Introduction

Improved yield with high quality and adaptability is the primary goal of wheat breeding programs to address global food security and sustainability. Wheat yield is a complex quantitative trait controlled by quantitative trait locus/loci (QTL) and affected by environmental factors. A large number of QTL have been detected for wheat yield under different environments, including irrigated, drought- and/or heat-stressed conditions. However, few have been successfully adopted for marker-assisted selection (MAS) in breeding. One reason lies in the low map resolutions in many studies, resulting in large QTL intervals with loose marker-gene linkages. Another reason is that many of these QTL have not been validated, as validation or fine mapping of the loci requires substantial effort and investment. Meta-analysis provides a good alternative for validating QTL and narrowing the QTL intervals, as it can detect consistent QTL by integrating QTL from studies using various environmental and genetic backgrounds, and reduce their genetic intervals, leading to the identification of candidate genes (Wu et al. 2016).

Genome-wide QTL meta-analysis approach has been applied for many traits in wheat. A meta-QTL (MQTL) analysis, based on a consensus map of major and consistent QTL for yield and yield-related traits, identified 12 significant MQTL on chromosomes 1A, 1B, 2A, 2D, 3B, 4A, 4B, 4D, and 5A, some of which contained important known genes such as Rht and Vrn (Zhang et al. 2010). However, candidate genes underlying the meta-QTL were not investigated due to the limited knowledge of the reference genome sequence at the time. Tyagi et al. (2015) conducted an MQTL analysis of grain morphological traits, including grain weight, and identified 17 reliable MQTL on chromosomes 1B, 2A, 2D, 3B, 4A, 6A, and 6B, with 16 of which related to grain weight. Again, no candidate genes were suggested in this study. Acuña-Galindo et al. (2015) conducted a meta-analysis of wheat QTL associated with drought and heat tolerance. They identified 66 MQTL distributed throughout the genome, many of which were co-localized with the yield MQTL identified by Zhang et al. (2010). The authors identified 41 candidate genes through SNP-gene association, including those involved in sugar metabolism, reactive oxygen species (ROS) scavenging, and abscisic acid-induced stomatal closure. Apart from yield-related traits, MQTL analyses have been undertaken for other traits in wheat, including flowering date (Hanocq et al. 2007) and preharvest sprouting tolerance (Tyagi and Gupta 2012). These MQTL studies surveyed relevant QTL studies and refined the positions of some major QTL. However, none investigated how the identified MQTL are distributed in a wide range of wheat genotypes or how they contribute to trait performance.

Reports identifying new yield-related QTL in wheat using high-resolution markers have increased in the past few years (Cabral et al. 2018; Su et al. 2018; Tura et al. 2020). An MQTL analysis incorporating these new studies is needed. Publication of the reference genome sequence in wheat (Appels et al. 2018) paved the way for more efficient gene identification and marker development. Many yield-related genes have also been identified in other cereal crops, including rice and barley (Huang et al. 2018; Nadolska-Orczyk et al. 2017), with some explaining a significant proportion of the phenotypic variation in their traits, such as OsGS3 in rice for grain weight (Fan et al. 2006). This newly available information will facilitate MQTL analysis on gene predictions and marker development.

As wheat requires significant local adaptation to succeed, breeders need to develop high-yielding cultivars adapted to their target environment. Contributions of genomic regions to yield may vary according to the target environment. Therefore, assessments of the major genomic regions responsible for wheat should be undertaken in the targeted region. This study assessed the MQTL of selected cultivars with extreme expressions of yield traits using association analyses of genotypic data obtained via ddRAD sequencing and phenotypic data from 2012 to 2016 Australian National Variety Trials (NVT) (https://www.nvtonline.com.au). Since Australian wheat cultivation mostly takes place under rainfed conditions, where drought and/or heat stress often occur, the QTL responsible for yield in drought- and/or heat-stressed environments were included in the MQTL analysis, as well as those in irrigated environments.

By integrating and summarizing the results from separate QTL mapping studies, this study aims to locate consistent QTL regions associated with wheat yield and its component traits. We reviewed the genomic regions involved in the control of wheat yield and its related traits and refined the positions of yield-related QTL. While many genes for small effects may contribute to yield, this study only focused on consistent or major QTL to identify MQTL and refine their chromosomal positions to improve MAS efficiency for yield-related traits in wheat.

Materials and methods

Dataset development for yield-related QTL

As Zhang et al. (2010) conducted an MQTL analysis on wheat yield for studies pre-2010, we reviewed and summarized 24 QTL studies published from 2010 to 2020 related to yield and yield components under different environments, including irrigated, drought- and/or heat-stressed conditions (Supplementary Table S1). We identified 381 QTL for inclusion in our meta-analysis, being major QTL with > 10% of the phenotype variation explained (PVE), and those that were consistent within each study. The MQTL analysis required the R2 (percentage PVE) and confidence interval (CI) for each QTL (Zhang et al. 2010). The most recent published study (Tura et al. 2020) did not provide PVE values, so the QTL were excluded from the MQTL analysis. However, the QTL were included as a separate reference for comparison on the MQTL map as the study used 3502 high-resolution markers and identified consistent yield QTL, with significant yet small main effects, across 10 environments on three continents over six seasons.

Projection of QTL on a consensus map and MQTL analysis

The major markers used for constructing genetic linkage maps in QTL mapping studies include simple sequence repeat (SSR), diversity arrays technology (DArT), and single nucleotide polymorphism (SNP) markers. A highly saturated consensus map—containing 52,607 markers, including 51,655 SNP, 667 SSR, 266 DArT, and 19 other markers—was simplified for use as a reference map (Wen et al. 2017). Simplification reduces the computer processing time for QTL projection, as it only retains selected markers and reduces the number of markers for the same chromosome positions. The selected markers included those that: (1) overlapped, at least once, with the markers used in the collected QTL studies; and (2) had different chromosome positions to the markers in (1), but only one marker was retained for each chromosome position. The simplification step did not reduce genetic distance on the consensus map because we retained at least one marker for every original map position. In addition, the removed markers were not used in the 24 QTL publications collected for this MQTL study, so their removal will not affect the MQTL results. The simplified consensus map contained 22,664 markers, including 21,909 SNP, 534 SSR, 214 DArT, and seven other markers, and was used as the reference map for QTL projection.

BioMercator V4.2 (https://urgi.versailles.inra.fr/Tools/BioMercator-V4) (Arcade et al. 2004) was used to project QTL from different populations in the collected studies onto the reference consensus map. The QTL projection was based on LOD scores, PVEs, CIs, and QTL positions. For those QTL lacking flanking markers and CIs, a 95% CI was calculated as 530/(N × R2), where N is the population size, and R2 is the proportion of phenotypic variance (Zhang et al. 2017); the positions of the closest markers to these intervals were selected as the QTL positions on the reference map. We eliminated QTL that did not have a tightly linked consensus marker from further analysis. Meta-analysis was performed on the QTL clusters for each chromosome using algorithms from the BioMercator software (Goffinet and Gerber 2000; Veyrieras et al. 2007). The lowest Akaike information criterion (AIC) value was used to select the best QTL model for identifying the number of MQTL on each chromosome (Lu et al. 2018).

Searching for putative candidate genes within the MQTL confidence intervals

The flanking markers of the MQTL were blasted with the reference genome IWGSC RefSeq v1.0 (https://wheat-urgi.versailles.inra.fr/) (Appels et al. 2018). Putative candidate genes for that MQTL with a physical interval of less than 20 Mbp were identified to extract gene information within or around the intervals; their functions were compared to find the best possible candidate genes.

Genotype–phenotype association analysis of Australian cultivars with contrasting yield performance

Field phenotypic data of 109 cultivars were collected from the 2012 to 2016 Australian NVT conducted at 108 trial locations across different environments, including Western Australia, Queensland, New South Wales, Victoria, and South Australia. The soil texture and weather conditions of each trial location are available on the NVT website (https://www.nvtonline.com.au). The mean yield performance of each cultivar across 5 years was compared across environments (Table S2). Phenotypic data from 2015 to 2017 CIMMYT Australia ICARDA Germplasm Evaluation (CAIGE) field trials were collected as a reference check (http://www.caigeproject.org.au/germplasm-evaluation/bread/data-compilations/). Cultivars that consistently (in at least two of the four environments listed in Table S2) fell into the top 10 highest or lowest yielding cultivars were chosen for further investigation. As a result, 14 cultivars were used for genotype–phenotype association analysis, including Cobalt, Tenfour, Beckom, Scepter, Scout, Suntop, and Trojan in the high-yielding group, and Sunvale, Crusader, Dart, Hatchet CL Plus, Impress CL Plus, Tungsten and Lang in the low-yielding group. Differences between the mean yields of the high- and low-yielding groups were analyzed using a t test (Table S2).

The 14 cultivars were genotyped using ddRAD sequencing, a two-enzyme genotyping-by-sequencing (GBS) technology. Specifically, genomic DNAs were extracted using leaf tissue from three-leaf stage seedlings using a modified CTAB method (Mia et al. 2019), and their quality examined on 1% agarose gel and a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific Inc., Australia). The samples were sent to the Beijing Genomics Institute (BGI), China, where they were double-digested using a rare-cutting EcoRI-HF and a frequent-cutting MseI enzyme. Sequencing libraries of 150 bp paired-end reads were prepared and sequenced using HiSeq X Ten (Illumina, San Diego, USA) according to the manufacturer’s standard protocols. Raw sequencing data were filtered following the procedure described in Mia et al. (2020). Genotyping of SNP and detection of indel markers were performed on the clean data. The GBS markers within the MQTL physical positions were blasted with the wheat reference genome to search for their associated genes.

Results

Identification of MQTL for yield and yield components under different environments

Major (R2 > 10%) or consistent QTL were selected for the meta-analysis, as the only QTL with these qualities can be used in MAS (Zhang et al. 2017). As such, 381 QTL for wheat yield and yield components were projected on different chromosomes (Table S1; Fig. S1). Chromosome 1B had the most QTL (58), and chromosomes 3D and 5D had the least (3) (Table 1). A total of 86 MQTL—integrated QTL from several experiments—were identified by meta-analysis; of which, 39 were responsible for yield under drought-stressed environments, 36 under heat-stressed environments, 23 under drought- and heat-stressed environments, and 34 under irrigated/non-stressed environments (or environments without a specific stress mentioned). MQTL1B.5 integrated the most (12) initial QTL, followed by MQTL3B.2 (10); both formed a significant peak on the density curves (Fig. S1), suggesting that they are hotspots for improving yield in wheat. Chromosomes 4A, 4B, and 7A had the most MQTL detected (7 each), whereas chromosomes 1A, 3D and 5D had the least MQTL (one each). The meta-analysis reduced the CIs of the original QTL from 12.7 cM on average to 5.2 cM for each MQTL. The CIs of MQTL ranged from 0.03 cM in MQTL1B.7 and MQTL1B.8 to 25.06 cM in MQTL6B.2. Some MQTL directly flanked markers for commonly known genes, including MQTL2D.1 with photoperiod loci Ppd-D1, MQTL4B.3 with reduced height loci Rht1, and MQTL4D.1 with Rht2 (Table 1). The physical length of these MQTL ranged from 0.17 to 71.36 Mb (Table 2).

Table 1 Yield-related MQTL detected in this study
Full size table
Table 2 Putative candidate genes of MQTL
Full size table

Putative candidate gene identification

Apart from the commonly known genes Ppd, Vrn, and Rht, the annotations of high-confidence genes reported in RefSeq v1.0 were screened for the target MQTL with a physical interval < 20 Mbp. The functions of genes within the physical intervals, or around the intervals if inadequate genes were inside each interval of the MQTL, were scrutinized for putative candidate genes. Those genes with functions previously reported as important for yield and stress tolerance traits, such as sugar transporter, sucrose synthase, ethylene response factor, stress-associated proteins, were considered putative candidate genes for the MQTL. Two hundred and ten putative candidate genes were identified using this method (Table 2), of which several genes with similar functions were identified repeatedly on different chromosomes, including 42 genes for functions as E3 ubiquitin ligases, 21 for heat shock proteins (HSPs), 21 for MYB transcription factors (TFs), 15 for ethylene-responsive genes, 14 for sugar transporters, 10 for NAC TFs, and 10 for WRKY family genes. Many of these genes acted as gene clusters with some consecutively arrayed in MQTL regions, such as consecutive HSPs in MQTL1A.1, MQTL2A.2, MQTL3B.3, MQTL 3D.1, and MQTL6D.1, consecutive MYB TFs in MQTL1B.4, MQTL3B.3, MQTL 3D.1, and MQTL7D.2, consecutive NAC TFs in MQTL2A.3, consecutive defensins in MQTL2B.2, consecutive E3 ubiquitin ligases in MQTL3B.1 and MQTL3B.3, consecutive MADS-box proteins in MQTL6B.4, consecutive WRKY TFs in MQTL7A.1 and MQTL7B.5, consecutive transcription elongation factors in MQTL7A.5, consecutive peroxidase in MQT7A.6, consecutive sugar transporters in MQTL7A.7, and consecutive NBS-LRR disease resistance proteins in MQTL7D.3 (Table 2).

Genotype–phenotype association of 14 cultivars with contrasting yield performance

Around 10 Gb of clean data per cultivar sample were obtained from the ddRAD sequencing for the 14 selected cultivars with contrasting yield performance. A total of 139,657 indel and 1,001,955 SNP markers were generated for genotyping. After filtering out markers with no polymorphism, 73,205 indels and 627,697 SNPs were used for final genotyping (Table S3). Variations in the GBS markers most closely located to the identified MQTL gene regions were compared between the two groups of cultivars (7 high-yielding and 7 low-yielding); the t-test identified a highly significant difference (p < 0.01) for group mean yield (Table S2). Markers with distinctive variations only in the high-yielding group were considered positive alleles for yield, while those only in the low-yielding group were considered negative alleles for yield. Table 3 lists the positive alleles, and some of the negative alleles if they fell inside the target genes. Generally, only the closest marker is shown for each candidate gene; however, markers showing additional variations, or significant variations between the two groups—which could be useful for MAS—were included in the list, even if they were not the closest marker to the gene. Examples of such markers are: (1) an indel marker on chromosome 2A showing variations in five of the 14 cultivars, 0.74 Mbp away from genes TraesCS2A01G136700 and TraesCS2A01G136800; and (2) an indel marker on chromosome 2B with distinctive positive and negative alleles for yield, 2.67 Mbp away from TraesCS2B01G110000 and TraesCS2B01G110100. Notably, an SNP marker on chromosome 2B, 0.01 Mbp away from the target gene TraesCS2B01G087400, showed a distinctive negative allele for yield in the low-yielding group, and a positive allele mostly in the high-yielding group (5 of 7 cultivars).

Table 3 Variations in gene-linking GBS markers among the 14 wheat cultivars contrasting in yield performance
Full size table

Putative candidate genes with distinctive positive alleles in the high-yielding group in at least two cultivars were deemed validated for their important role for wheat yield. Eighteen genes or gene clusters were validated, of which two were located in MQTL2A.2 (TraesCS2A01G136700 and TraesCS2A01G138100), two in MQTL2B.1 (TraesCS2B01G087400 and TraesCS2B01G089700), four in MQTL2B.2 (TraesCS2B01G105100, TraesCS2B01G105300, TraesCS2B01G110000, and TraesCS2B01G112600), three in MQTL7A.1 (TraesCS7A01G090700, TraesCS7A01G095100, and TraesCS7A01G096200), and one each in MQTL1D.1, MQTL2A.1, MQTL3A.2, MQTL3D.1, MQTL4A.7, MQTL6A.1, and MQTL7D.3 (Table 3).

Discussion

Key MQTL and associated genes can be target regions for improving yield in wheat

Although wheat yield has increased over time as a direct result of new varieties produced by breeders, there is substantial room to improve, compared to other major cereal crops, such as rice and barley. Genes or genomic regions responsible for high yield potential, adaptability, and stability are desirable. Meta-analysis can map QTL from different mapping populations in different experiments on the same linkage group and lower the QTL CIs for more effective identification of candidate genes (Goffinet and Gerber 2000). In this study, 381 QTL were integrated into 86 MQTL responsible for yield-related traits under irrigated, drought- and/or heat-stressed conditions in wheat. All of the MQTL had narrower CIs (95%) than the mean values for the original QTL. Six MQTL spanned a physical interval of < 0.5 Mbp each—MQTL1B.1, MQTL1D.1, MQTL4A.5, MQTL4D.1, MQTL5A.4, and MQTL6A.2. Two hundred and ten candidate genes were identified for the MQTL—those with small genetic and physical intervals or with candidate genes or gene clusters having robust effects on yield and stress tolerance (Table 2) are important regions for MAS, pyramiding, fine mapping, positional cloning, and functional analysis.

Wheat yield is controlled by many genes with small effects and significantly affected by environmental factors. Contributions of genomic regions to yield will vary according to the target environment; therefore, breeding strategies will differ for individual environments. Identification of major or consistent genes within a targeted environment is crucial for successful gene-stacking practices. In this study, phenotypic data of 14 cultivars were collected from 5-year Australian NVT in key wheat production areas. Genotype–phenotype association analysis between the different cultivar groups (high- or low-yielding) enabled us to detect the genomic regions responsible for high yield in Australian environments. Eighteen genes or gene clusters were validated and can be used as main targets for wheat breeding in these areas.

Drought and heat stress are major limiting factors to wheat yield, especially in rainfed farming areas. Breeding stress-tolerant varieties remain the best approach for increasing crop production (Zhang et al. 2017). This study identified major chromosome regions responsible for yield and yield components under drought and/or heat stress. The MQTL contained validated genes and showed combined drought and heat stress tolerance for MQT1D.1, MQTL2B.2, MQTL3D.1, and MQTL7D.3. These regions and their validated genes can act as favorable locus/gene combinations for pyramiding to create highly adaptable high-yielding cultivars.

The MQTL identified in this study were compared with the QTL detected in a recent QTL mapping study on wheat yield under drought and heat stress (Tura et al. 2020). The study was based on a linkage map containing 3502 markers and phenotypic data collected from 32 field experiments at 10 locations over six seasons. Two of the QTL were co-localized with MQTL with matching traits—QTgw.aww-7B.2 overlapped MQTL7B.3, and QYld.aww-7DS.2 overlapped MQTL7D.3. Tura et al. (2020) fine mapped the 1B QTL QYld.aww-1B.2 to a 2.2 Mbp physical interval of 659,988,745–662,154,351 bp, which was close (18,789,404 bp away) to MQTL1B.8 at an interval of 638,015,656–641,199,314 bp. Two other QTL were closely located to the MQTL—QTgw.aww-6B to MQTL6B.1 (5.95 Mbp away), and QYld.aww-7A.3 to MQTL7A.7 (1.09 Mbp away). These five MQTL are important for future breeding of wheat yield.

The MQTL associated with commonly known genes, such as Ppd, Vrn, and Rht, were further explored in this study to detect other genes that might be co-located in the genomic regions. No candidate mutation was found for Ppd-B1 alleles and Vrn-A1 alleles in Chinese Spring wheat (Díaz et al. 2012), so the search of putative candidate genes excluded these gene regions on RefSeq V1.0. Studies have suggested that other yield-related genes could co-exist in these regions; for example, Kadam et al. (2012) identified a 4BS QTL for grain yield under drought stress that explained up to 22% of the phenotypic variation, and its contributor was a tall cultivar C306. Although the QTL, flanked by markers Xbarc20 and Xgwm368, is only 27 Mbp away from chromosome 4B’s Rht1 gene (responsible for reduced height), it is a large-effect locus independent of Rht1. Fine mapping of QTL/genes with low Q × E effects, independent of phenology, should be focused on to identify the causal gene(s) controlling yield. Therefore, more genes in the major MQTL regions need to be identified, excluding known yield-related genes such as plant height (Rht) or phenology (Ppd or Vrn).

Orchestrated networks of gene families and gene clusters play important roles in determining yield performance in wheat

The genes associated with the MQTL could be classified into yield-related and stress-related genes. The most yield-related genes functioned as E3 ubiquitin ligases, which also underlie the known Grain Weight 2 (GW2) gene on chromosome 6D (Su et al. 2011). Nadolska-Orczyk et al. (2017) reviewed that the major genes determining yield-related traits in wheat often had functions in four categories, including TFs, carbohydrate metabolism, or signaling of growth regulators, cell division and proliferation, and floral regulators. They summarized some yield-determining genes in wheat with known orthologues in rice, maize, and barley, including cytokinin dehydrogenase (CKX), transcript elongation factor (TEF), GW2, thousand-grain weight 6 (TGW6), grain size (GS5 and GS-D1), sucrose synthase (Sus1 and Sus2), Nuclear Factor Y (NFYAs, NFYBs, and NFYCs), NAC TFs, and cell wall invertase (CWI). The TCP family genes also affect the development of spike and grain development (Zhao et al. 2018) and showed high frequencies in the MQTL regions identified in this study, which were considered putative candidate genes related to yield. The C4 enzymes genes, such as phosphoenolpyruvate carboxylase (PEPC), located in MQTL1B.2 and MQTL5D.1 in this study, are important for increasing yield in C3 plants like wheat. Over-expression of C4 enzyme genes PEPC and PPDK (pyruvate orthophosphate dikinase) increased yield and photosynthesis rate in wheat (Häusler et al. 2002; Khan et al. 2019; Zhang et al. 2014).

Many of the candidate genes identified in the MQTL belong to large gene families responsive to stress, including MYB TFs, NAC TFs, WRKY TFs, ethylene-responsive TFs, HSPs, NADPH family gens, and MADS-box genes. These genes have been found responsible for tolerances to various stresses in wheat (Dey and Corina Vlot 2015; Erdayani et al. 2020; Guérin et al. 2019; Hu et al. 2018; Mia et al. 2019, 2020; Schilling et al. 2020; Xue et al. 2011; Yousfi et al. 2017; Zhao et al. 2017, 2018). Kulkarni et al. (2017) reviewed the key genes responsive to drought stress for controlling root system and transpiration in wheat. Later, Khan et al. (2019) summarized the genes responsible for drought tolerance that were used for transgenic improvement in wheat, in which WRKY, NAC, and Dehydration Responsive Element Binding (DREB) genes were among the major responsible or targeted genes. Heat shock TFs (Hsfs), which modulate the expression of HSPs, are crucial for stress tolerance (Xue et al. 2014; Zhou et al. 2019). Other stress-responsive factors include proline (Mwadzingeni et al. 2016), mannitol (Abebe et al. 2003), trehalose (Ibrahim and Abdellatif 2016), defensin (Stotz et al. 2009), and late embryogenesis abundant (LEA) proteins that act as a molecular chaperone protein by stabilizing the protein or membrane structure (Chen et al. 2019). Chini et al. (2004) reported that some disease resistance proteins, such as NBS-LRR, play crucial roles in drought tolerance. These genes repeatedly appeared in the MQTL regions. The peroxide detoxifying system—antioxidant defense to detoxify ROS—is important for plant stress tolerance, in which ascorbate peroxidase (APX) is a key enzyme for scavenging hydrogen peroxide (H2O2) in chloroplasts (Caverzan et al. 2012; Janda et al. 2019). These genes were found in MQTL2B.1, MQTL2D.1, MQTL5A.3, and MQTL7A.6. Other genes occasionally found in the MQTL regions, such as fatty acid hydroxylase superfamily protein, have been related to stress (Kandel et al. 2005; Wang et al. 2016).

A noteworthy phenomenon is that many genes appeared as gene clusters in the MQTL regions. Gene clusters of functionally related genes are common in eukaryotic genomes (Yi et al. 2007). Dixit et al. (2015) hypothesized that gene clusters and multiple intra-QTL genes underpinned large-effect MQTL. Many enzymatic pathways in plants are encoded in gene clusters (Medema et al. 2015), which often locate closely (a few thousand base pairs away to each other) in a small genome region, encode for similar proteins, and collectively share a generalized function. In such cases, we suggest that the transgenic method using a single gene might not be as effective as MAS, where markers can target a much larger region where all the genes in gene clusters underlying the QTL play important roles. The genotype–phenotype association analysis of the 14 contrasting cultivars found that the most suitable markers for MAS may not necessarily be those closest to or only targeting a single gene. The indel marker on chromosome 2A linked to two genes, TraesCS2A01G136700 and TraesCS2A01G136800, and an indel on chromosome 2B linked to two genes TraesCS2B01G110000 and TraesCS2B01G110100, are such examples, as they showed more variation among the cultivars than those closer to the genes.

Conclusions

The MQTL validated in this study as major genomic regions controlling yield-related traits are MQTL1D.1, MQTL2A.1, MQTL2A.2, MQTL2B.1 MQTL2B.2, MQTL3A.2, MQTL3D.1, MQTL4A.7, MQTL6A.1, MQTL7A.1, and MQTL7D.3 (Table 4; Fig. 1). Of these, MQTL1D.1, MQTL2B.2, MQTL3D.1, and MQTL7D.3 stand out, as they can be used as target regions for improving wheat yield under combined drought and heat stress. MQTL2A.2 is a key region responsible for yield under heat stress, and MQTL3A.2 and MQTL7A.1 are responsible for yield under drought stress and can serve as prime targets for improving stress tolerance in wheat. Future studies, including the development of near-isogenic lines, fine mapping, and functional analyses of these regions, are required to pinpoint the key gene(s) for improving wheat yield and adaptability to stresses.

Table 4 Summary of key MQTL and their associated genes identified in this study
Full size table
Fig. 1
figure 1

Key MQTL regions with major or consistent QTL for yield and yield components on the wheat consensus map

Full size image

Author contribution statement

HL and GY designed and conceived the study. DM, CZ, and SZ helped with the phenotypic and genotypic data acquisition. HL performed data analysis and wrote the manuscript. DM, CZ, SZ, XL, AZ, ZL, YW, and GY critically reviewed the article and provided constructive feedback. All authors are participants of the Global Innovation Linkages Project.