Phenotypic screens as a renewed approach for drug discovery

Molecular target-based screening

The foundation for a molecular target approach of drug development started with advances in pharmacology, synthetic and medicinal chemistry that began in the early 20th century. The wealth and depth of research performed in the 1950s and 1960s on enzymes and enzyme kinetics provided a method for precise calculation of compound’s potency (IC₅₀ or EC₅₀) and efficacy (% maximal response) of an enzyme [1]. Hundreds of enzymes were discovered and purified during this period, later becoming important molecular targets of drug discovery [1]. The methodology of enzyme kinetics was extended to receptor pharmacology in 1970s [2], although the molecular entity of receptors was largely unexplored at this time. The progressive research in receptor pharmacology and the nature of druggability later made receptors the most popular targets for drug discovery [3,4]. Technological advances in molecular biology and genome science initiated a modern era of molecular target-based approach for drug discovery in late 1980s. Recombinant DNA technology enabled the generation of new assays for a wealth of molecular targets, allowing rapid screens of large chemical libraries using purified recombinant proteins or engineered cell lines [5-8]. This, along with developments in combinatorial chemistry, assay miniaturization and robotic automation, greatly facilitated the emergence and rapid development of high-throughput screening (HTS) in the 1990s [9,10].

The molecular target-based approach for drug discovery, also called ‘reverse pharmacology’ or ‘reverse chemical biology’ [11-13], generally starts with target identification of a disease of interest (Figure 2A). Molecular targets are often discovered in basic research, with studies involving animal disease models and clinical observations of patient phenotypes. For example, an abnormal function of a specific protein, a signaling pathway, or a mutation in a specific gene can be identified in basic research with connection to a disease. Once a suitable target has been identified and validated, assay development is initiated, followed by HTS of chemical libraries to identify hits such as enzyme inhibitors or receptor antagonists against the target. The most active compounds, usually compounds making up one to three lead series, are then confirmed and validated in orthogonal assays that are more physiologically related to the target. This is then followed by chemical optimization to characterize the structure–activity relationship (SAR) of the lead series and to enhance favorable absorption, distribution, metabolism, and excretion (ADME) and pharmacokinetic/pharmacodynamic properties of the compounds. In this paradigm, only a few lead compounds with a defined mechanism of action and demonstrated efficacy in disease models are able to move to preclinical drug development, toxicology studies, and hopefully, clinical trials. In the past 20 years, molecular target-based screening has become the major approach in early drug discovery. G protein-coupled receptors (GPCRs), ion channels and enzymes are the most common and successful molecular targets for drug discovery [5-8]. It is interesting to note that all the biologics approved for treatment of human disease are target-based therapeutics [14]. In contrast to some small molecule compounds, biologics such as proteins (e.g. enzymes, antibodies), hormones, peptides, vaccines, and blood components are made in biological process and their mechanism of action is dependent on a specific protein target.

There is no doubt that molecular target-based screening has some distinct advantages over phenotypic screening (Table 1). [RE1]For example, a molecular target and its related screening assay are often vital in guiding subsequent chemical optimization of lead compounds and necessary to fully characterize the SAR. Additionally, knowledge of a molecular target can help guide toxicology studies during preclinical development. Biomarker development, which is critical for evaluation of drug effects in animal disease models and clinical trials, may also be facilitated by the known molecular target and its signaling pathway.

It has been recognized recently, however, that target-based drug discovery may have its limitations. Recent analysis has revealed that high attrition rates in Phase II and III clinical trials are mainly due to lack of drug efficacy along with other factors [15,16]. Although the lack of drug efficacy in late stage drug development can be the result of multiple factors, including poor correlation of animal models with human diseases and genetic variation of patient populations, invalidated targets for disease is a significant factor for many failed drug candidates. Additionally, the numbers of validated druggable targets currently available for drug development are seemingly more limited than previously thought [8,17]. A recent review of FDA approved drugs indicated that there currently exists only 435 effective drug targets although the success of human genome program has revealed a total of approximately 20,000 human genes that encode approximately 500,000 proteins [8]. Thus, identification of new drug targets from the human genome remains an unmet biomedical research goal.

The success of target-based screens used for drug discovery has also recently come into question. Swinney and Anthony [14] analyzed the first-in-class small molecule drugs approved by the FDA between 1999 and 2008 and found that 28 of them were discovered using a phenotypic screening approach compared to 17 drugs discovered by a molecular target-based approach. This surprising discovery has contributed to growing interest and reconsideration of phenotypic screens for drug development in both pharmaceutical industry and academic research centers, with a hope that newly increased application of this traditional approach can rejuvenate early-phase drug discovery and improve the success rates in late stage drug development (Figure 1).

Phenotypic screening in drug discovery

Today, the main application of cell-based phenotypic assays is to screen large compound libraries, composed of 0.4–2 million compounds, to identify lead compounds for drug discovery projects. Historically, drug discovery was phenotypic by nature – with new drugs either accidently found, as in case of penicillin, or through designed bactericidal screens to discover additional antibiotics [18]. The phenotypic screening approach for drug discovery is also called ‘forward pharmacology’, ‘classical pharmacology’ or ‘forward chemical biology’ [11-13] and the molecular mechanism and protein target can remain unknown even after the drug’s activity and efficacy are determined. Generally, a characteristic associated with the disease is exploited to develop a cell-based assay for a modern phenotypic screen (Figure 2B). Compounds are then screened in the phenotypic assay to identify active lead compounds that ameliorate the disease phenotype, exemplified by selectively killing cancer cells [19], eliminating pathogens in culture [20], or reducing lysosomal cholesterol accumulation in Niemann Pick disease type C patient cells [21].

The phenotypic screen is usually more physiologically relevant and less artificial because intact cells and native cellular environment are used. Primary hits identified in the phenotypic screens can potentially target different types of proteins (receptors, enzymes, transcription factors, among others) and even different signaling pathways. Lead compounds can be further selected from the hits with or without knowledge of the target, although identification of the target can facilitate the SAR study. The phenotypic screen in this ‘forward pharmacology’ process enables lead discovery for many diseases in which a drug target has not been identified and/or validated. Therefore, this approach can have a useful role in drug discovery for many rare diseases which tend to be under studied, and with most lacking an effective drug therapy. Phenotypic screening can also be applied to the discovery of novel drug targets, which may prove useful for common neurological diseases such as Alzheimer’s and Parkinson’s diseases, for which there have been many failures of target-based drug candidates in clinical trials.

Recent retrospective analysis has found that many drugs approved by the FDA (especially those in 1970s) have an unknown mechanism of action or an unknown target [22]. Not surprisingly, many of these early approved drugs were discovered using phenotypic screens and were approved by regulatory agencies before their precise mechanism of action or protein targets were identified. A famous example of this is aspirin (acetylsalicylic acid) for which it took almost 100 years to determine the mechanism of action and molecular target [23]. Calcium channel antagonists [24] including 1-4 dihydropiridines (nifedipine, nicardipine and nimodipine etc.), verapamil and diltiazem were found and developed using phenotypic screens involving smooth muscle relaxation, vasodilatation and reduction of high blood pressure [25,26]. The precise mechanism of action for the treatment of hypertension and other cardiac indications was not clear when the first of these drugs were approved in 1980s. They originally had the generic name of ‘calcium blockers’ [27], while the first L-type calcium channel that these drugs act on was cloned in 1987 [28]. Ezetimibe (Zetia), a cholesterol absorption inhibitor, was also discovered in an animal model with a high cholesterol diet [29,30]. It got FDA clearance in 2002 as a cholesterol lowering drug without a known molecular target [31], which was reported later to be the NPC1L1 cholesterol transporter [32]. Even today, regulatory agencies around the world will approve a new drug without requiring the precise mechanism of action or a molecular target, as long as the drug is efficacious and safe for patients. It should be noted that in depth characterization of the drug properties including mechanism of action and molecular targets can aid in the design of improved next-generation compounds with reduced adverse effects.

Animal-based phenotypic screens

Historically, isolated tissues or animal models were involved in phenotypic screening, as described briefly above. In the past 10–20 years, many disease models of several small animals including Caenorhabditis elegans, zebrafish, Xenopus laevis, and Drosophila melanogaster have been developed and applied to compound screening to achieve relatively high screening throughput [33]. The phenotypic screens using in vivo model systems can provide rich information on compound absorption, distribution, metabolism and toxicity in addition to the valuable efficacy data on disease models. Although the throughput of compound screens in rodents or large animal models are limited, the screening capacity with C. elegans, D. melanogaster, Zebrafish and X. laevis has been improved by using 96-well plates [34-38] although it is still significantly less compared to cell-based assays. One disadvantage to screening with in vivo models is that potential lead compounds with properties of poor drug absorption, quick metabolism, limited cell membrane permeability and toxicity may not be active in the primary screens. The poor relevance of some animal models to human diseases, due to species difference and other reasons, can contribute to the failures in the late stages of drug development. Therefore, cell-based phenotypic screening seems more suitable for primary compound screens to identify physiological and disease relevant lead compounds for drug development.

Cell-based phenotypic assays

With advances in new assay technologies, the throughput of phenotypic screening has greatly improved in the past 10 years (Figure 2). Robotic screening platforms and highly sensitive detection systems have been developed which allow phenotypic assays to be miniaturized and used to rapidly screen large chemical libraries. In contrast to the lack of cellular content of many molecular target-based assays using purified recombinant proteins, cell-based phenotypic assays offer additional biological complexity, with the cellular milieu of interacting proteins and signaling networks, while still maintaining the capacity of HTS. Cell-based phenotypic assays usually use primary human cell lines, immortalized cell lines (primary or engineered), or, more recently, specific cell types differentiated from induced pluripotent stem cells (iPSCs) derived from patient or normal human cells (Table 2).

Cell viability assays, cell signaling pathway assays, and disease-related phenotypic assays are three types of cell-based phenotypic assays commonly performed in lead discovery. These assays can be miniaturized to a robotic screening platform and have higher screening throughput. Active compounds are identified that confer a change in a cellular phenotype such as killing pathogens or cancer cells, activating or inhibiting a signaling pathway, or normalizing a phenotypic change associated with human disease. Additionally, other types of phenotypic assays are also available including autophagy, apoptosis, cell cycle analysis, cell infection, cell motility, cell secretion, cytoskeletal rearrangement, nuclear translocation, receptor internalization and neurite outgrowth.

Application of primary cells and human cells derived from stem cells

Although recombinant cell lines and immortalized primary cells are commonly used in phenotypic screens to identify lead compounds, largely because they rapidly proliferate and can be expanded for the generation of large quantities of cells needed for HTS, primary human cells and patient derived cells are more desirable for phenotypic screens because of their biological insight and disease relevance. Primary human cells have been used in compound screens that are more biologically relevant for drug discovery [59]. However, limited availability of large amounts of cells and cell types has prevented the broad application of isolated primary cells in lead discovery. Embryonic stem (ES) cells and induced pluripotent stem (iPS) cells are capable of being differentiated into expandable progenitor cells that can be further differentiated to many types of mature cells such as neurons, cardiomyocytes and hepatocytes for drug screens [59,60]. In addition, the capability of generating iPS cells from a patient’s skin, blood or other cells allows establishment of disease models using patient cells that have better pathophysiological relevance to human disease [61,62]. While the methods for stem cell differentiation including the differentiation efficiency, scale-up, reproducibility and cost effectiveness are still being improved, several pilot compound screens using stem cell differentiated progenitor cells have been recently reported. High throughput screens with smaller compound collections have been performed in the precursor cells derived patient or normal stem cells, including neural crest stem cells (from iPSCs with familial dysautonomia, IKBKAP expression) [63], neural progenitor cells (from normal iPS cells, Wnt/β-catenin signaling) [64], neuroepithelial-like stem cells (from normal iPS cells, cell proliferation and viability) [65], and neurons (from ES cells, AMPA glutamate receptor) [66]. Additionally, several other types of human cells derived from stem cells have also been used to assess drug efficacy and evaluate compound toxicity for a small set of compounds [67,68].

Phenotypic screening to identify new indications and new targets of approved drugs

The second application of a phenotypic assay is to identify new indications of known drugs – an application that is particularly useful for diseases without an effective therapy. An approved drug collection has recently been established at our center [69] and has been used to identify lead compounds for new applications in different diseases including Giardiasis [20], NF-κB signaling [70], Niemann Pick disease type C [21], Chronic Lymphocytic Leukemia (CLL) [71], Chordoma [72], adrenocortical cancer [73] and thyroid cancer [74]. Similarly, a smaller collection of approved drugs has become available (http://www.nihclinicalcollection.com/) that has been used to identify lead compounds for the CaV1.3-selective L-type calcium channel and a lithium mimetic project [75,76]. The identification of new applications for approved drugs can save time and resources in drug discovery and development, while reducing the risk of failure in early clinical trials [77,78]. This approach is particularly useful in attempts to identify potential therapies for the vast number of rare diseases, as well as neglected diseases, in which it is imperative to find an effective drug quickly at the lowest possible cost (Figure 2C). Additionally, phenotypic screening of approved drugs may lead to the identification of new drug targets, because of the known pharmacological properties of the drugs on a specific enzyme, receptor or protein. The information obtained from the phenotypic screen can be used for a new drug development program once the new target is validated (Figure 2D).

Concluding remarks

It has been recognized that there is a genuine need for more biologically relevant screening platforms for drug discovery that may lead to the identification of high quality lead compounds. The new phenotypic screening assays should have great potential to meet this challenge as they are usually much more biologically and/or disease relevant. While the screening throughput and disease relevancy of animal models still needs to be improved, the new cell-based phenotypic screens including those utilizing primary cells and stem cell derived human cells have recently emerged for lead discovery in early drug discovery in parallel to the molecular target-based screening approach. The application of using differentiated cells derived from patients for phenotypic screening assays can greatly expand the types and numbers of cell-based disease models. Therefore, phenotypic screening using newly developed cell-based disease models may lead to a new era of lead discovery and contribute to development of personalized medicine.