Inferring biomarkers for Mycobacterium avium subsp. paratuberculosis infection and disease progression in cattle using experimental data

Abstract

Available diagnostic assays for Mycobacterium avium subsp. paratuberculosis (MAP) have poor sensitivities and cannot detect early stages of infection, therefore, there is need to find new diagnostic markers for early infection detection and disease stages. We analyzed longitudinal IFN-γ, ELISA-antibody and fecal shedding experimental sensitivity scores for MAP infection detection and disease progression. We used both statistical methods and dynamic mathematical models to (i) evaluate the empirical assays (ii) infer and explain biological mechanisms that affect the time evolution of the biomarkers, and (iii) predict disease stages of 57 animals that were naturally infected with MAP. This analysis confirms that the fecal test is the best marker for disease progression and illustrates that Th1/Th2 (IFN-γ/ELISA antibodies) assays are important for infection detection, but cannot reliably predict persistent infections. Our results show that the theoretical simulated macrophage-based assay is a potential good diagnostic marker for MAP persistent infections and predictor of disease specific stages. We therefore recommend specifically designed experiments to test the use of a based assay in the diagnosis of MAP infections.

Figure 1

(A) Disease progression stages and the corresponding MAP shedding kinetics that can be detected in CFU sample’s at each stage. (B) Model diagram: Macrophages kill free bacteria at rate k_m and get infected at rate k_i giving rise to infected macrophages. Uninfected and infected macrophages have death rates given by μ_m and μ_i, respectively. Infected macrophages burst at rate k_b and they release N_o bacteria at the same time. Th1 cells are assumed to kill infected macrophages at rate k_l. IFN-γ and antibodies are assumed to be Th1 and Th2 subset surrogates, respectively. Both the population of infected macrophages and free bacteria are assumed to be the source of bacteria excreted in feces at rates λ₁F_m(I_m) and λ₂F_B(B), respectively.

Figure 2. Classification of disease/infection groups.

Time series kinetics of CFU shedding, IFN-γ and ELISA assay expressions for each of the 57 cattle grouped into separate categories based on immune and shedding patterns. Group 1 is classified as silent infections. In this group no shedding is observed and immune responses are not distinct and differently expressed. In Group 2 (sub-clinical infections), intermittent shedding is observed, while IFN-γ and ELISA test show differential expression of immune responses. Animals with high consistent CFU shedding and high IFN-γ and ELISA are categorized as Group 3, and are assumed to be in the clinical state.

Figure 3. Immune markers as predictors of disease progression.

Fitting linear models to determine the level of associations between the immune variables and the level of shedding for animals in the different groups. In Group 1 and Group 2 animals, weak linear relationships between CFU vs Th1, CFU vs Th2 and Th1 vs Th2, are predicted. In Groups 3, the ELISA assay is shown to be strongly correlated with MAP shedding, while IFN-γ is also positively correlated with the CFU and the ELISA assays. The data in each group are represented by a different shape (i) Group 1 a circle, (ii) Group 2, a square and Group 3, a triangle. The model fitted lines are represented by (i) a solid continuous lines (the model y = β₁x) with different color shadings showing the 95% CIs and (ii) the broken lines (y = β₀ + β₁x) with the grey shading showing the 95% CIs.

Figure 4. Comparing the dynamic model to Group 2 and Group 3 data.

Model fitting to data using the ordinary differential equations dynamic model. The upper row shows Group 2 animals fitting while the lower row shows animals in Group 3. The solid lines represent the model fitting to the data (represented by the solid circles). The blue, green and red shadings represent the 95% CrIs while the grey shading represent the 50% Crls. Model parameters that were varied to generate the fits are given in Table 1 and the parameters that were fixed during model fitting are given in Table S1.

Figure 5. Comparing Group 2 and 3 animal data to the hybrid model.

The Figure shows data fitting for Group 2 (upper row) and Group 3 (lower row) animals using a hybrid model. The solid lines represent the model fitting to the data (represented by the solid circles). CFU shedding is shown by a piecewise continuous logistic function. The color shadings (blue, green and red) represent the 95% CrIs and the 50% CrIs are shown by the grey shading. See Table 1 for the parameters that were varied during model fitting. Table S1 gives parameters that were kept invariant during model fitting.

Figure 6. Assay simulation and comparison.

Panel (A) shows that the population of infected macrophages is higher than the bacteria population. And these potentially stimulate the Th1 and Th2 responses noticed in Group 2 animals and their sensitivity increase over time. Panel (B) shows that in Group 3, there are more infected macrophages compared to within host extracellular bacteria in the first 12 months of infection, but later there are more free bacteria than infected macrophages as infection progresses. This figure was generated using the ODE model and the estimated parameters obtained through model fitting to data shown in Fig. 4.

Figure 7. Theoretical assay evaluation.

Panels (A) and (B) show results of the theoretical assay simulated using the basic model. Panel (A) shows assay evolution for non-progression infection, while (B) shows how the assays predict fast progression infections. Assay values were simulated using a six month moving average. The model estimated parameters were used to evaluate the sensitivity score of the different assays for each group.

Figure 8. Assay summary diagnostic interpretation.

The ability of each separate assay to detect MAP infection and JD progression stages in non-progressing infections and fast progressing infections are shown. The blue color represents a negative assay result, the green color represents an unreliable assay result, and the colors, yellow, brown, and red, stand for intermediate, strong and very strong assay prediction results, respectively.