Suitability of Reversed-Phase High-Performance Liquid Chromatographic Separation of Wheat Proteins for Long-Term Statistical Assessment of Breadmaking Quality'

Cereal Chem. 67(4):395-399 The proteins of flours of 26 widely different bread wheat varieties grown using a different column of the same type. The integrated areas of the in Canada were extracted with 70% ethanol, with and without a reducing 21 regions for the 16 varieties were used to test the prediction equations. agent (20% mercaptoethanol). Replicate extracts with each solvent and Extensigraph extensibility was the most consistently predictable breadduplicate injections of each extract were analyzed by high-performance making quality parameter from HPLC analyses of ethanol extracts preliquid chromatography (HPLC) on a reversed-phase column over a period pared with and without the reducing agent. The extensibility appeared of two months. The HPLC chromatograms were compared with standards to be predicted from the overall protein composition rather than from (Neepawa chromatograms) determined at the same time. For statistical individual gliadins. Prediction of dough extensibility using prediction analysis, each chromatogram was divided into 21 regions. The areas of equations derived from HPLC analyses would complement SDS-PAGE these regions were used to generate prediction equations for several breadfor screening varieties for breadmaking quality in early generations of making quality parameters. Flour proteins of 16 different wheat varieties breeding programs. grown in Canada, extracted with the same two solvents, were separated In the conversion of wheat into bread, flour proteins play a major role in the unique viscoelastic properties of dough that produces bread of high quality. The main protein constituents are the gliadins and the glutenins, which together produce gluten, the three-dimensional matrix of a loaf of bread. Statistical analyses of protein components separated by electrophoresis link certain gliadins and glutenin subunits to breadmaking quality parameters (Wrigley et al 1982; Branlard and Dardevet 1985a,b; Campbell et al 1987; Payne et al 1987; Ng and Bushuk 1988). On the other hand, doubts about the reproducibility of results with different high-performance liquid chromatography (HPLC) columns (Goldberg 1982) and their long-term stability (Glajch et al 1987), have delayed the use of reversed-phase HPLC (RP-HPLC) to predict breadmaking quality on the basis of gluten protein components separated by this technique. In this study, the proteins of 26 diverse bread wheat varieties were analyzed by RP-HPLC and the results were used to develop equations for predicting breadmaking quality parameters. The equations were then used to predict the breadmaking quality parameters of 16 other wheat varieties analyzed on a different RP-HPLC column. MATERIALS AND METHODS Reagents Chromic acid and 2-mercaptoethanol (2-ME) were of reagent grade, and dimethyl sulfoxide (DMSO) was of ACS reagent grade. All other chemicals were of HPLC grade obtained from sources indicated previously (Scanlon et al 1989). Wheat Samples Grain of 26 varieties of the Uniform Quality Nursery (UQN) (described by Ng and Bushuk 1988), the Canadian bread wheat variety Marquis, 10 varieties of the 1987 Saskatchewan Wheat Pool (SWP) bread wheat test, and six varieties of the 1987 Parkland Wheat Cooperative Test (PWCT) was used in the present study. Grain of the Canadian bread wheat variety Neepawa was used as the reference standard (Sapirstein et al 1989). Presented in part at the AACC 74th Annual Meeting, Washington, DC, November 1989. Publication no. 165 of the Food Science Department, University of Manitoba, Winnipeg, Manitoba, Canada R3T 2N2. 2 Present address: Flour Milling and Baking Research Association, Chorleywood, Herts WD3 5SH, England. © 1990 American Association of Cereal Chemists, Inc. Sample Preparation Grain samples were milled into straight-grade flour on a Buhler pneumatic laboratory mill. Additionally, whole meal of the variety Neepawa was prepared by grinding on a Udy cyclone mill (Udy Analyzer Co., Boulder, CO). Two extracting solvents were used, 70% ethanol and 70% ethanol containing 20% 2-ME. The use of 20% 2-ME maximizes the amount of ethanol-soluble protein extractable from flour or ground wheat meal (Wren and Nutt 1967). Since heating the extracts did not increase protein extractability (from examination of the chromatograms, data not shown), extractions for this study were performed at room temperature. Flour (100 mg) was mixed with extracting solvent (400 ,ul) in 1.5-ml microcentrifuge tubes and vortexed at 5-min intervals for 15 min and then centrifuged for 15 min at 8,800 X g at room temperature. The clear supernatant was filtered through a 0.45-,um nylon filter (Millipore, Mississauga, ON) into a chromatography microvial (Hewlett-Packard, Palo Alto, CA). Chromatography Apparatus and solvents for HPLC were as described previously (Scanlon et al 1989). If the HPLC procedure is to be used to routinely monitor wheat quality, then changes in operating conditions should not affect (or only slightly affect) the results. Therefore two SynChropak C18, 300 A (25 cm X 4.6 mm i.d.) reversed-phase columns were used in conjunction with guard columns (5 cm X 4.6 mm i.d.) of the same packing. The column used to generate the prediction equations (UQN varieties) was supplied by Terrochem (Edmonton, AB). Subsequently, the analogous results for 16 varieties, obtained on a similar column supplied by Phenomenex (Rancho Palos Verdes, CA), were used to test the prediction equations. The second column was used to simulate the effect of another laboratory using the prediction equations to predict breadmaking quality parameters for a variety whose proteins had been analyzed by HPLC. Column temperature was 50.0°C; solvent flow rate was 1.0 ml/min; and 10-,ul aliquots were chromatographed. The elution gradients used are given in Table I. To generate data for the prediction equations for the extracts obtained with 70% ethanol only (unreduced), the 26 UQN varieties and the variety Marquis were analyzed in random order; the extracts were analyzed in groups of six. Each extract was injected twice. The 27 varieties were reextracted and the extracts analyzed again in duplicate but in a different random order. Thus for these extracts, two replicate extracts and two duplicate injections of each were analyzed for each of the 27 varieties. Vol. 67, No. 4,1990 395 Two replicate extracts with 70% ethanol containing 2-ME (reduced extracts) and two duplicate injections of each extract were analyzed similarly after all extracts without reducing agent had been analyzed. Each group of six extracts (varieties) was defined as a sequence. The samples of each sequence were chromatographed in the following order: extract of Neepawa reference; extracts of the first group of six samples; two-column cleanup runs (see below); reequilibration to the starting solvent composition (25 min); a second different extract of Neepawa; the duplicate injections of the six samples; an aliquot of a third extract of Neepawa; and finally, three cleanup runs. Each cleanup run consisted of a -Mul injection of DMSO at a solvent composition of 55% acetonitrile/45% water (both containing 0.1I% trifluoroacetic acid) at a flow rate of 0.1 ml/ min for 2 hr. For the 16 varieties used to test the prediction equations, only one replicate for each extracting solvent was prepared, and no duplicate injections were performed. The 10 SWP varieties were analyzed first. Two months later (to allow some column aging) the six PWCT varieties were analyzed similarly to obtain the data for 16 varieties for testing the prediction equations. Quantitation of HPLC Results Each chromatogram was divided into 21 regions (Fig. 1). Some of the regions represented distinct single peaks whereas others represented a number of peaks. Because of the selectivity and changes in retention time that occur during chromatography of a large group of samples (Sapirstein et al 1989, Scanlon et al 1989), absolute retention times were not used as limits of the regions. Rather, each chromatogram was divided into regions by comparing retention times of sample peaks with retention times of peaks in the Neepawa chromatogram acquired before and after in the same sequence (this is a form of visual normalization). The Hewlett-Packard 79994A software was used to integrate the chromatograms (Scanlon et al 1989) and the peak areas for the regions were recorded. Since not all peaks from other varieties conveniently overlaid the demarcation limits for those of Neepawa, some peak areas had to be estimated by comparing the size of the peak of interest with similarly sized peaks for which the area had been determined. Differences in the overall level of integration and in the amount of protein analyzed were eliminated by expressing each region as a percentage of the total peak area for each chromatogram. Technological Tests Farinograms were obtained according to the AACC method 54-21 (AACC 1983) using 50 g of flour to derive dough development time (in minutes) and mixing tolerance index (MTI, Brabender units). Extensigrams were obtained according to Holas and Tipples (1978) using a Brabender Extensigraph to derive extensibility (E, mm), maximum resistance (R, Brabender units), R/ E (ratio of R to E), and area under the curve (A, cm). The remix baking test for 100 g of flour was carried out according to Kilborn and Tipples (1981). Volumes of the resulting loaves (LV, cm) were measured using a pup loaf volumeter (National Mfg. Co., Lincoln, NE). The baking strength index (BSI) was determined according to Tipples and Kilborn (1974). These tests gave a total of eight breadmaking quality parameters. Statistical Analyses Data were analyzed on the University of Manitoba Amdahl TABLE I Gradients Used to Elute Proteins of Unreduced and Reduced Samples Unreduced Samples 6280 computer using Statistical Analysis System program packages (SAS 1985). The peak areas for each of the 21 regions of the chromatogram of each of the 26 UQN varieties were the independent variables used to generate a prediction equation for each dependent variable (breadmaking quality parameter) using stepwise multiple regression (STEPWISE procedure) with the maximum r2 improvement option. The prediction equation selected was the one that gave the maximum r value with the minimum probability value. Since duplicate injections for a given extract were almost exact overlays of each other, the duplicate results were not used in the statistical analyses. Analysis of the results for the first extract of the 26 varieties produced the first set of prediction equations. The analysis was repeated for the second set of replicate extracts, and likewise for the other extracting solvent. In this way, four prediction equations were generated for each breadmaking quality parameter; two for unreduced protein extracts and two for reduced. To test the efficacy of the 32 prediction equations (eight quality parameters, two solvents, two replicates), the linear regression procedure (REG) was used with the option of 95% confidence limits to predict the quality parameters from analogous RP-HPLC data for 10 SWP and six PWCT varieties. RESULTS AND DISCUSSION Figure 1 A and B show the separation achieved for proteins extracted from Neepawa by 70% ethanol without and with 2-ME (reducing agent), respectively. Although better separation could have been achieved in the 30-38 min region of Figure 1 B (Marchylo et al 1988), the expenditure of extra time and materials was deemed unnecessary since it appears likely that the peaks in this region represent proteins of a single gliadin or glutenin type (Bietz and Burnouf 1985). Stepwise multiple regression analysis showed highly significant relationships (P < 0.001) between certain gliadin components of the two replicate analyses of unreduced extracts and E, MTI, LV, and RIE. Likewise certain peak areas for the reduced extracts gave highly significant relationships (P < 0.001) to E, MTI, LV,