Journal of Cereal Science 38 (2003) 205–215 www.elsevier.com/locate/jnlabr/yjcrs
Wheat protein quality in relation to baking performance evaluated by the Chorleywood bread process and a hearth bread baking test K.M. Tronsmoa,b,*,1, E.M. Færgestada, J.D. Schofieldc, E.M. Magnusa b
a MATFORSK—Norwegian Food Research Institute, Osloveien 1, N-1430 A˚s, Norway Department of Chemistry and Biotechnology, Agricultural University of Norway, P.O. Box 5040, N-1432 A˚s, Norway c School of Food Biosciences, The University of Reading, Whiteknights, Reading RG6 6AP, UK
Received 19 June 2002; revised 17 February 2003; accepted 5 March 2003
Abstract The relationships between wheat protein quality and baking properties of 20 flour samples were studied for two breadmaking processes; a hearth bread test and the Chorleywood Bread Process (CBP). The strain hardening index obtained from dough inflation measurements, the proportion of unextractable polymeric protein, and mixing properties were among the variables found to be good indicators of protein quality and suitable for predicting potential baking quality of wheat flours. By partial least squares regression, flour and dough test variables were able to account for 71 – 93% of the variation in crumb texture, form ratio and volume of hearth loaves made using optimal mixing and fixed proving times. These protein quality variables were, however, not related to the volume of loaves produced by the CBP using mixing to constant work input and proving to constant height. On the other hand, variation in crumb texture of CBP loaves (54 – 55%) could be explained by protein quality. The results underline that the choice of baking procedure and loaf characteristics is vital in assessing the protein quality of flours. q 2003 Elsevier Ltd. All rights reserved. Keywords: Loaf volume; Form ratio; Crumb texture; Dobraszczyk/Roberts dough inflation system
1. Introduction The importance of the protein fraction of wheat on baking properties is well documented (Kasarda, 1989; Abbreviations: d.m., dry matter basis; m.b., moisture basis; HMW-GS, high molecular weight glutenin subunits; CBP, Chorleywood bread process; PLS, partial least squares; PSI, particle size index (a measure of grain hardness); SDS, sodium dodecyl sulfate sedimentation volume; Mixing properties: DDT, Farinograph dough development time at 126 rpm; MPT, Mixogram peak time; MPH, Mixogram peak height; SE-HPLC, sizeexclusion high performance liquid chromatography: %F1*,…, %F5, proportion (%) of chromatography fraction F1*,…, F5; %UPP, proportion (%) of unextractable polymeric protein; Dough Inflation: P, tenacity (maximum pressure); MP stress, stress at maximum pressure during bubble inflation; MP strain, strain at maximum pressure; W, deformation energy; n, strain hardening index; L, extensibility (drum distance at burst point); burst stress, stress at burst point; burst strain, strain at burst point. * Corresponding author. E-mail address:
[email protected] (K.M. Tronsmo). 1 Present address: Danone Vitapole, RD 128, F-91767 Palaiseau Cedex, France. Tel.: þ33-1-69-35-74-31; fax: þ33-1-69-35-76-98. 0733-5210/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0733-5210(03)00027-4
MacRitchie, 1984,1992; Pomeranz, 1988). The classical study by Finney and Barmore (1948) showed that for pan loaves produced by an optimised breadmaking process, loaf volume increased linearly with protein content, but the rate of increase differed between cultivars. The most important contribution to this difference in the slope of the regression line is protein quality, which relates to the composition of glutenin subunits (Payne et al., 1987; Shewry et al., 1992; Brett et al., 1993) and gliadins (Flæte, 2001, Sontag-Strohm et al., 1996), the relative proportions of different protein classes (Kasarda, 1989; Singh et al., 1990) and the molecular weight distribution of the glutenin polymers (Southan and MacRitchie, 1999). Interactions between flour quality and the baking process for different products are also of great importance for the characteristics of the final product (Finney and Barmore, 1948; Færgestad et al., 1999, 2000; Magnus et al., 1997; Roels et al., 1993). Quite different results are revealed in optimised baking processes compared with processes using fixed levels for process variables such as water
206
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
absorption, mixing time and proving time. Most studies reported in the literature concern the baking of pan bread. An earlier study in our laboratory (Færgestad et al., 2000) showed that a hearth bread method involving optimal mixing and fixed proving time was efficient in distinguishing between wheats of different protein quality. The method could also distinguish between the effects of protein content and protein quality. The form ratio (height/width) of the bread was positively affected by protein quality and slightly negatively related to protein content, whereas the loaf volume was positively influenced by protein content (Færgestad et al., 2000; Tronsmo et al., 2002). In the United Kingdom and in a number of other countries, the baking industry predominantly uses the Chorleywood Bread Process (CBP) or a similar process. The CBP is a mechanical dough development process involving intensive mixing, the use of oxidants, a short fermentation process, and the use of a pan for baking. The method is also largely used for the assessment of wheat samples for research purposes, with the loaf volume regarded as the most important bread characteristic since it provides a quantitative measure of baking performance. As the protein properties of wheat dough are very complex, depending on relative proportions of different protein classes, glutenin polymer size distribution and interactions with other flour components, no single test can be expected to describe the system comprehensively. Nevertheless, several tests exist which can give good indications of the baking properties. In an earlier study we compared the potential of several dough and gluten rheological tests for describing variation in protein quality (Tronsmo et al., 2003a). Among these was the dough inflation test. In the previous report, this test was performed on a limited sample set. The present work provided a larger sample set (20 cultivars grown under the same conditions) for further assessing the potential of the dough inflation test in predicting baking quality. The aim of the present study was to compare results from baking by the CBP and a hearth bread baking method developed in our laboratory, with respect to the influence of protein quality on bread characteristics. The study also evaluates the predictive power of various tests of protein quality, including the dough inflation test.
2. Experimental 2.1. Material The sample set comprised 20 Norwegian spring wheat ˚ s, Norway, (Triticum aestivum L.) genotypes grown in A in 1999. The genotypes represented a relatively wide variation in protein quality, as well as variation in endosperm hardness from relatively soft to hard wheats. The material included a few unusual combinations of protein composition and hardness, e.g. soft flour with
strong mixing properties, and hard flour with weak mixing properties. Analyses of glutenin and gliadin compositions for the 20 genotypes have been published elsewhere (Flæte, 2000, 2001). The effects of fertiliser levels and year-to-year variation have also been published for these cultivars grown in two earlier seasons (Uhlen et al, 2000; Tronsmo et al., 2003b). The wheat samples were cleaned and tempered before milling on a laboratory mill (Brabender Quadrumat Senior, Model 8802, Duisburg, Germany). 2.2. Grain and flour analyses Kernel hardness was analysed by near infrared reflectance (NIR) using an Infralyzer 500 (Bran and Luebbe, Germany) calibrated for particle size index (PSI) (using a calibration developed at Matforsk, Norway). SDS sedimentation volume (SDS) was determined according to AACC approved method 56– 70 on whole meal flour obtained by milling on a Falling Number 3100 hammer mill (Perten Instruments AB, Sweden). Protein content of flours was determined as Kjeldahl N £ 5.7 (expressed on dry matter basis) according to ICC standard no. 105. Flours were analysed by a Brabender Farinograph (ISO 5530-1) and the dough development time (DDT) determined for mixing at double speed (126 rpm) (Færgestad et al., 2000). Mixograms were recorded in duplicate on the 2-g direct drive Mixograph (National Manufacturing, Lincoln, NE), and the data collected and analysed by the software package MixSmart (National Manufacturing, Lincoln, NE, USA). Water was added according to Farinograph water absorption (500 BU) for each flour. Out of several parameters determined from the Mixograms, Mixogram peak time (MPT) and Mixogram peak height (MPH) were retained for the final analysis. Flours were sequentially extracted in an SDS solution without and with sonication, and the extracts were analysed by size-exclusion high performance liquid chromatography (SE-HPLC) as described earlier (Tronsmo et al., 2002). The chromatogram of the non-sonicated fraction showed 4 main peaks, denoted F1 –F4, and some minor peaks at higher Mr grouped together under the denomination F5. The sonicated fraction showed one main peak eluting at the void volume of the column, denoted F1*. In addition to determining the proportions of each fraction, the proportion of unextractable polymeric protein (%UPP) was calculated as %UPP ¼ F1* =ðF1* þ F1Þ* 100; analogous to the definition by MacRitchie and Gupta (1993). 2.3. Baking 2.3.1. Hearth bread Hearth bread was produced with complete replicates for each flour, and the baking order was randomised. Doughs were made from 300 g flour (14% moisture basis) with, as a
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
proportion (%) of the flour mass, 1% dry yeast (Saf-instant, S.I.Lesaffre, Marcq, France), 1.25% salt, 3.5% fat (Pals Matfett containing plant based fats and oils, A/S Pals, Oslo, Norway), 0.003% ascorbic acid, and water according to Farinograph absorption at 500 BU. Doughs were mixed to peak in a Farinograph equipped with a 300 g-bowl operated at 126 rpm to a final dough temperature of 27 ^ 0.5 8C as described earlier (Færgestad et al., 2000). Fermentation (10 min), rounding, proving and baking conditions were as in previous studies (Færgestad et al., 1999, 2000), but the three pieces obtained from each dough were given different proving times: 35, 45 and 55 min. Analyses of bread loaf characteristics were done as described previously (Færgestad et al., 1999). They included the determination of loaf volume (rapeseed displacement method), form ratio (height/ width), crumb texture (Dallmann’s pore table) (Dallmann, 1958), and the baker’s subjective evaluations of overall outer appearance and brittles on the crust. 2.3.2. Chorleywood bread process Pan loaves were produced at Campden and Chorleywood Food Research Association using the CBP (Axford et al., 1963). Flours no. 8, 17, 19 and 20 were baked without replication owing to limited availability of sample material, but the remaining samples were baked twice. The baking order was randomised. Doughs were made from 1400 g flour (14% moisture basis) with, as a proportion (%) of the flour mass, 2.5% compressed yeast, 2.0% salt, 1.0% fat (Ambrex, slip point c. 45 8C), 0.01% ascorbic acid, and water according to Farinograph absorption at 600 BU. Based on flour Falling Numbers, levels of alpha-amylase activity were adjusted to 80 Farrand units by addition of fungal alpha-amylase. Doughs were mixed in a Morton high-speed mixer (300 rpm) to a work input of 40 kJ/kg under atmospheric pressure. Dough temperatures were 30.5 ^ 1 8C. Unlidded pan loaves were produced from 454 g dough pieces as described earlier (Alava et al., 2001) involving 10-min fermentation and final proof to a constant height of 10 cm. Dough handling properties were assessed by the bakers after mixing, and oven rise was determined from the average height of the four loaves immediately after baking. After storage for 24 h at 21 8C, assessments were made of loaf volume (seed displacement), crumb colour (Minolta CR310 colorimeter) and crumb texture (expert assessment on a scale from 1 to 10, where the highest score would be given to a loaf considered as perfect, with a fine, even pore structure).
207
times as determined by Farinograph at 126 rpm. Salt and ascorbic acid were added at the levels used for baking of hearth breads, and the water bath temperature regulated to obtain dough temperatures of 27 ^ 0.5 8C. Doughs were rested for 10 min at room temperature before being rolled out. Discs were stamped out of the dough sheet, oiled, covered, and left to rest for 35 min. The discs were pressed in a sample retainer to a thickness of 2.67 mm and 55 mm diameter, and left covered for 10 min before inflation. The results were plotted in two ways, as pressure as a function of drum distance, and stress as a function of Hencky strain. The following parameters were read from the curves: tenacity, or maximum pressure ðPÞ; the stress and strain at which the maximum bubble pressure occurred (MP stress and MP Strain), the stress, strain and drum distance at the time of burst (burst stress, burst strain and L; respectively) and the deformation energy ðWÞ; obtained from the area under the pressure-distance curve. The strain hardening index ðnÞ was calculated by fitting an exponential curve to the stress – strain inflation curve between peak pressure and peak stress as described earlier (Tronsmo et al., 2003a). 2.5. Statistical analyses Analysis of variance, Tukey’s test for pairwise comparisons and simple correlations were carried out by the general linear model using Systat version 9.01 (SPSS, Inc. 1998). Only variables showing significant variation among cultivars were included in the subsequent multivariate analyses using the software Unscrambler v7.6 (Camo ASA, 2000). Partial least squares (PLS) regressions were performed using flour and dough characteristics as predictors (x-variables) and one or more bread characteristics from the two baking processes as response variables (y-variables). The variables were weighted to equal variance. Both full cross validation and random validation set-up was used for evaluating the models. Significant variables in the PLS predictions were identified by the modified Jack-knife uncertainty test (Martens and Martens, 2000), and the models were recalculated using only the significant variables. More detailed explanations of the methodology have been included in earlier cereal related studies (Færgestad et al., 2000; Tronsmo et al., 2002).
3. Results and discussion 2.4. Extensional rheology 3.1. Flour properties Biaxial extensional rheological measurements were performed with the Dobraszczyk/Roberts (D/R) Dough Inflation System (Stable Microsystems, UK) (Dobraszczyk and Roberts, 1994; Dobraszczyk, 1997). Doughs were prepared in a Farinograph operated at 126 rpm, using Farinograph water absorption (500 BU) and mixing
Selected analysis variables (Table 1) show that the sample set represented significant variation in protein quality. The proportion (%) of unextractable polymeric protein ranged from 24.3 to 45.1%, MPT from 2.1 to 4.8 min, and strain hardening index from 1.03 to 1.60.
208
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
Hardness ranged from 45.8 to 63.6 (PSI). Flours with a PSI below 55 were considered as hard, and flours of PSI . 55 as soft to conform with the classification done on the same cultivars from an earlier growing season, where the same groups were obtained by setting the limit at 40. Protein contents were low, ranging from 8 to 11% (dry matter basis).
situated among the hard, strong samples. This indicates that hardness was of less importance than protein quality for the baking results of these experimentally milled flours. In the corresponding loading plot (Fig. 1(b)), the x-variables known to be associated with protein quality, such as strain hardening index ðnÞ; unextractable polymeric protein (%UPP) and MPT, had high positive loadings along the first PLS axis. Only the first PLS factor was significant for the prediction, and the plot should therefore be interpreted in one dimension only. In materials designed to represent significant variation in protein content independently of the variation in protein quality, the second PLS factor has been associated with the effect of protein content (Færgestad et al., 2000; Tronsmo et al., 2002, 2003b). In the present material, variability in protein content had a minor influence on baking properties. Hearth loaf characteristics (volume, form ratio and porosity), as well as the crumb score and crumb colour of CBP loaves, were all located to the right of the plot, positively related to mixing properties, strain hardening index, SDS sedimentation volume, HMW-GS 5 þ 10 and unextractable polymeric protein (%UPP). In accordance with earlier studies (Færgestad et al., 2000), the form ratio of hearth loaves decreased and loaf volume increased with prolonged proving times. An optimal proving time would thus be the time giving the highest possible volume with an acceptable form ratio. As an approach to optimise proving in the present study, the ‘best’ of the three proving times were chosen for each sample. The average form ratio for the entire sample set was 0.58. Final
3.2. Prediction of bread characteristics The PLS regression model predicting bread characteristics for the two baking processes from flour and dough properties is presented in Fig. 1. Results for hearth bread made from doughs that had been proved for 45 min were used. The model explained between 69% and 84% of the variation in hearth bread characteristics and 52 –54% of CBP crumb characteristics. The explanation of CBP oven spring was weak and barely significant (24%), and the variation in CBP volume could not be explained (Table 2). When predicting only one response variable at a time, slightly higher figures were obtained, as shown in the last columns of Table 2. In the score plot of the samples (Fig. 1(a)), varieties containing the HMW glutenin subunits 5 þ 10 were located on the right along the first PLS factor, and those containing the subunits 2 þ 12 on the left. Samples of unusual combinations of hardness and protein quality were situated as expected from their protein quality. The hard, weak flour 15 was situated among the soft, weak samples, and the soft samples containing HMW-GS 5 þ 10 (flours 7, 19, 20) were
Table 1 High molecular weight glutenin subunit (HMW-GS) composition, protein content (dry matter basis), hardness, proportion of unextractable polymeric protein (%UPP), Mixogram peak time (MPT) and strain hardening index for the 20 flours Flour
HMW-GS
Protein content (%)
Hardness (PSI)
%UPP
MPT
Strain hardening index
01. Bastian 02. Tjalve 03. Polkka 04. Brakar 05. NK93608 06. T1022 07. NK93533 08. T9042 09. T92596-1 10. T92596-2 11. T91545-1 12. T91546 13. T91545-2 14. T91545-3 15. T92601-1 16. T92601-2 17. T92601-3 18. T92584 19. T91549-1 20. T91549-2
5 þ 10 5 þ 10 2 þ 12 5 þ 10 5 þ 10 2 þ 12 5 þ 10 5 þ 10 5 þ 10 5 þ 10 5 þ 10 5 þ 10 5 þ 10 5 þ 10 2 þ 12/5 þ 10 2 þ 12 2 þ 12 2 þ 12 5 þ 10 5 þ 10
10.3 9.7 9.8 9.6 8.0 9.4 9.8 9.9 9.0 8.5 9.4 10.0 9.8 10.9 8.5 8.3 9.3 9.7 10.1 10.2
47.3 45.8 48.8 51.5 48.0 62.0 56.7 52.9 47.4 49.1 48.2 48.0 47.1 53.5 54.4 59.4 62.5 60.7 62.4 63.6
37.8 42.3 32.6 41.8 34.2 24.7 37.0 36.6 38.5 36.6 39.4 40.9 45.1 43.3 25.3 28.0 26.1 24.3 35.0 33.2
3.3 3.5 3.3 4.0 3.6 2.4 3.8 3.5 3.6 3.7 3.3 3.0 4.8 4.2 2.7 2.7 2.9 2.1 3.3 3.3
1.52 1.57 1.47 1.57 1.29 1.15 1.43 1.49 1.50 1.39 1.45 1.47 1.59 1.60 1.03 1.07 1.15 1.18 1.45 1.40
3.4
0.4
0.03
LSD LSD, least significant differences.
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
209
Fig. 1. Partial least squares (PLS) regression predicting bread characteristics from flour and dough analyses. (a) Score plot of samples indicated by flour numbers followed by letters S or H (soft: PSI . 55 or hard: PSI , 55) and W or S (weak or strong gluten quality predicted from presence of HMW-GS 2 þ 12 vs 5 þ 10) (b) X-loading weights and Y-loadings (bold). Bread characteristics are indicated by the suffix H for hearth breads and by CBP for breads made by the Chorleywood bread process. Abbreviations are explained on the first page of the paper.
proof giving the form ratio closest to this value was chosen, and the volume at this proving time was included as ‘Volume opt. H’ in the PLS regression. When included in the model with the other y-variables, the explained validation variance of this volume at ‘optimal’ proving was 83% by one PLS factor, with a correlation coefficient between predicted and measured values of r ¼ 0:90: A model predicting only this volume gave an explained validation variance of 89% ðr ¼ 0:94Þ; using F1*, F1, %UPP, SDS, PSI and DDT as x-variables after employing
the uncertainty test to eliminate non-significant predictors. The simple correlation between the volume at ‘optimal’ proving and %UPP was r ¼ 0:93: 3.3. Dough inflation Analysis of variance showed that there was significant variation among cultivars for all the recorded variables of the dough inflation curves. Among these, the strain hardening index ðnÞ had the highest loading along the first
210
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
towards deformation. This implies that in an expanding gas bubble, the thinnest regions will be the stiffest, resisting further stretching. Consequently, stretching will occur preferentially in the thicker regions. This results in selfrepair of defects, and allows the bubble to reach greater expansion before rupture (Dobraszczyk and Roberts, 1994; Dobraszczyk, 1997; van Vliet et al., 1992). In a dough with a high strain hardening index, the gas cells will be able to expand to larger volumes before rupturing or coalescing with neighbouring bubbles. Assuming that the results from the expansion of large single bubbles can be extrapolated to the numerous small bubbles expanding and interacting in a dough, this method is a more direct measure of dough properties than for instance the Extensigraph test, which operates in uniaxial extension, as opposed to the biaxial extension experienced by the membranes of expanding bubbles.
Table 2 Explained validation variances (EVV) and correlation coefficients between predicted and measured values ðrÞ for y-variables predicted from flour and dough characteristics by partial least squares (PLS) regression using 1 PLS factor
Volume H Vol opt. H Form ratio H Crumb H Brittles H Overall H Volume CBP Oven spring CBP Crumb colour CBP Crumb score CBP
Combined PLS model
Single PLS models
EVV
r
EVV
r
84.3% 83.1% 70.1% 68.5% 74.7% 70.5% 26.5% 23.7% 54.2% 52.1%
0.91 0.90 0.82 0.81 0.85 0.82 20.25 0.42 0.70 0.69
92.6% 89.3% 80.4% 71.3% 88.9% 82.1%
0.96 0.94 0.88 0.83 0.94 0.90
27.0% 54.6% 54.3%
0.46 0.71 0.71
3.4. Size distribution of flour proteins
PLS factor, and thus could be taken as the variable best reflecting protein quality. The deformation energy ðWÞ; the stress, strain and drum distance at the time of burst (burst stress, burst strain and L; respectively) and the strain at which the maximum bubble pressure occurred (MP Strain) also had high loadings along PLS factor 1. The tenacity (P; maximum pressure during bubble inflation) was not significant for the prediction, and the stress at the same point (MP Stress) had a smaller loading along the first PLS factor. The greatest strain hardening index was found for cultivar no. 14, and the lowest for cultivar 15 (Table 1). Most samples containing HMW-GS 2 þ 12 had significantly lower strain hardening indexes than the cultivars containing HMW-GS 5 þ 10. The exception was cultivar no. 3 (Polkka), which contains the subunit pair 2 þ 12 but had a high strain hardening index, indicating that other proteins give this sample improved rheological properties. The simple correlation between strain hardening index and hearth loaf volume was r ¼ 0:89; and there were also significant correlations with other hearth and CBP bread characteristics (Table 3). The phenomenon of strain hardening is believed to be essential for the expansion of gas cells in dough. It is manifested by a curvature in the stress –strain relationship, such that at higher strains, there is a greater resistance
Analysis of variance revealed significant variation among cultivars for the proportions of the peak areas of fractions F1*, F1, F3 and F4, as well as the proportion of unextractable polymeric protein (%UPP). There were no significant variations among cultivars for the proportions of the smaller fractions F2 and F5 ðp . 0:05Þ: Fraction F2 appeared only as a shoulder on the peak of fraction F3, and fraction F5 represented multiple small late-eluted peaks (Tronsmo et al., 2002), where varietal differences would hardly surpass the experimental error. The percent unextractable polymeric protein appeared with a high loading in the multivariate analysis (Fig. 1), showing this to be a good indicator of protein quality. Values for dough properties, such as those obtained from the dough inflation test, may depend not only on the flour quality, but also the amounts of other dough ingredients, mixing conditions and resting time, and may thus be biased when correlated with variables from different baking processes. As SE-HPLC analyses were performed directly on flour extracts, %UPP is an unbiased variable for comparing with both baking tests. Scatter plots and linear regression for %UPP versus the most important bread characteristics showed that the relationships were more linear with the characteristics of hearth loaf characteristics than those of CBP loaves (Fig. 2). The simple
Table 3 Simple correlations ðrÞ between selected protein quality test variables and bread characteristics Hearth bread
SDS %UPP MPT N DDT
CBP bread
Hearth volume
Form ratio
Crumb texture
CBP volume
Oven spring
Crumb colour
Crumb score
0.94*** 0.92*** 0.69** 0.89*** 0.83***
0.85*** 0.91*** 0.84*** 0.80*** 0.66**
0.79*** 0.82*** 0.56* 0.88*** 0.72***
ns ns ns ns ns
0.51* 0.59** 0.46* 0.61** ns
0.71** 0.74*** 0.63** 0.70** 0.61**
0.77*** 0.78*** 0.66** 0.82*** 0.59**
ns, non significant; *p , 0:05; **p , 0:01; ***p , 0:001:
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
211
Fig. 2. Plots of proportion (%) of unextractable polymeric protein (%UPP) versus bread characteristics for CBP loaves (left) and hearth loaves (right). Correlation coefficients are found in Table 3.
212
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
correlation of %UPP with hearth loaf volume was r ¼ 0:92; with form ratio r ¼ 0:91 and with crumb texture r ¼ 0:82: %UPP showed no significant correlation with CBP volume, but a correlation coefficient of r ¼ 0:59 with oven spring, r ¼ 0:74 with crumb colour and r ¼ 0:78 with crumb score (Table 3). 3.5. Hearth loaf characteristics In the baking of hearth loaves, high volumes and form ratios and good crumb and crust characteristics were obtained for samples with strong gluten quality, as assessed by variables such as MPT, strain hardening index ðnÞ and unextractable polymeric protein (%UPP). Photographs of samples 18, 7 and 14 (Fig. 3(a)) illustrate the range in differences in loaf volume and form ratio for the 45 min proving time. 3.6. CBP loaf characteristics CBP loaves made from the same flours are shown in Fig. 3(b). Whereas for hearth breads, flour 14 gave the highest volume of all the samples, for CBP loaves flour 7 gave the highest volume. No significant correlations could be found between x-variables and CBP volume, and the simple correlations with oven spring were low. Crumb characteristics of CBP loaves showed stronger relationships with protein quality variables (Table 3). The CBP was developed with moderately high protein content (around 12 –12.5%, 14% moisture basis) ‘baker’s’ flours, which, at the time (early- to mid-1960s), were milled from grists containing considerable amounts of strong wheat from North America and other countries and which were then being used to produce bread commercially by
bulk fermentation procedures. It was quickly realised that bread of similar quality to that produced from those moderately high protein content flours could be produced in the CBP using flours of somewhat lower protein content (from 11.0 to 11.5%). This led to considerably larger proportions of UK-grown, lower protein content wheats of moderate strength being used in bread grists for flours intended for use in the CBP. Developments in enzyme improver technology (use of high levels of fungal alphaamylase and use of microbial arabinoxylanases) and flour protein supplementation using vital wheat gluten have accentuated that trend. To test whether medium strong wheats could be expected to perform better in the CBP than strong or weak flours, square effects were included in the PLS model. This produced relatively little improvement in the prediction of CBP crumb characteristics. Prediction of oven spring increased somewhat, however, from an explained validation variance from 27 to 49%. Plotting oven spring against the percentage of unextractable polymeric protein, the curved effect was not obvious (Fig. 2). A model with square effects could account for only 13% of the variation in loaf volume, which was below a significant level. The poor explanation of CBP loaf volume did not result from statistical error. Variance components from a nested analysis of variance showed that 77% of the variation in volume was associated with differences between flours, 16% with the variation between replicate doughs, and 7% with loaf volume measurements of breads from the same dough. The latter was thus less than one tenth of the variance component for flour. The simple correlation between oven rise and loaf volume was r ¼ 0:88: The crumb structure, both as evaluated by colour measurement and expert assessment, reflected variability
Fig. 3. Breads made from flours no 18, 7 and 14, ranked from low to high values of Mixogram peak time (MPT), strain hardening index (n) and unextractable polymeric protein (%UPP). (a) Hearth breads (from 150 g dough pieces) (b) CBP breads (from 454 g dough pieces).
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
in protein quality. The subjective evaluation requires a skilled person to perform the assessment of all samples to be compared. The crumb colour, which can be measured instrumentally, may be more convenient for routine use. Further information on crumb characteristics can be acquired using image analysis methods (Alava et al., 2001). 3.7. Comparison of results from the two baking procedures Whereas the volumes of hearth loaves showed an increasing trend with increased proportions of high molecular weight glutenin polymers (Fig. 4), the volumes of CBP loaves showed no relationship with either glutenin polymer size or mixing strength. The difference in specific volume between the smallest and largest loaf was small for CBP loaves (0.78 ml/g), compared with hearth loaves (1.36 ml/g). As the variation in protein content of this material was limited, the observed effects on baking quality were mainly related to protein quality, and not to protein content. The evaluations of crumb characteristics for the two baking methods showed common traits. They were located in the same area of the PLS plot. The simple correlations between the crumb texture of hearth breads and the crumb characteristics of CBP breads were r ¼ 0:66 for CBP crumb colour and r ¼ 0:68 for CBP crumb score ðp ¼ 0:001Þ: The three variables were all predicted by the same x-variables, with significant contributions from all the variables included in the PLS model in Fig. 1 except from %F4. 3.8. Relating process variables to differences in bread characteristics There were several differences between the two baking processes that may have contributed to the different results
213
for loaf volume by the two baking tests. These include mixing intensity, mixing time, dough temperatures and proving times, each of which will be discussed below. It has been shown that to reach optimal dough development, the mixing intensity must be above a minimum critical level dependent on flour quality and mixing action, and work input must be above a critical value for the specific flour (Kilborn and Tipples, 1972). As Z-blade mixers were used for both processes, the mixing actions were comparable. However, the Morton mixer used for the CBP was operated at higher speed (300 rpm) compared with the Farinograph used for hearth loaf preparation (126 rpm). An earlier baking test comparing mixing in a Farinograph operated at 126 rpm and at 63 rpm, the latter being the normal speed for the Farinograph test, showed that, at the lower speed, flours of strong protein quality were not fully developed (Færgestad et al., 2000). The present results, showing good correlations between protein quality and characteristics of hearth breads, indicate that the mixing at 126 rpm is above the critical level of intensity needed to exploit the full potential of the flour. It should also be noted that results from this small-scale baking test using Farinograph operated at 126 rpm have been shown to correspond well to results obtained in commercial scale baking using spiral mixers (unpublished results). Whereas optimal mixing times were used for hearth loaves, doughs were mixed to constant work input in the CBP process. As strong samples have greater resistance to mixing, the energy transfer to the dough is more rapid, and the mixing time to constant work input is shorter than for weak samples. In the baking process used for hearth loaves, where doughs were mixed to optimal dough development, the inverse was true: strong doughs required longer mixing times than weak doughs. Whereas the mixing times used for
Fig. 4. Specific volumes of hearth loaves proved for 45 min (B) and CBP loaves proved to constant height (33–42 min) (A). Flours are sorted according to the proportion of unextractable polymeric protein (%UPP). Error bars indicate standard deviation between doughs mixed in duplicate.
214
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
hearth bread doughs ranged from 1.2 to 4.0 min, the CBP mixing times ranged only from 3.2 to 3.5 min. There was a slight negative correlation between mixing times from the two processes ðr ¼ 20:57; p ¼ 0:009Þ: The mixing to constant work input by the CBP generally resulted in longer mixing times than for the hearth loaf baking, except for flour no 14. The CBP, using longer mixing times combined with a higher mixing speed, was thus likely to cause overmixing of the weakest doughs. Dough assessment after CBP mixing classified some samples (no. 1, 5, 6, 10, 15, 16, 17 and 18) as softer than optimal, which might indicate overmixing. These were samples with relatively short DDT in the Farinograph operated at 126 rpm. The only sample with a shorter CBP mixing time than its Farinograph DDT at 126 rpm, flour 14, was assessed as slightly tighter than normal after CBP mixing. It is generally known that to determine the full baking potential of a flour, an optimised baking test is required (MacRitchie, 1984). The Chorleywood process adapts mixing time for each flour with respect to a constant work input, which was originally calculated as an average of the individual optimum work inputs of around 60 commercial flours, and which was found to produce acceptable results from a broad range of flour types. If the aim was to obtain the full baking potential of a flour, however, the total work input would have to be optimized (Alava et al., 2001; Kilborn and Tipples, 1972). Three different constant proving times (35, 45 and 55 min) were used for hearth breads, whereas the CBP loaves were proved to a fixed height. These proving times varied from 33 to 42 min. Because of the fixed proving height, the differences in final volume observed were mostly a result of volume expansion during oven rise. Some prediction was obtained for oven spring, which is calculated from the height measured immediately after baking, whereas no prediction was obtained for loaf volume. The CBP specific loaf volumes (proof times 33 – 42 min) of all flour samples were greater than the specific loaf volumes of hearth breads proved for 45 min (Fig. 4). The intensive and long mixing, as well as the higher temperature in dough after mixing and in the proving cabinet, may have contributed to the rapid volume increase for CBP loaves. The dough temperature after mixing was 27 ^ 0.5 8C for hearth loaves, and 30.5 ^ 1.0 8C for CBP loaves. Hearth bread doughs were proved at 37.5 8C, and CBP doughs at 43 8C. The higher temperatures are likely to have contributed to increased rates of CO2 production and changes in dough structure during the fermentation process for CBP loaves, causing a more rapid increase in volume. The CBP was not developed specifically as a test method for distinguishing between varieties, but rather as a rapid process for the baking industry for producing acceptable loaves from flours covering a range of dough strengths and protein contents (Axford et al., 1963). The intensive mixing
and the use of high amounts of ascorbic acid probably contribute to masking differences between flours, resulting in a robust baking process with regards to variability in flour quality. This is undoubtedly an advantage for the baking industry. The process can be used for determining whether a flour can perform above a minimum acceptable level, but the process is less suitable for assessing the baking potential of wheat flours. Baking tests are often given large importance in wheat quality testing, and loaf volume is regarded as the key characteristic. Correlation studies of new quality tests versus baking characteristics are carried out without necessarily knowing whether the baking test has the potential for discerning between samples of weak and strong protein quality. The present results show that the baking process, as well as the bread characteristics used to evaluate breadmaking performance, must be chosen with care. The hearth bread test, using optimal mixing and fixed proving, has been tested on different wheat materials, and has been shown to give good distinctions between flours of differing protein quality (Færgestad et al., 2000; Tronsmo et al., 2002, 2003b). In materials expressing variation in protein contents as well as in protein quality, the method also permits distinction between the effects of protein content and quality (Færgestad et al., 2000; Tronsmo et al., 2002, 2003b).
4. Conclusions The present results underline that the choice of baking procedure is vital in assessing flour quality. The material included varieties of widely differing protein quality. In the CBP, the protein quality was not important for the loaf volume. As loaf volume is often considered the key quality tests against which other tests are evaluated, this is an important result. For the purpose of assessing differences in the protein quality of flours, the CBP volume was unsuitable as a reference. On the other hand, the crumb structure of CBP loaves, assessed both by colour measurement and subjective evaluation, was dependent on protein quality. However, stronger relationships were found between the flour and dough properties tested and the characteristics of hearth loaves. Differences in protein quality influenced both loaf volume, form ratio, crumb and crust structure of hearth bread, and up to 93% of the variation in loaf volume could be predicted from dough properties and protein size distribution. Optimisation of process parameters in the CBP, in particular work input during mixing and perhaps proving times, might be expected to give stronger relationships between protein characteristics and baking properties. The dough inflation test was shown to give good indications of baking quality. Particularly the strain hardening index of the inflating bubble was a good indicator of breadmaking quality, with a simple correlation with hearth loaf volume of r ¼ 0:89:
K.M. Tronsmo et al. / Journal of Cereal Science 38 (2003) 205–215
Acknowledgements The authors would like to thank Dr Anne Kjersti Uhlen for providing the wheat material, the bakers Hans Helge Raae Olsen and John Tore Syversen at MATFORSK for performing of the hearth bread baking, and Kim Little and the bakers Dereck Buttler and Andrew Keene at Campden and Chorleywood Food Research Association for performing the CBP baking experiment. This project was supported by a research grant from The Research Council of Norway (grants no. 118028/112 and 127202/130).
References Alava, J.M., Millar, S.J., Salmon, S.E., 2001. The determination of wheat breadmaking performance and bread dough mixing time by NIR spectroscopy for high speed mixers. Journal of Cereal Science 33, 71–81. Axford, D.W., Chamberlain, N., Collins, T.H., Elton, G.A.H., 1963. The Chorleywood process. Cereal Science Today 8, 265–270. Brett, G.M., Mills, E.N.C., Tatham, A.S., Fido, R.J., Shewry, P.R., Morgan, M.R.A., 1993. Immunochemical identification of LMW subunits of glutenin associated with bread-making quality of wheat flours. Theoretical and Applied Genetics 86, 442 –448. Dallmann, H., 1958. Porentabelle, Moritz Scha¨fer, Detmold, Germany. Dobraszczyk, B.J., 1997. Development of a new dough inflation system to evaluate doughs. Cereal Foods World 42, 516–519. Dobraszczyk, B.J., Roberts, C.A., 1994. Strain hardening and dough gas cell-wall failure in biaxial extension. Journal of Cereal Science 20, 265–274. Færgestad, E.M., Magnus, E.M., Sahlstro¨m, S., Næs, T., 1999. Influence of flour quality and baking process on hearth bread characteristics made using gentle mixing. Journal of Cereal Science 30, 61–70. Færgestad, E.M., Molteberg, E.L., Magnus, E.M., 2000. Interrelationships of protein composition, protein level, baking process and the characteristics of hearth bread and pan bread. Journal of Cereal Science 31, 309 –320. Finney, K.F., Barmore, M.A., 1948. Loaf volume and protein-content of hard winter and spring wheats. Cereal Chemistry 25, 291–312. Flæte, N.E.S., 2000. Allelic variation at the storage protein loci (Glu-1, Glu-3 and Gli-1) in Norwegian wheats (Triticum aestivum L.). Journal of Genetics and Breeding, 283–291. Flæte, N.E.S., 2001. Identification of gliadins and low-molecular-weight glutenin subunits in wheat (Triticum aestivum L.) and their relation to protein quality. Dr Scient. Thesis, Agricultural University of Norway, Department of Horticulture and Crop Sciences Kasarda, D.D., 1989. Glutenin structure in relation to wheat quality. In: Pomeranz, Y., (Ed.), Wheat Is Unique: Structure, Composition, Processing, End-use Properties and Products, American Association of Cereal Chemists, St Paul, MN, pp. 277 –302. Kilborn, R.H., Tipples, K.H., 1972. Factors affecting mechanical dough development. I. Effect of mixing intensity and work input. Cereal Chemistry 49, 34–47. MacRitchie, F., 1984. Baking quality of wheat flours. In: Chichester, C.O., (Ed.), Advances in Food Research, Academic Press, London, pp. 201–277.
215
MacRitchie, F., 1992. Physiochemical properties of wheat proteins in relation to functionality. In: Kinsella, J.E., (Ed.), Advances in Food and Nutrition Research, Academic Press, London, pp. 1–87. MacRitchie, F., Gupta, R.B., 1993. Functionality-composition relationships of wheat-flour as a result of variation in sulfur availability. Australian Journal of Agricultural Research 44, 1767–1774. Magnus, E.M., Bra˚then, E., Sahlstro¨m, S., Færgestad, E.M., Ellekjær, M.R., 1997. Effects of wheat variety and processing conditions in experimental bread baking studies by univariate and multivariate analyses. Journal of Cereal Science 25, 289–301. Martens, H., Martens, M., 2000. Modified Jack-knife estimation of parameter uncertainty in bilinear modelling by partial least squares regression (PLSR). Food Quality and Preference 11, 5 –16. Payne, P.I., Nightingale, M.A., Krattiger, A.F., Holt, L.M., 1987. The relationship between HMW glutenin subunit composition and the bread-making quality of British-grown wheat varieties. Journal of the Science of Food and Agriculture 40, 51– 65. Pomeranz, Y., 1988. Composition and functionality of wheat flour components. In: Pomeranz, Y., (Ed.), Wheat: Chemistry and Technology, American Association of Cereal Chemists, St Paul, MN, pp. 219– 370. Roels, S.P., Cleemput, G., Vandewalle, X., Nys, M., Delcour, J.A., 1993. Bread volume potential of variable-quality flours with constant protein level as determined by factors governing mixing time and baking absorption levels. Cereal Chemistry 70, 318 –323. Shewry, P.R., Halford, N.G., Tatham, A.S., 1992. High molecular weight subunits of wheat glutenin. Journal of Cereal Science 15, 105 –120. Singh, N.K., Donovan, R., MacRitchie, F., 1990. Use of sonication and size-exclusion high-performance liquid chromatography in the study of wheat flour proteins. II. Relative quantity of glutenin as a measure of breadmaking quality. Cereal Chemistry 67, 161–170. Sontag-Strohm, T., Payne, P.I., Salovaara, H., 1996. Effect of allelic variation of glutenin subunits and gliadins on baking quality in the progeny of two biotypes of bread wheat cv Ulla. Journal of Cereal Science 24, 115–124. Southan, M., MacRitchie, F., 1999. Molecular weight distribution of wheat proteins. Cereal Chemistry 76, 827 –836. ˚ ., Schofield, J.D., Magnus, Tronsmo, K.M., Færgestad, E.M., Longva, A E.M., 2002. A study of how size distribution of gluten proteins, surface properties of gluten and dough mixing properties relate to baking properties of wheat flours. Journal of Cereal Science 35, 235– 248. Tronsmo, K.M., Magnus, E.M., Baardseth, P., Schofield, J.D., Aamodt, A., Færgestad, E.M, 2003. Cereal Chem. (in press). Comparison of small and large deformation rheological properties of wheat dough and gluten. Tronsmo, K.M., Magnus, E.M., Færgestad, E.M., Schofield, J.D, 2003. Cereal Chem. (in press). Relationships between gluten rheological properties and hearth loaf characteristics. Cereal Chem. (in press). Uhlen, A.K., Magnus, E.M., Færgestad, E.M., Sahlstro¨m, S., Ringlund, K., 2000. Effects of genotype, N-fertilisation, and temperature during grain filling on baking quality of hearth bread. In: Shewry, P.R., Tatham, A.S. (Eds.), Wheat Gluten (Proceedings of the 7th International Workshop on Wheat Gluten Proteins), Royal Society of Chemistry, Cambridge, UK, pp. 484 –487. van Vliet, T., Janssen, A.M., Bloksma, A.H., Walstra, P., 1992. Strain hardening of dough as a requirement for gas retention. Journal of Texture Studies 23, 439 –460.