Genetic dissection of flag leaf morphology in wheat (Triticum aestivum L.) under diverse water regimes

Background Morphological traits related to flag leaves are determinant traits influencing plant architecture and yield potential in wheat (Triticum aestivum L.). However, little is known regarding their genetic controls under drought stress. One hundred and twenty F8-derived recombinant inbred lines from a cross between two common wheat cultivars Longjian 19 and Q9086 were developed to identify quantitative trait loci (QTLs) and to dissect the genetic bases underlying flag leaf width, length, area, length to width ratio and basal angle under drought stress and well-watered conditions consistent over four environments. Results A total of 55 additive and 51 pairs of epistatic QTLs were identified on all 21 chromosomes except 6D, among which additive loci were highly concentrated in a few of same or adjacent marker intervals in individual chromosomes. Two specific marker intervals of Xwmc694-Xwmc156 on chromosome 1B and Xbarc1072-Xwmc272 on chromosome 2B were co-located by additive QTLs for four tested traits. Twenty additive loci were repeatedly detected in more than two environments, suggestive of stable A-QTLs. A majority of QTLs involved significant additive and epistatic effects, as well as QTL × environment interactions (QEIs). Of these, 72.7 % of additive QEIs and 80 % of epistatic QEIs were related to drought stress with significant genetic effects decreasing phenotypic values. By contrast, additive and QEIs effects contributed more phenotypic variation than epistatic effects. Conclusions Flag leaf morphology in wheat was predominantly controlled by additive and QEIs effects, where more QEIs effects occurred in drought stress and depressed phenotypic performances. Several QTL clusters indicated tight linkage or pleiotropy in the inheritance of these traits. Twenty stable QTLs for flag leaf morphology are potentially useful for the genetic improvement of drought tolerance in wheat through QTL pyramiding.


Background
Wheat (Triticum aestivum L.), one of the most important foodstuff crops in the world, is grown under a broad range of environmental conditions in terms of water regimes, climatic factors, and soil types. As current changes in global climate have increased variability in precipitation with more frequent episodes of drought [1], wheat production in semiarid and arid regions is increasingly constrained due to erratic drought stresses [2]. In particular, terminal drought occurring during the reproductive phase is responsible for poor grain set and development and finally results in substantial reductions in grain yield [3]. Therefore, the improvement in drought tolerance, as well as grain yield, is of very importance in the selection of wheat cultivars in dryland environments.
Grain yield in cereal crops is due to complex physiological and biochemical processes but is essentially associated with the carbohydrate accumulation process of grain filling, which in turn is attributed to leaf functionalities [4]. By contrast to other leaves in the duration of reproductive phase, flag leaves are the main organ for photosynthesis, providing the major assimilate source required for plant growth and panicle development and also sensing environmental signals for adaptation [4,5]. For example, under favorable conditions and depending on wheat genotype, the wheat flag leaf contributes 45-58 % of photosynthetic performance [6] and 41-43 % of assimilates used in grain filling after flowering [7]. In this regard, key components underlying grain yield in cereal crops are positively correlated with flag leaf size estimated by length (FLL), width (FLW) and area (FLA) [8][9][10][11][12], flag leaf length to width ratio (FLWR) [13] and basal angle of flag leaf (BAFL) [14,15]. Based on this, improvement of flag leaf traits has led to a large increase in grain yield [16]. Under drought conditions, water deficit in cereal crops significantly decreases leaf areas and adjusts the BAFL to avoid excessive transpiration loss [17]. Positive adaptation may also delay leaf senescence induced by drought stress, thus maintaining photosynthetic capacity and a favorable supply of assimilates to the grain for a longer period of time to assure better grain yield [18]. As a result, wheat genotypes with smaller and more erect flag leaves are considered more able to roll their leaves to reduce water loss in response to drought stress than genotypes with lax leaves [19], resulting in higher yields [20]. Qian et al. [21] also found that, in wheat plants exposed to drought stress, FLW, FLL and BAFL during grain-filling were positively correlated with yield component traits, but the correlation coefficients were smaller than those under well-watered conditions [21]. Of course, it is indisputable that reduction in flag leaf area induced by drought stress is per se conductive to limited water use and could also result in lower productivity [22], whereas ideal flag leaf sizes and shapes in wheat are still beneficial for sustaining yield potential in water-deficit environments [15,19]. Therefore, obtaining optimal flag leaf morphology (FLM) could be an important target in breeding wheat for drought tolerance, especially under terminal drought stresses.
To better develop molecular marker-assisted selection and explore novel functional genes for FLM in improving drought tolerance in wheat, it is essential to dissect the molecular genetic basis. This understanding will provide knowledge on how genes/QTLs underlying phenotypic variation are modulated. Much effort has already been exerted to uncover the genetic mechanism for such traits in cereal crops [5,[23][24][25]. Early studies showed that FLM-related traits were under additive control combined with partial dominance and epistasis [14,26], or even predominantly controlled by complex epistatic interactions, dominance, and additive × dominance variation [7]. Furthermore, the phenotypic variation was governed by one gene with at least three distinct alleles [27]. With the recent availability of molecular markers and genetic maps, a large number of quantitative trait loci (QTLs) for FLM-related traits have been identified in wheat [15,24,[28][29][30][31][32], rice [10][11][12][33][34][35] and barley [23,25]. Two major QTLs (qFLL1 and qFLW4) for FLL and FLW in rice were fine mapped [11,35], and even some genes controlling FLW are cloned [36]. In wheat, putative QTLs with flexible expressions in various genetic populations and environments have been detected on almost all 21 chromosomes. For example, using a wheat recombinant inbred line (RIL) population, Fan et al. [32] identified 38 additive QTLs (A-QTLs) for FLW, FLL and FLA on 12 chromosomes, explaining 3.96-27.68 % of the phenotypic variance. However, only three A-QTLs were stable across environments [32]. Working on another RIL population, Wu et al. [24] found that just four of 61 A-QTLs were repeatedly expressed in all environments [24]. Isidro et al. [15] detected 30 A-QTLs for BAFL on chromosomes 2A, 2B, 3A, 3B, 4B, 5B and 7A in a double haploid (DH) population, individually accounting for 8.9-37.2 % of the phenotypic variance. That study confirmed that the pattern of QTL expression was dynamic and time-dependent during the ontogeny of BAFL [15]. Recently, one of the major QTL for FLW, QFlw.nau-5A.1, was fine mapped to a 0.2 cM Xwmc492-Xwmc752 interval in the chromosome 5AL 12-0.35-0.57 deletion bin [30], closely linked with Fhb5, a gene for type I Fusarium head blight resistance [29,30,37]. Some important chromosome regions with abundant QTL information for FLMrelated traits overlapped the marker intervals of QTLs associated with yield component traits [8,28,29,31,32]. These findings further confirmed that FLM is quantitatively inherited by ploygenes and is significantly influenced by environmental factors. However, few studies so far have been undertaken to fully dissect the variability in genetic components and QTL × environment interactions (QEIs) under the drought stress.
In this study, a RIL population of 120 F 8 -derived lines grown under drought stressed (DS) and well-watered (WW) regimes in four environments was employed to map QTLs for FLM-related traits FLL, FLW, FLWR, FLA and BAFL. The objectives were to identify A-QTLs and epistatic QTLs (AA-QTLs) underlying components of FLM-related traits and to analyze additive QEIs (A-QEIs) and epistatic QEIs (E-QEIs) of the traits in two water regimes across environments. The findings might provide a better understanding of the genetic mechanisms governing FLM-related traits in wheat under water-limited environments, and should benefit genetic improvement of drought tolerance in wheat by pyramiding favorable QTLs.

Plant materials
A RIL population of 120 F 8 -derived lines was developed from a cross between two Chinese winter wheat varieties, Longjian 19 and Q9086. Longjian 19, released by the Gansu Academy of Agricultural Sciences, Lanzhou, Gansu, is an elite drought-tolerant cultivar widely grown in rainfed areas (300-500 mm annual rainfall) in northwestern China. Q9086, released by Northwest Agriculture & Forestry University, Yangling, Shanxi, is a high-yielding cultivar suitable for cultivation under conditions of sufficient water and high fertility, but is prone to early senescence under terminal drought stress. The two parents differ significantly in several agronomical and physiological traits under terminal drought stress, such as plant height, grain weight and accumulation and remobilization of water soluble carbohydrates in stems [38][39][40].

Field trials
The RIL population and parents were grown at three locations in Gansu province, namely, at Yongdeng (103°1 8' E, 36°42' N, 2140 m above sea level) in 2011 (294.3 mm of annual rainfall, 1879.8 mm of annual evaporation capacity, 6.2°C of average daily temperature) and 2012 (309.6 mm of annual rainfall, 1906.2 mm of annual evaporation capacity, 6.4°C of average daily temperature); at Anning (103°51' E, 36°04' N, 1520 m above sea level) in 2012 (346.5 mm of annual rainfall, 1664.9 mm of annual evaporation capacity, 8.1°C of average daily temperature), and at Yuzhong (104°07' E, 35°51' N, 1900 m above sea level) in 2013 (328.4 mm of annual rainfall, 1495.8 mm of annual evaporation capacity, 7.2°C of average daily temperature). The environments were named E1, E2, E3 and E4, respectively. The experimental field in each year was divided into DS and WW sections. The DS treatment was equivalent to rainfed conditions with rainfall of 95.8, 98.6, 113.2 and 101.5 mm in E1 to E4, respectively, during the growing season (from early October in the sowing year to late June in harvesting year). The WW treatment involved irrigation with 750 m 3 ha -1 water supply at each of preoverwintering, jointing, and flowering stages, respectively. The field designs were randomized complete blocks with three replications. Each plot was 2 m long with six rows spaced 20 cm apart with approximately 160 plants per row. Nutrition supplied to all treatments was 180 kg ha -1 N, 20 kg ha -1 P 2 O 5 and 75 kg ha -1 K 2 O only at sowing. Other aspects of field management followed the local practices.
Five FLM traits, FLL, FLW, FLWR, FLA and BAFL, were evaluated in this study. For each plot, the main shoots from 10 plants in the centre of each row were randomly selected to measure FLL, FLW and BAFL at the milky ripe stage (Feeks 11.1) and to investigate the plant height (PH), spikelet number (SN), kernel number (KN), kernel weight per spike (KW) of main shoots and yield per plant (YP) at the kernel ripe stage (Feeks 11.4). The FLL and FLW measurements were made at the longest and widest parts of the flag leaf using a ruler. The BAFL from the peduncle to the midrib of the flag leaf surface was determined with a protractor. FLA and FLWR were calculated as follows: FLA = FLL × FLW × 0.75 and FLWR = FLL/FLW. Agronomic traits were determined by conventional methods. Trait means of 10 samples from each plot were used in the data analysis based on three replications.

Data analysis
Basic statistics and Pearson's correlation analysis were performed on the phenotypic data from each water environment. Analysis of variance (ANOVA) was employed to evaluate the total and residual variances among RIL progenies for each FLM-related trait. Broadsense heritability (h 2 B ) was estimated for each trait using ANOVA analysis and method proposed by Toker [41]. All analyses were performed using the SPSS version 18.0 statistical package and P values less than 0.05 were significant.
A genetic linkage map of 21 chromosomes, consisting of 524 simple sequence repeats (SSR) marker loci, was previously made for the RIL population [38,39]. The map spanned 2266.7 cM with an average distance of 4.3 cM between adjacent markers and average 24.9 SSR markers per each chromosome. To dissect the quantitative genetic basis of FLM-related traits in the RIL population, the phenotypic data for the trait under both water regimes (DS and WW) as a set of variants in each environment were subjected to QTL analysis using the software QTLMapper version 1.0 set for composite interval mapping of a mixed linear model [42]. The genetic model divided genetic effects into additive effects (A), epistatic effects (AA), and QEIs (AE and AAE) effects. QTLs with genetic effects indicated that genes in these genomic regions were expressed in the same way across environments. QTLs with AE and AAE effects suggested that gene expression at those loci was environmentally dependent [42]. The closest marker to each local log odds (LOD) peak (putative QTL) was used as a cofactor to control the genetic background while testing at a position of the genome. The threshold LOD score to declare the presence of a QTL was 2.50, and the significance level was P < 0.005 for identifying additive and epistatic effects of QTLs and QEIs effects. If a QTL for one trait was detected repeatedly in two or more environments, it was considered a stable QTL. The QTL nomenclature was according to the rule "QTL+ trait + lab designation + chromosome".

Phenotypic variations
The phenotypic means for five FLM-related traits from the RIL population and parents, along with basic statistics under DS and WW conditions in four environments, are summarized in Table 1. Except for FLWR, the parents Longjian 19 and Q9086 differed significantly in the measured traits. Phenotypic means of Q9086 for FLL, FLW, FLA and BAFL were much higher than those of Longjian 19. Across all treatments, the means of the RIL population were intermediate between those of the two parents, showing wide phenotypic variability. The corresponding coefficients of variation (CV) ranged from 13.51 to 38.25 % in DS and from 8.28 to 24.44 % in the WW conditions. Some lines had more extreme values than the parents, showing substantial transgressive segregation. All skewness and kurtosis values were less than 1.0 in all treatments, indicative of continuous variation and a quantitative genetic basis.
Results of ANOVA showed that the variances for phenotypic values in the RIL population reached the 0.05 or 0.01 significance levels, except for the interaction variances for both water regime × genotype and environment × water regime × genotype (  (Table 2). Hence water environments made a significant impact on phenotypic variation and heritability of FLM-related traits.

Additive QTLs and water environmental interactions
A total of 55 A-QTLs governing FLM-related traits in environments E1 to E4 were mapped on chromosomes 1B, 2A, 2B, 3A, 4A, 4D, 5A, 5B, 6A, 6B and 7A, individually explaining 0.68 to 12.92 % of the phenotypic variation (     and FLA were derived from Q9086. This implied that Q9086 contributed more genes regulating FLL and FLA in the RIL progenies, whereas Longjian 19 provided more genes controlling FLW and BAFL. The majority of A-QTLs (35 of 55, or 63.6 %) for FLM-related traits were identified in one environment. Among them, 11 loci (7 for FLWR) showed no water environmental interactions. This suggested that the A-QTLs for FLWR expressed only in one environment were more insensitive to water treatments than those for other traits. However, the other 24 loci also showed significant A-QEIs with water environments. Of these, 15 A-QEIs were associated with DS and their AE effects decreased phenotypic values, whereas 9 A-QEIs were involved with WW and their AE effects increased phenotypic values. The A-QEIs in both groups individually explained from 1.37 to 10.19 % and from 2.41 to 6.97 % of the phenotypic variation, respectively. This indicated that the capacity of DS to influence phenotypic variation in the traits was stronger than those of WW. In particular, Qfll.acs-5A.1 made a greater contribution to phenotypic variation in FLL not only by A effect (9.78 %) but also by AE effect (10.19 %), whereas the A and AE actions of other loci for corresponding traits were considerably lower.
Twenty of 55 (36.4 %) A-QTLs were detected in more than two environments, suggestive of stability. All of these loci were involved in A-QEIs with water environments to different extents, individually accounting for phenotypic variation of 1.83 to 7.90 % by A effects and 2.43 to 9.83 % by AE effects. Each of these loci even showed the same direction of A or AE effects in responding to different environments. Two loci, Qflwr.acs-3A.

Chromosomal distribution and genetic contributions of detectable QTLs
In this study, 55 significant A-QTLs for FLM-related traits in the RIL population were mapped on 11 chromosomes. They were more frequently located on chromosomes 1B, 2A, 2B, 3A, 4D, 5A and 5B (more than 5 A-QTLs). The highest number (9 or 16.4 %) was detected on chromosome 3A, whereas the lowest number (1 or 1.8 %) was on chromosome 4A. Chromosomes 2B and 3A possessed A-QTLs for all tested traits. An interesting feature was the highly concentrated distribution of A-QTLs in a few chromosomal regions and the existence of QTL hotspots, namely, the chromosomal regions shared by multiple QTL (Table 5, Fig. 1). For example, several A-QTLs underlying FLL, FLWR, FLA and BAFL were detected within the marker interval Xwmc694-Xwmc156 on chromosome 1B. Similarly, A-QTLs for FLL, FLW, FLA and BAFL were co-located in the marker interval of Xbarc1072-Xwmc272 on chromosome 2B. The other ten specific intervals, for example, Xmag2150-Xgwm339 on 2A, Xwmc695-Xgwm162, Xgwm162-Xmag3082 and Xwmc505-Xwmc343 on 3A, and so on, harbored A-QTLs controlling two to three traits. On the other hand, QTL clustering also occurred in several neighboring marker intervals. For example, the region flanking markers from Xwmc522 to Xgwm249 on chromosome 2A was shared by A-QTLs associated with FLW and FLWR. A-QTLs for all five traits shared neighboring intervals Xbarc1072 to Xksum248 on chromosome 2B and Xwmc695 to Xmag3082 on chromosome 3A. The other clustered A-QTLs involving two to four traits were mapped in five adjacent marker intervals Xbarc92 to Xgdm61 on chromosome 4D, Xgwm205 to Xmag694 on chromosome 5A, Xbarc164 to Xwmc376 on chromosome 5B, Xwmc341 to Xmag2276 on chromosome 6B, and Xwmc139 to Xgwm63 on chromosome 7A. This indicated that specific hotspot regions might carry genes controlling traits contributing to FLM.
The mean genetic component effects and phenotypic variations explaining genetic effects for all tested traits across environments E1 to E4 are given in Fig. 2. Both above mean values significantly differed from genetic components for each trait. Genetic effects generally acted to decrease phenotypic values. In this case, the highest values of genetic effects were highlighted in AE and/or AAE, although the A effects for FLWR, FLA and BAFL were also important. Thus based on effect magnitudes of genetic component effects, it could be perceived that genetic regulation of FLM was more ascribable to QEIs effects caused by DS, rather than additive and epistatic effects. In addition, the means of phenotypic variations explained by genetic effects also further illustrated the characteristics of QTL expressions for tested traits. By contrast, the contribution rates of phenotypic

Phenotypic variations in response to drought stress
The flag leaf is the most important source organ for synthesis and output of assimilates during the reproductive stage, and is responsible for regulating final plant growth and yield formation in cereal crops [4,5]. The morphological attributes of flag leaves, such as FLL, FLW, FLA and BAFL, are therefore critical factors in determining a desirable plant type [43], and also sense environmental signals for adaptation [4,5]. In this study, ANOVA clearly showed that phenotypic means of tested traits in a RIL population were more affected by both water regime and environment factors. The phenotypic means under the DS were significantly lower than those under the WW conditions (Tables 1 and 2). These indicated that flag leaves remained smaller sizes and erect postures when adapting to DS, in agreement with previous studies [19][20][21]. Obviously, reduced flag leaf size should be beneficial in limiting excessive water losses by transpiration [17], while maintaining assimilate synthesis and transport to grain as efficiently as possible [18].
Most traits related to FLM were positively correlated with each other in both water regimes, whereas correlation coefficients under DS (r = 0.31 * to 0.93 ** ) were generally higher those under WW conditions (r = 0.29 * to 0.81 ** ) ( Table 3). This suggested that all components related to FLM under DS might be more effectively coordinated by phenotypic reduction to withstand adverse conditions. By contrast, FLL appeared to be the main contributor to FLA and also influenced BAFL to some extent, as evidenced by higher correlation with each other. However, when working with a wheat RIL population (Kenong 9204 × Jing 411) under nitrogen stress, Fan et al. [32] found that the positive correlation between FLW and FLA (0.84 ** ) was stronger than that between FLL and FLA (0.57 ** ), suggesting a predominant contribution of FLW relative to FLA [32]. This indicated that water and nitrogen supply could affect flag leaf size and shape in different ways. Of course, this possibility cannot be excluded from the differences in the genetic backgrounds of the two populations. FLL and FLA showed higher and more significant positive correlations with PH, KW and YP than with other traits under both water regimes across environments ( Table 4), indicating that FLL and FLA contributed more to PH, KW and YP.

Genetic components and QTL-by-environment interactions
Although a wealth of information from previous studies considerably improved our understanding of the morphophysiological functions of flag leaves [4,5], as well as applications in wheat breeding programs [4,19], few studies considered the genetic basis of FLM-related traits under water-deficit conditions at the molecular level   main-effect QTLs and affects the accuracy of isolating main-effect QTLs [44]. In this regard, the importance of epistasis and QEIs in determining the quantitative genetic basis of other traits in wheat, such as yield-related and physiological traits, has been documented [45,46]. These studies showed that the actions of QTLs with additive effects were not completely independent, but varied depending upon their interactions with other loci and/or with environmental factors. Our study also confirmed that phenotypic variation of all traits was controlled by A and AA effects, as well as QEIs (AE and AAE) effects (Tables 5  and 6). As genetic main effects, A and AA effects were largely responsible for the genetic basis of FLM, but the cumulative contributions from AA effects were significantly lower than those from A effects for all tested traits (Fig. 2). The results were consistent with the previous findings involved in yield-associated traits in other cereal crops [47,48]. It was interpreted that low contributions to phenotypic variance explained by AA effects were due to large numbers of AA-QTLs with minor genetic effects [46]. On the other hand, we concluded that the phenotypic variation in FLM was predominantly controlled by additive and QEIs effects, depending on exclusive genetic contributions. Genotype × environment interaction is critical in determining the adaptation and fitness of genotypes in adverse environments [47], resulting in phenotypic variation referred to as phenotypic plasticity [49]. The phenotypic plasticity of quantitative traits arises in nature from interactions between QTLs and environments at the molecular level [50]. Numerous cases of such QEIs for agronomic and physiological traits showed that QTL expressions varied across environments [45,46,49,50]. In the present study, A-QEIs and E-QEIs for all five traits were also identified. For example, 80 % (44 of 55) of A-QTLs and 68.6 % (35 of 51) of AA-QTLs participated in QEIs, of which 72.7 % A-QEIs and 80 % E-QEIs were associated with DS, individually explaining 1.37 to 10.19 % and 1.93 to 6.02 % of the phenotypic variation, respectively (Tables 5  and 6). This indicated that DS influenced the phenotypic variation in these traits more strongly than WW conditions. Moreover, these QEIs effects under DS decreased phenotypic values of FLM. This also seemed to explain why FLM-related traits showed higher coefficients of variation (13.51 to 38.25 %) and lower phenotypic values under DS, compared to those under WW conditions ( Table 1). The present study also suggested that QTLs for FLM-related traits could have different expression patterns responsive to different environments, because a majority of them were detected in single environment. Similar results were obtained for other quantitative traits such as grain yield and related traits in rice [51] and wheat [45,46]. Li et al. [51] suggested that this phenomenon might occur in any of the following situations: (1) a QTL expressed in one environment but not in another, as reflected by inconsistent detection of QTL across environments; (2) a QTL expressed strongly in one environment but weakly in another, as indicated by variation in its effects across environments; and (3) a QTL expressed very differently and with opposite effects in different environments [51].

Chromosomal location and pleiotropy of QTLs
In accord with previous studies [15,24,29,30,32], the distributions of A-QTLs controlling FLM-related traits in the present work behaved in a highly uneven way (Fig. 1). They were more frequently located on chromosomes 1B, 2A, 2B, 3A, 4D, 5A and 5B (more than 5 A-QTLs for each chromosome). The highest number of QTLs (9 or 16.4 %) was detected on chromosome 3A, whereas the lowest (1 or 1.8 %) was on chromosome 4A. Chromosomes 2B and 3A possessed A-QTLs for all tested traits. Similar results were also observed by Wu et al. [24]. This indicated that these important chromosomes carried large numbers of genes controlling FLM. Furthermore, QTLs for FLM-related traits were likewise highly concentrated in a few chromosomal regions on the same chromosomes (Fig. 1). These QTL clusters were generally involved in correlated traits with higher correlation coefficients between traits (Table 3), similar to the previous studies [11,12,24,33]. It was hypothesized that the inheritance of component traits of FLM could be highly correlated with each other, and even with yield-related traits, because many specific or adjacent intervals with QTLs for traits associated with FLM share locations with QTLs for yield-related traits in wheat [28,29,31,32] and rice [5,11,12]. Using the same RIL population in our previous studies, some reported QTLs for PH [38] and thousand-grain weight (TGW) [39] were co-located or adjacent the locations of the present QTLs for FLM-related traits in particular marker intervals on chromosomes 2A, 2B, 5B and 7A. Moreover, some reported QTLs for heading date were shared the same marker interval Xbarc151-Xwmc630 on chromosome 5A with stable QTL for FLW, Xbarc109-Xwmc376 on chromosome 5B with QTLs for FLL, FLW and BAFL [52], and Xgwm408-Xwmc75 on chromosome 5B with QTL for BAFL [52,53]. However, it remains a puzzling question whether these clustered QTLs represent close linkages of multiple genes affecting different traits or have pleiotropic effects of regulatory genes that affect the related traits [12]. One particular interpretation is that the nature of QTL clusters in particular chromosomal regions might be resolved by increasing population size, or by using overlapping substitution lines. As a result, most of QTL clusters for correlated quantitative traits were proved to inherit as a linkage way, instead of pleiotropy [12].