February 1, 2016 (Vol. 36, No. 3)
DeeAnn Visk Ph.D. Founder and Principal Writer DeeAnn Visk Consulting
After Having Been Eclipsed by Target-Based Screening, Phenotypic Screening Is Finally Shining Again
In drug development, target-based screening embodies a sensible ready–aim–fire approach. The alternative approach, phenotypic screening, is a little different.
It’s a little like the game of Battleship®, wherein players blindly call out shot coordinates and then wait to hear if they’ve scored any hits. Whenever players learn of a hit, they can guide subsequent shots accordingly, firing until a pattern of hits finally reveals the target’s identity.
The problem with this alternate ready–fire–aim approach is that in drug discovery, target validation has been a anything but a game. Nevertheless, there has been progress in identifying the targets of phenotypic hits. Promising advances include chemical proteomics and computational analysis of biological pathways. These advances, combined with complex and biologically relevant cell-based screens, make it possible for drug developers to chase down disease-modifying effects too subtle or too elusive to be captured by reductionist approaches.
“Incorporation of cell-based screens, engineered tissues, 3D models, and whole-small-animal screens can more accurately reflect disease phenotypes,” says D. Lansing Taylor, Ph.D., director of the University of Pittsburgh Drug Discovery Institute. “Integrating these approaches with experimental models and computational models—as part of a holistic approach to understand the mechanisms of disease progression—proves to be extremely useful.”
When developers discover a desired effect in a phenotypic screen they can go back to a computational model of the pathways involved in disease progression to mathematically describe the disease process. The developers can then return to phenotypic experiments to test the new predictions made with the computational model.
“This iterative and integrated approach is referred to as quantitative systems pharmacology (QSP),” notes Dr. Taylor. “The origins of QSP go back to about 2010, when the National Institutes of Health sponsored a workshop to integrate the distinct fields of systems biology and pharmacology.”
Dr. Taylor indicates that his group is hopeful that applications of QSP can reinvent the field. In Dr. Taylor’s group, early applications include programs in metastatic breast cancer, hepatocarcinoma, and Huntington’s disease.
“One of the driving forces for implementing QSP is the brutally expensive road of traditional drug discovery,” explains Dr. Taylor. “I returned to academia because it offers the freedom to pursue innovations that are too fiscally risky for the pharmaceutical industry.”
Up Close and Personal—and Holistic
“At the beginning of a program, patient samples are analyzed by a variety of genomic, transcriptomic, and metabolomic methods, and the data that accrues is used to infer pathways of disease progression using computational biology,” explains Dr. Taylor. “For example, we compare primary cancer tumors to metastatic tumors, asking which pathways are involved in the disease progression. This marriage of patient samples and computational biology software tools is a powerful paradigm for understanding the mechanisms of disease progression.”
“The pathways, which are computationally inferred, contain potential molecular targets for the discovery process,” Dr. Taylor continues. “Next, the computational model predicts which known drugs would interact with the ‘list’ of potential targets from the inferred pathways. From there, we return to the phenotypic screen to determine which drugs show activity in our system. The determination of the ‘reversal’ of the disease phenotype activity is used to refine the model, as well as developing potential leads for drug candidates.”
Dr. Taylor acknowledges that this holistic approach is time-consuming. He adds, however, that in terms of revealing how a drug works, the approach has its own economy. That is, it works without requiring the drug to become the focus of a full-scale discovery effort.
Dr. Taylor believes this heavily front-loaded strategy will allow the selection of better molecular targets and drug candidates, lowering the attrition rate of drug candidates in later stages of the drug discovery and development pipeline. Once this innovative approach is matured and applied to commercial drug development programs, Dr. Taylor suggests, it will optimize the process of drug discovery.
The Right Combination
“We take the idea of using drug combinations to treat a disease and apply it to the phenotypic screen. In our particular case, we have successfully employed this approach to multidrug-resistant bacteria and the Ebola virus,” expounds Wei Zheng, Ph.D., a group leader at the National Center for Advancing Translational Sciences, part of the National Institutes of Health (NIH).
“The main goal of our lab is to translate academic research into applications that can be developed further by industry,” states Dr. Zheng. “We want to benefit patients by working with their doctors and academic scientists.
“With the outbreak of the Ebola virus last year, we were inundated with requests to rapidly screen compounds from many academic organizations. Thankfully, we were able to rapidly evaluate all the compounds in collaboration with Adolfo Garcia-Sastre, Ph.D., professor of medicine and microbiology at the Icahn School of Medicine at Mount Sinai. We recently published a paper detailing which compounds prevent Ebola entry into the cell.
“We employed a phenotypic screen to determine which compounds prevented Ebola entry into the cell,” Dr. Zheng continues. “Next, we evaluated them in two to three drug combinations to identify synergistic effects to block Ebola virus entry. A single drug may require such a high dose to be effective that patients cannot tolerate it. Combing two to three medications with different mechanisms of action at low doses can better block the Ebola virus entry. These lower doses are better tolerated by the patient.”
This cocktail strategy has been employed with another virus: the human immunodeficiency virus (HIV). By combining drugs with different mechanisms of action, the treatment regimen has better efficacy then either drug alone; hence, the synergistic effect.
In the search for new medications for multidrug-resistant bacteria like Klebsiella pneumoniae, Dr. Zheng’s laboratory screened a Food and Drug Administration–approved library of drugs in collaboration with the NIH-based researchers Karen Flank, M.D., Ph.D., and Peter Williamson, M.D., Ph.D. “Of these 2,800 medicines, some 20 to 30 have shown activity against multidrug resistant bacteria,” states Dr. Zheng. “Although most of them are antibiotics or anticancer drugs, none of these drugs had previously been shown to have an effect on Klebsiella pneumoniae.”
In a case still to be reported in the literature, Dr. Zheng worked directly with doctors of a patient infected with drug-resistant bacteria, both Gram-negative and Gram-positive. This very sick individual was admitted to the NIH clinical center’s intensive care unit and was not expected to live.
“In collaboration with the patient’s doctors, we isolated the bacteria that were infecting the patient,” Dr. Zheng reports. “Then, in real time, we did a screen to determine which antibiotics killed the patient’s bacteria.
“The patient responded to the antibiotic cocktail suggested by the screen. Later, the patient experienced liver toxicity in response to the drugs. We reviewed the data from the screen and recommended a change that worked. In the future, it would be wonderful to see this approach used to find treatment for all patients with life-threatening, drug-resistant bacterial infections.”
Thinking Inside the Black Box
When planning drug discovery and development, investigators should remember basic chemical biology. “Key questions should be posed about the characteristics of protein targets and small drug molecules,” states Erik Hett, Ph.D., a senior scientist at Biogen.
According to Dr. Hett, protein targets pose the following questions: Where does your target localize? Does it have multiple isoforms? And small drug molecules pose questions of their own: Where does the molecule bind? Where does it distribute? What effect does it have in binding to the target? How much occupancy is required to lead to the desired phenotype?
“Biogen is expanding its phenotypic screening facility. We hope to have some exciting results to reveal in the years to come,” continues Dr. Hett. “Phenotypic screening is a powerful way to get at complex biology. A small molecule can advance to clinical trials even if its mechanism of action is unknown—even if the underlying biological function of the drug is unclear.”
Nonetheless, knowledge of a drug target is valuable to medicinal chemists. This information can help them optimize the drug’s target-binding ability. Also, if a drug target is known, it is possible to determine different isoforms.
“Isoforms are different proteins made from the same gene, but with splice variations that can lead to long or short forms,” explains Dr. Hett. “Temporal and spatial expression of these different isoforms can also be explored. If your target is an isoform found only in the brain and you are going after a neuroactive compound, you don’t want to find out that the target of your drug candidate is a skeletal muscle–only isoform.”
According to Dr. Hett, phenotypic screening is a very “black box” phenomenon. Current practice is almost always done in human cells—either immortalized cell lines or primary cultures. This ensures that the target of the drug candidates are properly expressed, folded, and post-translationally modified. This includes processes such as glycosylation.
“In the past,” Dr. Hett points out, “problems have been encountered when going from a human cell culture system to a whole animal model in another species. Some of these difficulties can be traced to different isoforms between species.”
Dr. Hett also emphasizes the importance of localization. Specifically, the protein target’s localization should be understood not only at the tissue level, but also at the cellular level. For example, if a protein target is located in the mitochondria, then that particular organelle can be isolated. In principle, screens on different kinds of cell components may be countenanced.
“Remember the ‘Rule of 3’ for phenotypic screens,” urges Dr. Hett. This rule refers to the right cells, the right stimulus, and the right readout. The better the disease is understood, Dr. Hett adds, the better the model system that can be constructed.
“Occupancy of the target by the candidate is another topic to consider in drug development, as well as the kinetics of the drug binding to the target,” Dr. Hett continues. Characterizing occupancy and kinetics, he says, allows researchers to have confidence that the mechanism was adequately tested in the clinic. That is, researchers can consider whether the mechanism failed, or whether they need to return to the same mechanism with a different approach.
Another consideration raised by Dr. Hett is the half-life of the protein being studied. Long-lived proteins, he notes, give developers the opportunity to consider a covalent or very low off-rate drug candidate modality.
“If the target turns over in an hour, then compound may need to be on board much of the time,” indicates Dr. Hett. “Small molecules binding to a target can change the half-life of the protein.” Heat shock protein 90 is an example of this because it stabilizes many kinases, lowering the turnover of the bound kinase, and this interaction can be intentionally disrupted by a small molecule.
“Phenotypic drug discovery can improve the probability of translation of preclinical findings to patients,” concludes Dr. Hett. “So, ensure the screen has been optimized for the disease-relevant system, the stimulus, and the endpoint.”
Early-Stage Discovery Gets Real
“Over the last decade, phenotypic assays have become more popular. Gradually the focus has shifted from using easy-to-interrogate artificial cell systems, such as immortalized cell lines manipulated by insertion of reporter genes or protein overexpression, to using more biologically relevant assay systems,” states Susanne Heynen-Genel, Ph.D., director of High Content Screening Systems at the Sanford Burnham Prebys (SBP) Medical Discovery Institute. “From 2006 to 2010, about 40% of the screens my group completed looked at proteins or cellular function in the native (endogenous) context of the cell. From 2011 to 2015, the screens that qualified as phenotypic increased to 70%.”
“Essentially, we are moving toward more biologically relevant systems and away from artificial systems,” elaborates Dr. Heynen-Genel. “Reductionist screens often do not translate into viable drug candidates. Phenotypic screens can better represent the functionality found in nature, especially when cells are not perturbed from their normal activities.”
One trend Dr. Heynen-Genel sees is the move toward using human cells in phenotypic screens whenever possible. She points out that the cells employed her group’s screens are increasingly from patient-derived primary cultures or freshly harvested from patient-derived tumor xenografts. “In the recent past, from 2006 to 2010, about 50% of our screens were done with human cell lines while the remaining 50% employed mammalian cells,” she details. “Currently, cells of human origin are used almost exclusively in phenotypic screens at SBP.”
The human cells utilized in screens are increasingly from patient primary cultures or derived from induced pluripotent stem (iPS) cells instead of the traditionally used immortalized cell lines. Getting tissue from the brain or heart of living patients is not feasible; hence, deriving those cells from iPS cells allows tissue-specific testing. Getting iPS cells in large numbers is challenging and expensive, so the screens with iPS cells tend to be smaller—from 5,000 to 10,000 compounds.
“Goals for a particular screen are also changing,” Dr. Heynen-Genel explains. “For example, rather than target all brain cancers, investigators concentrate on a particular subgroup within a cancer. The objective of the studies is to stratify patients into groups, with specific treatments for each category.”
Another trend in phenotypic screens is the utilization of label-free systems for phenotypic screening. These systems, remarks Dr. Heynen-Genel, bypass the need for fluorescently labeled proteins or antibodies. Use of label-free assay technologies allows the cells to remain in an unperturbed, biologically relevant state.
“One novel approach we have taken can be thought of as fishing for a target,” says Dr. Heynen-Genel. “We study proteins’ subcellular localization during viral infection. This provides us with an unbiased look at a biological process. The hope is that we will find novel targets for small drug molecules.”
“With increased accessibility to technology and ease of use, anyone can generate numbers from an imaged assay plate. Nonetheless, how the numbers are derived must be carefully examined,” warns Dr. Heynen-Genel. “Attention must be paid to ensure the results accurately reflect the biology.”
The final trend Dr. Heynen-Genel cites is that academic research institutes are working more closely with the pharmaceutical and biotech industry. No longer do academics merely publish the results of a potential new drug target and leave them there in the hopes that a company will pick them up for further development. Now academic researchers, with the help of academic drug discovery centers, move on to develop and validate assay screens.
Toward Targets Known and Unknown
“[We have] experience in phenotypic screening and finding compounds that induce disease-related changes in biologically relevant cells. This is an alternative to target-based screens, which rely on hypotheses of the roles particular targets may play in disease,” maintains Ulrich Schopfer, Ph.D., executive director and head of Integrated Lead Discovery at the Novartis Institutes for BioMedical Research, the research arm of the pharmaceutical giant Novartis.
“When thinking about a screening strategy, Novartis looks at what each approach can bring to the problem, rather than favoring one technique over another,” explains Dr. Schopfer. “This allows us to integrate information from all approaches, knowing that some will bring one strength to the table while another will bring a different strength.
“This philosophy allows Novartis to continue on in the drug discovery path, even if the target is not definitively known. The phenotype must be predictive of a disease state, of course. However, bringing the project to a halt, until the target can be nailed down, is not necessary. Novartis views target identification as something that can be done in parallel with compound optimization and in vivo testing.
“When determining which disease to next study, Novartis evaluates two key areas: the understanding of the biology of the disease and the medical need for remedies for the disease. An example of this can be found in spinal muscle atrophy (SMA), which is a leading cause of congenital death in infants. The lack of a protein, SMN2 (survival of spinal motor neuron 2) was known to contribute to the condition.”
“When Novartis began its screen of 1.4 million compounds, we were only looking for candidates that promoted expression of the missing protein,” clarifies Dr. Schopfer. “We did not know what the target of the compound would be, only that it restored the presence of the missing protein.”
As the project progressed, they determined that the compounds increased SMN2 protein levels by stabilizing a splicing complex. By then the project had progressed to an in vivo model even though the precise details of the target remained unknown.
Novartis’ approach titrates the complexity of the assays to the right level. A natural tension exists between complexity (relevant biology) and simplicity (for large scale and easier data interpretation). Dr. Schopfer advocates using insights into disease biology to guide the complexity required in modeling the disease in a screen. This can sometimes lead to a surprisingly simple assay rather than extremely complex one, which lends itself well to phenotypic screening.
“This flexibility of integrated lead discovery gives Novartis freedom from a standardized approach to drug discovery,” asserts Dr. Schopfer. “In essence, projects can begin anywhere and go anywhere. For example, a project could start with a phenotypic screen using primary human cells. Finding a compound with a known target can help validate the assays. Then other hits with no target-hypothesis can also be explored with more confidence.”