New insight into the function of wheat glutenin proteins as investigated with two series of genetic mutants

Among the three major food crops (rice, wheat and maize), wheat is unique in accumulating gluten proteins in its grains. Of these proteins, the high and low molecular weight glutenin subunits (HMW-GSs and LMW-GSs) form glutenin macropolymers that are vital for the diverse end-uses of wheat grains. In this work, we developed a new series of deletion mutants lacking one or two of the three Glu-1 loci (Glu-A1, -B1 and -D1) specifying HMW-GSs. Comparative analysis of single and double deletion mutants reinforced the suggestion that Glu-D1 (encoding the HMW-GSs 1Dx2 and 1Dy12) has the largest effects on the parameters related to gluten and dough functionalities and breadmaking quality. Consistent with this suggestion, the deletion mutants lacking Glu-D1 or its combination with Glu-A1 or Glu-B1 generally exhibited strong decreases in functional glutenin macropolymers (FGMPs) and in the incorporation of HMW-GSs and LMW-GSs into FGMPs. Further examination of two knockout mutants missing 1Dx2 or 1Dy12 showed that 1Dx2 was clearly more effective than 1Dy12 in promoting FGMPs by enabling the incorporation of more HMW-GSs and LMW-GSs into FGMPs. The new insight obtained and the mutants developed by us may aid further research on the control of wheat end-use quality by glutenin proteins.

In the three environments (BJ, ZX and XX) in 2014/2015, ZSV and two Farinograph parameters, i.e., dough development time (DDT) and dough stability time (DST), were measured. Concomitantly, a key breadmaking quality parameter, loaf volume (LV), was also examined. Higher DDT and DST values are generally associated with stronger dough 56,57 , while a larger LV indicates better breadmaking quality 58 . From Table 2, it is apparent that the ZSV and the DDT and DST values of the six deletion lines were all significantly reduced when compared with those of Xiaoyan 81 in all three environments. Again, the strongest reduction was observed in DLGluD1, DLGluA1D1 and DLGluB1D1, with the decrease shown by DLGluA1B1 being intermediate and that by DLGluA1 and DLGluB1 being relatively low ( Table 2). The changes in LV exhibited by the six deletion lines were more complex. Nevertheless, the LV values of DLGluD1, DLGluA1D1 and DLGluB1D1 were significantly and consistently lower than that of Xiaoyan 81 in all three environments (Table 2). On the other hand, significant decrease in LV was observed in only two environments for DLGluB1 and DLGluA1B1 and merely one environment for DLGluA1 (Table 2).

Investigation of HMW-GSs in IG.
The level of HMW-GSs in IG was investigated for Xiaoyan 81 and the six deletion lines. As anticipated, the presence of HMW-GSs in IG was significantly reduced in all six deletion lines relative to that of Xiaoyan 81 in all five environments (Fig. 3). The reductions shown by DLGluD1, DLGluA1D1 and DLGluB1D1 were generally high, which was followed by DLGluA1B1; the decreases exhibited by DLGluA1 and DLGluB1 were comparatively low. In general, DLGluB1D1 exhibited the largest decrease, with only a minor amount of HMW-GSs (3.85-7.51% of that WT control, Fig. 3) present in IG.
The effects of lacking one or more HMW-GSs in the six deletion lines on the incorporation of the remaining HMW-GSs into IG were also examined. Generally, the lack of one or more HMW-GSs in the six deletion lines decreased the incorporation of the remaining HMW-GSs into IG ( Figure S1). Such effects were most pronounced in DLGluD1, DLGluA1D1 and DLGluB1D1, with the reductions ranging from 31.9% to 74.35%. These effects were lessened in DLGluA1B1 (reductions varying from 18.8% to 32.6%), and became relatively weak in DLGluA1 and DLGluB1 (reductions ranging from 3.65% to 13.2%) ( Figure S1).

Investigation of LMW-GSs in IG and SG.
Relative to WT control, the presence of LMW-GSs in IG was consistently and most severely decreased in DLGluD1, DLGluA1D1 and DLGluB1D1 in all five environments, but this decrease was less severe in DLGluA1B1, and relatively low in DLGluA1 and DLGluB1 (Fig. 4A) (Fig. 1). On the contrary, the existence of LMW-GSs in SG was generally and strongly enhanced in DLGluD1, DLGluA1D1 and DLGluB1D1 in all five environments, with presence of LMW-GSs in SG increased by 57-106% in the three mutants relative to WT control (Fig. 4B). This enhancement was, however, less pronounced for DLGluA1, DLGluB1 and DLGluA1B1 (increased by 1-26% relative to WT control, Fig. 4B). Table 4, the percentages of IG occupied by the HMW-GSs encoded by three Glu-1 loci differed significantly, 14.79-16.38% by the Glu-D1 subunits 1Dx2 + 1Dy12, 10.86-12.96% by the Glu-B1 subunits 1Bx14 + 1By15, and 4.26-7.02% by the Glu-A1 subunit 1Ax1. Obviously, the abundance of 1Dx2 + 1Dy12 in IG was substantially higher than that of 1Bx14 + 1By15 or 1Ax1, with the amount of 1Ax1 being the lowest. In line with this difference, the lack of 1Dx2 and 1Dy12 together caused the largest reduction in IG (by 47.93-53.52%), the strongest decrease of HMW-GSs in IG (69.51-72.08%), and the most severe reduction of LMW-GSs in IG (37.19-45.37%) ( Table 4). The three effects were lessened when 1Bx14 and 1By15 were missed, and tended to be small when 1Ax1 was absent ( Table 4).

Effects of 1Dx2 or 1Dy12 alone on gluten, dough and end-use quality parameters. The forego-
ing experiments highlighted the functional dominance of Glu-D1 over Glu-A1 and -B1. Because Glu-D1 encodes both 1Dx2 and 1Dy12, it became necessary and important to examine if the two subunits may act similarly or differently in wheat end-use quality control. To this end, we compared two EMS knockout mutants (md2-1 and md12-1) lacking the expression of 1Dx2 and 1Dy12, respectively ( Figure S2). The two mutants were developed  Table 3. Correlation coefficients between UPP, IG and the gluten, dough and breadmaking quality parameters of the samples collected from five environments. DDT, dough development time; DST, dough stability time; IG, insoluble glutenin; LV, loaf volume; UPP, unextractable polymeric proteins; ZSV, Zeleny sedimentation volume. ** Statistically significant at P < 0.01. using the common wheat cultivar Xiaoyan 54 45 , and their genetic backgrounds were made near-identical to that of Xiaoyan 54 through six rounds of backcrossing (see Methods). Xiaoyan 54 is one of the two parents of Xiaoyan 81, and expresses an identical set of HMW-GSs as Xiaoyan 81 ( Figure S2). Xiaoyan 54 and the two knockout mutants were cultivated in two crop seasons (environments) (2014/2015 and 2015/2016), with the grains harvested being used for measuring gluten, dough and end-use quality parameters. In the two environments, ZSV and the DDT and DST values of the two knockout mutants were generally and significantly decreased relative to those of Xiaoyan 54, and in four of the six cases, the reduction exhibited by md2-1 was significantly more severe than that by md12-1 (Table 5). In agreement with these results, the loaf volume values of the two knockout mutants were significantly lower than those of Xiaoyan 54 in both environments (Table 5). Moreover, the loaf volume of md2-1 tended to be smaller than that of md12-1 ( Figure S3), with the difference reached to a significant level (P < 0.05) in 2014/2015 (Table 5).
Alterations in IG content and composition caused by knocking out 1Dx2 or 1Dy12. The potential consequences of lacking 1Dx2 or 1Dy12 on IG content and the incorporations of HMW-GSs and LMW-GSs into IG were investigated as described above. The results obtained for the grain samples harvested in 2014/2015 are displayed in Fig. 5. Compared with Xiaoyan 54, IG content was significantly decreased in both md2-1 and md12-1, but the scale of the decrease was much higher in md2-1 (Fig. 5A). The knockout of 1Dx2 reduced the incorporation of the remaining HMW-GSs into IG, with the percentage of the reduction being 38.7%, 47.0%, 48.2% and 45.4% for 1Ax1, 1Bx14, 1By15 and 1Dy12, respectively (Fig. 5B). The knockout of 1Dy12 also decreased the incorporation of other HMW-GSs into IG, but the percentages of the reduction observed (21.5% for 1Ax1, 33.8% for 1Bx14, 44.4% for 1By15 and 32.8% for 1Dx2) were generally lower than those caused by the lacking of 1Dx2 (compare Fig. 5B and C). Lastly, the lack of 1Dx2 decreased the incorporation of LMW-GSs in IG by 39.3%, whereas the absence of 1Dy12 reduced the incorporation of LMW-GSs in IG by only 18.7% (Fig. 5D). The results gathered for the grain samples harvested in 2015/2016 ( Figure S4) were similar to those shown in Fig. 5, although the scales of the decreases in IG, LMW-GSs in IG, and the percentages of reduction of different HMW-GSs in IG tended to be smaller. These variations may be caused by differences in the growth environment between the two seasons.

Discussion
In this work, we investigated the function of glutenin proteins in wheat end-use quality control and the mechanism involved through analyzing two series of well-defined genetic mutants. Complementary sets of data were obtained using the grain samples harvested from multiple environments, which permitted an objective assessment of the genetic effects of lacking one or two Glu-1 loci on the examined gluten, dough and breadmaking quality parameters. The new insight obtained is discussed below.

Comparative analysis of single and double mutants of Glu-1 loci reinforces the dominance of
Glu-D1 in wheat end-use quality control. Previously, we found that the contribution of three Glu-1 loci to wheat gluten and GMP parameters can be ranked as Glu-D1 > Glu-B1 > Glu-A1 through analyzing single deletion mutants lacking individual Glu-1 loci 49 . Here, we substantially extended our investigation by including both single and double deletion mutants of Glu-1 loci, using the grain samples from multiple field environments, and testing more gluten, dough and end-use quality parameters. We consistently observed that the mutants lacking Glu-D1 (DLGluD1) or its combination with Glu-A1 (DLGluA1D1) or Glu-B1 (DLGluB1D1) showed the strongest reductions in the examined gluten, dough and breadmaking quality parameters (Tables 1 and 2). These observations reinforce the dominance of Glu-D1 in wheat end-use quality control. In most cases, the reductions displayed by DLGluA1D1 were larger than those by DLGluD1 but smaller than those by DLGluB1D1. This is consistent with the fact that the number of HMW-GSs lacked in DLGluB1D1 (i.e., 4) was more than that in DLGluA1D1 (3) or DLGluD1 (2) (Fig. 1). Clearly, there exist positive and additive interactions among the three Glu-1 loci studied in this work, with the functional effects of the interactions between Glu-A1 and Glu-D1 being comparatively weaker than those between Glu-B1 and Glu-D1. Past studies have also detected positive and additive interactions among the three Glu-1 loci 47,48,59,60 .
While comparing the doughs of Xiaoyan 81 and the single and double mutants using Farinograph test (Table 1), we focused on only two major parameters (DDT and DST) owing to the large number of samples needing to be assayed. However, other parameters of this test, i.e., width of Farinograph curve at peak consistency and rapidity of decline of Farinograph curve after peak consistency, also provide useful information on dough elasticity and cohesiveness 61,62 . Upon closer examination of the Farinograph curves ( Figure S5), the width of the Environment Subunit (Glu-1 locus)   Figure S4). curve in DLGluA1, DLGluB1 and DLGluA1B1 was not reduced as severely as that in DLGluD1, DLGluA1D1 and DLGluB1D1, and in general, the Farinograph curves of DLGluA1, DLGluB1 and DLGluA1B1 were declined less rapidly than those of DLGluD1, DLGluA1D1 and DLGluB1D1. Since DLGluA1, DLGluB1 and DLGluA1B1 all possessed a functional Glu-D1, these observations suggest that Glu-D1 is more important than Glu-A1 and Glu-B1 in maintaining the width of Farinograph curve at peak consistency and for slowing down the decline of Farinograph curve after peak consistency. Because of the presence of Glu-D1 in DLGluA1, DLGluB1 and DLGluA1B1, the functionality of the doughs of the three lines was not lowered as drastically as that of DLGluD1, DLGluA1D1 and DLGluB1D1 (all lacking Glu-D1). This may help to explain the less consistent decreases in LV observed for DLGluA1, DLGluB1 and DLGluA1B1 in different environments despite extensive reductions in their ZSV, DDT and DST values ( Table 2). The functional dominance of Glu-D1 over Glu-A1 and Glu-B1 had also been suggested by prior studies using recombinant wheat lines differing in the composition of Glu-1 loci and in genetic background [46][47][48]63 . For example, Lawrence and coauthors demonstrated that Glu-D1d (a different Glu-D1 allele encoding 1Dx5 and 1Dy10 subunits) was functionally superior to Glu-A1 and Glu-B1 46 . In contrast, our data were obtained by using six mutant lines with highly similar genetic background. Therefore, our work validated previous observation by more robust genetic data. The Glu-D1 allele studied by us is Glu-D1a, which is predominant in worldwide common wheat varieties 64,65 . Apart from Glu-D1a and Glu-D1d, there are several minor Glu-D1 alleles (Glu-D1b, -D1c, -D1e and -D1f) 64 . It will be interesting to investigate if these minor Glu-D1 alleles may also be functionally dominant over Glu-A1 and Glu-B1 in the future.

Decrease of LMW-GSs in IG (%)
Glu-D1 has the strongest potency to promote the incorporation of HMW-GSs and LMW-GSs into FGMPs. FGMPs play pivotal roles in gluten and dough functionality and end-use quality 20,66 . Their amount and polymerization characteristics are strongly affected by both the quantity and structural features of different HMW-GSs and LMW-GSs. Based on the changes in UPP, IG and the amount of HMW-GSs and LMW-GSs in IG among WT control and the six deletion mutants observed in this work (Figs 2-4), we suggest that the three Glu-1 loci differ significantly in the ability to control the accumulation of FGMPs through promoting the incorporation of HMW-GSs and LMW-GSs into FGMPs. Specifically, Glu-D1 has the strongest potency to promote the incorporation of HMW-GSs and LMW-GSs into FGMPs, and thus makes the largest contribution to FGMP accumulation. In contrast, Glu-B1 is less effective than Glu-D1, and Glu-A1 is weaker than Glu-B1 in these processes. From this suggestion and the existence of highly significant correlations between the changes in gluten, dough and breadmaking quality parameters and those in UPP and IG content (Table 3), we further propose that, for individual Glu-1 loci, the higher the potency to promote the incorporation of HMW-GSs and LMW-GSs into FGMPs, the stronger the contributions to FGMPs, gluten and dough functionality, and end-use quality performance.
The high potency of Glu-D1 in promoting the incorporation of HMW-GSs and LMW-GSs into FGMPs is also supported by two additional lines of evidence. First, the absence of Glu-D1 or its combination with Glu-A1 or Glu-B1 reduced the incorporation of the remaining HMW-GSs into IG (by 31.9-74.35%) much more strongly than that (18.8-32.6%) caused by lacking Glu-A1, Glu-B1 or both ( Figure S1). Second, in the absence of Glu-D1 or its combination with Glu-A1 or Glu-B1, the presence of LMW-GSs in SG was greatly enhanced (by 57-106%) relative to that (1-26%) due to the mutation of Glu-A1, Glu-B1 or both (Fig. 4B).
The reason(s) underlying the enhanced potency of Glu-D1 to promote the incorporation of HMW-GSs and LMW-GSs into FGMPs may be complex, because the subunits encoded by Glu-D1 (1Dx2 and 1Dy12) differ from those encoded by Glu-B1 (1Bx14 and 1By15) and Glu-A1 (1Ax1) in multiple aspects. Nevertheless, we noticed that the abundance in IG of 1Dx2 + 1Dy12 was significantly higher than that of 1Bx14 + 1By15 or 1Ax1, and nearly equaled to the amount of 1Bx14 + 1By15 + 1Ax1 in all five environments (Table 4). Furthermore, the reduction of IG and the decreases of HMW-GSs and LMW-GSs in IG brought about by lacking 1Dx2 + 1Dy12 were always more severe than those caused by missing 1Bx14 + 1By15 or 1Ax1 (Table 4). Therefore, the high abundance of 1Dx2 + 1Dy12 in IG (relative to that of 1Bx14 + 1By15 or 1Ax1) is likely an important factor for the functional dominance of Glu-D1 (over that of Glu-B1 or Glu-A1). Because of the existence of many amino acid substitutions among the deduced proteins of 1Dx2, 1Ax1 and 1Bx14 and between those of 1Dy12 and 1By15 67, 68 , the structural differences of these subunits may also contribute to the functional dominance of Glu-D1. Further work is needed to validate this possibility. 1Dx2 has a stronger function than 1Dy12. In common wheat, Glu-B1, Glu-D1 and their different alleles usually express two different HMW-GSs (one x-and one y-type) 2,8 . Consequently, uncovering functional difference between the two subunits is essential for more comprehensively understanding the action of HMW-GSs in controlling wheat end-use quality. Some information has been gained on the function of certain HMW-GSs (e.g., 1Dx5 and 1Dy10) in controlling wheat end-use quality through studying variety population differing in HMW-GS composition, transgenic overexpression or RNA interference [69][70][71][72] . However, there is still no report on the use of knockout mutants with a near identical genetic background in investigating functional difference between the two subunits encoded by a Glu-1 locus. In this work, we examined functional difference between the Glu-D1 encoded subunits 1Dx2 and 1Dy12 by comparing two knockout mutants, md2-1 (lacking 1Dx2 expression) and md12-1 (without 1Dy12 accumulation), with their WT progenitor Xiaoyan 54. Judging from the data presented in Table 5, the function of 1Dx2 is generally and considerably stronger than that of 1Dy12 with respect to the control of the examined gluten, dough and breadmaking quality parameters. The stronger function of 1Dx2 (relative to that of 1Dy12) is most likely caused by its higher contribution to FGMPs through promoting the incorporation of more HMW-GSs and LMW-GSs into FGMPs (Fig. 5). Thus, the ability to promote the incorporation of more HMW-GSs and LMW-GSs into FGMPs is a common reason for the functional superiority of both Glu-D1 and the 1Dx2 subunit encoded by it.
Scientific RepoRts | 7: 3428 | DOI:10.1038/s41598-017-03393-6 In line with our finding, earlier studies also revealed that x-type subunits had greater effects on dough functionality parameters than y-type subunits by analyzing transgenic lines and variety population 71,73 or through artificial incorporation of HMW-GSs into developing dough 74 . Thus, the function of x-type HMW-GSs may be generally stronger than that of y-type HMW-GSs in the control of wheat end-use quality. This raises the question what is the mechanism behind the stronger function of x-type HMW-GSs. In the current model on the structure of GMPs, y-type HMW-GSs interact covalently with LMW-GSs, with the resultant units linked by x-type HMW-GSs 43 . Although x-y and x-x linkages have been found among HMW-GSs, it is still uncertain if covalent interactions may happen between x-type HMW-GSs and LMW-GSs 23,39,40,42 . We speculate that x-type HMW-GSs may interact with LMW-GSs and form FGMPs. This speculation is based on the gluten, dough and breadmaking quality parameters obtained in this work for the double mutant DLGluB1D1. Although this mutant had only one x-type HMW-GS (i.e., 1Ax1) accumulated in the grains (Fig. 1), its bread volume was still 74.3-82.0% of that of WT control (Table 2), and its UPP and IG contents were still 20-30% and 30-40% of those of WT control, respectively (Fig. 2). Moreover, a substantial amount of LMW-GSs was present in the IG of DLGluB1D1 (47.3-55.7% of that of WT control, Fig. 4). Considering that there was no y-type HMW-GS present in DLGluB1D1, and the level of 1Ax1 in its grains was rather low (Fig. 3 and Figure S1), the interactions between 1Ax1 and LMW-GSs, if existed, may be fairly effective. Therefore, the actions of x-type HMW-GSs in FGMP formation are likely more extensive than currently thought. A better elucidation of these actions may help to explain the functional superiority of x-type HMW-GSs (over their y-type counterparts) in wheat-end use quality control.
In summary, we have generated new information on the functional difference among three Glu-1 loci and between two HMW- GSs (1Dx2 and 1Dy12). The three loci, as well as the two subunits, differ significantly in the efficacy to promote the incorporation of HMW-GSs and LMW-GSs into FGMPs, and these differences are largely responsible for the functional dominance of Glu-D1 over Glu-A1 and Glu-B1 and the functional superiority of 1Dx2 to 1Dy12. This insight increases our understanding of the function of HMW-GSs in controlling important gluten and dough properties and breadmaking performance. Moreover, the data from this work and our previous study 45 confirm that the Glu-1 locus deletion mutants and the EMS mutants lacking individual or combinations of HMW-GSs are valid materials for further research on wheat end-use quality. Continued analysis of these mutants with functional genomics approaches (e.g., using transcriptomic, proteomic and/or metabolic methods) may shed new light on the genetic and molecular basis of gluten and dough functionalities and lead to valuable strategies for improving wheat end-use traits.

Methods
Plant materials and growth conditions. Genetic crosses were conducted in between DLGluA1, DLGluB1 and DLGluD1, which have Glu-A1, -B1 and -D1 deleted, respectively 49 . Homozygous plants missing two Glu-1 loci (Glu-A1 and -B1, Glu-A1 and -D1 or Glu-B1 and -D1) were identified by checking HMW-GS composition in F2 seeds, and used to develop the three double deletion mutants DLGluA1B1, DLGluA1D1 and DLGluB1D1. Xiaoyan 81 (WT progenitor) and the six deletion mutants were grown in five field environments with normal supplies of irrigation water and chemical fertilizers 75 . The five environments were created by growing the materials in two locations (Zhaoxian and Xinxiang) in 2013/2014 and three locations (Beijing, Zhaoxian and Xinxiang) in 2014/2015. The knockout mutants md2-1 and md12-1, lacking the expression of 1Dx2 and 1Dy12, respectively, were backcrossed six times using their WT progenitor Xiaoyan 54 as recurrent parent 45 . In this study, the three isogenic lines (Xiaoyan 54, md2-1 and md12-1) were cultivated in Beijing in two wheat crop cycles (2014/2015 and 2015/2016) as described above. The deletion mutants and their grains were checked for agronomic traits using standard methods 49,75 . Flour samples were prepared for the different lines as reported before 49 .

SDS-PAGE.
HMW-GSs accumulated in the different experimental lines were extracted using 20 mg flour, and separated with 10% SDS-PAGE according to the method describe previoulsy 76 .

RP-HPLC.
IG and SG levels in the flour samples were assayed using RP-HPLC. The majority of the analysis steps were carried out at room temperature (RT, approximately 25 °C) except where noted. IG and SG were extracted following the method detailed previously 31 . Briefly, for each line (WT control or deletion mutant), the flour sample (50 mg) was extracted twice with 0.5 ml of 50% (v/v) 1-propanol. After each extraction, the sample was centrifuged at 2,200 g for 3 min. The pellet was washed with 0.5 ml 50% (v/v) 1-propanol for 1 min and centrifuged at 15,000 g for 3 min. The pellet was further extracted with 0.5 ml 50% (v/v) 1-propanol containing 1% (w/v) dithiothreitol (DTT) at 65 °C for 1 h, followed by the addition of 1.4% (v/v) 4-vinylpyridine for 30 min at 65 °C. The mixture was then centrifuged at 15,000 g for 10 min, with the supernatant retained as IG fraction. The three supernatants after the extraction with 50% (v/v) 1-propanol in the preceding steps were combined, and 1-propanol was added to 70%. After centrifugation at 12,000 g for 3 min, the precipitated proteins were dissolved in 0.5 ml 50% (v/v) 1-propanol containing 1% (w/v) DTT by incubating at 65 °C for 1 h. Subsequently, 4-vinylpyridine was added to 1.4% (v/v), and the mixture was maintained at 65 °C for another 30 min. Lastly, the mixture was centrifuged at 15,000 g for 10 min, with the supernatant kept as SG fraction. The SG and IG fractions were all filtered through 0.45 μm nylon filter before being analyzed by RP-HPLC. RP-HPLC analysis was accomplished with the Agilent 1260 infinity Quaternary LC System using a C18 column. The elution conditions were essentially those described by González-Torralba and coauthors 77 . For each RP-HPLC run, a volume of 15 μl of the filtered IG (SG) was analyzed. The amount of HMW-GSs and LMW-GSs were calculated by integrating the areas under the corresponding protein peaks of the chromatogram.

SE-HPLC.
The assay of UPP content by SE-HPLC was conducted at RT following the method described in a previous study 78 . Briefly, for each sample to be assayed for UPP, 10 mg flour was suspended in 1 ml extraction buffer (50 mM sodium phosphate containing 0.5% SDS, pH 6.9), and vortexed for 10 min. The mixture was centrifuged at 17,000 g for 15 min, and the resultant pellet was resuspended in 1 ml extraction buffer, followed by sonication in a SCIENTZ-IID sonicator (Scientz Biotechnology Co., Ningbo, China). The condition of sonication was 20% output power for 30 s using a 3 mm sonicator probe, with the probe placed at 1/3 distance from the bottom of the microfuge tube. Afterwards, the mixture was centrifuged at 17,000 g for 15 min, with the supernatant retained as UPP. It was filtered through 0.45 μm nylon filter, and analyzed by SE-HPLC with the Agilent 1260 infinity Quaternary LC System using a Biosep SEC-4000 column (Phenomenex, Torrence, CA, USA). For each filtered UPP sample, an aliquot of 15 μl was assayed by SE-HPLC, with UPP content calculated by integrating the areas under the corresponding peaks of the chromatogram.
Evaluation of ZSV and Mixograph, Farinograph and loaf volume parameters. ZSV was measured following the method described previously 52 . Mixograph parameters (MPT, MPH and MPW) were determined with a 10 g mixograph system (National Manufacturing Co., Lincoln, NE, USA) using the AACCI method 54-40.02 79 . Farinograph parameters (DDT and DST) and loaf volume were measured according to the AACCI methods 54-21.02 and 10-10.03, respectively 79 .
Statistical analysis. For the experiments described above, three separate tests were carried out for each sample. The data obtained were subjected to one-way analysis of variance (ANOVA) using IBM SPSS Statistics 19 software (IBM, New York, USA), followed by the least significant difference multiple comparison test. Pearson's correlation coefficients between IG, UPP and the gluten, dough and breadmaking quality parameters were calculated using the IBM SPSS Statistics 19 software.