Identification of Candidate Salivary, Urinary and Serum Metabolic Biomarkers for High Litter Size Potential in Sows (Sus scrofa)

Abstract

The selection of sows that are reproductively fit and produce large litters of piglets is imperative for success in the pork industry. Currently, low heritability of reproductive and litter-related traits and unfavourable genetic correlations are slowing the improvement of pig selection efficiency. The integration of biomarkers as a supplement or alternative to the use of genetic markers may permit the optimization and increase of selection protocol efficiency. Metabolite biomarkers are an advantageous class of biomarkers that can facilitate the identification of cellular processes implicated in reproductive condition. Metabolism and metabolic biomarkers have been previously implicated in studies of female mammalian fertility, however a systematic analysis across multiple biofluids in infertile and high reproductive potential phenotypes has not been explored. In the current study, the serum, urinary and salivary metabolomes of infertile (INF) sows and high reproductive potential (HRP) sows with a live litter size ≥ 13 piglets were examined using LC-MS/MS techniques, and a data pipeline was used to highlight possible metabolite reproductive biomarkers discriminating the reproductive groups. The metabolomes of HRP and INF sows were distinct, including significant alterations in amino acid, fatty acid, membrane lipid and steroid hormone metabolism. Carnitines and fatty acid related metabolites were most discriminatory in separating and classifying the HRP and INF sows based on their biofluid metabolome. It appears that urine is a superior biofluid than saliva and serum for potentially predicting the reproductive potential level of a given female pig based on the performance of the resultant biomarker models. This study lays the groundwork for improving gilt and sow selection protocols using metabolomics as a tool for the prediction of reproductive potential.

Keywords: LC-MS/MS; gilt selection; infertility; litter size; metabolomics; reproductive potential; saliva; serum; urine.

Figure 1

Flow of the study. Pigs were sampled for saliva, urine and serum as outlined in the methods section. After sampling, pigs were classified into their respective group based on their past reproductive performance. Samples were analyzed with targeted metabolomic quantification using LC-MS/MS techniques. Raw data from this analysis was appropriately pre-processed and analyzed using our developed metabolomics protocol. Metabolomics analysis results were then interpreted in the context of biological relevance of sow reproduction.

Figure 2

Distinct metabolic separation between HRP and INF pigs based on concentration of analyzed metabolites. PLS-DA plots clearly separated the HRP and INF groups using metabolite concentrations evaluated in urine (A), saliva (B), and serum (C). Cross validation and performance measures (D) of the PLS-DA models suggest that urine and saliva perform better than serum in the metabolic separation of the HRP and INF groups.

Figure 3

Volcano plots depicting the significant differences of metabolite concentrations in the urine (A), saliva (B), and serum (C) of HRP pigs in comparison to INF pigs. Each dot represents a metabolite. Red metabolites indicate an upregulation of the metabolite in HRP sows in comparison to INF sows, blue metabolites indicate a downregulation of the metabolite in HRP sows in comparison to INF sows, and grey metabolites indicate no change between HRP and INF sows. Fold-Change threshold = 2, p-value threshold = 0.05. A full list of p-values and FC values can be found in Supplementary Tables S4–S6.

Figure 4

Comparison of the metabolite changes of amino acid (A), fatty acid oxidation (B), lipid membrane (C) and steroid hormone (D) metabolite groups in serum, urine and saliva between HRP and INF pigs. Red metabolites indicate a decrease in the concentration of the metabolite in HRP compared to INF, green metabolites indicate an increase in the concentration of the metabolite in HRP compared to INF, and blue metabolites indicate metabolites that have variable changes in the metabolite across more than one biofluid, with the direction of their changes indicated by the arrow and corresponding biofluid underneath.

Figure 5

Selected features for ROC-AUC diagnostic classifier for urine (A), saliva (B), and serum (C). Features were selected via PLS-DA (VIP ≥ 1.25 + p < 0.05) and RFE.

Figure 6

ROC-AUC diagnostic predictors using the selected biomarker candidates for urine (A), saliva (B) and serum (C). The urine model appeared to be the most diagnostic (AUC = 0.98), followed by saliva (AUC = 0.93) and serum (AUC = 0.88). The red dotted line depicts a classifier that has no predictive ability (AUC = 0.5) and the blue solid line depicts the average predictive ability of the diagnostic predictor over five stratified K-fold cross-validations.