A set of seven parents and five hybrids was selected for this experiment (Table 1). The parental lines include two drought-tolerant inbreds, HKI1105 and CML425, and five drought-sensitive inbreds, HKI1128, HKI161, LM13, LM14, and LM17. The hybrid panel includes one drought-tolerant hybrid ADWLH2 (HKI1105 × CML425), two moderately tolerant hybrids HM8 (HKI1105 × HKI161) and HM9 (HKI1105 × HKI1128), and two sensitive hybrids PMH1 (LM13 × LM14) and PMH3 (LM17 × LM14).

Seven parental inbreds and the five hybrids were grown under controlled glasshouse conditions at the National Phytotron Facility, ICAR-IARI, New Delhi. The glasshouse conditions include, 30/26°C (day/night) with relative humidity of 50–55%. All the genotypes were sown in plastic cups (6”) filled with sandy loam soil in a randomized complete block design and were replicated three times with three plants per replication (Thirunavukkarasu et al., 2017). Plants were watered at field capacity until the emergence of the third leaf. Stress treatment was given to each genotype after the emergence of the third leaf by withdrawing water and keeping them without water for five consecutive days. A second set of genotypes was maintained simultaneously with regular watering at field capacity to serve as a control (Min et al., 2016; Thirunavukkarasu et al., 2017).

A set of 52 drought-responsive candidate genes distributed across the maize genome was selected to understand their expression in the parental inbreds and the hybrids under a drought stress condition, of which, 14 genes were selected from a genome-wide association mapping study owing to their association with drought tolerant functions in maize (Thirunavukkarasu et al., 2013). The remaining 38 genes known for drought tolerance were selected from the public data base. In the preliminary search, several genes were collected from the Plant Transcription Factor Database (PlantTFDB) and MaizeGDB. The domains of the collected genes were searched using the Pfam database using a cut-off value of 1.0. The genes that satisfy the E-value of 1 were further used for analysis. The genomic sequences of all the genes were retrieved from Ensembl Plants (http://plants.ensembl.org/index.html). For each gene, the gene ontology, under different categories, such as molecular, cellular and biological functions, was identified using Blast2Go (Conesa and Götz, 2008). Then, the annotations of all the sequences were manually checked for drought-related functions. Finally, 38 genes with drought-related functions were short-listed and were further cross-checked by the Stress Responsive Transcription factor Database (STIFDB), which classifies the genes according to different stresses.

Selected genes were categorized into different drought-related functional groups (Table 2), including stomatal regulation (10 genes), root development (10 genes), ROS scavenging (10 genes), hormone signaling (12 genes), photosynthesis (5 genes), and sugar metabolism (5 genes). Features of genes, such as length and position in the genome, were identified using Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html).

Root length (RL), chlorophyll content (CC) and relative water content (RWC) were measured in the stress and the control plants. Three plants per replication were used to measure the root length and other phenotypic traits. Utmost care was taken to avoid the damage to root system while removing the soil and measuring of root length. The RL was measured 5 days after the stress treatment in stressed and well-watered control seedlings and is expressed in centimeters. The CC (%) was recorded daily from day 1 to day 5 of the stress treatment period using a portable SPAD chlorophyll meter (SPAD-502, Minolta Camera Co. Ltd., Japan). The RWC was measured 5 days after the stress period using the following formula:

where FW is the fresh weight, DW is the dry weight, and TW is the turgid weight (Barrs and Weatherley, 1962).

After 5 days of stress treatment, the parent and hybrid seedlings were removed from the cup, and the shoot and root samples were collected. Total RNA from the shoots and roots was separately isolated using Qiagen RNeasy columns (Qiagen, Hilden, Germany). From the isolated RNA, 2.5 μg RNA was treated with DNase I (Promega, Madison, WI) as per manufacturer's protocol to remove any genomic DNA contamination. The concentration of the isolated total RNA was determined using a Thermo Scientific NanoDrop 1000 spectrophotometer (Thermo Scientific, Delaware, USA). One microgram of RNA was used to synthesize cDNA using an Affymetrix Kit (Santa Clara, California, USA).

Primers for each gene were designed using IDT PrimerQuest (http://www.idtdna.com/scitools/applications/primerquest/default.aspx), and the 18s RNA coding gene was selected as the internal control (Nakashima et al., 2007). The primer pairs for the genes were designed to produce only a single desired amplicon in the quantitative real-time PCR (qRT-PCR). The details of the primer pairs of the genes are given in Supplementary Table S1.

One-step real-time qPCR was performed in an Mx3005P qPCR system (Stratagene, La Jolla, CA) using a SYBR Green-based PCR assay (with ROX as the optional reference dye). qPCR was performed using a total reaction volume of 25 μl, which consisted of 12.5 μl of the SYBR green RT-PCR master mix (Affymetrix, Santa Clara, California, USA), 1 μl of reverse transcriptase, 0.5 μl of ROX dye, 1 μl of cDNA, 2 μl of each primer, and 8 μl of nuclease free water. Reverse transcription was performed at 50°C for 30 min and was then terminated at 95°C for 10 min; PCR was then performed with 40 cycles of 94°C for 3 s, 58–60°C for 1 min, and 72°C for 30 s with three technical replicates.

The expression data for the genes in the different treatments, tissues, and genotypes were normalized by subtracting the mean reference gene CT value from the individual CT values of the corresponding target genes (ΔCT). The relative abundance of the transcripts was derived using the expression 2^−ΔΔCT, where ΔΔCT is the difference between the ΔCT of the condition of stress treatment and the ΔCT of the control.

The quantitative measurement of the F₁ hybrid expression level of each gene related to the average of the two parents (mid-parental level) was determined using a d/a ratio method (Guo et al., 2003, 2006). Considering d is dominant, a is additive, and μ is the mid-parental value (average of the parental expression), the dominant (d) is measured by the difference between the F₁ (hybrid) and the average of the parents (μ), and the additive (a) is measured by the difference between the parent (either maternal or paternal) and the average of the parents (μ) i.e., d=F1-μ and a=Parent-μ In case of a complete dominant gene action of the P₁ (maternal) allele, F₁ = P₁, then d/a = 1. Similarly, d/a = −1 explains the complete dominant gene action of the P₂ (paternal) allele. In case of an additive gene action, if F₁ = μ, then d/a = 0.

Based on the above-mentioned concept, the transcript expression level is considered a phenotype of each gene, and the F₁ hybrid expression was measured relative to the parental expression (Guo et al., 2003). In the case of multiple loci, “a” and “d” represent the composite effect of the respective gene action. Genetically, the genome of an F₁ hybrid constitutes one dose of a genome from each parent, and an additive allelic expression in the hybrid would be expected to be equal to the mid-parental value [(maternal + paternal)/2].

From the actual expression data of the parents and the respective hybrids, the deviation of the F₁ hybrid from the average of the two parents was calculated as d = F₁ – μ. Then, the deviation of the maternal parent from the average was calculated as a = Parent _maternal – μ. The d/a ratio was then used to measure the gene action in the hybrids. If d/a = 0, this indicates additive gene action; if d/a = 1, this indicates the dominant gene action where the hybrid expression is skewed toward a maternal parent; if d/a = −1, this indicates a dominant gene action where the hybrid expression is skewed toward the paternal parent; if d/a > 1, this indicates an over-dominant gene action where the hybrid expression is skewed toward the maternal parent; and if d/a > − 1, this indicates over-dominant gene action where the hybrid expression is skewed toward the paternal parent, considering that the maternal parental expression is higher than the paternal parent expression. When the paternal parent expression is higher than the maternal parental expression, then the sign of the d/a ratio is reversed in over-dominant situations.

If the d/a ratio is between −0.5 and +0.5, this is considered an additive gene action, where both parental lines contribute equally to the gene expression, or the F₁ hybrid expression is equal to that of the parental mean. If the d/a ratio > ±0.5, this is considered a non-additive gene action that is skewed toward any one of the parental lines.

The reduction in the CC was a mere 2.5% in the parents HKI1105 and CML425 and was 3.2% in ADWLH2 on day 5 of stress compared to day 1 (Figure 1). The parents HKI161 and HKI1128 showed an 11.1 and 13.5% reduction, respectively, in the CC on day 5 compared with day 1 under the stress condition. Hybrids HM8 and HM9 showed more reduction in the CC to the level of 10.6 and 12.9%, respectively, on day 5 of the stress. A severe reduction in the CC (~50%) was noticed in the sensitive parents LM13, LM14, and LM17 and their hybrids PMH1 and PMH3 under the stress condition compared to their respective WW control. The growth of the roots under the stress condition in HKI1105 and CML425 and their hybrids was comparable to their respective WW control on day 5 since the reduction in the RL was ~2% (Figure 2A). An approximately 10% reduction in the RL was observed under stress in parents HKI161 and HKI1128 and in their respective hybrids HM8 and HM9 compared with the respective WW control. The RL was reduced more than 21% in HM8 and HM9 compared to ADWLH2 under the stress condition. LM14 showed a greater reduction in root growth (36%), followed by LM13 (34.1%) and LM17 (33.3%) compared with the respective WW control. The RWC, on day 5, was reduced to 3.5% in HKI1105, 6.2% in CML425 and 4% in ADWLH2 under the stress condition, and the RWC was further reduced in parents HKI161 and HKI1128 to the levels of 5.6 and 10.5%, respectively, under the stress condition compared to their respective WW control (Figure 2B). An approximately 30% reduction in the RWC was observed in all three of the sensitive parents (LM13, LM 14, and LM 17) as well their hybrids (PMH1 and PMH3).

Five hybrids and their respective parental lines were exposed to stress, and the expression level of 52 candidate genes was measured under the stress condition. These 52 genes represent six important functional categories, including stomatal regulation, root development, ROS scavenging, hormone signaling, photosynthesis, and sugar metabolism.

The CC, RL and RWC of the drought tolerant parents (HKI1105 and CML425) were on par or slightly lower under stress than their WW controls. The performance of the hybrid ADWLH2 was also better under stress since its parents (HKI1105 and CML425) were highly tolerant to drought stress. There was a gradual reduction in all three parameters in the parents HKI161 and HKI1128. Despite the sensitive paternal parents, the growth performance of the hybrids HM8 and HM9 was moderate under stress since the maternal parent (HKI1105) in both hybrids was stress tolerant.

A robust root system is very important to absorb water, to maintain a higher level of water potential in the aerial systems and for the ability to transpire for a longer duration under stress conditions (Hammer et al., 2009). The presence of a good root system in the tolerant lines provided an advantage for their growth and development under the water stress condition. In water-deficient soils, a tolerant plant adapts and modifies its root architecture in a variety of ways (Herder et al., 2010). Leaves with a good amount of water and nutrient content are essential to support the root system (Price et al., 2002). The high RWC with a high CC in the tolerant lines under the stress conditions maintained the water potential as well as provided an adequate level of photosynthesis and nutrients to the plant system.

The d/a ratio indicated that the expression of genes in the shoot and root tissues of the hybrids followed three patterns, including a mid-parental level (additive), one that was similar to that of the parental line (dominant) and one that fell outside of the parental expression (over-dominant) (Figure 5). The d/a ratio also showed a more or less equal distribution of dominance from both of the parents in the hybrids; however, the maternal influence was slightly higher (216) over the paternal (207) for all the genes, hybrids and tissues combinations.

In this experiment, the direction of influence by the parental lines is of great significance in all five hybrids. The direction of skewness did not affect the ADWLH2 performance under stress since the alleles from both parents contributed to the drought tolerance. However, the direction is very important for the hybrids HM8 and HM9, since one of the parents, HKI1105, was stress tolerant and the paternal parents, HKI161 and HKI1128, respectively, were not tolerant. Nearly 40% of the genes in HM8 and HM9 in the different functional categories showed dominance or over-dominance toward their respective sensitive parent HKI161 and HKI1128 in the shoots and roots. The influence of the sensitive parents in the hybrids reduced the transcript level in the respective hybrids, thereby affecting the crucial biological function under the stress condition. However, other genes showed either an additivity or non-additivity toward the tolerant parent HKI1105, which reduced the level of damage caused by stress and provided the necessary support for maintaining biological activities. PMH1 and PMH3 were sensitive to stress, and since their respective parents were stress sensitive, the transcript level in the hybrids also either followed the mid-parental level or was dominant toward the parents.

The gene action does not depend upon the parent per se, whereas it is characterized by the combination of the parental lines involved. For example, me2 and nced2 in the stomatal regulation category in AWDHL2 showed dominance toward the parent HKI1105 in the shoots and roots, whereas the genes in HM8 and HM9 showed dominance toward the other parent, HKI161 and HKI1128. A hormone signaling gene, camta5, explained the over-dominance toward LM14 in PMH1, but the same gene showed over-dominance toward the parent LM17 in PMH3. The variation in the direction and the magnitude of the same genes in the different hybrids, especially in the hybrids involving one common parent, suggested that the extent of the genetic diversity between the parents also decides the gene action (Stupar et al., 2008).

The gene action for a given gene may either be dominant toward any one of the parents or additive in nature. On a very rare occasion, the gene action was switched from dominant to additive for any one of the tissues. In addition, the genes that showed non-additive gene action were tissue-specific. For example, aldh3 showed non-additive gene action toward CML425 in the shoots, and the same gene skewed toward the maternal parent HKI1105 in the roots. The tissue-specific expression of genes is one of the factors influencing the nature of gene action (Guo et al., 2003, 2004). A variation in the gene sequence or an allelic polymorphism will produce differential levels of transcripts, which is a major reason for the presence of variations in the transcript level. The structural variations and polymorphisms at the nucleotide level are one of the reasons for the transcriptional variation of the candidate genes under stress conditions (Messing and Dooner, 2006; Stupar and Springer, 2006). The intra-specific variation in the transcripts is attributed to the cis-acting factors in the genome (Stupar and Springer, 2006; Springer and Stupar, 2007).

Nearly 80% of the drought-responsive candidate genes explained partial-, complete-, or over-dominance toward any one of the parents. Several studies in Drosophila (Gibson et al., 2004), Arabidopsis (Vuylsteke et al., 2005), and maize (Stupar et al., 2008) have found that the expression of genes falls outside of the parental range. The non-additive action outside the parental range in the hybrids could be due to novel gene regulation owing to complementation of alleles under stress conditions (Stupar et al., 2008). The genome-wide expression of genes in hybrids was predominantly additive in nature. However, a non-additive gene action from a few to several loci in the genome was observed in various crops (Auger et al., 2005; Guo et al., 2006; Swanson-Wagner et al., 2006; Uzarowska et al., 2007; Pea et al., 2008; Li et al., 2009).

A variation in non-additive gene action might be due to the tissue-specific expression of genes (shoots or roots), and the specific growth conditions in which the genotypes were exposed (seedling or flowering and control or stressed). A non-additive gene action, especially over-dominance of the crucial genes under the stress condition, is also the result of the linked-loci of those genes. Since, drought tolerance is a complex trait, and various component traits are involved in stress tolerance, these traits are supported by a cascade of pathways inter-linking with each other. Further, these pathways are triggered in response to stress and results in the abundance of specific transcripts that are needed for the survival of the plant and to maintain various biological functions. Linked loci also play a critical role in the overexpression of genes (Swanson-Wagner et al., 2006) under the stress condition. The over-dominant gene action of these genes under the stress condition was also contributed by a range of post-transcriptional processes, including splicing, translation, protein folding and stabilization. Additionally, the role of small RNAs, such as microRNA and small interfering RNA (siRNA) cannot be ignored (Swanson-Wagner et al., 2006).

Drought stress in plants triggered a series of stimuli and activated several genes to act in response to stress. The series of genes involved in the stress mechanism are the signaling genes that activate other down-stream genes, i.e., genes that protect the cellular components and functions (uptake of water, ions and nutrients, regulation of photosynthesis, and etc). Many of the genes that were part of the signaling cascade were activated in the tolerant lines under stress, which explained that these genes were actively involved in sensing the stress and were able to trigger other genes. One of the key hormones is ABA, which plays a variety of roles under stress conditions, including stomatal regulation. An increased accumulation of ABA during stress stimulates the closure of stomata and decreases the transpiration rate under stress conditions (Boursiac et al., 2013).

The ABA signaling pathway comprises three components, which are a group of ABA receptors, including protein phosphatase 2C and SnRK2. SnRKs are serine/threonine protein kinases that represent key regulators of plant responses to different stresses (Soon et al., 2012). SnRK2 is an important signaling molecule that phosphorylates its downstream targets, including the transcription factors NAC, bZIP, HSF, MYB, WRKY, and RAV1 (belonging to the AP2–ERF family) (Furihata et al., 2006; Fujita et al., 2009; Kim et al., 2012; Feng et al., 2014). ERF is another drought-responsive transcription factor that is stimulated under the effect of ABA but is integrated with other two hormones, jasmonic acid and ethylene, which induce the closing of stomata (Cheng et al., 2013).

Root development is strongly influenced by abiotic stress conditions. A subfamily of NAC transcription factors is involved in various abiotic stresses such as drought, cold, salinity (Mao et al., 2012). Under a water stress condition, vacuolar proton pumps reduce the water potential. The V–type PPase pump enhances ion homeostasis, which regulates the osmotic balance and, hence, copes with drought (Li et al., 2008; Pasapula et al., 2011). Oxidative stress commonly occurs when the plant undergoes any kind of stress; the production of antioxidants is one of the ways the plant copes with the stress. In stress conditions, ascorbate peroxidase functions as a H₂O₂ reductant in ROS detoxification (Foyer and Halliwell, 1977). Under oxidative stress conditions, ascorbate peroxidase reduces H₂O₂ to water and mono-dehydroascorbate. Under drought stress, glutathione reductase activity is increased in the roots and leaves, which might facilitate the defense against ROS (Gallé et al., 2013). Additionally, SOD and glutathione peroxidase are also involved in ROS scavenging (Dietz et al., 2006). Recent advances have also emphasized the role of MAPK under drought stress in ROS detoxification (Ning et al., 2010).

In response to drought stress, photosynthetic activity is reduced because of a decline in stomatal conductance as well as Rubisco activities, resulting in lower carbon fixation followed by an over reduction in the components of ETS system. The up-regulation of the Rubisco activase precursor under the drought stress condition reveals that a higher activation state could also have affected the Rubisco activity (Ramachandra Reddy et al., 2004). Another transcription factor, CAMTA, is identified as a regulator of the photosynthetic machinery where the T-DNA insertion line of AtCAMTAs is observed at a low photosystem II efficiency under drought stress (Pandey et al., 2013). Abiotic stresses lead to major alterations in carbohydrate metabolism and most likely modulate metabolism sugar signaling pathways, which interact with stress pathways. Drought stress affects sucrose metabolism and decreases the activities of soluble and insoluble forms of invertases (Zinselmeier et al., 1995, 1999). Sucrose synthase enzyme is a key enzyme involved in the synthesis and cleavage of sucrose. Under water deficit conditions, the activity of sucrose synthase is enhanced in the cleavage direction and increases the sucrose synthesis (Castrillo, 1992; Yang et al., 2000).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.