Fig 1.
Unsupervised analysis of serum metabolomes.
A) Principal component analysis. In the PCA score plot, points represent each sample. Each sample is colored according to the individual cat of origin, such that samples from the same cat are represented by the same color. The shape of each point represents the storage condition to which the samples were exposed. B) Statistical heatmap of serum metabolites. A statistical heatmap was used to visualize the relative abundance of serum metabolites. The color of each cell is proportional to the z-scaled relative abundance of each metabolite. Individual metabolites (rows) are organized by hierarchical clustering of Euclidian distances, but the dendrogram is not shown. Samples (columns) are organized by hierarchical clustering of Euclidian distances and the clustering dendrogram is present above along the x-axis at the top of the image.
Fig 2.
Volcano plots of statistical contrasts.
Volcano plots of the -log of the q-values (y-axis) and the log2 of the fold-change between conditions were generated for each statistical contrast in the post-hoc tests among storage conditions. A) Volcano plot for the comparison of samples stored at -20°C for 12 months and -80°C. B) Volcano plot for the comparison of samples stored at -20°C 6 Months and -80°C. C) Volcano plot for the comparison of samples stored at -20°C for 12 months and -20°C for 6 months. The red horizontal lines represent the threshold for statistical significance (q-value < 0.05) and all metabolites above this line were significantly variable.
Fig 3.
Significantly variable serum glutathione metabolites.
Relative abundances of serum glutathione metabolites were increased in serum samples stored at -20°C. Boxplots contains data from 8 individual cats with three biologic replicates from each cat representing the three different storage conditions. Linear mixed-effects models were used to identify serum metabolites that varied significantly among the storage conditions. Plots are annotated with the false discovery rate-adjusted p-values (q-values) derived from the post-hoc pairwise comparisons among storage conditions: q < 0.05 (*); q ≤ 0.01 (**); q ≤ 0.001 (***); q≤ 0.0001 (****). Non-significant comparisons are hidden.
Fig 4.
Significantly variable serum amino acid and gamma-glutamyl amino acid metabolites.
Relative abundances of serum amino acid and gamma-glutamyl amino acid metabolites were decreased in serum samples stored at -20°C. Boxplots contains data from 8 individual cats with three biologic replicates from each cat representing the three different storage conditions. Linear mixed-effects models were used to identify serum metabolites that varied significantly among the storage conditions. Plots are annotated with the false discovery rate-adjusted p-values (Padj) derived from the post-hoc pairwise comparisons among storage conditions: q < 0.05 (*); q ≤ 0.01 (**); q ≤ 0.001 (***); q≤ 0.0001 (****). Non-significant comparisons are hidden.
Fig 5.
Significantly variable serum polyunsaturated fatty acid metabolites.
Relative abundances of serum PUFA metabolites were decreased in serum samples stored at -20°C. Boxplots contains data from 8 individual cats with three biologic replicates from each cat representing the three different storage conditions. Linear mixed-effects models were used to identify serum metabolites that varied significantly among the storage conditions. Plots are annotated with the false discovery rate-adjusted p-values (q-values) derived from the post-hoc pairwise comparisons among storage conditions: q < 0.05 (*); q ≤ 0.01 (**); q ≤ 0.001 (***); q≤ 0.0001 (****). Non-significant comparisons are hidden.
Fig 6.
Analysis of metabolite stability.
A) Intraclass correlation (ICC) statistics. ICC analysis was used to classify metabolites according to the degree of correlation among measurement of samples from the same cats stored under different conditions. The counts and proportions of metabolites classified with “excellent,” “moderate,” “fair,” or “poor” stability are represented. B) Metabolite set enrichment analysis. Metabolite set enrichment analysis (MSEA) was used to detect metabolic sub-pathways enriched among metabolites with high (positive NES) and low (negative NES) stability. Sub-pathways presented here are those with NES P > 0.05. Bars are annotated by the false discovery rate-adjusted p-value (q-value) such that grey bars represent pathways with q ≥ 0.05 and orange bars representing sub-pathways with q < 0.05.