QUERCUS HUMBOLDTII (COLOMBIAN OAK): CHARACTERISATION OF WOOD PHENOLIC COMPOSITION WITH RESPECT TO TRADITIONAL OAK WOOD USED IN OENOLOGY

The use of new oak barrels and the demand for oak wood in oenology is increasing. Thus, it is necessary to search for new wood sources to supply the current demand in cooperage. The aim of this work was to study the composition of ellagitannins and low molecular weight phenolic compounds (LMWP) by HPLC-DAD of green wood Quercus humboldtii Bonpl . (Colombian), compared to the species typically used in cooperage: Quercus sessiliflora Salisb. (French and Romanian) and Quercus alba L. (USA). In Colombian oak wood, the same LMWP and ellagitannins as in the traditional species were identified. The most abundant LMWP was ellagic acid, and its concentration was lower in Q. humboldtii and Q. alba than in Q. sessiliflora , so these two presented a lower content of phenolic acids. As regards phenolic aldehydes, Colombian oak wood only showed significant differences in the content of coniferyl and sinapic aldehydes in relation to French oak. Finally, the total ellagitannin content of Q. humboldtii was lower than that of European oaks and similar to that of American oak ( Q. alba ). None of the ellagitannins studied presented significant differences compared to Q. alba . Therefore, Q. humboldtii wood was more similar in terms of phenolic composition to Q. alba than to Q. sessiliflora .


INTRODUCTION
Currently, the ageing trends for wines are changing towards the use of newer oak barrels, and thus the demand for oak wood in oenology is increasing. For this reason, in recent years an imbalance between the amount of oak wood available and the number of barrels produced in France has been detected (Cadahía and Fernández de Simón, 2004;Cadahía et al., 2008). This need for new sources of a quality wood supply for cooperage has led to looking into the possibility of utilising new alternatives to European (Q. sessiliflora and Q. robur) and American (Q. alba) oak wood, which are those most commonly used in oenology. More than 800 species exist; however, very few can be used in barrel making, because of their physical, mechanical and gustatory properties (Prida, 2002). Very few oaks fully meet these conditions, and those most used belong to the white oak group. In the last few years, other white oaks have been studied for use in oenology with the aim of comparing their characteristics with those of traditional oaks (Q. robur, Q. sessiliflora and Q. alba), as is the case of Q. pyrenaica and Q. faginea (Canas et al., 2000;Cadahía et al., 2001a,b;Fernández de Simón et al., 2006;Jordão et al., 2007;Gonçalves and Jordão, 2009;Cabrita et al., 2011;Gallego et al., 2012).
An alternative source could be "white oak" (Quercus humboldtii) from Colombia, which is currently used for the ageing of alcoholic drinks such as Rum or Brandy (González and Baleta, 2010). This white oak is a neotropical species found in the three mountain ranges, from 750 m to 3450 m above sea level, in 18 departments of Colombian Andes (Antioquia, Bolívar, Boyacá, Caldas, Caquetá, Cauca, Chocó, Cundinamarca, Huila, Quindío, Risaralda, Nariño, Norte de Santander, Santander, Tolima, Valle del Cauca, Cesar and Córdoba). It is one of the main species in Colombian forests (Avella and Cárdenas, 2010), and its wood is traditionally used for making posts, railroad ties, handles for tools, wooden rollers, charcoal, and firewood (Argoti et al., 2011). Currently, some cooperage companies that provide barrels made of this wood for ageing distilled beverages are proposing this type of wood to the wine industry, ensuring the same quality as the wood traditionally used in oenology. However, there is scarce cooperage information about Q. humboldtii, and previous studies have found that its evaluation could be of interest since its vanillin concentration is similar to that of Q. faginea, and it has a balanced syringaldehyde/vanillin relationship (González et al., 2008;González and Baleta, 2010), which are potentially good properties for wine ageing.
Moreover, Argoti et al. (2011) showed that this oak wood had a good radical-scavenging activity (DPPH and superoxide assays), even higher than commercial rosemary oleoresin (reference material), proposing this species for future investigation in order to identify the compounds responsible for their activity. On the other hand, to our knowledge there is no information characterising other phenolic compounds in Q. humboldtii oak heartwood, so this study can provide deeper knowledge of this wood.
Low molecular weight phenolic compounds (LMWP) are important components of heartwood and contribute to colour, astringency and bitterness in wines (Garde-Cerdán and Ancín-Azpilicueta, 2006). One of the most important extractable compounds besides LMWP are ellagitannins. In oak wood, these compounds are responsible for its hardness and the avoidance of microbiological alterations (Prida, 2002). During wine ageing, they behave as antioxidants thanks to their great ability to consume oxygen (Navarro et al., 2016, García-Estévez et al., 2017. In addition, they accelerate anthocyanin and tannin condensation, which promotes wine flavanol polymerization (flavanoellagitannins) (Cadahía et al., 2008). For this reason, they may contribute to colour stability, astringency, bitterness and wine flavour (Stark et al., 2010;Michel et al., 2011).
The first step to determine whether this new oak source may be of oenological interest is the study of its phenolic composition in green wood. Therefore, the aim of this work was to study the composition of ellagitannins and LMWP by HPLC-DAD of Q. humboldtii (Colombian) green wood, in relation to oak species traditionally used in cooperage, such as Q. sessiliflora (French and Romanian) and Q. alba (American).

Wood samples
To carry out this study four different kinds of oak (origin and/or species) were used: Q. sessiliflora Salisb. (French and Romanian oak), Q. alba L. (American oak) and Q. humboldtii Bonpl.
(Colombian oak). The geographical provenance of the different oaks used were: Vosges for French oak, Monte Semenic for Romanian oak, Pennsylvania for American oak, and North of the Colombian Andes for Colombian oak. French, Romanian, American and Colombian oaks were imported as green wood to the cooperage "Tonelería Nacional Chile LTDA" in Chile. The average annual growths of the wood used were: 1-2 mm in Q. sessiliflora oak, 1.5-2.5 mm in Q. alba and 3.5-4.5 mm in Q. humboldtii. Three lots were taken from each green wood, obtaining three repetitions per wood. Six staves of each lot were taken from the different parts of the pieces from the centre to the headboards. The selected staves were ground, sieved and mixed, taking the sawdust ranging in size from 0.80 to 0.28 mm. The samples were maintained at ambient room conditions (20°C, 55% relative humidity), so all of them had similar moisture content (MC =10%). A total of 12 sawdust samples were analysed (four kinds of wood with three repetitions).

Sample extraction
The sawdust samples (1 g) were extracted in 100 mL of methanol/water (1:1) in darkness for 24 h, following the method described by Fernández de Simón et al. (1996) for LMWP and by Fernández de Simón et al. (1998) for ellagitannins. After removal of the methanol in a rotary evaporator at a temperature below 40 °C, the aqueous solution was extracted with diethyl ether and ethyl acetate and then freeze-dried. The diethyl ether and ethyl acetate extracts were dried, dissolved in methanol/water (1:1) and used for the HPLC quantitative determination of LMWP, and the freeze-dried aqueous extract was used similarly for determining ellagitannins.

HPLC-DAD
analyses of LMWP and ellagitannins LMWP were analysed by the method described by Cadahía et al. (2001a)  LMWP and ellagitannins were detected at 325 nm (with a bandwidth of 150 nm). Chromatographic peaks were identified by comparing their retention time and the UV spectra with those of standards.
Quantitative determinations were carried out by the external standard method. Roburins B-D were expressed as roburin A because they are also dimers. Each green wood were extracted in triplicate (wood samples section) and each one was analysed in duplicate (n=6).

Statistical analysis
LMWP and ellagitannins data were processed using variance analysis (ANOVA) using the SPSS Version 21.0 statistical package for Windows (SPSS, Chicago, USA). Differences between means were compared using Duncan test at 95% probability level. Principal component analysis (PCA), using Stagraphics Centurion 16.2.04, StatPoint Technologies, Inc. (Warrenton, United States), was also carried out to assess the relationship between chemical data and the different oak wood origins. Only two principal components were extracted according to the Kaiser criterion (eigenvalue>1).

Low molecular weight phenolic compounds (LMWP)
Individual concentrations of LMWP are presented in Table I, which also shows the total concentration of phenolic acids and aldehydes from the different kinds of oak wood. Values are mean ± standard deviation (n = 6). Different letters in the same row indicate significant differences between different oaks (p<0.05).
Q. humboldtii green wood showed similar a concentration of total phenolic acids to Q. sessiliflora and Q. alba (Table I). The concentration of total phenolic aldehydes found in Q. humboldtii green wood was similar to that obtained in Q. sessiliflora and Q. alba. In addition, French oak presented lower total phenolic aldehydes than American oak (Table I), this result coincides with that observed by other authors (Cadahía and Fernández de Simón, 2004).
Focusing on the individual composition, Q. humboldtii wood presented the same compounds previously found in the oak wood traditionally used in oenology (Q. sessiliflora, Q. robur and Q. alba) (Miller et al., 1992;Cadahía et al., 2001a,b;Fernández de Simón et al., 2006;Prida and Puech 2006;Zhang et al., 2015), and in other oak wood studied over the last few years for their possible wine ageing potential such as Q. pyrenaica and Q.
The most abundant phenolic acid found in Q. humboldtii, as in the traditional oak wood, was ellagic acid followed by gallic acid. The concentration of ellagic acid was significantly higher in Q. sessiliflora (French and Romanian) green wood than in Q. alba and Q. humboldtii oaks (560.73, 556.16, 352.72 and 286.88 μg/g of wood, respectively). These results are similar to those observed by different authors in green wood (Cadahía and Fernández de Simón, 2004) or in dried or toasted wood (Prida and Puech, 2006;Cabrita et al., 2011) from Q. sessiliflora and Q. alba oaks. In addition, ellagic acid content was similar in Q. humboldtii and Q. alba green wood. However, the content of gallic acid was similar in all studied green oaks (Q. sessiliflora, Q. alba and Q. humboldtii) since no significant differences were observed. The concentration of syringic acid was statistically similar in all the green wood. However, the concentration of vanillic acid was significantly higher in French and American oak than in Romanian and Colombian wood.
The most abundant aldehyde found in all studied oak green wood was syringaldehyde. This result coincides with those observed in French, Romanian and American dry oak wood by Prida and Puech (2006). No significant differences were found in the syringaldehyde concentration among the species or origins of the studied oaks. Vanillin was the second most abundant phenolic aldehyde found in the samples. Its concentration in Q. humboldtii was statistically similar to those in Q. sessiliflora and Q. alba; however, its content was significantly higher in the American oak wood than in the French one (9.27 and 2.79 μg/g of wood respectively). Finally, coniferaldehyde and sinapic aldehyde contents were significantly lower in Q. sessiliflora from France, so the concentrations of these two compounds in Q. humboldtii were similar to those observed in Romanian and American oak wood.
As regards LMWP, the aldehyde content was higher than their respective acids. This result coincides with those observed by other authors in European oaks (Vivas et al., 1998;Fernández de Simón et al., 2006). As expected, the concentrations obtained showed a significant interindividual variability, since many factors have an influence on extractable composition even for the same oak wood origin. Several studies have shown great individual variability (Wilson and White 1986;Guilley et al., 1999). The factors contributing to this variability are the botanical species, the individual tree, the age of the wood and the geographical origin, among others (Guilley et al., 1999).

Ellagitannins
Concerning the green wood tannin composition from Q. sessiliflora (French and Romanian oaks), Q. alba, (American oak) and Q. humboldtii, (Colombian oak), Table II shows the total content and the detailed concentration. The main phenolic components analysed in Q. humboldtii were ellagitannins, with similar results to those found in other oaks traditionally used in oenology. Q. humboldtii green wood presented 1935.30 µg of ellagitannins versus 395.40 µg of low molecular weight phenols per gram of wood. In general, the ellagitannin concentration was lower than those recorded by Mosedale et al. (1996) and Fernández de Simón et al. (2006) but similar to those observed by Masson et al. (1995), Chatonnet andDubourdieu (1998), Fernández de Simón et al. (1999) and Doussot et al. (2000). The ellagitannins analyzed were castalagin, grandinin, roburins A, B, C, D and E and vescalagin. In the case of roburins A and B, they were evaluated together since the separation of chromatographic peaks was not always possible. These eight ellagitannins were isolated and identified in the wood of Q. sessiliflora and Q. Alba, and also found in Q. humboldtii. The quantitative differences are clearly perceptible in Table II, where it can be observed that this oak wood (Q. humboldtii) presented a significantly lower concentration of total ellagitannins than the wood from Q. sessiliflora. However, no significant differences were found between this wood and that of Q. Alba. The fact that Q. sessiliflora contained higher amounts of the individual ellagitannins than Q. alba has been widely reported (Canas et al., 2000;Cadahía et al., 2001b). In addition, the concentration of ellagitannins is affected by a large number of interdependent factors, including the species (Canas et al., 2000;Doussot et al., 2002;Jordão et al., 2007), forest origin (Doussot et al., 2002;Jordão et al., 2007), heartwood age (Matricardi and Waterhouse, 1999), grain coarseness (Jordão et al., 2007) among others. Some authors (Masson et al., 1995;Doussot et al., 2002) reported that species discrimination and especially the species-provenance combination are the main factors determining tannin levels.
Therefore, more studies about the different origin, anatomical aspect of this new oak source and ellagitannin concentration should be carried out.
The concentration of ellagitannins showed a strong interindividual variability, as occurred with the content of LMWP. This has also been described by other authors (Masson et al., 2000) and explained by the same influential factors mentioned above. The most abundant ellagitannins in all the studied wood were castalagin, followed by vescalagin, and roburin D. Grandinin was the only one of all the studied ellagitannins without any significant differences among the different species and/or geographical origins. Regarding the other ellagitannins (Cast, Vesc, Rob A+ B, Rob C, Rob D and Rob E), Q. humboldtii wood presented the lowest concentration of all of them. Moreover, Q. alba wood did not present significant differences with respect to the Colombian oak in any of the identified ellagitannins (Table II). Values are mean ± standard deviation (n = 6). Different letters in the same row indicate significant differences between different oaks (p<0.05).
Castalagin concentration was significantly higher in Q. sessiliflora green wood (French and Romanian oak) than in Q. alba and Q. humboldtii wood (3892.85, 4059.9, 1189.63 and 636.37 μg/g of wood, respectively, Table II). However, Q. humboldtii had a significantly lower vescalagin concentration than French green wood and its content was similar to that of Romanian and American oak. French oak wood showed a higher concentration of roburins A, B and C than the new wood (Colombian oak), without any significant difference to the rest of the studied wood. In addition, the contents of roburins D and E were higher in Romanian oak wood than in Q. humboldtii and Q. alba.

Principal component analysis (PCA)
Principal component analysis (PCA) was carried out to evaluate the discriminant power of these 16 compounds for the four kinds of oak wood studied.

CONCLUSIONS
If the results obtained so far are considered, it can be concluded that the phenolic composition in green wood from Q. humboldtii is qualitatively similar to that of Q. sessiliflora and Q. alba, only showing quantitative differences in some compounds. Q. humboldtii green wood presented statistically similar concentrations of total phenolic acids and aldehydes as Q. sessiliflora and Q. alba. In addition, this oak had a statistically similar concentration of total ellagitannins to Q. alba, although a significantly lower one than that of Q. sessiliflora. Colombian oak showed that the most abundant phenolic acids, aldehydes and ellagitannins were the same as in traditional oak wood (ellagic acid, gallic acid, syringaldehyde and castalagin), and without any significant differences in any of them with respect to American oak. Therefore, its wood phenolic composition was closer to that of American oak than to that of French oak. Finally, this wood seems to have an adequate composition as a new oak wood source for wine ageing, although it is necessary to continue the study of this wood species to confirm its potentialities for oenology.