Quantitative analysis of immune response and erythropoiesis during rodent malarial infection

Abstract

Malarial infection is associated with complex immune and erythropoietic responses in the host. A quantitative understanding of these processes is essential to help inform malaria therapy and for the design of effective vaccines. In this study, we use a statistical model-fitting approach to investigate the immune and erythropoietic responses in Plasmodium chabaudi infections of mice. Three mouse phenotypes (wildtype, T-cell-deficient nude mice, and nude mice reconstituted with T-cells taken from wildtype mice) were infected with one of two parasite clones (AS or AJ). Under a Bayesian framework, we use an adaptive population-based Markov chain Monte Carlo method and fit a set of dynamical models to observed data on parasite and red blood cell (RBC) densities. Model fits are compared using Bayes' factors and parameter estimates obtained. We consider three independent immune mechanisms: clearance of parasitised RBCs (pRBC), clearance of unparasitised RBCs (uRBC), and clearance of parasites that burst from RBCs (merozoites). Our results suggest that the immune response of wildtype mice is associated with less destruction of uRBCs, compared to the immune response of nude mice. There is a greater degree of synchronisation between pRBC and uRBC clearance than between either mechanism and merozoite clearance. In all three mouse phenotypes, control of the peak of parasite density is associated with pRBC clearance. In wildtype mice and AS-infected nude mice, control of the peak is also associated with uRBC clearance. Our results suggest that uRBC clearance, rather than RBC infection, is the major determinant of RBC dynamics from approximately day 12 post-innoculation. During the first 2-3 weeks of blood-stage infection, immune-mediated clearance of pRBCs and uRBCs appears to have a much stronger effect than immune-mediated merozoite clearance. Upregulation of erythropoiesis is dependent on mouse phenotype and is greater in wildtype and reconstitited mice. Our study highlights the informative power of statistically rigorous model-fitting techniques in elucidating biological systems.

Figure 1. Parasite and RBC dynamics.

Data on parasite density and RBC density, averaged across mice from individual treatments ().

Figure 2. Model fits.

Fits of baseline model to parasite density and RBC density for reconstituted mice infected with the AJ strain (top panels), or the AS strain (bottom panels). Crosses are data. Light-grey regions correspond to 95% posterior predictive intervals (PPI); dark-grey regions correspond to 50% PPIs. The solid lines give the best-fit (posterior mode) solutions.

Figure 3. Model fits.

Fits of baseline model to parasite density and RBC density for nude mice infected with the AJ strain (top panels), or the AS strain (bottom panels). Crosses are data. Light-grey regions correspond to 95% posterior predictive intervals (PPI); dark-grey regions correspond to 50% PPIs. The solid lines give the best-fit (posterior mode) solutions.

Figure 4. Model fits.

Fits of baseline model to parasite density and RBC density for wildtype mice infected with the AJ strain (top panels), or the AS strain (bottom panels). Crosses are data. Light-grey regions correspond to 95% posterior predictive intervals (PPI); dark-grey regions correspond to 50% PPIs. The solid lines give the best-fit (posterior mode) solutions.

Figure 5. Standardised residuals.

Assessment of baseline model fits to all mouse data by standardised residuals. Top panel: parasite density, bottom panel: RBC density. Each cross represents the standardised residual of a time point for an individual mouse. These have an approximately normal distribution with mean 0 and variance 1. The solid red line joins the means of the standardised residuals at each time point. The dashed lines delimit the 95% predictive interval for the expected mean standardised residual for the same number of residuals as the data, Bonferroni corrected for the number of time points (see Text S1 for details). The model systematically overestimates the data when the red line lies below the 95% predictive interval, and underestimates the data when it lies above this interval. The y-axis is scaled in units of standard deviations.

Figure 6. Predicted uRBC clearance rates.

Posterior predictive intervals of immune-mediated uRBC clearance rates for the baseline model. Light-grey regions correspond to PPIs; dark grey regions correspond to PPIs. The solid lines give the best-fit (posterior mode) solutions. Note the different scales for reconstituted, nude and wildtype mice.

Figure 7. Predicted pRBC clearance rates.

Posterior predictive intervals of immune-mediated pRBC clearance rates for the baseline model. Light-grey regions correspond to PPIs; dark grey regions correspond to PPIs. The solid lines give the best-fit (posterior mode) solutions.

Figure 8. Parameter estimates.

Marginal distributions of the fitted parameters for the baseline model. Histograms are for individual mice. Panels with white backgrounds correspond to AJ-infected mice; panels with grey backgrounds correspond to AS-infected mice.