January 1, 2005 (Vol. 25, No. 1)
Translating Clinically Relevant Genes into Good Quality Drugs
There are numerous steps between the mapping of the genome and the development of a drug for a particular disease. Someone must sift through the data, identify potential target genes, screen and validate those targets, then find compounds which antagonize them.
In the post-genomic era, strategies for doing this work have proliferated both in established pharmaceutical companies and in newer companies specializing in target discovery and validation.
Mining the druggable genome was a popular topic of discussion at scientific conferences throughout 2004, including the annual SBS conference in Orlando, FL, and a London conference hosted by The Royal College of Physicians dedicated to “Leveraging the Druggable Genome.”
The druggable genome can be defined as the set of all human genes that contribute to a disease phenotype and can be modified by a small moleculea drug. As recently as a decade ago, scientists estimated that the human genome contained more than a hundred thousand genes, and the task of screening and validating so many targets was viewed as too large for even the biggest organization.
However, early drafts of the map of the human genome, released in 2001, revealed that it was smaller than anticipated, perhaps only 30,000 genes. Final data published in the October 21, 2004, issue of Nature revises that figure yet again to barely more than 20,000 genes.
When requirements for druggability are applied to this figure, the number of useful genes drops to much more manageable levels. Andrew Hopkins, D.Phil., senior principal scientist at Pfizer Global Research and Development, U.K., estimates that a mere 1214% of genes are able to bind to a druglike molecule.
And, of these genes, Dr. Hopkins estimates, perhaps only 10% have clinical relevance, the rest being unrelated to disease or redundant. The challenge, then, is to translate these hundreds of druggable genes into good, quality drugs that will be safe and effective for human consumption.
Systematic Approach to Target Validation
Lexicon Genetics (The Woodlands, TX) specializes in target discovery through mouse knockout technology. The firm’s approach is encapsulated in a scientific review of the 100 top-selling drugs, authored by Brian P. Zambrowicz, Ph.D., executive vp, research, and Arthur T. Sands, M.D., Ph.D., president and CEO, at Lexicon (Nature 2:38-51, January 2003).
This retrospective study identified only 43 host proteins targeted by the top 100 drugs of 2001, and found that mouse knockout phenotypes have a high correlation with drug efficacy.
Therefore, Lexicon has undertaken to design mouse knockouts of five thousand genes from the human genome, a group of all secreted proteins as well as the conventional families of druggable genes: GPCRs, ion channels, nuclear hormone receptors, proteases, phosphodiesterases, kinases, phosphatases, etc.
Says Dr. Zambrowicz, “There’s a portion of the genes in the genome that are likely to be targets for drug development. You can think of it as prime real estate of the genome. Our approach has been to model knockouts for every one of those druggable genes.”
Over a one-month period, the knockout mice are tested to identify phenotypes that match human disease states. Because the candidate genes have not been previously characterized, no one knows whether a mouse line will falter at, for example, the water maze, or the electric shock memory test, or if it will survive boot camp with no observable phenotype.
Promising genes are then advanced to Lexicon’s drug development program. This approach has already turned up a number of promising drug targets, including LG653, which, when knocked out, produced mice that were healthy but leaner than their counterparts. Lexicon hopes to pursue this target for an anti-obesity drug.
Galapagos Genomics (Leiden, The Netherlands) pursues a similar mission, identifying useful drug targets out of the thousands of genes of the druggable genome. The company’s technology, however, relies on siRNA delivered by adenoviral vectors to primary human cells in vitro. 4,703 RNA transcripts corresponding to 3,700 loci are maintained in Galapagos’ RNA library and roughly half that number in a cDNA library used for “knock-in” experiments (where the gene product is overexpressed in cells).
There are several screening stages in the process which Helmuth Van Es, Ph.D., director of cellular and molecular biology, describes as “biology-based.” Treated cells are assayed for biochemical changes or examined for morphological changes. So far the library has yielded three targets for rheumatoid arthritis and have gone on to drug screening.
Dr. Van Es cites faster throughput and more biologically relevant results as an advantage of working with human cells over knockout mice. Additionally, mouse knockout experiments “will miss genes that are knocked out and give a lethal phenotype early on. We’d pick those up because we’re using cells from an adult human being.”
Multiple Orthogonal Methods
As promising as the mouse knock-out, RNAi, and cDNA knock-in screens are for target validation, there are concerns that basing a company’s entire drug portfolio on a single method is risky. Larry Hardy, Ph.D., associate director of pharmacology at Sepracor (Marlborough, MA), advocates strongly for combining multiple orthogonal approaches in target validation and screening.
“By combining small molecules and genetic approaches, one can get much more confidence in the validity of a target even before it reaches the clinic…trying to use a small molecule alone or a single mutation alone can be very misleading.”
Dr. Hardy advocates matching genetic or epigenetic methods such as knockout genes, antisense, or siRNA with a chemical or protein directed method such as small molecule inhibition, antibodies, or nutritional rescue.
The peril in pursuing a target identified only by a knockout gene or by siRNA is in spending money and time in development that is ultimately wasted on a false target. There is a relatively smaller risk of overlooking an important target.
According to Dr. Hardy, following up a genetic in vitro experiment that demonstrates compound binding to a target with an in vivo experiment demonstrating a physiological effect does not qualify as a multiple orthogonal approach, but rather a serial one that does not prove a causal relationship between the compound, the target, and the disease.
Rather, testing of antagonists in animals would be nicely complemented by in vivo target gene knockouts, and high throughput screening for small molecule antagonists would shore up data collected from cell lines overexpressing the target gene.
Dr. Hardy foresees a coming maturation of the druggable genome, with the majority of proteins identified, and exploitation of parts of the genome not currently at this time in the “druggable genome” such as RNA.
“The real significant advance is knowing what all of the players are in the protein world. The problem is that knowing the sequence of the entire genome does not tell us a priori what RNAs are getting made.
“Nor does it tell us the function of many of the proteins. I suspect that five years from now we’re going to know who all the players are in terms of druggable proteins, but there’s still going to be a lot that remains to be discerned with regard to druggable RNA targets.”
Predix Pharmaceuticals, based in Israel, is an example of a company exploiting the druggable genome for drug discovery purposes primarily, rather than target validation. Predix focuses on G-protein coupled receptors (GPCRs), which comprise about 23% of the genome, and by the company’s estimate 50% of drug targets.
Membrane proteins, because they are present on the surfaces of the cells, are highly accessible to drug compounds, and often act as “switches” for turning on and off signal pathways. Oren Becker, Ph.D., CSO, estimates that there are several hundred GPCRs in the human genome, and that 100200 of them are druggable. They are also readily identifiable by gene sequence because of the characteristic seven-transmembrane-domain structure.
Predix selects the targets from the genome, but does not engage in target validation. Once the targets are validated by other researchers, Predix creates 3-dimensional models of the GPCRs using a proprietary software algorithm, then screens a large library of compounds, in silico.
“We are shortcutting the time to discovery of the drug once we identify the target. First you go through this target validation process. That was one of the main bottlenecks in the nineties. Now that we have the genome, the bottleneck has been unclogged. Now we have targets.
“The part that can be more efficient is the discovery part. It usually takes at least five years. We are shortening the time of the discovery. We are getting to the clinic much faster by using structural info from the genome.”
In two years, Predix has three drugs that are either in the clinic or very close to it, one in Phase II for anxiety, one in Phase I for Alzheimer’s, and one for hypertension which is in the preclinical stage.
The identification of target families within the genome, such as GPCRs, has resulted in a number of organizations that focus on one particular family. Ionix Pharmaceuticals (Cambridge, U.K.) focuses on the treatment of acute and chronic pain. Ionix co-founder John Wood, of the University College in London, provided Nav1.8, a sodium channel expressed in a certain subset of sensory nerves.
Wood validated Nav1.8 as a pain target through the use of null mutant mice. The absence of Nav1.8 was associated with increased pain thresholds and reduced inflammatory pain responses (Proc Natl Acad Sci USA. 101(34):12706-11, August 24, 2004).
Work on Nav1.8 at Ionix led to an additional target, the associated regulatory protein p11. Ionix is currently engaged in lead discovery and optimization programs using Nav1.8, p11, and a voltage-gated calcium channel, Cav2.2.
This shows that not only can the genome be mined at random and discovery programs be based around the disease states that happen to arise, but that certain families of genes can be identified as being associated with particular disease states. Those families can be selected as targets and validated.
According to Dr. Hopkins, the drug industry produces four first-in-class drugs per year, and thus estimates that it will take another two generations to fully mine 300 druggable genes from the genome (a conservative estimate). However, this model of drug discovery is based on certain assumptions, chief among them the idea of one drug antagonist for one disease target.
It is likely that the future of drug discovery will include not only the systematic, one-by-one study of target proteins, but also parallel strategies such as exploring synergistic effects of two or more targets together, epigenetics and RNA targets, and biologically based research into the mechanisms of disease.