Relationships Between Flavour and Chemical Composition Of

Relationships between flavour and chemical composition of Abondance cheese derived from different types of pastures Christophe Bugaud, S Buchin, Agnès Hauwuy, Jean-Baptiste Coulon

To cite this version:

Christophe Bugaud, S Buchin, Agnès Hauwuy, Jean-Baptiste Coulon. Relationships between flavour and chemical composition of Abondance cheese derived from different types of pastures. Le Lait, INRA Editions, 2001, 81 (6), pp.757-773. 10.1051/lait:2001162. hal-00895376

HAL Id: hal-00895376 https://hal.archives-ouvertes.fr/hal-00895376 Submitted on 1 Jan 2001

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Lait 81 (2001) 757-773 757 © INRA, EDP Sciences, 2001 Original article

Relationships between flavour and chemical composition of Abondance cheese derived from different types of pastures

Christophe BUGAUDa, Solange BUCHINa*, Agnès HAUWUYb, Jean-Baptiste COULONc

a Technologie et Analyses Laitières, INRA, BP 89, 39801 Poligny Cedex, France b Groupement d’Intérêt Scientifique des Alpes du Nord, 11 rue Métropole, 73000 Chambéry, France c Unité de Recherches sur les Herbivores, INRA, Theix, 63122 Saint-Genès-Champanelle, France

(Received 12 February 2001; accepted 24 April 2001)

Abstract — The relationships between the flavour and composition of half-cooked pressed Abondance cheese (Haute-Savoie, France) were studied under natural cheese production conditions on the basis of different pastures exploited by 3 producers. Cheeses manufactured from milk produced from mountain pastures (M, n = 5, 1500–1850 m) were deemed to be more “fruity”, “animal”, “boiled milk” and “hazelnut” and less pungent and “propionic acid” than cheeses made from milk produced from valley pastures (V,n = 5, 850–1100 m). It was possible to partly attribute these differences in flavour to the presence of protein-based volatile compounds in the cheeses. The V cheeses had a greater variety of flavours than the M cheeses. Cheeses from gramineae-rich pastures had the most intense “cooked cabbage” odours, related to the greater amounts of sulphur compounds in these cheeses. Terpenes, which are more abundant in cheeses produced from dicotyledon-rich pastures, did not contribute directly to cheese aroma. The differences in flavour between the cheeses manufactured by the 3 producers were of the same magnitude as those observed between different pastures used by a same producer. The origin of the volatile compounds in the cheeses – whether of microbial origin or from the feed – was discussed. cheese / flavour / volatile compound / proteolysis / pasture

Résumé — Relations entre la flaveur et la composition chimique de fromages d’Abondance fa- briqués à partir de laits produits sur différents types de pâturages. Les relations entre la flaveur et la composition des fromages à pâte pressée demi-cuite de type Abondance (Haute-Savoie, France), ont été étudiées dans des conditions naturelles de production du fromage à partir de pâturages diffé- rents exploités par 3 producteurs. Les fromages fabriqués à partir de laits produits sur des pâturages de montagne (M, n = 5, 1500–1850 m) ont été jugés plus « fruité », « animal », « lait cuit » et « noisette »

* Correspondence and reprints Tel.: (33) 3 84 73 63 03; fax: (33) 3 84 37 37 81; e-mail: [email protected] 758 Ch. Bugaud et al. et moins piquant et « propionique » que les fromages issus de laits produits sur des pâturages de vallée (V, n = 5, 850–1100 m). Ces différences de flaveur ont pu être attribuées en partie à la présence dans les fromages de composés volatils issus du catabolisme des acides aminés. Les fromages V présentaient une plus grande diversité de flaveur que les fromages M. Les fromages issus de pâturages riches en graminées étaient les plus intenses en odeur « chou cuit », flaveur liée aux quantités plus importantes de composés soufrés dans ces fromages. Les terpènes, plus abondants dans les fromages issus de pâturages riches en di- cotylédones, n’ont pas contribué directement à l’arôme des fromages. Les différences de flaveur des fromages entre les 3 producteurs ont été du même ordre de grandeur que celles observées chez un même producteur entre différents pâturages. Les origines microbienne et alimentaire des composés volatils des fromages ont été discutées. fromage / flaveur / composé volatil / protéolyse / pâturage

1. INTRODUCTION forage, such as terpenes, may be found later in the dairy products [40] and thereby con- Although it is empirically accepted that tribute to cheese aroma [8, 9, 22, 24, 39]. the type of pasture influences the sensory Buchin et al. [13] also considered the possi- properties of cheese [27] and its flavour in bility that there could be mechanisms in- particular, few experimental studies on the volving milk proteolytic enzymes such as topic have yet been undertaken. It is never- plasmin. theless particularly important to identify These initial studies show that the type the relationship between forage type and of pasture does influence cheese flavour, cheese properties in the case of Protected but the relationships established between Designation of Origin (PDO) cheeses, the raw material (grass) and the final prod- since it constitutes one of the bases for their uct (cheese) remain incomplete. relationship with the “terroir” [28]. The aim of our study was to explore the Bosset et al. [9], Buchin et al. [13] and relationships between the composition and Verdier-Metz et al. [38], have demonstrated flavour of cheeses produced from different significant differences in cheese flavour de- types of pasture. This study is part of a pro- pending on the altitude of the pastures and ject aimed at understanding the influence of botanical composition of the forage con- pastures on the sensory properties of sumed by the animals, whether in the form cheese, of which the first results have al- of pasture or preserved forage. Neverthe- ready been presented [14–16]. The study less, many gaps remain in the existing was undertaken under real conditions of knowledge of the mechanisms involved in Abondance cheese production (Haute- the influence of forage type on cheese fla- Savoie, France), in order to benefit from a vour. The above-mentioned authors ob- wide variety of pasture types and be able to served differences in volatile compound monitor the entire production chain from composition in cheeses – whether these grass to cheese via milk. compounds were of microbial origin or originated from the feed – that may be linked to differences in flavour. On the one 2. MATERIALS AND METHODS hand, there may be a specific type of microflora in each production zone, as has 2.1. Experimental conditions been demonstrated [18, 21], which would lead to significant variations in the sensory The experiment was undertaken during quality of the cheese [7, 12, 20]. On the the spring and summer of 1998 in 3 farms other hand, certain compounds present in (X, Y, Z) producing Abondance cheese, Cheese flavour and pastures 759

producer Z producer X producer Y 38Montbeliardes 37 Abondances 28 Abondances

identical Hay (H) for each herd

170 169 128 15 April HZ 16 * HX 16 HY 9 ** Valley pastures (V) and mountain pastures (M)

15 May 199 VZ1 23 160 VX1 17 30 May 161 VX2 17

20 June 174 MX1 15

15 July 214 MX2184 MY1 203 VZ2 17 12 7

1 August 194 210 MX3 13 MY2 7

5 September 177 VZ3 22

Figure 1. Experimental design. a a . –1 *HZb where was the average stage of lactation (d) and b the milk yield (L d ). ** The Y studied milk provided exclusively from the morning milking. where each producer manufactured cheese scribed by Bugaud et al. [15]. The manufac- using the milk of his own herd. Ten pasture and ripening of Abondance cheese tures, described by Bugaud et al. [14], be- have been described by Bugaud et al. [16]. longing to these producers, were used as the The experimental design is presented in basis of this study: 5 valley pastures (V) Figure 1. located at altitudes of between 850 and 1100 m and 5 mountain pastures (M) located at altitudes of between 1550 and 2.2. Analyses 1850 m. For each pasture, and during the last 3 d of pasture, the cheesemaking pro- All the analyses were carried out at cess was monitored and a cheese was sam- the end of the cheese-ripening period pled each day for analysis. (24 weeks). As a preliminary stage, the 3 systems of 2.2.1. Sensory analysis production (herd + cheesemaking process), were compared by producing cheeses using Two panels of trained judges assessed milk derived from hay-based feeding (H), the flavour of the ripened cheeses. Panel 1 which was identical for each of the produc- comprised 14 judges trained to assess the ers. The herds and milks have been de- intensity of 9 descriptors on a scale of 1 to 7, 760 Ch. Bugaud et al. using commercial Abondance cheese. The 2.2.2.2. Volatile compounds descriptors were the following: intensity of taste and odour, “fruity” and “hazelnut” The other volatile compounds were ex- aroma, “sweet”, “salty”, “pungent” and tracted using a dynamic head-space tech- “bitter” taste, and “fruity” odour. The nique (Purge and Trap LSC 3000; Tekmar, descriptor “fruity” designated aromatic Cincinnati, OH), separated by gas phase richness with positive connotations (a term chromatography (chromatograph 6890, used by cheese production professionals) Hewlett Packard Agilent Technologies, Les rather than fruit aroma as such. Ulis, France) and identified by mass spec- trometry (MSD 5973, Hewlett Packard Panel 2 comprised 20 judges trained to Agilent Technologies). The volatile com- recognise the presence of 52 descriptors of pounds were analysed in two steps: (i) all odour and aroma for all types of cheese. volatile compounds except sesquiterpenes, These descriptors were marked on a pres- in all the cheeses, (ii) sesquiterpenes only, ence/absence scale. Those descriptors in one cheese per pasture. The correspond- whose detection frequencies (number of ing sample treatments were the following: “presence” responses out of the total num- (i) Ten grams of cheese were cut into 7 mm ber of responses) were below 10% on aver- cubes and placed in a cylindrical glass con- age for all the pastures were not kept for tainer (140 mm high and 30 mm in diame- statistical data processing. ter) connected to the “Purge and Trap” system. Volatile compounds were trapped 2.2.2. Cheese composition for 15 min at 25 oC with a helium flow of 2.2.2.1. Volatile fatty acids 40 mL.min–1. (ii) The cheese was grated, diluted at 10% (p/v) in deodorised milliQ Volatile fatty acids were solvent-ex- water (boiled for 15 min), and then ultra tracted under reflux in a Jalade extractor turrax-homogenised for 1 min. Ten milli- and then saponified as described by Ardö litres of the solution were introduced into a and Polychroniadou [1]. They were then U-shaped 25 mL glass vial (Sparger 5182- analysed by gas chromatography after re- 0849 Hewlett Packard Agilent Technol- lease by soap acidification. The ogies) connected to the “Purge and Trap” chromatograph (GC 8000, Carlo Erba, system. Sesquiterpenes were trapped for Thermoquest, Les Ulis, France) was 60 min at 55 oC with a helium flow of equipped with a split/splitless injector 40 mL.min–1. In both analyses, the purge o . –1 (240 C, split flow of 120 mL min ), a FID was operated with a MCS temperature of o detector (245 C) and a capillary column 225 oC, a desorb temperature of 225 oCand coated with a FFAP phase (Stabilwax DA, a desorb time of 2 min. The chromato- Restek, Evry, France, length 30 m × internal graphic conditions were the same for both diameter 0.53 mm × phase thickness 1 µm), analysis methods and are described by and preceded by an uncoated retention gap Buchin et al. [13]. Ions between 29 and (Supelco, L’Isle d’Abeau, France, length 1 206 amu (atomic mass units) were collected m × internal diameter 0.53 mm). The carrier by the mass detector. The relative amount gas was helium (column head pressure of extracted for each volatile compound was . –1 o 40 kPa, flow of 6 mL min at 120 C), the given by the surface area of the peak of o temperature gradient was 120 C for 1 min, its most specific ion (semi-quantitative o o . –1 o 120 to 240 Cat10 C min , 240 Cfor5 analysis). min, 240 to 250 oC at 20 oC.min–1. 2.2.2.3. Gross chemical composition The analysis methods of pH, fat in dry matter, moisture of skim cheese, sodium Cheese flavour and pastures 761 chloride in the moisture, nitrogen soluble in Due to technological problems, four tri-sodium citrate 0.5 mol.L–1 at pH 4.4 nor- cheeses were withdrawn from the result malised to total nitrogen (pH 4.4-SN/TN), processing procedure. One cheese for pe- nitrogen soluble in phosphotungstic acid riod VX1 and for period MX1 (the ferment- normalised to total nitrogen (PTA-SN/TN), ing agent had very little acidification effect and casein ratios αS1-I/αS1+S2, resulting in the cheeses being acidified only αSdeg/αS1+S2, γ/β, have been described by very little) and two cheeses for MY2 (cool- Bugaud et al. [16]. ing occurred too rapidly during cheese pressing leading to insufficiently drained 2.2.3. Statistical analysis of the results cheeses) were not taken into consideration in the statistical analyses. The differences in flavour and the differences in volatile fatty acid and volatile compound composition between the cheeses 3. RESULTS were studied by analysis of variance (Ver- sion 6.12, SAS Institute INC., 1996) in two 3.1. Cheese flavour stages: (i) comparison between H, V and M cheeses, (ii) comparison between H The differences in flavour between the cheeses, between V cheeses and between M cheeses are presented in Tables I and II. cheeses. The analysis of variance carried 8 descriptors of Panel 1 and 10 of Panel 2 out on the H cheeses allowed us to deter- made it possible to significantly discrimi- mine the combined effect of cheese tech- nate the cheese groups (P < 0.05). nology and herd. In the analysis of the Of the H, V and M cheeses, M cheeses sensory descriptors used by Panel 1, the for- were the least pungent, the most “animal” age type, the judges and the forage type x (P < 0.001), “boiled milk”, “hazelnut” judge interactions were considered as fac- (P < 0.01) and “fresh cream” (P < 0.05). tors. The V cheeses had the least “fruity” and the To describe synthetically the volatile most “propionic acid” aromas (P < 0.05). compound composition of the cheeses, a The H cheeses had the least “bread crust” principal components analysis (PCA) was aroma (P < 0.05). performed on the volatile compounds (ex- Numerous descriptors made it possible cept volatile fatty acids) that exhibited sig- to distinguish between the V cheeses. The nificant differences at the 5% threshold in at VZ1 cheeses had the most intense odour least one of the analysis of variance models. and aroma (P < 0.001), they were the most The relationships between cheese fla- “fruity”, “cooked cabbage” (P < 0.001) vour and composition were assessed using and “rancid” (P < 0.01). The VZ2 cheeses Partial Least-Squares (PLS) regression [36] were the most pungent and salty carried out using Splus 3.4® software. The (P < 0.001) and were similar to the VZ1 sensory variables were explained using the cheeses for “cooked cabbage” odour and following variables: volatile fatty acids, “rancid” aroma. The VZ3 cheeses were the volatile compounds and gross chemical most bitter and “fresh cream” (P < 0.001). composition of the cheeses. Bitter and salty The variations in sensory properties be- tastes were analysed by linear correlation tween M cheeses were just as great as be- between: (i) salty taste and sodium chloride tween V cheeses. Nevertheless, due to in moisture, (ii) bitter taste and gross chem- greater variability between the replicates, ical composition, because both descriptors there were significantly fewer differences were supposed not to be influenced by the between the M cheeses than between the V volatile compound composition. cheeses. The MY1 cheeses were distinctly 762 Ch. Bugaud et al. ns (1.6) (1.4) (1.2) (1.0) (1.0) (1.4) (0.8) (1.1) (1.0) (1.3) (1.9) (0.8) (1.7) a b b ab bc a c bc ab b a ab ab (1.6) 1.4 (1.4) 2.3 (0.9) 1.9 (1.4) 1.8 (1.3) 2.6 (1.6) 1.6 (1.4) 2.2 (1.4) 1.6 (1.7) 2.2 (1.4) 2.0 (1.5) 1.7 (1.7) 1.7 (0.9) 1.6 *** a a b c a d b c a abc ab c bc M

different to the others in terms of their more (P < 0.01) and “bread crust” aroma intense and fruitier odour, their less pun- (P < 0.05). The HZ cheeses were the most gent taste (P < 0.01), their more “fresh bitter and had the most intense aroma milk” odour and their less “butyric acid” (P < 0.01). The HX cheeses had the fruiti- aroma (P < 0.05). The MX3 cheeses were est odour (P < 0.001). the most bitter and the most salty (P < 0.01). The MX2 cheeses resembled 3.2. Cheese composition the MY1 cheeses in terms of the intensity of their odour and “fruitiness” and they were 3.2.1. Volatile fatty acids the most “butyric acid” (P < 0.05). On the basis of identical feeding (hay The volatile fatty acid concentrations of period), the HY cheeses were the least salty the Abondance cheeses are given in Ta- (P < 0.001) and pungent (P < 0.01) with a ble III. Propionic acid content (C3) was more intense “fresh cream” odour found to vary greatly between the cheeses:

Table III. Concentrations of volatile fatty acids in cheeses (mg.100 g–1 cheese).

cheeses n C2 C3 C4 iC4 iC5 C6 HZ 3 130a (18) 175a (43) 9.5 (1.4) 1.4 (0.3) 3.7 (0.4) 2.9 (0.7) Hay HX 3 79c (5) 27b (12) 10.5 (1.0) 1.0 (0.7) 2.9 (1.3) 3.1 (0.6) HY 3 103b (4) 112a (54) 9.4 (1.0) 1.1 (0.4) 2.4 (0.8) 2.5 (0.2) analysis of variance ** * ns ns ns ns VZ1 3 88 (19) 39b (11) 10.1b (1.0) 3.4a (1.1) 7.6a (2.1) 3.8 (0.3) Valley pastures VZ2 3 55 (20) 33b (25) 9.3bc (0.4) 1.9b (1.0) 5.2ab (3.3) 3.6 (0.2) VZ3 3 94 (28) 138a (63) 8.1c (0.7) 1.3b (0.1) 3.1ab (0.2) 3.6 (0.8) VX1 2 74 (3) 15b (9) 12.2a (0.3) 1.6b (0.5) 3.5ab (0.7) 4.0 (0.1) VX2 3 71 (11) 8b (3) 10.4b (0.8) 1.0b (0.2) 2.4b (0.3) 3.4 (0.2) analysis of variance ns ** ** * * ns MX1 2 102a (26) 5ab (2) 7.7b (1.0) 2.0b (0.8) 3.6b (1.0) 3.3 (1.7) Mountain MX2 3 96a (10) 8a (1) 9.9b (1.1) 4.3a (0.8) 8.6a (1.5) 3.0 (0.5) pastures MX3 3 99a (6) 4b (1) 8.9b (0.5) 1.7b (1.1) 3.7b (1.7) 2.5 (0.1) MY1 3 64b (9) 2b (2) 9.2b (0.1) 0.6b (0.3) 1.5b (0.3) 3.1 (0.3) MY2 1 87ab (-) 6ab (-) 14.1a (-) 0.4b (-) 1.2b (-) 2.1 (-) analysis of variance * * ** ** ** ns analysis of variance * *** ns ns ns ** between H, V, M V

2-heptanol 2-butanone 2,3-butanedione 3-m-1-b-acetate 3-buten-2-one 1-m-p-propionoate 2-m-propanal 2-pentanone 2-m-1-butanol e-acetate 2-butanol 2-m-1-b-acetate 2-butene 3-m-2-pentanone 2-m-propanol 1-m-p-acetate 1-m-p-butanoate 2-heptanone 3-m-butanal 2,5-dim-furan e-propionoate tridecane 2-m-1-butene dodecane 3-m-furan 2-m-butanal 2-nonanone phenol propenal T nonane 1,3,5-trim-bz 3,7-dim-2-octene 2-e-furan e-octanoate styrene 1-pentene p-propionoate decanal 1-m-thiopropane 5-undecene 2-b-furan octene 2,4-dithiapentane 2-heptanol p-acetate 2-p-furan T T T 1,2,4-trim-bz 2-m-e-butanoate pentanal T T 3-m-e-butanoate 2-pentanol 2-propanol T T T T propanol m-acetate T S-m-thioacetate T 2-m-furan carbon-disulfide 2,4-octadiene T T dim-disulfide T methanethiol undecane T T T 4-octene dim-trisulfide 2-p-furan m-benzoate T m-benzene T n-b-propionoate pentanol 2-octene 3-m-3-buten-1-ol hexanol m-hexanoate 1,3-octadiene b-butanoate butanol b-acetate butanal

HY b) HY Axis 2 = 16 %

H cheeses HX MX1 VZ3 MY1 HX M cheeses MX3 HZ HX MX1 MY1 MX3 HZ MY2 HZ MX2 MX3 MX2 MY1 VX2 VX2 MX2 VZ3VZ3 VX2 VZ1 V cheeses Axis 1 = 36 % VZ1 VX1 VZ2 VZ1 VZ2 VZ2 VX1

Figure 2. Principal component analysis of the volatile compounds of the cheeses: plot of principal axes 1 and 2. a) Correlation circle. T: monoterpenes, m: methyl, e: ethyl, p: propyl, b: butyl, bz: benzene In bold: esters, italic bold: aldehydes, italic bold underline: ketones, italic: alcohols, boxed: sulphur- containing compounds, underline: furans, normal: hydrocarbons, benzene derivatives and terpenes. b) Representation of cheeses. Cheese flavour and pastures 767

Cheeses H, V and M were separated on gen fractions PTA-SN/TN and pH 4.4- axes PC1 and PC2. The M cheeses were the SN/TN, branched chain volatile fatty acids most rich in monoterpenes, furans and un- (iC4 and iC5) and acetic acid esters. These saturated hydrocarbons and the least rich in descriptors were associated negatively with esters, sulphur-containing compounds and 2-methyl-1-butanol and its derivatives and alcohols. The V cheeses were the least rich with propionic acid esters. Moreover, the in ketones, aldehydes and saturated hydro- intense and fruity odours were positively carbons. The H cheeses were the most rich associated with decanal and negatively as- in 2- and 3-methyl butanol precursor or de- sociated with fat in dry matter and C3. rivative compounds. The “butyric acid” aroma was positively Of the V cheeses, VX cheeses were the associated with branched chain volatile most rich in monoterpenes and VZ cheeses fatty acids (iC4 and iC5), nitrogen fractions were the most rich in sulphur-containing PTA-SN/TN and pH 4.4-SN/TN and sul- compounds. phur compounds, and negatively associated Of the M cheeses, MX1 and MX2 with ketones of 4 and 5 carbons. cheeses were less rich in monoterpenes The descriptor “fresh milk”, with low than MX3, MY1 and MY2 cheeses. On the prediction accuracy (explaining 35% of PC3 axis which is not shown (7% of total variance), was positively associated with variability), MX cheeses contrasted with ketones and negatively associated with es- MY cheeses in terms of their greater ters, sulphur compounds and branched amounts of ethyl esters, 2- and 3-methyl- chain volatile fatty acids. “Fresh cream” butanol and secondary alcohols such as 2- aroma was positively associated with pentanol and 2-heptanol, and their smaller propionic acid esters, whereas “fresh amounts of decanal. cream” odour was associated with satu- On the basis of identical feeding, HY rated and unsaturated hydrocarbons and cheeses were distinct from HX and HZ with pentanal derivatives. These two cheeses on the PC2 axis. The HY cheeses descriptors were negatively associated with were richer in ketones, aldehydes and precur- nitrogen fraction PTA-SN/TN and with salt sors or derivatives of 2- and 3-methyl butanol. content in the moisture. The “cooked cabbage” odour and the “animal” aroma were positively associated 3.3. Relationships between with sulphur compounds and negatively as- cheese flavour and sociated with esters. The secondary alco- cheese composition hols also predicted a “cooked cabbage” odour whereas branched chain fatty acids The relationships between the sensory and furans predicted the “animal” aroma. descriptors and the chemical composition variables of the ripened cheeses were stud- The “hazelnut” aroma, with low predic- ied using PLS regression. The first 25 vari- tion accuracy (explaining 21%of variance), ables, predicting each of the descriptors, was positively associated with monoterpenes were taken into account for the interpreta- and unsaturated hydrocarbons of 8 carbons tion. Only the interpretable descriptor/ vari- and negatively associated with secondary able relationships are given (Tab.VI). With alcohols and C4. the exception of “hazelnut” aroma and It was not possible to interpret the other “fresh milk” odour, the predictions of the descriptors (pungent taste, “bread crust”, descriptors explained over 53% of variance. “propionic acid” and “rancid” aroma, and The odour, aroma and “fruity” intensi- “boiled milk” odour) on the basis of the ties were positively associated with nitro- cheese chemical composition variables. 768 Ch. Bugaud et al.

Table VI. Prediction of sensory descriptors on basis of cheese composition by PLS-regression. descriptor positive correlations negative correlations explained variance odour intensity PTA-SN/TN, esters (4), C2, F/DM, M/SC, C3 99% aldehydes (2), iC4 aroma intensity PTA-SN/TN, pH 4.4- esters (6), ketones (2), 53% SN/TN, iC5, iC4, esters (3) alcohols (2) fruity odour PTA-SN/TN, iC4, F/DM, esters (3), 81% esters (4) alcohols (3), C3 fruity aroma PTA-SN/TN, iC4, iC5, S compounds (1), 71% pH 4.4-SN/TN, esters (4) esters (3), F/DM butyric acid aroma iC5, iC4, S compounds (4), esters (2), ketones (4), 63% PTA-SN/TN, esters (2), alcohols (3) pH 4.4-SN/TN, NaCl/M fresh milk odour furan (1), ketones (6) S compounds (3), 35% esters (6), alcohols (3), iC4, iC5 fresh cream aroma esters (6), ketones (1) S compounds (1), 80% NaCl/M, PTA-SN/TN fresh cream odour unsaturated NaCl/M, esters (2), PTA- 78% hydrocarbons (4) SN/TN, C6 cooked cabbage odour S compounds (4), esters (2), M/SC, 75% alcohols (4) C3, pH animal aroma S compounds (4), F/DM, esters (3) 76% iC4, iC5, furans (3), esters (2) hazelnut aroma unsaturated hydrocarbons alcohols (2), 21% (3), terpenes (8) esters (2), C4

Between ( ): number of compounds of the chemical family, S compounds: sulphur-containing compounds, C2: acetic acid, C3: propionic acid, C4: butyric acid, C6: caproic acid, iC4: iso-butyric acid, iC5: iso-valeric acid, PTA-SN/TN: nitrogen soluble in phosphotungstic acid in total nitrogen, pH 4.4-SN/TN: nitrogen soluble in a solution of tri-sodium citrate at pH 4.4 in total nitrogen, F/DM: fat in dry matter, M/SC: moisture of skim cheese, NaCl/M: sodium chloride content in the moisture. The salty taste could not be correlated the possible origin of the compounds with the salt content in the moisture, nor the responsible for cheese flavour and particu- bitter taste with the primary and secondary larly the origin of volatile fatty acid and proteolysis variables. other volatile compounds.

4. DISCUSSION 4.1. Volatile compounds in cheese

In order to understand the relationships Volatile compounds in cheese can origi- that may exist between cheese flavour and nate from two different sources. Most of types of pasture, it is necessary to identify them result from the metabolism of cheese Cheese flavour and pastures 769 microorganisms, endogenous microflora of The other volatile compounds identified raw milk or starters. In some cases they in Abondance cheeses were mainly the come directly from the feed. products of microbial metabolism. Our results confirm that terpenes The sulphur-containing compounds (mono- and sesquiterpenes) detected in present in cheese mainly originate from the cheese were of plant origin [8, 9, 22, 24, degradation of sulphur-containing amino 39]. Cheeses M and VX, which were the acids, notably methionin, by the bacteria most rich in terpenes, were made from the present in the cheese [11]. The richness in most terpene-rich milks [15]. According to sulphur-containing compounds of VZ Bugaud et al. [15] and Viallon et al. [40], cheeses can therefore have been linked to the terpene composition of milk depends on the composition of this microbial flora, but that of the forage consumed, which in turn also to the physico-chemical conditions of depends on the botanical composition of cheese. Thereby, the higher pH value of the forage [14, 33]. Most notably, M and these cheeses may have promoted the pro- VX pastures where dicotyledons – in duction of sulphur-containing compounds particular umbelliferae, composites and [11]. Conversely, the greater amounts of plantaginaceae – were predominant, were terpenes in VX and M cheeses may have in- the most terpene-rich, both in terms of di- hibited the production of sulphur-contain- versity and quantity [14]. By contrast, VZ ing compounds, which have been found pastures, which were rich in gramineae, less abundantly in these cheeses. Indeed, were poor in terpenes and therefore resulted the hypothesis that terpenes have an inhibi- in terpene-poor cheeses. tory action on the production of sulphur- On the contrary, the unsaturated hydro- containing compounds, and possibly of carbons, which were found more abun- other compounds, has already been sug- dantly in M cheeses because they were gested [35], but to date no study has demon- distributed uniformly in milk, were un- strated it. likely to have originated directly from Propionic acid fermentation has been plants. According to Frankel [25], some of shown to occur to a lesser extent in these compounds, such as 2- and 4-octene, Abondance cheeses than in Swiss type may result from the oxidation of unsatu- cheeses (Comté, [3], Emmental [5]), in rated fatty acids which were found in particular due to ripening in a colder cellar greater proportions in mountain milks [15]. (10–11 oC). The high degree of propionic According to Dumont and Adda [23], acid fermentation of VZ3 cheeses was cer- the furans present in dairy products may be tainly related to the higher pH value of these the result of a Maillard reaction between an cheeses at the end of pressing [26], and a amino acid and a sugar. There is however possible relationship with green maize- another possible pathway for the produc- based feed is difficult to explain. It is also tion of furans, which has never been sug- possible that there was an inhibitory effect gested in dairy products, i.e. the oxidation of long chain unsaturated fatty acids, found of polyunsaturated fatty acids [17]. In our in greater proportions in M milks, on study, the total amount of furans in cheese propionic acid bacteria [10]. was correlated with the proportion of poly- Isobutyric acid and isovaleric acid result unsaturated fatty acids in the milk from the degradation of amino acids, valine (R2 = 0.62, P < 0.001). The higher polyun- and leucine respectively [23, 37]. In MX2, saturated fatty acid contents in mountain VZ1 and VZ2 cheeses, the formation of milk [15] may thereby explain the greater branched chain volatile fatty acids ap- amounts of furans in the corresponding peared to be related to the amount of free cheeses. amino acid precursors, the presence of 770 Ch. Bugaud et al. which was indicated by the PTA-SN/TN cheese, also contribute to the intensity and fraction. The catabolism of amino acids aromatic richness of odour and aroma. This also leads to the formation of aldehydes, al- is consistent with the results of Berdagué cohols and branched chain esters, such as et al. [4] who showed that the fruitiness of the derivatives and precursors of 2,3- Comté cheeses was correlated with methyl butanol and 2-methyl propanol isovaleric acid. This explains how the [30]. The fact that these compounds were higher concentrations of small peptides and present in greater quantities in H cheeses free amino acids in cheeses MX2, MY1 and may be due to a higher enzyme potential VZ1 led to the fruitiest cheeses. and/or a physico-chemical environment that was more favourable to their formation The intense “cooked cabbage” odour of and preservation in cheese. cheeses VZ1 and VZ2 was mainly related to sulphur compounds that was found more Long chain methyl-ketones and second- abundantly in these cheeses. This is consis- ary alcohols (of over 4 carbons) result from tent with the studies by Cuer et al. [19] and fatty acid catabolism whereas the shorter by Kubickova and Grosch [31] who demon- ones result from fermentation. In particular, strated the important role played by sulphur 2,3-butanedione, 2-butanone and 2-butanol compounds in the “cooked cauliflower”, result from citrate fermentation [23]. The “cabbage” and “garlic” aromas in cheese. use of a particular acidifying agent (calf Moreover, certain sulphur-containing com- rennet macerated in boiled acidified whey) pounds, associated with isobutyric acid, by producer Y may be the reason explain- isovaleric acid and furans, may contribute ing the higher concentrations of these com- to the “animal” aroma of cheeses. Given the pounds observed in Y cheeses. distribution of these different molecules between cheeses, it would appear that the “an- 4.2. Cheese flavour imal” note of VZ1 cheeses was related to the presence of sulphur-containing com- In spite of the fact that some compounds pounds and volatile fatty acids, whereas the that are potentially tasteful or aromatic “animal” note of M cheeses was related (amino acids, peptides, pyrazines, rather to the presence of furans. The latter furanones, amines, lactones) were not compounds, which are not very aromatic, taken into consideration in this study, it was may be labelling compounds of other mole- possible to identify relationships between cules that are aromatic and characteristic of cheese composition and cheese flavour. the “animal” note. Bosset et al. [9] observed that mountain Gruyère cheeses had a more The descriptors intensity and aromatic intense “animal” odour than cheeses from richness (fruity), both in terms of aroma plains, although they were unable to relate and odour, had very clear links to the sec- this odour to volatile compounds in cheese. ondary proteolysis indicators (PTA- SN/TN). These results, which confirm Milky aromas have been preferentially those found on cooked pressed cheeses [4, perceived in the absence of compounds 6] or on Cheddar cheese [2], show that free with strong aromatic notes, such as sul- amino acids and small peptides play an im- phur-containing compounds or branched portant role in the flavour of Abondance chain fatty acids. Therefore, the more in- cheese by contributing either directly to tense “fresh milk” and “fresh cream” cheese taste or indirectly as precursors of odours of Y cheeses may be related to mol- flavour compounds. Thereby, isobutyric ecules that are not very odorous, such as 2- acid and isovaleric acid, originating from butanone and 2-butanol [37] or characteris- amino acids, and which have been associ- tic of milky aroma such as 2,3-butanedione ated with the “butyric acid” aroma in [31]. Cheese flavour and pastures 771

The only aroma that seems to be related parameters in the milk and cheese deter- to terpenes was “hazelnut” aroma, which mining the way in which microorganisms was more intense in the M cheeses. How- develop and their enzyme activity. ever this note was only weakly perceived in This work completes three previous the cheeses (maximal note below 1.9 on a studies [14–16], and made it possible to 7 point scale) and its prediction rates by link the botanical composition of pastures terpenes was low (only 21% of variance ex- and the terpene composition of cheeses. plained). So, although Guichard [29] in However, the role played by these terpenes cooked pressed cheeses, and Moio et al. as well as fatty acids, which were also re- [34] in Gorgonzola cheese, identified a fruit lated to pasture properties, seems to con- note from limonene by olfactometry, the di- tribute indirectly to cheese flavour. rect role played by terpenes in the the over- all appreciation of cheese flavour appears It would be worthwhile to validate these to be limited. results by carrying out targeted studies. The retention of volatile compounds by However, these results should make it pos- fat could explain why odour intensity and sible to improve our understanding of the “fruity” odour were negatively correlated origin of the diversity in cheese flavours as to the fat in dry matter content of the well as provide additional information and cheeses. objectives for further thought concerning the evolution of the PDO cheese sector. Our study did not enable us to relate chemical composition to salty and bitter tastes in cheese and would suggest that the ACKNOWLEDGEMENTS latter were linked rather to the peptide and amino acid composition [6, 32] which were not analysed. The more intense bitterness This study was co-financed by the Northern of cheeses VZ3 and MX3 may also be ex- Alps Scientific Interest Grouping and the French National Institute for Agricultural Research. We plained by a delay in acidification observed would like to thank J.C. Salmon (INRA, during the manufacture [16]. These delays Poligny) for his help in analysing and identify- in acidification, probably related to the ing volatile compounds, L. Tessier (INRA, presence of natural starter inhibitors in the Poligny) for carrying out the PLS regressions, T. milk, may have led to defects in drainage Métais (Les Maisons du Goût, La Roche-sur- and the occurrence of off-tastes such as bit- Foron) for undertaking the sensory assessments terness [41]. and B. Martin (INRA, Theix) and Y. Noël (INRA, Poligny) for critically reviewing the manuscript. 5. CONCLUSION REFERENCES This study was undertaken under natural conditions of cheese production and so made it possible to demonstrate consider- [1] Ardö Y., Polychroniadou A., Laboratory manual for chemical analysis of cheese, Office for offi- able sensory differences between the cial publications of the european communities, cheeses depending on cow feeding (valley Luxembourg, 1999. pastures, mountain pastures or hay). Most [2] Aston J.W., Durward I.G., Dulley J.R., Proteoly- of the volatile compounds were microbially sis and flavour development in Cheddar cheese, produced and depended on both the compo- Aust. J. Dairy Technol. (1983) 55–59. sition of the microflora present (endoge- [3] Berdagué J.L., Jeunet R., Grappin R., Affinage et qualité du gruyère de Comté. III. Fermentation nous microflora in the raw milk and/or from lactique et teneur en acides gras volatils des starters) and on the physico-chemical fromages de Comté, Lait 67 (1987) 249–263. 772 Ch. Bugaud et al.

[4] Berdagué J.L., Grappin R., Affinage et qualité du [16] Bugaud C., Buchin S., Noël Y., Tessier L., gruyère de Comté. VI. Caractéristiques Pochet S., Martin B., Chamba J.F., Relationships sensorielles des fromages, Lait 68 (1988) between Abondance cheese texture, its composi- 189–204. tion and that of milk produced by cows grazing [5] Berdagué J.L., Grappin R., Delacroix-Buchet different types of pastures, Lait 81 (5) (2001) A., Chaillet B., Caractérisation de l’emmental 593–607. français « grand-cru » I. Composition physico- [17] Chang S.S., Smouse T.H., Krishnamurthy R.G., chimique, Lait 70 (1990) 1–14. Mookherjee B.D., Reddy R.B., Isolation and [6] Berdagué J.L., Grappin R., Chaillet B., Clément identification of 2-pentyl-furan as contributing J.F., Caractérisation de l’emmental français to the reversion flavour of soyabean oil, Chem. « grand-cru » II. Analyses sensorielles, Lait 70 Ind. 12 (1966) 1926–1927. (1990) 133–145. [18] Corroler D., Mangin I., Desmasures N., [7] Beuvier E., Berthaud K., Cegarra S., Dasen A., Gueguen M., An ecological study of lactococci Pochet S., Buchin S., Duboz G., Ripening and isolated from raw milk in the Camembert cheese quality of Swiss-type cheese made from raw, registered designation of origin area, Appl. En- pasteurized or microfiltered milk, Int. Dairy J. 7 viron. Microbiol. (1998) 4729–4735. (1997) 311–323. [19] Cuer A., Dauphin G., Kergomard A., Roger S., [8] Bosset J.O., Bütikofer U., Gauch R., Sieber R., Dumont J.P.,Adda J., Flavour properties of some Caractérisation de fromages d’alpages subalpins sulphur compounds isolated from cheeses, suisses : mise en évidence par GC-MS de Lebensm. –Wiss.-Technol. 12 (1979) 258–261. terpènes et d’hydrocarbures aliphatiques lors de [20] Demarigny Y., Beuvier E., Buchin S., Pochet S., l’analyse par « Purge and Trap » des arômes Grappin R., Influence of raw milk microflora on volatils de ces fromages, Schweiz. Milchw. the characteristics of Swiss-type cheeses: II. Forschung 23 (1994) 37–41. Biochemical and sensory characteristics, Lait 77 [9] Bosset J.O., Jeangros B., Berger T., Bütikofer U., (1997) 151–167. Collomb M., Gauch R., Lavanchy P., Scehovic [21] Desmasures N., Gueguen M., Monitoring the J., Sieber R., Comparaison de fromages à pâte microbiology of high quality milk by monthly dure de type Gruyère produits en région de sampling over 2 years, J. Dairy Res. 64 (1997) montagne et de plaine, Rev. Suisse Agric. 31 271–280. (1999) 17–22. [22] Dumont J.P., Adda J., Occurrence of sesquiterpenes [10] Boyaval P., Corre C., Dupuis C., Roussel E., Ef- in mountain cheeses volatiles, J. Agric. Food fects of free fatty acids on propionic acid bacte- Chem. 26 (1978) 364–367. ria, Lait 75 (1995) 17–29. [23] Dumont J.P., Adda J., Flavour formation in dairy [11] Bruinenberg P.G., de Roo G., Limsowtin G.K.Y., products, in: Land D.G., Nursten H.E. (Eds.), Purification and characterization of cystathione Progress in flavour research, Appl. Sci. Pub- g-lyase from Lactococcus lactis subsp. lishers, London, 1979, pp. 245–262. Cremoris K11: possible role in flavor compound formation during cheese maturation, Appl. Envi- [24] Dumont J.P., Adda J., Rousseaux P., Exemple de ron. Microbiol. 63 (1997) 561–566. variation de l’arôme à l’intérieur d’un même type de fromage : le Comté, Lebensm. –Wiss.- [12] Buchin S., Delague V., Duboz G., Berdagué J.L., Technol. 14 (1981) 198–202. Beuvier E., Pochet S., Grappin R., Influence of pasteurization and fat composition of milk on the [25] Frankel E.N., Lipid oxidation, Prog. Lipid Res. volatile compounds and flavor characteristics of 19 (1980) 1–22. a semi-hard cheese, J. Dairy Sci. 81 (1998) [26] Fryer T.F., Peberdy M.F., Growth of 3097–3108. propionibacteria in Swiss and Egmont cheese, [13] Buchin S., Martin B., Dupont D., Bornard A., N.Z. J. Dairy Sci. Technol. 12 (1977) 133–134. Achilleos C., Influence of composition of Alpine [27] Gagnaire S., Vanhaecke C., Quelle relation peut- highland pasture on the chemical, rheological on envisager entre la nature de la flore pâturée et and sensory properties of cheese, J. Dairy Res. la qualité sensorielle des fromages, Rev. ENIL 66 (1999) 579–588. 216 (1998) 14–15. [14] Bugaud C., Bornard A., Hauwuy A., Martin B., [28] Grappin R., Coulon J.B., Terroir, lait et fromage : Salmon J.C., Tessier L., Buchin S., Relation en- éléments de réflexion, in : 3rd Renc. Rech. Rum., tre la composition botanique de végétations de Institut de l’Élevage, Paris, 1996, pp. 21–28. montagne et leur composition en composés [29] Guichard H., Contribution au développement de volatils, Fourrages 162 (2000) 141–155. méthodes instrumentales pour la caractérisation [15] Bugaud C., Buchin S., Hauwuy A., Coulon J.B., sensorielle des fromages à pâte pressée cuite de Dupont D., Influence of the nature of Alpine pas- type Emmental, Comté, Beaufort : Étude de la tures on plasmin activity, fatty acid and volatile fraction volatile et de la fraction azotée soluble compound composition of milk, Lait 81 (3) dans l’eau, Thèse, Université de Bourgogne, (2001) 401–414. ENSBANA, 1995. Cheese flavour and pastures 773

[30] Hemme D., Bouillanne C., Métro F., de type l’Etivaz ou Gruyère, Rev. Suisse Agric. Desmazeaud M.J., Microbial catabolism of 30 (1998) 167–171. amino acids during cheese ripening. Sci. Ali- [36] Tenenhaus M., La regression PLS. Théorie et ments 2 (1982) 113–123. pratique, Ed. Technip, Paris, 1998. [31] Kubickova J., Grosch W., Quantification of po- [37] Urbach G., The flavour of milk and dairy prod- tent odorants in Camembert cheese and calcula- ucts: II. Chesse: contribution of volatile com- tion of their odour activity values, Int. Dairy J. 8 pounds, Int. J. Dairy Technol. 50 (1997) 79–89. (1998) 17–23. [38] Verdier-Metz I., Coulon J.B., Pradel P., Viallon [32] Lemieux L., Simard R.E., Bitter flavour in dairy C., Albouy H., Berdagué J.L., Effect of the bo- products. II. A review of bitter peptides from ca- tanical composition of hay and casein genetic seins: their formation, isolation and identifica- variants on the chemical and sensory character- tion, structure masking and inhibition, Lait 72 istics of ripened Saint-Nectaire type cheeses, (1992) 1289–1297. Lait 80 (2000) 361–370. [33] Mariaca R.G., Berger T.F.H., Gauch R., Imhof [39] Viallon C., Verdier-Metz I., Denoyer C., M.I., Jeangros B., Bosset J.O., Occurrence of Pradel P., Coulon J.B., Berdagué J.L., Desorbed volatile mono- and sesquiterpenoids in highland terpenes and sesquiterpenes from forages and and lowland plant species as possible precursors cheeses, J. Dairy Res. 66 (1999) 319–326. for flavor compounds in milk and dairy products, [40] Viallon C., Martin B., Coulon J.B., Berdagué J. Agric. Food Chem. 45 (1997) 4423–4434. J.L., Pradel P., Garel J.P.,Transfer of monoterpenes [34] Moio L., Piombino P., Addeo F., Odour-impact and sesquiterpenes from forages into milk fat, compounds of Gorgonzola cheese, J. Dairy Res. Lait 80 (2000) 635–641. 67 (2000) 273–285. [41] Weber F., Les germes utiles susceptibles d’être à [35] Scehovic J., Jeangros B., Troxler J., Bosset J.O. l’origine d’altérations, in : Hermier J., Lenoir J., Effets de la composition botanique des herbages Weber F. (Eds.) Les groupes microbiens d’intérêt pâturés sur quelques composants des fromages laitier, CEPIL, Paris, 1992, pp 371–394.

To access this journal online: www.edpsciences.org