NMR-based metabolomics of plasma from dairy calves infected with two primary causal agents of bovine respiratory disease (BRD)

Abstract

Each year, bovine respiratory disease (BRD) results in significant economic loss in the cattle sector, and novel metabolic profiling for early diagnosis represents a promising tool for developing effective measures for disease management. Here, ¹H-nuclear magnetic resonance (¹H-NMR) spectra were used to characterize metabolites from blood plasma collected from male dairy calves (n = 10) intentionally infected with two of the main BRD causal agents, bovine respiratory syncytial virus (BRSV) and Mannheimia haemolytica (MH), to generate a well-defined metabolomic profile under controlled conditions. In response to infection, 46 metabolites (BRSV = 32, MH = 33) changed in concentration compared to the uninfected state. Fuel substrates and products exhibited a particularly strong effect, reflecting imbalances that occur during the immune response. Furthermore, ¹H-NMR spectra from samples from the uninfected and infected stages were discriminated with an accuracy, sensitivity, and specificity ≥ 95% using chemometrics to model the changes associated with disease, suggesting that metabolic profiles can be used for further development, understanding, and validation of novel diagnostic tools.

Figure 1

Dairy calves’ responses to viral (n = 5) and bacterial (n = 5) infection. Calves were challenged with the infectious agent on D0. (a) Daily rectal temperature (TEMP, °C) displayed as Mean ± SD. (b) WBC (thousands per cubic millimeter, K/µl) presented as Mean ± SD. A characteristic increase after signs of disease (indicated by arrows) in each challenge can be seen due to the activation of defense mechanisms against BRSV and M. haemolytica (MH) virulence factors. Markers in blue are the Baseline days; the red markers point to the day of the pathogenic challenge.

Figure 2

Principal component analysis (PCA) correlation loadings plots for TPR and CBC data. The variables inside the outer circle (colored area) have the most influence on database variability and are positively or negatively correlated within each model, while the parameters within the inner circle have low or no influence. (a) Baseline (n = 55); two PCs explained 48% of the variance. (b) General infection (V + B) database (n = 47) combined data from both studies, two PCs explained 43% of the variation of the database. (c) Infected with BRSV (n = 21), two PCs explained 53% of the variation of the database. (d) Infected with M. haemolytica (MH) (n = 26), two PCs explained 50% of the variation of the database.

Figure 3

¹H-NMR spectra (0.8–9.0 ppm) showing the peak intensities of metabolites present in plasma after the controlled infections with the main causal agents of BRD. (a) BRSV sample 66 (D0, calf 4), (b) BRSV sample 29 (D9, calf 4), (c) M. haemolytica (MH) sample 38 (D0, calf 6), (d) MH sample 2 (D2, calf 6). To improve the visualization of the peaks, the size of the region between 5.1–8.5 ppm was increased 40X.

Figure 4

Principal component analysis (PCA) correlation loadings plots for the concentration of the selected ¹H-NMR metabolites (n = 72) in plasma from dairy calves. The variables inside the outer circle (colored area) have the greatest influence on database variability and are positively or negatively correlated during the Baseline or Infected stages; the points inside the inner circle are thought to have low or no influence. (a) Baseline (n = 35); two PCs explained 45% of the variance. (b) Infected (V + B) data points from both challenge studies (n = 35), two PCs explained 39% of the variation of the database. (c) Infected calves with BRSV (n = 20), two PCs explained 43% of the variation of the database. (d) Infected calves with M. haemolytica (n = 15), two PCs explained 36% of the database variation. Each plasma sample, on average, contained 43 ± 8 of the selected metabolites.

Figure 5

Principal component analysis (PCA) scores plots for ¹H-NMR spectra from Baseline and Infected plasma samples. (a) PCA scores plot (n = 70) from the combined V + B infection database (PC-2: R²X = 0.34, Q² = 0.26). (b) PCA scores plot (n = 40) from the BRSV challenge (PC-2: R²X = 0.34, Q² = 0.10). (c) PCA scores plot (n = 30) from the M. haemolytica (MH) challenge (PC-2: R²X = 0.45, Q² = 0.25). Labels above the scores indicate the sample ID (Table S1).

Figure 6

OPLS-DA scores plots resulting from ¹H-NMR spectra of plasma as well as the corresponding coefficient loading plots. The color map depicts the significance of spectral signals between the two categories (Baseline and Infected). (a) OPLS-DA scores plot from the combined V + B infection database (n = 70). (b) Coefficient loadings plot for general infection. (c) OPLS-DA scores plot from the BRSV challenge (n = 40). (d) Coefficient loadings plot from the BRSV challenge. (e) OPLS-DA scores plot from the M. haemolytica (MH) challenge (n = 30). (f) Coefficient loadings plot from the M. haemolytica challenge. Labels above the scores indicate the sample’s ID, and above the peaks show the metabolite’s ID. To improve the visualization of the peaks in the coefficient loadings plots, the size of the region between 6.0–9.0 ppm was increased 10X.