A framework for optimizing immunotherapy for long-term control of HIV

Abstract

HIV infects around 1 million people every year. Antiretroviral therapy (ART) is used to suppress viral replication and control disease progression but it cannot eradicate the virus. ART is therefore lifelong. Today, e ort is focused on using alternative strategies, particularly immune modulatory strategies, that would allow disease control after stopping ART. One such strategy is to administer HIV antibodies at the time of stopping ART to prevent viral resurgence and sustain the control established by ART over an extended duration. Antibody treatment, however, can fail due to viral mutation-driven development of resistance. Clinical trials document the failure of antibody therapy within weeks of its initiation despite the presence of high concentrations of the antibodies in circulation. A quantitative understanding of these observations is necessary to design therapies that would prevent such rapid failure. Here, we present a framework that provides such an understanding and facilitates treatment optimization. We recognize that the loss of control post-ART is associated with the stochastic reactivation of infected cells harboring latent virus because ART blocks all active virus replication. We rst adapt models of the development of resistance to antiretroviral drugs and show that the waiting time for the growth of antibody resistant strains due to the reactivation of latent cells would not capture the rapid failure observed clinically unless the latent cells already contained antibody resistant viral strains. Using ideas of population genetics, we then estimate the prevalence of such mutants before the start of antibody treatment. Using Gillespie simulations, we then estimate the distribution of the ii waiting time for the reactivation of the latent cells containing antibody resistant virus. Our simulations quantitatively capture clinical data of the rapid failure of antibody therapy. Our simulations suggest that combination therapy should be used to maximally prolong disease control. Finally, we considered the present stragies to identify optimal drug combinations. Synergistic drugs are preferred in combination therapies for many diseases, including viral infections and cancers. Maximizing synergy, however, may come at the cost of e cacy. This synergy-e cacy trade-o appears widely prevalent and independent of the speci c drug interactions yielding synergy. We present examples of the trade-o in drug combinations used in HIV, hepatitis C, and cancer therapies. We therefore believe that screens for optimal drug combinations that presently seek to maximize synergy may be improved by considering the trade-o .