Influence of dietary composition on the carbon, nitrogen, oxygen and hydrogen stable isotope ratios of milk


  • Presented at the annual meeting of the Stable Isotopes Mass Spectrometry Users' Group (SIMSUG), 20–22 June, 2007, hosted by the Institute for Research on the Environment and Sustainability (IRES) and the School of Civil Engineering and Geosciences, Newcastle University, UK.


The stable isotope ratios (13C/12C, 15N/14N, 18O/16O, D/H) of animal feed and milk were investigated, considering cows stabled in two farms and fed with diets made up of different kinds of C3 plants and different amounts of maize. Maize was characterised by δ13C, δ18O and δD values significantly higher than those of the C3 plants, while, for the C3 plants, Festuca arudinacea had significantly higher content of 13C and 15N. The δ13C and δ18O values of the overall diet and the δ13C of milk casein and lipids were shown to be significantly correlated with the percentage of maize in the animal diet. On the other hand, the δ18O values of milk water and the δ18O, δD and δ15N values of casein were shown to be only slightly influenced by the amount of maize in the feed, being probably more closely correlated with the geo-climatic and pedological characteristics of the area of origin and with the presence of fresh plant or silage in the ration. The δ13C value of casein was shown to be a suitable parameter for evaluating the amount of maize in the diet: each 10% increase in the maize content corresponded to a shift of 0.7‰ to 1.0‰ in the δ13C of casein. A threshold value of −23.5‰ for δ13C in milk casein, above which it is not possible to exclude the presence of maize in the diet, was suggested. The results obtained could be useful for determining mislabelling of dairy products declared to have been produced by pastured animals or of PDO cheeses with an established amount of maize in the diet and for verifying the unpermitted addition of exogenous components to milk. Copyright © 2008 John Wiley & Sons, Ltd.

Recently, several papers have been published about the application of stable isotope ratio analysis (SIRA) using isotopic ratio mass spectrometry (IRMS) to animal products. The aims of these studies were to obtain dietary reconstruction, to study animal movement patterns,1–5 and to characterise the geographical origin.6–16 Another important aim could be to study and quantify the alteration of stable isotope ratios (SIRs) in milk by substituting a C3 plant diet with maize. This could be important in order to protect milk and milk derivatives produced using traditional or ‘organic’ practices of pasture compared with products produced with cheaper diets based on maize derivatives. There have been recent reports about this in defatted meat,17–19 in related fat,17 in several animal tissues such as blood, plasma, liver, kidney, cow hair20 and in sheep faeces.21 In these papers it was shown that there is a strong correlation between the SIR of carbon and the amount of maize in the diet. It was also shown that the degree of this alteration depends on the component or body tissue considered, due to its biochemical composition and turnover rate.22 On the other hand, variations in the SIRs in milk and milk derivatives following suitable experimental design have not been so extensively explored. Metges et al.23 investigated the 13C/12C of CO2 breath, bulk milk and serum by exchanging a C3 plant with a C4 plant diet, and vice versa. Wilson et al.24 studied the contribution of body protein to milk protein, changing abruptly from a C3 to a C4 diet. More recently, Knobbe et al.25 followed the variation in stable carbon and nitrogen isotopic composition under different feeding regimes. It was concluded that, to obtain more detailed information on the influencing factors, the analysis of the SIRs of other elements such as oxygen and hydrogen may be necessary.

In this study we considered around 130 cows stabled in two farms and supplied with diets made up of different kinds of C3 plants and different amounts of maize. The different feed ingredients, the overall diet and the relevant milk (casein, lipid and some bulk samples) were subjected to analysis of 13C/12C and 15N/14N. As suggested by Knobbe et al.,25 we also investigated the 18O/16O and D/H of the whole diet, casein and some feed components, and 18O/16O in bulk milk. The aim of the work was to evaluate the influence of dietary composition (amount of maize and different kinds of C3 plants), on the SIRs of animal feed and milk, comparing the results with those previously published, discussing the possible reasons for the variations observed and checking for the presence of significant linear relationships. The final aim was to verify the possibility of establishing the presence and percentage of maize in the animal diet on the basis of multi-element (C, N, H, O) isotope ratio investigations of milk. Knowledge about the presence or amount of maize in animal diet could be very useful for protecting the consumer, who is willing to give additional value to milk and cheese produced using traditional pasture practices or to PDO products, for which the amount of maize or silage in the diet is regulated. Moreover, a study on the relationship between the SIRs of different milk fractions could help to determine the unpermitted addition of exogenous milk components, in the case of reconstituted dairy products.


Animals and experimental diets

Around 130 dairy cows from two farms situated near two cities (Cremona and Mantova) in the northern Italian region of Lombardia were considered. The two farms will be indicated in the text as farms C and M. Before the experimental period the cows were fed with a mixed C3 and C4 diet and approximately half of the energy was supplied by maize (silage and flour). During the experimental period, the cows from each farm were randomly divided into two groups, A and B. Each group from both farms was supplied successively with two of four diets with a different % of maize, corresponding to thesis 0, without maize silage and theses 1, 2, 3 with an increasing amount of maize, as shown in Table 1. Group A was first fed with diet 1 and then with diet 0, and group B was first fed diet 2 and then diet 3. The other dietary constituents were forage derived from C3 plants, such as Medicago sativa, Lolium multiflorum, Festuca arudinacea and concentrates based on cereals and soybean (Table 1). The forage and silage were obtained in the farms mentioned, whereas concentrates and maize flour were bought on the market. Feed and water were available ad libitum. Each cow ate about 22 kg of dry matter feed per day. As one week is sufficient time to reach a new isotopic equilibrium of milk elements following a change in the type of diet,24, 25 in this experiment each diet was adopted for at least 2 weeks before collecting three or four samples of milk produced on different days. The experiment was carried out from the middle of February up to the end of March 2005.

Table 1. Percentage composition of the feed ingredients in the diet of the animals (% of dry matter)
FARM 0123
FARM CMedicago sativa24151010
Festuca arudinacea2415118
maize silage0152736
maize flour0242015
cotton seed4666
concentrate 19888
concentrate 239000
concentrate 30161817
%Zea mays23425155
FARM MMedicago sativa34252014
Lolium multiflorum221495
maize silage0173044
maize flour28262219
soybean meal791012
%Zea mays28435363

Feed samples

During the experiment, samples of different feed ingredients, the overall diet, feed water and silage water were taken in each farm. The feed samples were ground with a 1093 Cyclotec sample mill (screen 1 mm; Foss, Padua, Italy). The silage water was obtained by freezing and pressing the silage and then filtering the liquid obtained with filter paper. All the samples were stored at −20°C until analysis. Comparing the chemical composition of the animals' diet in the two farms, as obtained by the Ersaf laboratory, the farm C maize silage showed relatively higher acidity, damp and fermentability, due to the high amount of lactic acid, a lower starch content and digestibility and a higher water content, whereas the farm C forage a higher protein content and digestibility, as compared to the farm M diet.

Milk samples

Milk production was about 26–30 L/cow day−1 and 23–27 L/cow day−1 in farms M and C, respectively. For each thesis, three or four pooled samples of milk (250 mL each) were collected on three or four different days, mixing the milk obtained from the two milkings on the same day. The milk samples were stored at −20°C until analysis. The chemical composition of the milk from farm M showed a mean% level of fat, protein and lactose of 4.00, 3.48 and 4.93, respectively, while farm C had levels of 3.80, 3.41 and 4.94.

Milk fat was removed by centrifugation (10 min at 4500 rpm). Casein was precipitated from the skimmed milk by acidification at pH 4.3 with 2 N HCl and then centrifuged. The precipitate was rinsed with water and freeze-dried.26


Measurement of the 13C/12C and 15N/14N ratios of feed, milk and casein and of the 13C/12C of fat was carried out using IRMS (Delta plus XP mass spectrometer; ThermoFinnigan, Bremen, Germany) following total combustion of the sample (≈0.5 mg of casein and ≈1.5 mg of feed) in an EA Flash 1112 elemental analyser (ThermoFinnigan), as described elsewhere.11 The D/H and 18O/16O ratios of feed and casein were measured in one run using IRMS following pyrolysis of the sample in a high-temperature conversion/elemental analyser (TC/EA; ThermoFinnigan). A high furnace temperature (1450°C), low instrumental H3 factor (<8, for correction of the contribution of [H3]+ to the m/z 3 signal)27 and special care to maintain dry conditions during analysis were ensured, to obtain reproducible results. The samples were weighed into silver capsules (around 0.100–0.300 mg), put in the carousel of the autosampler and stored in a desiccator above P2O5 for at least 24 h. Afterwards, the carousel was inserted in the autosampler equipped with a suitable cover. During measurement, dryness was guaranteed by flushing nitrogen continuously over the samples. The casein samples had previously been left in the laboratory atmosphere for at least 24 h, to allow the casein hydrogen to be in equilibrium with air humidity.28 In the TC/EA the oxygen and hydrogen in the sample were quantitatively converted into CO and H2, respectively, then separated in a gas chromatography (GC) column (80°C) and flushed into the IRMS instrument through a ConFlo III device (ThermoFinnigan). The mass spectrometer measured first the D/H ratio and then, following a magnet jump, the 18O/16O ratio, taking about 10 min per sample. The H2 gas reference material was calibrated against NBS22 (−120‰) and the CO gas against IAEA CH6 (+36.4‰); both standards were supplied by IAEA, Vienna, Austria. In the case of casein, the values were corrected against a casein reference material with an assigned value of D/H, according to the ‘comparative equilibration technique’.16, 28 We had in fact to take into account that part of the total hydrogen present in the casein is exchangeable with air humidity. The 18O/16O ratios of milk, drinking water and silage water were analysed on CO2 with an ISOPREP 18 (VG Isotech, Middlewich, UK) on-line preparation system that allows CO2/H2O equilibration, interfaced to a SIRA II isotope ratio mass spectrometer (VG Isogas, Middlewich, UK) according to the water equilibration method described for wine in the EU Regulation 822/97.

The values were expressed in δ11 against international standards (Vienna Pee Dee Belemnite for δ13C, Vienna Air for δ15N, Vienna – Standard Mean Ocean Water for δ18O and δD). The uncertainty of measurements was ±0.2‰ for the δ13C and δ15N analysis of milk, casein and fat, ±0.5‰ for the δ13C and δ15N of feed (due to the lower homogeneity of the samples), ±0.2‰ and ±0.5 ‰ for the δ18O of milk water and casein or feed, respectively, and ±3‰ for the δD of casein and feed.

Statistical analysis

Analysis of variance (ANOVA) and subsequent means comparison using the t-test and Unequal N Tukey HSD test were used to verify the existence of statistically significant differences between the groups. Linear regression analysis was used to establish the relationship between SIR and the percentage of maize in the diet. The data were processed using the statistical software package Statistica 7.1 (Statsoft Italia srl, Vigonza, Italy).


Isotopic values of feed

The isotopic values of different dietary feed ingredients of both the farms are presented in Table 2. The δ13C values corresponded with those in the literature,6, 11, 19, 20, 24 in particular with those for feed from Germany, probably due to the similar climatic conditions. The δ15N values appeared to be more variable, having been proved to be influenced by a greater number of factors including soil conditions, climate, distance from the sea, fixation of nitrogen from air, intensity of crop practices and the use of fertiliser.11, 16, 28

Table 2. Isotopic values of feed ingredients
  δ13C (‰, V-PDB)δ15N (‰, AIR)δ18O (‰, V-SMOW)δD (‰, V-SMOW)
forages - C3 plantsMedicago sativa−28.70.521.3−99
Medicago sativa−27.30.5  
Festuca arudinacea−−106
Festuca arudinacea−25.36.9  
mixture of Medicago + Festuca−25.71.4  
mixture of Medicago + Festuca−28.31.9  
C4 feedmaize silage−12.43.725.4−69
maize silage−12.54.8  
maize flour−−32
concentratescotton seed−26.07.0  
concentrate 1−18.23.724.6−57
concentrate 2−18.72.924.9−64
concentrate 3−22.50.923.4−79
forages – C3 plantsMedicago sativa−28.43.319.8−105
Medicago sativa−28.81.1  
Lolium multiflorum−29.50.721.1−123
Lolium multiflorum−29.60.4  
mixture of Medicago + Lolium−28.81.6  
mixture of Medicago + Lolium−28.90.8  
C4 feedmaize silage−−69
maize flour−−32
soybean meal−25.11.522.1−85

As expected, the C4 feeds (maize silage and maize flour) showed higher δ13C values than the C3 plants (hay of Medicago sativa, Festuca arudinacea and Lolium multiflorum, soybean, barley, cotton seed), due to the different photosynthetic CO2 fixation pathways used by plants.29 Among the C3 forages (δ13C values from −29.6 to −23.2‰), the hay of Festuca arudinacea had significantly (p < 0.05) higher content of 13C than the others. Possible reasons for this difference are variability in the water availability, the irrigation treatments adopted, the development level of the plant and plant habits, leading to changes in plant stomatal aperture, and, as a consequence, changes in internal CO2 concentration.30 The values of concentrates varied between −25.1 and −26.7‰ for the C3 plant derivatives and between −18.2 and −22.5‰ for those containing maize (concentrates 1 and 2: around 40% of maize; concentrate 3: around 5%).

For δ15N values, Festuca arudinacea (p < 0.001) and Zea mays (p < 0.01) showed significantly higher values than Medicago sativa and Lolium multiflorum, whereas as regards the concentrates, cotton showed a higher 15N content than soybean and barley. Soybean could have low 15N content because it is a nitrogen-fixing crop and therefore could have δ15N values close to that of atmospheric N2 (=0‰). Cotton had a high content of 15N, probably because it was produced in an area with a hot climate (as observed in Namibia,31 for example). For the other plants, the differences could be due to different types (e.g. organic/conventional) and intensity of fertilisation practices (e.g. maize is normally produced in intensively fertilised fields).6

With regard to the δ18O and δD values, it has been shown previously32, 33 that the compounds of C4 plants have significantly higher D content than those of C3 plants, whereas no differences were observed between the two kinds of plants as regards δ18O.34, 35 In this work both δD and δ18O values, measured in a subset of samples, were significantly (p < 0.001) lower in the C3 plant forage than in maize samples and concentrates with maize. The values of concentrates changed according to the content of maize. Maize silage showed lower δD values than maize flour, probably because it contained vegetal waters. The δ18O values of drinking water were −9.1‰ and −9.2‰, whereas the vegetal water of silage had δ18O values of −4.0‰ and −2.8‰ in farms C and M, respectively. As expected, the vegetal water was enriched in 18O in comparison with the source water, due to evaporation processes occurring in the plants.

Considering the δ13C variation of the whole diet with the percentage of maize silage (Fig. 1), strong linear relationships (p < 0.001) were shown. The regression lines for the two farms had similar slopes and different intercepts, due to the presence of different components with unequal 13C content in the two diets (e.g. Lolium in the diet of farm M and Festuca in the diet of farm C). The extrapolations of the δ13C regressions to 0% and 100% maize in the diet corresponded very well with the measured values of forage mixtures and of maize (Table 2).

Figure 1.

Relationship between the δ13C of the whole diet and the percentage of maize in the diet.

Contrary to expectations, at least for farm M, on the basis of the δ15N values of the feed components, the δ15N values of the overall diet did not change with the level of maize (Table 3), and were slightly higher in farm C (C: 2.8 ± 0.3‰; M: 2.5 ± 0.6‰).

Table 3. Isotopic values of overall diet and milk (bulk, casein, lipid) (mean ± SD)
THESIS% MAIZESAMPLENδ13C (‰, V-PDB)δ15N (‰, AIR)δ18O (‰, V-SMOW)δD (‰, V-SMOW)
  OVERALL DIET2−24.0 ± 0.22.9 ± 0.722.3 ± 0.3−91 ± 0
023%MILK3  −7.1 ± 0.1 
  CASEIN3−20.3 ± 0.05.8 ± 0.114.9 ± 0.3−84 ± 2
  LIPID3−23.1 ± 0.0   
  OVERALL DIET2−20.1 ± 0.22.8 ± 0.023.6 ± 0.3−78 ± 2
142%MILK4−20.3 ± 0.05.0 ± 0.0−7.4 ± 0.2 
  CASEIN4−19.2 ± 0.25.4 ± 0.110.8 ± 0.4−104 ± 1
  LIPID4−21.7 ± 0.2   
  OVERALL DIET2−18.8 ± 0.12.9 ± 0.124.1 ± 0.3−76 ± 2
251%MILK4−19.7 ± 0.35.3 ± 0.1−7.2 ± 0.1 
  CASEIN4−18.3 ± 0.25.7 ± 0.110.9 ± 0.5−105 ± 2
  LIPID4−21.2 ± 0.1   
  OVERALL DIET2−18.6 ± 0.22.7 ± 0.324.3 ± 0.0−77 ± 2
355%MILK3  −5.8 ± 0.3 
  CASEIN3−17.5 ± 0.25.5 ± 0.115.9 ± 0.3−79 ± 2
  LIPID3−20.5 ± 0.1   
  OVERALL DIET2−24.5 ± 0.72.7 ± 1.522.7 ± 0.2−85 ± 0
028%MILK3  −6.8 ± 0.2 
  CASEIN3−20.6 ± 0.14.5 ± 0.0413.8 ± 0.4−89 ± 3
  LIPID3−22.0 ± 0.1   
  OVERALL DIET2−21.6 ± 0.12.4 ± 0.223.3 ± 0.5−81 ± 1
143%MILK4−20.3 ± 0.14.4 ± 0.1−7.3 ± 0.1 
  CASEIN4−19.5 ± 0.14.7 ± 0.110.5 ± 0.5−113 ± 3
  LIPID4−20.9 ± 0.1   
  OVERALL DIET2−20.1 ± 0.82.4 ± 0.423.6 ± 0.3−77 ± 2
253%MILK4−19.5 ± 0.04.4 ± 0.0−7.1 ± 0.2 
  CASEIN4−18.7 ± 0.14.7 ± 0.111.2 ± 0.3−109 ± 2
  LIPID4−19.8 ± 0.3   
  OVERALL DIET2−18.7 ± 0.62.7 ± 0.424.0 ± 0.0−77 ± 2
363%MILK3  −5.8 ± 0.3 
  CASEIN3−17.5 ± 0.04.6 ± 0.115.0 ± 0.1−85 ± 7
  LIPID3−18.2 ± 0.0   

With regard to the δ18O and δD values of the diet as a whole, δ18O was shown to be significantly correlated (p < 0.01) (Fig. 2(a)) with the amount of maize in the diet, even with different intercepts and slopes for the two farms. The extrapolation of the δ18O regression to 0% and 100% maize in the diet corresponded with the measured values of forage and of maize. The δD values showed a trend to increase according to the amount of maize, although not according to a linear relationship (Fig. 2(b)). The δD and δ18O values of feed ingredients and the diet as a whole were significantly correlated (p < 0.001) according to the following linear relationship:

equation image
Figure 2.

Relationship between the δ18O (a) and δD (b) values of the whole diet and the percentage of maize in the diet.

Calculating the values of the diet as a whole on the basis of the isotope mass balance mixing model (sum of the isotopic values of the feed ingredients multiplied by the relevant weight, expressed as dry matter in the diet), the following mean differences between the measured and calculated values were observed: +0.3 ± 0.3‰, +0.7 ± 0.6‰, 0.0 ± 0.2‰ and 3 ± 1‰ for δ13C, δ15N, δ18O and δD, respectively.

δ13C of milk

As already observed,24 the fat fraction of milk showed a lower δ13C value than the casein fraction (Table 3), a consequence of isotopic fractionation occurring during the synthesis of lipids in plants and animals.36 The δ13C values of bulk milk, measured for two out of four theses, were between those of casein and lipids.

The difference between casein and fat was 2.8 ± 0.3‰ for farm C whereas for M it was 1.2 ± 0.3‰. Milk proteins and lipids are produced by different metabolic pathways. Milk proteins are mainly produced by the mammary gland, starting from the amino acids present in the blood deriving from intestinal absorption of the digested proteins of dietary and microbial origin. Moreover, as stressed by Wilson et al.,24 around 20 to 30% of the carbon of casein derives from the body protein reserves. A significant proportion of lipids are derived from the body triglycerides and 40–50% are produced by the mammary gland, starting from the acetic and butyric acids deriving from rumen fermentation.37 A possible explanation for the differences between the two farms could be that the maize silage in farm C had lower digestibility and therefore lipids derived to a lower extent from maize. This is supported by the fact that the δ13C of lipids in milk from farm C is less influenced by the increase of percentage of maize in the diet (Fig. 3).

Figure 3.

Relationship between the δ13C values of milk casein and lipids and the percentage of maize in the diet.

Comparing the values of casein (Table 3) with those of the diet as a whole, an increase of about 4‰ for thesis 0, based more on C3 plants, and of about 1‰ for the other diets with a higher maize content was shown in both farms. A similar deviation has previously been observed in beef, in bulk milk and in CO2 in breath,17, 23, 25 and could be due to two reasons. First, differences in the content of fibre, cellulose,38 lignin, starch, etc., in C3 and C4 plants might lead to different digestibility and therefore preferential assimilation of a certain nutrient by the animals.4 Secondly, in the case of thesis 0, there might be an effect of the previous diet (thesis 1) because, in contrast to the findings of Wilson et al.,24 2 weeks are not long enough to obtain a complete turnover of carbon in casein. This is probably because the body proteins, which could contribute to up to 30% of casein production, need more than 230 days to achieve complete carbon turnover.17, 18 This interpretation is supported by the fact that extrapolation of the casein δ13C regression (Fig. 3) to 0% maize in the diet is quite high (>−23.5‰) compared with values in the literature.6, 24

Considering the isotopic shift of the δ13C in milk lipids, different values were evident for the two farms: for milk from farm C the δ13C of fat was almost always lower than that of the feed values,24 whereas, for milk from farm M, the δ13C of fat was always higher than that of the feed values. Possible explanations are discussed above.

There was a strong linear relationship (Fig. 3) between the δ13C of casein and the maize% in the diet (p < 0.001). As observed for the feed, the regression lines for the two farms had similar slopes and different intercepts, due to the C3 plant composition of the diet. The regression analysis indicated (95% confidence) that each 10% increase in maize content corresponded to a 0.7–1.0‰ shift in the δ13C of casein. The extrapolation of the δ13C regression to 0% and 100% maize in the diet corresponded (95% confidence) to a minimum value of −23.5‰ and to a maximum value of −14‰.

On the basis of these results and published values for milk casein from grazing cows (<−23.5‰, ranging from −24.8‰ to −27.1‰6, 24), we suggest −23.5‰ as a threshold limit, above which it is not possible to exclude the presence of maize in the diet. It is worth noting that this experiment was based on stable-reared cattle previously fed with a diet containing around 50% maize. Considering that a relatively long time is necessary to obtain a complete turnover of carbon in casein following a change of diet (due to the contribution of body protein), this value could be more appropriately applied to milk produced by stabled cows. Further investigations regarding grazing cows and increasing the adaptation time to at least 1 month are needed to confirm this value.

The δ13C values of fat also increased linearly with the% of maize (Fig. 3). The regression lines for the two farms had different slopes and intercepts, as already explained.

It is interesting to note that the slopes of the regression lines for farm M for both the ‘protein’ and the lipid fractions corresponded with those observed in the defatted meat material and in meat lipids.17 The slope value found for milk lipids from farm C was significantly (p < 0.001) lower, as described above.

δ15N of milk

The δ15N values of bulk milk, measured in two out of four theses (Table 3), were about 0.4‰ lower than those of casein,15 probably due to the contribution of whey protein, which has a lower content of δ15N, as demonstrated in previous studies (data not published).

Comparing the casein values with those of the feed, we observed an increase in the δ15N values of about 2.8‰ for farm C and 2.1‰ for farm M. This situation is in agreement with what has been observed in meat protein,39 even with less individual variability,17 in cheese casein,11 and in bulk milk.25 The different 15N enrichment observed in the two farms led to significantly (p < 0.001) different δ15N values of the relevant milk casein (C = 5.6 ± 0.2‰; M = 4.6 ± 0.1‰), which were not evident between the two overall diets. It has been reported that the δ15N of animal tissue is usually less influenced by specific dietary input and more variable due to metabolic factors.1, 39 Moreover, it has been stressed that the nutritional quality of dietary ingredients could affect diet-animal shifts: in general, better quality diets (i.e. with a lower C/N ratio) lead to a lower δ15N enrichment.4 Thus, the main reason for this shift could be the different assimilation rate of the dietary ingredients according to their different digestibility (Festuca and cotton were probably preferred to the other components).

As observed in the diet as a whole, the δ15N of milk casein was not influenced by the amount of maize in the diet (Table 3).

δ18O and δD of milk

Comparing the δ18O values of milk water with those of drinking water, we observed an increase of about 2‰ for theses 0, 1 and 2 and of about 3‰ for thesis 3 in relation to a higher content of maize silage (containing vegetal water enriched in 18O compared with the drinking water), in agreement with literature values.40, 41 This increase was justified by the fact that the milk water is derived not only from the drinking water, but also from the water produced by the animal endogenously through oxidation, when the oxygen from air and from organic matter in food is enriched in 18O42 and from the vegetal water of fresh plants ingested by the animal.26

The isotopic shift of the δ18O of casein compared with the feed values was about −13‰ for theses 1 and 2 and about −9‰ for theses 0 and 3, whereas for δD it was about −30‰ for theses 1 and 2 and from +6 to −8‰ for theses 0 and 3.

We did not observe a clear trend for increasing values in the δ18O of milk water and the δ18O and δD of casein according to the amount of maize in the diet, in contrast to the findings for the δD and δ18O of feed. Indeed higher values were found for thesis 3 (with the highest maize content) and for thesis 0 (without silage). The δ18O and δD values of casein were not correlated with the relevant values of the diet, but they corresponded more closely with the δ18O values of milk water (Fig. 4). The 18O and D content of milk casein, due to the isotopic fractionation effects of metabolic pathways and to the continuing exchange of D with the body water and milk water, is therefore not so much connected to the feed composition but rather influenced by the isotopic characteristics of the water taken from drinking water and from fresh plants or silage.

Figure 4.

Correlation of δ18O between casein and milk water.


We have demonstrated that casein could be a suitable component for evaluating the alteration of SIR with dietary composition. The δ13C values of the lipid fraction could also be influenced by other factors such as the digestability of forage and silage, and those of bulk milk also by its composition (e.g. fat content).

The amount of maize in the animal diet was clearly reflected in the δ13C of casein, with each 10% increase in the maize content of the diet resulting in a 0.7‰ to 1.0‰ shift in the δ13C of casein, similar to the findings for the δ13C of lipid-free muscle. A threshold value of −23.5‰ for the δ13C of casein was found as the limit above which it is not possible to exclude the presence of maize in the diet and this was shown to be particularly appropriate for stabled cows. A suitably designed experiment should be planned for grazing cows, reared in stables with the adaptation time being increased to at least 1 month, in order to obtain more information for milk from pasture.

The δ15N, δ18O and δD values of milk were shown to be not significantly influenced by the dietary composition, probably being more correlated with the geo-climatic and pedological characteristics of the area of origin, and, for δ18O and δD, with the presence of fresh diet or silage in the diet.

The study also allowed investigation of the relationships between the SIRs of different milk components, which could be useful for determining the unpermitted addition of exogenous fat or casein to milk; for example, in the case of reconstituted products.


The work was funded within the context of the “Analisi innovative per la determinazione dell'origine geografica di prodotti lattiero-caseari” ERSAF project – Lombardia Region, D.G.R. 17703 of 04/06/2004. We thank G.A. Zapparoli, M. Marchesi, L. Loatelli, P. Preus, A. Allegretti for supplying samples and A. Tonon for technical assistance.