A Hybrid Adherent/Suspension Cell-Based Selection Strategy for Discovery of Antibodies Targeting Membrane Proteins

Abstract

Membrane proteins are favored drug targets and antibody therapeutics represent the fastest-growing category of pharmaceuticals. However, there remains a need for rapid and effective approaches for the discovery of antibodies that recognize membrane proteins to develop a robust clinical pipeline for targeted therapeutics. The challenges associated with recombinant expression of membrane proteins make whole cell screening techniques desirable, as these strategies allow presentation of the target membrane proteins in their native conformations. Here, we describe a workflow that employs both adherent cell-based and suspension cell-based whole cell panning methodologies to enrich for specific binders within a yeast-displayed antibody library. The first round of selection consists of an adherent cell-based approach, wherein a diverse library is panned over target-expressing mammalian cell monolayers in order to debulk the naïve library. Subsequent rounds involve the use of suspension cell-based approaches, facilitated with magnetic-activated cell sorting (MACS) or fluorescence-activated cell sorting (FACS), to achieve further library enrichment. Finally, we describe a high-throughput approach to screen target binding and specificity of individual clones isolated from selection campaigns.

Keywords: Antibody discovery; Biopanning; Cell panning; Directed evolution; Membrane proteins; Molecular engineering; Yeast surface display.

Fig. 1

Enrichment of specific (blue) versus nonspecific (red) scFv-expressing yeast following one round of adherent cell-based selections under four different selection conditions. An anti-programmed death-ligand 1 (PD-L1) scFv-expressing yeast clone was spiked into a naïve library of 10⁹ diversity [16], and 10¹⁰ total yeast were subjected to a single round of biopanning using the following conditions: (i) One 30 min negative selection (NS) + 60 min positive selection (PS); (ii) One 30 min NS + 30 min pre-block with soluble target-null mammalian cells at a 2:1 target-null:target-expressing cell ratio (PB) + 60 min PS; (iii) Two 30 min NS + 30 min PB + 60 min PS; and (iv) Two 30 min NS + 30 min PB + 120 min PS. Results showed that the additional NS and implementation of the PB did not significantly impact specific enrichment ratios, and in fact, implementation of PB with only a single NS actually decreased specific enrichment. However, lengthened PS incubation time (120 min) in combination with an additional NS and PB resulted in the highest specific enrichment ratio

Fig. 2

Minimizing enrichment of nonspecific yeast through rounds of suspension cell-based selections via MACS. An adherent cell-based Round 1 (R1) was conducted for a naïve yeast-displayed scFv library with diversity of 10⁹ [16], and different suspension cell-based sorting conditions via MACS were implemented for R2-R4: (i) 30 min negative selection (NS) + 60 min positive selection (PS); (ii) 30 min pre-block (PB) with soluble target-null mammalian cells at a 2:1 target-null:target-expressing cell ratio (PB) + 60 min PS; (iii) 30 min PB with 10:1 target-null:target-expressing mammalian cell ratio + 60 min PS; and (iv) 30 min NS + 30 min PB with 10:1 target-null:target-expressing mammalian cell ratio + 60 min PS. The graph presents the percent of target-null cells that are bound to yeast (i.e., nonspecific binders). We found that employing both the NS and PB leads to reduced nonspecific clone enrichment, although the target-null:target-expressing mammalian cell ratio does not significantly affect enrichment outcome

Fig. 3

Minimizing enrichment of nonspecific yeast in suspension cell-based selections via FACS. Flow cytometry histograms depict an enriched yeast-displayed scFv library binding to target-null mammalian cells. In each plot, the population at left represents unbound mammalian cells, whereas the population at right represents mammalian cells that are bound to yeast. Suspension cell-based selections were carried out using FACS under the following conditions: (i) 2:1 ratio of target-null:target-expressing mammalian cells in the pre-block step (Round 4, 2:1); and (ii) 10:1 ratio of target-null:target-expressing mammalian cells in the pre-block step (Round 4, 10:1). We found that use of the 2:1 and 10:1 ratios led to similar enrichment of nonspecific yeast binders

Fig. 4

Gating scheme and cell populations for FACS selections. The sorted sample contained 5 × 10⁶ anti-cmyc antibody-labeled yeast cells, 1 × 10⁶ Violet Cell Trace^™ dye-labeled target-expressing mammalian cells, and 2 × 10⁶ unlabeled target-null mammalian cells. Six populations arose from this cell mixture: (1) unbound scFv-expressing yeast cells; (2) yeast-bound target-null cells; (3) yeast-bound target-expressing cells; (4) unbound non-expressing yeast cells; (5) unbound target-null cells; and (6) unbound target-expressing cells. The upper right quadrant contains the yeast-bound target-expressing cells, which represents the desired sort population

Fig. 5

Titration of CellTrace^™ dye for yeast labeling. Specific scFv-expressing yeast were labeled with various concentrations of CellTrace^™ Violet dye and then incubated with unlabeled target-expressing cells. Flow cytometry analysis allows discrimination of unbound mammalian cells (left peak) from yeast-bound mammalian cells (right peak)

Fig. 6

Gating scheme for yeast/mammalian cell binding analysis. CellTrace^™ Violet-labeled PD-L1-expressing Chinese hamster ovary (CHO)-K1 cells were incubated with either yeast expressing an anti-programmed cell death protein 1 (PD-1) scFv (Control Yeast) or yeast expressing an anti-PD-L1 scFv (Specific Yeast). Fluorophore-conjugated anti-cmyc antibody was used for detection of scFv-expressing yeast cells. The left panel depicts the gating scheme for mammalian cells from all events (CellTrace^™ Violet histogram). The middle and right panels consist of the mammalian cell population, wherein a yeast-unbound (left peak) and a yeast-bound (right peak) population exist. Distinct profiles are observed for anti-PD-1 scFv-expressing (Control) versus anti-PD-L1 scFv-expressing (Specific) yeast binding to PD-L1-expressing CHO-K1 cells, demonstrating specific binding