Introduction

Wheat (Triticum aestivum L.) is one of the most important staple food grains on earth. After years of domestication in fertile crescent region, the present day wheat varieties evolved that are adapted to a wide range of environmental conditions ranging from high-humidity regions like South America to low-humidity regions like India, Nigeria, Egypt, and Australia (Pont et al. 2019). Wheat yields are affected by both biotic and abiotic stresses. Among the abiotic stresses, drought and heat are the most severe stresses that affect the life cycle of the crop (Zampieri et al. 2017). These two factors progressively became important due to global climate change (Akbarian et al. 2011; Zampieri et al. 2017). The global temperature has been presumed to increase by 0.18 °C per decade (Hansen et al. 2012). An estimated 6% loss in wheat production occurs globally for every 1 °C rise in temperature (Asseng et al. 2015). Higher temperatures have a direct influence on plant growth and crop yields owing to reduced opportunities for photosynthesis since the life cycle is truncated (Bita and Gerats 2013; Stocker et al. 2013). At grain-filling stages, the rise in temperature adversely affects the quantity and quality of wheat grains thereby inducing various cellular and metabolic changes.

The advances in biotechnology including recent progress in genomics and molecular breeding have enabled wheat researchers to use the technology in mitigating the detrimental effects of HS (Lamaoui et al. 2018). Molecular markers improve the efficiency of conventional plant breeding by indirectly selecting for the gene of interest. Simple sequence repeats (SSRs) or microsatellites are efficiently and frequently used for the identification of the quantitative trait loci (QTL) linked to drought and HS tolerance (Pinto et al. 2010; Paliwal et al. 2012; Mondal et al. 2015). Such QTL linked markers hold promise in forward marker-assisted selection for improved selection efficiency and development of stress-tolerant cultivars (Rai et al. 2018; Sandhu et al. 2019). QTLs for various physiological and morpho-agronomic traits have been studied internationally; significant progress has been made in the mapping of QTLs for yield and contributing traits under HS in wheat (Pinto et al. 2010; Sukumaran et al. 2018; Tadesse et al. 2018). However, the use and implementation of QTLs is carried out to a limited extent due to differences in genetic backgrounds, environments and poor understanding of expression and regulation of genes governing the trait. In addition, the small effect of a single QTL enforces the breeder to prioritize for stable and strong effect QTLs which limit the whole efforts (Tricker et al. 2018). A viable option is the development of markers from the genic regions of HS transcriptomes that may help in population genetics and association studies through the identification of genes/QTLs linked to component traits of HS tolerance.

Stress-associated genes are induced in response to heat stress (Chauhan et al. 2011; Rampino et al. 2012; Lamaoui et al. 2018). Though significant progress has been made in research on HS tolerance, information on the genes involved in HS response is limited. In the past decade, most of the studies were restricted to the identification and mapping of QTL/genomic regions for constituent traits of HS tolerance (Pinto et al. 2010; Talukder et al. 2014; Mondal et al. 2015; Tadesse et al. 2018) but stress-associated genes lying in such genomic regions were known to a lesser extent. In the recent past, Acuna-Galindo et al. (2015) identified QTL hotspots in eight major clusters using the meta-QTL approach on linked SSRs and reported a few clusters harboring agronomically important genes. Li et al. (2004) reported that the SSRs lying in the gene regions may be involved in regulating the expression of respective genes. The genic SSR markers from candidate genes have greater potential in identifying marker–trait associations in germplasm collections involving diverse backgrounds and environments. Marker–trait associations of these candidate genes-based SSRs could prove very helpful in future genetic diversity and MAS studies.

Although several gene expression studies have identified differentially expressed stress-responsive genes in contrasting wheat genotypes under HS, yet the association with component traits of HS tolerance still remains to be learned (Qin et al. 2008; Kumar et al. 2017; Li et al. 2019). Keeping this in view, microsatellite markers were developed from the RNA sequencing analysis of Indian heat-susceptible and -tolerant genotypes. A total of 1216 differentially expressed genes (DEGs) were observed that contain more than 2000 SSR motifs. Among these, the stress-associated genes (SAGs) including the chaperones, transcription factors, signaling factors, etc. harbored 243 SSRs. In the present study, we assess the performance of candidate gene-based novel microsatellite markers in understanding marker–trait associations in indigenous and exotic germplasm collections and gene expression pattern of selected genes in contrasting genotypes.

Materials and Methods

Plant Materials

An international core set for abiotic stress comprising of 145 lines, received from CIMMYT, Mexico under the Generation challenge Program was evaluated under timely (TS) and late sown (LS) environments for three years. The experiment was laid out in a randomized complete block design with two replications in three rows of one-meter length each. The core set was used for screening of 243 candidate gene SSRs newly developed from whole transcriptome sequencing of contrasting wheat cultivars under HS.

A subset of 85 lines including 34 lines from this larger international core set and 57 elite Indian lines developed for different agro-ecological conditions were utilized for corroboration of marker–trait associations of identified candidate gene SSR markers. The subset mainly included parents of Indian and International mapping populations for drought and HS tolerance (Table S1). The subset experiment was laid out in alpha-lattice design in two environments (timely and late sown) in the years 2018–19. Each environment has two replications and each replication constitutes 8 blocks. Every block had 11 genotypes. For the TS environment, planting was done during the second week of November with recommended irrigation; while for the LS, planting on a delayed date during first week of January was carried out. The crop was maintained using standard cultivation practices prescribed for wheat.

Development and Scoring of SSR Markers

SSR markers were developed from RNA sequencing data generated on heat stress-tolerant (HD2985) and a heat stress-susceptible (HD2329) genotype (for details see, Kumar et al. 2017). Among the identified SSR motifs, 243 SSRs overlay on the stress-associated genes (SAGs) such as chaperones, transcription factors, signaling factors, etc. Therefore, a total of 243 SSR markers were developed from differentially expressed heat-responsive transcripts obtained under control (22 ± 3 °C) and high-temperature stress (42 °C, 2 h) conditions using the Microsatellite identification tool (MISA; https://pgrc.ipk-gatersleben.de/misa/misa.html). The sequences, repeat motifs and amplification conditions of 243 SSR markers can be obtained from Kumar et al. (2017). These newly synthesized candidate gene-based SSR markers were evaluated for their performance in 145 lines of the international core set. The PCR reaction profile was: DNA denaturation at 95 °C for 5 min followed by 35 cycles of 94 °C for 1 min, 55 or 60 °C for 1 min (depending upon primer), 72 °C for 1 min and finally 72 °C for a final extension of 10 min. Amplified PCR products of each reaction were separated on 3% metaphor agarose gel (Lonza, Rockland ME, USA) and were photographed using a Gel Documentation System, by keeping the magnification constant. Manual gel scoring was done for each gel picture based on the bands of the standard 100 bp DNA ladder. Every allele was scored as present (1) or absent (0) for individual SSR marker. The markers that produced the expected size of the amplicon with clear bands and showed polymorphism were further validated on a smaller set of 85 lines.

Phenotypic Characterization of Germplasm Lines

Phenotypic characterization of international core set was carried out in three different growing seasons (2013, 2014, and 2016) during TS and LS conditions for various agronomic and physiological traits. The gross plot size of the TS experiment was 1.38 m × 3.0 m with rows at 20 cm apart, whereas for LS experiments, the gross plot size was 1.08 m × 3.0 m with a row-to-row spacing of 18 cm. Data was recorded for days to heading (DH), days to maturity (DM), plant height (PH), flag leaf area (LA), 1,000 kernel grain weight (TGW), yield per plot (YLD), canopy temperature at vegetative (CT1) and reproductive (CT2) stage, normalized difference vegetation index (NDVI) at vegetative stage (lateboot stage (Z49), NDVI1), grain-filling stage (early milk stage (Z73), NDVI2) and grain maturity stage (late milk stage (Z83–87), NDVI3) according to Zadoks scale (Zadoks, 1974). NDVI at different growth stages was recorded using GreenSeeker® (Trimble, Inc.). Hand-held infrared thermometer (Kane May Model Infratrace 8000, USA) was used for the measurement of CT. Data on DH, DM, CT and NDVI were recorded on a plot basis; whereas PH, LA were recorded on randomly chosen five plants per plot. At maturity, plants were harvested from the experimental plots individually to record the grain yield.

The experiment on a subset population was conducted during 2018–19 in TS and LS environments in alpha-lattice design as described above. Data on various agro-physiological traits viz. DH, grain weight per spike (GWPS), TGW, grain length (GL), grain width (GW), biomass (Bio), yield per plot (YLD), CT, NDVI were recorded in replications in both environments. DH, CT, NDVI, Bio, and YLD were recorded on a plot basis while GWPS and TGW were recorded on randomly collected 20 spikes from each plot. GL and GW were measured for a random sample of grains of each genotype. Grain characteristics were measured using a grain image that was processed using software GrainScan developed by CSIRO (www.plant-image-analysis/software/grainscan).

Phenotypic Data Analysis

Phenotypic data recorded on each genotype of the subset population are subjected to descriptive statistical analysis. The analysis included block, replication, and treatment for each variable. Best Linear unbiased estimates (BLUEs) for the phenotypic data were calculated using R 3.6.0. Analysis of variance on mean values of 2 replications per genotype were analyzed in alpha-lattice design using SAS 9.3. The BLUE values of different traits were further used to perform the correlation analysis.

Genotypic Data Analysis

The model-based (Bayesian) Structure version 2.3.4 was applied to identify clusters of genetically similar individuals on the basis of their genotypic data. The program was run five times independently for K value (number of subpopulations) ranging from 1 to 10, adopting the admixture model. The normal logarithm of the probability was calculated against each K value, and the optimal number of subpopulations was determined using the ΔK approach described by Evanno et al. (2005). The threshold for statistical significance was determined by running 10,000 permutations.

To study the candidate gene-based genetic diversity on subset genotypes, the binary data produced from the scoring of bands on randomly chosen nineteen polymorphic SSR markers from above (37 polymorphic markers) were used as input for further analysis. Genotypic data obtained from 85 germplasm lines were used as input in DARwin 6.0 program (https://darwin.cirad.fr/darwin), and a dendrogram was constructed using the unweighted neighborhood joining algorithm.

Marker–Trait Associations

The association of genotypic and phenotypic data was conducted using TASSEL 2.1 software (https://www.maizegenetics.net), via a general linear model (GLM) with 1000 permutations. The significance threshold of the association was determined by the p value (< 0.05). In GLM model, population structure of the germplasm was included as fixed effects, while the association was estimated by simultaneous accounting of the population structure (Q matrix).

qRT-PCR Analysis

Ten-day-old seedlings of three moderate to highly tolerant (Halna, Raj3765, and WH1021) and one susceptible (HD2009) varieties were used for the purpose of qRT-PCR validation of 28 gene specific SSRs. The candidate gene SSRs included: Four SSRs (SSR30, SSR32, SSR35, and SSR36) located in genes encoding heat shock proteins (HSPs), two SSRs (SSR100, SSR155) from gene involved in signaling, nine SSRs (SSR12, SSR13, SSR60, SSR64, SSR79, SSR166, SSR177, SSR179, SSR205) from genes encoding transcription factors (TFs), two SSR (SSR158, SSR223) belonging to gene encoding regulatory proteins and 11 SSR belonging to genes with miscellaneous function.

Total RNA was extracted using TRI reagent (Sigma-Aldrich, St. Louis, MO, USA) from the control (22 ± 3 °C) and heat shock-treated (42 °C for 4 h) leaf tissues of the four varieties at 10-day-old seedling stage. The first strand cDNA synthesis was performed using Verso cDNA synthesis kit (Thermo Scientific Inc., USA) following the manufacturer’s instruction. Expression of the 28 genes was analyzed using real-time PCR with SYBR Premix (ABI, USA). PCR primers were designed using the NCBI Primer design tool (https://www.ncbi.nlm.nih.gov-tools-primer-blast) (Table S1). 18S RNA was used as a reference gene for normalization of expression. The relative expression was determined using the comparative Ct method (Livak and Schmittgen 2001). The relative fold change expressions were calculated for normalized heat shock-treated versus control samples. The change in expression between control and heat stress treatment was statistically analyzed using Student’s t test at a 5% level of significance.

Results

Screening on International Core Set

A total of 243 genic SSRs were screened on an international core set. Out of 243 SSRs, 93SSRs amplified specific bands; 58 markers produced monomorphic bands and 37 markers were polymorphic. The remaining 148 SSRs did not amplify well and produced nil/fuzzy/multiple bands and were not used for further analysis. The 37 polymorphic SSR markers produced a total of 106 alleles ranging from 2 to 5 alleles per SSR (Table 1). These SSR markers belonged to candidate heat-responsive genes encoding transcription factors (11), heat shock proteins (4), regulatory proteins (7), signaling (3) and others (12) (Table 1).

Table 1 List of heat-responsive SSRs, their gene description, repeat motifs and number of alleles amplified in wheat
Full size table

To understand the association of 37 markers with phenotypic traits, the ancestral contribution of genotypes in the population was estimated; for this purpose data on 145 lines with 37 SSR markers were utilized. The model-based analysis with Structure identified an optimal number of subpopulations at K = 3 (Fig S1) as the maximum likelihood when K was set from K = 1 to K = 10 subpopulations. The number of 145 wheat accessions assigned to each of the three inferred clusters is 58, 36 and 51 when the membership proportion was set at more than 0.5 for each cluster. Fixation index (FST) values between all groups were significant (p < 0.001) suggesting a real difference among these clusters. The inferred cluster values were subsequently utilized for understanding marker-trait associations. The phenotypic data on 8 agronomic traits from different growing seasons (2013, 2014, and 2016) were used for association with SSRs (Table S2). A total of 22 SSR markers were identified to be associated with the 7 traits at the 0.05 probability level (Table 2), and phenotypic variation ranged from 5.11 to 18.01%. SSR markers viz., SSR30, SSR32, SSR35, and SSR36 (gene encoding heat shock proteins) were associated with the physiological traits, NDVI and CT. SSR64 (gene encoding ethylene-responsive TF) was associated with various traits including NDVI, CT and YLD. SSR141 was associated with LA and NDVI in the years 2013 and 2014, and YLD in the year 2016.

Table 2 Marker–trait associations (R2) of the 37 candidate gene SSR markers with various agro-physiological traits in international core set during three (2013, 2014 and 2016) different crop seasons under late sown conditions
Full size table

Phenotypic and Genomic Analysis on Subset Population

The 37 polymorphic SSR markers identified as above were used for validation on a subset population of 85 lines. The summary statistics of phenotypic data recorded under TS and LS environments on this subset is presented in Table 3. Analysis of the variance indicated significant variations for the majority of the traits (Table 3). The germplasm lines showed difference in adaptive traits under the two environmental conditions. The coefficient of variation (CV) ranged from 4.10 to 18.52 for TS, and 5.71 to 22.4 for LS environments. Heritability ranged from as low as 0.29 for CT to 0.70 for NDVI in LS environments. Significant genotypic differences were observed for all the traits. The average reduction due to HS for agronomically important traits in LS environments can be seen in Fig. 1.

Table 3 Analysis of variance and basic statistics for phenotypic traits under timely (TS) and late sown (LS) conditions in wheat
Full size table
Fig. 1
figure 1

Box-plots representing best linear unbiased estimates (BLUEs) for mean data on various traits in timely sown (control) and late sown (heat stress) environments. a Days to Heading. b Normalized difference vegetation index. c Grain yield. d Grain weight per spike. e Biomass. f Canopy temperature. g Grain length. h Grain width. i Thousand grain weight. In the timely sown environment, sowing was carried out in mid-November and in the late sown environment, sowing was carried out in first week of January

Full size image

To investigate the level of association of grain yield with various phenotypic traits, the correlation between physiological and agronomic traits with YLD was analyzed under normal and HS conditions (Table 4). There was a positive correlation of YLD with GL, GW, TGW and GWPS under LS environments. NDVI at different intervals showed significant positive correlation with GL, GW, and GWPS in TS and LS condition. A significant negative correlation of DH was observed with GL, GW, GWPS, and TGW under LS. On the other hand, during TS environments CT had a negative correlation with GW and, TGW.

Table 4 Correlation coefficients among various phenotypic traits under timely (TS) and late sown (LS) conditions. The significance of correlation is depicted by *(p > 0.05) and **(p > 0.01)
Full size table

Diversity analysis Using Candidate Gene SSRs

The polymorphic SSR markers obtained as above in the core set were used for generating polymorphism profiles of the selected subset population. A total of 50 alleles were generated with an average of 2.6 alleles per locus. The allele diversity data was used to estimate dissimilarity which was subsequently used for cluster analysis. Four major clusters were produced; Cluster I and II were predominantly represented by genotypes of Indian origin with the exceptions of Frontana, Chiriya3, Giza163, Baviacora, synthetic lines, and UASD lines whereas Cluster III involved the majority of the exotic genotypes representing HS tolerant cultivars (Fig. 2). Cluster IV comprised of only six genotypes; two among these were exotic.

Fig. 2
figure 2

Neighbor joining tree representing genetic relationships of 85 genotypes using candidate gene-based SSR markers. Four clusters were produced that represent exotic genotypes in red color and indigenous genotypes in black color

Full size image

Marker–Trait Associations

The allele diversity data of the subset population were utilized in estimating population structure for inferring marker–trait associations subsequently. The k was assumed from 1 to 10 subpopulations using the co-ancestry model. The estimated linkage probability revealed that eighty-five germplasm lines were grouped into two clear subgroups (Fig. 3). The genotypes with more than 70% of the proportion of similar genome (based on shared alleles) were assigned to a common subgroup. One group (Red) represented mostly exotic lines from Australia and CIMMYT and the other group (Green) represented lines from Indian wheat breeding programs (Fig. 3). The lines G5, G12, G16, G31, G32, G34, G46, G6, G62, G65, G70, G74, and G77 showed less than 70% majority with any single group and, therefore, carried mixed proportion of alleles both from Indian and exotic origin.

Fig. 3
figure 3

Population structure of 85 germplasm lines using allele diversity data on polymorphic SSRs. The proportion of genome assigned to each subpopulation is depicted on Y-axis. The bars in red and green colour for each genotype depict the proportion of genome assigned to the two subpopulations. Subpopulation1 (red) comprised of majority of exotic genotypes while subpopulation II (green) comprised of majority of Indian genotypes. The numbers 1–85 represents the serial number of the genotype given in Online Resource1 in the same order

Full size image

A total of 30 marker–trait associations was identified with 14 different SSR markers in both the environments using days to heading as a covariate (Table 5). The average phenotypic variation (R2) value ranged from 5.2 to 25.1% in the population. Few candidate gene SSR markers were associated with more than a single trait; SSR13 (stress transcription factor) was closely associated with CT, GL, GW, GWPS; SSR76 (barley stem rust resistance protein) was associated with NDVI1, NDVI3 and GWPS; SSR230 (peroxisomal biogenesis factor 11) was associated with GL under TS, LS and YLD under LS. Candidate genes belonging to SSR72, SSR73, SSR100, SSR158, SSR166, SSR170 were also associated with more than one trait. Five markers viz. SSR12, SSR50, SSR60, SSR205, and SSR240 associated with one trait; Bio, CT, Bio, Yld and CT, respectively, in only one (either TS or LS) environment (Table 5).

Table 5 Marker–trait associations (R2) of the identified polymorphic SSR markers with the agro-physiological traits in subset population of 85 genotypes
Full size table

Expression of Genes Encoding Heat-Responsive SSRs

Among the SSR overlaying candidate genes, 28 candidate genes that associated significantly with important phenotypic traits (Tables 2 and 5) were selected for analyzing expression. Two genes encoding SSR141 and SSR205 did not produce single amplification products in qRT-PCR and were not included in further analysis. The details of qRT-PCR expression analysis are as follows: (i) The two signaling genes harboring SSR100 and SSR155 were significantly upregulated in thermotolerant lines compared to the thermosensitive cultivar HD2009 (Fig. 4) (ii) Heat shock protein (HSP) genes harboring SSR30, SSR32, SSR35, SSR36, had higher expression in the tolerant lines, WH1021 and Raj3765 (iii) TFs carrying SSR12, SSR60, SSR64, SSR166, and SSR179 and genes overlaying SSR73, SSR122, SSR170 and SSR240 had higher expression in the thermotolerant lines (iv) Genes overlaying SSR13, SSR50, SSR76, SSR79, SSR158, SSR177, and SSR221 had low expression under HS in all the four lines (v) the remaining four genes (SSR59, SSR72, SSR223, SSR230) were neither over-induced in tolerant cultivars nor highly reduced in susceptible cultivar under HS. Overall, the genes displayed lower expression in the susceptible line, HD2009, and medium tolerant line, Halna, with an exception of genes carrying SSR12, SSR32, SSR36, SSR64, and SSR179. In Halna, these five genes displayed higher expression under HS. Despite our repeat efforts, no amplification was observed for Raj3765 and WH1021 for the candidate genes carrying SSR179 and SSR76, respectively.

Fig. 4
figure 4

Relative expression of the SSR harboring genes belonging to signaling (a), Regulatory proteins (b), HSPs (c), TFs (d), miscellaneous functions (e) in four wheat genotypes: HD2009 (thermosusceptible), Halna (medium tolerant), WH1021 (thermotolerant) and, Raj3765 (thermotolerant). For normalizing the data, the expression of the 18SRNA gene was used; 10-day-old untreated control (C) samples at 22 °C and treated samples (T) with heat shock at 42 °C for four hours were used for estimation of fold change in each genotype; vertical bars indicate SE. Comparison of means between control and treatment samples was carried out by Student’s t test and significant differences (p < 0.05) are represented by asterisk for each genotype

Full size image

Discussion

Heat stress has a huge consequence in limiting the total grain yield of wheat (Hatfield and Prueger 2015). High temperature during grain-filling stages induces a number of cellular and metabolic changes leading to reduced production of normal proteins. Plants employ various stress adaptive mechanisms to cope up with elevated temperatures. It has been widely accepted that morpho-physiological and yield contributing traits such as DH, DM, NDVI, chlorophyll content, chlorophyll fluorescence, CT, spikelets/spike, grain number/spike, spike length, biomass, tillers/plant and, harvest index had high correlation and heritability under HS and, thus can be effectively used in the breeding program as selection criteria for improvement of stress tolerance and for selection of best genotypes (Araus et al. 2008; Reynolds et al. 2001; Gupta et al. 2017; Pinto et al. 2010; Jain et al. 2018). We observed a high correlation between the majority of the studied morpho-physiological and component traits of yield under HS. Plants tend to have early anthesis and early maturity to avoid the effect of HS (Mondal et al. 2016). In the present study, genotypes showed early heading and a negative association of DH to yield and contributing traits under LS. Grain morphology is also found to be an important parameter under stress conditions. GW had a positive correlation with GL and a negative correlation with DH (Table 3). Kushwaha et al (2011) suggested that heat stress during the terminal stage of the crop growth inhibits the starch biosynthesis which leads to reduction in normal grain size.

Availability of superior and diverse alleles/genes is the starting point of genetic enhancement of crop plants including wheat, for the development of improved cultivars (Abouzied et al. 2013). In the present study, a large proportion of the newly developed candidate gene SSRs had monomorphic alleles (58), implicating that the genes are either conserved or involved in housekeeping among both the susceptible and tolerant genotypes. For the remaining polymorphic SSRs, the subpopulation structure grouped individuals based on their shared alleles into two subpopulations, differentiating genotypes of exotic origin from the Indian origin. This suggests the use of diverse germplasm in bringing out tolerant genotypes at International breeding programs and that of the Indian breeding programs. Lines from Australia and CIMMYT formed a distinct subpopulation representing diverse alleles imparting tolerance; however, exotic lines that have been used so far in Indian crossing program shared common alleles and were grouped with the Indian subpopulation. For instance, Kauz is a heat-tolerant genotype from CIMMYT and the lines carrying Kauz in their pedigree such as DBW173, DBW88, MACS6273, Baj and Kachu grouped together in cluster III suggesting the presence of common alleles. The neighbor joining cluster-based findings were in total agreement with the population structure analysis which suggested two subpopulations (exotic and Indian) based on > 70 percent of shared ancestry among individuals. Clusters I and II majorly comprised of heat-tolerant Indian genotypes including HD2932, WH1021, Raj3765, Halna, DBW71, DBW90 and PBW226. These Indian genotypes can themselves serve as a source as parents and could be used in breeding programs without much effort for adaptation. In cluster III, Indian and exotic genotypes did not form separate clusters, which reflect upon the similar expression of stress-responsive genes under HS.

Abiotic stress is a complex process that involves several factors like secondary metabolites, hormones, transcription factors and signaling systems (Lamaoui et al. 2018). Conventional breeding has had limited success in improvement for heat tolerance. Breeding for heat stress is a tough process as the component traits are quantitative in nature, hence the use of MAS and QTL mapping approaches may prove helpful (Collins et al. 2008). Information on QTL hotspots with significant marker-trait associations is being generated in several crops including wheat (Sukumaran et al. 2018; Tadesse et al. 2018; Sinha et al. 2018; Acuna-Galindo et al. 2015). The probability of finding significant marker–trait associations is further enhanced by the availability of genic SSRs. The markers identified from heat stress transcriptome revealed high phenotypic variance in this study either for the same or other correlated traits under stress. As many as 28 SSR loci revealed significant associations with various phenotypic traits. The marker-trait associations were further supplemented with information on heat-responsive genes that are induced under stress. A synergistic response of genes belonging to TFs, HSPs, signaling molecules is observed suggesting specific genomic locations for adaptation and acclimation.

Although transcriptomes have majorly been used for expression analysis, a number of SSRs derived from transcriptome sequencing had been extensively used in plant genetic diversity analyses such as in pigeon pea (Dutta et al. 2011) and chickpea (Kant et al. 2017). SSRs derived from transcriptional approaches are highly suitable for assessing functional diversity. In this study, the polymorphic genic SSRs from diverse TFs, HSPs, signaling and regulatory molecules were associated with important phenotypic traits. Such functional SSRs may regulate gene expression and function under HS. The expression of as many as 15 genes (SSR12, SSR30, SSR32, SSR35, SSR36, SSR60, SSR64, SSR73, SSR100, SSR122, SSR155, SSR166, SSR170, SSR179, and SSR240) were significantly induced under HS in either of the two thermotolerant lines compared to control (Fig. 4). Expression of only five genes (SSR12, SSR32, SSR36, SSR64, and SSR179) was elevated in Halna under high temperature. In fact, Halna is a medium tolerant line that follows the mechanism of heat avoidance by early completion of its life cycle and hence poor expression of candidate genes. The expression of the majority of the genes was reduced in the HD2009 suggesting under-expression of TFs, HSPs and regulatory molecules that might have led to lower expression of associated traits upon high-temperature stress. Heat stress-responsive TFs and proteins are known to be induced in wheat in several previous reports (Zhang et al. 2017; Xue et al. 2015; Kumar et al. 2017). Heat shock factor C2a was involved in heat protection in developing grains in wheat (Hu et al. 2018). Not only an individual gene is expressed but also a cluster of genes is expressed in response to stress that signifies the presence of several loci in a bonafide QTL. In the tolerant lines WH1021 and Raj3765, genes for HSPs, DREB transcription factor, ethylene-responsive TF, ring finger proteins and sugar metabolism genes were elevated coherently for adaptation to stress. Either such genes or the associated traits may be selected for precise improvement under heat stress.

Candidate genes harboring the newly developed SSR markers were linked to various phenotypic traits in this study. This knowledge of candidate genes governing phenotypic traits will help in a better understanding of the underlying mechanism for HS tolerance which is important to address the issue of heat stress through the use of molecular technology and MAS. The information generated on the traits governed by stress-associated genes is useful for the plant breeders who might be indirectly selecting for the superior traits/genes. The candidate gene-based SSRs have potential use in the transfer of traits for improvement of future breeding programs on HS tolerance.