The vertebrate adaptive immune system modifies the genome of individual B cells to encode antibodies that bind particular antigens1. In most mammals, antibodies are composed of heavy and light chains that are generated sequentially by recombination of V, D (for heavy chains), J and C gene segments. Each chain contains three complementarity-determining regions (CDR1–CDR3), which contribute to antigen specificity. Certain heavy and light chains are preferred for particular antigens2–22. Here we consider pairs of B cells that share the same heavy chain V gene and CDRH3 amino acid sequence and were isolated from different donors, also known as public clonotypes23,24. We show that for naive antibodies (those not yet adapted to antigens), the probability that they use the same light chain V gene is around 10%, whereas for memory (functional) antibodies, it is around 80%, even if only one cell per clonotype is used. This property of functional antibodies is a phenomenon that we call light chain coherence. We also observe this phenomenon when similar heavy chains recur within a donor. Thus, although naive antibodies seem to recur by chance, the recurrence of functional antibodies reveals surprising constraint and determinism in the processes of V(D)J recombination and immune selection. For most functional antibodies, the heavy chain determines the light chain.

Half a billion years of evolutionary battle forged the vertebrate adaptive immune system, an astonishingly versatile factory for molecules that can adapt to arbitrary attacks. The history of an individual encounter is chronicled within a clonotype: the descendants of a single fully rearranged adaptive immune cell. For B cells, reading this immune history for an individual remains a fundamental challenge of modern immunology. Identification of such clonotypes is a magnificently challenging problem for three reasons: The cell history is inferred rather than directly observed: the only available data are the sequences of V(D)J molecules occurring in a sample of cells.Each immune receptor is a pair of V(D)J molecules. Identifying these pairs at scale is a technological challenge and cannot be done with perfect accuracy—real samples are mixtures of cells and fragments thereof.These molecules can be intensely mutated during the optimization of the response to particular antigens, blurring distinctions between kindred molecules.It is thus impossible to determine clonotypes exactly. All solutions to this problem make a trade-off between sensitivity and specificity; useful solutions must address actual artifacts found in real data.We present enclone1, a system for computing approximate clonotypes from single cell data, and demonstrate its use and value with the 10x Genomics Immune Profiling Solution. To test it, we generate data for 1.6 million individual B cells, from four humans, including deliberately enriched memory cells, to tax the algorithm and provide a resource for the community. We analytically determine the specificity of enclone’s clonotyping algorithm, showing that on this dataset the probability of co-clonotyping two unrelated B cells is around 10-9. We prove that using only heavy chains increases the error rate by two orders of magnitude.enclone comprises a comprehensive toolkit for the analysis and display of immune receptor data. It is ultra-fast, easy to install, has public source code, comes with public data, and is documented at bit.ly/enclone. It has three “flavors” of use: (1) as a command-line tool run from a terminal window, that yields visual output; (2) as a command-line tool that yields parseable output that can be fed to other programs; and (3) as a graphical version (GUI).

The vertebrate adaptive immune system modifies the genome of individual B cells to encode antibodies binding particular antigens1. In most mammals, antibodies are composed of a heavy and a light chain which are sequentially generated by recombination of V, D (for heavy chains), J, and C gene segments. Each chain contains three complementarity-determining regions (CDR1-3), contributing to antigen specificity. Certain heavy and light chains are preferred for particular antigens2–21. We considered pairs of B cells sharing the same heavy chain V gene and CDRH3 amino acid sequence and isolated from different donors, also known as public clonotypes22,23. We show that for naive antibodies (not yet adapted to antigens), the probability that they use the same light chain V gene is ∼10%, whereas for memory (functional) antibodies it is ∼80%. This property of functional antibodies is a phenomenon we call light chain coherence. We also observe it when similar heavy chains recur within a donor. Thus, though naive antibodies appear to recur by chance, the recurrence of functional antibodies reveals surprising constraint and determinism in the processes of V(D)J recombination and immune selection. For most functional antibodies, the heavy chain determines the light chain.

Humans who have recovered from infectious disease possess a memory B cell pool that contains highly-potent antibodies against the cleared pathogenic agent. This pool is a valuable source of therapeutic antibodies but identifying these requires a method that can screen a large number of cells rapidly to identify promising candidates. We used Barcode Enabled Antigen Mapping (BEAM) to screen 100 million peripheral blood mononuclear cells (PBMCs) from a donor who had recovered from COVID-19. This method labels antigens of interest with unique reporter oligonucleotides and uses these labeled antigens to stain lymphocytes according to the binding specificity of their antigen receptors. Individual antigen-bound cells are then captured and analysed using the 10x Genomics Single Cell Immune Profiling Solution. This allows us to identify the antigen-specific cells while also generating transcriptomic profiles and natively paired, heavy and light chain full-length B-cell receptor sequences from each cell. We used BEAM to identify B cells that bound to the spike protein of SARS-CoV-2 and discovered 222 antibodies with high binding affinity confirmed by SPR. The majority of these antibodies bound to multiple variants of concern with some also recognising endemic coronaviruses. We then investigated the potential for these antibodies to neutralize live SARS-CoV-2 and confirmed that 55 exhibited potent neutralization activity. Finally, we performed epitope binning and discovered that the collection of antibodies bound to multiple different epitopes on the spike protein. Compared with hybridoma or other approaches, single cell methods have the potential to discover therapeutic antibodies more rapidly and with a higher diversity of leads.

No abstract