Complementarity-Determining Regions (CDRs)
These contain the residues most likely to bind antigen and thus they must
be retained in the humanised antibody.
They are defined by sequence according
to Kabat or by structure
The advantage of using the latter definitions is that the CDRs are shorter,
therefore the humanised antibody will have less xenogeneic fragments in it.
However, the experience of other groups
(e.g. antibody MaE11) has shown that using
the Kabat definition generally requires fewer iterations in the humanisation
Originally defined by Chothia, they have now been revised by Andrew Martin.
His web page allows input of the variable region sequences and automatic
identification of the canonical structure class and important residues.
Canonical residues should be retained in the humanised antibody if they are
different to those in the human frameworks. Numerous examples of the importance
of canonical residues can be found in the
list of Humanized Antibodies page.
Interchain Packing Residues
Defined by Novotny and Haber, and Chothia at the
interface between VL and VH. Generally,
unusual packing residues should be retained in the humanised antibody
if they differ from those in the human frameworks. Their importance is
illustrated in the humanisation of 1B4.
Unusual Framework Residues
These can be located using Andrew Martin's program. For mouse antibodies, it is better
to determine the subgroup according to Kabat and identify the residue positions which differ from the consensus.
These donor specific differences may point to somatic mutations which enhance activity.
Unusual residues close to the binding site may possibly contact the antigen
(antibody BMA 031 provides an example).
However, if they are not important for binding, then it is desirable to get
rid of them because they may create immunogenic
neoepitopes in the humanised antibody. Note that sometimes unusual residues
in the donor sequence are actually common residues in the acceptor
(see humanised anti-Tac). Unusual residues in
the acceptor frameworks are not desirable because of the possibility of
Potential N- or O-Glycosylation Sites
It has been known for a long time that potential N-glycosylation sites are
specific to the consensus pattern Asn-Xaa-Ser/Thr. It must be noted that the
presence of the consensus tripeptide is not sufficient to conclude that an
asparagine residue is glycosylated, due to the fact that the folding of the
protein plays an important role in the regulation of N-glycosylation. It
has been shown that the presence of proline between Asn and Ser/Thr will
inhibit N-glycosylation; this has been confirmed by recent statistical
analysis of glycosylation sites, which also shows that about 50% of the sites
that have a proline C-terminal to Ser/Thr are not glycosylated.
It must also be noted that there are a few reported cases of glycosylation
sites with the pattern Asn-Xaa-Cys.
Potential N-glycosylation sites can be located with the
Removal of a potential N-glycosylation site has not destroyed the binding
avidity of a humanised antibody thus far (see antibodies
AN100226m and SK2)
and in the case of M195 it
O-glycosylation sites are usually found in helical segments which means that
they are uncommon in the beta-sheet structure of antibodies. They have no
See humanised antibody
BrE-3 for thoughts on O-glycosylation.