The autochthonous microbiota of traditionally made cheeses elaborated from raw milk gives them their particular characteristics, and thus constitute an excellent source of new strains of microbes with phenotypic and genotypic diversity, which could be of technological interest (Rademaker et al., 2007). This microbial diversity is responsible for the production of compounds giving rise to flavour during ripening (Coolbear et al., 2008). The activity of the starter which is made up of lactic acid bacteria and could include strains producing aroma, could be reinforced by adding a secondary microbiota that will contribute to the ripening process of the cheeses. Non-starter yeast and molds (NSYM) population in cheese is very diverse and its role in the ripening is often underestimated (Lavoie et al., 2012).

The importance of using the filamentous yeast-like fungus, Geotrichum candidum as a co-starter in dairy products is well known because of its various properties related to the development of flavour and texture in semi-hard cheeses (Gaborit et al., 2001). In fact, in recent years, there has been increasing interest in the use of this microorganism, especially in the manufacturing of cheeses from pasteurized milk in order to reproduce the characteristics of cheese made from raw milk (Boutrou and Guéguen, 2005). G. candidum was the dominant yeast species in the first and second week of Armada ripening (Fresno et al., 1996; Tornadijo et al., 1998). Then, it could have important contributions to the flavour and texture of this one or similar cheeses. Nonetheless, G. candidum strains differ in their biochemical capacity to produce aromatic compounds in dairy products (Spinnler et al., 2001), which require a selection process on the basis of their technological suitability. In an earlier study, several properties of technological relevance were studied in G. candidum and the strains with greater technological capacities were characterized at a molecular level (Sacristán et al., 2012, 2013).

The aim of this study was to investigate the contribution of a natural starter composed of strains of Lactococcus and G. candidum as co-starter on the chemical, microbiological and sensorial characteristics of a goat’s milk cheese. The G. candidum strain that provides the best sensorial properties was selected to be included as co-starter in the Armada cheese manufacture.  

Cheese physico-chemical analysis

The contents of dry matter (DM), fat, protein, lactose and salt were determined in cheese according to standard methods (IDF, 2004, 2008, 2001, 1967; AOAC, 1990a).

The pH was determined potentiometrically from samples homogenized with 100 mL of distilled water warmed at 45-50°C so as to disperse the fat and then cooled down to 20°C ± 2°C. Water activity (a_w) was measured using an Aqua Lab CX-2 water activity meter (Decagon, WA, USA). The titratable acidity of the cheese and the fat acidity index (FAI) were determined according to standard methods (AOAC, 1990b; IDF, 1969).

Extraction and quantification of pH 4.4 soluble nitrogen (pH4.4-SN), trichloroacetic acid soluble nitrogen (TCA12%-SN) and phosphotungstic acid soluble nitrogen (PTA5%-SN) fractions were carried out using the method described by Bütikofer et al. (1993). All analyses were carried out in duplicate.

Sensory analysis

Cheeses were analysed after 15, 30 and 60 days of ripening by a panel of 20 trained tasters following the standard recommendations (ISO, 2005; 2012). Several parameters, related to appearance (mouldy rind, yellowish rind, white paste, mouldy spotted paste and cracked paste), taste (bitter taste, sweet taste, acid taste, salty taste, metallic taste, spicy taste, astringency, aftertaste and persistence), odour (fresh milky odour, mouldy odour, rennet odour, buttery odour and farmyard odour) and texture (hardness, buttery texture, grainy texture, crumbly texture and sticky texture) were evaluated on a 7-point intensity scale from 1 (dislike extremely) to 7 (like extremely), with 4 being an “acceptable” value. Finally, the cheeses were scored from 1 to 10 on the the basis of overall sensory impression.

Statistical analysis

In order to investigate possible significant differences among batches, the ANOVA/MANOVA analysis using Fisher's least significant difference (LSD) test  (Statistica  8.0  computer  program: Statsoft, Tulsa, Oklahoma, U.S.A.) was carried out with the confidence intervals set at 95% level and other higher levels (99 and 99.9%). The correlation between changes in the physico-chemical parameters and the log counts of the major microbial groups was also studied by the Pearson’s correlation coefficient, so as to discover what influence the ripening process had on microbial development.

The use of Lactococcus lactis subsp. lactis (TAUL 1292) strain selected for its acidifying and proteolytic activity (Herreros et al., 2003), caused a very marked fall in pH at early stages. The slight increase in pH observed after sixty days of ripening could have been due to the capacity of G. candidum to metabolize lactic acid, which would favour the implantation of another type of microbiota (Cosentino et al., 2001; Fadda et al., 2004). Titratable acidity (TA), expressed as g of lactic acid 100 g^-1 of DM, decreased during the ripening (p < 0.001), with no significant differences between batches. With regard to changes in lactose, significant differences (p < 0.01) were noted between the different batches. The lactose content in the curds and at the start of the ripening was similar to the content observed in Armada cheeses manufactured from pasteurized milk and the  LAB  strains (Herreros, 2010). The lactose content dropped sharply up to the fifteenth day of ripening. This decline is attributed to the rapid development of the lactic microbiota included in the starter, favoured by the low salt to moisture ratio (S/M) in two-day-old cheese. These results agree with those obtained by Herreros (2010) in batches made form pasteurized goat’s milk and the LAB strains.

The FAI showed no significant differences between batches. However, there were significant differences (p < 0.001) as a function of ripening time, with much higher values after fifteen days. The presence of cracks in the cheeses, especially after fifteen days of ripening, favoured the growth of fungi and hence that of lipolysis. After the fifteenth day of ripening, a fall in the FAI was observed until the end of ripening. The lipoprotein lipase in  milk  is  deactivated by pasteurization and the addition of a starter culture triggers off a marked decrease in pH at the beginning of ripening, which could negatively affect the lipases present.

The protein content decreased slightly during the first two weeks of ripening, and then remained stable until the end of the process. Significant differences were then observed with regard to the ripening time (p < 0.01), but in contrast there were no significant differences between batches (Table 2). In order to determine the degree of proteolysis in cheese, the changes in different nitrogen fractions were evaluated (Table 3). The nitrogen fractions which showed significant differences between batches were also illustrated in Figure 2.

The formation of pH 4.4 soluble nitrogen during the ripening of the cheese is mainly due to the rennet action on caseins and, to a lesser extent, to the plasmin and acid protease action in milk. No significant differences were found as a function of the batch or ripening time in the batches of Armada cheese elaborated using the various G. candidum strains. The pH 4.4-SN values remained stable at about 12.4%, which indicates that only slight proteolysis took place. These results were very similar to those reported by Herreros (2010) for Armada cheeses elaborated from pasteurized milk and the LAB strains. With regard to the trichloroacetic acid soluble nitrogen, significant differences were noted during ripening (p < 0.01). The TCA12%-SN increased slightly up to thirty days, but then stabilized. Significant differences (p < 0.05) were also observed between the various batches. The TCA12%-SN, measures the depth of proteolysis and it is nevertheless an indicator of proteolytic activity due to the microbiota present in cheese. In the case of phosphotungstic acid soluble nitrogen (PTA5%-SN), significant differences (p < 0.001) were found with regard to both ripening time and the various manufactured batches. In all cases, the average values detected at each sampling point were very low, which confirms the presence of low-level aminopeptidase activity. These results were lower than those found by Herreros (2010) in the same type of cheese elaborated without the addition of the relevant G. candidum strains. Polypeptide nitrogen (Polypeptide N), calculated from the difference between pH4.4-SN and TCA12%-SN, decreased over the ripening process, but without any significant differences being noted either in ripening time or in batches. As for peptide nitrogen (peptide N), calculated from the difference between TCA12%-SN and PTA5%-SN, no significant differences were seen either with regard to ripening time or between the batches (Table 3).

Changes in the main microbial group counts

Changes in the microbial counts during the ripening process of the various batches of Armada cheese are shown in Table 4. The microbial counts which showed significant differences between batches are also illustrated in Figure 3.

Most of the microbial groups had an increase of approximately two logarithmic units in two-day-old cheese due to the physical retention of microorganisms in curds, their multiplication during coagulation and drainage and the delay of salting process in this cheese. The highest counts of mesophilic aerobic microorganisms (PCAm) were recorded after two days of ripening, but thereafter there was a gradual decline in counts until at the end of ripening. The concentration of NaCl dissolved in the moisture in cheese, which increased during ripening, has an inhibitory effect on microorganisms (Beresford et al., 2001). All the batches showed a very similar evolution and no significant differences were observed with regard to the batch in the counts of mesophilic aerobic bacteria. Nevertheless, there were significant differences with regard to ripening time (p < 0.001).  The  counts  recorded  were  similar  to  those  in Armada cheese elaborated from pasteurized goat’s milk using the same lactic acid bacteria strains as starter culture (Herreros, 2010).

The counts observed on MRS and M17 agar followed a pattern practically identical to that observed on PCAm. On both culture media, significant differences were observed with regard to both ripening time (p < 0.001) and batch (p < 0.05) up to sixty days of ripening. Counts were also similar to those reported for other goat and cow’s cheeses (Arenas et al., 2004; González et al., 2003; Herreros et al., 2007). Counts on M17 agar reached their highest values between two and fifteen days of ripening because Lactococcus break down lactose and their counts increase rapidly (Williams et al., 2000). Then, slowly dropped until the end of ripening at which point significant differences were noticeable between batches. A significant positive correlation (p < 0.01) was found in the Lactococcus on M17 counts with regard to the a_w levels (r = 0.88) and moisture (r = 0.86) (Table 5).

Lactobacilli count on ROGOSA agar increased progressively as ripening proceeded (p < 0.001), with significant differences (p < 0.05) being observed between batches. The slower lactobacilli metabolism and its greater capacity to adapt to adverse conditions (acidity, low values for a_w or high NaCl concentrations) related to other LAB could contribute to its increasing predominance as the number of days of ripening increases. In fact, there was a significant negative correlation (p < 0.01) between the developments in counts on ROGOSA agar and titratable acidity (r = -0.52), a_w (r = -0.45) and moisture (r = ‑0.45). On the other hand, a positive correlation (p < 0.05) was found between the ROGOSA agar counts and the S/M ratio (r = 0.39) (Table 5).

Significant differences were detected in VRBGA medium with regard to the batch (p < 0.01) and ripening time (p < 0.01). After fifteen days of ripening, the counts for Enterobacteriaceae decreased up to undetectable count, and eventually they completely disappeared owing to the unfavourable growth conditions which gradually prevailed in the cheese as this process develops (Buffa et al., 2001). The detection of Enterobacteriaceae in the cheeses of one of the batches after thirty days of ripening may have been due to later contamination.

Counts for moulds and yeasts on OGYEA medium increased up to fifteen days of ripening, when they reached their highest levels. They then decreased until the end of ripening. The evolution of this microbial group was similar in all batches manufactured. However, there were significant differences (p < 0.001) with regard to ripening time. G. candidum was the predominant fungal species during the manufacturing and ripening of the cheeses elaborated from the lactic starter culture and the G. candidum co-culture. Thus, OGYEA counts constitute an indicator of changes undergone by G. candidum during the ripening. The highest counts were reached after fifteen days of ripening, without apparent inhibition of growth. Interactions of G. candidum and starter culture were reported by other authors (Šípková et al., 2015). From here on, the counts fell more or less steeply depending on the batch, until after sixty days when they reached values similar to those at the start of the ripening.

Lipolytic activity by LAB is generally very limited. Their lipases act primarily on mono- and di-glycerides previously formed by indigenous lipases of the milk, and their capacity to act on triglycerides is very slight (El Soda et  al., 1986).  Positive correlation between the values for the FAI and the counts obtained for OGYEA shows the involvement of fungal population in lipolysis (Table 5).

Sensory evaluation

The overall sensory evaluation of the Armada cheeses elaborated by adding an autochthonous starter culture and co-starter culture of G. candidum is shown in Table 6. These cheeses developed a notable flavour characteristic of goat cheese, which gives individuality and quality to this type of cheese (Gaborit et al., 2001). In fact, flavour is one of the main sensory attributes of cheese’s quality (Zabaleta et al., 2015). The evaluation of the texture and taste parameters in cheeses after sixty days of ripening is shown in Figure 4.

With regard to the overall sensory evaluation of the manufactured cheeses, no significant differences could be seen throughout the ripening period, but there were significant differences (p < 0.05) with regard to batches. Batches I and IV received the best scores after sixty and thirty days of ripening, respectively.

Because  of   the   particular   techniques   used  in  the manufacturing of these cheeses, in which salting is carried out after two days, kneading stage was required, and a dry grainy texture is a constant in cheeses after fifteen and thirty days of ripening. This was also seen in the cheeses elaborated with pasteurized milk using commercial and autochthonous starter culture in the test carried out by Herreros (2010). Generally, as the ripening process went on, hardness increased (p < 0.001) and soft buttery texture decreased (p < 0.001), as did the stickiness of the cheese paste in the mouth (p < 0.05). Batch I presented a hardness profile that differed from the others, as it was the batch with the softest texture after fifteen and thirty days as well as had the greatest increase in hardness. Stickiness decreased in all the batches, although somewhat irregularly, depending on the batch. 

The fresh milky odour characteristic of these cheeses declined slightly during the ripening period (p < 0.01), with significant differences observed among batches (p < 0.05). Another pleasant attribute of these cheeses was a buttery odour, which increased slightly over the ripening period in all the cheeses. A rennet odour and a farmyard odour   were two negative attributes of some of these cheeses, with significant differences (p < 0.05) arising as ripening progressed. In general, the two unpleasant parameters were observed more in thirty-day-old cheeses, then decreased after sixty days of ripening.

The cheeses were characterized by a fresh acid taste which decreased somewhat during the ripening period (p < 0.05). The desirable odour and aftertaste of goat which developed was a parameter for quality in these cheeses, associated with G. candidum. In fact, these cheeses received higher scores than those elaborated from pasteurized milk without the addition of G. candidum culture. Astringency was also affected by ripening time (p < 0.01), increasing from the fifteenth to the sixtieth day. The spicy taste characteristic of these cheeses also increased during ripening (p < 0.05). A further parameter that was affected was the persistency on the palate, which notably increased after sixty days of ripening period (p < 0.001), which indicates a greater potency of the mixture of tastes and odours at the end of the ripening process.

With regard to the assessment of batches as a function of ripening time, at fifteen days of ripening the taste was equal, milky and fresh in all batches, if not very intense. In some cheeses it was possible to note a certain mouldy taste, and in those from batch I a typical goat odour. In general, after thirty days the cheeses in batches II and III showed a drier and grainier texture and an unequal and metallic taste, with an unpleasant aftertaste and marked bitterness. In some cases, odour defects were observed, such as a mouldy or rennet odour. At sixty days of ripening, cheeses of the batch III presented a more acceptable texture (stronger buttery texture and less grainy) and a strong buttery odour, but it did not reach high   overall   scores.   Although,  cheeses  in   batch   IV achieved the best scores at thirty days of ripening, then they worsened in texture, becoming grainier and crumblier, as well as having a taste that was excessively acid and astringent. With regard to cheeses of the batch I, they obtained high scores at thirty days of ripening and were even the best evaluated at the end of ripening because of their odour, fresh balanced pleasant taste and creamy smooth texture.

The use of an autochthonous starter culture constituted lactic acid bacteria strains and G. candidum as co-culture affected the evolution of chemical and physico-chemical parameters in the case of Armada cheese, in particular lactose, pH and titratable acidity. The impact that this starter culture had on the sensorial characteristics of the Armada cheese was observed from the fifteenth day of ripening. Cheeses elaborated with the autochthonous lactic starter culture and G. candidum co-culture had a marked odour characteristic of goat cheese and a stronger buttery and creamy texture, as they did not undergo excessive drying-out and did not develop any residual rancid taste. Cheeses in batch I, elaborated with the Ge‑1886 strain (with high lipolytic activity and low proteolytic activity) were given the best assessments for presenting a wider and more intense range of pleasant odours, and a soft buttery and creamy texture.

The authors have not declared any conflict of interests.

The authors acknowledge the financial support from Junta de Castilla y León (Spain) (Project LE 010A06) for this work.