTL atlas that includes the major genes for GY and its associated traits (Zhou et al., 2020). In our study, we identified QTLs for the CW and GY traits on chr 2, which were physically located at bps 241,494, 911–241,497,812 and bps 14,676,667–14,679,369, respectively. These QTLs are different from the previously reported meta-QTLs (Zhou et al., 2020). We also identified another QTL for grain yield on chr 6, located at bps 179,555,644–179,566,050, which is also different from the meta-QTLs that have been previously reported on this chromosome. Additionally, we identified three minor QTLs for the TB trait: two on chr 2 (one under WW and the other under WD conditions) and one on chr 5 (under WW conditions). Similar studies have also identified QTLs for biomass production and leaf area. For example, Chen et al. (2011) identified seven QTLs for biomass production and leaf area in the marker interval bnlg1832–P2M8/j (bp 1.05) on chr 1. Rahman et al. (2011) identified a QTL for yield on chr 1 that was co-located with the QTLs for root traits, total biomass, and osmotic potential in a region of about 15 cM. These studies suggest that genetic control of biomass and yield is complex and involves multiple QTLs distributed across different chromosomes and genomic regions.

The missing proportion of phenotypic variance for a trait may be attributed to epistasis (Carlborg and Haley, 2004), a term that refers to non-allelic interaction that can modify the degree of phenotypic expression by suppressing or enhancing the expression of interacting genes (Mackay et al., 2014; Rahman et al., 2017). Although our study was limited by a smaller population size, it is possible that epistatic effects, in addition to a few major and minor loci, contributed to the variation observed in the morphophysiological and yield-related traits studied, such as A_net, g_s, TR, LT, ASI, CW, and TB. However, further validation using a larger mapping population is necessary to confirm these findings. To better understand the stability of QTLs across different environments, it would be beneficial to evaluate this population in multiple locations, across multiple years, and under different treatments, such as variable temperature and watering regimes (Boer et al., 2007). Although our study used a smaller population size, a joint analysis over multiple years was conducted to establish the stability of QTLs and to evaluate the interactions between QTLs and the environment. Our findings suggest that, with the exception of g_s with respect to the QTL located at the marker interval S8_152991912–S8_123369436 on chr 8, most of the traits were associated with non-significant effects of environment in terms of the expression of QTLs, which captured only a small proportion of the phenotypic variation explained by additive-by-environment effects.

In this study, analysis of the genomic region of the QTLs using in silico methods led to the identification of candidate genes associated with stress tolerance. The QTL qNDVI2–1 contained two important genes: Golden 2-like (GLK) transcription factor 27 (Zm00001d007962, GRMZM2G173882, and Zm00001eb118010) and cold-regulated 413 plasma membrane protein 2 (Zm00001d007968). The Golden 2-like (GLK) transcription factor 27 gene plays a crucial role in regulating chloroplast growth and development, and also contributes to the maintenance of stay-green traits (Chen et al., 2016; Qin et al., 2021), whereas the cold-regulated 413 plasma membrane protein 2 gene enhances osmotic stress tolerance through enhanced expression of AtCor78/AtRD29A (Zhou et al., 2018). In the QTL interval of qNDVI2–2, three genes were identified: barren stalk 2 (Zm00001eb084940), AP2-EREBP transcription factor 131 (Zm00001eb084810, GRMZM2G087059, and Zm00001d003884), and potassium high-affinity transporter 1 (Zm00001eb084630, GRMZM2G093826, and Zm00001d00386). These genes contribute to stress tolerance through their involvement in signaling (Yao et al., 2019), plant growth and development (Xie et al., 2019), and facilitation of potassium (K⁺) ion distribution in shoots (Qin et al., 2019). Overall, these genes in the QTL region for the NDVI trait play roles in maintaining stress tolerance through signaling and transporter activity.

In our study, we also found that the QTL interval of qA_net3–2 contained several genes that encode various transcription factors, such as ABI3/VP1 transcription factor 31 (Zm00001eb144270, Zm00001d042460), AP2-like ethylene-responsive transcription factor (GRMZM2G141638, ZM00001EB144510), ARR-B-transcription factor 6 (Zm00001eb144290, Zm00001d042463), bHLH transcription factor 132 (Zm00001d042482 GRMZM2G114873), AP2/EREBP transcription factor 53 (Zm00001d042492), and GRAS transcription factor 7 (GRMZM2G013016). These transcription factors play important roles in the regulation of downstream genes through binding to DNA elements in the promoter regions of the target genes. These genes also play vital roles in plant growth, development, hormone signaling, and stress responses (Kimotho et al., 2019). Additionally, we identified transporters, such as the magnesium (GRMZM2G139822, ZM00001EB144080) and proline transporters 1 and 2 (ZM00001EB144010, GRMZM2G078024), that are essential for maintenance of membrane homeostasis. Magnesium is a vital component of chlorophyll, which plays a crucial role in absorbance of sunlight during photosynthesis. It also acts as a phosphorus carrier in plants and is essential for phosphate metabolism (Hermans et al., 2013). Overall, our findings suggest that the qAnet3–2 QTL region plays a vital role in the regulation of transcription factors and transporters, which are crucial for various physiological processes in plants.

The QTL qg_s3–1 contains three genes that are critical for plant stress tolerance and development. The bHLH transcription factor 70 (GRMZM2G397755 and Zm00001d039459) is a key regulator of stress-related genes, enabling the plant to activate adaptive responses under various abiotic stresses. This transcription factor also plays a vital role in synthesis of flavonoids, which are essential for stress tolerance (Qian et al., 2021). The epidermal patterning factor-like protein (GRMZM2G077219, Zm00001d039470, and Zm00001eb121050) is involved in the development of stomatal cells in the upper epidermis of plant leaves (Lee et al., 2012). This protein is crucial for gaseous exchange in plants, which is vital for photosynthesis and respiration. The third gene, glutaredoxin 12 (Grx12) (Zm00001eb121030, GRMZM2G303044, Zm00001d039468, and Zm00001eb121020), is a stress-related redox sensor that plays a significant role in signaling through glutathione. Grx12 enables glutathione to play a signaling role through glutathionylation of target proteins (Zaffagnini et al., 2012). The minor QTL qg_s7–2 includes the gene Zm00001eb316940, which encodes for the NAD(P)-binding Rossmann-fold superfamily protein. Although the exact function of this protein is not well understood, it is known to bind to NAD and NADP and is predicted to play a role in metabolic processes.

The QTL qCH1–1 was expressed under both WW and WD conditions, and it is associated with the gene Zm00001eb011970, which encodes the LIM zinc-binding domain-containing protein DA1–2. This protein is related to ubiquitin binding and the expression of cell cycle genes, which contribute to long-distance phloem transport (Park et al., 2014). The LIM domain is a protein interaction domain that is involved in many cellular processes, including signal transduction, transcriptional regulation, and cytoskeletal organization. The QTL qCH1–2 encompasses the gene encoding calcium-dependent protein kinase 36 (CDPK 36), which translates elevated calcium concentration into enhanced protein kinase activity and subsequent downstream signaling events (Singh et al., 2018). Calcium is an important secondary messenger in plants that plays a crucial role in a wide range of signaling pathways, including stress responses, development, and growth.

The QTL qCH3–1 encompasses two important genes: receptor-like protein kinase (RLK) G-type lectin S-receptor-like serine/threonine-protein kinase (ZM00001EB151330), and homeobox transcription factors (Zm00001d043231 and ZM00001EB151130). The first gene, RLK, is a member of the cell-surface receptor-like protein kinase family, which is critical in perception of signals. RLKs have active functions in various physiological processes such as plant growth and development, and in responses to both biotic and abiotic stresses (Ye et al., 2017). The G-type lectin S-receptor-like serine/threonine-protein kinase is involved in signaling during plant reproduction and defense against pathogens. Homeobox genes play a crucial role in specifying cell identity and positioning during embryonic development (Khan et al., 2018). They regulate various developmental processes, such as organogenesis, morphogenesis, and differentiation. The QTL qCH3–1 includes the genes encoding RLK and homeobox-transcription factors, which play critical roles in plant growth, development, and responses to both biotic and abiotic stresses.

The QTLs qTB2–1 and qGY2–1 were found to be located within the same genomic region and contained several genes encoding for various transcription factors. ABI3/VP1 transcription factor 12 is involved in seed maturation and germination, and the AP2/EREBP transcription factor is known to regulate gene expression in response to various environmental cues. Calcium-dependent protein kinase 5 (CDPK 5) is a key component of the calcium signaling pathway and plays a crucial role in stress responses. The GRAS transcription factor is involved in various developmental processes, including root and shoot development, while MYB-related transcription factor 20 is involved in the regulation of secondary metabolism. The RING zinc finger protein-like RING/U-box superfamily protein and zinc finger C3HC4-type family of proteins are involved in protein degradation and the regulation of gene expression, respectively (Kimotho et al., 2019). In addition to these transcription factors, the QTL region also contained various stress-inducible genes, such as abscisic acid receptor PYL9, which plays a critical role in drought and salt stress responses, and the dehydration-responsive element-binding protein 1D, which is involved in the regulation of water stress-responsive genes. The guard cell S-type anion channel SLAC1 is involved in stomatal regulation, while the WD40 repeat-like superfamily of proteins are involved in various developmental processes, including cell division and differentiation. The naked endosperm, sucrose synthase, and xyloglucan galactosyltransferase genes are all involved in seed development and are essential for proper seed maturation.

Overall, this study has identified several candidate genes that play crucial roles in various physiological processes, including the perception of external signaling, expression of functional proteins involved in stress tolerance, and the regulation of gene expression in response to environmental cues. These findings have important implications for crop improvement, as they provide valuable insights into the molecular mechanisms underlying stress tolerance and growth and development in plants.