A research team led by Carolyn Williamson of the University of Cape Town, together with Penny L. Moore of the University of the Witwatersrand and James I. Mullins of the University of Washington, set out to answer a question that therapeutic trials cannot cleanly address: what happens to viral evolution when a broadly neutralizing antibody is already circulating in the blood at the moment of transmission? In chronic infection, resistance can emerge from archived reservoir viruses that carry pre-existing mutations, muddying any interpretation of cause and effect. In the AMP setting, by contrast, participants acquired HIV while receiving either VRC01 or placebo, and the virus that founded each infection was typically a single, genetically homogeneous lineage. That low diversity at transmission creates a high theoretical barrier to escape, because the virus must generate and select new mutations from a very narrow starting point. The trials therefore offered a uniquely clean window on de novo escape under defined antibody pressure.

The team focused on 47 participants who had acquired single-lineage infections with viruses that were sensitive or only intermediately resistant to VRC01, defined by an IC80 neutralization titre of 3 micrograms per millilitre or less. Twenty-three came from the Africa trial and 24 from the Americas trial, distributed across 21 placebo recipients, 15 low-dose recipients and 11 high-dose recipients. From each participant, the researchers analysed envelope sequences from two early timepoints, the first a median of just 18 days after the estimated date of acquisition. In total, 11,730 previously generated env sequences were examined, with a median of 108 sequences per participant per timepoint, generated using a Pacific Biosciences single-molecule real-time platform with unique molecular identifiers to ensure accuracy. This depth matters enormously: variants circulating at frequencies of one percent or lower would be invisible to conventional sequencing with around 20 reads, yet such low-frequency mutants are precisely where resistance begins.

To detect escape, the investigators tracked amino acid changes over time at 34 positions within the VRC01 epitope, using two complementary approaches. Escape logograms compared the amino acid composition at the second timepoint against the consensus of the founder lineage at the first, while maximum-likelihood phylogenetic trees paired with highlighter plots visualized which sequences had diverged and which mutations travelled together. Variant sites appeared in 18 of the 47 participants, and the pattern was strikingly asymmetric. Eleven variant sites were observed exclusively in VRC01 recipients and only one exclusively in placebo recipients, while three appeared in both groups. In the Africa trial, variant sites were significantly more common among treated participants, appearing in 9 of 13 VRC01 recipients versus 2 of 10 placebo recipients, and when the rate of change in variant frequency per day was pooled across trials, the high-dose group showed a significantly faster rate than placebo.

Identifying a variant site, however, is not the same as proving escape. To establish causality, the team used site-directed mutagenesis to introduce each observed amino acid change into a pseudovirus clone carrying the participant’s own transmitted founder envelope, then measured neutralization sensitivity in the standard TZM-bl assay. A change was classified as an escape mutation if it shifted the IC80 by more than threefold. The verdict was unambiguous: escape mutations emerged in 8 of 26 VRC01-treated participants, or 31 percent, compared with none of the 21 placebo recipients, a difference that held up statistically with a P value of 0.006. Notably, all eight escape cases were diagnosed within eight weeks of the last infusion, and no escape was seen in participants diagnosed after the trial endpoint, consistent with the idea that circulating drug levels were the selective force. Escape was also more frequent in the Africa trial, where all infections were subtype C, than in the Americas trial, where all were subtype B.

Structurally, the escape pathways converged on a familiar neighbourhood. The Loop D region of gp120 was the dominant hot spot, with mutations at residues 276, 279 and 280 detected in all eight participants, most often at positions 279 and 280. Additional mutations appeared in the N197 glycan motif and in the beta23 and V5 regions, including G458E, G459D and K460N. In two participants the virus appeared to toggle between alternative amino acids at the same residue, a pattern previously recognized as a signature of positive selection. Structural modelling onto a crystal structure of VRC01 bound to a subtype C gp120 offered a mechanistic explanation for several of these changes: the N280D substitution introduces a negative charge that electrostatically repels a negatively charged residue on the antibody light chain, while G458E and G459D insert bulky, charged side chains into a constrained binding interface. In contrast, D279A and D279G likely escape by shortening side chains and removing a negative charge, reshaping the contact surface in the opposite direction.

Perhaps the most clinically consequential finding concerns cross-resistance. The researchers tested 17 pseudovirus mutants against a panel of ten broadly neutralizing antibodies, including VRC01-class antibodies such as 3BNC117, VRC_CH31, VRC01.23LS and VRC07-523LS, as well as non-VRC01-class CD4 binding site antibodies N6, 1-18 and CH235.12, with PG9 and 10E8v4 as outgroup controls targeting the V2 apex and membrane-proximal external region. Pairwise correlations computed with random-effects generalized linear models showed that most VRC01-class antibodies clustered tightly with VRC01, meaning that escape from one largely compromised the others. Yet two antibodies stood apart. N6, whose distinct binding angle likely explains its resilience, was only moderately correlated with VRC01, and the highly potent antibody 1-18 was the most robust of all, retaining IC80 values below 1 microgram per millilitre against 75 percent of the VRC01 escape mutants and a median IC80 below 0.1 micrograms per millilitre overall. Intriguingly, some Loop D mutations actually increased sensitivity to CH235.12, by as much as 44-fold, illustrating that escape from one antibody can come at the cost of vulnerability to another.

Double mutations told an equally instructive story about how resistance compounds. In one participant, the N197T/K460N combination conferred more than tenfold greater resistance than either single mutation alone, pushing the IC80 above 25 micrograms per millilitre. In another, the double mutants N280D/G459D and D279A/G459D produced 42-fold to 76-fold increases in resistance relative to the single mutants. The effects were also strongly context dependent: the same N280D mutation conferred resistance to six of eight antibodies in one viral backbone, four of eight in another, and only one of eight in a third, underscoring that the genetic background in which a mutation arises shapes its antigenic consequence. One participant also carried an A318T substitution that co-evolved with the G458E escape mutation, possibly compensating for the fitness cost of inserting a charged, bulky residue into the CD4 binding interface.

That fitness cost is the study’s most reassuring thread. When the team searched 8,452 group M HIV-1 sequences from the Los Alamos National Laboratory database, every escape amino acid except K460N was present at a frequency below one percent of circulating viruses, and the K460N change was simply a reversion from one uncommon residue to a more common one. The rarity of these mutations in the wild suggests they impose a replicative penalty on the virus, which may explain why they have not spread despite the antibody’s clinical use. Whether resistant variants could persist and transmit if VRC01 exposure were more sustained remains an open question, and the authors argue that long-term surveillance of these variants in treatment and prevention cohorts should be a priority.

The broader implications reach directly into the design of next-generation prevention. Because 40 percent of the escape mutations detected here were present at frequencies of two percent or lower, standard sequencing approaches would have missed most of them, a gap the authors say argues for incorporating deep sequencing into resistance screening for prevention and cure trials. More strategically, the findings demonstrate that antibody-mediated pre-exposure prophylaxis, like antiretroviral-based PrEP before it, can select for viral escape after infection, and that escape pathways are common, convergent and predictable. As newer broadly neutralizing antibodies with greater potency, broader coverage and longer half-lives enter clinical evaluation, the map of VRC01’s vulnerabilities provides a blueprint: antibodies such as N6 and 1-18, which shrugged off these escape mutations, point toward combinations that could raise the evolutionary barrier and keep HIV’s adaptability in check.

Subject of Research: Viral escape from the broadly neutralizing antibody VRC01 during HIV prevention trials

Article Title: Prophylactic treatment with broadly neutralizing antibody VRC01 selects for escape mutants after infection in Antibody-Mediated Prevention trials

Article References: Williamson, C., Curry, L., Mkhize, N. N., Giorgi, E. E., Magaret, C. A., Lambson, B. E., Bhebhe, S., Kaldine, H., Moyo-Gwete, T., Rolland, M., Rossenkhan, R., Garcia, N. M. G., Moodley, C., Yssel, A., Huang, Y., Marsden, A. A., Reeves, D. B., Mayer, B. T., Bumgarner, R. E., … Mullins, J. I. (2026). Prophylactic treatment with broadly neutralizing antibody VRC01 selects for escape mutants after infection in Antibody-Mediated Prevention trials. Nature Microbiology. https://doi.org/10.1038/s41564-026-02467-4

Image Credits: AI Generated

DOI: 10.1038/s41564-026-02467-4

Keywords: HIV, VRC01, broadly neutralizing antibodies, viral escape, AMP trials, HIV prevention, CD4 binding site, Loop D, deep sequencing, viral evolution, antibody-mediated prophylaxis, Nature Microbiology