I’m using scripy to cluster B-cell receptor clonotypes from plasma cells using the tl.define_clonotype_clusters (specifying: receptor_arms = 'all', dual_ir = 'all', metric = 'alignment', sequence = 'aa', same_v_gene = True as arguments) and tl.clonotype_network.
The biggest clonotype cluster I’m getting is comprised of IgA, IgG, and IgM B-cell receptor classes, which have nearly identical variable and joining regions and CDR3 sequences.
Biologically, I’m not sure I understand what this means - are these B-cell receptors targeting the same antigen despite their different classes?
I guess there is no way to run tl.define_clonotype_clusters and having it cluster B-cell receptors only if they are from the same class, meaning I’d need to run this function separately for each B-cell receptor class. Is that the right thing to do?
The specificity of antibodies is not determined by their isotype (“receptor class”) but by the variable regions encoded in CDR3. Isotype classes are associated with maturation stage and may reflect effector functions/tissue location. For example, IgM antibodies are found in naive B cells, while IgG can be found in activated B cells after affinity maturation. During that process there is class switching from IgM to IgG. In other tissues, like intestine, IgA class can be more important.
If they antigen recognition regions are identical, they should probably target the same antigen although during affinity maturation there is somatic hypermutation process that causes the generation of higher affinity antibodies. So perhaps the VJ regions are identical, but the aa sequence is not?
Thanks for the discussion.
I’m running scirpy’s tl.define_clonotype_clusters based on CDR3 aa sequence similarity (sequence = 'aa'), in addition to requiring that both receptor arms are used (receptor_arms = 'all', dual_ir = 'all'), and having the same V gene (same_v_gene = True). So I’m getting clusters that have identical V genes and identical or nearly identical CDR3 AA sequences, but different classes. I wouldn’t be surprised if these only included IgM and IgG classes, because what you mention, but the fact that I’m also getting IgA in these clusters (even from the same individual sample) is surprising to me. Would you also expect that?
It all depends on whether the information from the constant region is considered as a measure of dissimilarity during clustering. Since that function uses CDR3 region, then constant region is not used and so it is not surprising that different isotypes clusters together. But I think there is no issue with that. I am not sure why do you want to have different isotypes in different clusters. You have your clusters by CDR3. You can also tell clearly which isotypes are which. Or filter out isotypes you are not interested in.
What I’m doing is clustering BCRs by cell-type (e.g, plasma cells separately from memory cells), however I’m getting clusters that are comprised of IgAs, IgMs, and IgGs that have the same V(D)J+CDR3AA sequences. At face value this suggests that these BCRs are recognizing the same antigen, which I find surprising, at least for IgA and IgG, although I’m not a B-cell immunologist so perhaps it’s only my lack of familiarity.