A low-complexity, tissue-function-guided engineering workflow for AlphaFold 3-assisted affinity maturation of a CD21 diagnostic nanobody

High-affinity antibodies with reliable tissue-staining performance are needed for reproducible immunohistochemistry (IHC), yet affinity maturation workflows often optimize recombinant-antigen binding and treat staining in formalin-fixed, paraffin-embedded tissue as a downstream validation step rather than as a selection criterion. We developed an AlphaFold 3-assisted, IHC-guided engineering workflow and applied it to an anti-CD21 nanobody, VHH4, whose parental VHH4-Fc format showed insufficient staining intensity for diagnostic use. Domain-resolved AlphaFold 3 modeling localized the VHH4 epitope to the short consensus repeat 8–9 region of CD21 and nominated nine residues in or near complementarity-determining regions for mutational analysis. Combined alanine-scanning and IHC analyses identified five positions at which alanine substitutions reduced tissue staining. Notably, G101A improved apparent recombinant-CD21 binding affinity but reduced IHC staining. A focused mutagenesis workflow then identified Y32 and G101 as plastic engineering hotspots after experimental testing of 64 single-mutant variants, providing a compact alternative to large display-library screening for this proof-of-concept case. Combining beneficial substitutions at these positions yielded the Y32V/G101W dual mutant, which improved binding affinity by 10.72-fold and increased staining intensity in formalin-fixed, paraffin-embedded tissue by 2.45-fold relative to parental VHH4-Fc and by 2.26-fold relative to the commercial anti-CD21 antibody control. Across the characterized variants, recombinant CD21 binding affinity correlated positively but incompletely with tissue staining intensity (Pearson’s r = 0.49, P < 0.0001), supporting direct tissue-staining assessment alongside binding measurements when selecting variants in this CD21/VHH4 engineering campaign. AlphaFold 3-based structural comparison and molecular mechanics-Poisson-Boltzmann surface area calculations suggested that Y32V improves intramolecular hydrophobic packing, whereas G101W strengthens hydrophobic and backbone contacts at the CD21 interface. This study provides a focused engineering workflow that integrates structure prediction, low-complexity mutagenesis, quantitative binding assays, and direct IHC screening of formalin-fixed, paraffin-embedded tissue for diagnostic nanobody maturation. This proof-of-concept case showed that computationally prioritized affinity maturation can reduce experimental burden and improve IHC performance while preserving tissue specificity, but it also highlights the need to select diagnostic antibodies in their intended tissue context rather than on recombinant-antigen affinity alone.