Production Process Effect on Mexican Agave Syrups Quality : A Preliminary Study

In this work, the quality parameters of commercial agave syrups produced in five different Mexican states and with different production process were evaluated. Regulated parameters (pH, Moisture, Ashes, and 5-(hydroxymethyl) furfural), as well as color and aw, were measured on 25 agave syrups, including traditional samples as controls. Traditional and semi-industrial syrups were samples obtained by thermal hydrolysis. Additionally, the semi-industrial process included control of process variables such as pH, °Bx, and temperature. The industrial process is technified. The agave syrups ranged from 70-76 °Bx, pH ranged from 3.2-6.7, and moisture from 20.2-28.6%. The aw values shown a wide variation as well as L* a* and b* color parameters. Some of those parameters shown significant differences in ANOVA analysis; however, most of the samples complied with the norm. General Discriminant Analysis (GDA) made it possible to discriminate between production process by using pH, % Ash, b*, 5-(hydroxymethyl)furfural and a* parameters. Further analysis of a wide range of syrups and the inclusion of non-regulated compounds such as volatile compounds and carbohydrates are needed to get more information for a deeper characterization of agave syrups.

Agave syrup is produced by three different processes (Fig. 1): Traditional (A) which consist on heating the sap in pots placed on the fire to hydrolyze carbohydrates and kept there until water evaporation.Empirical knowledge is mainly used for the control of the process.The traditional process is currently done as described by Spanish chroniclers (Pardo, 2007).Semi-industrial process (S) uses the thermal hydrolysis of sap, but only a few variables are controlled, for instance: pH, °Bx, and temperature; suggesting the use of basic quality control practices.Industrial (I) is the most sophisticated process of production and involves the use of the whole agave pine (raw or semi-cooked), the careful control of the process variables and also the use of more sophisticated technology for carbohydrate extraction and its subsequent hydrolysis.The main methods of carbohydrates industrial hydrolysis are: enzymatic, acid (García-Aguirre et al., 2009;Willems & Low, 2012) and the combination of them (García-Aguirre et al., 2009;Montañez-Soto, Gonzá lez-Herná ndez, Venegas-Gonzá lez, Bernardino Nicanor, & Gonzá lez-Cruz, 2013).process (sap, leaves or pine), generates a wide range of products sold under the name of agave syrup.
Fructose-rich syrups have received increased attention in recent years, becoming the most demanded sweeteners by the global food and pharmaceutical industries due to their functional and technological advantages over sucrose (García-Aguirre et al., 2009;Mellado-Mojica & López, 2015;Montañez Soto et al., 2011) and also by their beneficial health effects (García-Aguirre et al., 2009), related with the presence of bioactive compounds such as amino acids and fructanes (Muñiz-Má rquez et al., 2015), giving to it antioxidant capacity and antibacterial properties among others.
Mexican regulation NOM-003- SAGARPA-2016SAGARPA- (2016) ) establishes the food quality, authenticity, labeling and conformity assessment of agave syrup.Despite the existence of mandatory regulation, the characteristics and chemical composition of the products sold as agave syrup are wide, and there is no guarantee of the authenticity of the product.This study aims to evaluate the quality and the concordance with norms of commercial agave syrups.

Agave Syrup Samples
For this study, a total of 25 samples of agave syrup from six different states of Mé xico were analyzed.From them, 12 samples were from Jalisco (J), one from Coahuila (C), four from Estado de Mé xico (E), one from Hidalgo (H), six from Queré taro (Q), and one from Nuevo León (N).The samples were classified depending on its production process: 20 of the samples were Industrial (I), 3 were Semi-industrial (S), and 2 of the samples were Traditional or artisanal (A) (Figure 1).Traditional samples were included in the study as controls and were obtained from producers who guaranteed its inclusion in this category (process and raw material).Traditional samples were produced in Jalisco and Queretaro from A. tequilana Weber var.azul and A salmiana respectively.
pH measurements were performed using an Oakton 510 pH meter (Oakton Instruments, Vernon Hills, Il, USA), calibrated at pH 4 and pH 7 at 20±1 °C, in a solution made with 10 g of agave syrup and 100 mL of distilled water.
Total soluble solids (°Bx) and refractive index were obtained using an Abbé refractometer NAR-1T LIQUID (Atago Co., LTD, Minato-ku, Tokyo, Japan) at 20±1 °C, on homogenized samples.Moisture content (%M) was calculated according to the procedure in the Mexican regulation.
Ashes were quantified gravimetrically: 5-10 g of sample was placed in a porcelain crucible and then carbonized using a Bunsen burner; and placed on a muffle at 600±25 °C until constant weight.
The content of 5-HMF was determined using a 1200 series HPLC, equipped with a quaternary pump, autosampler, diode array detector, degasser and thermostated column compartment (Agilent Technologies, Santa Clara, CA, United States).A Zorbax Eclipse XDB C18 column of 4.6 x 150 mm packed with 5 µm particles size (Agilent Technologies) was used.A sample of 10 g was dissolved in MillyQ water until a volume of 50 mL, and filtered using a 0.45 µm PTFE filter, 13 mm diameter (Agilent Technologies).An aliquot of 20 µL of the sample solution was injected using a 0.5 mL min -1 isocratic flow of methanol-water (10:90 v/v).The detector was set at 285nm.Run was performed for 7 minutes.A calibration curve of 5-HMF (0.00 to 0.1 mg/kg) was used, with a minimum r 2 of 0.98.

Color
The agave syrup color was measured using a Minolta CM-5 spectrophotometer (Konica Minolta) standardized with both black and white standards on transmittance mode, using a Cell CM-A98 (10 mm), at 20±1 °C.The color was reported as Hunter L*, a*, b* tristimulus values.All the samples were analyzed using a D65 illuminant.

Water Activity
Water activity (a w ) was measured using an Aqualab dew point water activity meter 4TE, (Decagon devices, WA, USA), at 20±1 °C, calibrated with a water activity standard of 6 mol kg -1 NaCl (0.76 a w ).

Statistical Analysis
All the statistical analysis were carried out using the STATISTICA software 13.3 (TIBCO software INC).Full factorial ANOVA analysis was performed for all the physicochemical parameters analyzed by the state as well as by production process to identify significant differences (p<0.05).Additionally, the Fisherś least significant difference (LSD) test was performed for each significant physicochemical parameter for testing the population means to group the samples according to the evaluated factor.After that, a General Discriminant Analysis (GDA) was carried out to reduce the number of variables that made it possible to discriminate between samples; both, by state or production process.For both data sets forward stepwise analysis (p inclusion 0.05, p exclusion 0.05) was carried out to reduce the number factors for the discriminant model.

Quality Parameters
Results obtained for quality parameters of the analyzed samples are shown in Table 1.°Bx, pH, and %M showed values outside the range; for instance, °Bx had lower (70 °Bx), and %M had a higher value (28.6 %M) than the norm parameter.The other regulated parameters presented values between the limits (NOM-003- SAGARPA-2016SAGARPA- , 2016)).Lower values on °Bx and higher %M indicates a lower content of sugars than specified in the norm.A lower content of sugars could be associated with a higher susceptibility to microorganisms' growth.The obtained °Bx values were lower than those reported by Willems and Low (2012) and in the range of the obtained by Mellado-Mojica and López (2015) with commercial syrups (74.61-77.47 and 65-76, respectively).Wide variability in a w and color was observed (Table 1).Since °Bx and %M are related to a w , their variations could explain the wide range of values observed in this parameter, which is related to variations in shelf life.Montañez-Soto et al. ( 2013) obtained higher a w average value in agave syrups prepared under laboratory conditions.
Traditional syrups, with the darkest color, showed the highest values for L* (100), a* (48.01), and b* (100).Darker colors are mainly associated with caramelization reactions induced by high temperatures during the heating process, and in a lower degree by the occurrence of Maillard reactions (Mancilla-Margalli & López, 2002;White, 2009).It was found a greater variability on L*, a* and b* color parameters than those reported by Willems and Low (2012).Their low variability could be attributed to the fact that they only worked with industrial syrups.

Analysis of Variance (ANOVA)
Results of ANOVA are shown in Tables 2 and 3.According to the state of production (Table 2), pH showed highly significant differences (p<0.000),being Q the state with the highest value.5-HMF also shown highly significant differences (p<0.000),being C the state with the highest content.Regarding color, a* was the only parameter that showed significant differences (p<0.05),being N the state with the most noticeable difference in color.A change on the sign of the axis for this parameter (-a*, green) was observed, respect to the other regions which were (+a*, red), meaning differences on the tone of color of the samples.-a* suggests less production of caramelization pigments.According to the production process (Table 3), significant differences (p<0.05) were observed for RI, °Bx, a w , pH and b*.Samples of A process had the highest significant values a w and pH, and on the contrary, RI, °Bx, and b* had the lowest ones.Those differences could be associated with the lower content of sugars in A samples.On the other hand, I and S samples did not show significant differences between their quality parameters.

Correlation Analysis
Correlations between quality parameters are shown in Table 4.It was observed significant positive (p<0.05)correlation between RI and °Bx.Those parameters are dependent on the content of sugar in the samples: The higher the content of sugars, the greater the °Bx and RI values.Both RI and °Bx are significant inversely correlated with a w , which corresponds to the highly bounded state of the water and carbohydrates on samples.RI and °Bx are also positively correlated with color parameters: L * and b * .It was observed a significant positive correlation between pH, a * , and a w , and inversely correlated with % Ash, 5-HMF, °Bx.A significant negative correlation was found between 5-HMF and L * .Being 5-HMF faint yellow colored, as expected, a decrease in the dark color of the samples was observed.The a * is significant and positive correlated with b * and negatively with L * .0.454 # Significant (p<0.05);RI: Refractive Index; a w : Water activity, 5-HMF: 5-hydroximethyl furfural (g/100 g)

Discriminant Analysis
Forward stepwise analysis for the state of production let to the selection of tree parameters (p inclusion 0.05, p exclusion 0.05): 5-HMF, a* and b*.Only 60% of the samples were correctly classified according to the state of production.Cooman graphics of Mahalanobis distances (Figure 2A) did not classify all the samples according to the state of production.The best discrimination was obtained for C samples (Coahuila) where 100 % were correctly classified followed by J samples (Jalisco) and Q samples (Queré taro) with 83% and 80% of correctly classified samples, respectively.On the other hand, none of the samples from E (Estado de Mé xico), H (Hidalgo) and N (Nuevo León) were correctly classified.5-HMF and a* variables selected for the state of production GDA model were also significant on ANOVA.Forward stepwise analysis for production process let to the selection of five parameters (p inclusion 0.05, p exclusion 0.05): pH, % Ash, b*, 5-HMF and a*.With the obtained model 92% of the samples were correctly classified.Misclassification was observed for S (semi-industrial) samples where only 33.33% were correctly classified (Figure 2B).S misclassified samples were classified as I (industrial), this suggests that these samples were produced with a similar process than industrial ones.
From the variables % Ash, pH, and b* were also significant for the ANOVA for the production process.Since pH and temperature are controlled in the production of semi-industrial agave syrups, and the temperature is related to the production of 5-HMF, it was not surprising that those parameters were selected by the discriminating model.
It was reported that furanic compounds, i.e. 5-HMF are not present on raw agave, there are produced during a cooking process (Prado-Jaramillo, Estarrón-Espinosa, Escalona-Buendí a, Cosí o-Ramí rez, & Martí n-del-Campo, 2015).Lower contents of 5-HMF have been associated with enzymatic hydrolysis of pines (García-Aguirre et al., 2009) that is a regular process of hydrolysis on industrial production.It was not surprising that 5-HMF and some color parameters were selected as discriminant variables because of its formation is highly related with the thermal treatment.
Even when several authors have done studies with agave syrups with different states of production, those studies do not use statistical analysis such as ANOVA or GDA to find out differences among samples (Mellado-Mojica & López-Pé rez, 2013;Willems & Low, 2012).

Conclusions
A wide variability of physicochemical quality parameters was observed in the analyzed samples, this variability was confirmed by the ANOVA analysis.GDA made it possible the discrimination between samples according to the production process, but not by the state of production.Traditional control samples were well-discriminated form semi-industrial and industrial ones.Further analysis of samples ensuring the representativeness of each analyzed category, and the inclusion of the analysis of non-regulated compounds such as volatile compounds and carbohydrates are needed to get more specific information that made it possible a deeper characterization of agave syrups.

Table 2 .
ANOVA and Fisherś LSD Test ResultsAccording to the State of Production.

Table 3 .
ANOVA and Fisherś LSD Test ResultsAccording to the Production Process

Table 4 .
Correlations Between Analyzed Quality Parameters.