Fig 1.
Display of hemagglutinin on yeast and evaluation of binding of single domain antibodies.
(A) Schematic of the yeast display vector pNIBS-5. HA0 is full length hemagglutinin gene from A(H1N1) pdm09 comprising HA1 (blue) and HA2 (violet) domains fused in frame with yeast cell surface anchor protein AGA2 and a SV5 epitope tag. Expression and display is mediated by aga signal sequence (ss) and a galactose inducible promotor (GAL1). Display of HA0 is detected using a anti-SV5 monoclonal antibody (mAb). (B) Detection of A(H1N1)pdm09 HA0 display. Display and correct folding of full-length HA0 was confirmed by co-staining of yeast cells with an anti-SV5 mAb and the conformational specific IgG antibodies FC41 or RM10 control antibodies which bind to conformational epitopes in the HA stem or head domain respectively. (C) FACS plots of seven HA specific single domain antibodies (sdAbs) R1a-F5, R1a-G6, R2b-E8, R2b-D9, R1a-A5, R1a-B6, R2a-G8 [8] binding to yeast displayed HA0. Negative controls sdAb R1a-G2 and no sdAb control are shown. The vertical arrow indicates absence of binding.
Table 1.
Location of sdAb epitope to stem or head region of hemagglutinin.
Fig 2.
A general strategy for high-throughput epitope mapping of single domain antibodies to hemagglutinin.
Generation of panels of high affinity monoclonal single domain antibodies (sdAbs or nanobodies) to hemagglutinin using immunisation of alpacas with HA and phage display technology, (i) design of a library of HA variants, (ii) display library on yeast cell surface, (iii) selection using flow cytometric cell sorting to enrich hemagglutinin variants that lose binding to sdAbs but retain display of correctly folded HA on yeast cell surface, (iv) pools of enriched HA mutants are then analysed using deep sequencing, (v) mutations are enriched and their frequencies in the selected population relative to the non-selected population are identified using bioinformatic analysis. Functional loss of binding is experimentally determined to confirm residues are energetically important and contribute to the antibody epitope. This approach can be used to generate a database of epitopes corresponding to diverse collections of sdAbs specific for HA, which upon the emergence of a new viral strain can be used to predict which antibodies could be chosen as suitable binding reagents for applications in diagnosis, research, immune surveillance and vaccine potency testing.
Fig 3.
Mapping of antibody epitopes using conventional sequence analysis of a HA library selected with single domain antibodies binding the head domain.
(A) FACS plots showing the initial HA library and following outputs incubated with 100 nM of R1a-G6 (round 1 and 2) and 100 nM of R1a-B6 (round 3). HA display (anti-SV5) is shown on the x-axis and R1a-G6 antibody binding (anti-cMyc tag) is shown on y-axis. For each round the gated population for cell sorting is shown. (B) Summary analysis of epitopes of sdAbs R1a-F5 and R1a-G6 located to the head domain (HA1). For each antibody epitope, the key residues are shown in bold. The position of antibody epitopes relative to loop 130, Ca2, Ca1 and Sa antigenic sites are indicated [3]. Specific mutations identified are given in S1 Table. (C) Example flow cytometry histograms plots of sdAbs R1a-G6 and R1a-F5 binding mutant and wild-type A(H1N1)pdm09 HAs displayed on yeast cells isolated from cell sorting (the sdAb used for the isolation of the specific HA mutants is indicated in parenthesis). The vertical arrow indicates no binding as expected relative to mutants isolated using the specific sdAb. (D) Surface structure of hemagglutinin HA from A(H1N1)pdm09 (PDB structure 3AL4) showing the HA1 domain (blue) and HA2 domain (violet). Receptor binding site (RBS) is indicated in yellow and key residues comprising the epitopes of R1a-G6 and R1a-F5 are shown in red.
Fig 4.
Mapping of antibody epitopes using deep sequence analysis of a HA library.
(A) The table highlights residues predicted to be involved in sdAb binding and residues which have a minimal effect on antibody binding. Deep sequencing data from a single round of sorting were analysed and mutations identified. The enrichment factors for single amino acid mutations were calculated as the ratio of a given mutation after selection relative to its ratio prior to selection. Positions showing mutations which were highly enriched (E≥5x) were identified as mutational ‘hotspots’ (red squares) and predicted to have a direct involvement in antibody binding, whereas positions which did not yield any enriched mutations (E<5x) were predicted to have little role in antibody binding (green squares). The epitope map shows mutational hotspots and coldspots within the region Ile6 and Glu68 spanning the HA2 domain (region HA1-Gly303 to HA2-Ala5, which gave only ‘coldspots’ except R329G for R1a-B6, is not included in Fig 4A. Detailed listing of all mutations is given in S2 Table. Analysis of the unselected library showed all residues were mutated to between 3 and 8 different amino acids (S4 Table). Sequencing datasets are available for download through accession number PRJEB15301. (B) Highly enriched mutated residues after three rounds of selection are shown in relation to the HA2 domain and fusion peptide for sdAbs R2b-D9, R1a-B6, R2a-G8, R1a-A5, R2b-E8. The fusion peptide is shown with grey arrow (HA2-Gly1 to HA2-Gly23, H3 numbering). For each antibody epitope, the key residues were reported in bold and residues where antibody binding was unaffected by mutation are shown as dots. (C) Mutant HA genes carrying single amino acid mutations at seven different positions predicted by deep sequencing were tested experimentally to confirm their role in antibody binding. Flow cytometry histograms are shown for antibody binding to wild-type (WT) H1N1 HA and each of the single point mutations indicated within the HA2 domain. The sdAbs are grouped as head-binding (R1a-G6) and stem-binding (R2b-D9, R2a-G8, R1a-B6, R1a-A5, R2b-E8). Mutations that eliminate antibody binding are shown in red, those that reduce binding but do not completely eliminate it are shown in yellow and those that have no effect on binding are shown in green. Each mutations was shown to have no effect on HA display (grey histogram). We determined the extent of antibody binding as follows; the MFI value of each antibody-mutant pair was divided by the value of the wild-type H1N1 HA incubation, and the resulting ratio normalized to percentage values. Relative binding of sdAbs to each displayed mutant was categorized as follows; ≤20% no binding (red), between 20% and 40% intermediate binding (yellow) and ≥40% strong binding (green).
Fig 5.
R1a-B6 key residues on Hemagglutinin A(H1N1)pdm09 crystal structure.
(A) A total of 5151 full-length HA sequences corresponding to H1N1, H5N1, H2N2 and H9N2 viral subtypes were aligned and the relative diversity at non-conserved positions was evaluated and showed as a logo sequence (S3 Table). Alignment of HA2 Gly12-Asn60 is shown. Residues predicted to form part of the epitope footprint of our stem binding sdAb panel and identified by yeast display and deep sequencing indicated by black arrows (Fig 5B and 5C). Non-conserved residues either within subtype or across subtypes are highlighted by grey boxes. (B) Residues that show diversity but are buried in the HA structure, positioned on the reverse face of the HA monomer or at the interface of a HA trimer were not considered for mutagenesis and testing (black residues) (S3 Table). Residues that vary across viral subtypes are surface exposed and close in the structure to the binding footprint defined by deep sequencing and to residues tested in Fig 5 (residues highlighted in red) were chosen for experimental testing (orange residues).
Fig 6.
Experimental testing of naturally occurring mutations within epitope footprint.
(A) Table showing binding activity of sdAbs to a panel of HA mutants carrying naturally occurring mutations. The extent of antibody binding was determined by dividing the MFI value of each antibody-mutant pair by the value of the wild-type H1N1 HA incubation, and the resulting ratio normalized to percentage values. Relative binding of sdAbs to each displayed mutant were categorized as no binding (-, red) and strong binding (+, green). (B) Flow cytometry histograms showing binding of R2a-G8 and R2b-E8 to wild-type H1N1 and panel of HA mutants. Absence of binding for D19A,Y34M, I45F and D46N mutation are indicated with a black arrow.
Fig 7.
Relative binding footprint of individual within the HA stem epitope.
Surface structure models of hemagglutinin (HA) A(H1N1)pdm09 (PDB structure 3AL4) showing the two domains (HA1 in blue, HA2 in violet) and the key epitope residues of each sdAb. The key epitope residues of human antibody CR6261, defined by X-ray crystallography [22], is indicated with a dotted black line and demonstrates overlapping sdAb epitopes. The epitope footprint of each sdAb is shown in red relative to the HA stem and combines residues identified by deep mutational scanning (Fig 5) and rational mutagenesis with naturally occurring subtype specific substitutions (Fig 7). The final panel shows a exploded view of the R1a-B6 epitope with key Gly20, Trp21 and Ile45 residues shown in red. Helix A and fusion peptide are highlighted in yellow and green respectively.
Table 2.
Correlation of sdAb binding with the D46N mutation in H1N1 hemagglutinin strains between 1918 and 2010.