Comparison of mammalian and bacterial expression library screening to detect recombinant alpha-1 proteinase inhibitor variants with enhanced thrombin inhibitory capacity

Serpins are a widely distributed family of serine proteases. A key determinant of their specificity is the reactive centre loop (RCL), a surface motif of ∼20 amino acids in length. Expression libraries of variant serpins could be rapidly probed with proteases to develop novel inhibitors if optimal systems were available. The serpin variant alpha-1 proteinase inhibitor M358R (API M358R) inhibits the coagulation protease thrombin, but at sub-maximal rates compared to other serpins. Here we compared two approaches to isolate functional API variants from serpin expression libraries, using the same small library of API randomized at residue 358 (M358X): flow cytometry of transfected HEK 293 cells expressing membrane-displayed API; and a thrombin capture assay (TCA) performed on pools of bacterial lysates expressing soluble API. No enrichment for specific P1 residues was observed when the RCL codons of the 1% of sorted transfected 293 cells with the highest fluorescent thrombin-binding signals were subcloned and sequenced. In contrast, screening of 16 pools of bacterial API-expressing transformants led to the facile identification of API M358R and M358K as functional variants. Kinetic characterization showed that API M358R inhibited thrombin 17-fold more rapidly than API M358K. Reducing the incubation time with immobilized thrombin improved the sensitivity of TCA to detect supra-active API M358R variants and was used to screen a hypervariable library of API variants expressing 16 different amino acids at residues 352-357. The most active variant isolated, with TLSATP substituted for FLEAI, inhibited thrombin 2.9-fold more rapidly than API M358R. Our results indicate that flow cytometric approaches used in protein engineering of antibodies are not appropriate for serpins, and highlight the utility of the optimized TCA for serpin protein engineering.