Protein engineering by phenotypic screening in microfluidic droplets

This proposal addresses the engineering of functional proteins by directed evolution, where the selection criterion is the effect of this protein on a cell (instead of KD). G protein-coupled receptors (GPCRs) constitute an important drug target class, but raising antibody binders against them has been difficult because GPCRs can often only be expressed at low levels in cells, are relatively unstable and difficult to handle, making their purification and subsequent assays (e.g. affinity panning) difficult. Although these problems can occasionally be overcome, approaches that bypass the need to purify a target are attractive alternatives. We plan to carry out high throughput cell-based assays of GPCR effectors in picolitre water-in-oil emulsion droplets that can be produced at frequencies >1kHz in microfluidic devices (see Kintses et al. Chem Biol 2012,19, 1001-9.). We display candidate proteins on DNA using our in vitro SNAP-display method (see e.g. Kaltenbach et al. ChemBioChem, 2011 12,:2208-16; Diamante et al. Prot Eng Sel Design, 2013, 26,713). We will then co-compartmentalise one protein display construct with one (or a few cells). The co-compartmentalisation makes it possible to carry out a cell-based assay that probes the effect with a library well above 10e6 members that are rapidly evaluated on-chip within a few hours. Crucially the readout will reflect a desired cellular phenotype, not just a binding event (as in affinity panning approaches). Furthermore, fluorescence (or absorbance) assays in droplets are followed with a high precision optical setup and can be carried out on a single encapsulated cell, which allows valuable data that is masked in ensemble assays to be used for identification of interesting molecules. These features distinguish our approach from current flow cytometric analysis, which is also a single-cell method, but lacks the ability to co-compartmentalise the protein display construct. This means that a reaction volume > 1 Mu L is needed for one assay - compared to 100 pL in microdroplets. Thus the throughput of the droplet-based approach would be more than 1000-fold higher.

Specifically we will search for molecules that target the glucose-dependent insulinotropic polypeptide receptor (GIPR). Previous work at MEDI has identified an antagonist antibody for the GIP receptor (J Biol Chem. 2013, 288(27):19760-72; JBC 288:27), but an agonist would be potentially interesting therapeutically. Agonising compounds for GIPR (or the related GLPR and glucagon receptor) are potentially interesting therapeutically because they can stimulate insulin secretion and improve glucose control in diabetic patients. We are not aware of antibodies that can act in this way, so success in a phenotypic screening campaign would create uniquely useful molecules. We will miniaturise existing optical assays that respond to GIPR agonists by creating a detectable fluorophore and test the dynamic range with single (or multiple detector cells; see our previous work on single cells in droplets in J Am Chem Soc 2009, 131(42):15251-6). Then cells will be co-compartmentalised with libraries of antibodies (ScFvs) or ankyrin repeat proteins (DARPIns) displayed on their coding DNA. After identification of hits in cell-based assays in droplets for agonist phenotype, the corresponding DNA will be isolated, sequenced and the candidate proteins expressed in quantity for verification in cell-based assays and, if validated, in mouse models.