A comprehensive search of functional sequence space using large mammalian display libraries created by gene editing

Abstract

The construction of large libraries in mammalian cells allows the direct screening of millions of molecular variants for binding properties in a cell type relevant for screening or production. We have created mammalian cell libraries of up to 10 million clones displaying a repertoire of IgG-formatted antibodies on the cell surface. TALE nucleases or CRISPR/Cas9 were used to direct the integration of the antibody genes into a single genomic locus, thereby rapidly achieving stable expression and transcriptional normalization. The utility of the system is illustrated by the affinity maturation of a PD-1-blocking antibody through the systematic mutation and functional survey of 4-mer variants within a 16 amino acid paratope region. Mutating VH CDR3 only, we identified a dominant "solution" involving substitution of a central tyrosine to histidine. This appears to be a local affinity maximum, and this variant was surpassed by a lysine substitution when light chain variants were introduced. We achieve this comprehensive and quantitative interrogation of sequence space by combining high-throughput oligonucleotide synthesis with mammalian display and flow cytometry operating at the multi-million scale.

Keywords: CRISPR/Cas9; IgG antibody library; Mammalian display; TALE nuclease; affinity maturation; fluorescence-activated cell sorting; gene editing; gene targeting; human therapeutic antibody discovery; magnetic-activated cell sorting.

Figure 1.

Antibody display cell line generation by nuclease mediated gene integration (a) Recognition sequence for left and right TALE nucleases are shown underlined and in upper case either side of the spacer region (shown in lower case). Emboldened and italicized sequence represents sequence of CRISPR target sequence (b) Representation of genomic AAVS locus. The AAVS cleavage site is located within a 4428 bp intron between the first and second exons of the gene encoding protein phosphatase 1, regulatory subunit 12C, PPP1R12C. TALE nucleases or CRISPR/Cas9 nucleases directed to this region are used to cleave the genome at this site. Hatched boxes on the 5ʹ and 3ʹ side of the cleavage site represent the left and right homology arms (HA), respectively (c) Representation of pD2 donor vector used to insert and display human IgG-formatted antibody genes. The transgene region is flanked by left and right homology arms (left HA, right HA) representing the sequences which flank the cleavage site in the AAVS locus. The vector encodes a promoter-less blasticidin gene, a dual promoter antibody expression cassette with antibody light chain and a human IgG2 heavy chain driven by pEF and CMV promoters, respectively. Expressed antibodies are anchored on the cell surface by a PDGFR transmembrane domain (TM) (d) Flow cytometric analysis of HEK293 cells displaying an anti-lysozyme or anti-PD1 antibody on the cell surface. Non-transfected HEK293 cells (left panel) and cells displaying either anti-lysozyme (middle panel) or anti-PD1 337_1_C08 (right panel) were stained with anti-Fc PE and either lysozyme conjugated with DyLight 633 (top row) or PD-1-biotin/streptavidin-APC.

Figure 2.

Nuclease-directed integration of single copy antibody genes. HEK293 cells transfected with either plasmid αFGFR1-pD6 or αFGFR1-pD6 (50:50 mix) were introduced via TALE nuclease into the AAVS locus of HEK293 cells, selected for blasticidin resistance for 13 days and stained with either FGFR1-Dy633 (a) or FGFR2-Dy488 (b). Plots show single color histogram overlaying anti-FGFR1 and anti-FGFR2 transfected cells. Dot plot shows dual staining to detect binding to FGFR1 (y-axis) and FGFR2 (x-axis). Samples are: (c) 300 ng donor, CRISPR/Cas9 via Amaxa; (d) 100 ng donor, CRISPR/Cas9 via Amaxa; (e) 300 ng donor, TALEN via Amaxa; (f) 300 ng donor, TALEN via PEI transfection.

Figure 3.

Affinity maturation using mammalian display. Schematic illustration of the experimental workflow (a). At 9dpt transfected cells were sorted for antigen binding using 10 nM PD-L1. The gate within the plot represents the population that was sorted, and the number represents percentage of the total cells (b). The sorted population was grown for an additional 14days and stained with anti-Fc PE and PD-L1-biotin at a concentration of (c) 10nM, (d) 1nM, (e) 0.1nM followed by detection with Streptavidin-APC. The dot-plot shows antibody expression (x-axis) and antigen binding (y-axis). Affinity ranking by capture ELISA (f) of anti- PD-L1 IgG, FACS selected at 10 nM (light hatch) or 1 nM PD-1 (dark hatch).

Figure 4.

Anti- PD-1 VH and VL CDR3 library creation by “total synthesis”. Schematic diagram showing a representation of mutation of the VL CDR3 (a) and VH CDR3 (b) 8-amino acid windows by oligonucleotide “total synthesis” where every possible encoded 1-mer and 2-mer substitution was made. The libraries were cloned to the targeting vector pINT17-BSD (c) prior to transfection of HEK293 cells.

Figure 5.

Mammalian display of variant libraries. HEK293 cells transfected with pINT17-337_1_C08 (a & e), pINT17-NNS library (b & f), pINT17-VH library (c & g) and pINT17-VH:VL combined library (d & h) were introduced via CRISPR/CAS9 (a-d) or TALE nuclease (e-h) into the AAVS locus of HEK293 cells. 14dpt cells were dual stained with 10 nM human PD-1-Avi tag antigen followed by streptavidin APC (y-axis) and anti-human Fc PE (x-axis). Affinity ranking ELISA of the antibodies isolated from the “VH only” Selection (i), “VH:VL-1nM” selection (j) and “VH:VL-0.1nM” selection (k). 337_1_C08 and nivolumab are used as references and the line represents the antibodies with improved affinities.

Figure 6.

Inhibition of the interaction of PD-1 with PD-L1 in a cell reporter assay. PD-L1 artificial antigen presenting CHO-K1 (aAPC/CHO-K1) cells were co-cultured with PD-1 effector Jurkat cells, which stably express human PD-1 and a NFAT-RE-luciferase reporter, in the presence or absence of anti-PD-1 antibodies (50 nM final concentration). The luminescence signal in relative light units (RLU) is plotted on the y-axis for each antibody named on the x-axis.

Figure 7.

Concentration dependent inhibition of PD-1 blockade in a cell-based assay. PD-L1 artificial antigen presenting CHO-K1 (aAPC/CHO-K1) cells were co-cultured with PD-1 effector Jurkat cells, which stably express human PD-1 and a NFAT-luciferase reporter, in the presence or absence of varying concentrations of anti-PD-1 antibodies. The luminescence signal (y-axis) is plotted against the concentration of inhibitory antibody for K5T8-9, K5E3-61, nivolumab, and the parental “wild-type” clone 331_1_C08.