Introduction

Protein kinases are the workhorses of cellular regulation, playing a key role in almost every major pathway in eukaryotic cells, including those that control cell division, cell death, cell growth, and programs of differentiation. These proteins play key regulatory roles in plants and bacteria, including many pathogenic microorganisms. Mutations or overexpression of these proteins is implicated in a wide range of diseases, from cancer to diabetes to neurodegenerative diseases. Small wonder then that they are objects of intense interest to drug developers.

Protein kinases also form one of the largest protein families, and hundreds of examples have been identified, cloned, sequenced, and subjected to X-ray crystallography and other methods of structure determination. Much is known about the core elements of the catalytic subunits of these multidomain proteins, which phosphorylate protein targets using ATP as a phosphate donor. Researchers have identified key active site residues, as well as sites of allosteric regulation. In the process, much of the low hanging fruit has been picked in the quest to discover protein kinase inhibiting drugs that are selective, potent, and possess low toxicity.

On April 28, 2009, four investigators involved in this work presented their latest findings on kinase structure and function and on current methods of kinase inhibitor drug development in a symposium organized by Stephen Burley of Eli Lilly and Co. and George Zavoico of Westport Capital Markets, LLC.

Susan Taylor of the University of California, San Diego, described her group's extensive work on cAMP-dependent protein kinase A (PKA), a prototypical protein kinase from whose study much of our knowledge about protein kinases is derived. She and her group have been studying the macromolecular structure of the PKA holoenzyme and characterizing its interactions with regulatory and targeting proteins. These studies have revealed a number of potential new targets for kinase inhibitor drug development that may allow the design of more selective drugs, or drugs for kinases that have thus far been targeted unsuccessfully.

Vincent Stoll of Abbott Laboratories provided a wealth of insights into protein kinase drug development, tackling topics such as toxicity, selectivity, efficacy, and patentability. Although over 100 kinase inhibitors are currently in development, and many more have been studied, only nine candidates have thus far reached the market as drugs. His analysis of the pitfalls and promises of kinase inhibitor drug development shed considerable light on some of the reasons why kinase-targeted drug discovery has not been more productive, despite considerable effort by researchers around the world.

Harren Jhoti of Astex Therapeutics and Stephen Burley of SGX Pharmaceuticals, now a division of Eli Lilly, described research programs in their companies intended to develop kinase inhibitors using a relatively new method known as fragment-based drug design. This method, which relies heavily on structural information, uses small chemical fragments identified by screening as starting points to build larger, drug-like compounds with favorable physicochemical and clinical properties. The method seeks to avoid some of the downfalls of high-throughput screening-based methods, which tend to result in drug candidates that are too large and too hydrophobic to be successful as marketable therapeutics. Both research programs have developed promising kinase-based drug candidates that are in early clinical development, primarily for cancer applications.

Protein kinase inhibitors such as the blockbuster Gleevec are among the first successes of the era of targeted drug development, in which drugs are developed based on a detailed understanding of the underlying molecular processes of disease. Although additional progress in producing such drugs has been slow in coming, these researchers make it clear that kinase-targeted drugs still have significant potential to treat intractable clinical syndromes.

Structure-based Drug Design on Kinase Targets: Key Lessons Learned

Speaker:
Vincent Stoll, Abbott Laboratories

Highlights

• Drug researchers working on protein kinases frequently encounter problems related to issues of toxicity, efficacy, selectivity, and patentability.
• Selectivity for a specific kinase or group of kinases is particularly important, but desired selectivity may be difficult to discern and achieved selectivity may be difficult to measure.
• The wealth of available structural data on protein kinases can greatly assist drug development but must be used cautiously due to the high degree of flexibility in these structures.

Kinases and toxicity

Vincent Stoll of Abbott Laboratories followed Taylor's highly theoretical discussion of future kinase targets with a very practical discussion of protein kinases as drug targets in the here and now. Kinases are highly popular as drug targets, second only to the G protein-coupled receptors. There are over 100 kinase inhibiting compounds currently in phase 1 through phase 3 clinical trials. There are nine approved therapies, only one of which, Gleevec (imatinib) has achieved blockbuster status. Gleevec, which is approved for the treatment of chronic myeloid leukemia, had sales of over $3 billion in 2007. Although the emphasis has largely been on oncology, kinase inhibitors are under consideration in a number of other clinical areas, including pain, Alzheimer's disease, eye disease, and immunological applications.

Because kinases are involved in multiple signaling pathways, predicting effects can be difficult.

Stoll offered a primer on developing kinase inhibitors based on his extensive research experience in this area. He identified and explored a number of common issues that are faced by drug discovery teams who are pursuing protein kinase targets, including issues of toxicity, efficacy, selectivity, and patentability, among others.

Toxicity is often a primary concern when developing kinase inhibitors. It tends to arise because kinases are involved in multiple, complex, and often important signaling pathways. This creates a need to design inhibitors that are as selective as possible in order to avoid unwanted effects on other kinases. Kinase inhibitors also usually require heterocyclic constituents, which increases the level of difficulty since such molecules are highly likely to bind to other, non-kinase targets as well.

Improving efficacy

Once a nontoxic inhibitor has been identified, difficulties with efficacy may arise. Sometimes the target is invalidated; that is, inhibition of the targeted kinase turns out to have no effect on the condition it was intended to treat. Compounds may be too selective; some recently successful drugs—for example, sorafenib, which is used for liver cancer—actually inhibit multiple kinases. Sometimes the targeted pathway has a backup system in place that is activated when the targeted kinase is inhibited, precluding efficacy.

Because many inhibitors compete with ATP for the kinase binding site, the apparent potency of molecules under investigation may vary depending on the ATP concentration present in the assays and the relative affinity of the target kinase for ATP. This and other reaction conditions for screening must be chosen carefully to maximize the possibility of detecting activity and to increase the likelihood that the inhibitor will work under physiological conditions inside the cell.

If activity is detected, it is important to characterize the mode of inhibition. Inhibitors may be competitive, allosteric, or irreversible. There are a number of possible artifacts that can lead to spurious kinase inhibition, including changes in solubility properties, the presence of a reactive contaminant, or the presence of a contaminating kinase in the preparation that is used to screen for inhibitors. These false positives are often identified when a putative inhibitor fails to inhibit the kinase in subsequent assays using different conditions.

Stoll indicated that it is important to make sure that the intended inhibitor has drug-like properties; in other words, the inhibitor should be a relatively small molecule that is not too highly hydrophobic. Kinase inhibitors that are identified by in vitro screening assays tend to be flat, aromatic molecules with poor solubility and other unfavorable physicochemical characteristics. Such molecules must be optimized for improved oral bioavailability before further development.

Because kinase inhibition is such an active area of research it can be difficult to find new inhibitors that are truly novel and thus patentable and worthy of further development. One way to deal with this problem is to use a fragment-based approach, in which a novel scaffold molecule with good size and solubility properties is elaborated on by the addition of functional groups to create a novel inhibitor. As an example, Stoll noted that using a rarer starting molecule such as thienopyrazole, which is represented by only 68 structural derivatives in the chemical database, is more likely to result in a unique molecule than indazole, which is represented by over 22,000 derivatives.

Researchers can take advantage of structural information to model fragment binding to the kinase before proceeding, and may also be able to translate a successful scaffold to multiple kinases in different therapeutic areas. However, synthesizing a novel scaffold can be time consuming and labor intensive. In addition, targeting structural regions that are well understood may result in inhibitors that are not as selective as needed, because these are often regions that are highly conserved among the kinases.

The problem of selectivity

Once a novel, nontoxic, effective inhibitor with good drug-like properties has been identified, researchers must face questions of selectivity. First, it is important to know what the desired selectivity profile is; i.e., what kinases need to be inhibited to provide therapeutic efficacy. Second, it is not always a straightforward process to determine what the selectivity profile is for a newly discovered inhibitor, nor is it always easy to determine which parts of the inhibitor molecule are providing selectivity.

Inhibitors may have different binding modes on different kinases, making generalization difficult. However, Stoll pointed out that selectivity is a highly important feature for kinase inhibitors. He noted that the research compound staurosporine, which inhibits 104 out of 113 kinases, has annual sales in the thousands of dollars, while Gleevec, which inhibits only 5 of 113 kinases, has annual sales in the billions.

Hundreds of kinase structures are available in protein structure databases, but the literature is often not reproducible.

Stoll provided some insight into the role of structural information when investigating questions of selectivity. Since kinase structures are so dynamic, it can be difficult to predict whether a given compound will inhibit a given kinase based on a static crystal structure. This necessitates empirical studies. However, once a compound has been identified with the desired selectivity profile, the large amount of kinase structural data often allows rapid elucidation of the molecular basis for selectivity, and this knowledge can be exploited in further drug development.

Kinase structures can be difficult to work with due to the large number of states they can assume, depending on the presence or absence of ATP or other ligands, post-translational modifications, and differences in crystallization conditions. Stoll commented on the large number of kinase structures that are available in protein structure databases. These include over 800 structures for around 80 kinase targets. In addition, Abbott has an internal library of more than 500 structures for 20 kinase targets. He noted that the literature is not very reproducible in this area. Moreover, some kinases have proven very difficult to crystallize, necessitating the choice of a surrogate kinase for modeling purposes. The choice of surrogate is made more difficult because sequence similarity is not a good indicator of structural similarity in these highly flexible proteins.

Stoll's analysis of the promises and pitfalls of kinase drug discovery illuminated some of the reasons why only a few kinase targeted drugs have reached the marketplace, despite intense interest and the expenditure of many person-years of research effort.

SGX523 Is an Exquisitely Selective ATP-Competitive Inhibitor of the MET Receptor Tyrosine Kinase

Speaker:
Stephen Burley, Eli Lilly and Co.

Highlights

• Fragment-based drug discovery can be facilitated by using a fragment library whose members possess chemical "handles" that allow rapid further modification.
• Attention to the lipophilicity measure clogP can further assist in the development of drug candidates that more closely resemble successfully marketed drugs.
• This method has been used to develop inhibitors of the MET tyrosine kinase, a target in familial and sporadic cancers.

Getting a handle on drug discovery

Stephen Burley described another program of fragment-based drug design for protein kinases, this one performed at SGX Pharmaceuticals, which is now a part of Eli Lilly. Researchers at SGX cloned over 200 protein kinase catalytic domains and solved 60 crystal structures in the course of this work, then applied fragment-based methods to discover kinase inhibitors.

Like Jhoti, Burley explained that maintaining ligand efficiency is a guiding principle of this research program. But Burley and his colleagues also chose to incorporate a lipophilicity parameter into their work, in the hopes of further increasing the chances that their compounds would become effective oral agents. They combined calculated ligand efficiency with a biochemical parameter known as clogP, a measure of lipophilicity. Most currently approved drugs have a clogP value of 5–6. A lower clogP means less promiscuous binding and thus less chance of toxicity due to off-target effects in drug candidates.

In fragment-based drug design, multiple targets can be pursued at one time, using multiple potential binding sites.

Burley and his group developed a library of 1500 fragments, each designed to act as scaffold for further development. Each fragment included two or three chemical "handles" that would serve as positions for the addition of new functional groups. Overall, the molecules in this library included about 26 different handles that could support over 50 different chemistries, allowing the rapid generation of considerable diversity in the drug design process. The initial fragments were soaked into protein crystals, and X-ray crystallography, surface plasmon resonance, and high concentration biochemical screening were used to investigate the binding properties of each fragment. Then, rather than proceeding combinatorally, the SGX team explored the effects of adding new chemical groups using one handle at a time.

Burley described three general lessons that he and his group learned from their work with this library. The first was that hits tend to resemble the overall composition of the library that they come from. Thus, if specific properties are desired, those should be the dominant properties in the library that is screened. Second, they found that most kinase hits were already fairly selective, with 73% selectivity for a single kinase. Third, they found that hits with initial affinities in the millimolar range were readily optimizable. They achieved up to 550-fold increases in potency with some hits. For this reason, low initial potency should not deter researchers from continuing to pursue a fragment that binds the target of interest.

Burley described the advantage of fragment-based drug design as one of multiplicity: multiple targets can be pursued at one time, using multiple potential binding sites. In addition, multiple chemotypes are detected as hits, rather than the very similar molecules that tend to be detected again and again using traditional high-throughput libraries.

Applications with RTK

Researchers at SGX used fragment-based methods to pursue inhibitors of the MET receptor tyrosine kinase (RTK). Activating mutations in this kinase are found in human cancer germline mutations, such as hereditary papillary renal cell carcinoma, and also in sporadic cancers, such as non-small cell lung cancer. MET RTK activating mutations have been found in cancer metastases. In addition, some cancers show MET RTK gene amplification, including gastric cancer and EGFR inhibitor-resistant lung cancer. This work resulted in the development of the drug candidate SGX523. Preclinical assays showed that SGX523 had a favorable toxicity profile, promoting xenograft tumor stasis and regression. This molecule was also an attractive drug candidate because it was very selective, binding the nearest related RTKs with greater than 1000-fold lower affinity.

However, phase 1 trials of SGX523 did not go well. Patients who were administered the compound at doses greater than 80 mg per day experienced rapid onset renal failure. It was later determined that the metabolism of this compound was significantly different in intact humans compared with the human liver microsomes in which it had been tested. Studies in monkeys revealed that the compound formed insoluble crystals in the kidneys. Fortunately, researchers at SGX had developed a backup as well, a related compound that turned out to be metabolized differently. This and other compounds in the series are currently moving forward in development.