Modeling Within-Host Dynamics of SARS-CoV-2 Infection: A Case Study in Ferrets

Abstract

The pre-clinical development of antiviral agents involves experimental trials in animals and ferrets as an animal model for the study of SARS-CoV-2. Here, we used mathematical models and experimental data to characterize the within-host infection dynamics of SARS-CoV-2 in ferrets. We also performed a global sensitivity analysis of model parameters impacting the characteristics of the viral infection. We provide estimates of the viral dynamic parameters in ferrets, such as the infection rate, the virus production rate, the infectious virus proportion, the infected cell death rate, the virus clearance rate, as well as other related characteristics, including the basic reproduction number, pre-peak infectious viral growth rate, post-peak infectious viral decay rate, pre-peak infectious viral doubling time, post-peak infectious virus half-life, and the target cell loss in the respiratory tract. These parameters and indices are not significantly different between animals infected with viral strains isolated from the environment and isolated from human hosts, indicating a potential for transmission from fomites. While the infection period in ferrets is relatively short, the similarity observed between our results and previous results in humans supports that ferrets can be an appropriate animal model for SARS-CoV-2 dynamics-related studies, and our estimates provide helpful information for such studies.

Keywords: COVID-19; SARS-CoV-2; ferrets; viral dynamics; within-host modeling.

Figure 1

Model 1 fitting to the data. Prediction of the viral dynamics model 1 (solid line) along with the data (filled circle) containing the total viral load (RNA copies per mL) and the infectious virus titer (PFU per mL), collected from ferrets infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2.

Figure 2

Model 2 fitting to the data. Prediction of the viral dynamics model 2 (solid line) along with the data (filled circle) containing the total viral load (RNA copies per mL) and the infectious virus titer (PFU per mL), collected from ferrets infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2.

Figure 3

Model 3 fitting to the data. Prediction of the viral dynamics model 3 (solid line) along with the data (filled circle) containing the total viral load (RNA copies per mL) and the infectious virus titer (PFU per mL), collected from ferrets infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2.

Figure 3

Model 3 fitting to the data. Prediction of the viral dynamics model 3 (solid line) along with the data (filled circle) containing the total viral load (RNA copies per mL) and the infectious virus titer (PFU per mL), collected from ferrets infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2.

Figure 4

Sensitivity analysis. Partial rank correlation coefficients (PRCC) from the Latin hypercube sampling method of the total viral-peak size (1st column), infectious viral peak size (2nd column), time to viral-peak (3rd column), and cell loss (4th column) for Model 1 (1st row), Model 2 (2nd Row), and Model 3 (3rd row).

Figure 4

Sensitivity analysis. Partial rank correlation coefficients (PRCC) from the Latin hypercube sampling method of the total viral-peak size (1st column), infectious viral peak size (2nd column), time to viral-peak (3rd column), and cell loss (4th column) for Model 1 (1st row), Model 2 (2nd Row), and Model 3 (3rd row).

Figure 5

Role of Initial infectious viruses. Fitting data with initial virus at the infection site ($V_{i0}$) estimated vs fixed. Shown here are two representative animals from each of the two groups infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2. The best-fit curves corresponding to $V_{i0}=9.25$ PFU/mL (F13-E-2) and $V_{i0}=9.12$ PFU/mL (CTan-H-2) were obtained by allowing $V_{i0}$ to be estimated by data fitting process, while other best-fit curves were obtained by fixing the values of $V_{i0}$ at ${10}^2$, ${10}^3$, and ${10}^4$ PFU/mL.

Figure 5

Role of Initial infectious viruses. Fitting data with initial virus at the infection site ($V_{i0}$) estimated vs fixed. Shown here are two representative animals from each of the two groups infected with F13-E SARS-CoV-2 and CTan-H SARS-CoV-2. The best-fit curves corresponding to $V_{i0}=9.25$ PFU/mL (F13-E-2) and $V_{i0}=9.12$ PFU/mL (CTan-H-2) were obtained by allowing $V_{i0}$ to be estimated by data fitting process, while other best-fit curves were obtained by fixing the values of $V_{i0}$ at ${10}^2$, ${10}^3$, and ${10}^4$ PFU/mL.