Antibodies VRC01 and 10E8 Neutralize HIV-1 with High Breadth and Potency Even with Ig-Framework Regions Substantially Reverted to Germline

Recent years have seen an explosion in the number of broadly neutralizing Abs (bNAbs) against HIV-1 (1–10). Many of these bNAbs have been shown to protect from or to provide control of infection (11–13) and, therefore, are of interest for passive-immunization approaches (14). An underlying characteristic of anti–HIV-1 Abs is the substantially increased levels of somatic hypermutation (15). Somatic hypermutation is part of the diversification of Abs that occurs during affinity maturation: this process occurs in activated B cells exposed to Ag within germinal centers where high-affinity Abs are selected over their low-affinity counterparts (16). Generally, chronic viral infections are associated with the generation of Abs with increased numbers of mutations compared with acute viral infections, suggesting that persistent Ag exposure plays a role in stimulating repeated rounds of somatic hypermutation and selection (17, 18). In the case of HIV-1, bNAbs mostly show higher mutation levels compared with weakly neutralizing Abs (18). Moreover, V-gene reversion to germline of several anti-HIV bNAbs results in lack of neutralization activity (18–20), indicating that somatic hypermutation is important for neutralizing breadth and potency (18).

Although somatic mutations occur preferentially within the CDR regions of Abs (21), large numbers of mutations in anti–HIV-1 bNAbs are also found within the Ab-framework regions (18, 19). Klein et al. (18) analyzed a set of anti–HIV-1 bNAbs targeting diverse epitopes on the HIV-1 envelope glycoprotein and found that full framework reversion to germline substantially reduces or completely abrogates neutralization activity for many of these Abs; function is partly restored in some Abs by allowing framework mature mutations in positions that are in direct contact with the Ag (18). The results of that study underline the importance of framework maturation for broad and potent neutralization by anti–HIV-1 Abs. However, it has been unclear whether most or all of the framework mutations would be necessary for retention of Ab function or whether some of these mutations could be reverted to germline with minimal effects on function.

To investigate this question, we selected two bNAbs that target different sites of vulnerability on the HIV-1 Env glycoprotein: the CD4 binding site (CD4bs) Ab VRC01 and the membrane-proximal external region (MPER) Ab 10E8. These Abs neutralize ∼90% and 98% of HIV-1 strains at an average potency of 0.33 and 0.22 μg/ml, respectively (4, 10). The variable regions of both of these Abs exhibit high degrees of amino acid mutation: VRC01 V-gene, 42% heavy/28% light and 10E8 V-gene, 22% heavy/17% light. The putative germline-reverted versions of these Abs are incapable of neutralizing HIV-1 viral strains (19; M. Louder and J. Mascola, unpublished observations). For VRC01, the mature CDRs, alone or in combination with the Ag-contacting framework residues, are not sufficient for potent neutralization: i.e., they were able to only weakly neutralize 0 and 3 of 10 tested strains, respectively (18). These results confirm the importance of framework mutations in VRC01. However, we conjectured that not all mutations from germline would be necessary for retention of Ab-neutralization activity. To test this conjecture, we created a series of VRC01 and 10E8 variants with partial framework reversions to germline in both H and L chains and compared their neutralization activity to that of the mature Abs. This approach allowed us to explore the relationship between neutralization activity and the number of framework mutations in anti–HIV-1 bNAbs. Our results challenge the notion that extensive framework mutations are necessary for broad and potent neutralization by anti–HIV-1 bNAbs and suggest strategies for Ab redesign in the context of Ab-product optimization or immunogen design.

Different approaches for various degrees of Ab germline reversion, ranging from full V-gene reversion to selective point mutations, were described previously, with applications to a number of viruses (18, 19, 22–24). In our study, a CDR-grafting procedure typically used for humanization of nonhuman Abs (25) was used as the basis for our designs, in combination with structural modeling and mutation analysis. Specifically, the CDR regions (Kabat definition) of the mature Abs (VRC01 and 10E8) were grafted onto homologous human germline genes (in effect, this replaces the germline CDRs with their mature counterparts). Additionally, series of germline framework residues were selected for mutation to their mature counterparts (Fig. 1). Structural modeling, analysis, and visualization were used to determine which germline framework resides to mutate to mature. Specifically, mutations were selected based on a combination of several factors: whether the position is implicated in supporting CDR regions (25); the differences between germline and mature amino acid types (e.g., change in amino acid charge, hydrophobicity, or structural properties); and proximity to Ag or CDR regions, as determined by the Ag-bound crystal structure. The Ag-bound crystal structure of VRC01 was obtained from Protein Data Bank id: 3NGB (19). The Ag-bound crystal structure of 10E8 [Protein Data Bank id: 4G6F (10)] was not available at the time of the design and was therefore only used for retrospective analysis. Structural modeling, analysis, and visualization were performed using OSPREY (26, 27), MolProbity (28), and PyMOL (29). Throughout this article, the mAb-nfH/mfL nomenclature is used for designed partial framework revertants, where n is the number of framework mutations in the H chain and m is the number of framework mutations in the L chain: for example, VRC01-5fH/6fL refers to the VRC01 variant with grafted mature CDRs and 5 H chain and 6 L chain mature framework mutations. Designed H and L chains were paired in a sparse matrix that was deemed to sufficiently represent the diversity of the possible H–L chain combinations.

Site-directed mutagenesis or gene synthesis was used to generate the CMV/R plasmids encoding the H or L chain framework revertants. Full-length IgG proteins were expressed by cotransfection of H and L chain plasmids into 293F cells, and Abs were purified with affinity chromatography using either Protein A Fast Flow Resin or HiTrap Protein A HP Columns (both from GE Healthcare).

Neutralization was measured using single-round-of-infection HIV-1 Env pseudoviruses and TZM-bl target cells, as described previously (30–32). Neutralization curves were fit by nonlinear regression using a five-parameter hill slope equation, as previously described (31). IC₅₀ values were used for the computation of neutralization breadth and potency for the different Ab variants. Average neutralization potency for a given Ab was computed as the geometric mean of the IC₅₀ values for a set of 21 diverse viral strains (7), where values > 50 μg/ml were set to 50. Neutralization breadth for a given Ab was computed as the percentage of strains with IC₅₀ values < 50 μg/ml.

A computational design procedure was used to create partial framework reversion variants of Ab VRC01 (see Materials and Methods and Fig. 1). The H chain of the mature VRC01 Ab contains 30 mutations within framework regions 1–4 (Fig. 2A). To test whether a subset of these mutations could be reverted back to germline without substantially affecting Ab activity, we created two partial framework revertants of the H chain, with five and eight mature framework mutations each (in the context of mature CDR regions), as well as the CDR-grafted version of the H chain (with mature CDR regions and no framework mutations) (Fig. 2). The selected mutations in the VRC01-5fH variant were in proximity to the Ag (distances of 3.6–7.7 Å) (Fig. 2A, 2B). The additional three mutations in the VRC01-8fH variant were deemed to represent substantial changes in amino acid type that were still in relative proximity to the Ag (distances of 6.7–8.8 Å). The remaining 22 residue positions of maturation were more distal to the Ag (mean distance of 16.3 ± 4.7 Å; range 8.3–24.9 Å) and/or represented more conserved amino acid changes and, thus, were not selected for study. Similarly, the mature VRC01 L chain contains 19 framework mutations, so we created partial framework revertants with mature CDR regions and 0, 6, and 10 mature framework mutations (Fig. 2). None of the framework mutations in the VRC01 L chain was in substantial proximity to the Ag (mean distance of 21.7 ± 6.9 Å; range 9.8–32.3 Å), so the reversion analysis was focused on framework interactions with the CDR regions and/or within the overall Ab structure. The VRC01-6fL L chain variant included mutations in proximity to the CDR regions (distances of 2.9–5.5 Å), as well as a two-residue FWR4 insertion that is not found in most other HIV-1 bNAbs, including other Abs from the VRC01 class (9). Structural modeling identified an additional four mutations that could be of importance for the overall structure of VRC01, and the addition of these mutations formed the VRC01-10fL variant (Fig. 2).

A total of 10 H–L chain combinations using the 4 H chain and 4 L chain VRC01 variants was tested for neutralization activity. The mature VRC01 (VRC01-30fH/19fL, consisting of 30 H chain/19 L chain framework mutations) was among the best in terms of neutralization function (Fig. 3); however, the VRC01-30fH/10fL variant (with mature H chain and partially reverted L chain) consistently outperformed all other constructs when using neutralization potency-breadth curves as a metric (the percentage of neutralized viral strains at different IC₅₀ potency cutoffs) (Fig. 4A). Six of the nine partial framework revertants maintained good neutralization activity. In fact, with the exception of VRC01-0fH/6fL (which did not express), VRC01-5fH/0fL, and VRC01-0fH/0fL, the revertants generally had an average potency within an ∼10-fold range from mature (Fig. 4B). The VRC01-0fH/0fL CDR-grafted construct showed complete lack of neutralization activity, in concordance with previous findings (18). Generally, although a significant correlation (p = 0.0433) was found between the number of framework mutations and the neutralization potency for VRC01 variants (Fig. 4C), a small number of framework mutations appeared to be sufficient for retention of Ab function: for example, the VRC01-5fH/6fL construct had a <2-fold decrease in potency compared with the mature VRC01 (Fig. 4A, 4B).

The crystal structure of the 10E8–MPER complex was not available at the time of the designs and, thus, was used only for retrospective analysis. The H and L chains of mature 10E8 contain 19 and 16 mutations within framework regions 1–4, respectively (Fig. 5A). We created 3 H chain variants with mature CDR regions and 0, 2, and 10 framework mutations and 3 L chain variants with mature CDR regions and 0, 4, and 10 framework mutations (Fig. 5). The selected mutations in the 10E8-2fH and 10E8-10fH H chain variants and 10E8-4fL and 10E8-10fL L chain variants were deemed to represent more substantial changes in amino acid type and/or were found at positions that have been implicated in supporting CDR regions (Fig. 5C). Post hoc structural analysis revealed that the mutations in the 10E8-2fH H chain variant were in proximity to the Ag (distances of 3.6 and 6.9 Å), but the distances to the Ag or the CDR regions for the additional eight mutations in the 10E8-10fH H chain variant versus the remaining nine H chain mutations were not substantially different (18.1 ± 6.6 Å versus 20.6 ± 4.9 Å for Ag, 7.5 ± 3.4 Å versus 6.6 ± 2.1 Å for CDR regions) (Fig. 5A, 5B). None of the framework mutations in the 10E8-L chain was in substantial proximity to the Ag (mean distance, 22.4 ± 6.8 Å; range, 14.2–33.0 Å; Fig. 5A). The 10E8-4fL L chain variant included mutations in proximity to the CDR regions (distances of 3.1–5.9 Å), but the distances to the CDR regions for the additional six mutations in the 10E8-10fL L chain variant versus the six remaining residue positions of maturation were not substantially different (10.3 ± 3.2 Å versus 8.3 ± 4.0 Å) (Fig. 5A, 5B).

A total of 10 H–L chain combinations using the 4 H chain and 4 L chain 10E8 variants were tested for neutralization activity. The mature 10E8 (10E8-19fH/16fL), along with the 10E8-19fH/10fL and 10E8-10fH/16fL variants, displayed the best neutralization potency and breadth (Figs. 3, 6A). Unlike VRC01, the 10E8 variant with no framework mutations (10E8-0fH/0fL) retained significant, but reduced, neutralization activity (Fig. 6B). As with VRC01, a significant correlation was found between the number of framework mutations and the neutralization potency for 10E8 variants (Fig. 6C; p = 0.0022). Interestingly, however, the 10E8-0fH/4fL variant that contained no H chain and only four L chain framework mutations had improved neutralization potency compared with 10E8-0fH/0fL, with only an ∼3-fold decrease in potency relative to mature 10E8 (Figs. 6A, 6B).

The high rate of mutation and the numerous other immune-evasion mechanisms of HIV-1 serve as a constant stimulus for HIV-1–reactive Abs to continuously evolve and mature as they target the ever-changing viral envelope. Perhaps as a consequence, broadly neutralizing anti–HIV-1 Abs typically have extreme numbers of variable domain (CDR and framework region) mutations. In the case of VRC01, the importance of framework mutations is underscored by the fact that, although VRC01-5fH/0fL showed virtually no neutralization activity, the addition of six framework mutations in the L chain resulted in VRC01-5fH/6fL being within 2-fold potency of the mature Ab. The precise role for each of these six mutations was not clear; however, three of these mutations (G66R, T69P, and F71Y) clustered closely to CDR L1 (Fig. 2B, 2C), suggesting that these mutations may help to support the CDR L1 loop in a conformation appropriate for interaction with Ag. We note that these mutations are generally not conserved in VRC01-class Abs from different donors, suggesting that Abs in this class may utilize diverse pathways to achieve broad and potent neutralization (1, 9). In the case of Ab 10E8, a variant with no mature framework mutations showed reasonable, albeit ∼10-fold lower, neutralization activity. Thus, framework maturation in anti–HIV-1 Abs does not appear to be universally required for potent and broad neutralization activity.

Although the overall number of framework mutations can affect neutralization activity, some mutations appeared to have a more substantial effect on Ab activity, so the specific selection of residues for germline reversion is also important. For example, the 10E8-2fH/4fL and 10E8-2fH/0fL variants had an ∼4-fold and ∼2-fold decrease in potency, respectively, compared with the analogous variants without the two H chain framework mutations (10E8-0fH/4fL and 10E8-0fH/0fL). Similarly, the VRC01-5fH/6fL variant had an ∼2-fold better potency than the substantially more mutated VRC01-8fH/19fL. In the case of 10E8-0fH/4fL, post hoc analysis of the 10E8-MPER peptide structure indicated that additional reversions of mutations that do not directly interact with the Ag or CDR regions may be possible (e.g., S2Y; Fig. 5). Further, the reversion analysis was focused solely on framework residues; however, it should also be possible to selectively revert CDR mutations, because not all mutations within the CDR regions are important for Ag recognition or overall Ab structure (33, 34). This process could allow for the generation of partially reverted Ab variants with even lower degrees of mutation and with similar, if not better, neutralization activity as their mature bNAb counterparts. Interestingly, we also found that reversion of selected mutations resulted in improved Ab variants compared with the mature Abs. In the case of VRC01, a combination of the mature H chain and an ∼50% reverted L chain (VRC01-30fH/10fL) resulted in 2-fold improved potency over the mature Ab. For 10E8, two variants, 10E8-19fH/10fL (mature H chain/∼40% reverted L chain) and 10E8-10fH/16fL (∼50% reverted H chain/mature L chain), exhibited similar neutralization breadth and potency to the mature Ab. These results show the promise of partial framework reversion for Ab optimization.

Anti–HIV-1 bNAbs with lowered mutation levels may be of use both in vaccine design and as therapeutic agents. An effective prophylactic HIV-1 vaccine most likely will have to elicit bNAbs. Thus, the high mutation levels of most anti–HIV-1 bNAbs present a challenge for vaccine development and immunogen design, because the vaccine immunogen (or perhaps a series of different immunogens) would need to guide the Ab affinity–maturation process from germline activation through the appropriate intermediate stages of Ab development to the mature bNAb (35–37). However, bNAbs with lower levels of mutations, such as those presented in this article or from Ab ontogeny data from next-generation sequencing (1, 9), may be more easily inducible and can be suitable targets for immunogen design, alleviating the requirement for guiding the affinity-maturation process to extreme levels of hypermutation (alternatively, it may be that to achieve necessary CDR mutations, high overall levels of somatic hypermutation are required, resulting in a substantial number of “bystander” framework mutations). Abs with high mutation rates might also be problematic as clinical products because the likelihood of immunogenicity may be increased. Although, to our knowledge, no highly somatically mutated Abs have been evaluated for immunogenicity in humans, Abs that are closer to their germline ancestors (the genes that are commonly found in most individuals) may be more easily tolerated in the general population. Thus, partial germline reversion may be a useful addition to the armamentarium of technologies for translating anti–HIV-1 bNAbs into clinical products. Overall, our results suggest that at least some of the anti–HIV-1 Abs can retain broad and potent neutralization activity even when their framework regions are substantially reverted back to germline. These results should aid in efforts for immunogen design targeting the elicitation of bNAbs and for Ab-based therapeutics for the treatment or prevention of HIV-1.

We thank the Structural Biology Section, Structural Bioinformatics Core Section, Virology Laboratory, Vector Core Section, Humoral Immunology Section, and Humoral Immunology Core at the National Institutes of Health Vaccine Research Center for helpful discussions and comments on the manuscript. We are grateful to J. Stuckey for assistance with graphics and thank J. Baalwa, D. Ellenberger, D. Gabuzda, F. Gao, B. Hahn, K. Hong, J. Kim, F. McCutchan, D. Montefiori, L. Morris, J. Overbaugh, E. Sanders-Buell, G. Shaw, R. Swanstrom, M. Thomson, S. Tovanabutra, C. Williamson, and L. Zhang for contributing the HIV-1 envelope plasmids used in the neutralization panels.

The National Institutes of Health has filed patents relating to neutralizing Abs in this article: WO2013086533 A1, titled “Neutralizing antibodies to HIV-1 and their use,” and WO2013070776 A1, titled “Neutralizing gp41 antibodies and their use.” The authors have no financial conflicts of interest.