Towards Sustainable and Nutritionally Enhanced Flatbreads from Sprouted Sorghum, Tapioca, and Cowpea Climate-Resilient Crops

This study aimed to develop high-quality flatbreads for low-income countries by using composite flours from climate-resilient crops, i.e., sprouted sorghum, tapioca, and cowpea, as partial alternatives to imported wheat. Through the experimental design, several flatbread prototypes were developed that maximized the content of sprouted sorghum and cowpea flours and minimized the content of wholewheat flour. Three of them were chosen based on the best textural, nutritional (highest intake of energy, proteins, and micronutrients—iron, zinc and vitamin A), and economic (cheapest in Sierra Leone, Tanzania, Burundi, and Togo) features. The physicochemical properties, in vitro starch digestibility, total phenolic content, antioxidant capacity, and sensory acceptability were also measured for the samples. The experimental flatbreads showed lower rapidly digestible starch and higher resistant starch contents than the control (100% wholewheat based), and were also richer in phenolic content and higher in antioxidant activity. Moreover, one of the prototypes was perceived to be as acceptable as the control for texture and flavour properties. The ranking test, performed after explaining the nature of the samples, revealed that the flatbread meeting the nutritional criteria was the preferred one. Overall, the use of composite flour from climate-resilient crops was proven to be an efficient strategy to obtain high-quality flatbread.


Introduction
Developing bakery products by using composite flours is currently a research topic of great interest, as it represents a way to improve the nutritional value of standard wheatbased products [1]. If the composite flours are obtained from climate-resilient, nutritious, and affordable crops, the approach can be particularly efficacious in developing countries. This approach has the advantage of saving money by reducing the import of wheat flour and providing a better supply of macro and micronutrients in the diet. Moreover, many more business opportunities along the value chain can be created by the use of local agriculture [1,2].

Experimental Design
An experimental fractional factorial design (FFD) with Resolution V was used to investigate the effect of the SS, T, C, and W levels (input factors, independent variables) on the textural properties of the derived flatbread (responses, dependent variables). The experimental FFD was chosen for screening purposes to evaluate the preliminary significance of the variables and their interactions. Water level was included as an uncontrolled factor, as its use as a controlled/constant input produced a set of unfeasible experiments in preliminary trials, i.e., unworkable doughs due to too little or too much hydration. A centre point and two levels (minimum and maximum) for each input factor were selected in preliminary screening tests to meet different criteria. Firstly, the formulations had to include all four types of flour. Secondly, as much SS and C flours had to be included as possible and as little W as possible while ensuring the dough's workability for flatbread production. The experimental design (F1-F9) is reported in Table 1. The three-replicate centre point (Exp. F9-F11) allowed the calculation of experimental error in the analyses and predicted whether the experimental design would give a significant lack of fit [17]. The chosen responses were the texture flatbread quality attributes: force at rupture (f, N), extensibility (e, mm), moisture content (MC, g water/100 g sample), and water activity (a w ). The force at rupture and extensibility were measured by both a puncture test (P) and a one-dimensional extensibility test (E) (TA.XT2 Texture Analyzer, Stable Micro Systems, Godalming, UK).

Flatbread Production
Flatbreads were produced as follows: the dry ingredients were mixed at room temperature (R/T) in a home bread-maker (Backmeister 68511, UNHOLD, Hockenheim, Germany) for 2 min; then, sunflower seed oil and salt (2% w/w) were added and mixed for a further 2 min. Finally, distilled water was added in variable quantities until a dough with an optimal consistency (empirically evaluated), similar to that of the centre point, was obtained. After a resting period of 5 min at R/T, the dough was manually shaped into 50 g balls, which were rested for an additional 25 min at R/T and then laminated twice to obtain a circle of dough (1.85 ± 0.12 mm thickness). The shaped pieces of dough were then cooked at 200 • C on a glass-ceramic skillet (Schott Ceran, Mainz, Germany) for 1 min per side. The flatbread was left to cool at R/T prior to analysis.

Texture
The textural properties of the flatbreads were measured with a texture analyser (TA.XT2 Texture Analyser) equipped with a 25 kg load cell (Stable Micro Systems, Goldalming, UK) using a puncture test and a one-dimensional extensibility test [18,19]. The rupture force (N) and rupture distance (mm) were recorded for both tests. At least 5 replicates were performed for each test and each flatbread prototype.

Moisture Content (MC) and Water Activity (a w )
The moisture content was measured in triplicate by weight loss after forced-air ovendrying at 105 • C until a constant weight was achieved (ISCO NSV 9035, ISCO, Milan, Italy). The analyses were performed in triplicate.
Water activity was measured at 25 • C using an Aqualab 4 TE (Decagon Devices, Inc., Pullman, WA, USA). A minimum of 6 determinations were carried out for each flatbread prototype.

Choice of Formulations: Technological, Nutritional and Economic Criteria
The experimental design generated weak models characterized by high mathematical uncertainty and low predictivity (see Section 3.1). This prevented us from drawing firm conclusions and from choosing the best formulation from a technological point of view. We, therefore, decided to produce a 100% W-based flatbread taken as a reference (STD) for the evaluation of the textural properties of the composite flour flatbreads. The W flatbread was prepared by mixing 265 g of flour with 187.5 mL of water following the recipe described above (Section 2.2.2).
Consequently, the formulations produced by the experimental design were evaluated based on the textural properties of the derived flatbreads. Moreover, we decided to also consider their estimated nutritional value and costs following the methods described below. The formulations showing the best performance for these three criteria were selected for further characterization.

•
Physico-chemical properties The textural properties P_f (force at rupture from puncture test), P_e (extensibility from puncture test), E_f (force at rupture from one-dimensional extensibility test), and E_e (extensibility from one-dimensional extensibility test) of flatbreads were compared with those measured for the STD flatbread. The formulation with textural properties more similar to those recorded for STD was considered eligible to satisfy the technological criterion and therefore selected for further characterizations.

•
Nutritional value An estimated nutritional evaluation of the formulations was carried out. Firstly, the composition of macro and micronutrients expressed in 100 g of flour was calculated based on the nutritional information obtained analytically (Section 2.1, [16]) integrated with information obtained from the FAO/INFOODS Food Composition Table for Western Africa 2019 [20]. Then, for each nutrient, the range of variation identified within the experimental set (F1-F9) was established. For each range and nutrient, 4 target levels were defined: minimum intake, intake equal to 33% of the range, intake equal to 66% of the range, and maximum intake. Lastly, for each nutrient of each formulation, the following scores were assigned: 0 = an intake between the min value and ≤33% of the range of variation; 1 = an intake of >33% but ≤66% of the range; 2 = an intake of >66%. The individual scores were added up to obtain the final score for each formulation.
The most prevalent nutritional deficits among females in sub-Saharan Africa were estimated to be protein-energy malnutrition and iron, zinc, and vitamin A deficiencies [21,22]. Therefore, in a second evaluation, the scores obtained for the components mentioned above were multiplied by two. Eventually, the individual scores were added up and the formulation with the highest total score was considered eligible to satisfy the nutritional criterion and, therefore, selected to further characterization.

•
Economic sustainability The economic sustainability of the formulations was evaluated by calculating their prices ($/kg) in different African countries, Sierra Leone, Tanzania, Burundi, and Togo, based on the costs of the single raw materials and their proportion in the blend. This information was collected on-site by local partner institutions and collaborators of the University of Parma. The least expensive formulation that met the economic sustainability criterion was selected for further analysis.

Relevant Starch Nutritional Fractions
In vitro starch digestibility (rapidly digestible starch (RDS), slowly digestible starch (SDS), and resistant starch (RS)) was quantified with the Megazyme Digestible Starch and Resistant Starch assay procedure (K-DSTRS 11/19 commercial kit, Megazyme International Ireland Ltd., Wicklow, Ireland) following the manufacturer's protocol based on the method described by Englyst, Vinory, Englyst, and Lang [23]. Two batches of each bread type were produced on different days, and each one was analysed in triplicate. Six determinations of the RDS, SDS, and RS fractions, respectively, were acquired. Digestible starch (DS, RDS + SDS) and total starch (TS, DS + RS) were also calculated.

Total Polyphenol Content (TPC) and DPPH• Free Radical-Scavenging Activity Analysis
Flatbread phenolic extracts were prepared, starting from 1 g of ground bread added to 20 mL of a methanol/water (70:30 v/v) mixture, stirred at R/T for 1 h, and then filtered using a paper filter. The solvent was evaporated and the extract was dissolved in 1 mL of a methanol/water (70:30 v/v) mixture and centrifuged at 5040× g (6255 rpm) for 15 min at 4 • C.
For the total phenolic content (TPC) measurement, 50 µL of each extract was mixed with 1160 µL of distilled water, 300 µL of Na 2 CO 3 aqueous solution (20% w/v), and 100 µL of Folin-Ciocalteu's reagent. The absorbance of the solution was measured at 760 nm by a UV-visible spectrophotometer after 30 min of incubation in the dark at R/T. A calibration curve using gallic acid as an external standard (10-300 mg/L) was prepared for quantification. The results were reported in milligrams of gallic acid equivalents (GAE) per kilogram of bread on d.b.
Antioxidant capacity was determined using a DPPH• assay (2,2-diphenyl-1-picrylhydrazyl free radical) following the procedure proposed by Dall'Asta, Cirlini, Morini, Rinaldi, Ganino, and Chiavaro [24]. Briefly, 200 µL of each extract was mixed with 2.6 mL of methanol and 2 mL of a methanolic DPPH• solution (0.2 mmol/L). The absorbance of the solution was measured at 517 nm using a Perkin Elmer UV-Visible spectrophotometer after 30 min of incubation in the dark at R/T. A blank was prepared and analysed following the same procedure. The radical scavenging activity was calculated as follows (Equation (1)): where Abs 0 and Abs 1 are the absorbances of the blank and the sample, respectively. The TEAC value (Trolox-equivalent antioxidant capacity; µmol Trolox eq./g of bread on d.b.) was obtained from the 0.1-0.5 mmol/L calibration curve calculated by measuring the absorbance at 517 nm of Trolox ((±)-6-Hydroxy-2,5,7,8-tetramethylchromane-2carboxylic acid) methanolic solutions at different concentrations. Two batches from each bread type were analysed and the analyses were performed in duplicate.

Colour Analysis
Colour was measured with a CieLAB colourimeter (CM 2600d, Minolta Co., Japan) using a standard illuminant D65. The L* (lightness), a* (degree of redness), and b* (degree of yellowness) parameters were measured using a 10 • position of the standard observer. At least 15 determinations were performed by analysing the surface of three flatbread samples for each prototype. Differences in colour between flatbread prototypes and the STD sample were evaluated using the ∆E value, calculated as follows (Equation (2)): where ∆L* is the difference between L* values of flatbread prototype and STD; ∆a* is the difference between a* values of flatbread prototype and STD; ∆b* is the difference between b* values of flatbread prototype and STD.

Preliminary Sensory Evaluation
Consumers' sensory evaluation of flatbreads was assessed by means of an acceptability test, a preference ranking test, and a just-about-right (JAR) test.
The flatbreads were produced a few hours before the analysis. After baking and cooling, discs with a 5 cm diameter ( Figure S1) were obtained using a pastry cutter; then, they were divided in half and identified with random three-digit codes. Samples were simultaneously presented on a plate in randomized order and under blind conditions. Consumers were asked to taste and evaluate one sample at a time. Water was served to cleanse the palate between tastings. Prior to starting the sensory test, all interviewees were asked to read and voluntarily sign an informed consent form. A detailed description of the method, duration, and purpose of the study was provided. Moreover, the anonymity of the data and non-use for commercial purposes were guaranteed. Among the interested volunteers, people in a good and healthy state, without allergies nor dietary restrictions, and who did not have any aversions to specific food products were selected as panel members. The panel consisted of 27 untrained African consumers, of which 17 were students at the University of Parma and 10 were recruited among the workers of a local farm. Due to the difficulty in recruiting a larger number of consumers during the COVID-19 emergency period, the number of panellists was limited. Before the analyses, the consumers were asked questions about their age, gender, and country of origin. First, the panellists evaluated the acceptability of the flatbread by rating its overall acceptability, consistency, appearance and aroma by means of a 9-point hedonic scale (1 = dislike extremely, 2 = dislike very much, 3 = dislike, 4 = dislike slightly, 5 = neither like nor dislike, 6 = like slightly, 7 = like, 8 = like very much, and 9 = like extremely). Then, for the JAR test, consumers rated the samples on a 5-point JAR scale (1 = much too low, 2 = too low, 3 = just about right, 4 = too much, and 5 = far too much) for hardness, darkness, bitterness, and bean flavour. After tasting all the samples, the consumers were asked to answer a preference ranking test, before and after providing them with a brief explanation of the project and samples.

Statistical Analysis
All data were expressed as the mean ± standard deviation (SD). MODDE 12.1 Software (Umetrics AB, Malmö, Sweden) and the PLS method were used to design and analyse the experiments and analyse the results, respectively. The model validity and reproducibility were also evaluated. Reproducibility was evaluated by the variation in the response under the same conditions (pure error), often at the centre points, compared with the total variation in the response. To interpret the influence of terms (factors) on each model, the coefficient plot and effect plots and lists were evaluated. Each model was refined by removing outliers and/or insignificant terms, and/or adding significant/interaction terms. The reliability of the models was then evaluated by calculating the R 2 and Q 2 values, where R 2 is the percentage of the variation in the response explained by the model and Q 2 is the percentage of the variation in the response predicted by the model according to cross validation, expressed using the same units as R 2 .
One-way analysis of variance (ANOVA) followed by Duncan's post hoc test at 0.05 significance level (SPSS Statistical Software, Version 25.0, IBM SPSS Inc., Milan, Italy) were performed to assess significant differences between the samples. The Pearson correlation coefficient was also calculated to investigate the statistical correlations (p < 0.05) between the antiradical activity and TPC.
A Friedman nonparametric test followed by multiple pairwise comparisons using Nemenyi's procedure was used to determine the significance of the ranking preferences. For the data analysis of the JAR test, the impact of the JAR variables on the overall liking scores was calculated using Spearman's correlation coefficient. Then, penalty analysis (PA) was performed grouping the 5-point JAR scale into 3 levels (too little, JAR, and too much) for each attribute. Then, the mean overall liking score from the hedonic scale scores and the percentage of respondents represented in each of the 3 categories were calculated. Lastly, mean drops were determined by subtracting the mean overall liking score for the JAR group from the mean overall liking score of the "too much" or "too little" categories [25]. The penalty was then calculated as the weighted difference between the means (mean of liking for JAR-mean of liking for the two other levels taken together).
The ranking and JAR sensory data analyses were performed using XLSTAT Software (Version 2020.3.1, Addinsoft, New York, NY, USA).

Results and Discussion
In a previous study [15], a composite flour made up of SS, T, C, and W at 25% w/w flour basis (f.b.) was proven to offer a good compromise between technological and nutritional qualities. The aim of this work was to investigate the use of SS_C_T_W composite flour in breadmaking by evaluating the physico-chemical, nutritional, and sensory properties of the finished product while investigating a possible modulation of the ingredients to enhance the sustainability, texture, and nutritional quality of flatbread formulations.
After preliminary trials to increase the levels of SS, C, and T (locally sourced raw materials) and decrease the level of W (imported flour in Africa), and evaluating the workability of the derived dough, an experimental design was set for screening purposes to investigate the effect of the SS, T, C and W levels on the textural properties of the derived flatbreads.

Experimental Design
The experimental design allowed homogeneously testing a domain defined by four controlled factors (SS, T, C, W) and one uncontrolled factor (water), i.e., five dimensions. Very low reproducibility was found for the MC and a w responses; the variability in these data was, overall, comparable with the experimental error determined on the three-replicate centre point. For this reason, these responses were not investigated further.
For all the other answers, the reproducibility of the data was adequate to further study the obtained models; however, they were weak and characterized by high mathematical uncertainty. For all the results, the effects of the factors had to be read in probabilistic terms.
The models generated were generally characterized by low Q 2 values (Table S1; additional information about the significance of the models generated can be found in Appendix A). As drawing conclusions from a model with Q 2 > 0.5 is recommended [17], the low predictivity of the models and the high uncertainty of the effects prevented us from using the model's mathematical equations to make predictions. For this reason, with all the formulations proposed by the experimental design being workable, three of them were selected for complete characterization. The selection was performed based on three criteria: textural quality, nutritional value (estimated), and economic sustainability.

Choice of Formulations: Technological, Nutritional, and Economic Criteria
• Physico-chemical properties The textural properties, MC, and a w of flatbread samples are reported in Table 2. The STD flatbread exhibited the highest puncture rupture force (~8 N) and extensibility (~13 mm), indicating a more cohesive and extensible structure probably due to the presence of a higher amount of gluten proteins in this formulation when compared with flatbreads obtained from composite flours. Sample F6 was the only one that came close to STD with regards to P_f (~6 N) and, together with F4, also in terms of P_e (~13 mm for both F6 and F4). The data obtained from the one-dimensional extensibility test showed, for the STD sample, an intermediate rupture force (~6 N) when compared with the range identified for the other samples (~5 N (F2)-10 N (F5)). On the other hand, the extensibility of the STD (~7 mm) was significantly higher than those of all other formulations. F1 and F6 were characterized by extensibility more similar to that of STD, although significantly (p ≤ 0.05) lower (~3 mm for both samples). The moisture content of STD was 30.7 g water/100 g sample, which is within the range identified for composite flatbreads (26.1 (F9)-33.3 g water/100 g sample (F8)). Despite F1, F3, F5, F6, and F10 exhibiting similar MC when compared with STD, most of the samples showed significantly (p ≤ 0.05) lower a w , which may be related to stronger interactions between biopolymers and water in these samples, when compared with STD.
Overall, the changes in the textural attributes of flatbreads may be related to the different water-solids interactions developed as a function of the different compositions (gluten and non-gluten proteins, starch, sugars, fibre, etc.) of the used flours and to their different percentages in the blend. Polymers of different chemical and/or botanical natures can interact with water in different ways [26] and compete for the formation of molecular bonds with it, leading to different levels of polymer hydration. These phenomena significantly affect the water state and distribution in complex matrices, influencing the textural attributes [27]. F6 showed textural attributes more similar to those of STD; thus, it was identified as the sample meeting the technological criterion.

•
Nutritional value The estimated proximate composition of the flatbread formulations is presented in Table 3. An initial evaluation of the formulations was provided with scores (0, 1 or 2) based on the theoretical intake provided for each nutrient when compared with the range of variation within the experimental set. Based on this attribution, formulations F5, F6, and F7 exhibited the highest score for the energy and macronutrients (8) and micronutrients (26,28, and 30, respectively) contents. Overall, F7 was evaluated as the formulation with the best theoretical nutritional profile (total score = 38), followed by F6 (total score = 36) and F5 (total score = 34). Table 3. Estimated macro and micronutrient composition of flatbread formulations (F1-F9) and of a 100% W flatbread (STD), per 100 g of flour. In brackets, the score attributed to the theoretical intake of each nutrient provided by the formulation is presented. The total score is the sum of the partial scores. Previous findings revealed how the most prevalent nutritional deficits among females in sub-Saharan Africa were protein-energy malnutrition and iron, zinc, and vitamin A deficiencies [21,22]. Therefore, a second evaluation was then proposed by attributing a greater relevance to the contribution of the aforementioned components, multiplying the respective scores by two. The results are reported in brackets in Table 3. Formulations F5 and F6 received the highest rating (12) for the contribution of energy and macro components, followed by F7 (10), which differed from the other two mainly due to its lower energy intake. The theoretical highest score for micro components, on the other hand, was attributed to F7 (36), followed by F6 (34) and F5 (32). Overall, F6 and F7 were evaluated as the formulations with the best theoretical nutritional profile (total score = 46), followed by F5 (total score = 44). Their highest theoretical nutritional profile was probably the result of their high content of nutrient-rich flours (e.g., SS and C made up more than 85% of the flour in F7) [13][14][15][16]. Comparing F6 and F7 with STD, the former flours provided a theoretically higher intake of proteins and most micronutrients than the latter, along with lower intake of energy, carbohydrates, fat, fibre, iron, phosphorus, and vitamin B6. As a result, F6 and F7 were selected for further analysis based on the nutritional criterion. Interestingly, F6 was already chosen, given its compliance with the technological criterion.

•
Economic sustainability The prices ($/kg) of the raw flours (W, T, C, and unsprouted sorghum (S)) used for the creation of blends are shown in Table 4. For sorghum, we reported the price of the native flour, because the sprouted flour is not commonly sold at local markets, but it is produced domestically by subjecting sorghum kernels to sprouting and milling. As sprouting represents a feasible and inexpensive technology traditionally performed at the domestic level in Africa [8,28], a similar cost is expected for the raw and sprouted flours.  For all four considered African countries, T and S were the flours with the lowest price. The price of T varied between a minimum of 0.31 $/kg (Burundi) and a maximum of 1.10 $/kg (Togo); on the other hand, the price of S varied from 0.66 $/kg (Sierra Leone) to 1.29 $/kg (Tanzania).
In all countries, the cost of C was higher than that of T and S, varying from a minimum of 0.71 $/kg (Sierra Leone) to a maximum of 2.15 $/kg (Tanzania). Although cowpea is a widely cultivated and consumed legume in Africa [14], the higher cost of the derived flour could be associated with the fact that cowpeas are mainly consumed as a legume and not in the form of flour. The prices of W were also higher than those of S and T, as expected, as it is imported. W can be purchased for a minimum price of 0.71 $/kg (Sierra Leone) up to a maximum of 1.38 $/kg in Tanzania.
The costs of the formulations were a projection of the percentage of each ingredient in the blend and their prices in a specific country. The formulations proposed by the experimental design may have an average price of 0.68 $/kg (Sierra Leone), 1.62 $/kg (Tanzania), 0.67 $/kg (Burundi), and 0.93 $/kg (Togo). F4 was the least expensive composite formulation, thus meeting the economic sustainability criterion. Interestingly, all composite formulations were less expensive than STD (100% W) in Sierra Leone, Burundi, and Togo, while the opposite was observed for Tanzania.

Characterization of Selected Flatbreads
Based on the previous results, flatbreads F4, F6, and F7 were selected for further analysis and comparison with STD.

Relevant Starch Nutritional Fractions
Relevant nutritional starch fractions were identified (Table 5) based on the rate of glucose release and its absorption by the gastrointestinal tract during digestion. These include RDS, SDS, and RS, defined as the three sequential nutritional starch fractions determined by the reaction time when in vitro starch digestion is performed [29].  /100 g d.b.)   RDS is the starch portion digested in the mouth and in the small intestine, causing a rapid rise in the blood glucose concentration after the consumption of carbohydrates. The RDS fraction determined in vitro identifies the amount of starch digested within 20 min in a standard digestion reaction mixture [30]. The results for RDS were fairly close to each other and all significantly (p ≤ 0.05) lower in the blended flour flatbread samples (≈37 g/100 g d.b.) than in STD (≈48 g/100 g d.b.). SDS represents the starch portion digested slowly, but completely, in the human small intestine after RDS and identifies the starch portion digested within more than 120 min under standard in vitro substrate and enzyme concentration conditions [30]. SDS offers potential health benefits, e.g., stable glucose metabolism, control of diabetes, mental performance, and feelings of satiety [31]. The results for SDS in the composite flour-based breads ranged from ≈28 to ≈23 g/100 g d.b.; in all cases, they were higher than that in STD, which exhibited the lowest amount of SDS (≈17 g/100 g d.b.).
Focusing on the percentages of RDS and SDS expressed ion DS, i.e., the starch fraction available for digestion within 120 min [30], F7 exhibited the lowest RDS percentage (and consequently the highest SDS percentage), followed by F6 and F4, which presented intermediate results, and STD, most of whose starch was rapidly digested (73%).
RS represents the starch portion escaping digestion in the small intestine [30] that may undergo bacterial fermentation in the colon. The composite flatbreads contained RS amounts ranging from ≈0.93 (F4) to ≈0.69 (F6) g/100 g d.b., which resulted, in all cases, significantly higher values than the RS amount measured in the STD sample (0.48 g/100 g d.b.).
Overall, all composite flatbreads exhibited lower RDS and higher RS values than STD. These traits are of relevance from a nutritional point of view [23]. As T is a source of readily available carbohydrates [11], and sprouting has been proven to increase starch digestibility [16,32,33], the results observed for blended flour flatbread may be principally attributed to C. The incorporation of wholemeal legume flour into bread recipes appears to decrease starch hydrolysis, possibly due to the lower starch content, but greater fibre and protein contents. In fact, the reduced rates, as well as the overall reduced starch digestibility of legumes, are influenced by the cell wall structure, phenolic compounds, high amylose content, and viscous soluble dietary fibre components. Additionally, the high protein content of legumes can promote starch-protein interactions, which limits enzymatic attack [33,34].

Total Phenolic Content (TPC) and DPPH• Free Radical Scavenging Activity
The TPC of the flatbread samples is reported in Table 5. F7 showed the highest TPC (≈835 mg GAE/kg d. b), followed by F4 (≈771 mg GAE/kg d. b) and F6 (≈655 mg GAE/kg d. b), the latter not being significantly different (p ≤ 0.05) from STD (≈651 mg GAE/kg d. b). The highest TPC measured for flatbread F7 may be related to the highest presence of SS and C in the composite flour used, which possibly enriched the finished product with bioactive compounds with potential beneficial effects on human health. Indeed, as is known from the literature, like other legume flours, cowpea flour is a good source of phenolic compounds [14,35]. In addition, the positive function of sorghum sprouting on the phenolic content of wholemeal flour has been documented in detail in the literature [36][37][38][39][40]. It has been scientifically proven that the species of reactive oxygen and the free radicals generated in cellular metabolism or the peroxidation of lipids and other biological molecules play important roles in chronic diseases, such as coronary heart disease and cancer. Antioxidants in the diet fight against free radicals and probably help to prevent chronic disease risk factors [40]. In this work, the antioxidant capacity of the flatbreads was determined based on the DPPH• free radical scavenging activity and expressed as Trolox-equivalent antioxidant capacity (TEAC); the results are reported in Table 5. The TEAC values were 1.09, 1.03, 1.96, and 0.92 µmol Trolox eq./g d.b. for F7, F4, F6, and STD, respectively, the latter two not significantly differing from each other (p ≤ 0.05).
As expected, the Pearson correlation revealed a strong positive correlation (r = 0.90; p < 0.05) between TPC and TEAC. However, the improved TEAC observed in the composite flour bread samples could be due not only to the enhancement of phenolic compounds in the flatbread, but also to the documented increase in antioxidant vitamins after sorghum sprouting, along with the antioxidant potential of cowpea proteins. These factors are mainly related to the hydrophobic and aromatic amino acids' capacity to donate protons to free radicals [14] and to potential Maillard reaction products formed during baking [41].

Colour Analysis
The colour parameters (L*, a*, and b*) of the selected flatbreads and of the STD sample are presented in Table 5. STD exhibited the highest lightness (L* ≈ 67) and the lowest redness (a* ≈ 4) compared with the composite flour-based flatbreads, while its yellow hue (b* ≈ 11) was low, but not significantly different from that of F4 (b* ≈ 10.6). F4, in turn, had the lowest brightness (L* ≈ 53), while its red colour component was not significantly different from those of the other flatbreads (a* ≈ 6.5), which may have been due to the strong reddish-brown notes of the SS and the lighter reddish notes of the C. Among the flatbreads, F6 had lighter coloured notes (L* ≈ 57), probably due to the lower presence of SS in the blend, and more intense yellow notes (b* ≈ 14.6), while F7 had intermediate parameters (L* ≈ 54.6 and b* ≈ 13.1). Overall, the lower brightness and more accentuated yellow-red hues of the blended flour flatbreads compared with STD may be attributed to the SS and C flours, and possibly to the more extensive Maillard reaction occurring during cooking.
The ∆E values could estimate whether the colour differences between a sample made from blended flour and STD were perceivable to the human eye (the higher the value, the higher the perceived differences between the samples) [42]. A ∆E value of 11.2 was found for sample F6, leading to a strong colour difference from that of STD, while ∆E values > 12 revealed that samples F4 and F7 were perceived as having a different colour from the STD ( Figure S1).

Preliminary Sensory Evaluation
The consumer panel was composed of 15 men (56%) and 12 women aged between 18 and 45 years (37% aged ≤ 25 years, 44% aged between 26 and 35 years, and 19% aged ≥ 36 years) and coming from Ghana (6), Cameroon (5), Morocco (5), Togo (3), Gabon (3), Tanzania (1), Rwanda (1), Tunisia (1), Gambia (1), and Congo (1). As stated in the Material and Methods section, the number of judges was limited due to difficulties in consumer recruitment during the lockdown period related to the COVID-19 emergency. Nevertheless, given the great relevance of the assessment of sensory properties of these flatbreads among people who regularly consume this kind of product, some preliminary indications may be taken as a basis for further studies.
The average scores obtained from the blind test of flatbreads are presented in Table 5. Based on a one-way ANOVA, the flatbread formulation significantly affected (p ≤ 0.05) the acceptability of all measured sensory attributes (overall acceptability, consistency, appearance, and flavour). The scores recorded in the acceptability test for all attributes revealed that the different kinds of flatbread were generally appreciated by the assessors (scores > 5). The consistency of samples STD, F6, and F7 was evaluated in a similar way by the consumers (average score = 6.3), while the consistency of F4 obtained a significantly lower score (5.2). These results were in agreement with the textural properties measured analytically, especially concerning the similar textural perception recorded for F6 and STD, and the worse textural properties measured for F4 compared with STD. With regard to appearance, STD was the most appreciated (7), followed by F6 (6.5) and F7 (6.4); additionally, for this attribute, F4 exhibited the lowest score (5.2). The flavour results showed the highest scores for both the STD and F7 samples (6.9 and 6.8, respectively), followed by F4 (6.2) and F6 (5.4). The overall acceptability results revealed that STD was the most appreciated sample (7.2), followed by F7 (6.5) and by both F4 and F6 (5.7 and 6, respectively), mainly due to their texture and appearance (F4) or flavour (F6).
Penalty analysis (PA) was performed to define potential pathways to product improvement. After aggregating the JAR scores obtained for the attributes of each sample ( Figure S2a) into a three-level scale ( Figure S2b), PA was performed to underscore how many points had been lost due to a product dimension judged "too much" or "too little" by consumers [25]. Table S2 shows the representative PA result (F4 bread). The p-value (significance level = 0.05) calculated for all attributes was not significant, indicating that the attributes evaluated at a non-JAR level were not identified as a potential target for product optimization.
The preference ranking data, collected before providing the panellists a description of the flatbreads' main features, did not reveal any significant differences (p ≤ 0.05) between the samples. After an explanation of the project and samples, the Friedman nonparametric test showed significant differences among the preference ranking data (p ≤ 0.05). The multiple pairwise comparisons (Nemenyi's procedure) revealed that the sample meeting the nutritional criteria (F7) was the preferred flatbread, followed by F6, F4. and STD (Table S3), which was the least preferred, despite the high scores obtained in the acceptability test, possibly due to its perceived lower sustainability.

Conclusions
This work aimed to develop high-quality flatbread formulations for low-income countries by using climate-resilient crops. Wheat, tapioca, sprouted sorghum, and cowpea flour blends were used for this purpose. Among the flatbread prototypes proposed, three of them were chosen on the basis of textural, nutritional (based on the estimated composition of macro and micronutrients), and economic criteria. Then, the three selected prototypes were further characterized for their in vitro starch digestibility, total phenolic content, total antioxidant capacity, physico-chemical, and sensory properties, in comparison with a reference 100% wholewheat flatbread (STD). Overall, if the brief is to produce a flatbread with a physico-chemical profile similar to that of the 100% W reference, the use of a composite made of SS, T, C, and W at percentages of 29.2, 12.5, 49.7 and 8.6 w/w f. b.
(F6), respectively, should be preferred. Moreover, this formulation was less expensive than 100% W in Sierra Leone, Burundi, and Togo, and showed delayed starch hydrolysis and similar TPC and antioxidant activity compared with STD.
If the objective is to meet the nutritional criterion, a flour blend composed of SS, T, C, and W at proportions of 42.1, 7.4, 43.1, and 7.4% w/w f.b. (F7), respectively, should be chosen. Other than having delayed starch digestibility than the 100% W reference, it seemed to provide a higher content of phenolic compounds and antioxidant activity compared with the STD flatbread. Although the derived flatbread textural properties were found to be different from those of the STD when measured analytically, it appears that consumers did not detect significant differences in texture compared with the control, and even the flavour of this flatbread was perceived as acceptable as that of STD. However, consumers' sensory perceptions of flatbreads should be confirmed with a larger panel.
Finally, if the primary objective is economic sustainability, a composite made of SS, T, C, and W at percentages of 48.6, 12.5, 30.3, and 8.6% w/w f.b. (F4), respectively, should be preferred to others. Additionally, this formulation was proven to have a delayed starch digestibility, a higher content of phenolic compounds and antioxidant activity than STD, but worse textural properties.
Overall, the use of composite flour made from climate-resilient crops in breadmaking was proven to be an efficient strategy to promote the use of locally available raw materials and obtain economically sustainable, nutritionally enhanced flatbread with an acceptable technological and sensory profile.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/foods12081638/s1, Figure S1: Selected composite flour flatbreads based on technological (F6), nutritional (F6 and F7), and economic sustainability criteria (F4). STD, standard 100% W bread; Table S1: R 2 and Q 2 of the models generated by the controlled responses; Figure S2: Percentage of respondents (n = 27) providing consumer ratings for the breads' hardness, darkness, bitterness, and bean flavour a) based on 5-point JAR scale; b) collapsed on a 3-level scale; Table S2: Penalty analysis table for bread F4; Table S3: Multiple pairwise comparisons (Nemenyi's procedure) performed on second-preference ranking test data. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

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

Appendix A
Significance of the models generated for the controlled responses: The model generated for the P_f response was characterized by R 2 = 0.9 and Q 2 = 0.2. A probable linear dependence of the response on the SS and C factors was observed: an increase in SS resulted in a probable negative effect on the response (decrease in P_f), while an increase in C was more likely to exert a positive effect (increase in P_f). A negative effect from the interaction of the two factors (SS * C) was also possible.
The model generated for the P_e response was characterized by R 2 = 0.7 and Q 2 = 0.5. A probable linear dependence on factors T and W was observed: an increase in T resulted in a probable positive effect on the response (increase in P_e), while an increase in W more likely has a negative effect (decrease in P_e). The interaction coefficients revealed a positive quadratic dependence of the response on factors W and T (W * W and T * T), indicating the need for a second-order equation to describe the model. There was also an interaction effect of the two factors (W * T), which manifested as a distortion effect of the response surface.
The model generated for the E_f response was characterized by R 2 = 0.7 and Q 2 = 0.2. A probable linear dependence on the factors SS, T, and water was found: an increase in SS or T resulted in a probable negative effect on the response (decrease in E_f), while the increase in water was more likely to result in a positive effect (increase in E_f). As described before, and also in this case, the interaction coefficients revealed a positive quadratic dependence of the response on the factors SS and T (SS * SS and T * T). Finally, the model generated for the E_e response was characterized by R 2 = 0.6 and Q 2 = 0.5. Only the factor SS exerted an influence on the response, and a negative linear dependence was identified (as SS increased, E_e decreased).