From a cohort of 79 HIV-1 subtype C-infected women studied starting at the point of acute infection, we focused on two participants who developed Asn332-dependent BCN antibodies. Subject CAP177 produced antibodies by 3 years after infection that were capable of neutralizing 88% of a large multisubtype panel of 225 heterologous viruses (M. Lacerda, P.L.M., N. N., M.S.S., E.S.G. et al., unpublished data). The second individual, CAP314, neutralized 46% of 41 heterologous viruses after only 2 years of infection (Supplementary Fig. 1). Plasmas from CAP177 and CAP314 were unable to neutralize heterologous viruses that lacked the glycan at position 332 from the time when BCN responses were first detected and thereafter (Fig. 1a), confirming that their BCN activity depended on the 332 glycan, similar to the PGT128 mAb.

Single-genome amplification and gp160 sequencing from plasma viral RNA (Online Methods) was used to determine the amino acid sequence of the envelopes of circulating viral populations at multiple time points. This included the earliest available plasma samples, which were taken 2 weeks after infection for CAP177 (enabling inference of the transmitted/founder virus²⁰) and 3 months after infection for CAP314. In both cases, these acute viruses lacked the predicted N-linked glycan at amino acid 332, although almost all sequences contained an intact glycosylation site at position 334 (Fig. 1b,c). By 5–6 months of infection, a glycan at position 332 evolved in both CAP177 and CAP314 through an N334T or N334S substitution, which also resulted in the destruction of the neighboring glycan at position 334. By 12–15 months, when BCN antibodies became detectable, the 332 glycan was present in all sequences; however, at around 2 years the glycan reverted to position 334, most likely as a consequence of viral escape from the Asn332-dependent BCN response (Fig. 1b,c). Therefore, for both CAP177 and CAP314, the glycan that formed the basis of the BCN epitope was not present on the infecting virus but evolved shortly thereafter, coincident with the appearance of strain-specific antibodies (Fig. 1c).

The 332 glycan lies within the C3 region, which is highly immunogenic in HIV-1 subtype C and is often targeted by early strain-specific antibodies²¹. Indeed, we have previously shown that in CAP177 early neutralizing antibodies targeted the C3 region²². As viral escape often involves glycan rearrangements¹⁰, we postulated that the shift of a glycan from position 334 to 332 in CAP177 and CAP314 may have been the result of neutralization escape from early strain-specific antibodies. To test this, we cloned the acute virus and a representative 6-month virus from CAP177 and CAP314 and inserted the 332 glycan into the acute clones by mutating the asparagine at position 334 to a serine (which also deleted the glycan at 334). We tested all three viruses from both individuals for sensitivity to autologous plasma from 6 months after infection. We detected autologous neutralizing responses against the acute viruses, whereas the 6-month clones were resistant to contemporaneous plasma, as expected (Fig. 2a,b). Introduction of the glycan at position 332 resulted in almost complete resistance to neutralization by plasma from CAP177 and reduced neutralization sensitivity to plasma from CAP314, suggesting that the 332 glycan evolved to afford escape from early strain-specific antibodies.

The absence of the glycan at position 332 on the acute viruses from CAP177 and CAP314 suggested that they might be resistant to BCN mAbs that depend on this glycan, such as PGT128. Neutralization experiments showed that CAP177 and CAP314 acute clones were resistant to PGT128, whereas the 6-month clones that contained the 332 glycan were highly sensitive, with a half-maximum inhibitory concentration (IC₅₀) <0.06 μg ml⁻¹ (Table 1). We obtained similar data with a related mAb, PGT121, although the CAP177 acute clone showed moderate sensitivity (0.21 μg ml⁻¹) but was nevertheless 20-fold more resistant compared to the 6-month clone. Both N334S mutants (acute clones with a glycan introduced at position 332) were sensitive to PGT121 and PGT128, confirming the role of the 332 glycan in conferring sensitivity to these mAbs. None of the CAP177 or CAP314 clones was sensitive to the mAbs PGT135 or 2G12 (Table 1), which depend on additional glycans at other amino acids, that are uncommon among subtype C viruses^23-27. Thus, whereas shifting a glycan from position 334 to 332 allowed the virus to escape autologous neutralizing antibodies, this created a new neutralizing antibody epitope that provided the antigenic stimulus to elicit BCN antibodies targeting the 332 glycan.

We assessed whether this pattern of selection of the 332 glycan was evident at a population level using more than 7,300 single-genome amplification—derived gp160 envelope sequences from acute and chronic HIV-1 infections (Online Methods)^20,28. Sequences with evidence of dual infection were excluded. For each subject, the consensus sequence was generated and the frequency of the 332 glycan was determined. We assessed significance with Fisher’s exact test. In subtype C, the 332 glycan was significantly less common among transmitted/founder viruses (45/68, 66%) compared to unmatched chronic viruses (52/62, 84%, P = 0.0166, Fig. 3a). To ensure that this was not due to adaption of HIV to neutralizing antibodies over the course of the epidemic time²⁹, we performed the same analysis using a smaller data set of 502 matched sequences from 20 individuals, with similar results (P = 0.0457, Fig. 3a). Although we observed the same trend in subtype B sequences, it was not statistically significant (Fig. 3a). Taken together, these results suggest that the pattern of evolution we describe for CAP177 and CAP314 may be relatively common and that the absence of the 332 glycan on subtype C viruses may provide an advantage during transmission or early viral outgrowth.

We analyzed the phenotypes of 101 transmitted/founder subtype C viruses using envelope clones generated as part of the Vaccine Immune Monitoring Core Standard Virus Panel Consortium. For this, transmitted/founder envelope sequences were inferred from singlegenome amplified and sequenced envelope amplicons derived from plasma from acutely HIV-infected subjects³⁰ and cloned into a mammalian expression vector. Envelope clones were transfected into 293T cells with the HIV backbone construct pSG3ΔEnv to produce envelope pseudotyped particles, and neutralization assays were performed in TZM-bl cells as described in the Online Methods. Phenotypic analysis supported the genotypic analysis, with a high proportion (46%) of viruses resistant to PGT128 neutralization at the highest concentration tested (10 μg ml⁻¹) (Fig. 3b and Supplementary Fig. 2). Resistance strongly correlated with the absence of the 332 glycan (P < 0.0001) (Fig. 3c), although some viruses that contained the glycan were also resistant, consistent with the fact that additional residues are needed to form this epitope⁸. Of 31 viruses in which the glycan at position 332 was absent, only three showed neutralization sensitivity. Of these, two contained the glycan at position 295, which is very rare in subtype C viruses²⁶ but structurally proximal to the 332 glycan and shown by mutagenesis to affect the PGT128 epitope¹⁹. Although this virus panel was tested only against PGT128, resistance to this mAb generally extends to other Asn332-dependent PGT mAbs⁸. These data suggest that Asn332-dependent antibodies present either through passive immunotherapy or vaccination might be only partially effective in preventing subtype C infections and that combinations of antibodies targeting different epitopes may need to be tailored to match circulating viral variants^24,26,31.

In addition to the Asn332-dependent epitope, a second BCN antibody epitope that includes the glycans at amino acid positions 156 and 160 in the V2 region has been defined by the mAbs PG9, PG16 (ref. 18) and PGT145 (ref. 8). To assess whether epitope evolution was also associated with the appearance of Asn160-dependent BCN antibodies, we studied two individuals, CAP8 and CAP256, who developed this specificity¹⁵ (Supplementary Fig. 3a). In both cases, the glycan at amino acid 160 was absent from envelope sequences at transmission but appeared at 6 and 3 months in CAP8 and CAP256, respectively, suggesting a pattern of evolution similar to that observed with 332 glycan (Supplementary Fig. 3b). Furthermore, we found acute viruses to be resistant to BCN mAbs PG9, PG16 and PGT145, whereas later clones that contained the 160 glycan became sensitive to PG9 and PGT145, although not to PG16 (Supplementary Fig. 4). However, although insertion of the 160 glycan by site-directed mutagenesis into the resistant acute clones rendered them sensitive to PG9 and PGT145, unlike the case with the CAP177 and CAP314 mutants containing the 332 glycan, this single change did not mediate neutralization escape. It was also not possible to show selection of the 160 glycan using large sequence data sets, as it is highly conserved in both acute and chronic sequences (Supplementary Fig. 5).

Despite the fact that BCN epitopes are highly conserved, they only occasionally elicit BCN responses in humans. Those targeting epitopes centered around the 332 or 160 glycans develop more frequently in infected people compared to BCN antibodies directed at the CD4 binding site or membrane-proximal region^12-14. The data presented here raise the hypothesis that, at least in the case of Asn322-directed antibodies, viral evolution may facilitate their elicitation. However, the absence of these glycans on the acute virus is not a prerequisite for the development of breadth, nor does the evolution of a BCN epitope guarantee that these antibodies will arise. Indeed, within this cohort, examples of both scenarios exist (Supplementary Fig. 6a,b). Furthermore, in a study of a SHIV-infected macaque³² where the 332 glycan was present at transmission, potent and broad Asn332-dependent BCN antibodies developed within 9 months. In HIV-1 subtype C viruses, where the absence of the 332 glycan is favored at transmission, subsequent immune pressure exerted by strain-specific neutralizing antibodies results in the evolution of the 332 glycan. This process provides a mechanism for the evolution of BCN epitopes, with neutralization escape driving viral convergence toward glycan motifs that are highly conserved and serve as targets for BCN antibodies.

Supplementary