Accepted Manuscript A metabolomics comparison between sheep's and goat's milk
P. Caboni, A. Murgia, A. Porcu, C. Manis, I. Ibba, M. Contu, P. Scano PII: DOI: Reference:
S0963-9969(18)30863-9 doi:10.1016/j.foodres.2018.10.071 FRIN 8041
To appear in:
Food Research International
Received date: Revised date: Accepted date:
24 July 2018 11 October 2018 25 October 2018
Please cite this article as: P. Caboni, A. Murgia, A. Porcu, C. Manis, I. Ibba, M. Contu, P. Scano , A metabolomics comparison between sheep's and goat's milk. Frin (2018), doi:10.1016/j.foodres.2018.10.071
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
A metabolomics comparison between sheep’s and goat’s milk
P. Cabonia,*
[email protected], A. Murgiaa , A. Porcua , C. Manis a , I. Ibbab, M. Contub and P. Scanoc, d a
Department of Life and Environmental Science, University of Cagliari, via Ospedale 72,
09124 Cagliari, Italy Regional Association of Sardinian farmers, Milk Analysis Laboratory, Loc. Palloni,
PT
b
Nuraxinieddu, 09170 Oristano, Italy
Department of Chemical and Geological Sciences, University of Cagliari, Cittadella
RI
c
Universitaria, SS 554 km 4.5, 09042 Monserrato, Cagliari, Italy
Institute for Macromolecular Studies, National Research Council, Via Corti 12, 20133
SC
d
Milan, Italy
Corresponding author: Department of Life and Environmental Science, University of
NU
*
MA
Cagliari, via Ospedale 72, 09124 Cagliari, Italy
Abstract
Despite the worldwide consumption of bovine milk, dairy products from small
ED
ruminants, such as goat’s and sheep’s milk, are gaining a large interest especially in the
EP T
Mediterranean area. The aim of this work was to study the metabolite profiles of 30 sheep’s and 28 goat’s milk using an untargeted metabolomics approach by a gas chromatography coupled with mass spectrometry (GC-MS) analysis. Results showed
AC C
several differences in the metabolite profiles: arabitol, citric acid, α˗ketoglutaric acid, glyceric acid, myo˗inositol, and glycine were more abundant in sheep’s milk, while goat’s milk had higher levels of mannose˗6˗phosphate, isomaltulose, valine, pyroglutamic acid, leucine, and fucose. Associations between metabolite profile and milk compositional traits were also found. Predictive capabilities of statistical models indicated a good correlation between the metabolite profile and the protein content in sheep’s milk, and with the fat content in goat’s milk. This work leads to a better understanding of milk metabolites in
ACCEPTED MANUSCRIPT
small ruminants and their role in the evaluation of milk properties. Keywords Mannose-6-phosphate Citrate
PT
Arabitol GC-MS
RI
Metabolomics
SC
Multivariate analysis. 1. Introduction
NU
Nowadays, the increased request for healthier products is generating new interest for milk
MA
from different species, such as sheep and goats. Compared to the rest of the world, where bovine milk represents the most suitable dairy source, the Mediterranean area is
ED
characterized by a significant number of goat and sheep herds designated to milk rather than meat production. For this reason, the European Community decided to protect the
EP T
uniqueness of goat and sheep dairy (Scintu & Piredda, 2007). Sardinia (Italy) is the leading Mediterranean region in the production of goat’s and sheep’s milk. In this region,
AC C
the yield and the composition of sheep’s and goat’s milk is strongly affected by their grazing on natural pastures of native flora (Scintu & Piredda, 2007). It has been amply demonstrated that the overall composition of sheep’s and goat’s milk is very different showing the former with a higher content of proteins, fats and lactose (Park, Juárez, Ramos, & Haenlein, 2007; Raynal-Ljutovac, Lagriffoul, Paccard, Guillet, & Chilliard, 2008) compared to the latter. Although several researches have widely studied the different aspects of small ruminant dairy product (Scintu & Piredda, 2007; Park et al., 2007; Raynal-Ljutovac, et al., 2008), in literature there is still a little information regarding
ACCEPTED MANUSCRIPT
the milk metabolite profiles (Scano, Murgia, Pirisi, & Caboni, 2014; Caboni, Murgia, Porcu, Demuru, Pulina, & Nudda, 2016; Caboni, Manis, Ibba, Contu, Coroneo, & Scano, 2017). The pool of low molecular weight metabolites in milk and dairy products represents the final point of the gene expression during lactation and throughout several metabolic
PT
pathways (Yang et al., 2016). The study of the levels of these compounds has been very useful to understand the metabolic pathways and characteristics of milk (Klein et al., 2010;
RI
Melzer et al., 2013a; Melzer, Wittenburg, & Repsilber, 2013b; Scano et al., 2014;
SC
Sundekilde et al., 2014; Murgia et al., 2016). Moreover, correlation analyses between metabolites and milk compositional traits helped to understand the biochemical and
NU
technological characteristics of milk, highlighting those molecular traits that can be eligible
MA
as biomarkers (Klein et al., 2010; Melzer et al., 2013a; Melzer et al., 2013b; Caboni et al., 2016; Scano et al., 2016; Caboni et al., 2017).
ED
One of the most valuable approaches for the investigation of metabolite profiles in a biological system is metabolomics (Wang et al., 2005). This technique allows the
EP T
determination of metabolites that are present at a low concentration in biological matrices (Zhang, Sun, Wang, Han, & Wang, 2012). The ability of GC-MS metabolomics
AC C
in the study of the metabolite profiles of milk from different species has been thoroughly demonstrated (Klein et al., 2010; Melzer et al., 2013a; Melzer et al., 2013b; Scano et al., 2014; Murgia et al., 2016; Caboni et al., 2016). In this study, by a metabolomics approach, the milk metabolite profiles of sheep and goats bred in Sardinia were investigated also in association with measured main milk traits such as the contents of fats, proteins, caseins, lactose, and urea, together with freezing point, pH values, and the somatic cells count (SCC).
ACCEPTED MANUSCRIPT
2. Materials and methods
2.1. Chemicals and reagents Solvents, such as methanol, chloroform and hexane were all GC/MS or LC/MS grade
chloride,
N-methyl-N-(trimethylsilyl)
trifluoroacetamide
PT
(Sigma Aldrich, Darmstadt, Germany). Pyridine, methoxamine hydrochloride, potassium (MSTFA)
and
analytical
RI
standards like pyruvic acid, lactic acid, alanine, glycine, urea, valine, leucine, proline,
SC
succinic acid, glyceric acid, uracil, threonine, pyroglutamic acid, creatinine, α-ketoglutaric acid, glutamic acid, ribose, arabitol, fucose, orotic acid, glycerol-3-phosphate, citric acid,
NU
hippuric acid, galactose, lysine, tyrosine, sorbitol, palmitic acid, scyllo-inositol, myo-
MA
inositol, n-acetyl-glucosamine, mannose-6-phosphate, uridine and isomaltulose presented a purity greater than 95% (Sigma Aldrich, Darmstadt, Germany). Water was produced using
EP T
2.2. Collection of samples
ED
a MilliQ purification system (Millipore, Milan, Italy).
We analysed 30 individual milk samples from Sarda breed sheep and 28 from Saanen
AC C
goats; the samples came from 14 different farms. The morning milk samples were obtained from grazing sheep and goats in the winter season, after their lambs and kids were weaned. Samples were collected in sterilised tubes through manual milking and stored at -20 °C until analysis which was performed within 2 days from collection. The milk samples were provided by the Sardinian Regional Animal Farmers Association (Associazione Regionale Allevatori, ARA, Sardegna) thanks to the International Committee for Animal Recording (ICAR) program.
ACCEPTED MANUSCRIPT
2.3. GC-MS analysis Thawed milk (1 mL) was subjected to ultrasounds for 15 min, and 0.1 mL were extracted following the Folch procedure (Folch, Lees, & Sloane-Stanley, 1957) using 0.375 mL of a methanol and chloroform mixture (2/1, v/v). Samples were then vortexed every 15 min up
PT
to 1h, when 0.38 mL of chloroform and 0.09 mL of aqueous KCl 0.2 M solution were subsequently added to them. Samples where then vortexed again and centrifuged for 15
RI
min at 5139 xG at 4°C. Two hundred µL of the hydrophilic supernatant were dried in glass
SC
vials using a nitrogen stream and then derivatized using 0.05 mL of methoxamine chloride dissolved in pyridine at 10 mg/mL and left at room temperature overnight. The day after,
NU
70 µL of MSTFA were added to the samples and samples were left at room temperature up
MA
to 1 h, when 500 µL of a 5 mg/mL of 2,2,3,3-d4-succinic acid hexane solution were added as an internal standard. A Hewlett Packard 6850 gas chromatograph, a 5973 mass selective detector, and a 7683B series injector (Agilent Technologies, Palo Alto, CA) were used to
ED
analyse samples using Helium as the carrier gas at 1.0 mL/min flow. One μL samples were
EP T
injected in split-less mode and separated using a 30 m × 0.25 mm × 0.25 μm DB-5MS column (Agilent Technologies, Palo Alto, CA). The temperatures for the inlet, interface,
AC C
and ion source were 250, 250 and 230°C, respectively. The oven temperature was programmed to increase from 50 to 230°C by 5°C/min over 36 min and kept at this temperature for 2 min. The mass range was set between 50 and 550 m/z using and electron voltage of 70. MSD ChemStation software (Agilent Technologies, Santa Clara, CA) was used to elaborate the data. The GC-MS spectra deconvolution was performed by the AMDIS tool in the NIST08 library. For each metabolite, the GC-MS peak area related to the most abundant mass fragment, calculated using the MSD Chemstation, was choosed for the data tabulation. Retention times and relative mass spectra of each compound were
ACCEPTED MANUSCRIPT
compared with the related analytical standards with the aim to identify those compounds previously recognised by consulting the NIST08 library of the National Institute of Standards and Technology (Gaithersburg, MD), and the library developed at the Max Planck
Institute
of Golm (http://gmd.mpimp-golm.mpg.de/).
Retention indexes were
PT
calculated according to Kovats, using a C9-C24 series of alkanes. The relative levels of
RI
metabolites within each sample were normalized for the fat free dry matter.
SC
2.4. Milk composition analysis
Fat, proteins, caseins, lactose, and urea contents, freezing points (expressed in Horvet
NU
degree), and pH values were obtained with a Milkoscan FT6000 (Foss, Denmark)
MA
following the procedure ISO 9622:2013 (ISO 9622:2013 IDF 141:2013). The fatty acid (FA) content was determined by MIR spectroscopy (Caredda et al., 2016). Somatic cells
ED
were counted following the procedure ISO 13366-2-2006 (ISO 13366-2 IDF 148-2 2006)
EP T
by a Fossomatic 5000 (Foss, Denmark).
2.5. Statistical data analysis
AC C
To obtain descriptive information on the data, and to calculate the means and their standard error (SEM),
we performed
univariate analysis with the software OriginPro2016
(OriginLab, Northampton, MA). When the two groups of samples were compared, the null hypothesis (“the means are not significantly different among the two sets of samples”) was tested by using the unpaired samples two-tailed Student t-test. Linear correlations between two variables were calculated using the Pearson correlation coefficient r, and those having -0.75 > r > +0.75 and with a p-value < 0.05 were taken as significant. MVA was performed as implemented in SIMCA-P+ software (version 14.1, Umetrics, Umeå, Sweden).
ACCEPTED MANUSCRIPT
Variables were mean centered and scaled to unit variance. At first, for sample distribution overview, to detect outliers, deviating features and common trends, we performed a Principal Component Analysis (PCA). Subsequently, to identify discriminant metabolites between the two milks’ typologies, we performed an Orthogonal Partial Least-Squares -
PT
Discriminant Analysis (OPLS-DA) The obtained variable importance in projection (VIP) scores summarize the contribution of each variable to the model. In this work, we analyzed
RI
VIP scores in the predictive component of OPLS-DA and only those metabolites having
SC
VIP values > 1 were deemed as discriminant between the classes. Milk metabolite profiles were correlated to the measured main traits by a single-Y Orthogonal extension of Partial
NU
Least-Squares (OPLS) regression, in order to highlight the metabolites mostly involved in
MA
each milk trait. The quality of the models was evaluated based on the cumulative parameters R2 X, R2 Y, and Q 2 Y estimated by the default leave-1/7th-out cross validation
ED
and also tested for overfitting using y-table permutation test (permutations n = 400) as implemented in SIMCA-P+ program (Eriksson, Byrne, Johansson, Trygg, & Vikström,
AC C
Q 2 Y was < 0.50.
EP T
2013). The models were considered significant only when the difference between R2 Y and
3. Results and discussion By GC˗MS analysis of the methanol fraction of milk samples, we identified 38 low molecular weight metabolites (Table 1S). These compounds were mainly short chain hydroxylated carboxylic acids, organic acids, amino acids, polyols and sugars. Initially, the GC-MS dataset was submitted to a PCA. As shown in the score plot of Figure 1A, in the space of the first 2 principal components, sheep’s and goat’s milk samples clustered in two different groups, thus indicating a clear difference in their metabolite profiles (loading plot
ACCEPTED MANUSCRIPT
in Figure 1B). In particular, sheep’s milk samples showed a greater variability being, in the score plot, more spread than goat’s samples (Figure 1A). In order to find discriminating metabolites between sheep’s and goat’s milk, we performed a pair-wise OPLS-DA. The discriminant analysis correctly classified our samples (R2 Y=0.92, Q2 Y=0.88). As reported
PT
from the VIP’s analysis (Table 1), sheep’s milk was richer in α-ketoglutarate when compared to goat’s milk. α-ketoglutarate plays a crucial role in several metabolic pathways
RI
derived either from the Krebs cycle or from the rumen anaerobic bacteria activity
SC
(Bacteriodes) (Allison, Robinson, & Baetz, 1979). Increased levels of this compound have been associated with a number of health benefits (Andersen, Tatara, Krupski, Majcher, &
NU
Harrison, 2008) such as decreased cholesterol levels in rats (Radzki, Bieńko, &
MA
Pierzynowski, 2009), protection of gut mucosa and improved kidney function (Dąbek et al., 2005). Furthermore, α-ketoglutarate highly enhances the degradation of aromatic and branched-chain amino acids, which strongly contribute to the aroma of several cheese
ED
(Yvon, Berthelot, & Gripon, 1998). Furthermore, sheep’s milk contained higher level of
EP T
myo-inositol. This polyol is an important constituent of living cells, being the precursor to phosphatidylinositol, and is recognized as a lipotropic nutrient. Milk is a significant source
AC C
of myo-inositol and for its role in neonatal nutrition, this polyol is often added to infant formulas to prevent a potential deficiency during early neonatal growth stage (Woollard, Macfadzean, Indyk, McMahon, & Christiansen, 2014). Citrate, which was found in higher level in sheep’s milk samples, together with free amino acids, has been suggested as the nutrient source used by lactic acid bacteria (LAB) and nonstarter LAB during cheese making (Williams, Withers, and Banks, 2000). Sheep’s milk had also a higher content of arabitol, probably produced by Kluyveromyces lactis from lactose (Toyoda & Ohtaguchi, 2011), yeast strain found in Sardinian sheep’s milk and dairy products (Fadda, Mossa,
ACCEPTED MANUSCRIPT
Pisano, Deplano, & Cosentino, 2004). Compared to sheep’s samples, goat’s milk showed higher levels of mannose-6phospate
and
fucose.
Mannose-6-phosphate
is implicated
in the biosynthesis of
glycoproteins and glycophospholipids and participate to the glycolytic pathway (Kalhan,
PT
2009). Fucose is a monosaccharide which promotes the neonatal bifidobacterial growth, the intestinal mucosa cell protection against pathogens, and has a role in the neonatal brain
RI
development (Raynal-Ljutovac et al., 2008). Mannose-6-phospate and fucose are involved
SC
in the synthesis of oligosaccharides. Compared to other ruminant’s milk, goat’s milk is the richest in oligosaccharides (Urashima, Taufik, Fukuda, & Asakuma, 2013) which share
NU
similar structural elements with human milk (Urashima et al., 2013). The presence of
MA
fucosylated oligosaccharides was reported in milk from O/O genotype goats (α-s1-casein non-producers) (Meyrand, Dallas, Caillat, Bouvier, Martin, & Barile, 2013). In a previous study,
trying to
identify milk metabolite differences between goats with different
ED
expression of αS1-casein genotypes, we found that fucose and mannose-6-phosphate
EP T
contents were statistically increased in α-s1-casein weak null allele class when compared with the strong homozygous genotype AA, indicating changes in the carbohydrate
AC C
metabolism (Caboni et al., 2016). In the present work, we found that, in goat’s milk, fucose was correlated with galactose (r = 0.76), a milk sugar component of the oligosaccharide backbone.
Goat’s milk levels of valine and leucine resulted higher compared with sheep’s milk. These two essential branched amino acids are involved in the metabolic pathways for the production of branched-chain fatty acids (Massart-Leёn & Massart, 1981) contributing to the flavour of goat’s dairy products (Chilliard, Ferlay, Rouel & Lamberet, 2003). Pyroglutamic acid, produced by thermophilic LAB starters, also has a strong impact on
ACCEPTED MANUSCRIPT
cheese flavour. Pyroglutamic acid was correlated with glutamine (r = 0.78) and tyrosine (r = 0.77), suggesting that it is involved in the metabolism of aromatic amino acids. The level of isomaltulose, a naturally occurring disaccharide, composed of α-1,6-linked glucose and fructose, was higher in goat’s milk than in sheep’s milk, Isomaltulose is a digestible
PT
carbohydrate that produces glucose in a longer period (Lina, Jonker, & Kozianowski, 2002) and is completely hydrolyzed and absorbed by the small intestine (Holub et al.,
RI
2010). In our previous studies, comparing cow’s milk-based infant formula, human breast
SC
milk and donkey’s milk (Murgia et al., 2016; Scano et al., 2016), isomaltulose was
possible biomarker for ruminant’s milks.
NU
detected only in cow’s milk-based formula, leading us to consider this compound as a
MA
In this work, we studied the milk content of proteins, caseins, fat, lactose, fatty acids and other parameters, hereafter expressed as mean ± SEM (Table 2). Accordingly with previous results (Raynal-Ljutovac et al., 2008; Park et al., 2007), we found that
ED
sheep’s milk compared to goat’s milk had significantly higher contents of proteins (5.71 ±
EP T
0.18 and 3.80 ± 0.23 g/100 mL, for sheep and goat’s milk, respectively), caseins (4.20 ± 0.07 and 2.87 ± 0.07 g/100 mL) and lactose (5.02 ± 0.03 and 4.75 ± 0.04 g/100 mL).
AC C
Among fatty acids (FA), due to their health benefits (Júnior et al., 2012), a particular attention has been paid to rumenic acid (18:2 cis-9, trans-11), trans-vaccenic acid (18:1 trans-11) and α-linolenic acid (18:3 cis-9,12,15). Milk and milk-derivatives, are the only food products that naturally contain trans-vaccenic acid and rumenic acid. The ruminal biohydrogenation
and
the
mammary
lipogenic
and
Δ-9
desaturation
pathways,
considerably modifies the profile of dietary FA and the milk composition (Scintu & Piredda, 2007). The pasture plays a pivotal role in increasing levels of trans-vaccenic acid, rumenic acid and α-linolenic acid. It is well established that the concentration of rumenic
ACCEPTED MANUSCRIPT
acid is higher in the milk of free-grazing ruminants compared to animals fed on conserved forages (Scintu & Piredda, 2007). Usually, under the same diet, sheep’s milk has a higher rumenic acid content than goat’s milk (Zervas & Tsiplakou, 2011). According to these observations, we found that the contents of trans-vaccenic acid and rumenic acid, as well
PT
as α-linolenic acid, were statistically higher in sheep’s milk (3.36 ± 0.16, 1.52 ± 0.05, and 1.14 ± 0.04 % of FA, respectively) than in goat’s milk (1.48 ± 0.01, 0.91 ± 0.05, and 0.62 ±
RI
0.07 % of FA, respectively) (Table 2). Similar contents of trans-vaccenic acid and rumenic
SC
acid were reported for the Sardinian sheep by Nudda et al. (Nudda, McGuire, Battacone, & Pulina, 2005).
NU
The content and the characteristics of somatic cells in milk are related to the
MA
animal’s health (Pirisi et al., 2007), season, stage of lactation and parity (Raynal-Ljutovac, Pirisi, De Cremoux, & Gonzalo, 2007) and differ between cows, sheep and goats (Paape, Poutrel, Contreras, Marco, & Capuco, 2001). In our work, in agreement with the literature,
ED
goat’s milk showed a higher SCC mean value (789 ± 128 × 10 3 cell/mL) when compared
EP T
with sheep’s milk (423 ± 103 × 103 cell/mL) (Paape, et al., 2001). As already observed by Raynal-Ljutovac et al. (2008), goat’s milk contained
AC C
significantly higher concentration of chlorides compared to sheep’s milk (239.1 ± 8.2 and 109.4 ± 4.1 mg/100 mL), that, together with lower levels of lactose, confers a slightly salty taste to goat’s milk (Ribeiro & Ribeiro, 2010). As found in a previous study (Caboni et al., 2017), the content of chloride ions in goat’s milk samples was negatively correlated (r = 0.88) with the lactose content. Lactose and chloride are osmolytes and thus influence the milk’s freezing point. To keep milk isotonic to blood, the lower levels of lactose in goat’s milk compared to sheep’s milk were counterbalanced by higher levels of chloride ions (Caboni et al., 2017).
ACCEPTED MANUSCRIPT
By applying OPLS analysis, we tested the correlations between the GC-MS metabolite profile and the measured milk traits, and the results are reported in Table 2S and shown in Figure 2. Overall, the models for goat’s milk had better performances than those for sheep’s milk. The metabolite profile showed a stronger correlation to the protein
PT
content in sheep’s milk (Figure 2B), and to the fat content in goat’s milk (Figure 2D). The SCC is well predicted in both milks (Figure 2C and 2F). Interestingly, models for urea
RI
content failed to perform well (Table 2S). In Table 3, we reported the metabolites mostly
SC
correlated with protein, fat and SCC contents in sheep’s and goat’s milk, and we can see that they are quite different between the two typologies of milk, thus indicating different
NU
metabolic routes for milk production in these small ruminants. Metabolites associated with
MA
both protein and fat levels were glycine, glyceric acid, galactose, alanine, and fucose for sheep’s, and hippuric acid and inositols for goat’s milk. The presence of hippuric acid in
dietary
regimens
ED
goat’s milk has been proposed as a possible marker to discriminate between different (free-grazing
vs.
conserved
forage)
(Carpio
et
al.,
2013).
EP T
Phosphoethanolamine correlated with fat in both milks, and in goat’s milk, also scylloinositol and myo-inositol were associated with fat (Table 3). The levels of these
AC C
compounds can be linked with the pathway involved in the production of phospholipids, this is also supported by the correlation in goat’s milk between scyllo-inositol and the polyunsaturated fatty acid content (r = 0.75), which are components of phospholipids. The OPLS results indicated that the metabolite profile was also associated with the SCC (Table 2S). Ribose is a component of nucleic acids and its association to the SCC in sheep’s milk may reflect the presence of cell fragments in the milk. Moreover, in sheep’s milk, correlations of different amino acids (tyrosine, lysine, valine, glutamic acid) with the SCC model (Table 3) can be a consequence of the release of proteolytic enzymes from somatic
ACCEPTED MANUSCRIPT
cells (Li, Richoux, Boutinaud, Martin, & Gagnaire, 2014). In goat’s milk, metabolites associated with SCC values were pyruvic acid and lactic acid. A relationship between lactic acid levels and SCC values has been reported for cow’s milk (Hamann, 2002; Davis et al., 2004; Melzer et al., 2013a) and lactic acid was proposed as a biomarker candidate
PT
for mastitis even in the early onset of the disease (Farr, Prosser, Nicholas, Lee, Hart, &
RI
Davis, 2002).
SC
4. Conclusions
In summary, by this GC-MS based metabolomics approach, it was possible to highlight the
NU
metabolite differences between sheep’s and goat’s milk, and the possible different
MA
metabolic routes for milk production in these small ruminants. The associations between metabolites and milk main traits support the potential use of the metabolite profile as a
ED
predictive biomarker in different areas such as the dairy industry. Although not taking into account the effect of season variability in the levels of milk metabolites, the promising
EP T
results of this work encourage further studies in this direction. Acknowledgements
AC C
This study was supported by grants provided by Regione Autonoma della Sardegna within the frame of the project “Valorizzazione e innovazione delle produzioni lattiero-casearie caprine in Sardegna. Indagini chimico-fisiche e nutrizionali” (Legge Regionale 7 Agosto 2007, N.7). We also would like to thank Mariano Medda for proof editing. References 1. Allison, M. J., Robinson, I. M., & Baetz, A. L. (1979). Synthesis of alphaketoglutarate by reductive carboxylation of succinate in Veillonella, Selenomonas, and Bacteriodes species. Journal of Bacteriology, 140, 980-986.
ACCEPTED MANUSCRIPT
2. Andersen, N. K., Tatara, M. R., Krupski, W., Majcher, P., & Harrison, A. P. (2008). The long‐term effect of α‐ketoglutarate, given early in postnatal life, on both growth and various bone parameters in pigs. Journal of Animal Physiology and Animal Nutrition, 92, 519-528.
PT
3. Caboni, P., Murgia, A., Porcu, A., Demuru, M., Pulina, G., & Nudda, A. (2016). Gas chromatography-mass spectrometry metabolomics of goat milk with different
RI
polymorphism at the α S1-casein genotype locus. Journal of Dairy Science, 99,
SC
6046–6051.
4. Caboni, P., Manis, C., Ibba, I., Contu, M., Coroneo, V., & Scano, P. (2017), with a high somatic cell count: A
NU
Compositional profile of ovine milk
MA
metabolomics approach. International Dairy Journal, 69, 33-39. 5. Carpio, A., Bonilla-Valverde, D., Arce, C., Rodríguez-Estévez, V., Sánchez-
ED
Rodríguez, M., Arce, L., & Valcárcel, M. (2013). Evaluation of hippuric acid content in goat milk as a marker of feeding regimen. Journal of Dairy Science, 96,
EP T
5426-5434.
6. Caredda, M., Addis, M., Ibba, I., Leardi, R., Scintu, M. F., Piredda, G., & Sanna, G.
AC C
(2016). Prediction of fatty acid content in sheep milk by Mid-Infrared spectrometry with a selection of wavelengths by Genetic Algorithms. LWT-Food Science and Technology, 65, 503-510. 7. Chilliard, Y., Ferlay, A., Rouel, J., & Lamberet, G. (2003). A review of nutritional and physiological factors affecting goat milk lipid synthesis and lipolysis. Journal of Dairy Science, 86, 1751-1770. 8. Dąbek,
M.,
Kruszewska,
D.,
Filip,
R.,
Hotowy,
A.,
Pierzynowski,
Ł.,
Wojtasz‐Pająk, A., S. Szymanczyk, Valverde Piedra, J. L., Werpachowska, E., &
ACCEPTED MANUSCRIPT
Pierzynowski, S. G. (2005). α‐Ketoglutarate (AKG) absorption from pig intestine and plasma pharmacokinetics.
Journal of Animal Physiology and Animal
Nutrition, 89, 419-426. 9. Davis, S. R., Farr, V. C., Prosser, C. G., Nicholas, G. D., Turner, S. A., Lee, J., &
PT
Hart, A. L. (2004). Milk L-lactate concentration is increased during mastitis. Journal of Dairy Research, 71, 175-181.
RI
10. Eriksson, L., Byrne, T., Johansson, E., Trygg, J., & Vikström, C. (2013). Multi-and
SC
megavariate data analysis basic principles and applications (Vol. 1). Umetrics Academy, Umea, Sweden.
NU
11. Farr, V. C., Prosser, C. G., Nicholas, G. D., Lee, J., Hart, A. L., & Davis, S. R.
MA
(2002). Increased milk lactic acid concentration is an early indicator of mastitis. In Proceedings-New Zealand Society of Animal Production (Vol. 62, pp. 22-23). New
ED
Zealand Society of Animal Production; 1999. 12. Fadda, M. E., Mossa, V., Pisano, M. B., Deplano, M., & Cosentino, S. (2004).
EP T
Occurrence and characterization of yeasts isolated from artisanal Fiore Sardo cheese. International Journal of Food Microbiology, 95, 51-59.
AC C
13. Folch, J., Lees, M., & Sloane Stanley, G. H. (1957). A simple method for the isolation and purification of total lipids from animal tissues. Journal of Biological Chemistry, 226, 497-509. 14. Hamann,
J.
(2002).
Relationships
between
somatic
cell count
and
milk
composition. Bulletin of the International Dairy Federation, 372, 56-59. 15. Holub, I., Gostner, A., Theis, S., Nosek, L., Kudlich, T., Melcher, R., & Scheppach, W. (2010). Novel findings on the metabolic effects of the low glycaemic carbohydrate isomaltulose (Palatinose). The British Journal of Nutrition, 103,
ACCEPTED MANUSCRIPT
1730–1737. 16. ISO 9622:2013 (IDF 141:2013), Milk and liquid milk products - Guidelines for the application of mid-infrared spectrometry. 17. ISO 13366-2:2006 (IDF 148-2:2006). Milk - Enumeration of somatic cells - Part 2:
PT
Guidance on the operation of fluoro-opto-electronic counters. 18. Li, N., Richoux, R., Boutinaud, M., Martin, P., & Gagnaire, V. (2014). Role of
RI
somatic cells on dairy processes and products: a review. Dairy Science &
SC
Technology, 94, 517-538.
19. Júnior, O. S., Pedrao, M. R., Dias, L. F., Paula, L. N., Coro, F. A. G., & De Souza, E.
(2012).
Fatty acid content of bovine milkfat from raw milk to
NU
N.
MA
yoghurt. American Journal of Applied Sciences, 9, 1300-1306. 20. Klein, M. S., Almstetter, M. F., Schlamberger, G., Nürnberger, N., Dettmer, K.,
ED
Oefner, P. J., Meyer, H.H.D., Wiedemann, S., & Gronwald, W. (2010). Nuclear magnetic resonance and mass spectrometry-based milk metabolomics in dairy cows
EP T
during early and late lactation. Journal of Dairy Science, 93, 1539-1550. 21. Lina, B. A. R., Jonker, D., & Kozianowski, G. (2002). Isomaltulose (Palatinose®):
AC C
a review of biological and toxicological studies. Food and Chemical Toxicology, 40, 1375-1381.
22. Massart-Leën, A. M., & Massart, D. L. (1981). The use of clustering techniques in the elucidation or confirmation of metabolic pathways. Application to the branched-chain fatty acids present in the milk fat of lactating goats. Biochemical Journal, 196, 611-618. 23. Kalhan, S. C. (2009). Nonglucose carbohydrates and infant nutrition and metabolism. The Journal of nutrition, 139, 1611-1612.
ACCEPTED MANUSCRIPT
24. Melzer, N., Wittenburg, D., Hartwig, S., Jakubowski, S., Kesting, U., Willmitzer, L., Lisec, J., Reinsch, N., & Repsilber, D. (2013a). Investigating associations between milk metabolite profiles and milk traits of Holstein cows. Journal of Dairy Science, 96, 1521-1534.
PT
25. Melzer, N., Wittenburg, D., & Repsilber, D. (2013b). Integrating milk metabolite profile information for the prediction of traditional milk traits based on SNP
RI
information for Holstein cows. PloS one, 8, e70256.
SC
26. Meyrand, M., Dallas, D. C., Caillat, H., Bouvier, F., Martin, P., & Barile, D. (2013). Comparison of milk oligosaccharides between goats with and without the
NU
genetic ability to synthesize α s1-casein. Small Ruminant Research, 113, 411-420.
MA
27. Murgia, A., Scano, P., Contu, M., Ibba, I., Altea, M., Bussu, M., Demuru, M., Porcu, A., & Caboni, P. (2016). Characterization of donkey milk and metabolite
ED
profile comparison with human milk and formula milk. LWT-Food Science and Technology, 74, 427-433.
EP T
28. Nudda, A., McGuire, M. A., Battacone, G., & Pulina, G. (2005). Seasonal variation in conjugated linoleic acid and vaccenic acid in milk fat of sheep and its transfer to
AC C
cheese and ricotta. Journal of Dairy Science, 88, 1311-1319. 29. Paape, M. J., Poutrel, B., Contreras, A., Marco, J. C., & Capuco, A. V. (2001). Milk somatic cells and lactation in small ruminants. Journal of Dairy Science, 84, E237-E244.
30. Park, Y. W., Juárez, M., Ramos, M., & Haenlein, G. F. W. (2007). Physicochemical characteristics of goat and sheep milk. Small Ruminant Research, 68, 88113. 31. Pirisi, A., Lauret, A., & Dubeuf, J. P. (2007). Basic and incentive payments for goat
ACCEPTED MANUSCRIPT
and sheep milk in relation to quality. Small Ruminant Research, 68, 167-178. 32. Radzki, R. P., Bieńko, M., & Pierzynowski, S. G. (2009). Effect of dietary alpha‐ketoglutarate
on
blood
lipid
profile
during
hypercholesterolaemia
in
rats. Scandinavian Journal of Clinical and Laboratory Investigation, 69, 175-180.
PT
33. Raynal-Ljutovac, K., Pirisi, A., De Cremoux, R., & Gonzalo, C. (2007). Somatic
aspects. Small Ruminant Research, 68, 126-144.
RI
cells of goat and sheep milk: analytical, sanitary, productive and technological
SC
34. Raynal-Ljutovac, K., Lagriffoul, G., Paccard, P., Guillet, I., & Chilliard, Y. (2008). Composition of goat and sheep milk products: An update. Small Ruminant
NU
Research, 79, 57-72.
MA
35. Ribeiro, A. C., & Ribeiro, S. D. A. (2010). Specialty products made from goat milk. Small Ruminant Research, 89, 225-233.
ED
36. Scano, P., Murgia, A., Pirisi, F. M., & Caboni, P. (2014). A gas chromatographymass spectrometry-based metabolomic approach for the characterization of goat
EP T
milk compared with cow milk. Journal of Dairy Science, 97, 6057-6066. 37. Scano, P., Murgia, A., Demuru, M., Consonni, R., & Caboni, P. (2016). Metabolite
AC C
profiles of formula milk compared to breast milk. Food Research International, 87, 76-82.
38. Scintu, M. F., & Piredda, G. (2007). Typicity and biodiversity of goat and sheep milk products. Small Ruminant Research, 68, 221-231. 39. Sundekilde, U. K., Gustavsson, F., Poulsen, N. A., Glantz, M., Paulsson, M., Larsen, L. B., & Bertram, H. C. (2014). Association between the bovine milk metabolome and rennet-induced coagulation properties of milk. Journal of Dairy Science, 97, 6076-6084.
ACCEPTED MANUSCRIPT
40. Toyoda, T., & Ohtaguchi, K. (2011). D‐arabitol production from lactose by Kluyveromyces
lactis
at
different
aerobic
conditions. Journal of
Chemical
Technology and Biotechnology, 86, 217-222. 41. Urashima, T., Taufik, E., Fukuda, K., & Asakuma, S. (2013). Recent advances in
PT
studies on milk oligosaccharides of cows and other domestic farm animals. Bioscience, Biotechnology, and Biochemistry, 77, 455-466.
RI
42. Wang, M., Lamers, R. J. A., Korthout, H. A., van Nesselrooij, J. H., Witkamp, R.
SC
F., van der Heijden, R., Voshol, P. J., Havekes, L. M., Verpoorte, R., & van der Greef, J. (2005). Metabolomics in the context of systems biology: bridging Chinese
medicine
molecular
pharmacology. Phytotherapy
MA
Research, 19, 173-182.
and
NU
traditional
43. Weckwerth, W. (2010). Metabolomics: an integral technique in systems biology.
ED
Bioanalysis, 2, 829-836.
44. Williams, A. G., Withers, S. E., & Banks, J. M. (2000). Energy sources of non-
EP T
starter lactic acid bacteria isolated from Cheddar cheese. International Dairy Journal, 10, 17-23.
AC C
45. Woollard, D. C., Macfadzean, C., Indyk, H. E., McMahon, A., & Christiansen, S. (2014). Determination of myo-inositol in infant formulae and milk powders using capillary gas chromatography with flame ionisation detection. International Dairy Journal, 37, 74-81. 46. Yang, Y., Zheng, N., Zhao, X., Zhang, Y., Han, R., Yang, J., Shengguo, Z, Songli, L, Tongjun, G, Changjiang, Z, & Wang, J. (2016). Metabolomic biomarkers identify differences in milk produced by Holstein cows and other minor dairy animals. Journal of Proteomics, 136, 174-182.
ACCEPTED MANUSCRIPT
47. Yvon, M., Berthelot, S., & Gripon, J. C. (1998). Adding α-ketoglutarate to semihard cheese curd highly enhances the conversion of amino acids to aroma compounds. International Dairy Journal, 8, 889-898. 48. Zervas, G., & Tsiplakou, E. (2011). The effect of feeding systems on the
PT
characteristics of products from small ruminants. Small Ruminant Research, 101, 140-149.
RI
49. Zhang, A., Sun, H., Wang, P., Han, Y., & Wang, X. (2012). Modern analytical
AC C
EP T
ED
MA
NU
SC
techniques in metabolomics analysis. Analyst, 137, 293-300.
ACCEPTED MANUSCRIPT
Figure 1. PCA of GC-MS data of milk samples. (A) Score plot with samples projected in the first two principal components (PC). Explained variance=36%, the ellipse depicted in dashed line encloses the 95% Hotelling T2 confidence region. Sheep’s milk samples (n=30) are shown as grey boxes and goat’s milk samples (n=28) as empty circles. (B) Loading
PT
plot. Only identified metabolites are shown and abbreviated as proposed in Table 1S. Figure 2. OPLS correlation plots between experimental (y-axis) and predicted (x-axis) milk
RI
traits. A) fat, B) proteins, C) log somatic cell count (SCC), for sheep’s milk samples (n=30); D)
SC
fat, E) proteins, F) log somatic cell count (SCC), for goat’s milk (n=28). Results of OPLS
AC C
EP T
ED
MA
NU
models are reported in Table 2S.
ACCEPTED MANUSCRIPT
HIGHLIGHTS Goat’s and sheep’s milk are gaining interest for nutritional and health effects
Metabolomics has been applied to study dairy products
Milk metabolites can reflect the physiological mechanism of milk production
Sheep’s milk was found richer in myo-inositol
Metabolite profiles were associated to milk traits, such as fat, proteins and SCC
AC C
EP T
ED
MA
NU
SC
RI
PT