Organizers: Michael Foley (Tri-Institutional Therapeutics Discovery Institute), Ralph Garippa (Memorial Sloan-Kettering Cancer Center), David Mark (F. Hoffmann-La Roche), Lorenz Mayr (Astra Zeneca), John Moffat (Genentech), Marco Prunotto (F. Hoffmann-La Roche), and Sonya Dougal (The New York Academy of Sciences)Presented by the Biochemical Pharmacology Discussion Group
Reported by Robert Frawley | Posted January 12, 2016
There are two major methods for designing pharmaceutical drugs. In traditional drug discovery (TDD), or empiric design, researchers target a particular domain or protein after working to understand its mechanisms and molecular biology. In phenotypic drug discovery (PDD), many different compounds are tested on a system until one results in an observable phenotype of success, and the compounds' mechanisms of action are not considered. The Phenotypic and Biomarker-based Drug Discovery symposium, presented by the Academy's Biochemical Pharmacology Discussion Group on October 27, 2015, featured current work in PDD and highlighted the need to bridge commercial and academic research to improve phenotypic drug design.
Phenotypic drug discovery—screening of thousands of substances for functional cellular outputs such as gene expression, growth arrest, and cancer cell death—has led to the development of more commercial drugs than TDD, the more common method of discovery. Indeed, as Jonathan A. Lee of Eli Lilly noted, spending on TDD is out of sync with the rate of new drugs reaching approval; the number of new drugs per billion dollars spent dropped sharply in the last few decades. He argued that the need for functionally validated drugs could be met through a renewed focus on PDD.
Spending has increased while drug discovery has flattened, leading to historic reductions in new molecular entities (NMEs) per billion dollars spent. Noted are the introduction of expressed sequence tags (ESTs) and the mapping of the human genome sequence, which should have aided targeted drug discovery. (Image presented by Jonathan A. Lee)
Bruce A. Posner started the morning session with a discussion of a phenotypic screen conducted at the University of Texas Southwestern Medical Center which identified two chemical scaffolds that are effective in killing non-small cell lung cancer (NSCLC) cells but are harmless to the non-cancer cells tested. In further studies, the group showed that an optimized analog of one scaffold arrested tumor growth in a mouse xenograft model of NSCLC. Both chemical scaffolds appear to work through a novel mechanism targeting stearoyl-CoA desaturase (SCD), which is known to be important in unsaturated fatty acid synthesis. These compounds were found to be specific, effective, and potent in NSCLC cell lines that express elevated levels of Cyp4F11 and/or related Cyp family members. The group also showed that these scaffolds function as prodrugs that are activated only in cancer cells expressing these Cyp isoforms and that the Cyps produce metabolites of the prodrug that bring about cancer-specific cell toxicity. The group is working to improve these scaffolds and to develop a putative biomarker based on Cyp expression.
The Broad Institute's LINCS (Library of Network-based Cellular Signatures) database is designed to keep track of small-molecule therapeutics, collecting data on cellular responses to "perturbagens" (drugs, factors, and others stimuli). Data are generated using the L1000 assay, which assesses the expression of 1000 genes known to explain 80% of genetic variation in assayed cell lines. Aravind Subramanian explained that the technique can identify the majority of drug effects for a fraction of the cost of RNA sequencing. Although it examines only a subset of molecules and relies on measuring genetic responses, the technique can help predict the likelihood that new compounds will elicit desired effects.
Martin Main of AstraZeneca described phenotypic drug discovery at AstraZeneca. The company's model for discovery is to check phenotypic markers at every step, as drugs are moved from cell lines to patients. Main's team identified a molecule that enhances the regenerative function of cardiac myocytes after infarction. Using cells from several donors, the team validated a promising compound that increases proliferation of cardiac myocytes and drives epicardium-derived progenitor cells to assume a myocyte lineage. In another discovery, the team used islet β-cell regeneration as the phenotype, discovering a compound the researchers believe will reach clinical trials for type 2 diabetes.
Andras J. Bauer of Boehringer Ingelheim discussed a method to increase predictive strength in compound selection before phenotypic screening. By cataloging the structures of known target–reference compound binding pairs, the team can compare those structures to untested compounds, and then assess only the most promising compounds. The THICK (Target Hypothesis Information from Curated Knowledge bases) database gives interaction-probability scores to untested compounds on the basis of structure. Bauer also described a method to verify target–compound interaction without labeling the molecules, in which phenotypic results were verified with mass spectrometry.
In the afternoon session, Myles Fennell of Memorial Sloan-Kettering Cancer Center described his work testing small interfering RNA (siRNA) libraries to find siRNAs that alter macropinocytosis (MP), cell-surface ruffling that is seen in prostate cancer cells. The surface phenotype allows TMR-dextran uptake, which the researchers measured in the screen. MP is driven by RAS (a commonly affected gene family in cancers) and the pathways are already popular drug targets. The researchers tested two libraries of siRNAs, which block translation of specific proteins, using TMR as a marker to report MP severity, as well as sensitive single-cell assays to determine siRNA efficacy. The team identified promising target sequences and used a data-analysis pipeline called KNIME to define several hits, which the researchers are pursuing in therapeutic development.
TMR-dextran is able to work into cells undergoing macropinocytosis and thus these cells can be separated by phenotype as seen in the controls above. (Image courtesy of Myles Fennell)
Giulio Superti-Furga of the Austrian Academy of Sciences is a proponent of understanding the mechanisms of action (MOA) of candidate drugs. He began by explaining that the genome is an incomplete indicator of disease; epigenetics, altered protein function, metabolism, and other factors are also important. He then introduced pharmacoscopy and the "thermal shiftome" as methods to phenotypically screen compounds. Pharmacoscopy uses high-power automated microscopy to describe how compounds affect cell populations by using specific stains for different cell types; a computer then counts the cells expressing each stain, yielding results similar to those obtained via fluorescence-activated cell sorting but generated through an automated process. The thermal shiftome catalogs changes in thermal stability after protein binding in known reactions and is used to characterize the stability of new reactions. Superti-Furga offered a perspective that tempered the enthusiasm for pure PDD and advocated a mechanistic approach to drug discovery.
Michael R. Jackson, at one of the largest academic screening facilities, the Sanford Burnham Prebys Medical Discovery Institute, led a reexamination of drug screens performed by pharmaceutical companies. His team conducted millions of assays and accumulated a large data library with few new hits. However, the researchers were able to closely characterize the chemistry of one hit, an undisclosed interaction, and Jackson's group is proceeding to develop a drug to modulate nuclear receptor signaling. The researchers also have a procedure that can screen for the differentiation of human induced pluripotent stem cells (iPSCs) into neurons for potential neuro-regenerative therapies. They developed high-throughput morphology, endpoint-measurement, and proliferation assays that generate tightly clustered, repeatable data. The team has produced consistent results screening 10 immune modulators and various cytokines to assess the reactivity and stability of the cells, providing reliable compound characterization. This success in human cells shows that a disease-relevant patient-derived screening platform to characterize differentiation and immune response is possible with robust assays.
In the next set of talks, Friedrich Metzger and Susanne Swalley described the parallel work of Hoffmann-La Roche and Novartis, respectively, toward treating spinal muscular atrophy (SMA). A devastating disease that leads to loss of motor function and affects motor nerve cells in the spinal cord, SMA presents a unique drug development opportunity. The condition is caused by the loss of function of a single gene product called survival of motor neuron (SMN1). Humans encode an unstable gene product, called SMN2, which is nearly homologous to SMN1.
Metzger explained that the inactive SMN2 variant is largely the same as active SMN1 but, missing exon 7, cannot compensate in its absence. The group from Hoffmann-La Roche aimed to stabilize SMN2 by promoting the inclusion of exon 7. The researchers conducted a phenotypic screen seeking a compound that could change the splicing in patient fibroblasts in vitro and produce a stable, functional SMN2 protein including exon 7. In studies with an SMN2Δ7 mouse model (lacking exon 7), mice drugged with the compound experienced full phenotypic rescue. The compound has been shown to induce alternative splicing of SMN2 to include exon 7 in healthy human volunteers; it was well tolerated and is moving to human patient trials.
Swalley discussed the target identification and MOA of the Novartis compound. After a screening process similar to Roche's, Novartis moved its compound into animal models while also beginning parallel experimentation to find out why it worked. The group found that U1-snRNP, a spliceosome component required for the splicing process, is bound at two essential nucleotides by the compound. In the SMN2Δ7 mice, the compound improved survival and rescued full SMN2 protein expression. The Novartis compound stabilizes the appropriate spliceosome components to produce SMN2 with exon 7 intact. This novel mechanism demonstrates that a sequence-selective small molecule therapy can alter splicing activity to treat SMA. Together these talks demonstrated the power of PDD and the importance of validating drug mechanisms.
The final talk of the day was given by Hoffmann-La Roche's Jitao David Zhang, who suggested that pathway reporter genes, which are only modulated when a specific signaling pathway is activated or inhibited, can be used as phenotypic readouts. It is known that gene expression data can predict cell phenotype. Using transcriptomics as a surrogate for downstream phenotypes, for example by using expression data from a gene subset to predict outcomes, would save time and effort. In an iPSC cardiomyocyte model of diabetic stress, machine learning (guided by pathway information) characterizes the response of the iPSCs to a library of compounds, highlighting compounds and pathways worthy of further investigation. This new platform for molecular phenotyping using pathway reporter genes, sequencing, and early analysis speeds compound characterization.
Use the tabs above to find multimedia from this event.
Presentations available from:
Andras J. Bauer, PhD, PharmD (Boehringer Ingelheim)
Myles Fennell, PhD (Memorial Sloan-Kettering Cancer Center)
Jonathan A. Lee, PhD (Eli Lilly)
Martin Main, PhD (AstraZeneca)
Yao Shen, PhD (Columbia University)
Susanne Swalley, PhD (Novartis Institutes for BioMedical Research)
Jitao David Zhang, PhD (F. Hoffmann-La Roche)
American Chemical Society
How to cite this eBriefing
The New York Academy of Sciences. Phenotypic and Biomarker-based Drug Discovery. Academy eBriefings. 2016. Available at: www.nyas.org/PhenotypicDrug-eBLog in or Join Now to continue