Shining a light on phenotypic drug discovery

SUMMARY

In this issue of Cell Chemical Biology, Seneviratne et al. (2021) combine photoaffinity labeling and quantitative chemical proteomics to identify the molecular target of a lead compound discovered from a phenotypic drug screen. Their work showcase the power of coupling a photoreactive group to screening hits for rapid target deconvolution.

Drug discovery comes in different flavors. Target-centric approaches seek to identify small molecules that engage a molecular target with a defined (or hypothesized) role in disease pathogenesis (Swinney and Anthony, 2011). A target-focused strategy provides a specific biological hypothesis to be tested and a starting point for the identification of compounds to achieve therapeutic efficacy. The identification of potential drug targets is facilitated by the wealth of knowledge on molecular drivers of human disease that are readily available from genomics and other large-scale molecular profiling technologies widely used in academia and the pharmaceutical industry (Overington et al., 2006). Advances in high-throughput screening as well as structure- and computational-based tools have facilitated the identification and progression of a lead hit from a candidate compound to a drug that can engage a molecular target with efficacy.

Despite the predominance of target-centric strategies, a renewed interest in phenotypic drug discovery is driven by the exciting potential for discovery of first-in-class drugs. This target-agnostic approach does not rely on prior knowledge of the identity of a particular drug target or a hypothesis about its role in disease. Compounds are screened in cellular models that aim to reconstruct a cellular phenotype that closely resembles disease conditions. In contrast with target-based approaches where hit compounds bind the same target (and often the same site), hits arising from a phenotypic screen can mediate effects through different target proteins and mechanisms of action. Phenotypic drug discovery is thus well suited for testing a wide range of potential therapeutic mechanisms to discover novel treatments for diseases that are poorly understood by the scientific community and lack valid molecular targets (Moffat et al., 2017). In this issue of Cell Chemical Biology, the Pfizer team led by Uthpala Seneviratne and Martin Pettersson address a key step for enabling phenotypic drug discovery: target deconvolution (Seneviratne et al., 2020).

The Pfizer team established a phenotypic screen to identify small molecules capable of enhancing apolipoprotein E (apoE) secretion in human astrocytes. Human apoE is a 299 amino acid glycoprotein that is expressed predominantly in the liver and brain (Kim et al., 2009). In the central nervous system (CNS), nonneuronal cells including astrocytes and microglia (to a lesser degree) are the principal cell types that secrete apoE. The apoE protein functions as a ligand in receptor-mediated endocytosis of lipoprotein particles and thus plays an important role in lipid transport and cellular metabolism. The authors pursued identification of small molecule modulators of apoE secretion as chemical probes for studying apoE regulation in the context of neurodegenerative disease. The human APOE gene contains several single-nucleotide polymorphisms (SNPs) with the three most common isoforms leading to changes in the coding sequence: apoE2 (C112, C158), apoE3 (C112, R158), and apoE4 (R112, R158). These changes in amino acids at residues 112 and 158 result in profound alterations to apoE structure and function. Importantly, the 4 allele of the APOE gene was discovered to be a key genetic risk factor for both late onset Alzheimer’s disease (AD) and cerebral amyloid angiopathy (Kim et al., 2009). Strikingly, the risk factor for AD in individuals with one and two ε4 alleles is increased ~3- and ~12-fold respectively, compared to individuals without ε4 alleles. The deleterious effects from the APOE ε4 allele could be due to gain of toxic function and/or loss of neuroprotective function. Although the exact pathological mechanism is currently unknown, evidence supports a role for apoE isoforms in regulation of amyloid-β aggregation and clearance (Reiman et al., 2009). Thus, therapeutic strategies are directed principally at decreasing the toxic effects of apoE4 or increasing apoE expression and function to restore physiological functions. This study represents an important step towards testing the latter hypothesis by discovering a small molecule capable of enhancing apoE secretion in human astrocytes and identifying its protein target as the liver X receptor β (LXRβ) to understand mode of action (MOA).

A phenotypic screen was performed to identify enhancers of astrocytic apoE using human astrocytoma CCF-STTG1 as the model cell line because of its ability to spontaneously secrete lipidated apoE particles (Wahrle et al., 2008). A large compound library (>400K compounds) was screened to identify compounds that substantially increased (by three standard deviations from the mean) apoE levels resulting in a collection of ~5,900 hit compounds for secondary assays. A hit triage strategy was implemented to remove compounds that enhanced apoE secretion through mechanisms that could lead to toxicity including effects on the general secretory machinery and peripheral lipogenesis. The remaining hit compounds were further triaged by potency, physicochemical properties, and ability to enhance apoE secretion in primary human astrocytes to identify a pyrrolidine-containing sulfonamide lead compound (Compound 1, Figure 1). A key step towards understanding MOA for compounds identified by phenotypic screening campaigns is target deconvolution. To address this question, the Pfizer team developed a probe analog of compound 1 that contained (i) a benzophenone photoreactive group to facilitate covalent cross-linking with interacting proteins upon irradiation with UV light (Dorman et al., 2016) and, (ii) an alkyne handle to append reporter tags for visualizing (fluorophore) and LC-MS/MS identification (biotin for affinity chromatography) of probe-modified target proteins (Photoprobe 2, Figure 1). The authors performed quantitative chemical proteomics (Cisar and Cravatt, 2012) using SILAC light and heavy CCF-STTG1 cells, UV-mediated crosslinking in live cells, and LC-MS/MS to identify the protein-photoprobe 2 interactome (~700 proteins). They demonstrate using a competitive assay that only LXRβ probe labeling and enrichment was substantially competed in compound 1-treated cells. LXRβ was further validated as the molecular target of compound 1 by immunoblotting for endogenous LXRβ, recapitulation of compound 1 and photoprobe 2 activity using recombinant LXRβ, and cellular thermal shift assays (CETSA) to demonstrate target engagement of compound 1 with native LXRβ in CCF-STTG1 cells. Finally, the authors map the site of binding for photoprobe 2 in the ligand binding pocket of purified LXRβ and predict S278 or V279 as the site for covalent crosslinking (Figure 1).

Figure 1. Enabling target identification in phenotypic drug discovery by photoaffinity labeling and quantitative chemical proteomics.

A phenotypic screen was implemented to identify small molecules capable of enhancing apoE secretion in human astrocytes. Compound 1 was identified as a lead compound after a hit triage strategy. A probe analog of compound 1 (photoprobe 2) was developed to aid in target identification by embedding a photoreactive benzophenone group to cross-link amino acids in the binding site(s) of photoprobe 2-modified proteins in quantitative chemical proteomic studies. Competition studies with compound 1, which displaces photoprobe 2 labeling, facilitated target (liver X receptor β) and binding site (S278, V279) identifications. The crystal structure of the liver X receptor β shown is from protein data bank (PDB) accession number 1P8D.

In summary, the Pfizer team showcase the power of photoaffinity labeling (PAL) and quantitative chemical proteomics to enable target deconvolution in a phenotypic drug discovery program aimed at discovery of small molecule enhancers of apoE secretion in human astrocytes. Identification of the molecular target is a key step towards understanding MOA of hit compounds that not only de-risks further hit-to-lead efforts but can also initiate rationally designed target-based drug discovery efforts. In a recent study, PAL was combined with fragment-based ligand discovery to expedite discovery of functionally active hits and their respective molecular targets and binding promiscuity (Parker et al., 2017). Advances in target identification technologies, often through basic research in chemical biology groups, will continue to provide a path towards elucidation of MOA of hit compounds to enable phenotypic screening for drug discovery.

The rewards for a successful phenotypic screening campaign are high including, for example, the development of drugs for a disease lacking molecular targets and/or identification of compounds that operate through a novel mechanism. The studies presented by the Pfizer group (Seneviratne et al., 2020) provide an exciting example of how chemical biology is shaping and facilitating the phenotypic drug discovery landscape by providing an experimental path for target identification, which is an important criterion for selecting hit compounds to move forward for lead identification.