High-throughput specificity profiling of antibody libraries using ribosome display and microfluidics

Abstract

In this work, we developed PolyMap (polyclonal mapping), a high-throughput method for mapping protein-protein interactions. We demonstrated the mapping of thousands of antigen-antibody interactions between diverse antibody libraries isolated from convalescent and vaccinated COVID-19 donors and a set of clinically relevant SARS-CoV-2 spike variants. We identified over 150 antibodies with a variety of distinctive binding patterns toward the antigen variants and found a broader binding profile, including targeting of the Omicron variant, in the antibody repertoires of more recent donors. We then used these data to select mixtures of a small number of clones with complementary reactivity that together provide strong potency and broad neutralization. PolyMap is a generalizable platform that can be used for one-pot epitope mapping, immune repertoire profiling, and therapeutic design and, in the future, could be expanded to other families of interacting proteins.

Keywords: CP: Biotechnology; CP: Immunology; PolyMap; SARS-CoV-2 surface antigen; antibody-antigen interaction; cell surface protein; high-throughput screening; polyclonal antibody profiling; protein-protein interaction; ribosome display; single-cell analysis; viral neutralization.

Graphical abstract

Figure 1

Design of the PolyMap platform The PolyMap workflow begins with a library of antigens expressed on the surface of a mammalian cell, which are incubated with a soluble library of antibody scFvs in ribosome display format. Stained single cells are encapsulated with uniquely barcoded RNA capture beads (barcodes are represented by star, triangle, and diamond shapes) and lysed. All mRNA within each droplet (including scFv RNA from the ARM complex and antigen RNA from the cell) is captured on the bead. Beads are isolated and used to generate cDNA, which is then further amplified with gene-specific primers for sequencing. Analysis of the cell barcode, antigen barcode, and antibody CDRH3 is used to generate maps of antibody-antigen interactions.

Figure 2

Development of the PolyMap platform (A) The antigen expression construct includes CMV promoter with translation-enhancing element (2G), signal peptide (SP), transmembrane (TM) region for surface display, and a unique barcode (BC) in the 3′ untranslated region (UTR). An FRT site allows integration and translation of a glutamine synthetase gene (GS) for selection. (B) Surface expression of different spike mutants expressed stably in CHOZN cells as measured by a monoclonal antibody and flow cytometry. (C) The DNA fragment encoding the antibody scFv ribosome display library includes a T7 promoter, strep-tag II, spacer sequence (TolA), secM ribosomal stall sequence, 40-mer poly(A) tail, and T7 terminator. (D) RNA recovered from spike variant cells stained with clinical antibodies as ARM complexes as measured by RT-qPCR and reported as ARMs/cell. The table indicates the expected results based on literature and affinity studies. (E) Drop-seq binding data from PolyMap. The four antigen lines were pooled and stained with ARM complexes generated from an equimolar RNA pool of the five indicated antibodies and then put through the Drop-seq workflow. After merging lists of antigen and antibody cell BCs, 610 cells were analyzed for antibody binding. For each individual cell, the percentage of reads from each antibody is plotted. The horizontal lines represent the mean values. (F) Heatmap representation of the data from (E) showing the average percentage of reads for the antibodies within each cell line.

Figure 3

Binding profiles of 2020 and 2023 libraries of anti-SARS-CoV-2 antibodies against a library of circulating spike variants (A) Antibody repertoire and distribution used to generate ARM complexes were determined by sequencing heavy-chain fragments from the RNA mixture serving as input for the in vitro translation reaction. The individual and cumulative percentages of reads for the top 100 (2020 library) or 20 (2023 libraries) clones are plotted. (B) Distribution of spike antigen identity was determined by the percentage of sequencing reads corresponding to the known antigen barcodes amplified from genomic DNA of the cell line. (C) PolyMap profiling of the 2020 antibody library. The antigen cell line from (B) was stained with ARM complexes generated from the 2020 RNA mix in (A), and individual cells were encapsulated with barcoded beads and taken through the Drop-seq workflow. The percentage of normalized reads for each antibody across all cell lines is plotted (PolyMap score). The top 40 antibodies (based on the total number of antibody reads) are shown for a representative experiment. (D) PolyMap profiling of the pooled convalescent and vaccinated 2023 libraries. All 40 recovered clones are shown from a representative experiment.

Figure 4

Binding validation and functional testing of antibodies profiled by PolyMap (A) A subset of individual antibody clones from the 2020 library were validated by flow cytometry and compared to the previous data. For each antibody, CDR3H sequences and heatmaps for PolyMap profiling (top, from Figure 3C) and flow cytometry (bottom) are shown. For flow cytometry, full-length monoclonal antibodies were expressed in CHO cells, and the supernatant was used to stain cell lines expressing a spike variant. Heatmaps show MFI of the secondary antibody signal. (B) ELISA binding (EC₅₀ values) for antibody mixtures (1, 2, 3, 4, or 10 clones, or IVIG) binding to spike RBD from the indicated variant. (C) Pseudoviral neutralization (IC₅₀ values) of antibody mixtures for the indicated strain. n.d., not determined (no neutralization observed). For (B) and (C), the antibodies in the mixtures were as follows: 23.6 (1); 23.1 and 23.3 (2); 20.6, 23.5, and 23.7 (3); 20.4, 20.6, 23.4, and 23.5 (4); and 20.1–20.5 + 23.1–23.5 (top 10).