Untargeted metabolic profiling of ovine and bovine serum at mid-lactation
To detect the metabolic differences between cow, goat and sheep milk, an untargeted metabolomics analysis was performed and Human Metabolome Database (HMDB) was used to annotation. In total, 313 annotated metabolites from 1050 positive-ion feature and 173 annotated metabolites from 565 negative-ion feature were identified (Table S1). The results showed that the largest metabolic category was lipids and lipid-like molecules (188 metabolites), followed by organic acids and derivatives (98 metabolites) and organoheterocyclic compounds (55 metabolites) (Table S2).
In the positive-ion mode, the top 3 metabolites of ovine serum were Platelet-activating factor, Betaine and callystatin A and the top 3 metabolites of bovine serum were Hippuric acid, callystatin A and Platelet-activating factor. In the negative-ion mode, the top 3 metabolites of ovine serum were Oleic acid, Stearic acid and Ethyl myristate and the top 3 metabolites of bovine serum were Stearic acid, Ethyl myristate, Cholic acid. The results showed that high level of long-chain fatty acid at serum which could supply the formation of butterfat.
Multivariate Statistical Analysis
Principal Components Analysis (PCA) was used to determine the sample separation and aggregation between three milks. Each point on the PCA score graph represents a single sample. Aggregation of points indicates that the observed variables are highly similar, and discrete points represent significant differences (VIP ≥ 1; ratio ≥ 2 or ratio ≤ 1/2; q ≤ 0.05) in the observed variables. In the positive-ion mode, the PCA scores illustrated that PC1 and PC2 were responsible for 53.25 and 17.79% of the variation, respectively (Figure 1A). In the negative-ion mode, the PCA scores revealed that PC1 and PC2 were responsible for 54.35 and 18.51% of the variation, respectively (Figure 1B). The results demonstrated that serum from different species had different metabolic characteristics.
To identify specific differences between groups, partial least squares discrimination analysis (PLS-DA) was used. Higher values for PLS-DA model parameters (R2 and Q2) denote greater reliability for the PLS-DA model. In the positive-ion mode, R2 of the PLS-DA model was 1.00, and Q2 was 0.99 (Figure 2A). Coincidentally, R2 of the PLS-DA model was 1.00 and Q2 was 0.99 in the negative-ion mode (Figure 2B). The results indicated that both R2 and Q2 were high and subsequent analyses were credible.
Differential metabolites analysis
Next, we subjected the metabolomics data to univariate analysis of fold changes and T statistical testing to perform Benjamini-Hochberg correction and obtain the P-value. This was combined with multivariate statistical analysis of the VIP obtained via PLS-DA to screen for differential metabolites. Differential ions were defined as follows: VIP ≥ 1; ratio ≥ 2 or ratio ≤1/2; P ≤ 0.05. 269 and 143 metabolites were identified as differential metabolites in positive-ion and negative-ion modes, separately (Figure 3). In the positive-ion mode, 113 metabolites present higher level in ovine serum while 156 metabolites present higher level in bovine serum (Table S3). And 38 metabolites present higher level in ovine serum while 105 metabolites present higher level in bovine serum in the negative-ion mode (Table S3). The top 5 significant abundant metabolites of ovine serum were LAPPAOL C, 2-ETHYL-4,5-DIMETHYLOXAZOLE, N-C18:0 Phytoceramide, (2S)-2-Amino-8-hydroxyoctanoic acid and carisoprodol (Table 1). The top 5 significant abundant metabolites of bovine serum were 4-Ethyl-2,6-dihydroxyphenyl hydrogen sulfate, (2R)-1-(Nonadecanoyloxy)-3-(phosphonooxy)-2-propanyl docosanoate, Epinephrine, DG(16:1(9Z)/22:0/0:0) and tak-475 (Table 1).
Interestingly, much of metabolites which present higher level in ovine serum were associated with anti-microbico, antiviral or anticancer, such as Prunin, etravirine and Luteolin. While some of metabolites which present higher level in bovine serum were associated with contraception, such as gemeprost and Loxoprofen, which indicated that it may not suitable for pregnancy at this period. Notably, hernandezine, which is a novel AMPK activator, may play a role in the formation of lactoprotein of ovine milk.
Pathway enrichment of differential abundant metabolites
KEGG pathway enrichment showed that 18 and 10 functional pathways of differential metabolites were enriched at positive and negative ion mode, separately (Table 2). The most five enriched pathways of differential metabolites at positive-ion mode were Steroid hormone biosynthesis, Pathways in cancer, Prostate cancer, Purine metabolism and Oxidative phosphorylation (Table 2). The most five enriched pathways of differential metabolites at negative-ion mode were Carbohydrate digestion and absorption, Prion diseases, Insect hormone biosynthesis, Regulation of lipolysis in adipocytes and Aldosterone synthesis and secretion (Table 2). The results indicated that there may be different biological effects between two species serum.