Isolation of anti-CD22 Fv with high affinity by Fv display on human cells

Abstract

In vitro antibody affinity maturation has generally been achieved by display of mouse or human antibodies on the surface of microorganisms (phage, bacteria, and yeast). However, problems with protein folding, posttranslational modification, and codon usage still limit the number of improved antibodies that can be obtained. An ideal system would select and improve antibodies in a mammalian cell environment where they are naturally made. Here we show that human embryonic kidney 293T cells that are widely used for transient protein expression can be used for cell surface display of single-chain Fv antibodies for affinity maturation. In a proof-of-concept experiment, cells expressing a rare mutant antibody with higher affinity were enriched 240-fold by a single-pass cell sorting from a large excess of cells expressing WT antibody with a slightly lower affinity. Furthermore, we successfully obtained a highly enriched mutant with increased binding affinity for CD22 after a single selection of a combinatory library randomizing an intrinsic antibody hotspot. Important features are that one display selection cycle requires only 1 week, and transfection of cells in a single 100-mm dish produces 10(7) individual clones so that a repertoire of 10(9) is feasible under current experimental conditions.

Fig. 1.

Mammalian cell surface display. (A) Cell surface display of scFv on mammalian cells and comparison with other display systems. In the mammalian (HEK 293T) cell surface display system, the scFv (anti-CD22) containing VH-linker (L)-VL is fused to the N-terminal portion of the PDGFR transmembrane domain. Murine Ig κ-chain signal peptide (SP) and c-myc epitope tag (T) are used (see Materials and Methods). In a M13 phage display system, the scFv with the g3p signal peptide is fused to the N-terminal of g3p (10). In a typical yeast cell surface display system, the scFv with A-agglutinin signal peptide fused to the C-terminal portion of S. cerevisiae a-agglutinin yeast adhesion receptor (Aga2) (see Materials and Methods). VH, Ig heavy-chain variable region; VL, Ig light-chain variable region. (B) Schematic illustration of surface display on mammalian cells. An additional 10-aa epitope tag (c-myc) was fused to the C terminus of the scFv (based on the anti-CD22 RFB4 Fv structural model) (10), allowing quantitation of fusion display with mAb 9E10 independent of antigen (CD22) binding. Fusion to the N-terminal portion of the PDGFR transmembrane domain was used to anchor scFv on the mammalian (HEK 293T) cell surface. In the present study, Gly-91 and Asn-92 located in VL are mutagenized for affinity maturation.

Fig. 2.

Confocal microscopic images of HEK 293T cells displaying scFv. HEK 293T cells transfected with a plasmid directing surface expression of anti-CD22 scFv were grown on coverslips. (A–C) Transfected cells were fixed with DAPI nuclear staining (A) followed by detection with biotinylated CD22-Fc (B), and anti-c-myc antibody (C) followed by Steptavidin-Alexa Fluor-594 (red) and anti-mouse IgG-Alexa Fluor-488 (green). (D) Merged staining patterns (yellow) are shown. (Scale bar: 25 μm.)

Fig. 3.

Flow cytometric analyses of mammalian (A and B) and yeast cells (C and D) displaying anti-CD22 scFv. HEK 293T (A and B) and S. cerevisiae (C and D) cells displayed the same anti-CD22 scFv construct (HA22) at the cell surface. (A and B) In mammalian display, HEK 293T cells (A) or transfected cells (B) were labeled with biotinylated CD22-Fc proteins followed by streptavidin-R-PE conjugate and mAb 9E10 (anti-c-myc epitope tag) followed by an anti-mouse IgG-FITC conjugate. (C and D) In yeast display, noninduced (C) and induced (D) yeast cells were labeled with biotinylated CD22-Fc proteins followed by streptavidin-R-PE conjugate and mAb anti-V5 epitope tag followed by an anti-mouse IgG-FITC conjugate. About 40% of yeast cells do not express scFv on their surface, resulting in two histogram peaks.

Fig. 4.

Enrichment of mammalian cells displaying an improved scFv variant by kinetic selection and flow cytometric sorting. (A and B) The dot plot shows only 50,000 cells of the total cell population (10⁷). Each dot represents an individual observed cell. This is the 1:400 mixture of HA22:BL22-displaying cells, labeled with no antigen (A) or 1 nM antigen (CD22-Fc) (B) concentration for separation. FITC fluorescence (binding to the c-myc tag) represents the number of surface expressions on an individual cell. The PE fluorescence represents binding to the antigen (CD22). (C) A sort window was drawn to include the top 0.1% of total cells in terms of ratio of PE/FITC fluorescence (29). Cells that fell within the window were collected.

Fig. 5.

Interaction between CD22 and HEK 293T cell surface-displayed scFv. (A) Four individual clones (PT.1, PT.2, PT.3, and PT.4) of scFvs with the same mutation (PT) isolated from a single-pass enrichment were tested for the scFv expression (scFv) on HEK 293T cells with anti-myc mAb detected by FITC and for the binding to the antigen with CD22-Fc (CD22) detected by PE. Numbers of fluorocromes on a HEK 293T cell were calculated by using calibration beads (see Materials and Methods). The numbers of CD22-bound scFvs and total scFvs and their ratios are shown with those of the original HA22 scFv on the surface of HEK 293T cells. (B) Equilibrium binding titration curves to determine dissociation constants, K_D. MFI (%) of PE is plotted versus the various concentrations of biotinylated CD22-Fc used to label surface-displayed BL22 (WT, K_D = 5.8 nM, B_max = 453), HA22 (K_D = 2.5 nM, B_max = 293), or mutant PT (K_D = 1.2 nM, B_max = 275) antibody. Data are fit with a nonlinear least-squares regression, as described in Materials and Methods.