Binding affinity landscapes constrain the evolution of broadly neutralizing anti-influenza antibodies
- Department of Organismic and Evolutionary Biology, Harvard University, United States
- NSF-Simons Center for Mathematical and Statistical Analysis of Biology, Harvard University, United States
- Quantitative Biology Initiative, Harvard University, United States
- Department of Physics, Massachusetts Institute of Technology, United States
- Department of Physics, Harvard University, United States
- Department of Applied Physics, Stanford University, United States
- Laboratoire de physique de ÍÉcole Normale Supérieure, CNRS, PSL University, Sorbonne Université, and Université de Paris, France
Figures
Figure 1 with 7 supplements
Binding landscapes.
(A) Sequence alignment comparing somatic heavy chains to reconstructed germline sequences. Mutations under study (purple, numbered) and excluded mutations (black) are indicated; residues are …
-
Figure 1—source data 1
CR9114 library -logKD to H1, H3, and influenza B.
Biological triplicates, mean, and standard error reported.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig1-data1-v2.csv
-
Figure 1—source data 2
CR6261 library -logKD to H1 and H9.
Biological triplicates, mean, and standard error reported.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig1-data2-v2.csv
-
Figure 1—source data 3
Isogenic flow cytometry measurements of –logKD for select CR9114 and CR6261 variants.
Inferred –logKD and standard deviation for each replicate of isogenic FACS, alongside inferred –logKD mean and SEM from Tite-Seq using the mean bin and maximum likelihood (ML, shown only for CR9114) inference methods.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig1-data3-v2.xlsx
Figure 1—figure supplement 1
Experimental design and Tite-Seq workflow.
Experimental design and Tite-Seq workflow. (A) Experimental design. Amino acid sequence percent identity of the entire HA ectodomain and the stem epitope (Dreyfus et al., 2012) are shown between …
Figure 1—figure supplement 2
Tite-Seq data quality.
(A, C) Correlation of (A) CR9114 and (C) CR6261 KD measurements between biological replicates. (B, D) Validation of (B) CR9114 and (D) CR6261 Tite-Seq KD measurements by isogenic flow cytometry …
Figure 1—figure supplement 3
Antibody-antigen co-crystal structures.
(A) Alignment of co-crystal structure of CR9114 with H5 (light hues; PDB ID 4FQI) and CR9114 with H3 (dark hues; PDB ID 4FQY). Mutated residues shown as gray spheres. (B) Co-crystal structure of …
Figure 1—figure supplement 4
Force-directed graph for CR6261.
Force-directed graph for CR6261. (A, B) Force-directed graph for CR6261 H1 –logKD, as in Figure 1G. Nodes are colored by binding affinity to (A) H1 and (B) H9.
Figure 1—figure supplement 5
Expression of antibody libraries.
(A) Correlation of mean expression across Tite-Seq biological replicates for CR9114 library (left, red) and CR6261 library (right, blue). (B) Correlation between Tite-Seq mean expression and …
Figure 1—figure supplement 6
Tite-Seq gating strategy.
First, single yeast cells were gated by forward scatter (FSC) and side scatter (SSC) (step 1). Single cells were then either gated by scFv expression or HA binding. For the expression sort (step …
Figure 1—figure supplement 7
Reversions of excluded mutations.
(A) Reversion of A24S and E46D in CR9114 (somatic-16) does not substantially impact binding affinity compared to the fully somatic version of CR9114 (somatic-18) to either H1 (orange) or H3 …
Figure 2 with 1 supplement
First- and second-order effects.
(A, B) First-order effects inferred in best-fitting epistatic interaction models for (A) CR9114 and (B) CR6261. Mutations required for binding (present at over 90% frequency in binding sequences) …
-
Figure 2—source data 1
Interaction model coefficients for CR9114.
Coefficients are reported with standard errors, p-values, and confidence intervals (95% with Bonferroni correction by the number of parameters).
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig2-data1-v2.csv
-
Figure 2—source data 2
Interaction model coefficients for CR6261.
Coefficients are reported with standard errors, p-values, and confidence intervals (95% with Bonferroni correction by the number of parameters).
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig2-data2-v2.csv
-
Figure 2—source data 3
Tabulated contact surface area, number of HA contacts, and pairwise distances for mutations in CR9114 and CR6261.
For each mutated position, the contact surface area with HA (as plotted in Figure 2C and Figure 2—figure supplement 1) and the number of HA residues within six angstroms is tabulated for CR9114-H5 (4FQI), CR9114-H3 (4FQY), and CR6261-H1 (3GBN). Distances between alpha-carbons plotted in Figure 2F and Figure 2—figure supplement 1 are also tabulated here alongside the corresponding second order effects.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig2-data3-v2.xlsx
Figure 2—figure supplement 1
Structural context of first and second order effects.
(A) Left: first-order effects for each site, colored by effect size and plotted against the contact surface area between the corresponding somatic residue and HA (top, CR9114 with H3; bottom, CR6261 …
Figure 3 with 3 supplements
High-order epistasis for CR9114.
(A) Total higher order epistatic contributions of CR9114 mutation pairs for binding H1. Color bar indicates the sum of absolute values of significant higher order interaction coefficients involving …
Figure 3—figure supplement 1
CR9114: interactions between five key sites.
(A) CR9114 force-directed graph, as in Figure 3D, colored by mutation groups of sites 30, 57, 65, 82, and 83 (32 total groups). The dashed line emphasizes the observed separation of genotypes with …
Figure 3—figure supplement 2
CR9114: interactions between other sets of five sites.
(A) CR9114 force-directed graph, as in Figure 3D, but colored by mutation groups of a different set of five sites with fewer strong epistatic interactions (35, 36, 64, 66, and 85). (B) CR9114 …
Figure 3—figure supplement 3
High-order epistasis for CR9114 binding to H3.
(A) Higher order significant epistatic contributions of CR9114 mutation pairs, as in Figure 3A, for binding H3. Light yellow columns indicate required mutations (sites 57, 82, and 83). Significance …
Figure 4 with 2 supplements
High-order epistasis for CR6261.
(A) Total significant epistatic contributions of CR6261 mutation pairs for binding H1, as in Figure 3A. Significance is given by Bonferroni-corrected p-value < 0.05, see Materials and methods. (B) …
Figure 4—figure supplement 1
CR6261: interactions between four sites.
(a) CR6261 force-directed graph, as in Figure 1—figure supplement 4, colored by mutation groups of sites 29, 35, 82, and 83 (16 total groups). (B) Top, coefficients for terms in the epistatic …
Figure 4—figure supplement 2
High-order epistasis for CR6261 binding to H9.
(A) Higher order significant epistatic contributions of CR6261 mutation pairs, as in Figure 4A, for binding H9. (B) Scatterplot of significant epistatic interaction model coefficients for binding to …
Figure 5 with 3 supplements
Antigen selection scenarios and likely mutational pathways.
(A, B) Maximum binding affinity achievable for sequences with a given number of mutations. For each antigen for (A) CR9114 and (B) CR6261, the maximum observed (circles) and model-predicted …
-
Figure 5—source data 1
Total probability of mutational trajectories for CR9114 under different antigen selection scenarios.
Mean and standard error across 10 bootstrap samples are reported for moderate, weak, and strong selection strengths.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig5-data1-v2.csv
-
Figure 5—source data 2
Total probability of mutational trajectories for CR6261 under different antigen selection scenarios.
Mean and standard error across 10 bootstrap samples are reported for moderate, weak, and strong selection strengths.
- https://cdn.elifesciences.org/articles/71393/elife-71393-fig5-data2-v2.csv
Figure 5—figure supplement 1
Selection models.
(A) Functional form of mutation step probability, illustrated for parameters chosen to represent strong, moderate, and weak selection models. (B, C) Total log probability of the mutational …
Figure 5—figure supplement 2
Variant probabilities for CR9114 under the optimal (’O’) selection model.
(A) Histogram of the total probability of all pathways passing through each variant in the optimal selection scenario, divided by the total probability in a model with no selection, transformed to …
Figure 5—figure supplement 3
Probability of mutation order assuming moderate selection, under other antigen selection scenarios.
H1 (A) and 'R' (B) for CR9114 and 'O' (C) and 'R' (D) for CR6261, as in Figure 5I,J. For the random mixed scenario 'R', the representative cases from Figure 5G,H are shown.
Appendix 1—figure 1
Correlation between from ML inference on Tite-Seq data vs. from isogenic flow cytometry.
to H1 (salmon), H3 (green), and Flu B (blue) shown for select variants, identical to those shown in Figure 1—figure supplement 2B. Pearson’s r = 0.97.
Appendix 1—figure 2
Distributions of PE-A fluorescence (HA binding) for isogenic CR9114 strains incubated with H3.
PE-A fluorescence distributions from flow cytometry of isogenic CR9114 germline (left) and somatic (right) strains following incubation with 1 μM, 100 nM, and 10 nM H3, as described in Methods. …
Appendix 2—figure 1
Results from statistical epistasis models for CR9114.
Appendix 2—figure 2
Results from statistical epistasis models for CR6261.
Appendix 2—figure 3
Variance partitioning of statistical epistasis models.
(A) Variance partitioning for CR9114 binding to H1 (left) and H3 (right). (B) Variance partitioning for CR9114 binding to H1 and H9, denoted by colors as indicated.
Appendix 2—figure 4
Epistasis inference at full order.
(A,B) Numbers of significant coefficients for the full-order inference compared to optimal truncated regression models for (A) CR9114 and (B) CR6261. Significance for both model types is determined …
Appendix 2—figure 5
Results from epistasis models with nonlinear transformations.
(A), Fitting logistic functions to additive predicted phenotypes. Red lines indicate the optimized logistic function , with as indicated. (B), Scatterplot of coefficients from the optimal …
Tables
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Strain, strain background (Saccharomyces cerevisiae) | EBY100 | ATCC | Cat#:MYA-4941 | |
| Cell line (Spodoptera frugiperda) | Sf9 | ThermoFisher | Cat#:B82501 | Cell line for production of baculovirus |
| Cell line (Trichoplusia ni) | High-Five | ThermoFisher | Cat#:B85502 | Cell line for HA expression |
| Antibody | Anti-cMyc-FITC (Mouse monoclonal) | Miltenyi Biotec | Cat#:130-116-485 | FACS (1:50) |
| Recombinant DNA reagent | pCT302 (plasmid) | Addgene | Cat#:41845 | |
| Recombinant DNA reagent | pCT302_CR9114 _germline (plasmid) | This paper | Plasmid map in Supplementary file 4 | |
| Recombinant DNA reagent | pCT302_CR9114 _somatic (plasmid) | This paper | Plasmid map inSupplementary file 5 | |
| Recombinant DNA reagent | pCT302_CR6261 _germline (plasmid) | This paper | Plasmid map inSupplementary file 6 | |
| Recombinant DNA reagent | pCT302_CR6261 _somatic (plasmid) | This paper | Plasmid map inSupplementary file 7 | |
| Recombinant DNA reagent | pET21a-BirA (plasmid) | Addgene | Cat#:20857 | |
| Sequence-based reagent | CR9114 golden gate dsDNA fragments | IDT | Sequences listed inSupplementary file 2 | |
| Sequence-based reagent | CR6261 Golden Gate primers | IDT | Sequences listed inSupplementary file 3 | |
| Sequence-based reagent | Illumina sequencing primers | IDT | Sequences listed inSupplementary file 1 | |
| Peptide, recombinant protein | Streptavidin-RPE | ThermoFisher | Cat#:S866 | FACS (1:100) |
| Peptide, recombinant protein | Biotinylated A/New Caledonia/99 (H1) ectodomain | This paper | Plasmid sequence inSupplementary file 8 | |
| Peptide, recombinant protein | Biotinylated A/Hong Kong/99 (H9) ectodomain | This paper | Plasmid sequence inSupplementary file 9 | |
| Peptide, recombinant protein | Biotinylated A/Wisconsin/05 (H3) ectodomain | This paper | Plasmid sequence inSupplementary file 10 | |
| Peptide, recombinant protein | Biotinylated B/Ohio/05 (Flu B) ectodomain | This paper | Plasmid sequence inSupplementary file 11 | |
| Commercial assay or kit | Bac-to-Bac Kit | ThermoFisher | Cat#:10359016 | |
| Commercial assay or kit | Zymo Yeast Plasmid Miniprep II | Zymo Research | Cat#:D2004 | |
| Software, algorithm | Custom code | This paper | https://github.com/klawrence26/bnab-landscapes (copy archived at swh:1:rev:61c1673a101ea739d5b7e9b282f6bcfad41d7e90, Phillips, 2021) | |
| Software, algorithm | Interactive CR9114 data browser | This paper | https://yodabrowser.netlify.app/yoda_browser/ |
Additional files
-
Supplementary file 1
Primer sequences for sequencing library preparation.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp1-v2.xlsx
-
Supplementary file 2
Fragment sequences for Golden Gate construction of the CR9114 library.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp2-v2.xlsx
-
Supplementary file 3
Primer sequences for Golden Gate construction of the CR6261 library.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp3-v2.xlsx
-
Supplementary file 4
Plasmid map of pCT302 with CR9114 germline sequence.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp4-v2.dna
-
Supplementary file 5
Plasmid map of pCT302 with CR9114 somatic sequence.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp5-v2.dna
-
Supplementary file 6
Plasmid map of pCT302 with CR6261 germline sequence.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp6-v2.dna
-
Supplementary file 7
Plasmid map of pCT302 with CR6261 somatic sequence.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp7-v2.dna
-
Supplementary file 8
Plasmid map of pFastBac with influenza A/New Caldeonia/1999 H1 ectodomain.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp8-v2.seq
-
Supplementary file 9
Plasmid map of pFastBac with influenza A/Hong Kong/1999 H9 ectodomain.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp9-v2.seq
-
Supplementary file 10
Plasmid map of pFastBac with influenza A/Wisconsin/2005 H3 ectodomain.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp10-v2.seq
-
Supplementary file 11
Plasmid map of pFastBac with influenza B/Ohio/2005 HA ectodomain.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp11-v2.seq
-
Supplementary file 12
Inferred CR9114 VH germline nucleotide sequence.
- https://cdn.elifesciences.org/articles/71393/elife-71393-supp12-v2.dna
-
Transparent reporting form
- https://cdn.elifesciences.org/articles/71393/elife-71393-transrepform-v2.pdf
