Humanization

Humanization takes a binding site raised in another species and puts it into a human framework, to reduce the chance of an immune response against the therapeutic. The binding site is what you keep; everything around it is what you change.

The approaches

CDR grafting moves the complementarity determining regions onto a human germline framework. It is the standard method. The complication is that framework residues support loop conformations, so a pure graft often loses affinity, and selected framework residues have to be changed back to the parental ones.

Resurfacing changes only the surface-exposed framework residues to their human counterparts, leaving the core intact. It disturbs the structure less and removes less of the non-human sequence.

Germlining takes an already human or humanized sequence and moves individual residues toward the closest germline, reducing unusual features that could be immunogenic.

For VHHs, humanization means converting a camelid framework toward human VH3, and the hallmark framework 2 residues that keep the domain soluble without a light chain need care, because those are exactly the positions a naive graft would change.

Back mutations are the real work

The choice of acceptor framework matters less than which parental residues you retain. The usual candidates are the Vernier zone residues that pack under the loops, residues at the VH and VL interface, and anything unusual that a structure shows contacting a CDR.

A practical approach is to build a small series: the straight graft, the graft plus a minimal set of back mutations, and one or two intermediate versions. Express them all, measure affinity and stability, and pick from data. Humanization is one of the few places where making six constructs instead of one reliably saves time.

Humanness scores

Several scores exist, from simple germline identity to model-based measures of how typical a sequence looks against human repertoires. They are useful for comparing candidates and for catching outliers.

They are not predictions of immunogenicity. Clinical immunogenicity depends on T cell epitopes, on the patient population, on dose and route, and on aggregation, which is often a bigger driver than sequence. Where immunogenicity risk is a real decision point, T cell epitope prediction and ex vivo assays address it more directly than a humanness number.

What to measure afterward

Every humanized variant is a new molecule. Re-measure binding in the final format, thermal stability, aggregation and expression. A humanized variant that keeps affinity but loses several degrees of stability has moved the problem rather than solved it.