Effect of Humanizing Mutations on the Stability of the Llama Single-Domain Variable Region

Abstract

In vivo clinical applications of nanobodies (VHHs) require molecules that induce minimal immunoresponse and therefore possess sequences as similar as possible to the human VH domain. Although the relative sequence variability in llama nanobodies has been used to identify scaffolds with partially humanized signature, the transformation of the Camelidae hallmarks in the framework2 still represents a major problem. We assessed a set of mutants in silico and experimentally to elucidate what is the contribution of single residues to the VHH stability and how their combinations affect the mutant nanobody stability. We described at molecular level how the interaction among residues belonging to different structural elements enabled a model llama nanobody (C8WT, isolated from a naïve library) to be functional and maintain its stability, despite the analysis of its primary sequence would classify it as aggregation-prone. Five chimeras formed by grafting CDRs isolated from different nanobodies into C8WT scaffold were successfully expressed as soluble proteins and both tested clones preserved their antigen binding specificity. We identified a nanobody with human hallmarks that seems suitable for humanizing selected camelid VHHs by grafting heterologous CDRs in its scaffold and could serve for the preparation of a synthetic library of human-like single domains.

CDR1 Composition Can Affect Nanobody Recombinant Expression Yields

“…There are several examples demonstrating the feasibility of this approach, but these are still the results of case-by-case projects that require a major input from specialized researchers and often require access to structural information [9][10][11][12]. On the other hand, we recently confirmed that publicly accessible algorithms that require minimal data inputs can rapidly and effectively help in the identification of protein variants that exhibit astonishingly higher yields than the original sequences [13].…”

“…The A10 mutG0 clone (6.9 mg/L medium) produced roughly 30% more than A10 wt (5.2 mg/L), but both A10 mutKG and A10 mutKG.1 had higher production yields (9.6 and 12.8 mg/L, respectively). Interestingly, whereas the mutations from “human-like” to the canonical “llama” residues in the framework of A10 wt did not result in higher production yields [ 12 ], the mutant with glycine insertion (A10 mutKG.1) profited from the “camelization” of the original partially humanized sequence (A10 mutKG). In contrast, the further mutation present in A10 mutKG.2 (K/L in CDR3) had a remarkable negative effect on the nanobody yield, suggesting that the wt CDR3 was already stable and did not require further optimization.…”

“…The other major innovative result of this work is having demonstrated the relevance of CDRs as potential instability hot spots. Indeed, CDR grafting experiments performed in the past have demonstrated that molecular scaffolds had variable capacity to host heterologous CDRs differing for length and antibody composition in their framework [ 4 , 12 , 39 , 40 ]. The consequence is that several constructs are not stable and are eliminated during selection procedures.…”

“…The possibility of improving clone stability, increasing their yields, and maintaining or improving the sequence humanization by rational mutagenesis would represent a substantial advancement, and community efforts have already resulted in the development of successful approaches. There are several examples demonstrating the feasibility of this approach, but these are still the results of case-by-case projects that require a major input from specialized researchers and often require access to structural information [ 9 , 10 , 11 , 12 ]. On the other hand, we recently confirmed that publicly accessible algorithms that require minimal data inputs can rapidly and effectively help in the identification of protein variants that exhibit astonishingly higher yields than the original sequences [ 13 ].…”

“…To reduce the risk of disrupting the protein’s biological function, the user is recommended to specify active-site regions or residues. We previously used the anti-Her2 ectodomain nanobody A10 [ 4 , 17 ] with affinity in the low nM range as a model to evaluate in silico optimization approaches [ 12 ] and therefore also considered it to be the most suitable candidate to undergo a PROSS-mediated mutagenesis strategy. At the time at which this project started, no structural model was yet available to describe the interaction between A10 and the Her2, and it was assumed that the nanobody paratope was restricted to the amino acids in the CDRs.…”