Meta-QTLs, ortho-MQTLs and candidate genes for the traits contributing to salinity stress tolerance in common wheat (Triticum aestivum L.)

Abstract

A meta-analysis of QTLs associated with the traits contributing to salinity tolerance was undertaken in wheat to detect consensus and robust meta-QTLs (MQTLs) using 844 known QTLs retrieved from 26 earlier studies. A consensus map with a total length of 4621.56 cM including 7710 markers was constructed using 21 individual linkage maps and three previously published integrated genetic maps. Out of 844 QTLs, 571 QTLs were projected on the consensus map which gave origin to 100 MQTLs. Interestingly, 49 MQTLs were co-located with marker-trait associations reported in wheat genome-wide association studies for the traits contributing to salinity stress tolerance. Five potential MQTLs associated with the major salinity-responsive traits were also identified to be utilized in the breeding programme. In the resulted MQTLs, the average confidence interval (CI, 3.58 cM) was reduced up to 4.16 folds compared to the mean CI of the initial QTLs. Furthermore, as many as 617 gene models including 81 most likely candidate genes (CGs) were identified in the high confidence MQTL regions. These most likely CGs encoded proteins mainly belonging to the following families: B-box-type zinc finger, cytochrome P450 protein, pentatricopeptide repeat, phospholipid/glycerol acyltransferase, F-box protein, small auxin-up RNA, UDP-glucosyltransferase, glutathione S-transferase protein, etc. In addition, ortho-MQTL analysis based on synteny among wheat, rice and barley was also performed which permitted the identification of six ortho-MQTLs among these three cereals. This meta-analysis defines a genome-wide landscape on the most stable and consistent loci associated with reliable molecular markers and candidate genes for salinity tolerance in wheat.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12298-021-01112-0.

Introduction

Salinity is a major constraint to crop productivity worldwide and represents a big challenge for food security (Tuteja et al. 2012), causing yield losses up to 60% particularly in wheat (Triticum aestivum L.) (Hasanuzzaman et al. 2017). Around 25% of the global land area (or more than > 1 billion hectares) is affected by salinity, which is continuously expanding by up to 10 million hectares per year owing to inappropriate irrigation practices or natural salinization (Shahid et al. 2018). On the other hand, it is also expected that population growth will account for at least a 50% increase in the demand for wheat by 2050 (Shiferaw et al. 2011).

In wheat, salt stress adversely affects different agronomic traits (Azadi et al. 2015; Hasanuzzaman et al. 2017; Nezhad et al. 2019); several physiological traits involving nutrient imbalance, interruption of cell membranes, diminished ability to detoxify relative oxygen species, reduced photosynthetic activity, differences in the antioxidant enzymes and reduction in stomatal aperture (Hasan et al. 2015; Genc et al. 2014; Ilyas et al. 2020), eventually leading to plant death which severely constrains crop productivity and total crop production. Therefore, to meet the ever-increasing demands of wheat consumption, there is an urgent requirement to develop salinity-tolerant wheat cultivars. Since salinity tolerance is a highly complex quantitative trait including several plant-specific morphological, physiological and metabolic processes controlling mechanisms to combat salinity stress, little success has been achieved so far (Gupta and Huang 2014).

Quantitative trait loci (QTL) analysis has been proved a very powerful tool for providing information of genomic regions controlling complex quantitative traits (Holland 2007; Saini et al. 2020). Identification of molecular markers associated with the QTLs provides plant breeders with a way to improve selecting favourable recombinants from superior varieties and accelerating breeding programs, a strategy, widely known as marker-assisted breeding. Same to other complex traits, hundreds of QTLs have been identified for the traits contributing to salt tolerance in wheat as well (Online Resource 1). To make these previously identified QTLs more useful to wheat breeding and for the basic research which is generally aiming to get better insights into underlying genetics of traits conferring salinity tolerance, a thorough analysis of these QTLs need to be carried out. Because most mapping experiments vary in several aspects, such as type and size of mapping population, type and number of molecular markers employed and designed, and even the basic quality of the experiment; an efficient method is needed for this purpose. Regarding this, QTL meta-analysis has been evidenced as an efficient and reliable approach (Goffinet and Gerber 2000).

Meta-QTL analysis is made achievable through a software, BioMercator V4.2, where a specific set of refined algorithms have been furnished for reliable evaluation and re-estimation of the genetic positions of given QTLs (Arcade et al. 2004; Veyrieras et al. 2007; Sosnowski et al. 2012; Oliveira et al. 2014). Using the maximum likelihood method, these algorithms test the number of possible MQTLs resulting from the projection of initial QTLs (taken from different studies) in a linkage group and employ statistical criteria to find out the best model to be employed for deciding the real number of these MQTLs (Goffinet and Gerber 2000; Sosnowski et al. 2012).

Generally, QTL meta-analysis aims to confirm whether QTLs identified in different studies correspond to different loci or whether they represent a common place/position on the linkage map of the concerned species and vice versa i.e. if a QTL previously considered as identified in a common region in different genetic backgrounds exhibits different positions. Ultimately, the analysis aims to establish the occurrence of “QTL hotspots” or “MQTLs” in a consensus genetic map, which corresponds to the more precise region where different loci are mapped together (Goffinet and Gerber 2000). These refined genomic regions are supposed to harbour important candidate genes related to the traits (Saini et al. 2021c). Several meta-QTL analyses have already been conducted for the traits of different biotic and abiotic stresses in wheat (Online Resource 2). To the best of our knowledge, to date, no meta-QTL analysis has been conducted for salinity stress tolerance in wheat. Therefore, the present study is believed to be the first of its kind to report the most robust and consistent MQTLs for salinity tolerance in wheat.

With the advancement of DNA sequencing technology, high-throughput genotyping based on SNP arrays or next-generation sequencing has also made genome-wide association studies (GWAS) of complex quantitative traits possible. This association analysis method has also been used for unraveling the genomic regions associated with salinity responsive traits in wheat (Alotaibi et al. 2021; Chaurasia et al. 2021; Hu et al. 2021). In addition, several important genomic regions/marker-trait associations (MTAs) detected by GWAS for different traits have been further validated by linkage mapping studies (Cao et al. 2020; Zhang et al. 2018). All of this suggests that combining meta-QTL analysis and GWAS can effectively integrate the original QTL results from various studies in order to mine the key genomic regions and candidate genes that affect important quantitative traits.

Overall objectives of this study were (i) to construct a new highly dense consensus linkage map; (ii) to identify the most robust and stable MQTLs with small supporting intervals compared to the initial QTLs; (iii) to validate the MQTLs with the results obtained from GWA studies; (iv) to refine the initial QTLs for predicting the promising candidate genes; (v) to provide reliable markers of the MQTLs to be used in marker-assisted breeding. Furthermore, to assess the transferability of genetic information to other cereals, ortho-MQTLs were also investigated based on genomic synteny and collinearity among the wheat, rice, and barley (Sorrells et al. 2003; Mayer et al. 2011; Ahn et al. 1993; Kumar et al. 2009). This work, we believe, will be useful to improve selection efficiency for yield stability, potential, and performance under salinity stress conditions in wheat and other cereals.

Materials and methods

A detailed bibliographic investigation was performed on PubMed (http://www.ncbi.nlm.nih.gov/pubmed) and Google Scholar (https://scholar.google.com/) using suitable keywords for wheat QTLs related to the traits contributing to salinity stress published from 1996 to 2020. Subsequently, a total of 844 QTLs for different morphological, agronomical, and physiological traits measured under salinity stress conditions from 21 different bi-parental wheat populations (also included one durum wheat population, Dubcovsky et al. 1996) were retrieved from 26 studies (Online Resource 1, 3). From each study, the following information was retrieved: (i) trait name, (ii) chromosome, (iii) flanking markers or closely linked marker of QTLs, (iv) peak position and confidence interval of individual QTLs, (v) type and size of the mapping population, (v) LOD score, and (vi) phenotypic variation explained (PVE) or R² value of QTLs. All the QTLs were given unique identities for the purpose of analysis.

The size of the mapping populations ranged from 87 to 350 lines of different types including 13 sets of recombinant inbred lines (RILs), 5 doubled haploid (DH) populations, 2 F₂ populations, and one set of 1 recombinant chromosome substitution lines (RCSLs) evaluated at different locations, each for more than one year (Online Resource 1). Whenever the information about peak position was not available from a certain study, the most likely position of loci was estimated as the central point between the two flanking markers (Kumar et al. 2020). While in some cases where LOD score was not given, test statistic given in the respective source paper was used to calculate it using the following formula: LOD score = likelihood ratio statistics/4.6, and where it was merely mentioned that a threshold LOD score of 3 was considered to map the QTLs, the LOD 3 was assumed for all the QTLs identified in the concerned study. All QTLs were accommodated to the following six trait categories: (i) agronomic traits (ATs), (ii) mineral concentration in different tissues such as root, shoot, and leaves (MCs), (iii) shoot related traits (SRTs), (iv) root related traits (RRTs), (v) leaf related traits (LRTs), and (vi) plant performance metrics related traits (PMRTs) (Online Resource 4).

Construction of consensus map

For the construction of consensus map, the following three integrated genetic maps were utilized as the reference maps: (i) the ‘Wheat_Composite_2004’ map with 4403 marker loci, available at GrainGenes database (http://wheat.pw.usda.gov); (ii) the ‘Wheat, Consensus SSR, 2004’ with 1235 marker loci (Somer et al. 2004); (iii) an integrated map for durum wheat with 3669 markers (Marone et al. 2013). Twenty-one linkage maps generated using different mapping populations (Online Resource 1) were also included in this consensus map to further increase the marker densities on the chromosomes. The R package LPMerge was employed for this purpose (Endelman and Plomion 2014). Following Hubert and Hedgecock (2004), the genetic length of the consensus genetic map was calculated. It involved two steps: (i) the average marker spacing (χ) was first calculated using the formula given below:

$$\chi \left(\text{cM}\right)=\frac{\text{total genetic length (cM)}}{\text{number of total markers - number of linakge groups}}$$

and then, (ii) the genetic length of the consensus genetic map was determined by adding 2χ to the length of each linkage group to account for terminal chromosomal regions (Fishman et al. 2001; Shariatipour et al. 2021).

Projection of QTLs and meta-QTL analysis

The QTLs with the complete required information (CIs, peak position, original LOD score, and R² or PVE value) were projected onto the consensus map. Whereas, QTLs with the incomplete required information (with no PVE or unknown genetic position, etc.), which were initially considered during the QTL data collection, were not used for QTL projection. As reported in earlier MQTL studies (Venske et al. 2019; Khahani et al. 2020, 2021), to ensure consistency among the different QTL mapping studies, the CI (95%) was estimated for each locus, through different population-specific equations:

$$\text{CI}\left(95\%\right)=\frac{287}{\text{PVE x N}}$$ 1

$$\text{CI}\left(95\%\right)=\frac{163}{\text{PVE x N}}$$ 2

$$\text{CI}\left(95\%\right)=\frac{530}{\text{PVE x N}}$$ 3

where, N is the population size, and PVE is the phenotypic variation explained (or R²) by QTLs. Equation (1) was utilized for the QTLs identified using DH populations (Visscher and Goddard 2004) and Eq. (2) was utilized for the QTLs identified using RIL populations (Guo et al. 2006). Equation (3) was applied to calculate CI (95%) of the QTLs identified using F₂ populations (Darvasi and Soller 1997). All projections of QTLs on the consensus map were carried out using the software BioMercator V4.2 based on common markers between initial and consensus maps (Sosnowski et al. 2012; Oliveira et al. 2014).

The meta-analysis was performed using the same software BioMercator V4.2, for each chromosome or linkage group, separately, using the two different methods. When there were more than 10 projected QTLs available for a certain chromosome, Veyrieras two-step algorithm (Veyrieras et al. 2007) was employed for the meta-analysis. The best model of MQTLs was selected based on the prevalent value among Akaike Information Criterion, corrected Akaike Information Criterion, Bayesian Information Criterion, and Average Weight of Evidence criteria. When total QTLs projected on a certain chromosome were 10 or less, meta-QTL analysis was performed using the method proposed by Goffinet and Gerber (2000). Subsequently, the consensus QTLs, ascertained from the selected best model, was considered as an MQTL or “real” QTL. The statistical procedures and algorithms implemented in this software have been well-described elsewhere (Arcade et al. 2004; Veyrieras et al. 2007; Sosnowski et al. 2012).

Comparison of MQTLs with known MTAs

Genome-wide association studies (GWAS) available for the traits associated with salinity tolerance were reviewed (Alotaibi et al. 2021; Alsamman et al. 2020; Chaurasia et al. 2020, 2021; Hu et al. 2021; Liu et al. 2018; Oyiga et al. 2018; Quan et al. 2021; Yu et al. 2020). The physical positions of markers significantly associated with the salinity-responsive traits identified in these GWA studies were either obtained from the source papers or different databases (such as CerealsDB, JBrowse-WHEAT URGI) and compared with the physical coordinates of the MQTL regions on the same chromosome; an individual MTA occurring within a specific MQTL region was considered co-located. The details of population size, traits, markers used, and marker-trait associations (MTAs) identified in these studies are summarized in Online Resource 5. These GWA studies used phenotypic data recorded from different sets of winter and spring wheat genotypes (with panel sizes ranging from 70 to 370) grown in four different countries.

Ortho-MQTL analysis

Considering the high synteny among the cereals; the most stable and robust wheat MQTLs were investigated for mining of ortho-MQTLs for salinity tolerance among the wheat, barley, and rice. For this purpose, two studies from Zhang et al. (2017) on barley, and Islam et al. (2019) on rice reporting MQTLs associated with salinity tolerance were used. The following steps were followed for mining of ortho-MQTLs: (a) gene models detected in selected most stable and robust wheat MQTLs were BLASTed against rice and barley genome databases available at EnsemblPlants, (b) the rice and barley genes corresponding to wheat genes were extracted with their genomic positions from the database, (c) genomic positions of these rice and barley genes were compared with physical coordinates of rice MQTLs (Islam et al. 2019) and barley MQTLs (Zhang et al. 2017), respectively, (d) corresponding MQTLs harbouring the investigated genes were designated as ortho-MQTLs.

Candidate gene identification within MQTLs

The nucleotide sequences of flanking markers of the most promising MQTLs were retrieved using the following databases-GrainGenes database (https://wheat.pw.usda.gov/GG3) and CerealsDB (https://www.cerealsdb.uk.net/cerealgenomics/CerealsDB/indexNEW.php). Then, using these nucleotide sequences, BLASTN searches (with the criteria, maximum E-value = 1E-100, minimum 95% identity of the sequence) were performed against the wheat reference genomic sequence available in the EnsemblPlants database (http://plants.ensembl.org/index.html) to get the physical positions of the concerned markers. Physical positions of GBS-SNPs markers were directly obtained from the URGI database (https://wheat-urgi.versailles.inra.fr/). Among the selected MQTLs, the ones which had ≤ 2 Mb physical interval distance were directly explored for the identification of the gene models. In the remaining cases, where the physical interval was more than 2 Mb; first, the physical position of the MQTL peak was calculated using the formula given below and then the total 2 Mb physical region (1 Mb from each left and right side of the MQTL peak) was considered for mining of gene models.

$$\text{Peak position}\left(\text{bp}\right)=\text{start position}\left(\text{bp}\right)+\frac{\left(\text{end position}\left(\text{bp}\right)-\text{start position}\left(\text{bp}\right)\right)}{\left(\text{end position}\left(\text{cM}\right)-\text{start position}\left(\text{cM}\right)\right)}\times \frac{\text{CI}\left(\text{95},\%\right)}{2}$$

Either original or estimated physical position range for each selected MQTL was entered into the search field in the ‘BioMart’ tool of EnsemblPlant database for obtaining details of the gene models (locus ID information and functional descriptions) present in the corresponding MQTL regions.

Results

Main features of QTLs associated with salinity stress tolerance

The number of QTLs per trait ranged from 18 (RRTs) to 472 (ATs) with an average of around 93 QTLs per trait (Fig. 1a; Online Resource 6). The available QTLs were unevenly distributed on the 21 chromosomes, ranging from a minimum of 21 QTLs on 3D to a maximum of 75 on 3B with an average of approximately 40 QTLs per chromosome. Of the total QTLs, 34.5% were identified in sub-genome A, 38.03% in sub-genome B, and 27.36% in sub-genome D (Fig. 1b). About 67.65% (as many as 571 QTLs) of the collected QTLs had complete information required for meta-analysis (Fig. 1a). The CI of a single QTL ranged from zero to 67.4 with an average of 14.92 cM (Fig. 1c). The phenotypic variation explained by a single QTL ranged from 0.01 to 64.69% (with an average of 8.6%), with most of the QTLs exhibiting PVE < 10% (Fig. 1d).

Fig. 1.

a Trait-wise distribution of QTLs, b chromosome-wise distribution of the QTLs, c CI of the QTLs (used for projection), d PVE of the QTLs (used for projection). ATs: agronomic traits, MCs: mineral concentration in different tissues such as root, shoot and leaves, SRTs: shoot related traits, RRTs: root related traits, LRTs: leaf related traits, PMRTs: plant performance metrics related traits

Features of newly generated consensus map

The newly generated consensus map contained 7,710 markers, which includes a variety of markers: SNPs, DArT, SSR, AFLP, RAPD, STS, EST-SSR, ISSR, and KASP markers along with a few genes e.g. Vrn, Ppd, Rht, and Glu loci (Online Resource 7). The total length of the map is 4621.56 cM, with a length ranging from 139.29 cM (6D) to 337.09 cM (4B) across the 21 linkage groups (Fig. 2, Online Resource 8). The average number of markers carried by a single chromosome was approximately 367 (Fig. 2). The marker densities (markers per cM) for individual linkage groups ranged from 1.32 to 2.84 for sub-genome A, from 0.92 to 3.24 for sub-genome B, and from 0.55 to 1.59 markers for sub-genome D (Online Resource 8).

Fig. 2.

Distribution of the markers on the consensus genetic map used for meta-QTL analysis including genetic length of the chromosomes, and the number of markers in each chromosome

Salinity MQTLs and their distribution on the wheat genome

Out of total 844 QTLs, only 571 QTLs were successfully projected onto the consensus map which gave origin to 100 MQTLs, derived only from 470 QTLs. Eighty-eight QTLs were not projected because of either of the following reasons: (i) they lacked common flanking markers between initial and consensus maps, and (ii) the QTLs showed comparatively low R² values, causing a large CI, while, thirteen QTLs were singletons (single QTLs), not overlapping with any MQTL. Of the 100 MQTLs, 38 belonged to sub-genome A, 36 to sub-genome B, and 26 to sub-genome D (Online Resource 9). The MQTLs were named sequentially for each chromosome, separately (e.g. MQTL1A.1 to MQTL1A.4). The number of MQTLs ranged from 2 on several chromosomes to 8 each on chromosomes 2A and 3B. Almost all the MQTLs were observed as associated with at least 2 different salinity-responsive traits owing to the multi-trait and multi-gene effects on salinity tolerance. Among the 100 MQTLs, 74 MQTLs included QTLs of ATs. Likewise, 51 and 24 MQTLs included QTLs of MCs and SRTs, respectively. Thirty-one MQTLs included QTLs of ATs and MCs, 17 MQTLs contained QTLs of ATs and SRTs, the same number of MQTLs (17) contained QTLs of ATs and PMRTs, 9 MQTLs included QTLs of ATs and LRTs, and again 9 MQTLs contained QTLs of ATs and RRTs. Twenty-nine MQTLs contained QTLs of at least 3 major traits. In addition, 23, 14, and 10 MQTLs contained QTLs of PMRTs, LRTs, RRTs, respectively. Details regarding the different traits governed by MQTLs are presented in Online Resource 9.

The MQTL with a larger number of initial QTLs is considered a more reliable and stable MQTL independent from the phenotyping environment and genetic background. MQTL2D.1 with 20 initial QTLs had the largest number of QTLs derived from different experiments followed by 2D.4, 6A.1, 4B.1, 3B.8, 2D.6, 4A.5, 6A.3 with 19, 18, 14, 13, 11, 10, and 10 QTLs, respectively. These QTLs appeared as the most robust and stable QTLs in different locations and years. The MQTLs explained a significant percentage of the phenotypic variation ranging from 0.21 to 23.15% for different traits, whereas LOD values of the MQTLs ranged from 1.86 to 12.42. Overall, in the present study, a maximum number of initial QTLs was projected on chromosome 2D, followed by 2A, 5B, and 4A (as shown in Fig. 3 and 4). The genetic length (CI) of these MQTLs ranged from 0.01 to 22.64 cM with an average of 3.58 cM, showing a significant reduction from the initial QTLs ranging from 0.0 to 67.4 cM with an average of 14.89 cM. Overall, the average CI of MQTLs was 4.16-fold less than that of initial QTLs, and there were substantial differences in average CIs of MQTLs among different chromosomes. For instance, the average CI of MQTLs on chromosomes 2D and 1D reduced by 11.31 and 8.76 times, respectively, followed by 7.53 and 7.17 on chromosomes 7B and 4B (Fig. 5). More importantly, 46 MQTLs showed ≤ 1 cM CI only (Online Resource 9).

Fig. 3.

MQTLs for salinity tolerance in wheat for chromosome a 2A and b 2D. The different colors in the projected QTLs (left) and inside the chromosomes only indicate the different MQTL generated and how each initial QTL contributed to the formation of it

Fig. 4.

MQTLs for salinity tolerance in wheat for chromosome c 4A and d 5B. The different colors in the projected QTLs (left) and inside the chromosomes only indicate the different MQTL generated and how each initial QTL contributed to the formation of it

Fig. 5.

The fold reduction in confidence intervals of QTLs after meta-QTL analysis

MQTLs co-located with MTAs reported in previous GWAS studies

The MQTLs were anchored to the physical map of the wheat reference genome; however, physical coordinates for only 5 MQTLs could not be deduced because these MQTLs were flanked by markers such as AFLP, which lacked the sequence data required for BLAST searches. The physical positions of MQTLs were used for comparison with MTAs for the traits associated with salinity tolerance reported in earlier GWAS conducted using spring and winter wheat. For this purpose, only 95 physically mapped MQTLs could be considered; forty-nine (49) of these MQTLs (Table 1) each overlapped one or more known MTAs from among 638 MTAs that were available for comparison (Online Resource 10). As many as 26 MQTLs were co-located with MTAs obtained from GWAS involving spring wheat and 44 MQTLs were co-located with MTAs derived from winter wheat populations. However, 21 MQTLs were verified using GWAS involving both spring and winter wheat. The number of co-localized MTAs for each MQTL also varied, so that as many as 11 MQTLs each co-located with at least 10 MTAs identified in 9 GWA studies; of these, MQTL5A.2 matched with 122 MTAs, followed by MQTL5A.3, 5B.5, 6B.3, and 4A.1 with 115, 90, 35, and 34 MTAs, respectively (Fig. 6). Some of the MQTLs (e.g., MQTL4A.1, 5A.2, 5A.3, and 5B.5) each involving 4 or more initial QTLs co-located with more than 30 MTAs (Online Resource 10).

Table 1.

A summary of GWAS-validated MQTLs identified in the present study

S. No	MQTL name (CI, in cM)	Flanking markers	Number of QTLs involved (avg. LOD score)	Traits (avg. PVE)a
1	MQTL1B.1 (33.8–34.4)	AX-108725289/AX-109875509	2 (3.25)	TGW, GW (3.02)
2	MQTL1B.2 (65.06–67.79)	AX-94525557/AX-111129114	4 (2.65)	SOD, SPW, TGW, SK/Na ratio (9.71)
3	MQTL1B.3 (78.91–86.2)	Xswes953/AX-109306278	3 (2.34)	Pro, SL, CC (11.41)
4	MQTL1B.4 (94.19–100)	Xgwm608/wPt-8744	3 (3.49)	SL, RL, AL (4.83)
5	MQTL1D.3 (56.89–62.64)	AX-110385713/AX-110369731	7 (4.01)	GY, GNPP, SPS, root K conc., GNPS, Ca conc., Cl conc. (23.15)
6	MQTL2A.3 (34.54–36.2)	Xbcd1184b/AX-111526140	2 (3.95)	GW, TGW (4.36)
7	MQTL2A.6 (82.52–84.83)	AX-94417710/AX-108777020	3 (2.73)	SVI, GI, G% (13.34)
8	MQTL2A.8 (116.82–116.93)	XksuH16/wPt-1615	5 (2.68)	Root Fe conc., shoot K conc., Root Na exclu., root Mn conc.. root Zn conc. (4.08)
9	MQTL2B.1 (25.33–26.92)	Xwmc382/wPt-0100	5 (3.01)	PH, Ca conc., GY, Na conc., K/Na ratio (5.52)
10	MQTL2B.2 (44.33–44.48)	Xcdo405/Xcfd11	7 (4.55)	SPS, GL/GW, shoot K conc., Cl conc., SL, GW (5.53)
11	MQTL2B.4 (93.83–98.17)	wPt-11586/AX-111181035	2 (2.75)	root B conc., Cl conc. (7.55)
12	MQTL2D.1 (72.06–73.2)	AX-110484736/Xbarc297	20 (5.11)	DTH/F/M, TN, HI, SK/Na ratio, RL, SL, PH, RDW, TGW, K/Na ratio(16.65)
13	MQTL2D.3 (84.35–86.47)	Xwmc111/AX-110638095	2 (4.45)	BY, SPS (5.24)
14	MQTL2D.5 (106.64–110.57)	Xcdo1379/AX-108854020	3 (3.23)	TGW, SFW, PH (12.9)
15	MQTL3A.2 (78.71–81.81)	Xbarc45/AX-94831084	4 (4.1)	Cl conc., Mg conc., Cl conc. (9.4)
16	MQTL3A.3 (85.62–87.63)	AX-94831084/AX-110928333	6 (3.03)	GR, Ca conc., TDW, RNa conc., SH, LFW (9.08)
17	MQTL3B.3 (82.26–97.32)	AX-108869068/AX-108752433	2 (2.55)	Cl conc., RNa conc. (3.26)
18	MQTL3D.1 (3.48–11.73)	AX-109311768/AX-110348822	2 (4.85)	Root B conc., root K conc. (0.21)
19	MQTL3D.2 (43.24–47.57)	AX-109324006/Xgdm136	3 (3.41)	GL, GL/GW (4.26)
20	MQTL4A.1 (18.49–19.09)	AX-111621169/Xbcd1738	4 (2.6)	SV%, SH, CC (9.73)
21	MQTL4A.2 (25.84–29.66)	Xfba43/Xksuf8a	3 (2.81)	SK/Na ratio, GR, Ca conc. (5.35)
22	MQTL4A.5 (65.82–69.28)	wPt-6515/AX-108914639	10 (3.28)	RWC, GNPS, SPS, TGW, SL, SNPS, GN, SW (11.03)
23	MQTL4B.1 (14.77–17.34)	AX-109993293/tPt-5519	14 (12.23)	Shoot K conc., SPS, root K/Na ratio, GW, GNPS, GNPP, HI, SCC, SH, GP, PH, TGW, GL/GW (12.68)
24	MQTL5A.1 (26.81–28.57)	AX-111625534/AX-110980253	4 (4.39)	CC, HI, RL, root Mg conc. (6.28)
25	MQTL5A.2 (42.29–45.45)	AX-110542873/wPt-1903	9 (7.06)	SL, SPS, shoot Na conc., Cl conc., RDW, SH, GNPP (13.19)
26	MQTL5A.3 (52.29–58.59)	AX-109331965/Xgwm205	5 (3.2)	TDW, SDW, SH, SL, GNPS (7.53)
27	MQTL5A.4 (75.76–76.46)	AX-111498021/Xcfd2	2 (2.83)	TN, root K conc. (5.55)
28	MQTL5A.5 (119.98–127.36)	Xksum024/Vrn1A	3 (3.26)	SK/Na ratio, shoot K conc., shoot Na conc. (7.87)
29	MQTL5B.1 (27.09–27.65)	AX-111024537/AX-111616360	3 (6.29)	GP, GL/GW, GL (6.35)
30	MQTL5B.2 (49.56–54.75)	AX-109321350/AX-111678599	6 (2.84)	SPP, GW, GL, GL/GW, GP (5.88)
31	MQTL5B.3 (61.82–63.49)	AX-111627941/AX-111619122	2 (4.78)	SPS, SFW (6.06)
32	MQTL5B.5 (118.3–126.19)	AX-110942912/AX-111462836	4 (4.28)	SL, CC, SPW, TGW (10.88)
33	MQTL5B.7 (151.96–152.45)	Xcfa2121.1/Xfbb121.1	3 (2.77)	total Chl. content, Mg conc., SPW (10.3)
34	MQTL6A.1 (40.88–43.56)	wPt-0689/wPt-4589	18 (10.78)	GL, SPS, GL/GW, GW, GP, PH, TKW, HI, root Zn conc., Root Na exclu., root Fe conc., shoot Ca conc., shoot K conc., GNPP, GNPS, SA (7.76)
35	MQTL6A.2 (47.04–53.65)	XksuH4/AX-111624010	2 (3.74)	SL, AL (3.3)
36	MQTL6B.1 (51.46–53.76)	AX-110375424/AX-108853850	6 (12.42)	Shoot Ca conc., root S conc., shoot Mg conc., root Fe conc., root Ca conc., root Mn conc. (7.32)
37	MQTL6B.3 (79.56–80.17)	Xwmc486/Xbarc76	3 (3.56)	GY, GNPS (4.93)
38	MQTL6B.4 (86.31–91.24)	AX-109018181/Xcfd1	4 (3.25)	SH, SFW, CC, SDW (6.05)
39	MQTL6B.5 (116.3–119.15)	wPt-8554/Xfbb377	4 (4.09)	CC, Ca conc., SFW, GNPS (6.61)
40	MQTL6D.1 (8.27–9.42)	Xfba307/wPt-667726	4 (2.80)	SH, DTF, RL, TDW (9.26)
41	MQTL6D.2 (18.72–35.72)	wPt-667726/Xpsr106	2 (4.23)	Total Chl. content, SII (14.69)
42	MQTL6D.3 (50.07–56.03)	Xgwm132/Xbarc23.1	2 (2.7)	TN, SOD (12.58)
43	MQTL7A.5 (64.65–65.6)	wPt-7830/wPt-9314	4 (2.61)	Root Zn conc., Root Na exclu., shoot Na conc., Cl conc. (13.56)
44	MQTL7A.6 (67.74–70.59)	wPt-9555/AX-111651290	3 (3.57)	SPS, SV%, RWC (10.03)
45	MQTL7A.7 (71.16–76.2)	Xbarc154/Xbarc219	2 (3)	BY, shoot Na conc. (4.73)
46	MQTL7A.8 (92.92–115.56)	AX-111694025/wPt-2084	2 (2.7)	Pro, OP (13.5)
47	MQTL7B.3 (61.87–64.98)	wPt-7295/AX-109949434	4 (3.98)	BY, Ca conc., SPS, GY (6.03)
48	MQTL7B.4 (141.77–142.43)	AX-94727938/AX-109783423	5 (2.54)	OP, SVI, root S conc., GI, G% (10.39)
49	MQTL7D.1 (0.24–0.63)	AX-109420634/wPt-743790	4 (3.65)	MSI, TGW, CC (10.7)

^atrait abbreviations: GY grain yield, TGW thousand grain weight, DTH/F/M days to heading/flowering/maturity, BY biomass yield, HI harvest index, PH plant height, SPP spikes per plant, SL spike length, SPS spikelet number per spike, SPW spike weight, GNPS grain number per spike, GL grain length, GW grain width, GP grain perimeter, GL/GW length/width of grains, SA surface area of grains per plant, AL awn length, TN tiller number, TDW total dry weight, conc concentration, SH shoot height, SG shoot growth, SDW shoot dry weight, SFW shoot fresh weight, SCC shoot chlorophyll content, RL root length, RDW root dry weight, LFW leaf fresh weight, LDW leaf dry weight, LS leaf symptoms, CC chlorophyll content, FV/FM chlorophyll fluorescence, SII salt injury index, G% germination percentage, GI germination index, SVI seedling vigor index, GR growth rate MSI membrane stability index, %WC percentage of water content, RWC relative water content, WP water potential, OP osmotic potential, Pro proline content, WSC water-soluble carbohydrate content of stems at anthesis, and SOD super oxide dismutase. conc. concentration, exclu. exclusion

Fig. 6.

Frequencies of known GWAS-MTAs which co-located with each of 49 MQTLs. Number of MTAs co-located with individual MQTL is given in parenthesis

Ortho-MQTLs in barley and rice

For identification of ortho-MQTLs, initially, 11 most stable and robust wheat MQTLs, each based on > 8 QTLs, were selected. Of these 11 MQTLs, ortho-MQTLs involving only 6 wheat MQTLs could be identified. Ortho-MQTLs of three of these wheat MQTLs were detected in syntenic regions of barley, while, only two ortho-MQTLs were identified in syntenic regions of rice. One ortho-MQTL (Ortho-MQTL3D.3) was cross-species in all three crops. This confirms the occurrence of wheat MQTLs orthologues to rice and barley MQTLs on known syntenic chromosomes among the three species. Based on the number of orthologs matched between wheat and rice MQTLs, and between wheat and barley MQTLs, two ortho-MQTLs of the following two wheat MQTLs could be considered most promising: (i) MQTL3D.3 on 3D, which was syntenic with rice MQTL positioned on chromosome 1 (with 7 corresponding genes matched) and barley MQTL located on chromosome 2 (with 10 genes matched) (Fig. 7); (ii) MQTL6A.1 on 6A, which was syntenic with MQTL on rice chromosome 2 (with 5 genes matched). In the remaining cases, only one to three genes were found to match between MQTLs of wheat and rice and wheat and barley, these ortho-MQTLs are described in Table 2. The traits associated with these ortho-MQTLs and underlying orthologous genes along with their functional descriptions are presented in Online Resource 11.

Fig. 7.

Syntenic region of one most promising ortho-MQTL (ortho-MQTL3D.3) among the wheat, barley and rice

Table 2.

Ortho-MQTLs in wheat, rice and barley based on the syntenic analyses

S. No	Ortho-MQTL	Wheat MQTL (genomic position in Mb)	Original MQTL name (chr., genomic position in Mb)	Reference
1	Ortho-MQTL2A.1	MQTL2A.1 (50.31–57.99)	Barley: MQTL2H.2 (2H, 35.77–48.45)	Zhang et al. (2017)
2	Ortho-MQTL3D.3	MQTL3D.3 (338.87–384.65)	Rice: MQTLSIS1.4(1,28.34–29.32), Barley: MQTL3H.2 (3H, 76.09–516.13)	Islam et al. (2019); Zhang et al. (2017)
3	Ortho-MQTL5A.2	MQTL5A.2 (37.63–393.47)	Barley: MQTL5H.3 (5H, 166.67–476.78)	Zhang et al. (2017)
4	Ortho-MQTL6A.1	MQTL6A.1 (11.78–46.48)	Rice: MQTLSNK2.2 (2, 5.90–9.88); MQTLSNK2.3(2, 14.99–24.53)	Islam et al. (2019)
5	Ortho-MQTL6A.3	MQTL6A.3 (30.89–32.66)	Rice: MQTLSNK2.3 (2, 14.99–24.53)	Islam et al. (2019)
6	Ortho-MQTL7B.2	MQTL7B.2 (12.58–18.20)	Barley: MQTL7H.3 (7H, 299.56–553.38)	Zhang et al. (2017)

Candidate genes for wheat MQTLs

The physical interval of individual MQTL ranged from 0.61 (MQTL2D.2) to 651.97 Mb (MQTL5B.5) with a mean of 100.26 Mb. Forty-seven (47) MQTLs each had physical intervals less than 20 Mb. The reason behind the long physical intervals of some of the MQTLs was the unavailability of exact flanking markers. Thirty-two (32) most stable and consistent MQTLs (each involving at least 4 initial QTLs from the independent experiments) were selected for the candidate gene mining. A total of 617 gene models were identified from these 32 MQTL regions (Online Resource 12). Several gene models with similar function descriptions were detected repeatedly on different chromosomes, including 17 gene models for F-box protein, 13 for pentatricopeptide repeat, 11 for cytochrome P450, 11 for UDP-glucuronosyl/UDP-glucosyltransferase, 8 for very-long-chain 3-ketoacyl-CoA synthase, 6 for F-box-like domain superfamily and 3 basic-leucine zipper proteins. Besides, many of these gene models identified as gene clusters with some consecutively arrayed in MQTL regions, for instance, such as consecutive zinc finger C2H2-type in MQTL1D.2, consecutive jacalin-like lectin domain in MQTL2B.1, consecutive pentatricopeptide repeat MQTL4B.2, and consecutive F-box proteins in MQTL6A.1 (Online Resource 12). Some genes encoding unpredicted or uncharacterized proteins were also detected in some MQTL regions. These genes deserve further attention, to unveil their possible roles in the regulation of salinity tolerance in wheat.

Discussion

Breeding for salinity tolerance is particularly difficult due to the complexity of the target environments and the stress-adaptive mechanisms that plants use to mitigate the detrimental impacts of salt stress. As a result, identifying and introgressing favourable QTLs/genes associated with different salinity-responsive traits are desirable strategies in order to develop a sound breeding programme for salinity tolerance. During the past two decades, numerous studies have been published on QTL mapping for the traits contributing to salinity tolerance in wheat, each of them using different sets of traits, genetic backgrounds and/or environmental conditions. To minimize redundancies and determine consensus genomic regions harbouring the most robust and stable QTLs among the mapping populations, in the present study, a meta-analysis of QTLs was performed. The use of genetic markers that account for a significant portion of the variability for traits conferring salt tolerance, such as those investigated in the present study, may expedite genetic gain by improving selection efficiency in segregating populations and offering breeding programmes a significant advantage in terms of selection for adaptation to salinity.

MQTLs and their validation with the related GWA studies

In the present study, a total of 844 QTLs were collected from the published literature as being associated with salinity tolerance traits in wheat. A total of 571 QTLs having complete required information were included in the meta-analysis which generated 100 MQTLs, i.e., regions statistically proved as unique. Two hundred and seventy-three QTLs were not considered for projection because their map positions or PVE values were not available from the respective studies. The QTLs density is thought to be primarily related to gene density and polymorphism rate (Martinez et al. 2016). Observations in the present study revealed that the majority of MQTLs and QTLs were found in subtelomeric areas with significant gene density. This is in agreement with previous reports in rice, barley, and maize where QTLs and MQTLs were found to be densely located at the subtelomeric regions (Martinez et al. 2016; Khahani et al. 2020; Tavakol et al. 2016). Out of total projected QTLs, only 31.9% QTLs exhibited PVE value > 10%, showing that the proportion of phenotypic variation contributed by individual QTL was very less. This suggests that salinity tolerance in wheat is primarily governed by numerous loci of small/minor effect and represent a complex genetic architecture. The refinement achieved in the present study was quite variable among loci. In some cases, such as MQTL2D.1, a total of 20 QTLs (whole QTL or parts) were grouped into one single locus, which is a significant reduction in redundancy. On the other hand, as many as 29 MQTLs were the result of the analysis of only two initial QTLs.

More than 50% (51.57%, 49/95) of the physically anchored MQTLs were validated with MTAs from GWAS, which indicated that the impact of these genomic regions on salinity stress tolerance may be less limited by genetic background. In addition, the contribution of wheat genomic regions to salinity stress tolerance varied greatly depending on the environment. As a result, breeding strategies may differ based on the environment. GWAS studies on spring wheat and winter wheat populations validated 26 and 44 MQTLs, respectively. These different MQTL regions may be more effective in improving tolerance to salinity stress for corresponding wheat regions and can be an important target of wheat breeding in these different wheat-growing areas. A comparison of MQTLs with GWAS-based MTAs, as mentioned above, has been previously conducted in wheat for the different traits including leaf rust resistance (Aduragbemi and Soriano 2021), multiple disease resistance (Saini et al. 2021a), grain yield and component traits (Yang et al. 2021), and nitrogen-use efficiency and root system architecture (Saini et al. 2021b). Only 38.66, 63.33, 61.37, and 78.57% of MQTLs could be verified with GWAS results in these earlier studies, respectively. Our results have been in consonance with these values of MQTLs verified by GWAS in earlier studies.

From a breeding perspective, Loffler et al. (2009) proposed a criterion, which can be taken into consideration for MQTL selection to be used in breeding programs: (i) MQTLs should have a small supporting interval (SI), (ii) the MQTL should include a large number of original QTLs and (3) the original QTLs should have high PVE or R² values (Saini et al. 2021c). Based on these criteria, we identified five potential MQTLs, namely, MQTL1D.2, 1D.5, 2A.4, 2D.1, 7D.1 (each with < 2 cM SI, involving 4 initial QTLs from different studies, and explaining > 10% of phenotypic variation) to be used for the improvement of salinity tolerance in wheat. Detection of a maximum number of promising MQTLs on the D genome shows its potential involvement in salinity tolerance. These MQTLs are the potential targets for marker-assisted breeding for the improvement of salinity tolerance in wheat, although, breeders should take into consideration the quantitative nature of these traits, therefore, more than one MQTL should be selected for breeding programmes.

Ortho-MQTLs for salinity in barley and rice

Despite the broad interest in identifying the genes involved in salinity tolerance in wheat and barley as the two most important staple crops, the associated genes have mostly remained unknown owing to their complex genomes. As the grass genomes are evolutionarily related to each other; synteny analysis of wheat, barley, and the model crop i.e., rice can enable broadening our genetic information among these important species. Identification of ortho-MQTLs among these related species may expand their usefulness and it may assist in identifying regulatory genes with conservative function. Here, we selected the 11 most promising wheat MQTLs to investigate their evolutionary conserved syntenic regions reported in similar meta-QTL analysis studies on the same traits in barley and rice. Consequently, we identified six ortho-MQTLs for salinity tolerance conserved at orthologous positions in wheat, rice, and barley. Ortho-MQTLs were detected on four barley chromosomes 2H, 3H, 5H, and 7H; each of which comprised only one previously detected MQTLs, while in rice, ortho-MQTLs were identified on chromosomes 1 and 2.

The genes located at the two most promising syntenic regions (viz., ortho-MQTL3D.3 and 6A.1) were further investigated. Ortho-MQTL3D.3 harbours the seven important genes, including 2 genes encoding for protein kinases and 1 gene each for armadillo-like helical, zinc finger (RING-type), AP2/ERF TF, glycoside hydrolase, and phosphatidic acid. Association of these genes with salinity tolerance has been reported in different crops including wheat and rice (Sharma et al. 2014a, b; Jin et al. 2016; Agarwal et al. 2020; Serra et al. 2013; Cao et al. 2017; Munnik et al. 2000), for instance, overexpression of a protein kinase gene (TaCIPK25) resulted in hypersensitivity to Na + and superfluous accumulation of Na + in transgenic wheat lines, thereby negatively regulated the salt response (Jin et al. 2016), overexpression of a RING zinc finger protein in wheat conferred increased tolerance to salinity. Moreover, transgenic lines performed well under salt stress conditions in terms of better growth, increased fresh weight, higher chlorophyll accumulation, increased proline content, and higher membrane stability (Agarwal et al. 2020). Two AP2/ERF transcription factors (OsEREBP1 and OsEREBP2) are known to regulate a receptor-like kinase gene (OsRMC), a negative regulator of salt stress response in rice (Serra et al. 2013). Another ortho-MQTL (ortho-MQTL6A.1) harbours the 3 salinity responsive genes for F-box proteins and 1 gene each for protein kinase and leucine-rich repeat domain-containing protein. Transgenic rice plants with overexpression of OsSIK1 (a putative RLK gene, with extracellular leucine-rich repeat) exhibited significantly improved activities of peroxidase, superoxide dismutase, and catalase contributing to higher tolerance against both salt and drought stress (Ouyang et al. 2010). Overexpression of the F-box gene (MAIF1) reduced rice abscisic acid sensitivity and abiotic stress tolerance and improved rice root growth (Yan et al. 2011).

In the present study, only a few salinity responsive genes were detected at corresponding genomic regions in rice and barley; although, a large number of salinity-responsive genes have been characterized in these two cereal crops (Negrao et al. 2011; Chen et al. 2021; Mwando et al. 2020). Similar to other reports on ortho-MQTL analysis (Khahani et al. 2020, 2021), in the present study, only a small set of wheat genes (identified only in the MQTL peak region i.e., 2 Mb) was utilized for mining of ortho-MQTLs or for uncovering similar MQTLs in the genomic regions of rice and barley. The original physical regions of wheat MQTLs may harbour a large number of genes that may correspond to a large set of genes in the studied cereals, including different salinity stress-responsive genes. Therefore, the ortho-MQTLs identified in the present study may be further investigated for the identification of other promising orthologs among the studied cereals.

Orthologous gene sets identified in the present study may be functionally validated and precise molecular markers such as ‘conserved orthologous set’ markers may be developed which may enable the identification of diagnostic SNP appropriate for selection programs. Similar ortho-MQTL analyses have also been conducted for other traits in wheat and other cereals (Quraishi et al. 2011; Khahani et al. 2020, 2021; Saini et al. 2021b, c; Kumar et al. 2021). Overall, this approach provided a thorough understanding of potential genomic regions regulating investigated traits under salinity stress conditions with close evolutionary history and conserved function among the cereals.

Candidate genes available in salinity MQTL regions

MQTLs are considered potential genomic regions that harbour high confidence CGs for the traits. In the present study, gene mining was performed in the most robust and stable MQTL regions and a total of 617 gene models were listed. The genes detected in the present study are considerably different from the genes identified in the earlier individual genetic mapping studies (Genc et al. 2014; Asif et al. 2018, 2020), with only a very few genes overlapped, which encoded proteins belonging to different gene families namely: V-type proton ATPase, ABC transporter, peptide transporter, etc. These results are not surprising, as the meta-QTL analysis prioritized only those genomic regions that are most frequently involved in trait variation and narrowed down the QTL CIs (both genetic and physical), significantly, thus facilitating the identification of most promising CGs available in high confidence genomic regions. Moreover, in the present study, the genes were identified only in MQTL peak regions, which further strengthened the confidence in identified genes. Among the gene models detected in MQTL regions, as many as 81 most likely CGs (putative salinity-responsive genes) were selected based on the criteria i.e., the gene has a known physiological role in providing tolerance to salinity in any crop species or the gene affects the trait in question-based on studies of knock-outs, mutations, or transgenics in any crop species, preferably in cereal crops (Table 3).

Table 3.

Most likely candidate genes of some robust and stable MQTLs

S. No	MQTL name (Total predicted CGs)a	Putative Gene ID	Gene Function Description
1	MQTL1A.3 (14)	1A02G329400	Phosphatidylinositol 3-/4-kinase
		1A02G329900	Basic-leucine zipper protein
2	MQTL1D.2 (36)	1D02G039200	Very-long-chain 3-ketoacyl-CoA synthase
		1D02G040800	Histidine kinase/HSP90-like ATPase superfamily
		1D02G042800	Germin
3	MQTL1D.3 (3)	1D02G135500	P-type ATPase
4	MQTL1D.5 (31)	1D02G349800	UDP-glucuronosyl/UDP-glucosyltransferase
		1D02G349100	Pentatricopeptide repeat
		1D02G347700	Protein kinase domain
		1D02G348900	Pyruvate phosphate dikinase, AMP/ATP-binding
5	MQTL2A.1 (68)	2A02G097300	Pentatricopeptide repeat
		2A02G099900	Ethylene insensitive 3
		2A02G103200	Glycosyl transferase
		2A02G101100	NAC domain
		2A02G099900	Ethylene insensitive 3
		2A02G104800	WRKY domain
		2A02G105300	tRNA methyltransferase
6	MQTL2A.4 (17)	2A02G292400	Cytochrome P450
7	MQTL2A.8 (21)	2A02G293200	Histone deacetylase family
		2A02G439800	Cytochrome P450
		2A02G441100	O-methyltransferase protein
		2A02G439400	Alternative oxidase
		2A02G440800	Hypoxia induced protein
		2A02G441900	TPR and ankyrin repeat-containing protein 1
		2A02G440900	O-methyltransferase domain
8	MQTL2B.1 (47)	2B02G047400	Heat shock protein Hsp90 family
		2B02G047700	Dirigent protein
		2B02G049200	Thiolase
		2B02G049400	Cytochrome P450
9	MQTL2B.3 (8)	2B02G301000	Nodulin-like
10	MQTL2D.1 (9)	2D02G386300	B-box-type zinc finger
		2D02G386500	C2 domain containing protein
		2D02G386000	Cytochrome P450
11	MQTL3A.3 (23)	3A02G315900	Pentatricopeptide repeat
		3A02G316700	Diacylglycerol kinase
12	MQTL3B.2 (12)	3B02G077900	Inorganic pyrophosphatase
13	MQTL3B.4 (13)	3B02G384300	Pentatricopeptide repeat
		3B02G385500	V-type ATPase
		3B02G384100	Myb/SANT-like domain
14	MQTL3D.3 (14)	3D02G258900	Methyltransferase type 11
		3D02G260100	AP2/ERF domain
		3D02G260400	Phosphatidic acid phosphatase type 2/haloperoxidase
15	MQTL4A.5 (12)	4A02G248300	Pentatricopeptide repeat
		4A02G248400	Phospholipid/glycerol acyltransferase
16	MQTL4B.1 (7)	4B02G064200	Dehydrin
		4B02G064800	F-box-like domain superfamily
17	MQTL4B.2 (21)	4B02G342900	Tetratricopeptide-like helical domain superfamily
		4B02G342800	Transcription factor GRAS
		4B02G343000	Cation/H + exchanger
		4B02G342600	Heat shock protein 70 family
18	MQTL5A.2 (3)	5A02G111500	NIF system FeS cluster assembly
19	MQTL5A.3 (13)	5A02G049200	Methyltransferase protein
20	MQTL5B.2 (10)	5B02G272500	F-box-like domain superfamily
		5B02G272700	Methyl-CpG DNA binding protein
21	MQTL5B.4 (15)	5B02G439900	Alpha/beta hydrolase fold-3
		5B02G440700	Heat shock factor (HSF)-type, DNA-binding
		5B02G439700	UDP-glucuronosyl/UDP-glucosyltransferase
22	MQTL5D.1 (7)	5D02G210000	Glycosyl transferase
23	MQTL5D.2 (28)	5D02G320800	AP2/ERF domain
		5D02G321000	Heat shock factor (HSF)-type, DNA-binding protein
		5D02G321800	Small GTPase
24	MQTL6A.1 (24)	6A02G054100	F-box domain
		6A02G056600	Small auxin-up RNA
		6A02G056800	Leucine-rich repeat domain superfamily
25	MQTL6A.3 (24)	6A02G058400	Cytochrome P450
		6A02G059300	UDP-glucuronosyl/UDP-glucosyltransferase
		6A02G059600	Glutathione S-transferase protein
		6A02G060700	Glycoside hydrolase protein
26	MQTL6B.1 (12)	6B02G053700	F-box-like domain superfamily
27	MQTL6B.2 (13)	6B02G232000	Zinc finger, GRF-type
28	MQTL7A.3 (26)	7A02G107300	Dual specificity phosphatase
		7A02G107500	Very-long-chain 3-ketoacyl-CoA synthase
29	MQTL7B.1 (23)	7B02G009500	Transcription factor, MADS-box
		7B02G010200	CHAPERONE-LIKE PROTEIN OF POR1-like
30	MQTL7B.2 (23)	7B02G017700	GDSL lipase/esterase
		7B02G018400	AP2/ERF domain
		7B02G018800	Pectinesterase
31	MQTL7D.1 (39)	7D02G019200	UDP-glucuronosyl/UDP-glucosyltransferase
		7D02G019300	Ankyrin repeat
		7D02G019400	Glutathione S-transferase
		7D02G019600	Expansin

^aAfter removing gene models with no function descriptions

The MQTL2D.1 peak corresponded to nine gene models involving two most likely CGs encoding B-box-type zinc finger and cytochrome P450 protein. Over-expression of a gene, MdBBX10, encoding for B-box-type zinc finger protein in Arabidopsis significantly increased the tolerance to salinity stress, with positive effects on germination and root morphology-related traits (Liu et al. 2019). Cytochrome P450s are versatile catalysts that play a significant role in the biosynthesis of antioxidants, metabolites, and phytohormones in plants. The potential role of cytochrome P450s enzyme family in the regulation of salinity and other stresses has been recently reviewed in the literature (Pandian et al. 2020).

A total of 12 gene models, including two most likely CGs encoding for pentatricopeptide repeat and phospholipid/glycerol acyltransferase, were detected in the MQTL4A.5 region. Overexpression of a gene, SOAR1 encoding for pentatricopeptide repeat protein provided high tolerance to salinity stress and other stresses including drought and cold without any negative effect on the growth and development of plants (Jiang et al. 2015). Transgenic lines of Arabidopsis carrying a gene, SsGPAT encoding for glycerol acyltransferase improved salinity tolerance and further alleviated the photoinhibition of PSI and PSII under salinity stress (Sui et al. 2017).

The MQTL4B.1 corresponded to 7 gene models including two most likely CGs encoding for dehydrin and F-box-like protein. Transgenic rice plants carrying a dehydrin gene i.e. OsDhn1-OXs, showed enhanced tolerance to salinity and drought stress as evaluated by dry and fresh weight, chlorophyll fluorescence, chlorophyll, and water content and survival ratio (Kumar et al. 2014). F-box proteins are known to play important roles in the specific and selective degradation of protein through the 26S proteasome. Studies have revealed the possible response of the F-box proteins to salinity and other abiotic stresses in plants (Song et al. 2015).

As many as 24 gene models, including three most likely CGs encoding for F-box protein, small auxin-up RNA, and leucine-rich repeat protein were detected in the MQTL6A.1 region. The cell wall leucine-rich repeat extensins were observed as important for plant salinity tolerance in Arabidopsis (Zhao et al. 2018). It was suggested that the leucine-rich repeat extensins along with other proteins such as RAPID ALKALINIZATION FACTOR and receptor-like protein kinase FERONIA regulate the transduction of cell wall signals thereby controlling the salt stress tolerance and plant growth (Zhao et al. 2018). A small auxin-up RNA gene i.e. SAUR39 was observed as a negative regulator of auxin synthesis in rice. The expression of this gene exhibited fast induction by transient alteration in different environmental factors such as salinity, indicating that this gene mediates stress response in plants (Kant et al. 2009).

The MQTL6A.3 corresponded to 24 gene models including four most likely CGs encoding for cytochrome P450, UDP-glucuronosyl/UDP-glucosyltransferase, glutathione S-transferase protein, and glycoside hydrolase protein. A gene, UGT71C5 encoding for UDP-glucosyltransferase was characterized in Arabidopsis, which plays a crucial role in abscisic acid (ABA) homeostasis. Abscisic acid is a hormone that plays a crucial role in controlling plant growth and development as well as mediating adaptations to different abiotic stresses including salinity (Liu et al. 2015). In Arabidopsis, overexpression of a gene i.e. OsGSTU4 encoding for glutathione S-transferase (GST) protein exhibited good growth and higher GST activity and hence improved tolerance towards salt and oxidative stresses (Sharma et al. 2014a, b). Glycoside enzymes are known to participate in several developmental processes, plant hormone deactivation or activation, and responses to stresses including salinity in Arabidopsis and rice (Cao et al. 2017).

Details on the remaining most likely CGs identified in different MQTL regions are given in Table 3. Figure 8 indicates the possible mechanisms and most-likely CGs involved in the wheat salt tolerance; only some of the important class of genes associated with membrane stabilization, hormone biosynthesis, Na + extrusion (at root level), osmotic exclusion (at shoot level), and tissue tolerance (at leaf levels) are shown. Once characterized, the CGs may be exploited through biotechnological approaches such as transgenesis and gene editing to develop novel salinity tolerant cultivars. However, in some cases, where gene clusters regulate the target trait, the transgenic method using a single gene may not be as effective as marker-assisted selection, where flanking markers can target a much larger region encompassing all the genes of a cluster.

Fig. 8.

Wheat salt tolerance mechanism-overview of important salinity-responsive genes involved at root, shoot and leaf levels

Conclusion

Salinity tolerance is a complex quantitative trait governed by numerous QTLs of small effects. The meta-analysis of a large set of QTLs gives valuable information for robust and stable QTLs to be targeted for breeding. In this study, we collected 844 QTLs related to salinity tolerance, of which 571 QTLs were projected giving origin to 100 MQTLs. Forty-nine of these MQTLs were also validated with GWAS-MTAs. The MQTLs identified and selected in the present study as potential MQTLs associated with salinity stress tolerance include the following: MQTL1D.2, MQTL1D.5, MQTL2A.4, MQTL2D.1, and MQTL7D.1. The robust markers flanking these MQTLs can be used for marker-assisted breeding programs for improving salinity tolerance in wheat. As many as 617 gene models underlying 32 MQTLs were also identified which included 81 most likely CGs underlying different MQTLs with known functions as being reliable for salinity tolerance. The results provide a framework for future genetic and basic studies involving, fine mapping, positional cloning, and functional analysis of the potential genomic regions as well as validation of CGs by gene editing and similar techniques. Moreover, the identification of ortho-MQTLs among the wheat, rice, and barley shows the potential of ortho-MQTL mining strategy for the detection of evolutionarily conserved genomic regions and CGs regulating complex traits like salinity tolerance.

Supplementary Information

Below is the link to the electronic supplementary material.

Author’s contribution

D.K.S. and S.K. conceived and designed the project; S.K. supervised the study; D.K.S. and N.P. conducted the analysis and wrote the paper and S.K. corrected the final draft. All authors read and approved the final manuscript.

Funding

This research received no external funding.

Data availability

Data generated or analyzed during this study are included in this published article (and its Supplementary Material).

Declarations

Conflict of interest

The authors declare that they have no competing interests.