The HEATHER cohort was an observational study looking at the impact of ART on the viral reservoir in people with Primary HIV Infection (PHI) who started ART within 3 months of diagnosis (20). Samples from 157 participants were used to study the prevalence of resistance to 10-1074 based on sequence analysis of HIV env. All participants were on ART at the time of sampling for an average of 5 years (range: 0-18 years). The majority of participants (64.4%) seroconverted between 2013 and 2016 (range: 2001-2020). The median age of the participants at the time of ART initiation was 35 years (range: 18-64). All but two participants in this cohort were male (98.7%), and 91 of 92 recorded transmission events were men having sex with men (MSM). An average of 16 single, proviral env sequences per participant (range: 1-72 sequences) was assessed for sensitivity to bNAbs. In total, 2593 sequences were analysed in this study. Most participants had B clade HIV (69.4%), followed by F (5.7%), C (4.4%), CRF02-AG (4.4%), CRF01-AE (3.1%), A1 and D/A1 (2.5% each), with other clades representing less than 2% of the cohort ( Figure 1A ).

The key determinants of HIV sensitivity to 10-1074 are the presence of a sequon at amino acid positions 332-334 of the HIV Envelope protein (a sequence of N-x-T/S, where x can be anything but P), which allows glycan attachment on the asparagine, and an intact motif on the protein,³²⁴G(D/N)IR³²⁷, which allows CDRH3 binding (4, 17). 28.6% of the HEATHER cohort participants had at least one sequence with mutations associated with 10-1074 resistance. No participants had previously been exposed to 10-1074 or had received treatment with bNAbs. Co-existence of sensitive and resistant sequences to 10-1074 in the same person was detected in 5.7% of participants in the HEATHER cohort. For this group with mixed sequences, the proportion of resistant sequences varied between12.5 and 85.7%. Subtype was an important determinant of resistance - most participants with clade B HIV in this cohort were predicted to be sensitive to 10-1074 (81/109; 74.3%). The highest rate of predicted 10-1074 resistance in the cohort was detected in CRF01-AE, although this was only five participants (5/5; 100%) ( Figure 1B ).

Mutations that interfere with the glycan attachment on the PNG at HIV Envelope position 334 were the most common among sequences predicted to be resistant (35 of the 45 sequences with resistance-associated mutations, 77.7%) in the HEATHER cohort. More than half of the potentially resistant sequences (25 of the 45 sequences with resistance-associated mutations, 55.5%) carried mutations at both positions 332 and 334 ( Figure 2 ) and furthermore, 64.4% of likely resistant samples in the cohort had a PNG site shift from position 332 to position 334 ( Figure 2 ). Notably, 89.2% of those with a PNG at 334 in this dataset also carried a 336T (³³⁴N-x-T³³⁶), which has a 100% likelihood of glycan occupancy (21). However, ³³²N-x-T³³⁴ was very rare in the sequences with an intact 332PNG. Furthermore, mutations at 330, which has been described as a critical 10-1074 binding site on env, most often occur with mutations that affect the glycan binding at position 332 (50% of 4 sequences with 330 mutations in the HEATHER cohort) ( Figure 2 ).

To explore further the broader prevalence of resistance to 10-1074, 6516 HIV env protein sequences sampled between 1983 and 2019 were downloaded from the Los Alamos sequence database (dataset Year 2020, all HIV-1 subtypes included; one sequence per individual) and tested for mutations associated with resistance to 10-1074. Of these, we focused on subtype B, and any sequences for which the time of sampling was unknown (43 sequences) were excluded from the analysis. The resulting analysis of 2425 B clade HIV env sequences revealed a trend of increasing 10-1074 resistance over the course of the HIV pandemic (16) ( Figure 3A ). Further analysis in the subset of sequences with 10-1074 resistance-associated mutations, reveals that mutations impacting the 332PNG (in positions 332 and/or 334) are becoming more frequent with time ( Figure 3B ), potentially driving the 10-1074 resistance pattern identified.

The variable loops in HIV Env evolve to evade the host immune response by incorporating insertions and deletions which have been associated with bNAb sensitivity (18). Alignment-independent methods were used on the HEATHER cohort sequences to measure the variable regions’ lengths (V1, V2, V4 and V5) as well as the number of PNGs within these regions ( Supplementary Table 2 ). A multivariable logistic regression with mixed effects, accounting for intra-patient variability and clade, showed that only the length of variable V1 loop was significantly associated with the presence of resistance-conferring mutations (OR=1.57, p-value<0.01, Figure 4 ). No statistically significant relationship between 10-1074 resistance-conferring mutation and the number of PNGs within the variable regions was observed.

In the absence of longitudinal data and relevant dating, ancestral state reconstruction can be used to infer the sensitivity phenotype of the founding virus. Here, ancestral state reconstruction was performed from the nine participants who had both sensitive and resistance sequences using bootstrapped maximum likelihood nucleotide trees. The median average pairwise distance between nucleotide sequences within samples with more than 1 sequence was 0.009 substitutions per site (range: 0-0.098 substitutions per site). The ancestral state is displayed as a pie chart at the root of each tree and the likelihood of sensitivity and resistance is indicated by different colours. The analysis revealed that in five of the samples ( Figures 5A-E ), the inferred root was very likely sensitive, indicating that the ancestral strain in these patients was most likely susceptible to 10-1074. The most common mutations observed in these samples were N334, with a PNG emerging at 334. In the remaining four trees ( Figures 5F-I ), a large amount of uncertainty was associated with the status of the most recent common ancestor of the sequences, although these ancestral strains were slightly more likely to be resistant to 10-1074. As viral evolution is unlikely on suppressive ART, this suggests that the ancestral strain in these participants may have already possessed genetic mutations associated with 10-1074 resistance, consistent with transmission of bNAb resistant strains, or that evolution of these mutations occurred shortly after transmission before ART was started.

Blood samples used in this study were collected from 157 participants on ART, enrolled in the HEATHER study, a prospective observational cohort study of individuals who commenced ART (and remained on uninterrupted therapy) within 3 months of the date of HIV diagnosis during PHI. Individuals were considered to have PHI if they met any of the following criteria: HIV-1 positive antibody test within 6 months of an HIV-1 negative antibody test, HIV-1 antibody negative with positive PCR (or positive p24 Ag or viral load detectable), RITA (recent incident assay test algorithm) assay result consistent with recent infection, equivocal HIV-1 antibody test supported by a repeat test within 2 weeks showing a rising optical density or having clinical manifestations of symptomatic HIV seroconversion illness supported by antigen positivity. Participants were all adult men who have sex with men and had started ART within 6 months of seroconversion. All research was performed following relevant guidelines and regulations, and all patients gave written informed consent to participate. Recruitment to the HEATHER cohort was approved by the West Midlands-South Birmingham Research Ethics Committee (reference 14/WM/1104).

PBMCs were isolated from blood and genomic DNA was extracted using the QIAmp DNA blood midi kit (QIAGEN). HIV env was amplified with single genome amplification in 96-well plates. To achieve a Poisson distribution, where a maximum of 30% wells yields a product, the appropriate dilution factor was calculated for each sample. Nested PCR with Platinum Taq (Invitrogen) was used to amplify env, with 2 sets of primers: envB5out 5′-TAGAGCCCTGGAAGCATCCAGGAAG-3′ and envB3out 5′-TTGCTACTTGTGATTGCTCCATGT-3′ in the first round and envB5in 5′-CACCTTAGGCATCTCCTATGGCAGGAAGAAG-3′ and envB3in 5′-GTCTCGAGATACTGCTCCCACCC-3′ in the second round. If no product was amplified, an alternative set of outer primers (R3B6R 5′- TGAAGCACTCAAGGCAAGCTTTATTGAGGC-3′ and B3F3 5′-TGGAAAGGTGAAGGGGCAGTAGTAATAC-3′) was used in the first round. The first round PCR was run at 94°C for 2 min; 94°C for 15 s, 58.5°C for 30 s, and 68°C for 3 min × 35; and 68°C for 15 min. For the second round, 1 μl of the first-round product was used as a template and the mix was run at 94°C for 2 min; 94°C for 15 s, 61°C for 30 s, and 68°C for 3 min × 45; and 68°C for 15 min. The end products were run on 1% 96-well E-gels (Invitrogen) in a 1:5 dilution. Amplicons were then pooled in a sequencing library and were sequenced using MiSeq Nano kits V3 (Illumina). On average 20 sequences (range: 2-76) were sampled from each participant.

A custom bioinformatics pipeline was used to assemble raw sequences and evaluate their sensitivity to 10-1074, as described elsewhere (8). Briefly, non-HIV sequences were removed, the raw sequences were aligned to HXB2, and the consensus was used as a reference to re-align raw reads. Viral sequences were translated to proteins and functionality was assessed based on env length and premature stop codons. Resistance to 10-1074 was defined by the absence of the following amino acids: N332; 333: not P; S/T334; D/N/T325 and H/Y330. These critical residues were associated with 10-1074 resistance using an adapted model by West et al. (14). HIV subtyping was done with REGA HIV Subtyping Tool version 3.46 (31).

The time series analysis to model the frequency of 10-1074 resistance-associated mutations was done and visualised using R. To model the frequency of individual mutation patterns associated with 10-1074 resistance, the subset of sequences with 10-1074 resistance-associated mutations was resampled 1000 times to account for the small dataset size and the proportion of each mutation pattern, PNG, GDIR and Both, were calculated per year for all datasets. Linear models of the proportion of each mutation pattern as a function of Year were fitted for all datasets and an average intercept and Year coefficient value were calculated. Using this data, an average line and its confidence intervals were plotted to illustrate the mutation pattern frequency trend in time.

Multiple alignment of the Los Alamos B clade sequences and of the HEATHER cohort B-clade sequences was performed with mafft (version 7.490). The lengths of variable loops as well as the number of PNGs within these regions in the aligned amino acid env sequences were measured using the variable region characteristics tool on the Los Alamos HIV database (https://www.hiv.lanl.gov/content/sequence/VAR_REG_CHAR/index.html). A mixed-effects logistic regression was then performed to identify the effect size of V1, V2, V4 and V5 lengths on 10-1074 resistance. The length of V3 was found to be constant across the HEATHER cohort sequences with a very small variability (1-3 amino acids compared to the average length) in very few sequences. Taking these observations, as well as similar findings in the literature (32) into account, the length of V3 was not included in the explanatory variables.

Neighbour-joining (NJ) phylogenetic trees of all protein sequences in the HEATHER cohort were constructed using R package ape to check for inter-sample contamination. Maximum likelihood trees were built for each sample with both sensitive and resistant nucleotide sequences, using Fastree on NGphylogeny.fr. The trees were rooted on an outgroup sequence, which was then pruned while maintaining the tree structure and the position of the root. Ancestral state reconstruction was performed to estimate sensitivity to 10-1074 of the internal nodes and root. Ancestral state reconstruction was done using ape and trees were visualised using ggplot2 (33).