Relationships Among Dietary Roasted Soybeans, Milk Components, and Spontaneous Oxidized Flavor of Milk 1

Transcription

1 J. Dairy Sci. 84: American Dairy Science Association, Relationships Among Dietary Roasted Soybeans, Milk Components, and Spontaneous Oxidized Flavor of Milk 1 J. S. Timmons,* W. P. Weiss,* D. L. Palmquist,* and W. J. Harper *Department of Animal Sciences, Ohio Agricultural Research and Development Center, The Ohio State University, Wooster, OH Department of Food Science and Technology, The Ohio State University, Columbus, OH ABSTRACT Relationships among dietary roasted whole soybeans (RSB), milk fatty acid profile, and the development of spontaneous oxidized flavor of milk were investigated by using 20 commercial dairy herds. Diets contained 0 to 15.3% of dry matter as RSB. Concentrations of dietary RSB were correlated positively with concentrations of C18:2 and C18:3 in milk fat. Concentrations of α-tocopherol, β-carotene, and ascorbic acid in milk decreased from 0 to 3dofstorage (4 C), and oxidized flavor in milk increased linearly between 0 and 8dof storage. Milk fatty acid profile did not change during storage. The development of oxidized flavor at 8 d postsampling was correlated (r) with increased concentrations in milk fat of C18:2 (0.49), C18:3 (0.55), total polyunsaturated milk fatty acids (0.50), and dietary concentrations of RSB (0.38). Multiple regression was used to quantify relationships between variables and oxidized flavor (samples stored 8 d). All significant models included milk concentrations of Cu and dehydroascorbic acid. Concentrations of C18:2, C18:3, or total polyunsaturated fatty acids in milk fat, or dietary RSB concentrations, and interactions of those variables with Cu were included in individual models. Milk with high concentrations of polyunsaturated fatty acids and Cu were most susceptible to oxidation. Feeding RSB increased polyunsaturated fatty acid concentrations in milk fat, which increased the likelihood of oxidized flavor, especially when milk had high concentrations of Cu. (Key words: roasted soybeans, milk fatty acids, spontaneous oxidized flavor, copper) Received December 7, Accepted June 12, Corresponding author: W. P. Weiss; weiss.6@osu.edu. 1 Salaries and research support provided by state and federal funds appropriated to the Ohio Agricultural Research and Development Center, The Ohio State University and by Dairy Management Inc., Rosemont, IL. Abbreviation key: AA = ascorbic acid, DHAA = dehydroascorbic acid, PUFA = polyunsaturated fatty acid, RMSE = root mean square error, RSB = roasted soybeans, SOF = spontaneous oxidized flavor. INTRODUCTION Oxidized flavor of milk is described as being metallic, cardboardy, tallowy, or fishy and reduces the acceptability of dairy products by the consumer. Spontaneous oxidized flavor (SOF) develops without the addition of exogenous oxidants (e.g., Cu) or without exposure to light. The SOF is a result of the production of volatile aldehydes and other carbonyl compounds after the formation of hydroperoxides (Frankel, 1991) from the oxidation of pentadienyl groups in polyunsaturated fatty acids (PUFA). The oxidative process is autocatalytic, requiring only one initiating radical to start the production of lipid hydroperoxides; thus, SOF intensifies as storage time increases. The susceptibility of milk to SOF varies; some milk develops SOF very rapidly (detectable in the bulk tank), but other milk may not develop SOF until after several days of storage. Most commercial dairy herds have only a small number of cows (<10%) producing milk susceptible to SOF, but SOF is detectable in bulk tank milk only when >30% of the herd is affected (Nicholson and Charmley, 1991). Numerous studies have characterized SOF and extensive efforts to define the causes of SOF have been undertaken (Barrefors et al., 1995; Bruhn et al., 1976; Charmley and Nicholson, 1994; Dunkley and Franke, 1967). Milk with high concentrations of linoleic acid is more susceptible to autoxidation (Sidhu et al., 1975; Smith et al., 1963). Roasted soybeans (RSB) are high in PUFA, and feeding RSB increases the proportion of those fatty acids in milk fat, especially 18:2 and 18:3 (Dhiman et al., 1999; Tice et al., 1994). Because of the effect that feeding RSB has on milk fatty acid profile, the relationship between feeding RSB and the occurrence of SOF was investigated. The change in milk fatty acid profile in combination with the balance between prooxidants and antioxidants has not been stud- 2440

2 SOYBEANS AND OXIDIZED FLAVOR OF MILK 2441 Table 1. Nutrient composition of diets fed to the sample herds (DM basis). 1 Mean SD Minimum Maximum Roasted soybeans, % CP, % NDF, % Fatty acids, % C18:2, 2 % C18:2, % of total fatty acids C18:3, 2 % C18:3, % of total fatty acids Ca, % P, % Cu, mg/kg Fe, mg/kg Mn, mg/kg Zn, mg/kg α-tocopherol, mg/kg β-carotene, mg/kg n = 40 (20 herds 2 samples per herd). 2 Fatty acid number of carbons: number of double bonds. ied extensively in a field situation. Therefore, the objectives of this study were as follows: 1) to determine factors that are related to the development of SOF in milk when whole RSB are fed; 2) to quantify changes in antioxidant concentrations in milk during storage; and 3) to model factors involved with the development of SOF in milk from commercial herds. Herds and Samples MATERIALS AND METHODS In a preliminary survey, samples of bulk tank milk from commercial herds located near Wooster, OH, were taken in September 1998 and again in October 1998, for flavor analysis and milk fatty acid profile measurements. Based on those results, 20 herds were selected that represented a wide range in organoleptic flavor and milk fatty acid profile, in particular linoleic acid. For the 20 selected herds, herd size ranged from 35 to 430 cows with a rolling herd average of 6600 to 14,000 kg/yr. Concentrations of milk fat ranged from 3.4 to 4.8% and milk crude protein from 3.0 to 3.6%. Eighteen herds were comprised of Holstein cows, one herd was comprised of Jersey cows, and one herd consisted of a mix of Holstein and Jersey cows. Cows in all herds were milked twice daily and were fed one (5 herds), two (13 herds), or three (2 herds) times daily. Milk from seven herds was shipped to processors daily, and milk from 13 herds was shipped on alternate days. The concentration of RSB in the dietary DM ranged from 0 to 15.3% (Table 1). Diets were composed primarily of corn silage, alfalfa silage, high-moisture corn, and protein and mineral supplements. No fresh forage was consumed by the cows. Other than RSB, eight herds were fed no additional supplemental fat, four herds were fed tallow (0.6 to 1.1% of diet DM), five herds were fed calcium salts of fatty acids (1.0 to 1.5% of diet DM), two herds were fed whole cottonseed (3.7 and 9.0% of diet DM), and one herd was fed 1.5% calcium salts of fatty acids and 8% whole cottonseed. Information on nutrient composition of the diets is in Table 1. Concentration of Cu in water from the farms averaged 0.02 mg/l (SD = 0.04). Samples of milk were taken from each herd in December 1998, January 1999, February 1999, and April Milk samples were collected from the bulk tank after the morning milking before milk pickup in 2- L acid-washed glass jars. Samples were immediately placed on ice, protected from light, and arrived at the laboratory within 2 h. Upon arrival, milk was pasteurized at 62.7 C for 30 min and then immediately placed in an ice bath for 20 min. A 1-L subsample was transported on ice to another laboratory for organoleptic flavor analysis. Feed and water samples were taken in January 1999 and April Milk Assays Day 0 milk samples were analyzed by DHI Cooperative, Inc. (Powell, OH) for fat and crude protein percentage per approved procedures (AOAC, 1995) with a B2000 Infrared Analyzer (Bentley Instruments, Chaska, MN), for SCC by flow cytometry, and for urea N by the diacetyl monoxime procedure (automated version of Sigma Procedure 535-A, Sigma Aldrich Chemical Co., St. Louis, MO). Milk fatty acid profile was determined on 0, 3, and 8 d postsampling (Sukhija and Palmquist, 1988). Polyunsaturation index was calculated as the sum of each milk PUFA multiplied by its number of pentadienyl groups (i.e., 18:2 = one group, 18:3 = two groups).

3 2442 TIMMONS ET AL. α-tocopherol and β-carotene were extracted (in reduced light) from milk on 0 and 3 d postsampling (Indyk, 1988). Extracts were frozen at 80 C until assayed for α-tocopherol and β-carotene by using HPLC with a Supelcosil LC-18 column (25 cm 4.6 mm i.d., 5 µm particle size, no ; Supelco, Bellefonte, PA). For α-tocopherol, the solvent was an isocratic mixture of 95% methanol and 5% butanol at a flow rate of 1.8 ml/min and detection was at 294 nm. For β-carotene, the solvent was an isocratic mixture of 75% methanol and 25% butanol, and detection was at 450 nm. Ascorbic acid (AA) in milk was measured in d 0 and 3 samples by using the procedures of Barrefors et al. (1995) and Sapers et al. (1990) with slight modification. Ascorbic acid was determined by adding 1 ml of freshly made 1.12% metaphosphoric acid (ph 6.5) to1mlof37 C milk. Samples were centrifuged (4 C) at 12,000 g for 30 min in 2-ml filtration vials (Ultrafree-Cl centrifugal filters, no. UFC4 LTK; Millipore Corp., Bedford, MA) in reduced light. The supernatant was analyzed immediately for AA by using an HPLC system that consisted of a Hibar column (25 4 mm, i.d.; Maxil NH 2 ;5µm particle size; Phenomenex, Torrance, CA) and an isocratic mixture of 70% acetonitrile and 30% 0.05 M KH 2 PO 4 (ph 7.5) at a flow rate of 1.5 ml/min with detection at 254 nm. The sum of AA and dehydroascorbic acid (DHAA) was measured by using the same procedure except that the metaphosphoric acid contained 0.05% dithiothreitol (ph 6.5), which converts DHAA to AA. The concentrations of DHAA was calculated by subtraction of AA from the sum of DHAA plus AA. Xanthine oxidase was determined by the procedure of Granelli et al. (1995) on 0-, 3-, and 8-d samples. One enzyme unit was defined as the amount of xanthine oxidase needed to form 1 µmol of urate/min at ph 7.4 and 25 C. Concentrations of Cu, Zn, and Mn in milk were determined by using a modified procedure of AOAC (1995). Milk was warmed to 37 C and weighed (20 g) into 50-ml acid-washed ceramic crucibles that had been soaked in distilled water for 2 h. Blanks (HPLC grade water) and a standard (National Institute of Standards and Technology, Gaithersburg, MD) were analyzed with each run. Contents of crucibles were evaporated to dryness in a sand bath with frequent stirring by using a glass rod. Concentrated HNO 3 (10 ml) was added to the crucibles, evaporated to dryness, and the residue was ashed overnight at 420 C. Concentrated HNO 3 (5 ml) was added to cooled samples, and contents were evaporated to near dryness on a hot plate (100 C). Another 10 ml of 1 N HNO 3 was added and evaporated to near dryness. The residue was transferred to a 10-ml volumetric flask, brought to volume with 1% HNO 3, and analyzed for Cu, Zn (after 1:6 dilution with 1% HNO 3 ), and Mn (after 1:1 dilution with an aqueous solution of 0.05% CaCl 2 and 1% NaCl) by using flame atomic absorption spectrophotometry (Varian Spectra AA 200, Palo Alto, CA). A trained panel of six judges conducted organoleptic flavor analysis of samples at room temperature that were stored (4 C) for 0, 3, and 8 d. Milk sample identities were masked, and they were evaluated according to the following scale: 0 1, no oxidized flavor; 1 2, slight; 2 3, moderate; 3 5, strong; and >5, intense oxidized flavor. Judges worked independently and evaluated seven samples in a session. Feed and Water Analysis The TMR samples were hammer-milled, and a portion of each was lyophilized and ground through a 1- mm screen (Wiley Mill, Arthur Thomas, Philadelphia, PA). Samples were stored at 20 C in the dark until analysis to preserve vitamins. Dry matter of TMR was determined by oven-drying at 100 C for 24 h. Lyophilized TMR samples were analyzed for NDF (Van Soest et al., 1991), N (Kjeldahl Procedure 7.015; AOAC, 1995), and long-chain fatty acids (Sukhija and Palmquist, 1989). Total α-tocopherol (tocopheryl acetate was saponified) and β-carotene in TMR samples were extracted (McMurry et al., 1980) and analyzed by using the same HPLC system as for α-tocopherol and β- carotene in milk. Minerals in TMR and water samples were digested by perchloric acid and assayed by inductively coupled plasma spectroscopy (Star Laboratory, Wooster, OH). Statistical Analysis The effect of day postsampling on concentrations of vitamins (0 and 3 d) and milk flavor (0, 3, and 8 d) was determined by using the general linear model of SAS (1988). The model included month, herd nested within month, day, and day-by-month interaction. Day was included as a repeated measure. The herd nested within month was used to test the effect of month. Correlations were analyzed on d 0 for all variables by using Proc CORR (SAS, 1988). Multiple regression analysis was run by stepwise backward elimination (P < 0.10) using Proc REG (SAS, 1988). The dependent variable was the flavor score of d 8 samples. All independent variables and the interactions between Cu content of milk and proportions of fatty acids were included in the model except those that were highly correlated (P < 0.01) with each other. When the interaction term was significant, the main effect components of the interaction were forced into the regression

4 Table 2. Milk composition (0 d milk) and flavor scores. 1 SOYBEANS AND OXIDIZED FLAVOR OF MILK 2443 Mean SD Minimum Maximum Fat, % Protein, % Urea N, mg/dl SCC, 1000/ml Xanthine oxidase, U/ml Milk fatty acids 2 (% of total fatty acids) 4: : : : : : : : : trans 18: cis 18: : : CLA Other Total polyunsaturated Polyunsaturated index Minerals and vitamins, mg/l Cu Mn Zn β-carotene α-tocopherol Ascorbic Acid Dehydroascorbic acid Flavor score 3 0 d of storage d of storage d of storage n = 80 for all measures except ascorbic acid and dehydroascorbic acid (n = 66) and milk flavor score (n = 79). 2 Number of carbons:number of double bonds. CLA = Conjugated linoleic acid; polyunsaturated index = 18:2 + (18:3 2). 3 Score: 0 = normal, 5 = intense oxidized flavor. model. The total number of samples was 80 (20 herds 4 mo); however, one sample for organoleptic analysis was lost when the container broke, therefore, only 79 observations were available for most analyses. Another 13 samples could not be analyzed for AA and DHAA because of an instrument failure; therefore, 66 observations were used for the multiple regression analysis. RESULTS AND DISCUSSION Milk Composition and Dietary RSB Simple statistics describing milk variables are shown in Table 2. The concentrations of milk components other than fatty acids (i.e., vitamins and minerals) were not correlated with the concentration of dietary RSB. Concentrations of many milk fatty acids were correlated with the concentration of RSB in the diet (Table 3). In general, the concentrations of 18 C fatty acids and unsaturated fatty acids increased and concentrations of medium-chain length fatty acids (10:0 to 16:1) in milk fat decreased with increasing dietary concentrations of RSB. The short-chain milk fatty acids (4:0, 6:0, and 8:0) were not correlated (P > 0.05) with RSB. The strongest negative correlations between RSB and milk fatty acids were found for 16:0 ( 0.70) and 16:1 ( 0.79), and the strongest positive correlations were found for 18:2 (0.86) and 18:3 (0.69). The higher variation in 18:3 concentrations at low RSB (Figure 1) is attributed to the high proportion of 18:3 in hay crop forage and to differences among farms in the amount of hay crop forage included in the TMR. Based on linear regression, a one percentage unit increase in concentration of dietary RSB resulted in an

5 2444 TIMMONS ET AL. Table 3. Linear relationships between dietary concentrations (% of DM) of roasted soybeans and certain milk fatty acids (% of total fatty acids). 1 Model b 1 variable 2 b 1 b 1 SE Intercept SE r 1 16: : : : : C UNSAT PUFA n = :0, 16:1, 18:0, 18:2, and 18:3 fatty acids with number of carbons:number of double bonds. C18 = Sum of milk fatty acids with 18 carbons; UNSAT = sum of all unsaturated milk fatty acids; PUFA = polyunsaturated fatty acids (milk fatty acids 18:2 + 18:3 + conjugated linoleic acid). increase of 0.2 and 0.04 percentage units for 18:2 and 18:3 in milk fat (Figure 1) and a decrease of about 0.5 percentage units for medium-chain length fatty acids (data not shown). The equations relating dietary RSB to concentrations of 18:2 and 18:3 in milk fat were tested by using independent data (Chouinard et al., 1997; Dhiman et al., 1999; Tice et al., 1994; Timmons, 1999). Both equations produced reasonable estimates for 18:2 and 18:3 concentrations in milk fat (Figure 1). Increasing dietary RSB had no effect on the dietary concentration of 18:3 when expressed as a percent of dietary DM or dietary fatty acids (Figure 2). The lack of a relationship between dietary RSB and dietary 18:3 occurred because forages were the predominant source of dietary 18:3. The concentration of dietary 18:2 as percentage of total dietary fatty acids increased with increasing RSB (Figure 2), but the relationship was only moderately strong (r 2 = 0.25; P < 0.01). When dietary 18:2 was expressed as a percentage of dietary DM (Figure 2), the relationship with dietary RSB was stronger (r 2 = 0.53; P < 0.001). The RSB was related to increased milk 18:2 (Figure 1) possibly because of the increased intake of 18:2 (Figure 2), but the relationship between RSB and milk 18:3 does not appear to be simply a result of increased dietary concentrations of 18:3. Some chemical or physical characteristic of RSB may influence biohydrogenation of the unsaturated fatty acids in RSB (Morales et al., 2000). Changes in Milk During Storage The intensity of oxidized flavor increased linearly (P < 0.05) with days of storage (Table 4). Fatty acid Figure 1. Relationships between dietary concentrations of roasted soybeans (RSB) and (A) linoleic acid (18:2) and (B) linolenic acid (18:3) concentrations in milk fat (n = 80). The open circles represent data from this experiment that were used to generate the regression lines (linoleic acid = X; linolenic acid = x). Solid triangles represent data from the literature (Chouinard et al., 1997; Dhiman et al., 1999; Tice et al., 1994; Timmons, 1999) and were not used in the regression analysis.

6 SOYBEANS AND OXIDIZED FLAVOR OF MILK 2445 Figure 2. Relationships between dietary concentrations of roasted soybeans (RSB) and concentrations of linoleic acid (18:2, open circles) and linolenic acid (18:3, open triangles) in the TMR as a percentage of DM (A) and as a percentage of total fatty acids (FA) in the TMR (B). Concentrations of 18:3 in the DM or FA were not significantly (P > 0.15) related to RSB concentrations. Concentration of 18:2 in DM = RSB (P < 0.01; r 2 = 0.53; SE slope = 0.013). Concentration of 18:2 in FA = RSB (P < 0.01; r 2 = 0.25; SE slope = 0.17). composition, however, was constant during the 8-d storage period (data not shown). Antioxidants were measured in milk only on d 0 and 3 and were lower on d 3 (Table 4). The largest decrease (P < 0.05) was found for AA (70% decrease by d 3); α-tocopherol and β-carotene decreased by 21 and 26% over 3 d of storage. The concentration of DHAA, an oxidation product of AA, did not change over 3 d but made up a much larger proportion (P < 0.05) of the AA plus DHAA on d 3 (42%) than on d 0 (19%). Because DHAA did not accumulate as AA decreased, DHAA was oxidized to other compounds during storage. On a molar basis, milk (d 0) contained approximately 90 times more AA than α- tocopherol. However, AA is water soluble and α-to- copherol is fat soluble. The molar concentration, relative to the size of the appropriate phase (water or fat), was only four times higher for AA than α-tocopherol. After 3 d of storage, milk contained only about six times more AA than α-tocopherol (on a molar basis), and when molar ratio was expressed relative to the size of the appropriate phase, AA and α-tocopherol were essentially equal after 3dofstorage. Antioxidants prevent or terminate the oxidative chain reaction initiated by pro-oxidants such as Cu (Hailliwell and Gutteridge, 1995; Richardson and Korycka-Dahl, 1983). α-tocopherol was probably oxidized while protecting PUFA from lipid oxidation (Richardson and Korycka-Dahl, 1983) by reacting with lipid Table 4. Changes (averaged across all herds and sampling times) in concentrations of vitamins and flavor score of milk during storage (4 C). 1 Storage time, d SEM α-tocopherol, mg/l ** ND β-carotene, mg/l * ND Ascorbic acid (AA), mg/l ** ND 0.3 Dehydroascorbic acid (DHAA), mg/l ND 0.3 AA + DHAA, mg/l *** ND 0.4 DHAA, % of total AA *** ND 2.5 Flavor score *, **, ***Concentrations on d 3 differ from d 0 at P < 0.10, P < 0.05, and P < 0.01, respectively. 1 n = 80 for tocopherol and carotene; n = 66 for AA and DHAA; n = 79 for flavor score. 2 Not determined. 3 Flavor score: 0 = normal taste, 5 = intense oxidized flavor. Linear increase (P < 0.05) with time.

7 2446 TIMMONS ET AL. Table 5. Pearson correlation coefficients between milk flavor and milk and dietary constituents. 1 Variable 2 C16:0 C16:1 C18:2 C18:3 PUFA PI MUN RSB Flavor d 3 Flavor d C16: NS C16: NS C18: C18: NS PUFA PI MUN NS 0.32 RSB Only variables that were correlated (P < 0.01) with milk flavor are shown; n = Flavor d 8 and flavor d 3 = flavor score for milk stored 8 or 3 d (0 = normal taste; 5 = intense oxidized flavor); C16:0, C16:1, C18:2, and C18:3 are milk fatty acids (number of carbons:number of double bonds); PUFA = polyunsaturated fatty acids (18:2 + 18:3 + conjugated linoleic acid); PI = polyunsaturated index = 18:2 + (18:3 2); MUN = milk urea N; and RSB = roasted soybeans (% of dietary DM). free radicals, generating a tocopherol radical. The tocopherol radical might be regenerated by AA, producing tocopherol and DHAA at the lipid-water interface (Niki, 1987). This could explain the more rapid decrease in AA than α-tocopherol over storage time. The decrease in AA over time suggests that AA was not regenerated by reducing systems, such as glutathione, NADPH, and NADH, quickly enough to prevent the nonreversible reduction to 2,3-diketogulonic acid (Levine and Morita, 1985). β-carotene is a scavenger of peroxyl radicals and singlet oxygen in oxidation systems (Krinsky, 1989) and may help prevent lipid oxidation; however, concentrations were extremely low in milk (<10% of α-tocopherol on a molar basis). Milk Flavor On d 0, most (92%) milk samples had a flavor score of 1, and all samples had a score of <1.5. After 3 d of storage, 37% of the samples had a score of 1, 86% of samples had a score of 2, and only 2% of samples had a score of 3. At 8 d of storage, 13% of samples had a score of 1, 51% of samples had a score of 2, and 17% of the samples had a score of 3. Adequate variation in flavor score among d-8 samples existed to determined statistical relationships among milk flavor and milk composition. All relationships described below are for d-8 flavor score and d-0 milk composition data. Day 0, rather than d 8, composition data were used to determine what factors in fresh milk are related to the subsequent development of SOF. All variables that were correlated (P < 0.01) with milk flavor are shown in Table 5. Flavor score on d 3 was correlated with flavor score on d 8. Correlations between milk components and flavor score on d 8 were higher than those for d-3 scores probably because of the wider range in scores on d 8. No single variable accounted for more than 30% of the variation in d-8 flavor score. Increased d-8 flavor score (higher SOF) was related most strongly (ca. r = 0.5) with increased concentrations of 18:2, 18:3, and total PUFA in milk fat and the polyunsaturated index. A weaker positive correlation was found between flavor score and milk urea N. Concentrations in milk fat of total mediumchain fatty acids (data not shown) and 16 C fatty acids were negatively correlated with milk flavor. Those correlations were most likely caused by the negative correlation between concentrations of those fatty acids and 18:2 and 18:3. The concentration of dietary RSB were correlated positively with d-8 flavor score (0.38). The strength of that correlation was as expected, based on the correlation between RSB and 18:2 and 18:3 in milk and the correlation between 18:2 and 18:3 and milk flavor. The relationships between concentrations of 18:3 and 18:2 in milk fat and the development of SOF have been reported previously (Barrefors et al., 1995; Granelli et al., 1998; Smith et al., 1963). Linolenic acid (18:3) is less stable than linoleic (18:2) or more saturated fatty acids because two pentadienyl groups are more susceptible to extraction of an electron by free radicals, resulting in formation of a lipid free radical (Benzie, 1996). Multiple regression was used to quantify relationships between milk composition data and flavor score. All variables were included initially in the stepwise regression except those variables that were strongly correlated with each other. Only a single milk fatty acid or a logical grouping of fatty acids (e.g., total PUFA) was used in each regression because of colinearity among many of the fatty acids (e.g., 16:0 and 18:2). Although milk urea N was positively correlated with milk flavor score, it was not included in the regression models because it was correlated with 18:2, total PUFA, and polyunsaturated index. Because of the correlation between RSB and milk fatty acids, RSB was not included in models containing milk fatty acids.

8 SOYBEANS AND OXIDIZED FLAVOR OF MILK 2447 Interaction terms were included only if a chemical reason existed for the interaction to affect milk flavor. For example, interactions between prooxidants and unsaturated fatty acids (oxidative substrate) and interactions between antioxidants and oxidative substrates were tested. Only interactions between substrates and Cu were significant (P < 0.05). The five best fitting models (Table 6) included 18:2, 18:3, or total PUFA concentrations in milk fat, polyunsaturated index, or dietary RSB individually in each separate model. In addition, all models included DHAA and Cu concentrations and an interaction term between Cu and milk fatty acid measure or RSB. The multiple regression models had higher coefficients of determination (R 2 ) and lower root mean square error (RMSE) terms than did the linear regression equations, but the best multiple regression model still accounted for only about 50% of the variation in flavor score (Table 6). Based on the coefficient of determination and RMSE, the models that included 18:2, 18:3, total PUFA, or the polyunsaturated index were similar in their ability to estimate flavor score (R 2 = 0.40 to 0.46; and RMSE = 0.65 to 0.67); the model that included RSB had a poorer fit (R 2 = 0.31; and RMSE = 0.74). All models included DHAA with a positive coefficient of about 0.1 (i.e., flavor score would increase about 0.1 units for every 1 mg/dl increase in DHAA in milk at d 0). The effect of DHAA is probably not an indirect effect of AA, because AA and DHAA were not correlated. Milk that had higher initial concentrations of DHAA (d 0 sample) may have contained higher concentrations of unmeasured water-soluble oxidants that caused conversion of AA to DHAA. Also, possibly DHAA was converted into other products that led to the development of SOF. The effect of DHAA on milk flavor warrants additional research. Both Cu and measures of unsaturated fatty acids in milk had negative coefficients (Table 6), which suggests that increased concentrations of those would improve milk flavor. However, those coefficients cannot be examined in the absence of the interaction term. The interaction term implies that off-flavor is not likely to develop in milk with high concentrations of PUFA when milk Cu concentrations are low. Similarly, high milk Cu did not lead to SOF unless the concentration of unsaturated fatty acids also was high. Oxidation in milk is a function of the concentrations of antioxidants, prooxidants, and oxidizable substrate. Copper is a prooxidant and unsaturated fatty acids are oxidizable substrates. Milk Cu concentrations are related with development of SOF in milk (Granelli et al., 1998; Haase and Dunkley, 1970; King and Dunkley, 1958). Copper is primarily bound to the casein; however, Cu migrates to the fat phase when milk is heated (Samuelsson, 1967), as with pasteurization. The accumulation of reduced copper in close association with unsaturated fatty acids in the milk fat globule membrane should increase the development of SOF. In our study, dietary Cu was not correlated (P > 0.8) with milk Cu concentrations. However, when cows were fed diets with high concentrations of Cu (80 mg/kg of DM), milk Cu concentrations were higher than when cows were fed diets with more typical (11 mg/kg of DM) Cu concentrations (Dunkley et al., 1968). Copper concentration in milk is more likely increased by mineral buildup from wash water in milking systems (Nicholson and Charmley, 1991). The suppression of SOF development is dependent on a continuous supply of antioxidants to scavenge free radicals and terminate the oxidation reaction; however, concentrations of several antioxidants (i.e., α-tocopherol, β-carotene, and AA) at d 0 were not corre- Table 6. Significant (P < 0.01) multiple regression models describing the relationship between milk flavor (8 d of storage) and milk concentrations of dehydroascorbic acid (DHAA), Cu, and certain fatty acids (or dietary roasted soybeans). 1 Fat Model Type 2 b 1 SE Cu SE Cu*Fat SE DHAA SE b 0 SE RMSE 3 R 2 1 RSB : : PUFA PI n = 66. Flavor score: 0 = normal, 5 = strong oxidized flavor. Copper and DHAA expressed as mg/l. 2 RSB = roasted soybeans (% of dietary DM); 18:2 = linoleic acid (% of total milk fatty acids); 18:3 = linolenic acid (% of total milk fatty acids); PUFA = polyunsaturated fatty acids (18:2 + 18:3 + conjugated linoleic acid); and PI = polyunsaturated index = 18:2 + (18:3 2). 3 Root mean standard error.

9 2448 TIMMONS ET AL. lated with milk flavor score on d 8. The lack of a relationship between β-carotene was expected because concentrations in milk were very low. High concentrations of milk α-tocopherol have been related to reduced SOF (Barrefors et al., 1995; Charmley et al., 1993; St-Laurent et al., 1990). Generally, several thousand international units of supplemental vitamin E must be fed to reduce SOF in milk (Nicholson et al., 1991). Total vitamin E intake (supplemental and basal) by herds on this study averaged about 1000 IU/d. The concentration of α-tocopherol in d-0 milk in our study may not have been high enough to impact d-8 flavor score. α-tocopherol concentrations in milk decreased with storage (Table 4), implying that oxidation was being quenched by α-tocopherol. After 8 d of storage, the amount of α-tocopherol may not have been adequate to prevent additional oxidation. This premise is supported by the negative correlation between the concentration of α-tocopherol in d-0 milk with flavor score on d3( 0.27; P < 0.01). CONCLUSIONS The development of SOF in milk was related to high concentrations of PUFA in the milk fat (needed to propagate the oxidative chain reaction) and Cu (prooxidant that initiates oxidation). Feeding diets with a high concentration of RSB increased the PUFA in milk fat and SOF in milk. Equations were developed that accurately predicted concentrations of 18:2 and 18:3 in milk fat when RSB were fed. Those equations can be useful to estimate the potential for development of SOF in milk when milk Cu concentrations are known. These data suggest that feeding RSB or other fat sources that increase the concentrations of 18:2 and 18:3 in milk fat should be limited when milk contains high concentrations of Cu to prevent development of SOF and consumer rejection of dairy products. ACKNOWLEDGMENTS This study was supported in part by Dairy Management Incorporated. Special thanks to the 20 farms involved with the study, the technical assistance of S. Wagmiller, D. Wyatt, and S. Sohn, and to Reiter Dairy, Inc., Dean Foods, and area nutritionists for help in the selection of herds and provision of ration information. REFERENCES Association of Official Analytical Chemists Official Methods of Analysis. Vol. 1, 15th ed. AOAC, Arlington, VA. Barrefors, P., K. Granelli, L. A. Appelqvist, and L. Bjöerck Chemical characterization of raw milk samples with and without oxidative off-flavor. J. Dairy Sci. 78: Benzie, I. F. F Lipid peroxidation: A review of causes, consequences, measurements and dietary influences. Int. J. Food Sci. Nutr. 47: Bruhn, J. C., A. A. Franke, and G. S. Goble Factors relating to development of spontaneous oxidized flavor in raw milk. J. Dairy Sci. 59: Charmley, E., and J. W. G. Nicholson Influence of dietary fat source on oxidative stability and fatty acid composition of milk from cows receiving a low or high level of dietary vitamin E. Can. J. Anim. Sci. 74: Charmley, E., J. W. G. Nicholson, and J. A. Zu Effect of supplemental vitamin E and selenium in the diet on vitamin E and selenium levels and control of oxidized flavor in milk. Can. J. Anim. Sci. 73: Chouinard, P. Y., V. Girard, and G. J. Brisson Performance and profiles of milk fatty acids of cows fed full fat, heat-treated soybeans using various processing methods. J. Dairy Sci. 80: Dhiman, T. R., E. D. Helmink, D. J. McMahon, R. L. Fife, and M. W. Pariza Conjugated linoleic acid content of milk and cheese from cows fed extruded oilseeds. J. Dairy Sci. 82: Dunkley, W. L., and A. A. Franke Evaluating susceptibility of milk to oxidized flavor. J. Dairy Sci. 50:1 9. Dunkley, W. L., A. A. Franke, J. Robb, and M. Ronning Influence of dietary copper and ethylenediaminetetraacetate on cooper concentration and oxidative stability of milk. J. Dairy Sci. 51: Frankel, E. N Review: Recent advances in lipid oxidation. J. Sci. Food Agric. 54: Granelli, K., L. Bjöerck, and L. A. Appleqvist The variation of superoxide dismutase (SOD) and xanthine oxidase (XO) activities in milk using an improved method to quantitate SOD activity. J. Sci. Food Agric. 67: Granelli, K., P. Barrefors, L. Bjöerck, and L. A. Appleqvist Further studies on lipid composition of bovine milk in relation to spontaneous oxidized flavor. J. Sci. Food Agric. 77: Haase, G., and W. L. Dunkley Copper in milk and its role in catalyzing the development of oxidized flavor. Milchwissenschaft 25: Halliwell, B., and J. M. C. Gutteridge The definition and measurement of antioxidants in biological systems. Free Radic. Biol. Med. 18: Indyk, H. E Simplified saponification procedure for the routine determination of total vitamin E in dairy products, foods, and tissues by high-performance liquid chromatography. Analyst 113: King, R. L., and W. L. Dunkley Relation of natural copper in milk to incidence of spontaneous oxidized flavor. J. Dairy Sci. 42: Krinsky, N. I Antioxidant functions of carotenoids. Free Radic. Biol. Med. 7: Levine, M., and K. Morita Ascorbic acid in endocrine systems. Vitam. Horm. 42:1 64. McMurry, C. H., W. J. Blanchflower, and D. A. Rice Influence of extraction techniques on determination of α-tocopherol in animal feedstuffs. J. Assoc. Off. Anal. Chem. 63: Morales, M. S., D. L. Palmquist, and W. P. Weiss Effects of fat source and copper on unsaturation of blood and milk triacylglycerol fatty acids in Holstein and Jersey cows. J. Dairy Sci. 83: Nicholson, J. W. G., and E. Charmley Oxidized flavour in milk: A Canadian perspective. IDF Bull. 257: Nicholson, J. W. G., A. M. St-Laurent, R. E. McQueen, and E. Charmley The effect of feeding organically bound selenium and α-tocopherol to dairy cows on susceptibility of milk to oxidation. Can. J. Anim. Sci. 71: Niki, E Antioxidants in relation to lipid peroxidation. Chem. Phys. Lipids 44: Richardson, T., and M. Korycka-Dahl Lipid oxidation. Pages in Developments in Dairy Chemistry 2. P. F. Fox, ed. Appl. Sci., London, New York.

10 SOYBEANS AND OXIDIZED FLAVOR OF MILK 2449 Samuelsson, E. G The distribution of copper to the fat globules of milk. Report no. 76, Milk and Dairy Research. Swedish Univ. Agric. Sci., Alnarp, Sweden. Sapers, G. M., F. W. Douglas, Jr., M. A. Ziolkowski, R. L. Miller, and K. B. Hicks Determination of ascorbic acid, dehydroascorbic acid, and ascorbic acid-2-phosphate in filtered apple and potato tissue by high performance liquid chromatography. J. Chromatogr. 503: SAS/STAT User s Guide, Release SAS Inst., Inc., Cary, NC. Sidhu, G. S., M. A. Brown, and A. R. Johnson Autoxidation in milk rich in linoleic acid. I. An objective method for measuring autoxidation and evaluating antioxidants. J. Dairy Res. 42: Smith, L. M., W. L. Smith, and M. Ronning Influence of linoleic acid content of milk lipids on oxidation of milk and milk fat. J. Dairy Sci. 46:7 13. St-Laurent, A. M., M. Hidiroglou, M. Snoddon, and J. W. G. Nicholson Effect of α-tocopherol supplementation to dairy cows on milk and plasma α-tocopherol concentrations and on spontaneous oxidized flavor in milk. Can. J. Anim. Sci. 70: Sukhija, P. S., and D. L. Palmquist Rapid method for determination of total fatty acid content and composition of feedstuffs and feces. J. Agric. Food Chem. 36: Tice, E. M., M. L. Eastridge, and J. L. Firkins Raw soybeans and roasted soybeans of different particle sizes. 2. Fatty acid utilization by lactating cows. J. Dairy Sci. 77: Timmons, J. S Contribution of dietary roasted soybeans and milk components to the development of spontaneously oxidized flavor. M.S. Thesis. Ohio State Univ., Columbus, OH. Van Soest, P. J., J. B. Robertson, and B. A. Lewis Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74: