Flavors and Off-flavors in Dairy Foodsq

R Marsili, Rockford College, Rockford, IL, USA

Ó 2016 Elsevier Inc. All rights reserved.

Desirable Flavor Compounds in Milk 1

Flavor Development in Cultured Dairy Products 3
Buttermilk and Sour Cream 3
Yogurt 3
Common Causes of OFs in Milk and Dairy Products 5
Light-Induced OF 5
Oxidized OFs from Contact with Prooxidant Metals 7
Microbial OFs 8
Feed-Related OFs 9
Packaging-Related OFs 10
Techniques for Analyzing Flavors and OFs 11
Electronic Nose Applications for Measuring OFs in Milk 13
OFs Generated in Dry Dairy Ingredients 14
Problem with 2,4,5-Trimethyloxazole in Spray-Dried Dairy Products 14
Problem with Musty Trihaloanisole Malodors in Casein Powders 14
OFs from Catabolic Reactions of Aromatic Amino Acids 16
Conclusion 18
Further Reading 19

Desirable Flavor Compounds in Milk

Milk is such a bland-tasting food that consumers have a difficult time describing its taste attributes. Indeed, dairy researchers have
devoted far more time and effort studying chemicals responsible for off-flavors (OFs) in milk than in understanding what chemicals
are most important to good flavor quality.
Although processed milk of good flavor quality has a bland taste, it does, nonetheless, have a characteristic flavor, described by
some as a slight salty-sweet taste. To some degree, this flavor quality can probably be attributed to milk salts and lactose, but the
reason for milk’s characteristic taste is far more complicated than this.
The sensory perception of milk is significantly impacted by its pleasant mouthfeel, which is due primarily to the emulsion of fat
globules in an aqueous colloidal protein phase, but the volatile constituents in milk make the strongest contribution to its aroma
and flavor. Dimethyl sulfide was one of the first important aroma-contributing chemicals identified in milk (in 1956). Significant
improvements in analytical techniques and instrumentation have occurred since then, and dozens of other odoriferous chemicals
have been identified in milk, including aldehydes, ketones, alcohols, fatty acids, lactones, esters, sulfur compounds, nitrogen
compounds, and aromatic hydrocarbons. Complicating the understanding of what odorants are most important to good milk fla-
vor are three factors: (1) the large number of volatiles present in milk, (2) their low concentration levels, and (3) in some cases, their
low odor-detection threshold levels.
Recently, researchers identified 80 neutral volatiles in raw milks from different species using vacuum distillation and liquid–liquid
extraction followed by high-resolution gas chromatography (HRGC). In studying the flavor constituents of bovine, ovine, caprine, and
water buffalo fresh raw milks, these researchers found similar volatiles in milk from the four species. Different levels of odorants from
species to species were probably responsible for the different odors and taste of the milk. Dimethylsulfone alone comprised approx-
imately 25% of the volatile components in bovine, caprine, and ovine milk, but only 4% in buffalo milk. Pentanal and nonanal were
the prevalent aldehydes, while 3-methylbutanal was found only in buffalo milk, and phenylacetaldehyde and benzaldehyde were
present in large quantities in caprine milk but only in trace amounts in the other milks. The concentrations of ketones, predominantly
2-methyl ketones, were higher in buffalo milk than in the other three types. Phenylethanol, a component with a floral-rose odor, was
not found in ewesʼ and goatsʼ milk; its concentration in buffalo milk was 100 times higher than in bovine milk. The level of 1-octene-3-
ol, a compound with a metallic note, was highest in buffalo milk, followed in order by ewe, goat, and cow milks.
Although a significant number of volatiles were observed, results of this work did not identify which compounds were actually
the most important to good milk flavor. As an important follow-up, the researchers added an olfactometry technique to their
HRGC–MS (mass spec-trometry) method. To determine which of the 80 volatiles observed in milk samples contributed odor notes,
individual chromatographic peaks were sniffed as they eluted from the GC column. With this approach, 14 different odorants were

q
Change History: July 2015. R. Marsili added a new section on Off-Flavors from Catabolic Reactions of Aromatic Amino Acids and added Table 6.
Update of R. Marsili. Flavors and Off-Flavors in Dairy Foods. Encyclopedia of Dairy Sciences, 2nd Edition, 2011, Pages 533–551.

Reference Module in Food Sciences http://dx.doi.org/10.1016/B978-0-08-100596-5.00784-8 1

2 Flavors and Off-flavors in Dairy Foods

Table 1 Volatiles most commonly present in raw and processed milks of normal, good flavor quality

Volatilea Possible origin Comments

Acetone Feed; bovine ketosis Volatile present in highest concentration

2-Butanone Feed Volatile usually present at second highest concentration
Ethyl butanoate, Microbial metabolites (ethanol þ fatty acid / ester) Significantly more in raw milk than in pasteurized milk;
ethyl hexanoate most potent odorants in raw milk
Dimethyl sulfide Thermal decomposition of methionine and cystine; Present in trace amounts; higher levels in pasteurized milk
microbial metabolite compared with raw milk
Toluene Degradation of b-carotene Trace amounts; packaging materials can also contribute
toluene
Limonene Forages Trace amounts
Methyl ketones (C5–C12) Microbial metabolites; thermal degradation of b-keto acids Trace amounts
naturally found in milk
3-Methylbutanal In raw milk, because of microbial growth (Streptococcus Trace amounts; highest levels typically found in raw milk
lactis var. maltigenes); in heated milk as a result of
nonenzymatic browning reaction involving leucine
(Strecker degradation)
Chloroform Unknown (perhaps from chlorinated sanitizers used to Trace amounts; highest levels typically found in raw milk,
wash teats before milking) less in pasteurized milk
Fatty acids (C4–C10) Incomplete triacylglycerol synthesis in the mammary Trace amounts
gland; possibly also lipase activity
Xylene isomers, styrene Contaminants from packaging materials Trace amounts in processed milk; not normally present in
raw milk
Hexanal Autoxidation of linoleic acid Trace amounts
a
Except for ethyl butanoate and ethyl hexanoate in raw milk, all volatiles are at concentrations below the taste-odor detection threshold level.

identified. To determine which of these were ʻodor-impactʼ compounds, the researchers used the CharmAnalysis technique. In this
olfactometry method, stepwise dilutions (typically 1:2 or 1:3) of the prepared extract with a solvent are made, followed by evalu-
ation of each dilution by chromatographing and sniffing column effluents. Using CharmAnalysis, the odor-impact volatiles in the
four types of milk were determined. For raw bovine milk, the two most important odor-impact chemicals were ethyl butanoate and
ethyl hexanoate.
It is interesting to note that significantly higher levels of ethyl butanoate and ethyl hexanoate are detected consistently in raw milk
than in pasteurized milk. In fact, these esters are frequently not detected at all in pasteurized milk when analyzed by purge-and-trap
gas chromatography or other GC-based techniques capable of mg kg
1 detection levels. The majority of these esters, which apparently
are the primary odor-impact chemicals in raw bovine milk, are lost or destroyed during pasteurization and processing.
Other popular gas chromatograpy–olfactometry (GC–O) techniques include aroma extraction–dilution analysis (a dilution
technique) and Osme, a cross-modal matching technique. GC–O techniques are invaluable for identification of potent odorants
in dairy products, including desirable odorants as well as those that contribute OF. (Further discussion of GC–O is beyond the scope
of this article.)
Table 1 lists some of the volatiles most commonly present in normal-tasting raw and pasteurized milk, and describes their
possible origin. Figure 1 is a GC–MS chromatogram of fresh, pasteurized milk (2% milk fat content) stored in a plastic (high-
density polyethylene) bottle. This particular sample had a normal flavor profile, free of. Some of the components, such as hexanal

Figure 1 A GC–MS chromatogram of fresh, pasteurized milk (2% milk fat) of normal, good-tasting flavor quality stored in a plastic (high-density
polyethylene) bottle. Extraction of volatiles performed by SPME using a 75 mm Carboxen-PDMS fiber. Peak identities are as follows: (1) acetone;
(2) dimethyl sulfide; (3) 2-butanone; (4) chloroform; (5) 3-methyl butanal; (6) pentanal; (7) ethyl butyrate; (8) toluene; (9) hexanal; (10) and
(11) xylene isomers; (12) 2-heptanone; (13) styrene; (14) limonene; (15) 2-nonanone; (B) blank, chemical from GC septum.
 Flavors and Off-flavors in Dairy Foods 3

and styrene, have been known to cause OF in milk, but in this particular sample these compounds were present at concentrations
below their flavor-odor threshold detection level.

Flavor Development in Cultured Dairy Products

Modification of milk components by starter culture enzymes is responsible for the flavor of cultured dairy products. Lactic acid
fermentation is critical to the characteristic flavor development of cultured dairy products, including buttermilk, sour cream, and
yogurt.

Buttermilk and Sour Cream

Sour cream and buttermilk depend for their flavor on the combined action of strains of Lactococcus lactis subsp. lactis and Lc. lactis
subsp. cremoris to produce lactic acid and on Leuconostoc mesenteroides subsp. cremoris and citrate-positive (Citþ) strains of Lc. lactis
for production of aroma chemicals (especially diacetyl). In recent years, some dairies have stopped using Citþ Lc. lactis subsp.
lactis biovar diacetylactis because it generates a rather harsh flavor profile.
Lactic acid imparts a pleasant acid flavor. Excessive lactic acid from over-ripening or improper incubation temperature (or
delayed cooling/under-cooling) makes the finished product taste acrid and gives it a coarse mouthfeel. Furthermore, excessive
acidity masks the delicate flavor notes contributed by other important aroma chemicals. Only about 18% of the naturally available
lactose in milk is fermented to lactic acid. The desired lactic acid level in good-flavored buttermilk is 0.75–0.85%.
Trace amounts of acetic, formic, propionic, and/or pyruvic acids also are formed in secondary metabolic reactions in cultured
buttermilk and sour cream; these relatively volatile acids impart a pleasant acid flavor. The critical ʻbutteryʼ nutlike flavor character-
istic of most cultured dairy products is attributed to diacetyl, a potent fragrance compound.
Diacetyl is derived from the fermentation of citric acid by Ln. mesenteroides subsp. cremoris and Citþ Lc. lactis. The pathway for the
metabolism of citrate has been elucidated (Figure 2). High-performance liquid chromatography (HPLC) has been used to monitor
changes in organic acids during buttermilk fermentation (Figure 3), while static headspace gas chromatography has been used to
monitor volatile organics (Figure 4).
Chemical changes that occur during storage of buttermilk at refrigerated temperatures are shown in Table 2. The increase in acet-
aldehyde and the decrease in diacetyl are perhaps the most noteworthy changes. Loss of diacetyl is a result of its conversion (enzy-
matic reduction) to acetoin by diacetyl reductase (from Ln. mesenteroides subsp. cremoris). The increase in acetaldehyde and the
decrease in diacetyl can result in an OF that is commonly described as the ʻgreen flavor defectʼ.

Yogurt
Most commercial yogurt is produced by the action of two dissimilar organisms, Streptococcus thermophilus and Lactobacillus delbrueckii
subsp. bulgaricus, which grow simultaneously to give the desired flavor, mouthfeel, and body to the yogurt. It is generally agreed that
lactic acid, acetic acid, and acetaldehyde are among the most important contributors to yogurt flavor and that the symbiotic action
of Sc. thermophilus and Lb. delbrueckii subsp. bulgaricus is necessary to obtain the desired flavor profile and acid balance. However,
other secondary metabolites contribute significantly to the delicate flavor nuances of natural yogurt.

Figure 2 Pathway for the conversion of citrate to diacetyl and other compounds. The reactions are catalyzed by the following enzymes: (1) citrate
lyase; (2) oxaloacetate synthase; (3) pyruvate decarboxylase; (4) acetolactate synthase; (5) acetolactate decarboxylase; (6) diacetyl reductase;
(7) acetoin reductase. Reaction 8 is the oxidative conversion of a-acetolactate to diacetyl. Adapted from Jenness, R., Patton, S., 1959. Principles
of Dairy Chemistry. John Wiley, New York.
4 Flavors and Off-flavors in Dairy Foods

Figure 3 Changes in the concentration of organic acids in cultured buttermilk during a typical fermentation, as determined by HPLC. Reproduced
with permission from Marsili, R.T., 1981. Monitoring bacterial metabolites in cultured buttermilk by high-performance liquid chromatography and gas
chromatography. J. Chromatogr. Sci. 19, 451–456.

Figure 4 Changes in the concentration of volatile organic compounds in cultured buttermilk during a typical fermentation as determined by head-
space GC. Reproduced with permission from Marsili, R.T., 1981. Monitoring bacterial metabolites in cultured buttermilk by high-performance liquid
chromatography and gas chromatography. J. Chromatogr. Sci. 19, 451–456.
 Flavors and Off-flavors in Dairy Foods 5

Table 2 Changes in concentration of organic acids (high-performance liquid chromatography)

and volatile organic compounds (headspace gas chromatography) in buttermilk made from skim
milk, at end of fermentation, and after an additional week of refrigerated storage

Concentration (mgg
1)
Chemical 0 h at 23  C 18 h at 23  C 168 h at 3  C

Orotic acid 79.1 42.2 34.7

Citric acid 730 176 20
Pyruvic acid 1.3 33 <0.2
Lactic acid 45 7410 5820
Uric acid 20.4 18.8 19.7
Acetic acid <5 610 730
Propionic acid <10 90 40
Acetaldehyde 0.02 0.77 1.46
Acetone 2.10 1.98 2.14
Ethanol 1.00 24.2 37.3
Diacetyl <0.05 11.5 1.90

Reproduced with permission from Marsili, R.T., 1981. Monitoring bacterial metabolites in cultured buttermilk by high-
performance liquid chromatography and gas chromatography. J. Chromatogr. Sci. 19, 451–456.

Dozens of flavor compounds have been identified in yogurt. Headspace and simultaneous distillation–extraction have been
commonly used for the analysis of yogurt flavor volatiles. Because of the low intensity of yogurt odor, much of the research per-
formed in the past involved heating samples to increase the volatility of flavor components and improve the sensitivity of the analyt-
ical test. Heating samples, however, can alter the composition of sensitive aroma compounds and lead to artifact peaks that can be
misinterpreted as being potent odor-flavor contributors that occur naturally in the product. Therefore, much of this work may be
invalid.
Recently, a variety of low-temperature headspace procedures have been used to extract volatiles from yogurt. In one low-
temperature headspace technique, aroma volatiles were allowed to equilibrate between the matrix and the sample headspace at
a controlled temperature (30  C) before being swept from the headspace into a Tenax trap. The volatiles were then thermally des-
orbed from the Tenax, cryrofocused, and finally injected into a GC–MS equipped with an olfactometry detector.
One unique aspect of this work was the olfactometry method used to detect odor-impact chemicals. This GC–O technique is
referred to as the detection frequency method and is neither a dilution nor a cross-modal matching technique. Only one dilution
level was used, but GC–O experiments are repeated several times (e.g., 8–10). Aromagrams of individuals were then ʻaveragedʼ to
eliminate reproducibility problems. Resulting aromagrams permitted odor profile comparisons as peak intensities were related to
the frequencies of odor detection.
Using this technique, researchers detected some 91 different volatiles in the yogurt headspace, with 21 of these exhibiting
significant odor impact. These odor-impact chemicals are shown in Table 3. Five compounds – 1-nonen-3-one, methional,
2-methyltetra-hydrothiophen-3-one, 2(E)-nonenal, and guaiacol – with intense odor were reported for the first time in yogurt fla-
vor. Identification of 1-nonen-3-one is especially significant. This component appears to be one of the most potent flavorings iden-
tified to date. Its low odor threshold justifies its impact role in yogurt despite its low concentration.

Common Causes of OFs in Milk and Dairy Products

Light-Induced OF
Light-induced OF in milk is common and of major concern to the dairy industry. It has been estimated that exposure of milk in
high-density polyethylene bottles to fluorescent lights in supermarket dairy cases is responsible for the development of light-
induced OF in some 80% of samples sold in US supermarkets.
Light-induced OFs have two distinct components. Initially a burnt, activated sunlight flavor develops and predominates for
approximately 2 or 3 days. Degradation of sulfur-containing amino acids of the serum (whey) proteins is probably responsible
for this reaction. The exact reaction products responsible for this so-called light-activated flavor (LAF) have not been elucidated
clearly. Methional (3-(methylthio)propanal), however, has been implicated as a possible contributor. One mechanism that has
been postulated involves the light-induced reaction of methionine and riboflavin, producing methional.
Understanding the true impact that methional has on LAF is difficult, because it is relatively unstable and breaks down into more
stable components, including mercaptans, sulfides and disulfides. Some researchers have suggested that methanethiol, dimethyl
disulfide, and dimethyl sulfide also contribute to LAF. A recent dynamic headspace multivariate analysis study of in milk showed
a correlation between increasing levels of dimethyl disulfide with increasing exposure time of milk to fluorescent light. In two recent
studies, when light-exposed samples with LAF were analyzed by dynamic headspace GC–MS, methional was not detected at measur-
able levels. This has inspired researchers to propose an alternative mechanism, one that does not involve methional, to explain the
6 Flavors and Off-flavors in Dairy Foods

Table 3 Compounds contributing to the aroma of yogurt

Retention indexa Compound Odor descriptors

716 Acetaldehyde Fresh, green, pungent

757 Dimethyl sulfide Milk, lactone-like, sulfury, warm
995 2,3-Butanedione Butter, diacetyl, vanilla
1082 2,3-Pentanedione Butter, vanilla, milk
1120 2-Methylthiophene Gasoline, plastic, styrene
1221 3-Methyl-2-butenal Metallic, aldehydic, herbaceous
1322 1-Octen-3-one Mushroom, earthy
1406 Dimethyl trisulfide Sulfury, hydrogen sulfide, fecal
1424 1-Nonen-3-one Mushroom, earthy
1462 Acetic acid Pungent, acidic, vinegar
1479 Methional Soup, cooked vegetables, pungent, sulfury
1551 (cis, cis)-Nonenal þ Green, leather, sulfury
2-methyltetrahydrothiophen-3-one
1680 2-Phenylacetaldehyde Flowery
1684 3-Methylbutyric acid Sweaty, cheese, soy sauce, flowery
1715 Unidentified Flowery, warm, caramel
1750 Unidentified Metallic
1882 Caproic acid Rancid, flowery
1896 Guaiacol Bacon, phenolic, smoked, spicy
2002 Benzothiazole Burnt, rubbery
2043 Unidentified Hydrocarbon, chemical, burnt rubber
a
Retention index on DB-Wax phase, using headspace injection method.
Reproduced with permission from Ott, A., Fay, L.B., Chaintreau, A., 1997. Determination and origin of the aroma impact compounds of
yoghurt flavor. J. Agric. Food Chem. 45, 850–858.

formation of dimethyl disulfide. According to this mechanism (Figure 5), dimethyl disulfide is formed by singlet oxygen oxidation
of methionine.
The second component of light-induced OF in milk is attributed to lipid oxidation. This OF, often characterized as metallic or
cardboardy, usually develops after 2 days and does not dissipate. Aldehydes (especially pentanal and hexanal) and, to a lesser
degree, ketones (e.g., 1-hexen-3-one and 1-nonen-3-one), alcohols, and hydrocarbons have been observed to form in milk as a result
of light-induced lipid oxidation reactions. The unsaturated aldehydes and ketones have the lowest sensory thresholds and are
usually considered the primary sources of oxidized OF. When milk is exposed to light, various carbonyl compounds form from
the reaction of light and oxygen with unsaturated fatty acids in the milk fat triacylglycerols and other milk fat components. Autoox-
idation of unsaturated fatty acids involves a free radical reaction, forming fatty acid hydroperoxides that degrade to various
malodorous compounds, for example, hexanal, the predominant lipid reaction byproduct in light-exposed milk, in the case of lino-
leic acid.

Figure 5 Postulated mechanism for the formation of dimethyl disulfide by singlet oxygen oxidation of methionine. Adapted from Jung, M.Y.,
Yoon, S.H., Lee, H.O., Min, D.B., 1998. Singlet oxygen and ascorbic acid effects on dimethyl disulfide and off-flavor in skim milk exposed to light.
J. Food Sci. 63, 408–412.
 Flavors and Off-flavors in Dairy Foods 7

(a) 100 000 CH3(CH2)4CHO

80 000 CH3SSCH3

Total ion count

60 000
40 000
20 000
0
200 400 600 800 1000 1200 1400 1600
(b)
100 000
80 000

Total ion count

60 000
40 000
20 000
0
200 400 600 800 1000 1200 1400 1600
Retention time (s)

Figure 6 Chromatograms (obtained by purge-and-trap GC–MS) of (a) vanilla ice cream with severe putrid flavor defect caused by light-induced
production of dimethyl disulfide, and (b) control vanilla ice cream with no off-flavor.

Other dairy products besides milk have been known to develop significant LAF OF. One example is a vanilla ice cream sample
that developed a strong putrid flavor defect (Figure 6). Several samples developed the flavor defect, which was concentrated on the
surface of the ice cream nearest the lid; several different production runs developed the defect. The ice cream was packaged in a round
1.5 gallon paperboard carton fitted with a clear plastic lid. Analysis of complaint samples by purge-and-trap GC–MS revealed that,
compared with control samples, the complaint samples had significant concentrations of dimethyl disulfide and hexanal. Records
showed that the flavor defect first became noticeable when samples were stored in a new warehouse. Careful inspection of the
freezer warehouse storage conditions revealed that samples were stored in proximity to high-intensity light. Lighting adjustments
were made, and the problem was easily and quickly remedied.

Oxidized OFs from Contact with Prooxidant Metals

The presence of pro-oxidant metals like copper, iron, and nickel can significantly accelerate the rate of lipid oxidation OF develop-
ment in milk and processed dairy products. This problem has been observed, for example, in sour cream (Figure 7). (Note that the
size of the aldehyde peaks C5–C8 in normal-tasting sour cream would be similar to the size of these peaks shown in the 2% milk
chromatogram in Figure 1.) In this example, a copper valve was used in the processing equipment. The high level of hexanal is an

Figure 7 Chromatogram (obtained by solid-phase microextraction GC–MS using a 75 mm Carboxen-PDMS fiber) of a sour cream sample severely
oxidized by contact with processing equipment containing a copper valve. Oxidation of unsaturated fatty acids in milk fat generates numerous
malodorous aldehydes. Peak identities are as follows: (1) pentanal; (2) hexanal; (3) heptanal; (4) octanal.
8 Flavors and Off-flavors in Dairy Foods

excellent indicator of lipid oxidation catalyzed by pro-oxidant metal contamination. One interesting aspect of this problem was that
dairy technologists were unable to recognize the OF as an oxidation OF. Although the dairy technologists were experts at distin-
guishing oxidation OF in milk, the strong culture flavor background of the sample altered the perception of the oxidized flavor
so dramatically that oxidation was not even considered as a potential cause of the OF. This example shows how valuable good
analytical support can be in resolving OF issues with dairy products.

Microbial OFs
Milk is an excellent medium for the growth of microorganisms. Malodorous chemicals produced as metabolites during the bacterial
growth phase are a common cause ofacid, malty, fruity, bitter, putrid, and unclean OF in dairy products.
Post-pasteurization contamination of milk with psychrotrophic bacteria can cause milk OF. Some characteristic OF notes in milk
are clearly associated with specific types of psychrotrophs. For example, contamination by Pseudomonas fragi, a psychrotrophic
Gramnegative organism, often causes ʻfruityʼ off-notes in milk. Pseudomonas fragi lipase and esterase hydrolyze short-chain fatty
acids from milk fat and convert the acids to ethyl esters by reaction with ethanol. Strains of Bacillus spp. have also been observed
to produce fruity OF in milk.
Another common psychrotroph associated with a particular flavor defect is Lc. lactis var. maltigenes. The malty aroma of milk
cultures of Lc. lactis var. maltigenes is caused by the production of 3-methylbutanal, which at concentrations as low as 0.5 mgl
1
in milk generates the characteristic malty flavor defect. The action of bacterial enzymes on leucine produces 3-methylbutanal.
Another organism capable of producing 3-methylbutanal is Lb. maltaromicus.
Other common metabolites include ethyl acetate, dimethyl sulfide, dimethyl disulfide, ethanol and other alcohols, methyl
ketones, C4–C10 free fatty acids, lactic acid pyruvic acid, and bitter peptides.
One problem that complicates understanding of the causes of in milk is that many of these chemicals can be produced by two or
more mechanisms. For example, the production of methyl ketones and secondary alcohols (Figure 8) and lactones (Figure 9) in
milk can result from microbial reaction pathways as well as from high-heat treatment. Dimethyl sulfide is generated (but dissipates
after a few hours) by pasteurization and by microbial spoilage. Dimethyl disulfide is produced when milk is irraditated with fluo-
rescent light, as well as by enzyme-induced reactions on milk proteins.
Lactones can impart an undesirable burnt, fruity, stale, and coconut-like OF to dairy products. Elevated levels of hydroxy acids
sometimes occur in milk when cows are fed stored forages rather than pasture. Hydroxyacids may be incorporated into the triacyl-
glycerols. These esters of hydroxy acids are relatively unstable and are readily hydrolyzed from the triacylglycerol in the presence of
water and heat. They then cyclize to form lactones, which can cause a stale flavor in milk.
As long as they are bound to the triacylglycerol, the hydroxyacids do not impart OF to milk. However, thermal processing frees
the acid, leading to the formation of undesirable-tasting lactones. Various microbial reactions can also generate lactones. For
example, yeasts such as Candida spp. and Saccharomyces cerevisiae and molds such as Penicillium notatum and Cladosporium butyri
are known sources.
Raw milk contaminated with psychrotrophic bacteria before pasteurization can develop OF and malodor problems, even
though most of the bacteria are killed by pasteurization. Although pasteurization destroys the bacteria, their heat-stable lipases,
proteases, and other enzymes sometimes survive, and the action of these extracellular enzymes on milk components can also
generate OF. These OF chemicals can be monitored by GC–MS (Figure 10). (See Figure 1 for comparison with chromatogram
of normal-tasting 2% processed milk.) To minimize the chances of this problem, raw milk should never be held for longer than
48 h in refrigerated silos before pasteurization.

Figure 8 Formation of methyl ketones and secondary alcohols in milk by microbial and thermally induced reaction mechanisms. (a) By action of
enzymes from fungi, molds and some types of bacteria on butterfat triacylglycerols; (b) by thermal degradation of b-keto acids in triacylglycerols.
Adapted from Jenness, R., Patton, S., 1959. Principles of Dairy Chemistry. John Wiley, New York.
 Flavors and Off-flavors in Dairy Foods 9

Figure 9 Formation of lactones in milk by microbial pathways and by thermally induced reactions. Adapted from Jenness, R., Patton, S., 1959.
Principles of Dairy Chemistry. John Wiley, New York.

Microorganisms can generate OF in unexpected and interesting ways. Potassium sorbate is often used as a fungistatic agent for
cultured dairy products and cheeses. It has been shown that a large number of molds in the genus Penicillium can grow in the pres-
ence of up to 7100 mg l
1 potassium sorbate. The occurrence of a plastic, paint, or kerosene OF in Feta cheese, cottage cheese, and
other dairy products has been attributed to the production of trans-1,3-pentadiene, a strong odorant produced from the enzymatic
decarboxylation reaction of sorbic acid (trans, trans-2,4-hexadienoic acid) (Figure 11).

Feed-Related OFs
What cows eat can profoundly impact the flavor of the milk they produce. Feeds that are known to cause OF in milk include fermented
musty silage (maize, legumes, and grass), alfalfa (green or hay), clover hay, brewersʼ grain, and green barley. The actual chemicals
responsible for the OF flavor have been identified in some cases. Freshly cut alfalfa hay contains high levels of trans-2-hexenal,
10 Flavors and Off-flavors in Dairy Foods

200 000 B 1 6 7
180 000
160 000
140 000 2

Total ion count

120 000
100 000
10 B
80 000 89
18
60 000 12
40 000 3 11 14 B
5 16
1315 19
20 000 4 17
0
100 200 300 400 500 600 700 800 900 1000 1100 1200
Retention time (s)

Figure 10 Off-flavor chemicals formed in processed milk by exogenous enzymes. Raw milk held for too long before pasteurization. Processed milk
made from this raw form developed elevated levels of numerous bacterial metabolites 4 days after processing, even though microbiological plate
counts were normal. Peak identities are as follows (* indicates chemical is a metabolite): (1) acetone; (2) methyl acetate*; (3) extraneous chemical
from SPME fiber; (4) 2-methyl-2-propenal*; (5) diacetyl*; (6) 2-butanone; (7) ethyl acetate*; (8) chloroform; (9) 3-methyl butanal*; (10) 2-
pentanone*; (11) 3-methyl pentane; (12) propyl acetate*; (13) 3-methyl-2-pentanone*; (14) isoamyl alcohol*; (15) propanoic acid*; (16) hexanal;
(17) 4,4-dimethyl heptane; (18) 2-heptanone*; (19) 2-nonanone*; (B) blank, chemical from GC septum.

Figure 11 Mechanism of formation of 1,3-pentadiene from sorbate. 1,3-Pentadiene imparts a kerosene, garlic-like off-flavor to dairy products at
low concentrations.

trans-3-hexenals, and trans-3-hexenols, which impart a green, grassy flavor to milk. Many types of weeds – especially wild onion, garlic,
and related plants – impart serious OF to milk when ingested by dairy cows (predominantly because of various sulfur-containing
organic compounds). (Note that there is another good reason to avoid feeding dairy cows musty, fermented silage: aflatoxin contam-
ination. Mold-contaminated feed runs the risk of transmitting aflatoxins B1, B2, G1, and G2 to the cow, which are converted to aflatoxin
M1 in the cow’s liver and transmitted in that form to the milk. Aflatoxins are among the most potent carcinogens known.)
The cow’s diet can affect flavor in unexpected, subtle ways. In one study, for example, dozens of milk samples from one
geographical region of the United States were analyzed by purge-and-trap GC–MS to determine the source of the severe OF char-
acterized as ‘oxidized’ by dairy technologists. Testing revealed unusually high concentrations of hexanal, suggesting autoxidation of
linoleic acid in milk fat. The OF problem had occurred during the late winter months for the previous 4 years but was getting
progressively worse each successive year. Because the level of hexanal was so high, copper contamination from processing lines
or from diet supplements was suspected initially as the source of activation catalysis. Previous studies on milk samples intentionally
abused by exposure to fluorescent light and contact with copper revealed that hexanal levels tend to be significantly higher when
pro-oxidant metals are involved than when milk is light-abused without added prooxidant metal.
Further testing of ʻbadʼ milk samples by atomic absorption spectrometry ruled out contamination by copper, nickel, or iron.
Another suspected contributing factor was lack of naturally occurring antioxidants (e.g., a-tocopherol) as a result of changes in
winter feed regimens. However, HPLC analysis showed that samples contained normal levels of a-tocopherol.
When fatty acid profiles of complaint and normal samples were compared, samples with intense oxidized OF consistently con-
tained 200–300% more linoleic acid than normal-tasting milk from different geographical areas. Linoleic acid (C18:2) is highly
susceptible to photoxidation.
The high linoleic acid content was traced to the cows’ diet. Dairy farmers in this region had been feeding cows increasingly high
levels of soya beans each year. Feeding high levels of soya beans has been shown to increase the linoleic acid content of milk fat and
make milk unusually susceptible to oxidation. This problem was corrected by regulating levels of unsaturated lipids in feed to
decrease the proportion of unsaturated fat in milk fat and/or by supplementing feed with an antioxidant like a-tocopherol.

Packaging-Related OFs
Ironically, packaging materials, which are designed to preserve the freshness and flavor of foods and beverages, can actually
be directly responsible for causing OF defects. Although plastic packaging material consists primarily of nonvolatile
 Flavors and Off-flavors in Dairy Foods 11

Figure 12 Chromatograms (obtained by purge-and-trap GC–MS) of (a) control (normal tasting) half-and-half and (b) half-and-half with chemical
off-flavor because of contamination by propyl acetate solvent from packaging film.

high-molecular-weight polymers, volatile low-molecular-weight compounds are often added to improve the functional properties
of the materials: plasticizers to improve flexibility, antioxidants to prevent oxidation of the plastic polymers or the food inside the
packaging and UV blockers to prevent ʻyellowingʼ of polymeric material when it is exposed to light. Additional additives include
polymerization accelerators, cross-linking agents, antistatic chemicals, and lubricants.
Occasionally packaging materials are not cured before they are used, and a small amount of solvent associated with the
manufacturing of the packaging materials remains. Residual solvent in packaging materials can migrate into the dairy product,
imparting malodors and OF. Residual styrene monomer from polystyrene packaging is a common example.
One example of a dairy product that developed a severe OF problem because of packaging occurred with half-and-half packaged
in 13-ml polystyrene cups. (Note that half-and-half is a mixture of milk and cream with a milk fat content ranging from 10.5% to
18.5%; it is commonly used in the United States to flavor coffee and tea beverages.) Normal-tasting control samples and complaint
samples with a ʻchemical, solvent-likeʼ OF were analyzed by purge-and-trap GC–MS. The most significant differences in the chro-
matograms of control and tainted samples were an increase in the acetone levels and the presence of a significant proplyacetate peak
in the tainted samples (Figure 12). The odor of the propylacetate peak was characterized by GC–O experiments (using a sniff port at
the end of the GC column) and judged to be similar to the sampleʼs odor defect.
The packaging lid-stock consisted of three layers: an outer paper layer impregnated with water-based inks, a middle metal foil
layer, and an inner plastic film used for heat sealing. The inner heat-seal film, which is actuated by heat and pressure to seal the lid to
the cup, is applied to the foil as a slurry and dried. The solvent used by the packaging supplier to form the slurry was propylacetate.
After application of the plastic film to the foil, sheets of the lid-stock material pass through drying ovens to remove all residual pro-
pylacetate solvent. According to the packaging supplierʼs specifications, the finished lidstock should contain less than 1000 mg pro-
pylacetate per 280 m2 of material. After the problem was reported to the packaging supplier, quality control specifications were
monitored more closely and no further OF occurred.

Techniques for Analyzing Flavors and OFs

Gas chromatography with mass spectrometry detection is the method of choice for analyzing volatile and semivolatile odor-active
compounds in dairy products. The volatile and semivolatile compounds in the headspace are of interest because they can travel to
the nose during eating and stimulate the olfactory receptors in the nasal cavity.
Perhaps the most critical and challenging step in the analytical process is the sample preparation technique used to isolate and
concentrate the flavor compounds from the dairy matrix. Because it is not uncommon for the chemicals responsible for food
12 Flavors and Off-flavors in Dairy Foods

malodors to be present at mg kg
1 and even ng kg
1 levels, the extraction technique must collect as many molecules of chemicals as
possible. It is important that the extraction technique does not introduce or create volatiles that are not in the dairy product being
tested. For example, sample preparation techniques that involve heating the sample to a high temperature (e.g., steam distillation)
can generate artifact peaks in sample chromatograms, and these odoriferous artifacts may be misinterpreted as the cause of the mal-
odor OF problem.
Although some excellent flavor work has been done with HPLC, supercritical fluid extraction, supercritical fluid chromatog-
raphy, and other analytical methods, GC techniques have been the most useful. Dairy chemists now have a wide variety of sample
preparation methods that they can use to isolate and concentrate odor-active compounds before GC analysis. The most frequently
used techniques include vacuum distillation, simultaneous steam distillation–extraction (also referred to as the Likens and Nick-
erson extraction procedure), static headspace, dynamic headspace purge-and-trap, direct thermal desorption (for solid and semi-
solid samples), and solidphase microextraction (SPME).
Each sample preparation method has its own advantages, disadvantages, and biases. Selection of the most appropriate method
for a specific application depends on the number of samples to be tested, the type of sample, the specific analytes of interest, the
desired sensitivity, the nature of the information desired, cost of instrumentation and equipment, and so on.
One increasingly popular analytical extraction technique used before GC–MS analysis is SPME. SPME utilizes a short, thin, solid
rod of fused silica (typically 1 cm in length with an outer diameter of 0.11 mm) coated with an absorbsent/adsorbent polymer. The
coated fused silica (the SPME fiber) is attached to a metal rod, and both are protected by a metal sheath that covers the fiber when
not in use. The assembly is placed in a fiber holder. The system resembles a modified syringe (Figure 13).
Two sampling methods can be used with SPME depending on the placement of the fiber relative to the sample – immersion or
headspace sampling. For dairy products, which contain high levels of fat, carbohydrates, and protein, the headspace technique is
preferred. In SPME headspace analysis, a fiber is placed in the headspace above an equilibrated sample. For example, when
analyzing volatiles in a milk sample, 3 ml of milk can be placed in a 9-ml glass GC vial containing a small stirring bar and sealed
with a septum closure. The sample is then heated (e.g., to 50  C). After allowing approximately 5 min for the sample to reach

Figure 13 Solid-phase microextraction (SPME) fiber with holder. Courtesy of Supelco, Bellefonte, PA, USA.
 Flavors and Off-flavors in Dairy Foods 13

800 000 2
700 000
600 000

Total ion count

3
500 000
4
400 000
300 000
200 000 1
100 000
0
100 200 300 400 500 600 700 800 900 1000
Retention time (s)

Figure 14 Chromatogram (obtained by SPME GC–MS using a 75 mm Carboxen–PDMS fiber) of a milk sample (2% milk fat) contaminated with
1300 ppm Matrixx sanitizer. Peak identities are as follows: (1) acetic acid; (2) chlorobenzene (internal standard); (3) heptanal; (4) octanoic acid.
Acetic acid, heptanal, and octanoic acid were not detected in control milk (normal tasting) 2% milk before spiking with sanitizer.

thermal equilibrium, the needle assembly is inserted through the septum and into the headspace above the milk sample while the
sample is stirred with the stirring bar. The fiber is then exposed to the headspace gases. After an additional 10 min, the fiber is
retracted into the needle assembly and removed. The extracted volatiles are thermally desorbed from the fiber in the heated GC
injector port and into the capillary GC column.
Several types of fiber with varying affinities for specific classes of compounds are now available. As a result, the type of SPME fiber
can be selected to optimize results for a particular analyte class. If lipid oxidation is being studied, for example, the analyst could
select a CarboxenÔ–PDMS (polydimethylsiloxane) fiber to measure aldehydes in the 1–500 mg kg
1 range. If concentrations of
aldehydes higher than 500 mg kg
1 are present, the Carboxen–PDMS fiber will become saturated, and a 100 mm PDMS fiber would
be a better choice. Two fibers that work particularly well for the analysis of volatiles in dairy products are 75 mm Carboxen–PDMS
and 65 mm PDMS–divinylbenzene. The fibers are relatively inexpensive, and they are simple and easy to use. The technique is rela-
tively rapid, easy to perform, sensitive, reproducible, quantitative, and automatable.
SPME is particularly well suited to the analysis of dairy products. The technique is more capable of extracting a broader range of
analytes than most other sample preparation methods. For example, SPME is capable of mg kg
1 detection levels of both low-
molecular-weight, highly volatile compounds like acetaldehyde, acetone, and 1,3-pentadiene, as well as high-molecular-weight,
high-boiling-point compounds like vanillin, lactones, and dodecyl aldehyde. Furthermore, it can be used to quantify free fatty acids
(C4–C14) in dairy products. This important class of flavor compounds can be particularly challenging and time consuming to extract
by other techniques.
As an example of the capabilities of SPME, consider the analysis of milk contaminated with sanitizer. One popular sanitizer used
by some dairies is MatrixxÔ (Ecolab, St Paul, MN, USA). Matrixx has the following composition (approximate): 4.4% peroxyacetic
(PAA), 6.9% hydrogen peroxide, and 3.4% octanoic acid. Initial samples of milk off the bottling line may be contaminated with
Matrixx if the processing lines are not flushed thoroughly (Figure 14). Comparison of chromatograms of a control milk sample
(no OF) and a sample with a severe OF that was believed to be caused by contamination with Matrixx shows that peaks for acetic
acid, octanoic acid, and heptanal are good indicators that the sample was contaminated with Matrixx sanitizer. (Note that acetic acid
is a decomposition product of PAA.) Heptanal is a lipid oxidation product. The following conditions were used for the analysis:
Sample: 2 ml of low fat milk þ1 ml 0.033 M phosphoric acid þ1 g salt in a 9-ml vial; SPME fiber: 75 mm Carboxen–PDMS; extrac-
tion method: headspace (with stirring) for 12 min at 40  C; desorption: 2 min at 250  C. The analytical capillary column was FFAP
(30 cm  0.25 mm with a 1.0 mm film thickness). Figure 1 shows a chromatogram of a typical non-contaminated 2% processed
milk sample.

Electronic Nose Applications for Measuring OFs in Milk

A potentially significant development for quality control monitoring of flavors and OFs in dairy products is the recent introduction
of so-called ‘electronic nose’ (e-nose) instruments that employ an array of solidstate chemical sensors based on conducting poly-
mers, metal oxides, surface acoustic wave devices, quartz crystal microbalances, or combinations of these devices. E-noses have been
used to monitor the quality of edible oils and to distinguish different heat treatments of milk.
While e-nose instruments are tools for visually comparing the aromas of samples, they do not provide the same type of
specific detailed chemical information that is possible with GC–MS methods, and the technique is probably not sensitive enough
for some analytes and some types of matrices. The primary advantage of e-noses as a quality assurance tool for the dairy industry
is speed of analysis, in terms of both data generation as well as data interpretation. Rapid, meaningful data interpretation is
possible with various chemometric (multivariate analysis) techniques. As another alternative, data systems offered by some
instrument manufacturers incorporate artificial neural networks for interpretation of data. Neural networks are data-
processing algorithms based loosely on the structure of the human brain. Quality control models can be developed by ‘training’
14 Flavors and Off-flavors in Dairy Foods

the e-nose, and then routine samples can be tested against the model, providing a goodness-of-fit approximation or a sample
accept/reject answer.
To date, commercial e-nose instruments, in general, have not lived up to expectations. Memory problems (carryover), irrevers-
ible poisoning of the chemical sensors and the need for frequent sensor recalibration have limited their usefulness.
Recently, e-noses instruments that use mass spectrometers as the chemical sensor array seem to have overcome many of the defi-
ciencies of previous solid state-based sensor instruments. For example, one analytical system, consisting of SPME to isolate and
concentrate volatiles from the dairy product, MS as a chemical sensor, and multivariate analysis to decipher meaningful trends
in the MS output, has been applied to the study of OFs in milk. Using the multivariate analysis method of principal component
analysis, this technique has been used successfully to differentiate control, good-tasting 2% milk, and 2% milk samples with various
types of flavor defects, including samples contaminated with sanitizer, samples subjected to light abuse, samples that have been
oxidized because of contact with prooxidant metals, and samples that have spoiled because of microbial growth. Preliminary
studies have shown that the SPME–MS multivariate analysis system can even be used to accurately predict the shelf life of processed
milk. The multivariate analysis method used in this study was partial least squares regression.

OFs Generated in Dry Dairy Ingredients

Dairy-based ingredients are widely used in various food and beverage segments because of their high nutritional, functional, and
taste appeal. Therefore, monitoring dairy-based ingredients for OFs and malodors is important to avoid expensive product recalls.

Problem with 2,4,5-Trimethyloxazole in Spray-Dried Dairy Products

Reactions arising from carbonyl compounds can generate OF chemicals even at low temperatures. Several interesting malodorous
compounds can be generated from sulfur amino acids, in particular cysteine, such as pyrazines, alkylpyrazines, methylthiazoles,
acetylthiazole, acetylthiazolidine, trimethyloxazole, and dimethylethyloxazoles. These various compounds produce flavor notes
of sulfur, popcorn, hazelnut, toasted, roasted, and ripe fruits. Because of their OF character and low odor and taste thresholds, these
compounds can lead to serious taste and odor defects.
One OF problem that has occurred in spray-dried dairy products through this mechanism is the formation of 2,4,5-
trimethyloxazole in spray-dried buttermilk, sour cream, and yogurt products. For example, Figure 15 is a chromatogram of
a spray-dried sour cream powder sample, showing a 2,4,5-trimethyloxazole peak at 43 mg kg
1. This contaminant was responsible
for the rejection of the entire production lot of the sour cream powder by sensory paneling. Analysis was performed by solid-phase
microextraction (SPME) GC–MS using the PDMS–divinylbenzene fiber. One gram of sour cream powder and 5 ml of NaH2PO4
(25% w/v) solution were added to a 20 ml GC vial, vortexed for 1 min, preincubated for 4 min at 50  C with agitation and extracted
with the SPME fiber for 20 min at 50  C with agitation before GC–MS analysis.
L. Pripis-Nicolau and colleagues have shown that the release of acetaldehyde by cysteine in the presence of a-diketones can
generate alkyloxazoles. With diacetyl, the authors identified the formation of trimethyloxazole and with pentane-2,3-dione the
formation of two oxazole isomers (2,5-dimethyl-4-ethyloxazole and 2,4-dimethyl-5-ethyloxazole) in model system
studies. A possible mechanism for the formation of 2,4,5-trimethyloxazole from diacetyl, acetaldehyde, and ammonia is shown
in Figure 16. 2,4,5-Trimethyloxazole has a taste threshold in water of 5 mg kg
1. It has been detected in wine and dairy products
and has an OF characterized as melonlike with pungent ripe kiwi notes.

Problem with Musty Trihaloanisole Malodors in Casein Powders

Another dried dairy powder that can develop OFs and malodors is casein. Rennet casein, produced by enzymatic (rennet) coagu-
lation from pasteurized milk, is used in both industrial and food applications. Other types of caseins are produced by isoelectric

Abundance

500 000

400 000

300 000 43 µg kg–1 TMO

200 000

100 000

0
Time 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00 18.00 19.00

Figure 15 2,4,5-Trimethyloxazole (43 mg kg
1) in spray-dried sour cream powder.

 Flavors and Off-flavors in Dairy Foods 15

Figure 16 Possible mechanism for formation of 2,4,5-trimethyloxazole from diacetyl, acetaldehyde and ammonia.

precipitation by addition of acid or fermentation. The resulting range of acid and rennet casein products is used for nutritional and
medical foods. Because of their flavor stability and functional properties, casein ingredients have wide appeal in cheese analogs,
bakery, meat, confectionary products, desserts, nutritional beverages, and energy bars.
Casein powders have a characteristic unpleasant stale flavor. E. H. Ramshaw and E. A. Dunstone described the flavor as stale,
gluelike, and burnt feathers. Others have described the typical odor of rennet casein as ʻwet-dog.ʼ o-Aminoacetophenone has
been identified as an important odor-active component in stored skim milk powder. It is important to understand the nature of
the flavor defects of casein powders because of their wide application in the food industry; the flavor of casein is a critical quality
parameter.
Y. Karagul-Yuceer and colleagues have identified the primary odorants responsible for the typical odor of rennet casein powder.
Results of aroma extraction– dilution analysis (AEDA) indicated that o-aminoacetophenone is a potent odorant. However, sensory
descriptive analysis of model aroma systems revealed that the typical odor of rennet casein was caused principally by hexanoic acid,
indole, guaiacol, and p-cresol.
To isolate volatiles prior to GC–MS analysis, Y. Karagul-Yuceer and colleagues used direct solvent extraction followed by high-
vacuum distillation. This analytical technique to isolate volatiles is often referred to as solvent-assisted flavor evaporation (SAFE).
Although it is an excellent technique for isolating volatiles from complex sample matrixes before GC–MS, SAFE analysis is time
consuming and requires the use of large amounts of organic solvent (typically diethylether).
Recent advances in analytical extraction technologies have significantly simplified the extraction procedure. One example is stir
bar sorptive extraction (SBSE) developed by E. Baltussen and colleagues and marketed by GERSTEL Inc.
SBSE is a simple, solvent-free technique that permits the extraction and concentration of analytes in a single step. SBSE is an
excellent technique for studying odor-active compounds in casein powders. It was applied recently to the study of samples of rennet
casein that had developed an unusual strong musty odor. SBSE is based on mixing a sample solution with a GERSTEL TwisterÔ,
a magnetic stir bar sealed in glass and coated with PDMS. The advantages of PDMS phase as an extraction medium have been
described previously (M. A. Baltussen). Once the odor-active chemicals are extracted by SBSE, they are desorbed from the Twister
and onto a capillary GC column using a thermal desorption unit (GERSTEL TDU) equipped with an MPS-2 auto-sampler and a CIS
4 programmed temperature vaporization (PTV) inlet (Gerstel, Mulheim an der Ruhr, Germany) installed on an Agilent 6890 gas
chromatograph with a 5973 mass-selective detector (Agilent Technologies, CA, USA).
The sample preparation technique involved mixing 1 g of casein and 25 ml distilled water with a Twister stir bar
(2 cm  0.5 mm) for 3 h at 900 rpm. The Twister is removed from the sample, rinsed with distilled water, dried with a lint-free
paper towel, and thermally desorbed in a Gerstel TDU. Musty casein samples from one supplier showed that the samples contained
2,4-dichroloranisole; 2,4,6-trichloroanisole (7.2 mg kg
1); 2,3,6-trichloroanisole; and 2,4,6-trichlorophenol. Another set of musty
casein samples was found to contain 2,4,6-tribromophenol and 2,4,6-tribromoanisole (215 mg kg
1). Three major international
food companies had produced nutritional beverages and nutritional bars using these trihaloanisole-contaminated caseins, resulting
in expensive product recalls. Microbial (fungal enzymatic) o-methylation of the trihalophenols formed the strong musty trihaloa-
nisoles which were responsible for the musty OFs. Trichlorophenol is used as a pesticide and wood preservative, and tribromophe-
nol is used as a flame retardant.
In addition to measuring low concentration levels of trihalophenols and trihaloanisoles, the technique is useful for studying
odor-active chemicals in casein and could provide a better understanding of the cause of the common wet-dog malodor of rennet
casein. To get high recovery of both hydrophilic and hydrophoboic malodorous chemicals from casein, a technique called sequen-
tial SBSE was used. The technique was developed by N. Ochiaia and colleagues (GERSTEL KK, Tokyo, Japan). One gram of casein is
vortexed with 10 ml water for 3 min in a 20 ml GC vial. A GERSTEL TwisterÒ stir bar (1 cm  0.5 mm) is added, and the vial is
sealed with a screw cap. The sample is extracted for 1 h by stirring with the Twister at 1500 rpm. The Twister is removed, dried
with a lint-free paper towel and added to a thermal desorption tube. A second Twister (1 cm  0.5 mm) and 3.0 g NaCl are added
to the same casein/water sample. The vial is sealed with a screw cap, and the sample is re-extracted an additional hour while stirring
at 1500 rpm. This second stir bar is removed, rinsed in distilled water, dried with a paper towel, and added to the first stir bar in the
16 Flavors and Off-flavors in Dairy Foods

Figure 17 Experimental setup for sequential SBSE. Sample is 1 g of casein in 10 ml of distilled water.

same desorption tube. Both Twisters are then simultaneously desorbed in the Gerstel TDU. The experimental setup is shown in
Figure 17.
With the sequential SBSE, hydrophobic analytes are extracted with maximum recovery in the initial aqueous extraction, and
hydrophilic analytes are extracted with maximum recovery with the salting-out procedure used in the second extraction.
Thirty-two odor-active compounds were analyzed as standards. Five levels of each standard were prepared in the 20–
6000 mg kg
1 range (spiked in control casein sample) and analyzed by sequential SBSE. The average linear least squares correlation
coefficient for these standards was 0.9955. A mid-range standard was analyzed in triplicate for each of the 32 standards; the average
standard deviation of triplicate determinations for the 32 samples was 4.3%.
Quantitative results for 21 analytes in various types of caseins are reported in Table 4. Table 5 lists other odor-active analytes that
were detected in the caseins by sequential SBSE but were not quanitated.

OFs from Catabolic Reactions of Aromatic Amino Acids

Microbial catabolism of amino acids generated from the degradation of milk proteins during cheese maturation is critical to the
development of desirable cheese flavor. Examples of beneficial cheese flavors produced in this manner include the conversion of
methionine to methional, dimethyl sulfide, methanethiol, and other sulfur-containing compounds. However, not all catabolic reac-
tions produce desirable flavor compounds. Chemicals generated from the catabolism of aromatic amino acids (AAA) have been
implicated in the development of numerous types of significant OFs. For example, the phenylalanine catabolites phenylacetalde-
hyde and 2-phenylethyl alcohol can impart undesirable floral, rose-like OF and the tyrosine catabolite p-cresol imparts barny, cowy,
fecal, and medicinal notes to cheese and other dairy products. Previous research has shown that a-keto acids produced from tryp-
tophan, tyrosine and phenylalanine by amino-transferase enzymes are chemically labile and may spontaneously degrade into
a variety of OF chemicals including benzaldehyde, phenylacetaldehyde, phenylethyl alcohol and other objectionable aroma
compounds. 2-Aminoacetophenone (2AA) imparts a ‘gluey’ and ‘foxy’ character to milk powder and casein through degradation
of tryptophan. 2AA is an intermediate in the biosynthetic pathway for quinazolines, a pathway branching from the tryptophan cata-
bolic pathway. Evidence shows that 2AA is consistently produced by P. aeruginosa (Table 6).
Chemical analysis of AAA catabolites that cause OFs can be challenging. For example, p-cresol has been linked to barny/cow-like
odors in dairy products. Casein powder, a common food ingredient in cheese analogs, bakery, meat, confectionery products,
 Flavors and Off-flavors in Dairy Foods 17

Table 4 Levels (mg kg
1) of selected odorants in various caseinate samples by sequential SBSE

Ca cas control Ca cas musty Acid casein Rennet casein Na casein Ca cas musty
Name source A (mg kg
1) source A (mg kg
1) source A (mg kg
1) source A (mg kg
1) source A (mg kg
1) source B (mg kg
1)

Hexanal 549 480 2578 538 1835 719

2-Heptanone 0.0 0.0 52.5 17.6 29.0 3.4
Heptanal 1392 514 539 271 588 55
Hexanoic acid, 6.4 0.0 2.3 0.0 0.0 0.0
ethyl ester
Octanal 319 150 236 103 231 131
Phenol 1589 2292 0.0 0.0 0.0 0.0
2-Nonanone 46.4 67.0 30.7 8.6 24.0 15.8
Nonanal 1658 1138 1474 552 1023 7447
Maltol 0.0 296 0.0 0.0 0.0 0.0
Phenylethyl alcohol 1282 0.0 0.0 0.0 0.0 0.0
2-Nonenal, (E)- 0.0 42.1 146 31.1 71.5 0.0
g-Nonalactone 4.5 0.0 0.0 0.0 0.0 0.0
Decanal 132 179 417 123 183 1997
Benzothiazole 0.0 0.0 0.0 0.0 0.0 28.7
o-Aminoacetophenone 0.0 464 430 177 310 0.0
Indole 284 0.28 3.10 0.03 0.00 0.03
2,4,6-Trichloroanisole 0.0 7.2 0.0 0.0 0.0 0.0
d-Nonalactone 125 0.0 815 0.0 654 0.0
2,4,6-Tribromoanisole 0.0 0.0 0.0 0.0 0.0 215
2,4-Di-t-butylphenol 0.0 0.0 202 384 431 0.0
g-dodecalactone 211 1624 2168 135 0.0 0.0

Table 5 Some additional odorants detected in caseinate by sequential SBSE but not
quantitated

Name Retention time (s) CAS registry number

Butanoic acid 407 107-92-6

3-Heptanone 434 106-35-4
Isovaleric acid 457 503-74-2
Pyrazine, 2,5-dimethyl- 475 123-32-0
Pyrazine, 2,3-dimethyl- 485 5910-89-4
2-Heptenal, (E)- 524 18829-55-5
Benzaldehyde 538 100-52-7
Dimethyl trisulfide 544 3658-80-8
2,4-Pentanedione, 3-methyl- 556 815-57-6
Furan, 2-pentyl- 560 3777-69-3
Pyrazine, trimethyl- 584 14667-55-1
3-Octen-2-one 622 1669-44-9
Hexanoic acid 626 142-62-1
1-Hexanol, 2-ethyl- 625 104-76-7
3,5-Octadien-2-one 663 38284-27-4
Acetophenone 667 98-86-2
Pyrazine, tetramethyl- 685 1124-11-4
3,5-Octadien-2-one, (E,E)- 692 30086-02-3
Acetic acid, phenylmethyl ester 774 140-11-4
Octanoic acid, ethyl ester 796 106-32-1
2,4-Dichloroanisole 816 120-83-2
Octanoic acid 831 124-07-2
2,4-Nonadienal, (E,E)- 830 5910-87-2
Acetic acid, 2-phenylethyl ester 875 103-45-7
Decanoic acid, ethyl ester 995 110-38-3
Thiazole, 2-ethyl 999 15679-09-1
n-Decanoic acid 1022 334-48-5
2,4,6-Trichlorophenol 1025 88-06-2
2,3,6-Trichloroanisole 1026 50375-10-5
2,4,6-Tribromophenol 1121 118-79-6
Dodecanoic acid 1158 143-07-7
g-dodecalactone 1201 2305-05-7
18 Flavors and Off-flavors in Dairy Foods

Table: 6 Examples of aromatic amino acid (AAA) catabolites

that contribute off-flavors to dairy products

AAA Catabolite Precursor: Off-flavor Description

Acetophenone Phenylalanine: Almond, musty, glue

Benzaldehyde Phenylalanine: Almond, bitter almond
Phenyl acetaldehyde Phenylalanine: Rosy, violet-like
Phenylethyl alcohol Phenylalanine: Unclean, rose, violet-like, honey
Phenyl acetic acid Phenylalanine: Flowery, rosy, plastic
Phenol tyrosine: Plastic, rubber, medicinal
p–OH–phenyl aldehyde tyrosine: –
p–OH–phenyl lactate tyrosine: –
p–OH–phenyl acetate tyrosine: –
p-Cresol tyrosine: Barny, medicinal, unclean
Indole tryptophan: Unclean, mothball
Skatole tryptophan: Unclean, mothball
Benzaldehyde tryptophan: Almond
2-Aminoacetophenone Tryptophan: Foxy character, grape

nondairy creamers, nutritional bars and nutritional beverages can suffer from this malodor defect. Since the odor threshold for p-
cresol is approximately 50 mg l
1, an effective analytical method for monitoring cresols should demonstrate a sensitivity of 50 mg l
1
or less.
In our laboratory, SPME and SBSE sample preparation methods prior to GC–MS did not meet the required sensitivity. The mole-
cule is too hydrophilic to be extracted at high recovery by PDMS SBSE (log Ko/w is 1.98). To circumvent the problem, p-cresol was
derivatized to p-cresol acetate, which is nonpolar enough to be extracted with PDMS stir bars with high recovery.
Sample pretreatment is required. Casein samples are treated with methanol and 6-M HCl to free the cresols that are bound to the
casein. Following this pretreatment, the sample is acylated with acetic anhydride. The sample preparation method is as follows:
1. One gram of casein plus 20 ml of 6 M HCl are added to 3 ml distilled water in a 40 ml GC vial. The sample is stirred vigorously
and sonicated for 10 min to hydrate the casein.
2. Deuterated o-cresol (10 ml of a 0.04 mg ml
1 standard solution) is added.
3. Six ml of methanol are added, and the sample is stirred and sonicated for 10 min to free bound cresols from the casein.
4. The sample is filtered through a medium filter paper, washing filtrate with 3 ml of water.
5. The methanol is gently evaporated off with a nitrogen stream at 55  C.
6. One gram potassium carbonate is slowly added with gentle stirring.
7. One ml acetic anhydride is added, and the sample is stirred for 15 min (acylation step).
8. Three grams of NaCl are added.
9. An SBSE PDMS Twister (2 cm  0.5 mm) is added.
10. The sample is stirred for 3 h at 1000 rpm.
11. The SBSE Twister is rinsed with distilled water, patted dry with a lintless cloth, and thermally desorbed in a GERSTEL TDU
(thermal desorption unit).
12. Volatiles are analyzed in SIM at m/z 108, 115 and 150 amu with a DB-5MS capillary column (30 m  0.25 mm  0.25 mm).
Stir bar sorptive extraction followed by thermal desorption and GC–MS analysis (SBSE-TD-GC/MS) can be useful for quantita-
tion of methyl phenols and other AAA catabolites in dairy products. The technique using two types of Twisters was investigated for
phenols in drinking water (Y. Nie and T. Albinus, Comparison of EG-Silicone-SBSE and Derivatization-PDMS-SBSE for the Analysis
of Phenolic Compounds and OFs in Water, Gerstel AppNote, 2012). SBSE based on the PDMS Twister used in combination with
derivatization (acylation) and SBSE based on the EG-Silicone Twister without derivatization both gave similar results: Low limits of
detection and quantification, good linearity and repeatability, and negligible carryover. Based on the low limits of detection,
simplicity of use, low cost and high quality of results, SBSE-TD-GC/MS is a highly suitable analytical tool for the determination
of phenols and off flavor compounds in water. Limits of detection (LODs) for phenolic compounds were in the range 0.007–
0.036 mg l
1 for the EG-Silicone Twister (without derivatization) and 0.011–0.053 mg l
1 for the PDMS Twister (with derivatiza-
tion). This solventless technique should have application for quantifying low concentrations of phenolics in milk and other dairy
products.

Conclusion

The advent of new sensitive and rapid analytical methods in conjunction with olfactometry techniques and traditional sensory taste
paneling approaches has greatly improved our understanding of flavor-impact chemicals in dairy products. Application of new
 Flavors and Off-flavors in Dairy Foods 19

analytical technologies, as well as emerging analytical technologies such as electronic nose instruments, will allow dairy processors
to better understand why OF development occurs in dairy products and how to minimize it. In the near future, enhanced analytical
tools will likely move beyond dairy research laboratories and into production plant quality control laboratories as instrumentation
becomes less expensive, easier to use, and automated.

Further Reading

Azzara, C.D., Campbell, L.B., 1992. Off-flavors of dairy products. In: Charalambous, G. (Ed.), Off-flavors in Foods and Beverages. Elsevier, Amsterdam, pp. 329–374.
Baltussen, H.A., 2000. New Concepts in Sorption Based Sample Preparation for Chromatography (Ph.D. thesis). University of Technology, Eindhoven, The Netherlands.
Blank, I., 1997. Gas chromatography–olfactometry in food aroma analysis. In: Marsili, R.T. (Ed.), Techniques for Analyzing Food Aroma. Marcel Dekker, New York, pp. 293–329.
Harmon, A.D., 1997. Solid-phase microextraction for the analysis of flavors. In: Marsili, R.T. (Ed.), Techniques for Analyzing Food Aroma. Marcel Dekker, New York, pp. 81–112.
Jenness, R., Patton, S., 1959. Principles of Dairy Chemistry. John Wiley, New York.
Jung, M.Y., Yoon, S.H., Lee, H.O., Min, D.B., 1998. Singlet oxygen and ascorbic acid effects on dimethyl disulfide and off-flavor in skim milk exposed to light. J. Food Sci. 63,
408–412.
Karagul-Yuccer, Y., Drake, M.A., Cadwallader, K.R., 2001. Aroma-active components of nonfat dry milk. J. Agric. Food Chem. 49, 2948–2953.
Marsili, R.T., 1981. Monitoring bacterial metabolites in cultured buttermilk by high-performance liquid chromatography and gas chromatography. J. Chromatogr. Sci. 19, 451–456.
Marsili, R.T., 1997. Off-flavors and malodors in foods: mechanisms of formation and analytical techniques. In: Marsili, R.T. (Ed.), Techniques for Analyzing Food Aroma. Marcel
Dekker, New York, pp. 237–264.
Marsili, R.T., 1999. Comparison of solid-phase microextraction and dynamic headspace methods for the gas chromatographic-mass spectrometric analysis of light-induced lipid
oxidation products in milk. J. Chromatogr. Sci. 37, 17–23.
Marsili, R.T., 2000a. SPME-MS-MVA as an electronic nose for the study of off-flavors in milk. J. Agric. Food Chem. 47, 648–654.
Marsili, R.T., 2000b. Shelf-life prediction of processed milk by solid-phase microextraction, mass spectrometry, and multivariate analysis. J. Agric. Food Chem. 48, 3470–3475.
Marsili, R.T. (Ed.), 2001. Flavor, Fragrance and Odor Analysis. Marcel Dekker, New York.
Mistry, B.S., Reineccius, T., Olson, L.K., 1997. Gas chromatography–olfactometry for the determination of key odorants in foods. In: Marsili, R.T. (Ed.), Techniques for Analyzing
Food Aroma. Marcel Dekker, New York, pp. 265–292.
Moio, L., Langlois, D., Etievant, P., Francesco, A., 1993. Powerful odorants in bovine, ovine, caprine and water buffalo milk by means of gas chromatography-olfactometry. J. Dairy
Res. 60, 215–222.
Ochiai, N., Sasamoto, K., Kanda, H., Pfannkoch, E., 2008. Sequential stir bar sorptive extraction for uniform enrichment of trace amounts of organic pollutants in water samples. J.
Chromatogr. A 1200, 72–79.
Ott, A., Fay, L.B., Chaintreau, A., 1997. Determination and origin of the aroma impact compounds of yoghurt flavor. J. Agric. Food Chem. 45, 850–858.
Pripis-Nicolau, L., de Revel, G., Betrand, A., Maujean, A., 2000. Formation of flavor compoents by the reaction of amino acid and carbonyl compounds in mild conditions. J. Agric.
Food Chem. 48, 3761–3766.
Ramshaw, E.H., Dunstone, E.A., 1969. Volatile compounds associated with the off-flavor in stored casein. J. Dairy Res. 36, 215–223.
Roberts, D.D., Pollien, P., Milo, C., 2000. Solid-phase microextraction method development for headspace analysis of volatile flavor compounds. J. Agric. Food Chem. 48,
2430–2437.
Shipe, W.F., Bassette, R., Deane, D.D., et al., 1978. Off flavours of milk: nomenclature, standards, and bibliography. J. Dairy Res. 61, 855–869.