Development of gluten‐free bread formulations containing whole chia flour with acceptable sensory properties

Abstract Increasing the variety of better‐tasting and healthier gluten‐free products is important for consumers with gluten‐related disorders. This work aimed to develop a gluten‐free bread formulation containing whole chia flour with acceptable sensory properties. A mixture design for three ingredients and response surface methodology were used to identify the proportions of potato starch, rice flour and whole chia flour to achieve the best physical properties and result in sensory‐accepted products. The physical properties and visual appearance showed that whole chia flour alone is not suitable for bread production. Nevertheless, it is possible to add up to 14% whole chia flour to a rice flour‐based gluten‐free bread formulation while negligibly diminishing the loaf volume, crumb firmness and crumb moisture. The best formulations were prepared from rice flour blends with 5, 10, and 14% whole chia flour, which received overall acceptability scores of 8.7, 8.1 and 7.9 on a 10‐cm scale, respectively, similar to those of their white gluten‐free bread and wheat bread counterparts. Incorporating 5%–14% whole chia flour in the formulation increased the levels of ash, lipid, protein and dietary fiber compared to those of the white gluten‐free bread.

challenges imposed by a strict gluten-free diet (Capriles, Santos, & Arêas, 2016). Thus, more research and development are required to increase the variety of better-tasting and healthier GF products. This can be done by incorporating natural raw materials rich in nutrients and bioactive compounds, such as chia seed, into GFB formulations (Capriles, Santos, & Arêas, 2016;Torres, Arufe, Chenlo, & Moreira, 2017).
Thus far, few studies have been performed on the use of chia seeds in GF bread-making. Torres (2012, 2013) incorporated 2.5%-7.5% whole chia flour (WCF) into a gluten-free chestnut flour-based dough. These authors concluded that the addition of 7.5% WCF improved the dough rheological properties of stability, viscosity and elasticity. Costantini et al. (2014) replaced common and tartary buckwheat flour with 10% WCF, and they observed an improvement in the protein, lipid, dietary fiber, ash, α-linolenic acid, and phenolic compound contents as well as in the antioxidant capacity of the formulations. Steffolani, de la Hera, Perez, and Gomez (2014) observed that the replacement of rice flour with 15% WCF or 15% chia seeds darkened the GFB, reduced the specific volume, and increased the hardness, but it does not reduce the overall acceptability (for scores of approximately 5 -neither like nor dislike, on a 9-point hedonic scale). Huerta, Alves, Silva, Kubota, and Rosa (2016) replaced rice and soy flour with 2.5, 5.0 and 7.5% WCF, and they observed that bread with 2.5% WCF showed no significant differences in relation to the control for the specific volume and baking loss as well as for the color, aroma, taste, texture, and appearance acceptability (scores ranging from 4.5 to 5.5, on a 7-point hedonic scale).
These studies showed the potential use of WCF in GFB, but to the best of our knowledge, there have been no reports to date on the optimization of the WCF proportions in GFB formulations.
Considering that, the objective of this study was to use a mixture design to define the optimum WCF proportions in a GFB formulation with acceptable sensory properties. A mixture design for three ingredients and response surface methodology were used to identify the proportions of WCF in various blends with potato starch (PS) and rice flour (RF) achieving the best physical properties. Subsequently, the physical properties and sensory acceptability scores of the best formulations were compared to those of their white GFB and wheat bread counterparts.
A straight dough process was performed using a stand mixer (BPS-05-NSkymsen, Metalúrgica Siemsen Ltda., Brazil) with a paddle attachment. All ingredients were mixed at speed 4 (on a 1-10 mixer scale) for 4 min. The resulting dough (400 g) was then spread into previously greased and floured baking pans (19 × 7.5 × 5 cm) and proofed in a proofing chamber at 40°C and 85% relative humidity for 45 min (CFK-10, Klimaquip S/A -Tecnologia do Frio, Brazil). Baking was performed in an electric oven at 160°C for 22 min (HPE-80, Prática Produtos S.A., Brazil). After baking, the loaves were depanned and cooled for 2 hr on cooling racks at room temperature. The loaves were then stored in polyethylene bags to prevent moisture loss at room temperature (approximately 25°C). All analyses were performed within 3 hr.
Six loaves for each of the GFB trials were prepared from one batch. Three random loaves were used for the specific volume and crumb moisture analyses, and three random loaves were used for the crumb texture evaluation and photographs. An extra six loaves from the selected treatments were produced for sensory evaluation.

| Experimental design
The simplex-centroid design for mixtures of three ingredients was used to study the effects of pure and binary and tertiary mixtures of RF (x 1 ), PS (x 2 ), and WCF (x 3 ) on the physical properties of GFB.
The experiment was performed on three centroid point replications and included three axial points, for a total of twelve trials (Table 1), prepared using a previously randomized execution sequence (Cornell, 2002). The flour/starch blend corresponded to 35.8% of the dough for all formulations.
The crumb moisture was evaluated in triplicate, according to AACC method 44-15A (AACC 2010). The crumb firmness was determined according to AACC method 74-09 (AACC 2010) using a texture analyser (TA.XTplus, Stable Micro Systems, Surrey, UK). Texture measurements (six values) were performed on two bread slices that were taken from the centers of three different loaves.

| Physical property optimization and quality verification
The bread physical properties were used as response variables for the mixture design regression models. The following Scheffé canonical polynomial models were applied: where Y is the response variable; β 1 , β 2 , β 3 , β 12 , β 13 , β 23 , and β 123 are regression parameters; and x 1 , x 2 , and x 3 are the proportions of RF, PS, and WCF, respectively, in the flour/starch blend.
Each response in the linear model represents the effects of a pure ingredient. The quadratic model adds the effects of the binary mixtures, and the special cubic model includes the effects of the ternary blends.
Positive values for the binary coefficients (β 12 , β 13, and β 23 ) or the ternary coefficient (β 123 ) indicate synergistic effects, and negative values represent antagonistic effects between the ingredients (Cornell, 2002). Based on the regression model significance, contour plots were then produced to determine the optimal blend regions, and the best formulations were properly selected to achieve the best physical properties. These GFBs were prepared and experimentally analyzed, and the results were statistically compared to the predicted values from the fitted models.

| Sensory evaluation for acceptance
The sensory acceptability of the selected GFB formulations was evaluated by 50 untrained panellists (32 females and 18 males, aged 19-54 years) recruited from the campus via internal announcements. All the panelists agreed to taste the samples before the tests occurred and attested that they had bread-consuming habits and did not have any allergy or intolerance to any of the ingredients present in the products. They had no gluten-related disease and were made aware that they were tasting Informed consent was obtained from all individuals in the study.
The panellists scored the appearance, color, aroma, texture, taste, and overall acceptability of the formulations on a 10-cm hybrid hedonic scale (Villanueva, Petenate, & Da Silva, 2005). The bread slices (12.5 mm thick) were separately offered in a random sequence in polyethylene bags coded with 3-digit numbers. The evaluation was conducted in a climate-controlled (20°C-25°C) sensory evaluation laboratory equipped with separate booths. The panellists rinsed their mouths with water between samples to minimize any residual effects.

| Statistical analysis
Differences in treatment means were identified by one-way analy-  This negative effect of WCF on the GFB′s physical properties could be a consequence of the WCF particle size distribution and composition and also the hydration level effects on the dough properties.

| RESULTS AND DISCUSSION
The coarse WCF with bran particles probably disrupted the gas cells and starch gel uniformity in the dough, resulting in bread with a low specific volume and crumb softness. The formulation prepared with 66%-100% WCF (trials 3 and 12, Table 1) presented a higher dough consistency, making it difficult to mix and then incorporate gas cells during the mixing step. These effects likely occur because of the chia protein, dietary fiber, and mucilage water-binding capacity, and the starch dilution effect. These factors could limit starch swelling and gelatinization, which together with the bran particle effects impaired the GFB expansion, structure and texture (Capriles & Arêas, 2014).
Additionally, the high levels of fat present in the WCF may have implications for the GF dough and bread properties.
The water levels were fixed during this mixture design study. This variable could impair the GF dough and bread properties because increasing the amount of water is usually necessary in formulations that are enriched with fiber or fiber-rich flours. Increasing the amount of water allows for the adequate dough viscosity, starch gelatinization, and protein denaturation required during bread-making (Capriles & Arêas, 2014). Further studies could evaluate the effects of water level adjustments on chia-containing GF dough and bread.
It is possible to prepare a GFB made from 100% WCF. However, as shown in Figure 1 and Table 1, it is clear that using 100% WCF impaired the structure, texture, appearance, and color of GFB, and it also presented a poor mouthfeel and flavor. Similar technological limitations related to the use of whole-grain flour in GFB were reported by some researchers, including changes in the appearance, color, texture, aroma, and taste, which can easily impair consumer acceptability (Hager et al., 2012;Onyango, Unbehend, & Lindhauer, 2009;Schober, Messerschmidt, Bean, Park, & Arendt, 2005). Because of its own gray color, WCF darkened the GFB crumb. This darkening effect was also reported in other studies on baked products (Coelho & Salas-Mellado, 2015;Costantini et al., 2014;Pizarro et al., 2013;Steffolani et al., 2014).
The physical properties and visual appearance show that WCF alone is not suitable for bread production. Nevertheless, it was noted that the 17%-50% WCF blend with RF and PS resulted in a GFB with better physical properties and appearance (Table 1 and Figure 1).
The mixture regression models for the physical properties of GFB are given in Table 2. All the models were significant, and presented no lack of fit and high adjusted coefficients of determination (R 2 adj ), with 72% to 97% of the variation being explained by the models. These well-adjusted models were used to generate the contour plots for the physical properties of the GFB (Figure 2). Figure 2a shows that GFB prepared with higher proportions of RF and PS presented higher bake losses than those made with higher proportions of WCF. Because of antagonistic effects, the GFBs made from blends of RF and WCF present lower bake losses than breads made from the pure ingredients. Steffolani et al. (2014) also observed that the addition of WCF tended to produce a reduction in bake loss, and this effect can be related to a loaf volume with a lower surface area for exchange with the exterior and also to chia mucilage because of its water-holding capacity.
The loaf-specific volume was inversely correlated with the crumb firmness (r = −0.80, p < .01), and thus a lower loaf-specific volume results in a greater firmness because of the denser crumb and more compact cells. Figure 2b and c show that GFBs containing higher proportions of WCF exhibit a lower loaf volume and higher crumb firmness, while GFBs made with blends of RF and PS exhibit a higher volume and lower crumb firmness. The focus of this study was to verify the suitability of GFBs containing WCF. Considering that the loaf volume is directly related to the crumb softness and texture acceptance (Capriles & Arêas, 2014), promising formulations were selected considering the models fitted to these physical properties. GFB formulations prepared with blends of RF and WCF were selected from models presenting high loaf volume values and lower crumb firmness values, which could result in sensory-accepted products. Confirmatory experiments were performed, and the results show that the loaf volume and crumb firmness of GFBs made from RF blends with 5%, 10% and 14% WCF corresponded well with the predicted values. No differences were detected in the loaf volume or crumb firmness between these GFB formulations ( Table 3).
The results of the mixture design experiments showed that GFBs with good physical properties could be prepared with 5%, 10% and 14% WCF. These formulations present loaf volumes similar to those of two white GFBs, which were prepared with 100% RF and with 50% RF and 50% PS, but they had slightly higher crumb firmness values (trials 1 and 4 from Table 1).  Only the coefficients significant at the p < .05 level were selected for the predicted model construction.
3 R 2 adj adjusted coefficient of determination.
The GFB formulations made from RF blends with 5%, 10% and 14% WCF were accepted, with scores for appearance, color, aroma, texture, taste, and overall acceptability ranging from 7.3 to 8.7 on a 10-cm hybrid hedonic scale, as shown in Table 4. However, the GFBs containing 10% and 14% WCF presented darker crust and crumb colors, which diminish the appearance and color acceptability compared with those of the white GFBs that were prepared with 100% RF and with a 50% RF and 50% PS blend (fwb) and received sensory scores ranging from 8.2 to 8.5 according to the results recently reported by Capriles, Santos, and Arêas, (2016). No significant differences were observed between the aroma, texture, taste and overall acceptability scores of the chia-containing GFB and the white GFB, with scores ranging from 7.6 to 8.2, and neither with the standard wheat bread counterpart (scores ranging from 7.6 to 8.1) (Capriles, Santos, & Arêas, 2016).
The GFB formulations made from RF blends with 5%, 10%, and 14% WCF, for which the proximate compositions are presented in

| CONCLUSION
The application of a mixture design allowed finding that it is possible to add up to 14% WCF to an RF-based GFB formulation while negligibly diminishing the loaf volume, crumb firmness and crumb moisture. The best formulations were prepared from RF blends with 5, 10 and 14% WCF, and they received overall acceptability scores similar to those of their white GFB and standard wheat bread counterparts.
Incorporating 5%-14% whole chia flour in the formulation increased the levels of ash, lipid, protein and dietary fiber compared to those of the white GFB.
F I G U R E 2 Contour plots for the physical properties of gluten-free bread based on mixture design regression models. Ya= bake loss (%), Yb= loaf-specific volume (cm 3 /g), Yc= crumb firmness (N), and Yd= crumb moisture (%)

Potato starch
This research highlights the potential of WCF for producing nutrient-dense and acceptable GFB, which is important for consumers with gluten-related disorders because those products often lack nutrition content and acceptability.
T A B L E 3 Predicted and measured values for the loaf specific volume and crumb firmness of the optimized gluten-free bread formulations containing whole chia flour x 1 = Rice flour, x 2 = Potato starch, x 3 = Whole chia flour .

3
Values are the means and 95% confidence intervals.

4
Measured values from the confirmatory assay.

5
Values followed by different superscripts in each row are significantly different (p < .05).
T A B L E 4 Sensory acceptability scores of optimized gluten-free bread formulations containing whole chia flour Values are means ± standard deviations (n = 50) of acceptability scores on a 10-cm hybrid hedonic scale. Values followed by different superscripts in each row are significantly different (p < .05).
T A B L E 5 Compositions of optimized gluten-free bread formulations containing whole chia flour