Naïve and memory cell turnover as drivers of CCR5-to-CXCR4 tropism switch in human immunodeficiency virus type 1: implications for therapy

Abstract

Early human immunodeficiency virus infection is characterized by the predominance of CCR5-tropic (R5) virus. However, in many individuals CXCR4-tropic (X4) virus appears in late infection. The reasons for this phenotypic switch are unclear. The patterns of chemokine receptor expression suggest that X4 and R5 viruses have a preferential tropism for naïve and memory T cells, respectively. Since memory cells divide approximately 10 times as often as naïve cells in uninfected individuals, a tropism for memory cells in early infection may provide an advantage. However, with disease progression both naïve and memory cell division frequencies increase, and at low CD4 counts, the naïve cell division frequency approaches that of memory cells. This may provide a basis for the phenotypic switch from R5 to X4 virus observed in late infection. We show that a model of infection using observed values for cell turnover supports this mechanism. The phenotypic switch from R5 to X4 virus occurs at low CD4 counts and is accompanied by a rapid rise in viral load and drop in CD4 count. Thus, low CD4 counts are both a cause and an effect of X4 virus dominance. We also investigate the effects of different antiviral strategies. Surprisingly, these results suggest that both conventional antiretroviral regimens and CCR5 receptor-blocking drugs will promote R5 virus over X4 virus.

FIG. 1.

Schematic of the model. The model includes naïve (N), naïve infected (N_I), memory (M), and memory infected (M_I) CD4⁺ T cells. New cells emigrate from the thymus at rate λ and enter the naïve cell pool. Naïve cells die at rate δ_N, become infected with viral strain i at rate β_NiV, and divide at rate r_N. When naïve cells divide, a proportion (f) convert to memory cells, and the rest maintain their naïve phenotype. Memory cells die at rate δ_M, and replicate at rate r_M. Infected cells produce virus at rate p. Free virus is cleared at rate c and binds to naïve or memory T cells (at rates bK_NiN and bK_MiM, respectively). Infected cells die at a higher death rate (δ_I) than uninfected cells due either to immune-mediated killing or viral cytopathic effect.

FIG. 2.

Model outcome without therapy: CD4 T-cell count (in cells per microliter) (black), total viral load (red), R5 (blue), and X4 (green) viral loads. (a) Model assuming that cell division is required for efficient viral production, (b) model assuming that cell division increases the level of viral infection. Parameter values used in the simulations are given in the Appendix. When there is only R5 virus, the total virus curve (red) is hidden under the R5 curve (blue).

FIG. 3.

Relationship between CD4 count and cell division (Ki67⁺ cells). The experimental relationship between total CD4⁺ T-cell count (cells per microliter) and Ki67⁺ naïve (red) and memory (blue) CD4⁺ T cells is shown (data from reference 23). (Bottom) The statistical results of fitting the data to an inverse relationship (y = A/x + B), as well as the r² for a linear regression (y = a + mx) are shown below the graph. SE, standard error.

FIG. 4.

Predicted effects of antiretroviral therapy (purple bar at the top of each panel) introduced before (1,000 days) (a) or after (1,800 days) (b) the phenotypic switch to X4 virus. Predicted effects of CCR5-blocking drugs administered prior to (c) or after (d) the switch. CD4⁺ T-cell count (cells per microliter) (black), total viral load (red), R5 (blue), and X4 (green) viral loads are shown. Parameter values used in the simulations are given in the Appendix. When there is only R5 virus, the total virus curve (red) is hidden under the R5 curve (blue). After antiretroviral therapy, in panels a and b, the total viral load also includes noninfectious virus, which would be measured in an assay.

FIG. 5.

Predicted effect on the time for the phenotypic switch of changing the parameter A for a memory cell in the relationship between T-cell numbers and division rates (Fig. 3) (a) and the initial inoculum of X4 virus (note the logarithmic x axis) (b). All other parameters were held constant at the values shown in the Appendix, and the dashed vertical line shows our original simulations (Fig. 2a).