3.5.3. Dough Stability Evaluation

The 2FI predictive model results from the regression analysis showed that both PS and the addition level of BF had a remarkable influence (*p* < 0.05) and fitted well the experimental data for dough ST (*R*<sup>2</sup> = 0.51) (Table 3). From the addition level effect point of view on dough ST, a significant decrease (*p* < 0.05) was observed, while fraction size

did not significantly influence ST, which varied from 7.05 to 10.17 min. The response surface plot showed that ST decreased when addition BF increased, depending on the fraction size (Figure 1d). Dough ST increased with particle size rise. This behavior may be associated with the composition of particle size fractions. Probably, at a particular ratio of fiber and gluten, gluten–fiber interactions can occur, enhancing dough stability. This result is consistent with the report by Han et al. (2013) [49], which showed that ST increased when a certain fraction of tartary buckwheat flour was added in wheat flour compared to other milling fractions due to gluten–fiber interactions that increase dough stability. However, the decrease in dough ST with the addition of BF can be explained by the gluten dilution when wheat flour was supplemented with gluten-free flour, indicating a rising degree of softening. The weakening of the gluten network was probably related to the dissimilarity between the proteins of buckwheat and the wheat proteins. Additionally, the buckwheat grain protein consists mainly of albumin and glutelins [60] and it cannot form structures such as wheat protein. Mohamed et al. (2014) [61] reported similar findings regarding the incompatibility between the protein spectrum of pulses and the semolina gluten protein. Similar changes in dough ST when whole buckwheat flour supplemented wheat flour in different amounts were reported by Stefan et al. (2018) [62] and Sedej et al. (2011) [22], revealing decreases in ST with supplementation increase.

#### 3.5.4. Protein Weakening Evaluation during Heating

In the heating stage, the mechanical shear stress and heating constraint led to the destabilization of proteins when the dough consistency decreased and C2 torque to the minimum value. For C2, the quadratic model was found to fit well with the experimental data, showing a high *R*<sup>2</sup> value (0.86) (Table 3). The C2 torque was influenced (*p* < 0.05) by the level of BF added in wheat flour, while the particle sizes had a non-significant influence (*p* > 0.05). It was found that the BF level had a highly significant negative effect (*p* < 0.01) on C2, showing a decrease in C2 torque with the increase in the BF level added in wheat flour (Figure 2a). This behavior was probably due to the gluten dilution effect by the addition of gluten-free buckwheat flour, which caused the formation of a weaker protein network. It is well known that buckwheat flour proteins do not have structure forming ability compared to wheat flour proteins. From Figure 2a, a decrease of C2 as the fraction of BF decreased can be seen, and in the medium fractions this fact was most evident, which can be correlated with either a release of water molecules in dough or to the presence of high proteins and the BF interfering with the protein unfolding [32]. The presence of a high protein in the medium fraction compared, to large and small particles, was noticed (Table 2). C2 torque reduction for small particle sizes may be correlated with the lower water availability in the dough because of high carbohydrates contents from this particle (Table 2). Lower C2 torque denotes too weak of gluten network and lower resistance of gluten with rising temperature.

**Figure 2.** 3D response surface plots showing interaction between buckwheat flour particle size and addition level on (**a**) C2 torque (C2) and (**b**) difference between torques C1 and C2 (C1-2) during protein weakening.

The difference between the C1 and C2 torques (C1-2), which represents the rate of protein thermal weakening, showed that the quadratic model was adequate (Table 3) for the prediction of the effects of factors on C1-2. The BF addition level influenced (*p* < 0.05) the parameter C1-2, while the milling fractions did not affect the C1-2 torque, as is shown by the ANOVA results (Table 3). It can be observed from the response surface plot (Figure 2b) that the BF level added to obtain composite flour had a positive influence on the C1-2, revealing an increase in C1-2 as the BF level increased, whereas from the fraction size effect point of view on C1-2, a decrease was observed. This variation can be explained by the changes in protein network structure due to the kneading and temperature effects. Buckwheat presents relatively small starch granules, which can occur in aggregates and singly with the same range of diameter (3–8 μm) [53]. In this case, probably the presence of enzymatic attacking points is favored by the proteins, which become less compact in the heating, increasing in the speed of protein weakening, causing a decrease of C1-2.

#### 3.5.5. Starch Gelatinization Evaluation

When the dough is heated above 60 ◦C, the starch granules are quickly broken, making them to raise the water absorption and swelling capacity, conducting to leaching of the amylose molecules that increase the viscosity and therefore the torque. The torque represented by C3 indicates peak viscosity and quantifies the degree of starch gelatinization. The effect of milling fractions and the level of BF added in wheat flour on the maximum viscosity of hot gel at 90 ◦C, measured by the C3 torque, showed that the quadratic model obtained was found to fit well the experimental data of C3 (*R*<sup>2</sup> = 0.87) (Table 3). C3 was significantly (*p* < 0.05) influenced by the linear term of the BF addition level in wheat flour, the interaction between factors, and the quadratic term of particle size. The addition of BF caused a decrease of C3 torque in composite flour and might be because buckwheat amylose forms with lipids complex compounds [54]. Thus, these complexes can decrease the swelling power and solubility of buckwheat starch, thus affecting C3 torque. Moreover, nonstarch components (lipids, proteins, and dietary fibers) present in BF milling fractions could restrict swelling and gelatinization during cooking, in addition to a diluting effect due to the interaction with starch polymers (lipids and proteins), and to the competition for water (proteins and dietary fiber) [32] interfering with starch swelling [63]. The greater negative effect on C3 was given by the BF addition level, showing a decrease of C3 as the BF amount added in wheat flour increased. The lower swelling and gelatinization of starch granules was probably explained by the morphological characteristics of buckwheat starch, which restricted water quantity in the composite flour dough could decrease the peak of C3 torque. These results were in accordance with those reported by Filipˇcev et al. (2015) [64]. As can be seen from Figure 3a, C3 decreased with BF supplementation, and when the BF addition in wheat flour is 20%, the C3 torque decreased at 1.60 N·m for the large particle size value. These changes, most likely, are related to the higher enzyme activity and the amylose-lipid complexes formed during the heating of starch slurries. Qian et al. (1998) [54] related that to the small granule of buckwheat-starch and their pores, and this starch is more susceptible to fungal α-amylase than wheat starch. The viscosity of the gel formed is largely influenced by granule shape and swelling power, amylopectin-amylose entanglement, and amylose and amylopectin granular interaction [65]. The differences among the assessed fraction sizes might be due to the particle microstructural and chemical features of the buckwheat-starch that governs water absorption and swelling.

The difference between C3 and C2 torques (C3–2) is significantly (*p* < 0.01) affected by the milling fractions and BF addition levels. The quadratic regression model represented well the experimental data of C3-2 and the *R2* value (0.86), confirming that this model is suitable. The increase in the amount and the BF milling fractions led to C3-2 (Figure 3b) decrease, the interaction between milling fractions and the BF amount being more pronounced (Table 3). α-amylase activity along with the composition of the BF milling fractions from the composite flour could explain these changes.

**Figure 3.** 3D response surface plots showing interaction between buckwheat flour particle size and addition level on (**a**) C3 torque (C3) and (**b**) difference between torques C3 and C2 (C3-2) during starch gelatinization.

#### 3.5.6. Cooking Stability Evaluation

The cooking stability is measured by the minimum torque C4 obtained due to the rupturing of the swollen starch granules, thereby causing a decrease in the consistency of the hot-formed starch gel. The quadratic regression model for C4 torque was found to be significant and the *R*<sup>2</sup> value of 0.87 confirmed the suitability of the model (Table 3). The ANOVA results revealed that there was a significant effect (*p* < 0.0001) of the BF level added to wheat flour, and of the interaction (*p* < 0.05) between particle size and BF level on C4. The combined effect of factors on C4 torque is noticed in the response surface plot (Figure 4a). It can be observed that, with the addition of BF, the C4 torque decreased. Consequently, the cooking stability is diminished as a result of the paste resistance to disintegrate when the temperature is raised. The C4 torque indicated lower values for all samples formulated compared to the control, exhibiting a low resistance of starch against the enzymatic hydrolysis by amylase. The results obtained are in line with those reported by Filipˇcev et al. (2015) [64], which found a decrease in hot gel stability for buckwheat supplemented doughs. The lowest C4 value was achieved for a large fraction at a higher level of 20% BF. The C4 torque decrease when the amount of BF in composite flour increased might be correlated with the fraction composition, more likely because of the soluble fiber that can bind water by hydrogen bonds, causing a decrease in available water for the starch granules [31]. These results may be related to the protein denaturation from BF, which according to Janssen et al. (2017) [66] shows a minor and a major denaturation peak at about 80 ◦C and about 102 ◦C. The protein denaturation will change the dough network, thus determining a decrease in viscosity.

**Figure 4.** 3D response surface plots showing interaction between buckwheat flour particle size and addition level on (**a**) C4 torque (C4) and (**b**) difference between torques C3 and C4 (C3–4) during cooking stability.

The difference between C3 and C4 torques peak values (C3-4) reflects the hot-gel stability or cooking stability of dough. C3-4 was influenced (*p* < 0.05) by BF addition in composite flour and the interaction between fraction size and BF level. The quadratic regression model obtained fitted well to the experimental data of C3-4 with a high coefficient of determination (*R*<sup>2</sup> = 0.86) (Table 3). The simultaneous effect of particle size and addition level on C3-4 is shown in Figure 4b, revealing their capacity to increase the C3-4. This increase of C3-4 may be attributed to the compounds from BF fractions, which changed the α-amylase starch interaction modifying the α-amylase hydrolytic activity on starch.

#### 3.5.7. Starch Retrogradation Evaluation

The decrease in temperature during the cooling stage to 50 ◦C resulted in a final viscosity associated with a higher dough resistance, and consequently with the C5 torque, which reflects starch retrogradation. This is the process in which the leached amylose chains undergo recrystallization, conditioned by the ratio of amylose and amylopectin from starch [67,68], because amylose recrystallizes with a high speed compared to the amylopectin. Evaluation of C5 torque and C5–4 is used to characterize starch retrogradation. The quadratic regression model achieved for C5 torque was statistically significant (*p* < 0.001) and was a very good one, with a high value (0.92) of *R*2, which showed that only 8% of the total variation was unexplained by the model (Table 3). Both studied factors influenced (*p* < 0.05) the C5 torque. With the rise of fractions and the level of BF added in wheat flour, the C5 values decreased, indicating a low starch retrogradation and recrystallization. This result can be associated with the particular structure of starch and the granules of buckwheat presenting a low tendency toward retrogradation. Moreover, the high amounts of phenolic compounds can contribute to retrogradation lowering. The graphical representation of the effect of milling fractions and the level of buckwheat flour added on the C5 torque is presented in Figure 5a. From the response surface plot, it can be observed that, as the fraction size and the BF addition increased, the C5 torque decreased, suggesting that fraction size composition affected α-amylase activity in buckwheat-wheat flours, a lower C5 value being caused by the high activity. However, lower starch gels, which have low C5 torque, are usually linked to lower amylose content. The initial retrogradation is mostly attributed to the re-association of amylose, and to the long-term retrogradation associated with the development of gel structure and crystallinity, e.g., during staling of bread, being due to the recrystallization of amylopectin side chains [69]. As reported in some studies [31,70], the structure reorganization of starch may be favorable for the hardening of bakery goods during storage.

**Figure 5.** 3D response surface plots showing interaction between buckwheat flour particle size and addition level on (**a**) C5 torque (C5) and (**b**) difference between torques C5 and C4 (C5–4) during starch retrogradation.

The difference between C5 and C4 torques (C5–4) indicates starch retrogradation capacity, influenced by different factors such as the botanical source of the starch, amylose/amylopectin ratio and average chain length of amylose and amylopectin [71]. The

quadratic model, which was obtained for the C5–4, was found to have statistical significance (*p* < 0.05) with a high *R*<sup>2</sup> (0.88), which confirms the adequacy of the model (Table 3). C5–4 was affected (*p* < 0.0001) by the addition of BF in wheat flour, but the fraction size had a non-significant influence (*p* > 0.05).

The level of BF in the composition of dough decreases the C5–4 torque in comparison to wheat flour dough (Figure 5b). This decrease could be attributed to the higher lipid content of buckwheat flour, different structure of amylose and amylopectin fraction of buckwheat starch compared to wheat starch, and lipid-amylose complex-forming ability [22]. Amylose molecules cannot reassociate and recrystallize in a free manner as in systems with low lipid content. This could have a positive effect on bread staling made from buckwheat flour, once the starch retrogradation is one of the key-bread staling factors. These results are in agreement with those reported in other studies [22,72].

### *3.6. Optimization of Buckwheat-Wheat Composite Flour*

The models fitted in this study for dough rheological proprieties were used for simultaneous optimization by using numerical optimization and desirability function approaches. The numerical optimization allowed for the establishment of promising composite flour formulation, which was then compared to the control sample. The results highlighted that the optimal values of factors to achieve the most appropriate composite flour would be 10.75% buckwheat flour that were 280 ìm in particle size. Based on this optimal formulation, the predicted values for the evaluated responses, in terms of the FN index and Mixolab parameters, are shown in Table 4. The results revealed that the optimal values for WA, ST and C2 torque were slightly smaller than the values of the control sample, with no statistical difference.

**Table 4.** Features of wheat dough control and optimized buckwheat-wheat samples.


These results suggested that, during the first stage of bread making, the dough resulting from the optimization process could retain a CO2 similar to the wheat dough. The high rise of DT indicated an increase in the gluten network strength compared to the control, suggesting that the optimal dough can sustain the mechanical treatment for a longer period during the bread-making process, and these were findings that were in line with those previously reported [31,32].

In respect to starch behavior, a decrease in the C2, C3–2, C4, C5, and C5–4 Mixolab parameters was obtained, which highlighted that the optimal buckwheat-wheat composite flour may be appropriate for bread making. The baked goods that can be obtained from this composite flour will present a longer shelf-life in the absence of problems related to staling.

#### **4. Conclusions**

The particle size of buckwheat flour revealed variations in chemical characteristics and functional properties. The highest content of protein, lipids, and ash was found in medium particle sizes, followed by a large particle size and a small particle size, which are rich in carbohydrates. Medium particle size had more water and swelling properties, whereas the small particle sizes presented a higher volumetric density. The milling fractions and amount of buckwheat incorporated in wheat flour remarkably influenced dough rheological properties, predicted by the suitable regression models. The multi-response

optimization study allowed for the proposal of the optimal milling fraction size and buckwheat flour amount added in wheat flour for obtaining composite flour with the best rheological properties. The results allowed us to confirm that the best formulation composite flour has 10.75% buckwheat flour of 280 μm particle size and 89.25% wheat flour, and was the closest for the requirement to create new products in the bakery industry, without altering the dough matrix.

Future studies are required to identify the impact of buckwheat flour milling fractions on bread and sensory characteristics that would provide further evidence and confirmation for milling fractions and bread making.

**Author Contributions:** Conceptualization, I.C. and S.M.; methodology, I.C.; software, I.C.; validation, I.C. and S.M.; formal analysis, I.C. and S.M.; investigation, I.C.; resources, I.C. and S.M.; data curation, I.C. and S.M.; writing—original draft preparation, I.C.; writing—review and editing, I.C. and S.M.; visualization, I.C. and S.M.; supervision, S.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by Stefan cel Mare University of Suceava, Romania.

**Acknowledgments:** The authors thank Arcada Research laboratory, Arcada Mill, Romania for technical support.

**Conflicts of Interest:** The authors declare no conflict of interest.
