In vitro evolution of an influenza broadly neutralizing antibody is modulated by hemagglutinin receptor specificity

Abstract

The relatively recent discovery and characterization of human broadly neutralizing antibodies (bnAbs) against influenza virus provide valuable insights into antiviral and vaccine development. However, the factors that influence the evolution of high-affinity bnAbs remain elusive. We therefore explore the functional sequence space of bnAb C05, which targets the receptor-binding site (RBS) of influenza haemagglutinin (HA) via a long CDR H3. We combine saturation mutagenesis with yeast display to enrich for C05 variants of CDR H3 that bind to H1 and H3 HAs. The C05 variants evolve up to 20-fold higher affinity but increase specificity to each HA subtype used in the selection. Structural analysis reveals that the fine specificity is strongly influenced by a highly conserved substitution that regulates receptor binding in different subtypes. Overall, this study suggests that subtle natural variations in the HA RBS between subtypes and species may differentially influence the evolution of high-affinity bnAbs.

Figure 1. Schematic representation of the yeast display screening.

(a) Schematic representation of the dual promoter yeast expression plasmid encoding the C05 Fab is shown. The arrows indicate the locations of the SfiI sites employed for cloning. (b) Saturation mutagenesis was applied to the six amino acids of C05 CDR H3 that interact closely or are buried in the receptor-binding site of influenza hemagglutinin (HA). A C05 Fab plasmid mutant library, which consists of at least 10⁸ clones, was created. The plasmid mutant library was transformed into yeast strain EBY100 to generate a yeast surface display library. Different variants of C05 Fab were displayed on yeast cells. This yeast surface display library was then subjected to selection for HA-binding affinity. We used fluorescence-activated cell sorting (FACS) to enrich yeast cells that were able to interact with PE-conjugated influenza HA. The post-selection pool was then expanded and subjected to another round of selection. For each of H1, H3, and H5 HAs, three rounds of selection was performed. Variants that were able to bind to the HA would enrich in occurrence frequency throughout the screening process. Variants with higher affinity would enrich to a higher frequency. The plasmid mutant library and each of the post-selection mutant libraries were next-generation sequenced to monitor the frequency change of each variant.

Figure 2. Monitoring the selection process by next-generation sequencing.

(a) The occurrence frequency of each amino acid at each of the 6 residues of interest is shown as a heatmap. (b) For each round of selection, a sequence logo was generated using the amino-acid sequences of the top 10 variants with the highest occurrence frequency.

Figure 3. Validation for C05 variants binding against H1 HA and H3 HA.

(a) Supernatant of 293T cells transfected with plasmids encoding the indicated C05 Fab variants was used to estimate the affinity of C05 Fab variants against immobilized SI06 HA (H1) and A/Perth/16/2009 (Perth09) HA (H3). The relative K_d of WT C05 Fab was set as 1, which is indicated by the green line. For visualizing purpose, the relative K_d is capped at 100. Variants that had an H3-specific binding profile and cited in the main text are boxed. (b) The effect of changing position 100d from Ala to Ser on binding kinetics was investigated. Four pairs of C05 variants, in which variants within each pair were related by an Ala to Ser substitution at position 100d, were included in our validation experiment (a). We computed the fold change in kon and koff by comparing the binding kinetics in the variant that carried a Ser at position 4 against the variant that carried an Ala at position 4. For visualization, for a given pair of variants (i→j), the fold change in kon was calculated as on rate of j divided by on rate of i, whereas fold change in k_off was calculated as off rate of i divided by off rate of j, such that a positive value at the log scale indicates a positive effect in binding. (c) The occurrence frequencies of all observed variants in the sequencing data (a total of 848,630) in round 3 post-selection library and round 1 post-selection library are compared and shown as a scatterplot. Variants with >0.4% occurrence frequency in H1 round 3 post-selection library and <0.003% in H3 round 3 post-selection library were classified as ‘H1 specialists' (shaded in red in the scatterplot). Variants with >0.3% occurrence frequency in H3 round 3 post-selection library and <0.001% in H1 round 3 post-selection library were classified as ‘H3 specialists' (shaded in lilac in the scatterplot).

Figure 4. Interaction between VPGSGW and HK68/H3 HA1.

(a) The crystal structure of VPGSGW in complex with HA1 exhibits a similar conformation to that of C05 WT in complex with HA1 (PDB: 4FP8). (b) There are two VPGSGW-HA1 complexes in the asymmetric unit of the crystal structure. The orientations of the serine at 100d differ between those two complexes. Intramolecular hydrogen bonds are shown for the two different conformations of CDR H3 of VPGSGW. (c) Interaction of the HA RBS (grey) with CDR H3 (pink) is shown. Hydrogen bonds are represented by green dashed lines. Water molecules are represented by red spheres. A putative sodium ion is represented by the purple sphere. For clarity, only G100c, S100d, G100e, and W100f on the CDR H3 of VPGSGW are displayed.

Figure 5. E190D in the HA RBS favors binding against VPGSGW.

The hydrogen bond network involving S100d of the CDR H3 of VPGSGW is shown for (a) binding against HA RBS of HK68/H3, and (b) binding against HA RBS of HK68/H3 that carried mutation E190D, which is modelled based on the crystal structure of VPGSGW-HK68/H3 HA1. A putative sodium ion is represented by the purple sphere. For visual clarity, only G100c, S100d, G100e, and W100f on the Fab are displayed. (c) The affinities of WT C05 Fab (VVSAGW) and a C05 Fab variant (VPGSGW) against the HA from A/Hong Kong/1/1968 (wild type: E190; E190D mutant: D190) are shown as a bar chart. (d,e) The neutralizing activity of WT C05 and VPGSGW in IgG format against (d) SI06-HA/WSN virus, and (e) A/Aichi/2/68 virus were measured by cell viability assay. SI06-HA/WSN virus was generated based on WSN, in which the HA ectodomain was replaced by that from SI06 (see Methods). Colour code is the same as that of c. Mean value across three replicates is shown and the error bar represents the s.d. The large error bar in VPGSGW at 100 μg ml⁻¹ against A/Aichi/2/68 virus is due to complete protection in one but not in the other two replicates.