Genetic Dissection of the Mixing Properties of Wheat Flour (Triticum aestivum L.) Using Unconditional and Conditional QTL Mapping

Wheat (Triticum aestivum L.) flour mixing properties are essential quality parameters in the dough development process. Limited research on superior alleles for mixing properties has restricted their molecular improvement, and other factors related to the complex traits have been ignored. A molecular map of 9576 polymorphic markers in the RIL population (F8:9) (Shannong01-35/Gaocheng9411) was constructed to evaluate mixing property effects in three environments. The parents were selected with markedly distinct high-molecular-weight glutenin subunits (HMW-GS). This study not only evaluated mixing properties using conventional unconditional QTL mapping but also evaluated the relationships between protein-related traits using conditional QTL mapping. The analyses identified most additive QTLs for major mixing properties on chromosomes 1A, 1B, and 1D. Two major loci (1A.1-15 and 1D-1) associated with mixing properties have confirmed the important contributions of Glu-A1 and Glu-D1 to wheat quality at the QTL level, which were mainly affected by the gluten index. Another important locus, 1B.1-24 (associated with midline peak value and midline peak width, with high phenotypic variations explained), might represent a new variation distinct from Glu-B1. The favored alleles came from Gaocheng9411. Several mixing properties shared the same QTLs (1B.1-6 and 1A.1-15), indicating tight linkage or pleiotropism. Genotype-by-environment (G×E) interactions were also investigated in the present study. The QTL results in our study may improve our understanding of the genetic interrelationships between mixing properties and protein-related traits.

To date, little information on QTL mapping for mixing properties is available, and larger differences among previous reports were mostly caused by genetic populations or maps. Conventional QTL studies do not actually evaluate closely related traits and the genetic contribution to a single trait. Protein content and quality are strongly impacted by mixing properties. Zhu and Keightley (1995) proposed the conditional genetic analysis method [8]. It could study the contributions of single trait to complex traits and the genetic relationships among more than two traits. This method has been widely used to investigate the genetic basis of complex traits in many crops [9][10][11][12]. In wheat, conditional QTL analyses have also been conducted on plant height [13], kernel weight [14], and protein content [15]. At present, few conditional QTL analyses have evaluated the genetic relationships between mixing properties and protein-related traits.
In the present study, both unconditional and conditional QTL mapping were conducted to investigate the underlying genetic basis of mixing properties and to dissect the genetic relationships between these protein-related traits at the QTL level. The results could contribute to marker-assisted breeding for wheat quality, especially for mixing properties.

Field trials
The RIL population was harvested at Tai'an (116°36E, 36°57N) in 2009 (E1), 2010 (E2) and Suzhou (116°58E, 33°38N) in 2010 (E3). The materials were planted in a randomized complete block design with two replicates in each environment. Every pot had three rows, and each row was two meters long with a row-to-row distance of 21 centimeters. Crop management was performed according to local practices. The average value of the mixing properties in the above environments was also calculated (AE).

Measurement of mixing properties and protein-related traits
Harvested seeds were grinded into powder by a Buhler experimental mill (Buhler, Buhler-Miag Co., Germany). The flours were then sealed and kept at 8-10℃ for analysis. Mixing properties were determined by Mixograph (National Mfg. Co., Lincoln, NE) using a 10 g sample of flour system according to AACC-54-40. The mixograph curves (two envelopes and one midline) were computed with Mixsmart software (National Mfg. Co.). Parameters were recorded as follows: MPV (midline peak value), MPW (midline peak width), MPT (midline peak time), EW (end width), MTxV (midline time-8 min value), IT (integral of curve tail), CTV (curve tail value), CTW (curve tail width), SL (left slope of peak) and SR.
Traits PC (protein content), SV (sedimentation volume), WG (wet gluten), DG (dry gluten), and GI (gluten index) were measured. The measurement methods and phenotypic values among the above traits could be viewed in a previous study [16].

Genotyping and linkage map construction
A total of 9576 polymorphic molecular markers, including 9072 SNPs [17,18], 442 DArT markers (http://www.triticarte.com.au), 59 SSR markers, 2 biochemical markers and 1 CAPS marker, were used to genotype the RIL population. Genetic linkage groups were constructed using Mapmaker/Exp 3.0. The threshold for marker loci assignment was an LOD score of ≥3.0. The linkage map was ultimately generated using MapChart 2.1 software [19]. The genetic map spanned approximately 4048.5 cM, and the average interval of markers was 8.13 cM [16].

Statistical analysis and QTL mapping
Phenotypic values were analyzed using SPSS 17.0 software. QTL analyses were conducted by QTL Network 2.0 using the mixed-linear model approach. QTLs were designated as follows: Q, followed by the abbreviated trait name, then the wheat chromosome on which the QTL was identified; the last number indicated the marker interval on the chromosome. For example, QMPV1B.1-6 indicated a QTL for MPV identified on the sixth interval of chromosome 1B.1.

Phenotypic analysis of mixing properties
The phenotypic values were listed in Table 1. The two parents had marked differences in MPT, CTV, MTxV, CTW and EW. GC9411 exhibited higher parameter values than SN01-35 in all environments. Strong transgressive segregation was detected. This phenomenon may be affected by differences in HMW-GS subunits between parents. Differences in the CV (coefficients of variation) were found for all the traits in all three environments. Among these, CTW possessed the highest CV of 80.32% and a mean of 41.86%, followed by EW, SL and SR, with CV ranges of 37.83-49.27%, 39.46-58.24% and 26.82-36.97%, respectively. The CVs of MPV and MTxV were less than 10%, indicating that these three parameters had little potential for improvement. According to the analysis, the traits followed a normal distribution and were suitable for QTL analysis.

QTLs for mixing properties
Additive QTLs and genotype-by-environment interactions (G×E) for mixing properties MPV, CTV, and MTxV Four QTLs for MPV were detected (primarily on 1B) with net additive effect values of 0.83-2.17%, and the phenotypic variation explained (PVE) value of 6.9-29.6% (Table 2, Figure 1). The effect value of QMPV1B.1-4 was obviously changed when conditioned on SV, GI and PC, indicating that QMPV1B.1-4 was partially influenced by SV and GI and PC; however, it was not detected when conditioned on WG and DG, indicating that it was completely influenced by them. The PVE of QMPV1B.1-6 (PVE of 26.1%) was changed when conditioned on SV (PVE of 9%), but it showed no significant differences when conditioned on PC (PVE of 26.2%), GI (PVE of 25.8%) or DG (PVE of 25.6%), indicating that QMPV1B.1-6 was not influenced by PC, GI or DG but was partially influenced by SV. Notably, QMPV1B.1-24 was identified when conditioned on WG and DG in three environments, with PVEs of 8.9-27.1%. The positive alleles of the abovementioned QTL were derived from GC9411.
Six QTLs on chromosomes 1A, 1B, 3A and 5A were identified to associate with CTV, with PVEs of 1.1-21.4%. QCTV1A.1-15 was not identified when conditioned on SV, GI, WG or DG, indicating that QCTV1A.1-15 was completely influenced by the above related traits. QCTV1B.1-31 showed positive additive effects with a higher PVE of 17.3%, indicating that favorable alleles were contributed by GC9411. QCTV1B.1-26 was only identified when conditioned on PC, WG and DG, indicating that this QTL was detected when the influence of the mentioned traits was excluded. QCTV5A.1-5 was only identified when conditioned on GI, with a PVE of 5.2%, indicating that QCTV5A.1-5 was identified when the influence of GI was excluded. The favorable alleles of all the detected QTLs for CTV were contributed by GC9411.
Five QTLs associated with MTxV were identified on 1A, 1B and 1D; these QTLs were not related to DG, WG, or PC. Of these, QMTxV1B-31 exhibited the maximum PVE of 16.4%, which was strongly influenced by GI and SV.
QMTxV1D-1 (Glu-D1-WPT-667287) was not identified when conditioned on SV, GI, DG or WG, with a PVE of 7.6-10.4%; this indicates that the QTLs associated with MTxV were completely influenced by SV, GI, DG and WG. QMTxV1A.1-15 (Glu-A1-WPT-672089) was identified with a PVE of 8.07%; this QTL was not related to PC, SV, WG or DG, but it was crucially affected by GI.
Genotype-by-environment (G×E) interactions were investigated in the present study. Of the QTLs identified in the G×E analysis, QMPV1B.1-4, QCTV1B.1-26 and QMTxV1B.1-26 showed A×E% effects, with PVEs of 0.1%, 0.4% and 0.1%, respectively. Although the PVEs of the G×E QTLs were relatively low, these QTLs warrant attention for their possible role in wheat quality improvement [19].  Additive QTLs and genotype-by-environment interactions (G×E) for mixing properties MPW, CTW and EW Important QTL clusters associated with MPW were identified on chromosomes 1B, 4B, and 1D (Table 3, Figure 1). QMPW1B.1-4 was identified in unconditional analysis, with PVEs of 4.7-12.1% and net synergistic effects of 0.94-1.48%. The data indicated that cluster QMPW1B.1-4 was completely influenced by SV, WG, and DG and was not detected when conditioned on these traits. QMPW1B.1-24 was identified only in the conditional analysis, with PVEs of 6.1-15.9%, indicating that this QTL was identified when the influence of PC and WG was excluded. QMPW4B.  was identified only when conditioned on WG and DG. QMPW1D-1, with the PVE of 11.2%, was identified only when conditioned on WG and DG.
Five QTLs were identified as associated with CTW. These QTLs, with PVEs of 3.8-11.2%, were identified only in the conditional analysis, indicating that they were influenced by the quality and quantity of protein. QCTW3B.3-31 was a stable QTL identified in multiple environments. QCTW1D-1 (PVE of 13.04%) was detected in two environments, with an increased PVE compared with the corresponding unconditional QTL (unconditional PVE of 7.5%). QCTW1D-1 was not identified when conditioned on SV, GI, DG or WG, indicating that QCTW1D-1 was completely influenced by the above traits and was partially dependent on PC.
QTLs associated with EW were located on chromosomes 1A, 1D and 3B. QEW1D-1 showed the greatest contribution to EW variation, with PVEs of 16.7%. The PVEs of QEW1D-1 when conditioned on protein content did not differ significantly, and its PVE was decreased when conditioned on DG (PVE of 11.9%) compared to the corresponding unconditional QTL (unconditional PVE of 15.6%). QEW1D-1 was not identified when conditioned on SV, GI or WG, demonstrating that this QTL was independent of PC, completely dependent on SV, GI and WG and partially influenced by DG. QEW1A.1-8 was identified in multiple environments and showed PVEs of 6.63-10.04% in both conditional and unconditional mapping. QEW3B.3-31 was the major QTL identified in multiple environments (E1 and AE), with PVEs of 2.4-11.2%; this QTL was independent of SV.
A G×E interaction of QEW3B.3-31 was identified with an AE of 1.2%.   a b c d Same as Table 2.

Additive QTLs and genotype-by-environment interactions (G×E) for mixing properties MPT, SR and IT
Four QTLs were detected as associated with MPT (Table 4, Figure 1). The PVE of QMPT1A.1-14 was changed when conditioned on PC and SV, indicating that this QTL was partially influenced by PC and SV and completely influenced by GI. QMPT1A.1-15 was not identified when conditioned on SV, GI, WG or DG. QMPT1D-1 explained a PVE of 13.2-41.0%. Compared to the corresponding unconditional QTL, the PVE of QMPT1D-1 was not significantly different when conditioned on PC, but it showed a high PVE of more than 20% when conditioned on SV, WG and DG. QMPT1D-1 was not identified when conditioned on GI. The above results indicate that QMPT1D-1 was completely influenced by GI, partially influenced by SV and gluten content, and not influenced by PC. QMPT1A.1-14 and QMPT1D-1 showed positive additive effects, indicating that favorable alleles were contributed by GC9411.
Four QTLs on chromosomes 1A, 1B, 1D and 2B were discovered to be related to SR. Among them, QSR1A.1-14 was partially influenced by PC, SV and DG, whereas it was completely dependent on GI and WG. QSR1B.1-6 and QSR2B.7-34 were identified in multiple environments with a high PVE of more than 10%, indicating that these QTLs were major loci with strong stability. QSR1D-1, which was identified by both conditional and unconditional QTL mapping, was influenced by PC. The favorable alleles of all the QTLs, except for QSR1B.1-6 and QSR2B.7-34, were contributed by GC9411.
One major QTL cluster on chromosome 1B was identified as associated with IT. The three QTLs were located close to each other, with a maximum PVE of 40.9% and a net additive effect of 15.9% min. QIT1B.1-4 was detected only when conditioned on GI, WG and DG, which should attract more attention in breeding programs.

Discussion
Although several QTLs for mixing properties have been identified in previous research, there is no consensus on the merits of mixing properties in the selection of optimal dough properties. Furthermore, most of these studies only focused on additive effects using unconditional QTL mapping methods without considering the epistatic or G×E interaction effects. In this study, additive and epistatic-interaction QTLs for mixing properties, as well as their environmental interaction effects, were analyzed. In addition, we combined conditional and unconditional QTL analyses to investigate the molecular mechanisms governing mixing properties.

Important QTLs associated with mixing properties
According to the present results, several QTL clusters were identified with high PVEs and strong stability, particularly those distributed on chromosomes 1A, 1B and 1D (Fig. 1). Locus 1A.1-14 within the genomic region WPT-8455-GLU-A1 was detected in multiple environments; this QTL was related to MPT and SR and explained a PVE of 7.41-12.69%. The effect was speculated to be closely related to Glu-A1. A comparison of map positions revealed that 1A.1-8 was 13.9 cM away from 1A.1-14 (WPT-8455-GLU-A1), which was predicted to be different from Glu-A1. Further investigations are needed to completely rule out any such possibility.
QTLs related to mixing parameters (MPV, CTV, MTxV, MPW, SR, and IT) were detected on chromosome 1B and have been reported in several previous studies [2][3][4]. Here, using SDS-PAGE analysis, we confirmed that GC9411 and SN01-35 have the same HMW glutenin subunits, 7+8, which are controlled by Glu-B1 on 1BL. Therefore, the QTLs on chromosome 1BS identified as controlling mixing properties in the present study are most likely not associated with the Glu-B1 locus in the population. Consensus results were obtained by Echeverry-Solarte [20] and Liu [16], who detected loci for increased IT located on 1BS. Further investigations are needed.

Conditional QTLs for mixing properties
Conditional analysis has the ability to discern the contribution of each component trait to a complex trait. Based on the QTL analysis of mixing properties, locus 1A.1-14, which controlled MPT and SR, was identified as partially influenced by PC, SV, and DG and completely influenced by GI. An important QTL, 1D-1, that was associated with MPT, EW, CTW and MTxV was completely influenced by GI, partially influenced by SV, WG and DG, and slightly influenced by PC. A QTL that controlled MPV and MPW, 1B.1-4, was partially influenced by PC, SV, and GI and was crucially affected by WG and DG. It is worth noting that 1B.1-24, an additional locus controlling MPV and MPW, was identified in multiple environments when the influence of the examined traits was excluded, but it was not identified in the unconditional analysis. QCTV1A.1-15, which was completely influenced by protein-related traits, was detected only in unconditional mapping. QCTV5A. [1][2][3][4][5] was identified only when the influence of GI was excluded.

Conclusion
1A.1-15 and 1D.1-1 were major loci identified as associated with rheological properties influenced only by SV and GI. This finding demonstrates that these two QTL loci provide an important contribution to wheat quality at the genetic level. The combination of conditional and unconditional QTL mapping in the present study provided an opportunity for the detection of the mechanisms underlying the mixing properties of wheat.