*2.5. Statistical Analysis*

The e ffects of experimental variables (factors) and their interactions were evaluated with an analysis of variance (ANOVA). In ANOVA, the *F*-value indicates the e ffect of the independent variable on the response variable. Thus, if F is equal to one, the independent variable has no e ffect, while if *F* > 1 the independent variable has an e ffect and its e ffect is lager as *F* increases above 1. The *p*-value represents the probability of an *F*-value large enough for influencing the experiment; if this value is equal or lower than the significance level, then the assumption of the influence of the independent variables on the response variable is correct. The probability of rejecting the previous assumption even when it is true, is given by the significance level.

A quadratic model with second-order interactions and main e ffects were used to explain a relationship between the given continuous variables as indicated in Equation (2):

$$Z = \alpha\_0 + \Sigma \alpha\_i \mathbb{X}\_{\text{i}} + \Sigma \alpha\_{\text{ii}} \mathbb{X}\_{\text{i}\text{i}}^2 + \Sigma \alpha\_{\text{ii}} \mathbb{X}\_{\text{i}} \mathbb{X}\_{\text{j}} \tag{2}$$

where, Z represents the response variable (content of resveratrol), Xi and Xj are the factors (temperature, and concentration of maltodextrin) and α0, αi, αii and <sup>α</sup>ij are the linear regression coe fficients of the model.

In the process of selecting a model, some parameters of the complete model were first adjusted with Equation (2). Based on a normality test of Anderson Darling for the response variable (R), a transformation of the response was made by the Box-Cox analysis when it was necessary to stabilize the variance. Then, for simplification, the model was hierarchically pruned, and used only with significant factors. Here we present the results obtained with the pruned model, and transformed into response variables.

#### **3. Results and Discussion**

#### *3.1. Content and Retention of Resveratrol*

From 25 experiments, 18 runs were successfully spray-dried, while in seven runs the microstructure collapsed. Since the results for resveratrol were discussed herein, results of physicochemical characterization and yield can be consulted in the work previously reported for quercetin 3-D-galactoside [22], since both works are parts of the same experiment. Table 2 shows the detailed experimental description of the 25 runs and results of content, and retention (R) of resveratrol at di fferent processing conditions. Minimum and maximum values for content, and retention, of 18 spray-dried samples were 0.0–0.47 μg/g, and 0.0–10.24%, respectively. The corresponding average values for these two parameters were 0.28 μg/g, and 6.25%. The highest retention was obtained for run 10 with a value of 10.24%. Overall, average concentration of resveratrol in dried experiments diminished about 96%. This was calculated considering the initial concentration of resveratrol in fresh BJ, and the content of antioxidant determined in dried samples. On the other hand, content and retention values were relatively lower than those obtained for quercetin 3-D-galactoside [22]. Therefore, based on these observations, it is evident that resveratrol is less prone to interact chemically with maltodextrins, being more exposed to thermal degradation during the spray drying process. While some qualitative relations among the independent and categorical variables may be inferred from Table 2, the quantitative analysis of the e ffect of processing variables on content of resveratrol will be discussed in following section.


**Table 2.** Content and retention of resveratrol in spray drying of blueberry juice-maltodextrin (BJ-MX).


**Table 2.** *Cont.*

aAverage and standard deviation values calculated from three repetitions.

#### *3.2. ANOVA and Response Surface Plots Analysis (RSP)*

Table 3 shows ANOVA results calculated for the content of resveratrol. ANOVA results showed that concentration (C) of MX was the variable with the most important effect, while the type of maltodextrin, and inlet temperature had a negligible effect. These observations were confirmed by the *p*-value at a significance level of 0.05. The interactions between the same variable (intra) showed that only the concentration (C2) had an effect on content, but 2.4 times less than the single concentration (C). The rest of interactions between variables (inter and intra) such as T·C, T·MX, C·MX and T<sup>2</sup> showed a *p*-value higher than the significance level, thus their effects were negligible. Therefore, concentration was the independent variable with the larger effect on the content of resveratrol.

These observations indicated that concentration was the main processing variable affecting the content of resveratrol, while the type of maltodextrin and inlet temperature showed no effect on the response variables, suggesting that content of resveratrol was unresponsive to inlet temperatures exerted in the experimental design, and to differences in molecular weight distribution of MXs.


**Table 3.** ANOVA results determined for content of antioxidant in BJ-MX.

a Sum of squares; b mean squares; \* calculated at a significance level of 0.05; T = inlet temperature, C = maltodextrin concentration, MX = type of maltodextrin.

Figure 1 shows RSP for the content of resveratrol as a function of the type of MX. The surfaces shape was similar regardless of the type of MX. In all cases, the highest content of resveratrol was observed at a concentration of MX of 23%. At higher or lower concentration values, content of resveratrol decreased rapidly. However, slight differences were observed with inlet temperature. CM showed a maximum content of resveratrol at 170 ◦C, and a decrement at 210 ◦C. M10 showed the opposite behavior, with a maximum value at 210 ◦C and a decrement at 170 ◦C. High molecular weight maltodextrins (M20 and M40) showed two maximum values at opposite temperatures. All of these observations allowed setting the optimal processing conditions for spray drying of BJ-MX. Highest values for content of resveratrol were obtained at a concentration of MX of 23%, and inlet temperature

of 170 ◦C for CM or 210 ◦C for M10. In addition, RSP results confirmed the utilization limits of MXs, where low molecular weight MXs produced a higher content of resveratrol than high molecular weight MXs. These observations demonstrated that depending on the molecular weight distributions of MXs, these polysaccharides might be employed selectively as carrying agents in spray drying of diverse sugar-rich systems. Indeed, chemical interactions between maltodextrins and antioxidants (i.e., resveratrol) may be a ffected by other variables rather than by the degree of polymerization of maltodextrin. Other variables that may a ffect this interaction into a greater or lesser extent are the volumetric flow of solution injected into the dryer, wet bulb temperature (i.e., relative humidity of air), type of nozzle used (for example, regular versus ultrasonic) and adjuvant agents (i.e., soy protein and sodium alginate). However, in the present work, ANOVA was focused on two spray drying variables (concentration of maltodextrin and inlet temperature) and one categorical variable (type of maltodextrin). In this sense, Ameri and Maa, indicated that increasing the total content of solids in feed solution, increased the recovering of powders in spray drying [32]. Nadeem et al. concluded that yield was related to concentration of maltodextrin, rather than to drying temperature [33]. Caliskan and Gulsah found that increasing the concentration of maltodextrin resulted in an increment in the yield of dried powder [34]. Bhusari et al. attributed this behavior to an increment in the Tg of mixtures [35]. Peng et al. indicated that above 30% of the carrying agen<sup>t</sup> was detrimental for product quality [36]. Saavedra-Leos et al. determined an inverse relation between Tg and DE of maltodextrins [21]. Through several works reported by the Saavedra-Leos group [14,22,37–39], we have observed that in some juices such as orange and blueberries, their physicochemical properties varied with the di fferences in the type and distribution of molecular weight of the carrier agent. In these studies, we have attributed this behavior mainly to: (i) Di fferences in the molecular weight distribution of the carrier agent, (ii) the arrangemen<sup>t</sup> of polymer chains (i.e., entangled or linear) and (iii) the type of molecule in these chains (for example, glucose for maltodextrin or fructose for inulin). Additionally, according to Darniadi, Ho and Murray, when mixing low molecular weight sugars and high molecular active compounds, the active compound tends to segregate in some extent to the surface of dried particles [26]. While this behavior prevents the particles sticking on the dryer walls, it also exposes the active ingredient to a faster degradation.

Table 4 shows the predictive equations for the content of resveratrol as a function of the type of MX. These equations were extrapolated from SRP and in consequence were only valid within the interval of conditions tested herein. Inlet temperature presented a negative e ffect and concentration a positive e ffect. From both processing variables, the extent of concentration was about 1.9 times larger than that of temperature. The values for interactions (T·C), and square of temperature (T2) showed a relatively low positive value, but their numerical contributions were similar to that of concentration. Although the square of concentration (C2) showed a negative e ffect, its value was still lower than that of single concentration, thus indicating a little contribution on the content of resveratrol. Interactions (T·C, T<sup>2</sup> and C2) showed a constant value indicating that these interactions were insensitive with respect to the type of MX. Numerical calculations employing these equations supported the results found from RSP. CM showed the largest content value when using a temperature of 170 ◦C, while for M10 the higher content was obtained with a temperature of 210 ◦C.

**Figure 1.** Response surface plots analysis (RSP) for the content of resveratrol as a function of the type of MX: (**A**) CM, (**B**) M10, (**C**) M20 and (**D**) M40. The symbols (\*) in each RSP represent the center and start points for estimation of second order effects.

**Table 4.** Predicting equations extrapolated from SRP for the content of resveratrol in BJ-MX, as a function of type of MX.


Q = content of resveratrol (μg/g), T = inlet temperature (◦C), C = concentration of maltodextrin (wt. %).

#### *3.3. E*ff*ect of the Chemical Structure in the Content of Antioxidants*

In this section we compared the results of the content of resveratrol reported herein, against those for quercetin 3-D-galacoside previously reported [22]. The effect of spray drying processing conditions showed similar behavior for both antioxidants, but the main difference relied on the values of the content of each antioxidant. In general, the content of quercetin 3-D-galactoside was 2–8 times higher than that of resveratrol. Conversely, according to Shrikanta, Kumar and Govindaswamy, the total content of polyphenols was 1–14 times higher than the total content of flavonoids in underutilized Indian fruits [8]. In this sense, Araujo-Díaz et al. reported a concentration of resveratrol 1.5 times higher than for 3-D-galactoside when employing maltodextrin in spray drying of blueberry juice [37]. Figure 2 shows a schematic representation of the chemical structure of both antioxidants, and a maltodextrin repetitive unit. Resveratrol is a stilbene with a C6-C2-C6 structure and three hydroxyls (OH−). While quercetin is a flavonoid with a C6-C3-C6 structure, five hydroxyls, one alkoxy group (ether) and one carbonyl group (ketone); the galactoside refers to the galactose molecule containing four hydroxyls. On the other hand, maltodextrins are polysaccharide molecules consisting of glucose units linked by glycosidic α-(1-4) and α-(1-6) bonds, with a variable length of its polymeric chains expressed as DE [40]. The glucose molecule contains three hydroxyl groups, one alkoxy group (ether) and one hydroxyl group at each extreme of polymeric chains. All these functional groups are responsible for carrying out molecular chemical interactions such as hydrogen bonding, and Van der Waals interactions. In several works it has been reported that these chemical interactions (inter and intramolecular) are responsible of the adsorption of water on different carrying agents such as inulin and maltodextrins [21,38]. In these works, water adsorption was promoted with increasing the molecular weight of carrying agents. The set of maltodextrins employed as carrying agents in this work, was similar to that reported by Saavedra et al. and presented a degree of polymerization (DP) of: CM 2-12, M10 2-16, M20 2-21 and M40 2-30, units of glucose [21]. Although a high DP may indicate a larger number of hydroxyl groups exposed for chemical interactions, the polymeric chains in MXs may arrange in different configurations rather than linearly, forming entangled branches, thus reducing the availability of active sites. Additionally to the arrangemen<sup>t</sup> of polymeric chains, the stearic hindrance between the adsorbing molecules and MXs, is another aspect affecting the final content of antioxidants. Evidently, the size of quercetin 3-D-galactosie molecule is larger than that of resveratrol, suggesting that stearic hindrance was not the main factor influencing larger chemical interactions with MX, but the number of functional groups available such as hydroxyls, alkoxys and carbonyls. Based on these arguments, it is possible to infer that: (i) The availability of functional groups is the main cause of chemical interactions, since quercetin has more of these groups than resveratrol; hence its greater interaction with maltodextrin, and (ii) in maltodextrins, the availability of these functional groups is limited by branching and entangled of polymeric chains. For this reason, in both cases i.e., the content of resveratrol and quercetin, low molecular weight MXs presented higher antioxidant content than high molecular weight MXs, indicating that polymeric chains in these carrying agents are less branched and entangled, thus promoting more chemical interactions.

**Figure 2.** Schematic representation of chemical structures of resveratrol, quercetin 3-D-galactoside and the maltodextrin repetitive unit.
