Abstract

Bakery foods are the main sources of dietary exposure of acrylamide and 5-hydroxymethylfurfural (5-HMF). Applicable and convenient strategies for mitigation the formation of these neo-contaminants are in great demand. To better understand the potential of combined mitigating strategies on acrylamide, 5-HMF, and browning formation, three additives (sodium bicarbonate [SBC], sodium metabisulfite [SMBS], and citric acid [CA]) were optimized using a Box-Behnken design, and their effects as well as the resulting sensory properties were evaluated. SBC significantly mitigated acrylamide and 5-HMF generation (P < 0.05), and also increased total color changes (ΔE) (P < 0.05), while CA exerted an opposite effect as it promoted the formation of acrylamide and 5-HMF but decreased ΔE (P < 0.05). SMBS limited the formation of acrylamide and ΔE (P < 0.05). In addition, SBC and SMBS reduced acrylamide formation synergistically, and SMBS was the primary factor in decreased browning intensity. Interestingly, SBC inhibited 5-HMF formation via an improved pH value, whereas its inhibition on acrylamide formation were not obstructed by its effect of raising pH. With optimized levels of SBC (1.64%), SMBS (0.11%) and CA (0.50%), the concentrations of acrylamide and 5-HMF in cracker samples were 91.27 ± 11.26 μg/kg and 15.58 ± 0.35 mg/kg respectively, demonstrating 97.60% and 62.72% inhibition rates. Cracker samples with ΔE (4.41 ± 0.11) showed the desirable bright yellow color. This study presents a concrete example of how the control and optimization of selected operative parameters may mitigate multiple specific natural/process contaminants in final food products, while preserving a satisfactory sensorial range.

1. Introduction

Potato (Solanum tuberosum L.) is the world’s most important non-grain food crop and is central to global food security. Increasing global potato consumption drives a demand for processed potato-based products. Among them, potato-based snacks (e.g., chips, crackers, French fries) comprise the largest snack food sector in all markets ¹, due to their desirable taste, texture, convenience, and long shelf life. However, they are also blamed for their contribution to the dietary intake of acrylamide and 5-hydroxymethylfurfural (5-HMF), which pose potential health risks.

Acrylamide and 5-HMF can be generated from a sequence of thermal-triggered chemical reactions during snack food manufacturing. Acrylamide (C₃H₅ON) can be formed as a result of the Maillard reaction between asparagine and carbonyl compounds (reducing sugars), especially in asparagine- and carbohydrate-rich foods ^{2, 3}. In vivo and in vitro studies of acrylamide toxicity confirm that it is mutagenic, carcinogenic, and neurotoxic ^{4, 5}; and can disrupt skeletal development ⁶. Moreover, dietary exposure of acrylamide has been identified as a public health concern due to its association with multiple cancers ^{7, 8, 9}. 5-hydroxymethylfurfural (5-HMF, C₆H₆O₃) is another notable compound formed by Amadori products degradation during the Maillard reaction, or by the acid-catalyzed thermal dehydration of fructose and sucrose ³. From a toxicological point of view, 5-HMF exhibits potential genotoxic and mutagenic activities following metabolic activation to 5-sulphoxymethylfurfural ^{10, 11}. Although the toxicity of 5-HMF to humans is still questionable, 5-HMF reduction practices are gaining importance.

Several methods have been explored by either decreasing chemical precursors or by impeding Maillard reaction pathways. Processing strategies play a critical role, as acrylamide formation occurs during heat treatment. Reducing thermal intensity during processing is an expedient approach for the mitigation of acrylamide and 5-HMF formation, but may decelerate the Maillard reaction rate and degrade the sensory properties of the product ¹². Alternatively, additive agents have been used to inhibit the chemical reactions that generate acrylamide and 5-HMF. Suman et al. reported a significant (50-70%) reduction in acrylamide content in wholegrain and cocoa biscuits by using sodium bicarbonate (SBC) leavening agent. This reduction was attributed primarily to the consequent pH variations ¹³. Other pH-related acidulant such as lemon juice solution (540 ± 0.58 μg/kg) was more effective in reducing acrylamide content in fried potato chips than ginger garlic (720 ± 0.43 μg/kg) and mint leaf (770 ± 0.33 μg/kg) solutions ¹⁴. Amino acids such as lysine and glycine have also been recommended for reducing acrylamide formation in potato-based snacks, as they either form adducts with acrylamide through the Michael addition reaction or compete with asparagine for carbonyls during the Maillard reaction. However, these studies focused primarily on improving inhibition efficiency. A limited number of reports have examined related harmful components (e.g., 5-HMF) in the reactions. On the other hand, compared to single mitigation strategies, joint efforts should be considered for sugar- and asparagine-enriched products such as potato and cereal, as the overall efficacy of combined mitigation strategies may exceed the sum of single measures. Nevertheless, the interactive effects among different mitigating measures on acrylamide and 5-HMF formation have been reported infrequently. Moreover, despite recent progress in elucidating methods to decrease acrylamide content, most studies have been conducted on potato model systems instead of real food products, thus raising questions regarding the practical applicability of their findings.

SBC, sodium metabisulfite (SMBS), and citric acid (CA) are the legal additives for bakery foods in China. SBC and CA are usually used as the leavening agents and acidity regulators in biscuits and bread, while SMBS is frequently used as the bleaching agent and antioxidant in biscuits and starch products. They are also effective inhibitors of Maillard reaction and the consequent formation of acrylamide, 5-HMF and browning. SBC and CA protonate the nucleophilic non-protonated amine (–NH₂) of asparagine to non-nucleophilic protonated amine (–NH₃⁺) and block the nucleophilic addition of asparagine with electrophilic carbonyl carbon of reducing sugar that inhibits Maillard reaction and subsequently the acrylamide formation ¹⁵. SMBC is frequently used as an inhibitor of browning in Maillard reaction because the neucleophilic sulfur atom binds to the carbonyl group of sugars before the amino groups ¹⁶.

The aim of our study was to examine the levels of reducing acrylamide and 5-HMF formation in commercial potato-based crackers, with a focus on the effects of combining additive treatments (SBC, SMBS, and CA). Response surface methodology was performed to evaluate additive parameters with an experimental Box–Behnken design, and the interaction effects among these three agents were evaluated. Meanwhile, total color changes (ΔE) and browning index (BI), based on the Commission Internationale de l’Elcairage (CIE) values (L*, a*, and b*), were used to quantify the browning intensity resulting from Maillard reaction. The partial sensory properties of the crackers, including hardness and fracturability, were also characterized.

2. Materials and Methods

Food-grade cracker materials of potato flakes, wheat flour, cassava starch, sucrose, salt, dried milk powder and shortening were kindly gifted by Zhejiang Little Prince Food Co., Ltd. Food-grade SBC, SMBS and CA were purchased from a food additive market in Hangzhou. Standard preparations (≥95 % by HPLC) of acrylamide, acrylamide-d3, and 5-HMF were purchased from Sigma-Aldrich (Merch, Germany). All other chemical reagents were of analytical grade and obtained from Aladdin Reagent Corporation (Shanghai, China).

Cracker samples were prepared at laboratory scale according to the following optimized recipe: 600.0 g potato flakes, 160.0 g wheat flour, 60.0 g cassava starch, 63.0 g sucrose, 7.0 g salt, 50.0 g milk powder, 60.0 g shortening (1000 g in total). It is the recipe of a cracker produced by Zhejiang Little Prince Food Co. Ltd., using 0.82% ammonium bicarbonate as the leavening agent. Ammonium bicarbonate has been shown to contribute, promote, and accelerate acrylamide formation in bakery products. Therefore, based on the results of our preliminary experiments, ammonium bicarbonate was replaced with SCB (0.82 -1.64%), SMBS (0.05 - 0.25%), and CA (0.1-0.5%) at levels determined according to a three-factor, three-level Box-Behnken Design (Table 1). Cracker samples produced from the optimized recipe with 0.82% ammonium bicarbonate were used as the control.

All ingredients in each recipe were mixed with 500 mL distilled water using an electronic mixer (HM-945, Donlim, China) for 5 min, followed by 5 min of hand kneading. Dough was then cut into round discs with a height of 2 mm and a diameter of 5 cm and baked at 200 °C for 5 min in an oven (T4-L326F, Media, China). Three consecutively replicated bakings were conducted to obtain cracker samples from each recipe. Baked cracker samples were collected for texture analysis, and then ground and passed through a 60 mesh sieve, sealed tightly, and stored at -18 °C.

pH values of dough preparations before baking were measured according to the reported method ¹⁷. Briefly, 5 g dough samples were soaked in 45 ml distilled water for 1 h. The suspensions were homogenized with a handled ultrasonicator (Ultrasonic processor FXID: 6m2003, Fangxu, China) for 5 min. The pH values of the suspensions were determined with a pH meter equipped with a glass electrode (FE28-Standard, Mettler, Switzerland).

To extract acrylamide, 2 g ground cracker samples, 0.1 ml acrylamide-d3 aqueous solution (5 μg/ml) as internal standard, and 9 ml ultrapure water were homogenized in a 50 ml centrifuge tube. Then, 10 ml acetonitrile and 5 ml n-hexane were added to the homogenate, followed by vortex mixing for 1 min. After adding 1.5 g anhydrous sodium acetate and 6 g anhydrous magnesium sulfate to the centrifuge tube, extracting was conducted at room temperature for 1 h with shaking (60 rpm). The extract was centrifuged at 10000 rpm, 4 °C for 8 min to separate the acrylamide into the acetonitrile phase. An aliquot of 3 ml acetonitrile phase was cleaned with an MCX solid phase extraction cartridge. The filter was collected and dried under a nitrogen stream. The residue was re-dissolved in 0.3 ml ultrapure water and filtered through a 0.45 μm filter before the chromatograph analysis.

The chromatographic separation of acrylamide was performed by using the UPLC-ESI-MS system (Waters Acquity, USA). The UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm) was maintained at 30°C. The injection volume for each sample was 10 μl. Using 0.1% formic acid in water (A) and methanol (B) as mobile phase, a gradient elution was performed with flow rate of 0.2 ml/min and gradual changing of the mobile phase as follows: 0 ~ 3.0 min, 0% B; 4.0 ~ 5.0 min, 95% B; 5.2 ~ 9.0 min, 0% B. Electrospray ionization mass spectrometric detection was conducted in positive ion mode. The capillary voltage was 3.0 kV and the cone voltage was 30 V. The source temperature was 150 °C. Nitrogen was used as the desolvation gas at 1000 L/h with the temperature of 500 °C. For the transitions, m/z 72 (acrylamide) and m/z 75 (acrylamide-d3), m/z 55 (acrylamide) and m/z 58 (acrylamide-d3) were used with collision energies all set to 15 V.

5-HMF was extracted by suspending 1 g ground cracker samples in 10 ml ultrapure water and ultrasonic homogenizing for 30 min. The extracts of 5-HMF were clarified by adding 0.5 ml each of Carrez I and Carrez II solutions, standing by for 5 min, then centrifuging at 8000 g for 10 min. Supernatants were collected and filtered through 0.45 μm filters for chromatograph analysis.

A Shimadzu HPLC system (LC-2030C 3D Plus) was employed to quantitatively analyze 5-HMF concentrations in cracker sample extracts. The mixture of methanol: water (15% v/v) was used as the mobile phase, and the Nucleosil C18 column (250 × 4.6mm, 5 μm) was maintained at 25 °C. The injection volume for each sample was 10 μl. An isocratic elution program was performed with the flow rate of 1.0 ml/min. The Diode Array detector was set at 280 nm. Quantification was performed by an external standard calibration of 5-HMF (0 ~ 50 μg/ml).

Color of the cracker samples (in powder) was measured with a tristimulus colorimeter (ColorFlex EZ, HunterLab, USA). Ground cracker samples were loaded into a quartz sample cup and then tested in duplicate. Three color parameters, L*(lightness/darkness), a*(redness/greenness), b*(yellowness/blueness) were directly displayed in the color model. The ΔL (L* - L*₀), Δa (a* - a*₀) and Δb (b* - b*₀) were calculated to describe color changes in three color scales, where L*, a*, b* were the color values from cracker samples with different recipes, and L*₀, a*₀, b*₀ from the control sample of crackers. Total color changes (ΔE) were calculated as follows: ΔE = (ΔL² + Δa² + Δb²)^0.5. Browning index was also calculated based on the CIE values of L*, a*, b*, for describing the browning intensity of cracker samples: Browning index = 100 × (k - 0.31)/0.17, where k = (a* + 1.75L*)/(5.645L* + a* - 3.012b*) ¹⁸.

Cracker samples with uniform size and shape were selected for texture analysis. Puncture tests were conducted using a texture analyzer TA-XT2 (Shanghai, China) equipped with a spherical probe (P/0.25S) and a 10 kg load cell. The parameters for analysis were pretest speed of 10 mm/s, test speed of 1 mm/s, post-test speed of 1 mm/s, compression distance of 10 mm, and threshold force of 5 g. The force exerted to break the cracker is defined as hardness (g), and the linear distance between the first and the last fracture registered is termed fracturability (mm). The mean value was calculated from 6 independent assays of each sample.

Residual concentrations of sulfur dioxide in cracker samples were analyzed according to the method of Pharmacopoeia of the P. R. China 2020, using a distillation apparatus consisting of a flask with a liquid adding funnel and inert gas inlet, a back-condenser, and two successive distillate collecting tubes. 10 g ground cracker samples and 350 ml redistilled water were added to the flask. 50 ml of 3% hydrogen peroxide solution was used to collect the distillate (SO₂), by adding 3 drops 2.5% alcoholic methyl red solution as an indicator. The apparatus was conditioned by exposure to N₂ (0.2 L/min) for 10 min. After 10 ml HCl (6 M) was added, the contents of the flask were gently boiled and distilled under 0.2 L/min N₂ for 90 min. The distillate was titrated with 0.01 M NaOH solution. The residual concentrations in the cracker samples were then calculated based on the equivalent weight rule between reactants. A control experiment was performed to deduct the sulfur dioxide contents in background.

The Box-Behnken design is a suitable methodology for ascertaining the effects of formulation ingredients (factors) and their associated effect on responses (dependent variables) ¹⁹. In the present study, the levels of SCB (X₁), SMBS (X₂), CA (X₃) in cracker recipes are the input factors, while the concentrations of acrylamide and 5-HMF, ΔE, hardness, and fracturability of cracker samples are reported as the response variables. The layout of the Box-Behnken design, composed of 17 trials including 5 central points was given by the Design Expert software program. The response variables were fit to the quadratic model, which was expressed with the following polynomial equation:

(1)

Where Y_Z are the response variables (Y₁, acrylamide; Y₂, 5-HMF; Y₃, ΔE; Y₄, hardness; Y₅, fracturability), while X₁, X₂, and X₃ are the input factors. β₀ is the intercept and β₁ ~ β₉ are the regression coefficients.

Analysis of variance (ANOVA) was done to evaluate the fitness of response variables, and to investigate and validate the factors which mitigate the formation of acrylamide and 5-HMF, and affect the organoleptic properties of the cracker samples. Mathematic equations were developed to predict the response variables. Then, optimization was performed by using multi-response desirability function.

Response surface experimental design, data analysis, and generation of surface plots were performed by using Design Expert® 11 (Stat-Ease, Minneapolis, MN, USA). Pearson’s two-tailed correlation coefficient analysis was also conducted to determine the correlation among variables, by using the software package IBM SPSS statistics 19.0 (IBM Corporation, Armonk, NY, USA). All statistical parameters were evaluated at P < 0.05.

3. Results

Remarkable changes in the concentrations of acrylamide (Y₁), 5-HMF (Y₂), and ΔE (Y₃) were observed among cracker samples with different levels of SBC, SMBS and CA. As shown in Table 2, Y₁, Y₂, and Y₃ ranged from 15.37 ± 1.79 ~ 3283.99 ± 265.30 μg/kg, 9.37 ± 0.63 ~ 35.79 ± 1.36 mg/kg and 2.39 ± 0.11 ~ 11.41 ± 0.15, respectively. The browning indices of the cracker samples lied in 50.67 ± 0.71 ~ 64.77 ± 0.86 (as shown in Table S1.). ΔE was in accordance with the browning index, because significant positive correlation was determined between these two variables (r = 0.904, P = 0.000). Herein, ΔE was used in this study to measure the browning development in the cracker samples.

The ratios of maximum to minimum for Y₁, Y₂, and Y₃ response were 213.66, 3.82, and 4.78, respectively. A ratio ˂ 10 is usually acceptable. The ratio of 213.66 suggests high heterogeneity and non-normal distribution of the acrylamide data, which was confirmed by the Kolmogorov-Smirnov test (P < 0.05). High variation of response variables may result in the poor performance and low quality of response surface methodology optimization ²⁰. Therefore, Y₁ was transformed into the natural log scale (signed as Y₁'), to stabilize variance and to achieve log-normal distribution. This can be written as follows: Y₁' = ln (Y₁ + a), where a is constant and taken to be zero. Differences in hardness (Y₄) and fracturability (Y₅) of the cracker samples were much less than Y₁ ~ Y₃, which ranged from 419.78 ± 31.69 ~ 675.53 ± 107.86 g and 0.41 ± 0.07 ~ 1.10 ± 0.23 mm, respectively, as shown in Table S1. The ratios of maximum to minimum were 1.61 and 2.68, respectively.

To explain the relationship between input factors and response variables, data of Y₁', Y₂, Y₃, Y₄, and Y₅ were collected to fit the quadratic model. The statistical ANOVA results for the quadratic polynomial model are provided in Table 3. Herein, the higher F-value and lower P-value (< 0.05) of the model associated with the larger P-value of lack of fit (P > 0.05) and the multiple correlation coefficient (R²) close to 1 indicate a well-fitting regression model ¹⁹.

As shown in Table 3, the P-value of models of these three response variables were < 0.0001, and the P-value of lack of fit test for Y₁', Y₂, and Y₃ were 0.5814, 0.2567 and 0.0673, respectively. The high R² values of 0.9974, 0.9781, 0.9843 indicated that the models failed to explain only 0.26%, 2.19%, and 1.57% of total variations of Y₁', Y₂, and Y₃, respectively. The values of adjusted R² (0.9940, 0.9750, and 0.9870) were closed to the values of R². All of these suggested that Y₁', Y₂, and Y₃ fit the quadratic model well.

However, the data of hardness (Y₄) and fracturability (Y₅) failed to fit the quadratic model. Neither fit the two-factor interaction or linear models, suggesting that the changes of texture data were not significantly related to the input factors within the experimental ranges.

The following equations (2) – (4) were developed to predict response Y₁', Y₂, and Y₃ in coded factors.

(2)

(3)

(4)

The linearity relationship between the actual and the predict data (Figure 1a - Figure 1c), associated with high values of signal to noise ratio (60.92, 28.91 and 22.60) demonstrated adequate precision of the models for predicting Y₁', Y₂, and Y₃ within the variable ranges chosen in this study.

The variance contribution rate was calculated to reflect the importance of input factors. The larger the calculated values, the larger its weight on the response. Additionally, a positive sign of coefficient signifies the collegial effect while a negative sign signifies the opposite effect of the factor on response variables.

As reported in Table 3, for Y₁' response, the variance contribution rates of the linear terms X₁ and X₂ were 39.62% and 40.33% respectively, much larger than that of X₃ (7.61%). Associated with the negative signs of the coefficients of X₁ and X₂, it suggested that SBC and SMBS were the main factors which mitigated acrylamide formation in the cracker samples, while CA promoted the formation of acrylamide to some extent. SBC was the dominant factor that inhibited the formation of 5-HMF, with evidence that the linear term X₁ contributed 92.17% of total variations of Y₂. While adding CA favored 5-HMF generation in cracker samples. The effect of SMSB on 5-HMF formation was negligible, with the variance contribution rates of 0.08. Adding SMBS and CA both limited the development of browning in cracker samples, while adding SBC favored the browning development.

Three-dimensional (3D) response surface plots were drawn to illustrate the interaction effects between the two factors when the third factor was kept at the center point, as shown in Figure 2.

Y₁' decreased with increasing SBC and SMBS levels in the experimental range, when CA was kept at 0.3% (Figure 2a). Adding CA attenuated the effects of SBC and SMBS on acrylamide migration, as lower Y₁' was observed when CA was at the lower level (Figure 2a and 2b). Additionally, in our preliminary experiments, acrylamide concentrations in cracker samples were respectively 2667.91 ± 188.30 μg/kg and 274.06 ± 23.75 μg/kg, when SBC was added alone into the optimized recipe at levels of 0.82% and 1.64%, respectively. The acrylamide concentration was reduced to 15.37 ± 1.79 μg/kg with the combined using of SBC and SMBS in the Run 14 of Box-Behnken design (Table 2), despite adding CA might promote acrylamide formation. These results suggested the synergy of SBC and SMBS on the mitigation of acrylamide formation.

The 3D plots of Y₂ are shown in Figure 2d-Figure 2f. Significant interaction was only detected between X₂ and X₃ (P < 0.05). When SBC level was kept at the central point, the response Y₂ increased with the increasing levels of CA when SMBS was at the low level (0.05%), but decreased with increasing CA levels when SMBS was at the high level (0.25%) (Figure 2f).

The 3D plots of response Y₃ are shown in Figure 2g - 2i. Lower ΔE was identified when X₁ < 1.23% and X₂ > 0.15% (Figure 2g and 2h). The effect of CA on ΔE was more obvious when SBC was at high levels (1.64%) and SMBS was at the low level (0.05%) (Figure 2h and 2i).

In the present study, the pH values of dough samples were in the range of 6.96 ~ 8.05 with combined using SBC, CA and SMBS. Spearman’s correlation analysis was conducted to determine the main factors related with dough pH, and the effects of dough pH on the response variables. The results showed that the SBC level was the main factor related with the dough pH (r = 738, P = 0.001), followed by the CA level (r = - 0.504, P = 0.040). The dough pH was negatively related with Y₁' (r = - 0.816, P = 0.000) when using the SMBS level as control variable, because the SMBS level was not significantly related with the dough pH (r = - 0.230, P = 0.375) but contributed the most to acrylamide reduction (Table 3). Dough pH was also negatively correlated with Y₂ (r = - 0.708, P = 0.001), but was positively related with Y₃(r = 0.548, P = 0.023). These results suggested that raising dough pH might reduce acrylamide and 5-HMF formation, but promote the browning development.

Desirability function of Design Expert software was employed to determine the optimum levels of SBC, SMBS and CA for lowering acrylamide and 5-HMF concentrations and browning intensity. Constraints were given to the factors and the measured responses, as shown in Table 4. Responses Y₁', Y₂, and Y₃ were targeted for minimization. Because SMBS may cause sulfur dioxide residue and thereby increase the risk of asthma in sensitive individuals, the level of SMBS was set as low as possible to diminish the residual concentration of sulfur dioxide in cracker samples.

The optimum levels of SBC, SMBS and CA for minimum formation of Ln[Acrylamide], 5-HMF and ΔE were 1.64%, 0.11% and 0.50%, respectively. The desirability values of Y₁', Y₂, Y₃ and combination were 0.70, 0.79, 0.78 and 0.74, respectively. The desirability bar graph is reported in Figure 3.

To validate the optimization, cracker samples were made in triplicate with SBC, SMBS and CA at the optimized levels, as shown in Figure 4. Acrylamide and 5-HMF concentrations and ΔE were measured and compared with the predicted values, as shown in Table 5. The biased percentages were less than 5% for all response variables. While the bias was 12.95% on the base of actual acrylamide concentration.

The formation of acrylamide and 5-HMF in the verification test was inhibited by 97.27% and 62.72% respectively, compared with the concentrations of acrylamide and 5-HMF in the control cracker sample (Table 2). The cracker samples from the verification test were in bright yellow, which is a desirable color for bakery foods. The hardness and fracturability were 436.38 ± 37.82 g and 0.95 ± 0.10 mm respectively, which were within the experimental ranges. The residual content of sulfur dioxide was 67.37 ± 1.86 mg/kg. More effort is needed to diminish the level of residual sulfate, though the present amount was less than the benchmark level for biscuit products (0.1 g/kg) stipulated in the Chinese food criterion GB 2760.

4. Discussion

pH value is the strong modulator for Mallard reactions and impacts the formation of acrylamide and 5-HMF. Acidity regulators, such as CA, acetic acid, and sodium acid pyrophosphate has been widely used to limit acrylamide levels in various foods ²¹. Acidification might mitigate acrylamide formation by decreasing the generation of methylglyoxal and glyoxal, but promote 5-HMF formation by enhancing the generation of 3-dexoyglucosone ²², while alkalization favored the increasing of α-dicarbonyl compounds formation and the decreasing of 3-dexoyglucosone ²³. However, when using SBC as the leavening agent, over 50% acrylamide mitigation was achieved in various foods, though adding SCB increasing the pH values of the matrix before thermal-processing ^{13, 22, 24}. It seemed that pH values were not the determined factors for acrylamide formation when SBC adding into the formulation. Recently, Gülcan et al. reported that over 50% of acrylamide formation in bread was reduced by using CO₂ as the baking atmosphere, while the acrylamide formation was almost blocked with SO₂ as the baking atmosphere ²⁵. During baking, both SBC and SMBS decomposed and generated CO₂ and SO₂, respectively. The CO₂ filled in the baked foods, made the porous texture, and might have acted as an inert and inhibiting atmosphere that mitigated acrylamide formation in the cracker samples. The observed synergistic effect of SBC and SMBS on acrylamide mitigation in the present study might be due to the collegial effects of the generated CO₂ and SO₂.

Tas & Gökmen reported that alkalization favored the decreasing of 3-dexoyglucosone ²³, the important intermediate related with 5-HMF formation ^{26, 27}. Kavousi et al. also found that a higher amount of 5-HMF was formed at a lower pH at frying temperature in model system ²⁸. Sung & Chen reported that pH of cookies increased to 10.33 during 10 min-baking when 0.8% SBC was used as a leavening agent, due to SBC decomposing into sodium carbonate ²⁹. All of these results suggested that alkalinization caused by SBC drastically reduced 5-HMF formation.

During the thermal processing of foods, browning develops primarily as the result of Maillard reaction and caramelization ³⁰. Decreasing total color changes and increasing the brightness of carbohydrate-rich foods by adding SMSB or using a SO₂ baking atmosphere have been documented previously ^{28, 31, 32}, because the nucleophilic property of sulfur atoms facilitated the preferential binding of SO₂ to the carbonyl groups of sugars rather than amino groups ²⁵. Previous studies have reported that a pH increase from 4.0 to 8.0 strongly enhanced the polymerization of carbonyl compounds generated during caramelization ^{33, 34}. Therefore, the mitigation of browning in the cracker samples probably resulted from sulfur-mediated inhibition of the Maillard reaction and pH-mediated inhibition of caramelization.

5. Conclusions

By employing crackers as a model analyte, a strategy for mitigating the formation of acrylamide, 5-HMF and browning by combining three food additives was proposed by a Box-Behnken design and a desirability function approach. Optimal mitigation was obtained with SBC (1.64%), SMBS (0.11%), and CA (0.5%), without compromising the sensory quality of the product. The combined additives mitigated 5-HMF and browning formation by sulfur-mediated inhibition of the Maillard reaction and pH-mediated inhibition of caramelization. The mitigation of acrylamide formation resulted from the synergistic effects of SBC and SMBS. The inhibition of SBC on acrylamide formation was not obstructed by its effect of raising pH. More efforts are needed to understand the mechanism.

Acknowledgements

The authors would like to thank Jianlong Han for the excellent technical support in the acrylamide concentration analysis.

References

Supplement