Fifty-eight cheeses were collected from supermarkets of 5 different Chilean cities (Santiago, Rancagua, Melipilla, Viña del Mar and Puerto Natales) including 34 sheep cheeses, 7 from goat's milk, 11 from cow's milk, 4 from a mixture of sheep, goat and cow's milk and 2 from a mixture of sheep and cow's milk. The sheep cheeses were produced in Chile (23 cheeses) and imported from France (2 cheeses) and Spain (9 cheeses). After collection, the cheeses were cut in four sections at which time two cores of each cheese were obtained and stored in re-sealable bags at –80 °C for one month until further analysis.

Two cores of each cheese were used for the FA analysis. The lipids from the cheese were extracted according to a chloroform/methanol (2:1, v/v) by Folch et al. (1957) method and methylated with the modifications of Sukhija and Palmquist (1988) using hexane as an organic solvent instead of benzene. Analyses of fatty acid methyl esters (FAME) were performed using a gas chromatograph (GC; Shimadzu Scientific Instruments AOC-20s, Columbia, MD, USA) equipped with a flame ionization detector, an autoinjector and an Rtx column (30 m × 0.32 mm × 0.20 µm column). The GC conditions were similar to those reported by Vargas-Bello-Pérez et al. (2013b). FA peaks were identified by using two FAME standards (GLC 60; Nu-Check-Prep, Elysan, MN, USA), and the Food Industry 37 FAME mix, 35077 Restek Co, Bellefonte, PA, USA). C18:1 trans isomers were detected by comparing retention times as reported by Ledoux et al. (2000).

The samples were assigned to four groups: cow, sheep, goat and mixtures. The samples were analyzed using ANOVA and multiple comparisons of means. In order to determine which FA was responsible for the differentiation between cheeses that were made from different animal species and geographical origin a multivariate analysis was carried out using a correlation matrix between individual FA to discard those that showed very high correlations (r = >0.9) as well as those without correlations. Bartlett's test of sphericity was applied to examine the hypothesis that the variables were uncorrelated in the population and the Kaiser-Meyer-Olkin index was used to measure sampling adequacy. Hierarchical clustering was then performed and a discriminant analysis was used to verify the extent to which samples were correctly assigned to the clusters identified in the previous analysis. To check for possible differences among cheeses associated with country of origin (Chile, Spain and France) an individual analysis of variance and multiple means comparison were performed. The SPSS statistical software for Windows was used (version 15.0.0; SPSS Inc., Chicago IL, USA).

From the cheese samples analyzed, only C17:0 was not different among animal species (Table 1). Regardless of cheese type, C10:0, C14:0, C16:0, C18:0, C18:1 t11 and C18:1c9 represented around 78% of the total FA, and these results agree with those reported previously (Partidário et al., 2008; Walther et al., 2008). As reported by Park et al. (2007), sheep and goat cheese contained higher contents of caproic (C6:0), caprylic (C8:0), capric (C10:0) and lauric (C12:0) acids than cow cheese. Fatty acids which are formed by 4 to 14 carbons arise chiefly from de novo synthesis whereas FAs with more than 16 carbons are derived from the uptake of circulating lipids. Ruminants utilize acetate produced by the ruminal fermentation of carbohydrates (contained in dietary fiber) as the major carbon source (Bauman and Griinari, 2003). Fatty acids with less than 12 carbons are associated with the typical flavors of small ruminant cheese and can also be used to detect mixtures of milk from different species (Park et al., 2007).

Compared with cow and goat, sheep cheese had higher contents of C18:1 trans isomers. In ruminant fat, vaccenic acid (VA; C18:1 t11) is the most common C18:1 trans isomer accounting for 60 to 80% of the total trans FA (Vargas-Bello-Pérez and Garnsworthy, 2013). VA can be converted to C18:2 c9, t11 (rumenic acid; RA) through the action of stearoyl coenzyme A desaturase and it has been estimated that 20% of VA can be converted to RA in humans. The RA has been identified as anti-carcinogenic, anti-atherosclerotic, antioxidant, and immunomodulator (Young et al., 2013). The C18:1 trans isomer contents in sheep cheese are similar to those reported in Italian (Nudda et al., 2005) and Spanish (Luna et al., 2005) cheeses. This may be due to the feeding practices.It is possible that those cheeses came from farms where the sheep were fed with fresh pastures or supplemented with oilseeds. When pasture inclusion increases in the total diet, linear increases in C18:3n3, C18:1 t11 and C18:2 c9t11 and decreases in C10:0–C16:0 are observed (Chilliard et al., 2007). On the other hand, oilseeds are rich sources of polyunsaturated fatty acids (PUFA), which, after being biohydrogenated in the rumen, can increase the concentrations of health promoting FA such as linolenic (C18:3n3) and rumenic acids (Zhang et al., 2006).

Interestingly, the C18:1 trans isomers and C18:3n3 contents were higher in Chilean sheep cheese than in the French and Spanish cheese (Table 2). As mentioned before, those FA can be improved when the ewe diet is based mostly on pastures. In Chile, sheep production is based largely on grazing systems, however, in Spain for example, sheep cheese is produced throughout the year leading to an intensification of reproductive management, and a reduction in grazing periods (Riedel et al., 2007). This type of system results in a substitution of on-farm natural resources (forages and pastures) with external inputs (conserved forages and concentrates) (Vera et al., 2009). No differences were observed between French and Spanish sheep cheese so it is possible that these countries have similar conditions among their production systems including feeding, breeds and level of intensification of farm management (Riedel et al., 2007).

Nutritional recommendations are based on different ratios such as PUFA n6/ PUFA n3 and PUFA/SFA; these values are used to evaluate the nutritional value of fat for human consumption. The values for PUFA/SFA found in our cheese samples were below the recommended amount (above 0.45) for the human diet whereas only the values for sheep cheese were below the recommended level (less than 4.0) for the PUFA n6/PUFA n3 ratio (British Department of Health, 1994). Compared to cow and goat, sheep cheese had a lower thrombogenic index and intermediate concentrations of SFA, MUFA and PUFA (Table 1). Saturated FA are known to increase the risk of coronary heart disease (CHD; Williams et al., 2000), on the contrary, milk MUFA and PUFA (e.g., C18:2 c9, t11) can reduce the risk of CHD and prevent some types of cancer (Mohammadzadeh et al., 2013). However, there is a lack of scientific evidence that other SFA such as C4:0-C10:0 (short- and medium- chain FA; SMFA) have an effect on blood cholesterol and CHD risk (Astrup et al., 2011). In this study, compared to cow and sheep, goat cheese had higher contents of SMFA.

From the human health perspective, in this study, Chilean sheep cheese could be considered healthier than those imported from Europe; this is supported by the higher PUFA n3 and lower PUFA n6 contents (consequently lower PUFA n6/ PUFA n3 ratio) found in the Chilean sheep cheese (Table 2). The PUFA n3 play an important role in the prevention and control of coronary artery disease, hypertension, diabetes, arthritis, autoimmune disorders and cancer (Simopoulos, 2000).

The correlation matrix of FA discarded C4:0, C6:0, C18:0 and C20:1n9 (c11). The principal component analysis (PCA) carried out on the 16 remaining FA resulted in 4 principal components (PCs) which explained 0.84 of the overall variance in the data. PC1 and PC2 explained 0.34 and 0.28 of the variation in FA contents (Table 3). The KMO index was 0.62, whereas the Bartlett's test of sphericity showed the suitability (P<0.001) of those variables selected for analysis.

Three groups were found after PCA was carried out in order to obtain data clusters related to the fatty acid profiles of the cheeses. Group I was formed by 26 samples from which 23 were Chilean sheep cheeses, 2 were Chilean mixture cheeses (90% sheep and 10% cow milk) and 1 Chilean goat cheese. Group II was composed of 13 samples from which 11 were Chilean cow cheeses and 2 were Spanish mixture cheeses (proportion of milks was not found in the label). Finally, Group III was formed by 19 samples from which 9 were Spanish sheep cheeses, 2 were French sheep cheeses, 2 were Spanish mixture cheeses and 6 were Chilean goat cheeses.

PC1 consisted of the typical FA that can be found in the milk of small ruminants (C8:0 and C10:0) (Partidário et al., 2008) whereas C14:1c9 and C16:1c9 are usually detected in the milk and cheeses of cows (O'Donnell et al., 2010). Figure 1 shows a clear difference among animal species particularly between Groups I and II. PC2 comprised C16:0, C18:1t, C18:3n3 and C18:2n6, and therefore, the cheese in this PC may have a high MUFA/SFA ratio. Group I (mainly Chilean cheeses) had higher proportions of these FA.

The different data clusters obtained in this study showed that some FA vary depending on animal species, however, there are some confounding factors related to the type of feed, production system and breed of animals (Carta et al., 2008). For example, in Group III, where most of the European cheeses were grouped, most of the production systems may be intensive and their feeding management relies on high-concentrate diets (Sanz Sampelayo et al., 2007). These particular characteristics of production systems, animals, feedstuffs, etc., are closely related to the organoleptic characteristics of cheese that lead to the creation of the Protected Denomination of Origin (PDO) or Protected Geographical Identification (PGI) regimes (Martínez et al., 2011) and therefore become useful tools for creating strategies aiming at improving the nutritional characteristics of sheep cheese in Chile.