Meta-QTLs, Ortho-MetaQTLs and Candidate Genes for Grain yield and Associated Traits in Wheat (Triticum aestivum L.)

The present study involved meta-QTL analysis based on 8,998 QTLs, including 2,852 major QTLs for grain yield (GY) and its following ten component/related traits: (i) grain weight (GWei), (ii) grain morphology related traits (GMRTs), (iii) grain number (GN), (iv) spikes related traits (SRTs), (v) plant height (PH), (vi) tiller number (TN), (vii) harvest index (HI), (viii) biomass yield (BY), (ix) days to heading/owering and maturity (DTH/F/M) and (x) grain lling duration (GFD). The QTLs used for this study were retrieved from 230 reports involving 190 mapping populations (1999–2020), which also included 19 studies involving durum wheat. As many as 141 meta-QTLs were obtained with an average condence interval of 1.37 cM (reduced 8.87 fold), the average interval in the original QTL being > 12.15 cM. As many as 63 MQTLs, each based on at least 10 original QTLs were considered to be the most stable and robust with thirteen identied as breeder’s meta-QTL. Meta-QTLs (MQTLs) were also utilized for identication of as many as 1,202 candidate genes (CGs), which also included 18 known genes. Based on a comparative genomics strategy, a total of 50 wheat homologues of 35 rice, barley and maize yield-related genes were also detected in these MQTL regions. Moreover, taking the advantage of synteny, a total of 24 ortho-MQTLs were detected at co-linear regions between wheat with barley, rice and maize. The present study is the most comprehensive till date, and rst of its kind in providing stable and robust MQTLs and ortho-MQTLs, thus providing useful information for future basic studies and for marker-assisted breeding for yield and its component traits in wheat.


Introduction
Wheat provides ~20% of protein and calories in human diets worldwide. Due to continuous efforts of plant breeders, the global annual production of wheat has been increasing continuously during the last ve decades and reached 765 million metric tons in the year 2020. It has also been estimated that the annual production of wheat should increase by 50% by the year 2050 (https://www.openaccessgovernment.org/demand-for-wheat/83189/) to meet the demand of the growing population, which is expected to reach 9 to 10 billion in the year 2050. However, the rate of global annual wheat production declined in recent years to 0.9% from ~3% in the early years of the two postgreen revolutions (Shiferaw et al. 2013;Yadav et al. 2019). Therefore, further improvement in yield potential is needed to improve and sustain the required annual growth rate of ~2%. This will be possible only through a further detailed understanding of the genetic architecture of grain yield associated with the use of molecular approaches with conventional wheat breeding (Gupta et al. 2020).
Grain yield is widely known to be a complex quantitative trait, which is controlled by a large number of QTLs/genes involved in plant growth and development, contributing towards yield either directly or indirectly. Yield is also known to be in uenced by the environment, thus making its visual selection rather di cult. Major yield contributing traits include grain number, grain weight, grain morphology-related traits, tiller number, spike-related traits, harvest index, plant height, heading date, etc. (Gupta et al. 2020;Hu et al. 2020). Therefore, these traits are continuously targeted in wheat breeding programmes for developing novel high-yielding varieties (Zhou et al. 2007). Since these yield-related traits are each controlled by a number of quantitative trait loci (QTLs), the markers associated with these QTLs can be exploited in molecular breeding using marker-assisted selection (MAS).
During the last two decades, a large number of QTLs for grain yield and its component traits have been identi ed (Table S1). However, the practical use of these QTLs seems to be minimal in actual wheat breeding. In order to make use of these QTLs in wheat breeding, and in basic research, a detailed analysis of these loci needs to be undertaken. Meta-analysis of the QTLs is an approach that has been shown to provide more robust and reliable QTLs (Go net and Gerber, 2000;Salvi and Tuberosa, 2015). Since, Meta-QTL analysis also allows identi cation of "QTL hotspots" (if any), we undertook the task of nding such regions for use by the wheat breeders.
Efforts have also been made to improve algorithms used for meta-QTL analysis (Arcade et al. 2004;de Oliveira et al. 2014;Veyrieras et al. 2007). Most of these algorithms are available in the software 'Biomercator' (Sosnowski et al. 2012). Using the maximum likelihood method, these algorithms test the number of possible meta-loci (MQTLs), which result from the projection of initial/original QTLs (taken from different studies) on a consensus genetic map. Criteria are also available to nd out the best model for con rming the actual number of these MQTLs (Go net and Gerber, 2000;Sosnowski et al. 2012).
Wheat MQTLs have already been identi ed for several traits including the following: ear emergence (Hanocq et al. 2007), (ii) pre-harvest sprouting tolerance (Tyagi and Gupta, 2012), (iii) Fusarium head blight resistance (Venske et al. 2019), (iv) heat stress tolerance (Acuña-Galindo et al. 2015;Kumar et al. 2020) and (v) yield and quality-related traits (Liu et al. 2020;Quraishi et al. 2017;Soriano et al. 2017;Zhang et al. 2010). However, new QTLs are being regularly added to the database for QTLs. Most of these new QTLs have been identi ed using SNP arrays that have recently become available (Hu et al. 2020;Kuang et al. 2020;Lin et al. 2020;Xin et al. 2020;Yu et al. 2020). Keeping this in view, during the present study, meta-QTL analysis was conducted on grain yield (GY) and a number of its contributing traits listed above, using the QTL data from 230 studies published during 1999 to 2020 utilizing 190 different mapping populations (Table S1).
The MQTLs identi ed during the present study were also used to identify candidate genes (CGs). Ortho-MQTLs across gramineae family were also detected using synteny among cereals (Mayer et al. 2011;Hirsch et al. 2014;Kumar et al. 2009). Genomic regions associated with MQTLs that are homologous to known genes for yield in other cereals (rice, barley, and maize) were also identi ed. We believe that this work should prove useful not only for genomic selection based molecular breeding, but also for basic research on structural genes and regulatory elements involved in grain yield and associated traits not only in wheat, but also in other cereals.

Collection of QTL Data for Yield and Associated Traits in Wheat
The literature related to QTL mapping of grain yield and its component traits was collected from PubMed (http://www.ncbi.nlm.nih.gov/pubmed) and Google Scholar (https://scholar.google.com/) using appropriate keywords. A total of 8,998 QTLs were available from 230 studies, which involved 190 mapping populations. The size of the mapping populations ranged from 32 to 547 DH/RIL lines; these mapping populations also included 26 F 2 /BC populations. As many as 19 studies involved durum wheat, mainly published during 2015-2020. A summary of these QTLs studies is available in Table S1. For each QTL, the following data was retrieved: (1) QTL name, wherever available, (2) anking markers or closely linked marker, (3) peak position and con dence interval, (4) type and size of the mapping population, (5) LOD score, and (6) phenotypic variance explained (PVE) or R 2 value. Wherever, information about peak position was missing, mid-point between the two anking markers was treated as the peak. Secondly, where LOD score was not available, the same was calculated using the available test statistic. In case, when actual LOD score for an individual QTL was not available, a threshold LOD score of 3.0 was chosen for the study.
If names of QTLs were not available, names were assigned following the standard nomenclature (letter "Q" followed by abbreviated name of the trait, the institute involved and the chromosome). Different QTLs on the same chromosome were distinguished by using numerical identi ers following the chromosome. All QTLs were accommodated in the following traits: (i) 'grain weight' (GWei), recorded as thousand-grain weight, 50-grains weight, mean grain weight, hundred-grain weight, single grain weight, grain weight per plant, and test weight, (ii) 'grain number' (GN) recorded as average grain number per spike, grain number per spike, grain number per square meter, grain number per spikelet, grains per spikelet, and grains per fertile spikelet, etc; (iii) 'grain morphology related traits' (GMRTs) recorded as grain length, grain width, grain length-width ratio, grain thickness, grain thickness-length ratio, grain area, and grain diameter, grain volume weight, etc.; (iv) 'spike related traits' (SRTs) recorded as spike length, spikes per plant, spikes per square meter, spike compactness, spike formation rate, spike layer uniformity, basal sterile spikelet number, top sterile spikelet number, fertile oret per spike, and spikelets per spike, etc. (v) 'biomass yield' (BY) recorded as total biomass, tiller biomass, and plant biomass; (vi) 'tiller number' (TN) recorded as effective tiller number, tiller number per plant, and tiller number per square meter. Other remaining traits were treated individually by the available name of the trait (e.g., heading, days to owering, days to maturity, grain lling duration, earliness per se, and plant height).

Construction of Consensus Map
A consensus map (described as Wheat_Consensus_2021) was developed using the following integrated genetic maps: (1)  other independent studies were also included for developing the consensus map. The R package LPMerge was employed for the construction of the consensus map (Endelman and Plomion, 2014).

QTL Projection and Meta-QTL analysis
Only 7,275 QTLs from 8,998 available QTLs had the information required for QTL projection. Although, from these 7,275 QTLs, only 2,852 major QTLs (explaining ≥ 10 percent of phenotypic variance for the target trait), each with the required information (estimated CIs, peak positions, original LOD scores, and R 2 or PVE value) could be used for projection on the consensus map using BioMercator V4.2 (Sosnowski et al. 2012) (Table S1).
To ensure consistency in different studies, the con dence interval (95%) was estimated for each locus, through different population-speci c equations: for RILs: CI = 163/ (population size x R 2 ); for F 2 and backcross populations: CI = 530/(population size x R 2 ); and for DH: CI = 287/(population size x R 2 ) (Darvasi and Soller, 1997;Venske et al. 2019). Following projection, meta-analysis was performed, for each chromosome individually, via the Veyrieras two-step algorithm available in the software. The Akaike (AIC) statistics were used to nd the best QTL model for ascertaining the actual number of MQTLs on each wheat chromosome. The statistical procedures and algorithms furnished in this software have been well-described elsewhere (Sosnowski et al. 2012).

Candidate Gene mining within MQTLs and Expression Analysis
For identi cation of CGs, the sequences for anking markers of individual MQTLs were retrieved from GrainGenes (https://wheat.pw.usda.gov/GG3) and CerealsDB (https://www.cerealsdb.uk.net/cerealgenomics/CerealsDB/indexNEW.php) and used for BLASTN searches against wheat reference genome sequence available in the EnsemblPlants (http://plants.ensembl.org/index.html). Physical positions of the GBS-SNPs were directly obtained from URGI database (https://wheat-urgi.versailles.inra.fr/). Among the MQTLs obtained, those with a CI of ≤ 2Mb were directly used for identi cation of the candidate genes (CGs). In the remaining cases with longer CI, a 2Mb genomic region (including 1 MB region on either side of the MQTL peak) was examined for possible candidate genes. Gene models present within the original or estimated physical regions were retrieved using 'BioMart' of EnsemblPlant database. The function annotations of the identi ed gene models were explored to nd the best candidate genes within each MQTL.
For the in-silico transcriptional analysis, the identi ed putative candidate genes (CGs) were analyzed via the 'Wheat Expression Browser-expVIP' (expression Visualization and Integration Platform) Wagner and co-workers (2013), only CGs showing at least 2 transcripts per million (TPM) expression were considered in this study. Heat maps for expression data were constructed using the same online resource 'expVIP'.
CGs were further examined using information from cloned, and characterized wheat genes for the traits of interest. Nucleotide sequences of these collected genes were obtained from the NCBI database (https://www.ncbi.nlm.nih.gov/) using accession IDs given in the respective published studies and then BLASTN searches were carried out against the genomic database (available in EnsemblPlants) of wheat to nd the physical positions of these genes in the genome.

MQTL Regions homologous to known genes in other cereals
Information on rice, barley, and maize genes associated with grain yield and related traits was collected from the literature and used for retrieval of the corresponding protein sequences for the identi cation of homologous MQTL genomic regions. Amino acid sequences for relevant genes were retrieved from the NCBI (https://www.ncbi.nlm.nih.gov/) and used for BLASTP search to identify the wheat protein (available in EnsemblPlants) at an E-value of <10−10, with 60% coverage, and >60% identity. Physical positions of the corresponding genes and wheat MQTLs were then compared to detect the MQTL regions homologous to known genes in other cereals.

QTLs Associated with Grain Yield and Related Traits in Wheat
The trait-wise distribution of total QTLs, along with QTLs having information required for meta QTL analysis and major QTLs associated with different yield traits is presented in Table 1. The distribution of these QTLs on different wheat sub-genomes is shown in Fig. 1. Among the three sub-genomes, the subgenome A carried 3,457 (38.42%) QTLs, sub-genome B carried 3,566 (39.63%) QTLs and the sub-genome D carried only 1,975 (21.94%) QTLs.

Construction Of The Consensus Map
The integrated consensus map contained 2,33,856 markers, which included a variety of markers including the following types: SNPs, DArT, SSR, AFLP, RAPD, STS, EST-SSR, SRAP, ISSR and KASP markers. Following important genes are also included on this consensus maps: Vrn, Ppd, Rht, and Glu loci ( Table S2). The total length of the consensus map is 11,638.76 cM; the length of individual chromosome ranged from 281.26 cM (4D) to 763.08 cM (4A). The average number of markers carried by a single chromosome was 11,136 (Table S3). The marker densities for individual linkage groups ranged from 12.76 to 48.27 markers per cM for A and B sub-genomes, and 7.74 to 18.13 markers for D sub-genome.

Ortho-mqtls In Barley, Rice And Maize
The next question, that is natural outcome of MQTL study is what will be the status of other crops, will they be harbouring similar MQTLs in their genomic regions. The study was further advanced to search such orthologues in Gramineae family. For identi cation of ortho-MQTLs, 27 stable and robust wheat MQTLs were selected, each, based on > 20 QTLs. As many as 24 corresponding MQTLs (so called ortho-MQTLs) of these wheat MQTLs were identi ed in rice, barley and maize genomes; ortho-MQTLs for three wheat MQTLs (MQTL1A.5, MQTL2A.2, and MQTL3A.4) were not identi ed on any syntenic chromosome of the studied species; the details are as follows: (i) ve ortho-MQTLs mainly associated with traits GY, GWei, and PH in wheat and maize, (ii) ortho-MQTLs of 11 wheat MQTLs in syntenic regions of rice and maize, and (iii) two ortho-MQTLs (ortho-MQTL2D.5 and ortho-MQTL5A.3) in syntenic regions of maize and barley. Among these, 6 ortho-MQTLs (ortho-MQTL2B.3, ortho-MQTL4A.3; ortho-MQTL4B.1; ortho-MQTL4B.5, ortho-MQTL7A.3, and ortho-MQTL7B.1) involved MQTLs in all the four crops (Table 3).
Conclusively, ortho-MQTLs were detected on ve barley chromosomes 2H, 4H, 5H, and 7H; each of which comprised only one previously detected MQTLs. In rice, ortho-MQTLs were identi ed on all the chromosomes except chromosomes 1, 9, and 10, and underlying MQTLs ranged from 1 (on chromosome 2) to 8 (on chromosome 4) MQTLs. In maize, ortho-MQTLs were detected on all the ten chromosomes, number of MQTLs within the ortho-MQTL regions ranged from 1 to 16 MQTLs. The remaining information about ortho-MQTLs identi ed between wheat with barley, rice, and maize are given in Table 3 and the traits associated with these ortho-MQTLs and underlying orthologous genes along with their functional descriptions are presented in Table S5.

Candidate Genes For Wheat Mqtls
Gene mining in genomic regions carrying individual MQTL allowed the identi cation of 2,953 gene models; this number was reduced to 2,298, when following classes were removed: (i) duplicated genes from overlapping MQTLs; (ii) genes with no information available regarding molecular function and gene ontology (GO) terms. The in-silico expression analysis of the above gene models (except the gene models underlying three MQTLs) allowed the identi cation of 1,202 gene models, each showing at least 2 transcripts per million (TPM) expression (highlighted with yellow in Table S6). The gene models showed spatio-temporal gene expression, expressing in different plant tissues such as grains, spikes, leaves, shoots, and roots, etc. at speci c times during development (for instance, see Fig. 3). Detailed information regarding these gene models, hereafter, termed as 'putative CGs' detected in different MQTL regions is provided in Table S6. These 1,202 CGs mainly belonged to the ve major gene classes, including, (i) transcription factors, (ii) genes involved in metabolism and/or signalling of growth regulators-gibberellins, cytokinins and brassinosteroids, (iii) genes regulating cell division and proliferation (iv) oral regulators, and (v) genes involved in regulation of carbohydrate metabolism (Table  S7).

Wheat MQTLs with Homology to Known Genes from Other Cereals
Known genes for yield and its components from other cereals including rice, barley, and maize were also used for the identi cation of putative wheat CGs in MQTL regions; this allowed identi cation of 24 putative wheat CGs of the 48 available rice genes, 3 of the 7 available barley genes and 8 of the available 13 maize genes. Of these 24 rice genes, as many as 12 genes (viz., D2, DEP1, An-1, GW2, GIF1, qGL3, SMG1, OsLG3, OsALMT7, GS9, OsPK2, and FZP) showed more than one homologues, while, remaining twelve genes showed only one homologue each in wheat MQTL regions (Table 4 and S10). No wheat homologues in MQTL regions were available for four barley genes (Ert-m, HvAPETALA2, HvLUX1, and INT-C), while, two barley genes (vrs4 and COM1) each gave more than one wheat homologues and one gene showed a single homologue (Table 4 and S10). For 8 maize genes (FASCIATED EAR2, ramosa2, ZmFrk1, bs1, BIF1, ZmGS3, KNR6 and vt2), 12 wheat homologues were available in different MQTL regions. In some cases, more than one homologues were available within the same MQTL regions, for instance, homologues of rice genes An-1, GIF4, GW2, and OsPK2 were detected in the MQTL2A.2 region, homologues of rice genes D11, GIF1, OsPK2, and FZP rice genes were available in MQTL2D.8 region and homologues of Vrs4 (barley), ramosa2, vt2 (maize) were available in MQTL3A.3 region (Table 4).
Overall, 33 MQTL regions contained 50 wheat homologues involving 35 alien yield genes from other three cereals (rice, barley, and maize). To our knowledge, homologues of only 8 of these 34 alien genes (GW2, GIF1, GS3, DEP1, CKX2, OsSPL14, FZP, and ZmVT2) have already been cloned and characterized in wheat; following are the details of these wheat homologues: TaGW2

Discussion
During the past two decades, starting with the rst QTL studies on yield-related traits published by Araki et al. (1999) and Kato et al. (1999), a large number of of studies have been conducted on QTL mapping for grain yield and its component traits in wheat (Table S1). The studies involving development of MQTLs were largely motivated by the fact that only a small fraction of QTLs identi ed by interval mapping are major QTLs, and majority of QTLs are each associated with a large con dence interval, with anking markers often located away from the QTLs, thus making these QTLs not very useful for plant breeding. Also, QTLs identi ed using one bi-parental population may not be effective for a breeding programme involving other parents, without prior validation, unless the markers are functional markers located within the QTLs. These problems can be largely overcome through development of MQTLs, which are robust with a reduced con dence interval, thus increasing the utility of these MQTLs not only in crop improvement programmes, but also for basic studies involving cloning and characterizing QTLs/genes for the trait of interest.
Meta-QTL analysis has been conducted for a variety of traits in all major crops. In wheat, meta-QTL analysis has been conducted for a number of traits including yield, but the information on MQTLs soon becomes out-of-date. This is largely because, a large number of studies on QTL analysis for yield traits in wheat are regularly conducted, thus creating a need for conducting further studies on MQTLs periodically to obtain improved MQTLs. The present study is one such attempt, conducted to improve upon MQTLs reported so far for grain yield and associated traits in wheat (Bilgrami et al. 2020;Gegas et al. 2010;Gri ths et al. 2009;Gri ths et al. 2012;Liu et al. 2020;Quraishi et al. 2017;Zhang et al. 2010). The maximum number of QTLs used for these earlier meta-QTL studies on yield traits in wheat was 381 for hexaploid wheat (including the QTLs detected under heat and drought stress conditions) and 1,162 (including the QTLs for disease resistance, grain yellow pigment content, grain quality traits, and root architecture-related traits) for durum wheat leading to identi cation of a maximum of 55 MQTLs for hexaploid wheats (Liu et al. 2020;Zhang et al. 2010) and 71 for durum wheats (Maccaferri et al. 2019). In contrast to this, the number of available QTLs that we listed were 8,998, of which 2,852 major QTLs were used for identi cation of as many as 141 MQTLs suggesting that the present study is so far the most comprehensive study for identi cation of MQTLs in wheat.
With the availability of improved algorithms, software, and recently available resources such as high quality genomic and transcriptomic data (Appels et al. 2018;Borrill et al. 2019;Choulet et al. 2014;Clavijo et al. 2017;Pearce et al. 2015;Ramírez-González et al. 2018) the study was more demanding but effective in identi cation of better MQTLs. Hopefully, some of the MQTLs identi ed during the present study will be relatively more robust, since the number of initial QTLs per MQTL could be as high as 71 (MQTL5A.2). As many as 24 MQTLs identi ed during the present study had their genetic positions almost overlapping those occupied by MQTLs reported in two recent studies (Bilgrami et al. 2020;Liu et al. 2020), so that these MQTLs can be used in future studies with a higher level of con dence (Table S4). On a critical evaluation of these 24 MQTLs, we selected 15 MQTLs each involving atleast 10 initial QTLs, which should be tried in molecular breeding and future studies for cloning and characterization of QTLs/genes (Table S4).

Breeder's Mqtls
Following the criteria laid down by Lö er et al. (2009), from a total of 141 MQTLs identi ed during the present study, we selected thirteen most promising MQTLs, and described these MQTLs as "Breeder's MQTLs", each of which had a CI of < 2cM. Also, these MQTLs explained a signi cant proportion of phenotypic variance (ranging from 20.73 to 49.16 %) associated with high LOD values (ranging from 14.05 to 62.67) ( Table 2). For example, one MQTL having high PVE (49.16%) can be called a mega breeder MQTL located on 3D.1 for traits GWei, GMRTs and PH. Therefore, these MQTLs have the potential to serve as most e cient targets for their direct use in MAS for wheat improvement or can be treated as major genes. Another important feature of the present study is the availability of clusters of MQTLs, which included one cluster each on seven different chromosomes (Fig. 2). These clusters of MQTLs may be treated as hotspots and should be utilized for breeding and for future basic research with high level of con dence.

Ortho-mqtls For Cereals
For the present analysis, we took a step ahead for identi cation of Ortho-MQTLs, since these are conserved and may be recommended for use in all cereals. Conserved nature of these orthoMQTLs also suggest that these may represent some regulatory genes (Jin et al. 2015;Quraishi et al. 2011). Keeping this in view, we selected a total of twenty-seven most promising wheat MQTLs to investigate their evolutionary conserved syntenic regions (or so called ortho-MQTLs) reported in similar meta-QTL analysis studies on the yield-related traits in maize, barley, and rice (Table 3). Consequently, 24 ortho-MQTLs were identi ed for yield and associated traits conserved at orthologous positions in the barley, rice and maize. As many as six of these 24 ortho-MQTLs identi ed in the present study were crossspecies in all the four crops, revealing the high level of conservation of wheat with barley, maize and rice. Of the CGs underlying these ortho-MQTLs, precise orthologous gene sets can be considered as direct potential candidates for further homology-based cloning, functional validation or at least as a source of accurate molecular markers such as conserved orthologous set (COS) markers for use in cereal breeding programs.
Among earlier studies on ortho-MQTLs in wheat, conducted using the above cross-genome map based strategy, an ortho-MQTL associated with nitrogen use e ciency was earlier discovered. This ortho-MQTL was common for all cereals including wheat, maize, sorghum, and rice, which enabled characterization of a structurally and functionally important conserved gene 'glutamate synthase' (GoGAT) at orthologous positions in a number of species (Quraishi et al. 2011). In another study, ortho-MQTLs associated with grain iron and zinc were identi ed using rice MQTL. Further dissection of these ortho-MQTL regions led to the identi cation of two maize genes of rice, namely GRMZM2G178190, and GRMZM2G366919. These genes were characterized as natural resistance-associated macrophage protein genes and considered to be the best candidate genes associated with grain iron and zinc in maize (Jin et al. 2015). The ortho-MQTLs identi ed during the present study can also be subjected to similar studies leading to identi cation of orthologs.

Candidate Genes For Mqtls
The QTLome for grain yield and its component traits is also believed to have a signi cant correlation with gene density in the wheat genome (Maccaferri et al. 2019). In the present study, gene mining was performed in all MQTL regions (in ≤ the 2Mb region) and a total of 2298 gene models were identi ed. Of these, 703 gene models/putative CGs produced 2 to 5 TPM expressions (highlighted with yellow in Table   S6), whereas, as many as 489 gene models showed more than 5 TPM expression (highlighted with blue in Table S6) in different plant tissues at different times (spatio-temporal gene expression). As mentioned earlier, these putative CGs mainly belonged to ve major categories of the genes which are known to be involved in controlling the grain yield and associated traits in cereals (Daba et al. 2020;Nadolska-Orczyk et al. 2017) (Table S7).
MQTLs with genes/gene families involved in grain yield and its component traits has also been shown in several studies (Daba et al. 2020;Gautam et al. 2019;Gunupuru et al. 2018;Jia et al. 2020;Li and Wei, 2020;Ma et al. 2017;Nadolska-Orczyk et al. 2017;Niño-González et al. 2019;Sakuma et al. 2018). In the present study, several such genes/gene families with similar functions were detected repeatedly in different MQTL regions, including 114 genes for leucine-rich repeat proteins, 63 genes for serine/threonine-protein kinases, 33 genes for cytochrome P450 proteins and 14 genes each for WD40/YVTN repeat-like containing proteins, UDP-glucuronosyl/UDP-glucosyltransferases, FAD/NAD(P) binding proteins, and E3 ubiquitin ligases, etc (Table S6). Moreover, some genes encoding unpredicted or uncharacterized proteins also showed signi cant expression in different plant tissues (Table S6). These genes deserve further attention, to explore their possible roles in the regulation of yield and its component traits in wheat.
For many MQTL regions, it was also possible to identify several F-box-like domain superfamily, UDPglucuronosyl/UDP-glucosyltransferase, etc. (Table S6). These gene clusters are quite common in plant genomes and are known to encode proteins involved in many enzymatic pathways in plants (Medema et al. 2015;Yi et al. 2007). It has also been shown in these studies that members of a gene cluster can be located in close proximity (only a few thousand base pairs far from each other) in a small genomic region, and also encoding similar products/proteins, thus together sharing a generalized function.
We selected as many as 162 promising putative CGs which had more than 5 TPM expression in different tissues (Table 5). An extensive survey of available literature also shows association of these selected genes with the traits of interest in different plant species. These CGs may be cloned and further characterized and then can be exploited through biotechnological approaches such as transgenesis and gene editing. In a more recent study, it was observed that over-expression of the expansin gene in developing seeds minimizes the trade-off between grain number and grain weight, and ultimately improves the grain yield. Transgenic plants with enhanced expression of the Expansin gene yielded 12.3 % higher grain weight compared to the control, which nally translated into an 11.3% increase in grain yield under eld conditions (Calderini et al. 2020). In the present study, we also identi ed many putative CGs, including genes for Expansin proteins in some MQTL regions (Table 5, Table S6). In future, the targeted transgenic approach using these potential CGs may allow improvement for grain yield in wheat. However, in some cases, where gene clusters regulate the expression of target trait, the transgenic method using a single gene may not be as effective as MAS, where anking markers can target a much larger region encompassing all the genes of a cluster.
MQTL regions were also examined for genes already known to be associated with yield and its component traits; as many as 18 such wheat genes were identi ed. These genes encode a variety of proteins including the following: cell wall invertase, sucrose non-fermenting 1-related protein kinase, squamosa promoter binding protein-like protein, E3 ubiquitin ligase, APETALA2/AP2/ERF transcription factor, cytochrome P450 protein, phosphatidylethanolamine binding protein, and transmembrane protein, etc. Similar proteins/products are also encoded by many other CGs, identi ed in the present study. Therefore, these CGs can be used for further functional analysis. Some of the MQTLs also had anking markers associated with known genes including Vrn, Ppd, and Rht genes, that are widely known to regulate plant phenology, ultimately in uencing the grain yield and other component traits in wheat (Gupta et al. 2020;Kamran et al. 2014).

Wheat homologues of rice, barley and maize genes
During the present study, within the wheat MQTL regions, we also identi ed several wheat homologues of genes that are known to control grain yield and related traits in rice, barley, and maize genes; many of these genes have not yet been cloned and functionally characterized in wheat; these genes include the following: (i) rice genes: An-1, Bsg1, D11, D2, LP, PGL1, qGL3, SMG1, OsOTUB1, OsLG3, OsDHHC1, OsY37, qWS8, OsALMT7, GS9, GSN1, OsPS1-F, and OsPK2; (ii) barley genes: vrs4 and COM1 and (iii) maize genes: FASCIATED EAR2, ramosa2, ZmFrk1, bs1, KNR6, and BIF1 (Table 4, Table S9). Using comparative genomics, orthologs of these genes can be characterized in wheat and their functional markers can be developed and validated. For instance, in a study conducted in 2018, a meta-analysis of QTLs associated with grain weight in tetraploid wheat resulted in the identi cation of one important locus, mQTL-GW-6A on chromosome 6A. Further analysis identi ed and characterized a wheat homolog of the rice gene, OsGRF4 within this MQTL region (Avni et al. 2018). This suggests that integrating an MQTL study with a well-annotated genome can rapidly lead to the detection of CGs underlying traits of interest.
This was an in-silico approach for improvement of our understanding of the genetic architecture of grain yield and its component traits in wheat through identi cation of MQTLs, orthoMQTLs, and candidate genes. The study involved an integration of the available information about QTLs that were identi ed in earlier studies along with genomic and transcriptomic resources of the wheat. As many as 141 MQTLs, each associated with a narrow con dence interval and 1,202 putative CGs were identi ed. Thirteen of these 141 MQTLs regions given as breeder's QTL, we recommend for use in MAS for grain yield improvement in wheat. Based on a comparative genomic approach, several wheat homologs of rice, barley, and maize yield-related genes were also detected in the MQTL regions. The ortho-MQTL analysis demonstrated that MQTLs of yield-related traits appear to be transferable to other cereal crops that that may assist breeding programs in cereals. As many as 162 of 1,202 putative CGs are also recommended for future basic studies including cloning and functional characterization. However, after any in-silico analysis of this type, the in-vivo con rmation and/or validation of any of these loci, speci cally the CGs identi ed, is needed, which can be accomplished via further approaches, such as gene cloning, reverse genetic approaches, gene silencing followed by transcriptomics, proteomics, and metabolomics. We hope that the results of the present study will help in developing a better insight into the genetic architecture and molecular mechanisms underlying grain yield and its component traits in wheat and other cereals.
The information on the molecular markers linked with the MQTLs and CGs occupying the MQTL region may also prove useful in breeding for grain yield improvement in wheat.
Kumar A, Saripalli G, Jan I, Kumar K, Sharma PK, Balyan HS, Gupta PK (2020) Meta-QTL analysis and identi cation of candidate genes for drought tolerance in bread wheat (Triticum aestivum L).  Tables   Table 1 The trait wise distribution of total QTLs, along with QTLs having information required for analysis and major QTLs associated with studied traits   TraesCS7D02G398900 Glycoside hydrolase MQTL7D.5 TraesCS7D02G455900 Small auxin-up RNA TraesCS7D02G456600 Ribosomal protein S13 TraesCS7D02G457400 SANT/Myb protein TraesCS7D02G458200 Papain-like cysteine peptidase Figure 1 Distribution of collected QTLs for grain yield and associated traits on A, B and D sub-genomes of wheat.

Figure 2
Diagrammatic representation of MQTL clusters detected on chromosomes 1A, 1B, 2B, 4B, 5B, 6B and 6D; Only desired parts of the chromosomes are shown for better visualization; different colours correspond to different MQTLs on each chromosome.