Antibody Clonality Assessment
Antibody Clonality Background
The serological compartment is simple to access and contains a multitude of proteins, a large fraction of which are circulating antibodies. At Abterra Bio, we developed Alicanto which utilizes B cells from a variety of sources, including peripheral blood mononuclear cells (PBMCs), to help deconvolute the sequences of serum antibodies. However, it is still difficult to determine the primary sequences of the circulating antibodies as their unique somatic rearrangements are not always encoded in accessible B cells or plasma cells.
Relevant B cells are particularly challenging to obtain in the case of autoimmune diseases, infectious disease response from years prior, or other latent or chronic immune responses. In these cases a de novo sequencing approach must be taken – such as Griffin.
While Griffin delivers full-length antibody sequences, in this blog post we will focus on characterizing the clonal complexity of different samples by analyzing CDR3 sequences. CDR3 sequences influence the antibodies specificity, and define a clonal lineage. CDR3 cataloging is the first stage of Griffin, but can also be valuable by itself for tracking a clone across time; across tissue; similar clones across individuals; as well as other use-cases.
Antibody Clonality Methods
We utilized datasets of varying complexity to showcase the clonality assessment performed as part of Griffin. Namely we used:
- a monoclonal antibody (NISTmAB)
- a mixture of human monoclonals
- a polyclonal from a hyperimmunized camelid
- a human polyclonal from an infectious disease response
Griffin proceeds by sequencing CDR3s from the antibody protein sample, and conducting an analysis to understand sequence similarity, overlap of samples/replicates, and overall quality.
Antibody clonality assessment workflow is the first stage of the Griffin polyclonal antibody sequencing service.
Antibody Clonality Results
The first questions to ask of each sample are:
- How many CDR3s are found?
- How expanded are the clones?
We answer these questions by clustering CDR3s by the edit distance of their sequences. Generally, the number and complexity of the networks correlated with the expected complexity of the sample. Note the monoclonal (NISTmAb) has a deamidation variant shown.
Clustering complexity per-sample. The monoclonal shows a dominant node and deamidation PTM variant. The mix of mAbs shows a small number of clones. The camelid immunization was higly purified and simplified. The natural human disease response was antigen enriched and shows the most complexity and clonal depth.
We can delve deeper into each CDR3 to understand how many times it was sampled, and the confidence in each amino acid using the per-residue score distribution. This helps in determining which CDR3s are the most confident in the sample. In the plot below, the left is NISTmAb – sampled highly with high confidence, partly due to the lower complexity of the sample. On the right is a human infectious disease response, and the selected CDR3 is still high confidence, but has certain residues with less certainty than others.
Per-residue qualities of CDR3s. Per-residue distributions are scores from our deep learning model to assess correctness of each residue. The distributions are aggregated over multiple samplings of each full-length sequence.
Furthermore, we can associate isotype subclasses to specific CDR3s directly. Griffin achieves this by partially assembling the region surrounding the CDR3 – including the downstream J region and beginning of the constant region. Isotyping can be of particular importance given specific indications that are being investigated which may have isotype bias (see Volkov et al., 2022 for examples in autoantibodies). We show the variants of human IgG1/2/3/4 below, demonstrate that we can distinguish between IgG1/IgG24/IgG3, but not between IgG2 and IgG4.
Isotype association of CDR3s for human sample. Top: Clonal assignment of isotypes; Bottom: human IgG isotypes aligned showing variants with respect to IgG1. Of the 7 clones shown, 2 can be assigned to IgG1, 2 to IgG2, IgG3, or IgG4, and the other 3 cannot be assigned a particular subclass. IgG2 and IgG4 cannot be associated with clones as variants are further downstream. If the isotype could not be determined it is noted as simply IgG.
Conclusion
Antibody clonality from only serum samples can be assessed, quantified, and analyzed easily with Griffin. In particular we showed:
- Many clones can be identified and tracked in a variety of input samples, species, and contexts.
- A correlation with complexity of sample and complexity of clustered serum clones.
- CDR3s can be assessed at a per-residue scale for quality.
- Isotype subclasses can be associated with a subset of CDR3s for human samples.