Identification of novel viral mutations in HIV-1 subtype C conferring resistance to the broadly neutralizing antibody, VRC01

Abstract

Human immunodeficiency virus type 1 (HIV-1) subtype C remains a global concern with HIV-1 subtype C being the most prevalent form of HIV-1, accounting for approximately 50% of infections worldwide. HIV-1 bNAbs are the most promising preventive HIV vaccine, however, there are challenges in the elicitation of bNAbs by vaccination, and thus the field has moved towards assessing passive immunization using isolated bNAbs. South Africa was one of the participating countries in the recent vaccine trial HVTN 703/HPTN 081 of the Antibody Mediated Prevention (AMP) studies, which showed that the CD4bs antibody VRC01 effectively protected against the acquisition of HIV-1 in only 30% of virus isolates that were highly sensitive to VRC01, with no protection against resistant isolates. One of the challenges in using VRC01 for passive immunization is resistance to neutralisation due to naturally occurring mutations in the HIV envelope (Env) protein. The aim of this study was to identify mutations in HIV-1 subtype C viruses that contribute to VRC01 resistance. Seven Env sequences of VRC01-resistant HIV subtype C viruses were analysed to identify putative neutralisation escape mutations. HIV LANL database and Aliview software were used to align the resistant HIV Env sequences to HXB2. Putative escape mutations were identified in four regions of the Env sequences: the C1 region, loop D, and the β23 and β24 regions. Nine Env- pseudotyped viruses were generated from mutated envelopes and were analysed for their neutralisation sensitivity to VRC01 using a pseudovirus-based neutralisation assay. The mutant pseudoviruses were also assessed for sensitivity to four other bNAbs, that is, VRC07-523LS and 3BNC117 which bind on the same epitope as VRC01 and, 10E8 and PG9 that bind to different epitopes on the Env gene. Neutralisation data was compared with in silico data obtained from machine learning analyses performed at the Fred Hutchinson Cancer Research Center (FHCRC) (Seattle, Washington, USA) which predicted the probability of sensitivity and IC80 values for VRC01. Four Envs bearing single mutations (H0902_K279D, V0217_E279D, V1298_E455T, and V1255_D99N) became sensitive to VRC01 and 3BNC117 compared to their respective wildtype strains. Both VRC01-resistant clones and reversion mutations remained sensitive to VRC07-523LS except for H0902_K279D, highlighting possible cross-resistance among antibodies binding to the same epitope. When machine learning was used to predict the impact of each of the identified mutations on VRC01 sensitivity, the predictions matched the experimental findings in 3/5 cases. Generally, 3 mutations that conferred VRC01 sensitivity in the neutralisation assay, were predicted to be sensitive by machine learning as well, excluding V703_1255_D99N which was found slightly sensitive to VRC01 by neutralisation assay, and resistant by machine learning prediction. Interestingly the single mutation, E455T in the V703_1298 mutant conferred complete sensitivity to VRC01. However, in the H703_1798 Env, the same mutation (E455T) did not change the sensitivity to VRC01. Similarly, a single mutation, D99N in V703_1255 caused partial sensitivity to VRC01, while V703_1298_D99N was resistant. Thus, our neutralisation assay showed partial concordance with the in-silico predictions. In conclusion, more neutralisation data is needed to prove that machine learning may be a useful tool to identify and screen potential mutations that contribute to VRC01 resistance. Our data identified some specific residues important for VRC01 resistance in HIV-1 subtype C. Some of these features may result in cross-resistance to other antibodies that bind to CD4bs which is clinically significant for the testing of other CD4bs antibodies. Moreover, the sensitive mutant V703_1298_E455T warrants further investigation as this is the first report on mutation E455T to show VRC01 sensitivity in HIV-1 subtype C isolates.