EXECUTIVE SUMMARY

The formation of the Developmental Therapeutics Program (DTP) by the National Cancer Institute (NCI) 43 years ago was a crucial step in initiating the search for drugs useful in the treatment of cancer. This occurred at a time when there were essentially no major programs whose focus was on drug development in the area of cancer therapeutics. The program has been held in high esteem throughout the world and has made major contributions to the discovery and development of cancer chemotherapeutic agents. In fact, many of the drugs used in the clinic today came through this program.

In the ensuing years, interest in cancer therapeutics has burgeoned with academic laboratories, cancer centers, and pharmaceutical and biotechnology companies actively participating in the search for new cancer drugs. One incentive for this activity relates to the explosion of knowledge in cancer biology and the availability of new molecular targets for drug development. At the same time, it has become abundantly clear that cancer represents a number of different diseases and that the workings of each cancer cell are complex and devious. To date, a reliable and effective formula for searching for, synthesizing, and screening new drugs remains elusive. In 1997, the Director of NCI formed the Developmental Therapeutics Review Group and charged it with the task of defining the future of NCI with respect to the discovery and development of new chemical and biological therapies for the treatment of cancer.

Over the course of a year, the Review Group worked to develop a series of recommendations that it believes will enhance the ability to discover new and useful antitumor drugs during the next decade. The goal is to foster the discovery of drugs which are not simply antiproliferative agents, but rather have unique and novel mechanisms of action. To attain this goal, the Review Group adopted four recommendations that will result in far-reaching and major changes in the activities of the DTP.

Allocation of Funds and Roles of the Extramural and Intramural Programs

The Review Group strongly recommends that the in-house portion of the DTP budget be limited to approximately 15 percent of the total budget. An important role of DTP should be to assume a leadership position in informatics to facilitate the development of cancer therapeutics. It should assume responsibility for coordinating and disseminating information through an expansion of its current operations. Resources such as the natural product repositories, select chemical libraries, engineered cell lines, hybridization assay technology, standardized reagents for cancer immunotherapy, and information databases useful in the determination of protein structures, should be made available by DTP to qualified investigators. At the present time there is an insufficient flow of new information to cancer investigators working in the area of drug development. DTP should play a leadership role in facilitating such information flow. Extramural funding should be used to support cooperative groups such as the National Cooperative Drug Discovery Groups, National Natural Products Discovery Groups, contracts, and Centers of Excellence. Such centers should be sited in those areas of the country where there is a population of academic scientists who have the relevant expertise and the ability to collaborate with colleagues in government laboratories, and in the biotechnology and pharmaceutical industries. The goal is to assure that nationwide the most talented and best-trained scientists are working on cancer therapeutics.

Monitoring and Oversight of the DTP Research Portfolio

The Review Committee was enthusiastic about adopting a plan that would enhance dramatically the ability of DTP to make decisions concerning its resource allocations through the development of a mechanism to continuously monitor DTP activities. A major goal of this plan would be to create a discovery and development process for any drug target or drug candidate by bringing together and coordinating distinct proposals from different laboratories or institutions. This ability becomes particularly important when considering the multidisciplinary nature of drug development and the necessity of collaborations and coordination if progress is to be made in this area. An example of a glaring lack of strength at DTP, as well as in many academic laboratories interested in drug development, is in medicinal chemistry, an area that is crucial to modern drug discovery and development. By coordinating different proposals in different areas of research, for example in medicinal chemistry and mechanisms of action, progress in drug development would be enhanced.

The Review Group believes that NCI must have a flexible and rapid response mechanism in place to deal with changing research objectives and resource requirements that arise on a month-to-month basis. This could be accomplished by creating a committee of scientists (5 to 8) chosen from among the leadership of DTP, academia, and industry. This committee would be charged with the mandate to evaluate and fund or invest contract resources in projects that are submitted to DTP through the extramural program or proposed by in-house DTP personnel. The budget for this committee should be no less than $50 million that should come from the existing DTP budget. The Review Group felt that a lower budget would not give the committee a reasonable chance of demonstrating success.

The proposed committee would prioritize the distribution of resources available to DTP and be responsible for the discovery of novel therapeutics and novel drug discovery technologies as well as the development of candidate therapeutics. This committee would meet on a regular basis, at least four times a year, and would be empowered to issue Requests for Applications (RFAs), create Task-Order Agreements, or conclude contracts. The committee would further be empowered to recommend use of additional consultants to advise its membership on the quality of the scientific proposals it receives. However, the committee would retain the right to dispense funding to any proposals that it deemed useful whenever such proposals came to the attention of the committee. This approach would involve a "rolling approval" process for submitted proposals rather than the current system of joint review of all proposals at a single meeting.

The proposed committee would also have the authority to match scientists from different institutions that might possess complementary technologies and to encourage these individuals to work together by offering them seed money to conduct joint investigations. The committee would also be responsible for the composition of DTP and could dictate that the DTP scientific staff be changed over time to include, for example, more pharmacologists, chemists, or drug metabolism personnel, so as to open up any bottle necks in the drug development pipeline that may occur. Last, the committee would be empowered to recommend tenure and promotions to DTP staff based on their contributions to drug discovery and development projects, irrespective of a scientist's publication record. In many ways, the committee would function as an independent entity within NCI and would be responsible for its decisions solely to the NCI Director. In turn, the Director would be responsible for appointing the members of the committee and for evaluating the productivity of DTP under this new format.

The Decision Network Committee

There was agreement among the Review Group that the NCI Decision Network Committee, responsible for prioritizing drugs for clinical development, was not functioning properly. The membership of this committee should be expanded to include representatives of academia, including cancer centers, as well as NCI staff. A majority of the Review Committee believed that there should be a single Decision Network Committee with broad representation, including outstanding scientists in the area of biologics. Depending on the drugs being introduced at each meeting, scientists with expertise in specific areas should be invited to participate in the discussions. With a single broad-based Decision Network Committee, all drugs, including biologics, should be reviewed and compared in a fair manner. In opposition, however, the Review Group's Subcommittee on Biologics believed strongly that biologics should be analyzed for feasibility and prioritized by its own Biological Resources Branch (BRB) advisory board, and not by the Decision Network Committee.

Special Role of the Developmental Therapeutics Program Related to Drug Screening

The Review Group recommended four activities related to drug screening in which the DTP role is appropriate and relevant: 1) a focused screening program for active compounds using assays for which it has developed expertise and capacity; 2) providing public access to its repository of compounds, research tools, and information databases; 3) working with the government, academic, and industrial communities to develop, evaluate, and deploy new assays in both the internal and external scientific communities; and 4) fostering a more collaborative approach to screening by serving as a matchmaker between chemists and biologists for the analysis of novel agents.

The Review Group recognizes that some of the assays will be conducted by DTP in-house but most will be conducted extramurally at sites nationwide where relevant expertise exists. In this context, NCI should serve as a facilitator of external scientific study in these areas. To achieve these goals, the Review Group recommends that NCI establish an ongoing review mechanism (relying on internal and external expertise) to continuously evaluate the status of screening assays and advise the NCI Director on the best use of resources in this area.

Major Recommendations of the Review Group Subcommittees

The Review Group was divided into five subcommittees, each of which addressed a specific topic in detail. The groups were: the Subcommittee on Small Molecule Diversity and Screening Technology; the Subcommittee on Structural Inventory of Potential Drug Targets; the Subcommittee on Animal Models; the Subcommittee on Pharmacology and Toxicology, and Formulation; and the Subcommittee on Biologics. The full report is organized around the deliberations of the subcommittees; each chapter contains a series of recommendations that are programmatic in nature and will result in changes in existing programs. The entire Review Group discussed all of the recommendations. Detailed recommendations appear in the full report; major recommendations are as follows.

• NCI should support a chemical diversity program with the explicit goal of finding small molecules that can manipulate the function of all proteins or processes relevant to cancer.

• NCI should undertake a major new interdisciplinary initiative to acquire structural information on cellular targets that are potentially relevant to cancer. This would include establishment of an instrumentation and education resource dedicated to making X-ray crystallography broadly accessible to members of the cancer research community, in addition to vigorous participation in efforts to determine the structure of all proteins encoded in the human genome-alone and in complex with interacting cellular partners-that may be involved in the malignant process.

• NCI should reconfigure its program for screening compounds for anti-tumor activity to ensure responsiveness to changes in science and technology. The current 60-cell-line screen should be reduced to 3 cell lines focused on the identification of lead compounds based on inhibition of cell proliferation. The COMPARE program has been very valuable for identifying drug targets and should be maintained for selected compounds. However, NCI should establish a network of extramural sites that have expertise in the development of new assays to assess the effects of compounds on the biochemical, cell biological, and tissue physiological parameters that govern cancer cell pathogenesis and pathophysiology.

• NCI should establish Centers of Excellence in a variety of scientific areas, for example, pharmacology/toxicology core facilities with the technology in place to do state-of-the-art drug metabolism, pharmacokinetics, and drug absorption studies and simulations. NCI should develop methods to allow more accurate a priori determination of the potential for metabolism and/or toxicity.

• NCI should expand the scope of the Biologic Resources Branch by augmenting the categories of biological reagents that are currently being produced, and developing capabilities for production of recombinant vectors and novel production technologies.

Throughout the Review Group's deliberations there was a continual emphasis on the need for the DTP to be flexible, agile, and responsive to the needs of rapidly changing science and technology. New basic knowledge and technologies must be integrated into the process of discovering and developing anticancer drugs. It is only in this way that the DTP will be able to meet the challenges of drug development in the next decade.

INTRODUCTION

The development of new drugs for the treatment of cancer is a difficult task, yet the need for anticancer drugs has never been greater. Approximately 8 million Americans alive today have a history of cancer. This year about 1.2 million new cases will be diagnosed. Over half a million Americans will die of cancer this year, more than 1,500 people a day. This number will tend to increase as our population ages. In the United States, cancer is the second leading cause of death-responsible for one out of every four deaths-exceeded only by heart disease. The financial costs of cancer are staggering at over $107 billion a year.

While prevention is the ideal route to cancer control, there will always be cases of cancer which cannot be prevented and for which aggressive intervention is needed. Drugs that can prevent the initiation of cancer, slow its growth, or stop it altogether, are sorely needed.

In 1955, Congress recognized the need for public investment in the discovery and development of anti-neoplastic agents. The Developmental Therapeutics Program (DTP) of the National Cancer Institute (NCI) was established to support programs necessary for the development of new therapies for cancer. It currently supports programs necessary for:

• the preclinical development of novel therapeutic modalities for cancer (including chemotherapeutic agents, antibodies, and vaccines)
• acquisition, synthesis, and definition of activity in in vitro and in vivo models of cancer and HIV disease; and
• advancement of active agents in preclinical models toward clinical trials by establishing workable pharmaceutical formulation and definition of clinical pharmacology and toxicology.

During the past 43 years, NCI has been involved in the discovery and development of many of the anti-neoplastic agents currently in use. Over that period of time, the program has become an international resource: It has been the largest such public investment in the world and one that is without parallel in any other therapeutic area. However, full realization of the complexities of the malignant cell and rapid advances in cancer biology and chemistry plus the growth of a large and active biotechnology and pharmaceutical industry have renewed the debate about the suitable role of government in the drug development process. Is NCI investing in the right areas of innovation? Are its drug screening and acquisition programs state-of-the-art and appropriate?

CHARGE TO THE DEVELOPMENTAL THERAPEUTICS REVIEW GROUP

In 1997, the Director of NCI formed the Developmental Therapeutics Review Group (hereafter referred to as the Review Group, or the Group) to consider the role of NCI in drug development activities. Specifically, the Director asked the Review Group to:

• Evaluate how NCI conducts and facilitates drug discovery and the development processes that turn molecules into drugs suitable for human testing;

• Review comprehensively NCI's major tool for screening and characterizing anti-cancer activity (the 60-cell-line panel);

• Assess NCI's process for compound acquisition;

• Evaluate natural products discovery;

• Examine the various mechanisms for investigator-initiated discovery such as the National Cooperative Drug Discovery Groups and the National Natural Product Drug Discovery Groups.

• Review the contract activities by which lead improvement, bulk synthesis, toxicology, and pharmacology are accomplished.

• Examine the production facilities in Frederick, Maryland, including the fermentation plant and the monoclonal antibody and recombinant protein facility.

• Assess the role and outcomes of the Decision Network; and

• Evaluate how well NCI's development capabilities serve the needs of intramural scientists who are trying to bring new therapies to the clinic.

More generally, the Review Group was asked to review the status of NCI's drug development activities in a contemporary and futuristic context. Given the level of private activity in the development of new anticancer drugs how can NCI assist in the translation of biological and chemical discovery into agents with potential therapeutic importance? What should be NCI's involvement in screening? And is it appropriate to continue searching for natural products as anti-neoplastic agents?

To organize its review, the Group formed subcommittees in the following areas: 1) small molecule diversity and screening technology; 2) structural inventory of potential drug targets; 3) animal models; 4) pharmacology, toxicology, and formulation; and 5) biologics (see Appendix A for membership and meeting dates). This report is organized around those themes.

SMALL MOLECULE DIVERSITY AND SCREENING TECHNOLOGY

INTRODUCTION

Recent scientific progress in understanding signaling pathways and cell cycle regulation has provided a wealth of potential new targets for anti-cancer drugs. Targeting the signaling pathways involved in cancer cell growth may make it possible to treat cancer with far fewer side effects than are caused by conventional cytotoxic therapies. But in order to exploit these pathways fully, researchers must be able to find small molecules that can manipulate protein function and protein-protein interactions.

The traditional source of anticancer agents, as well as other drugs, has been natural products obtained from living organisms and identified through biological assays. To date, natural products have proven to be the most effective small molecules in terms of their ability to alter the function of proteins relevant to cancer. The DTP plays a special role in the collection of natural products both because of its extensive experience in organizing collection efforts and its long history of ensuring the intellectual property rights of the country from which a product originates. By continuing the discovery of new anticancer natural products and by applying the insights derived from a study of these natural products to the design and construction of libraries of synthetic molecules, it should be possible to develop new anti-cancer agents.

A general method for synthesizing small-molecule ligands that modulate protein functions would be extremely valuable both in cancer research and cancer drug design. Promoting the development of these molecules and exploiting their properties should be high priorities for NCI. With a focused effort, NCI should be able to facilitate the development of methods that will make the discovery of such ligands routine and to apply these methods to the discovery of ligands that modulate processes relevant to cancer.

It is important to note, however, that improved screening methods are needed to identify lead compounds. Each source will require its own effective means of screening. In addition, because new sources of compounds will yield large numbers of new chemical entities, new miniaturization and automation techniques will be needed to perform large numbers of assays in a short period. NCI can play a vital role in the development of high throughput "smart" assays, compatible with the new techniques for small molecule production and capable of detecting molecules that can regulate the newly discovered pathways critical to the etiology and maintenance of cancerous states.

In its deliberations, the Review Group assessed optimal strategies for incorporating advances in molecular diversity and screening into NCI programs.

MOLECULAR DIVERSITY

The Use of Small Molecules in Studying and Treating Cancer

Chemotherapy is an essential component in the treatment of cancer today. But current drugs are both toxic and ineffective against many common cancers. More effective drugs could be developed through two different approaches. One would be to develop improved cytotoxic agents that lack major toxicities of existing agents but continue to kill cancer cells by related mechanisms. An example would be an anti-mitotic agent that lacks the neurotoxicity of Taxol and the vinca alkaloids. The second approach, perhaps more difficult, but ultimately more effective, would be to develop agents that directly target the causes of cancer, such as the uncontrolled activation (mitogenic) or malfunctioning inactivation (checkpoint) pathways that initiate the disease. Both approaches require small molecules that specifically inhibit (or activate) novel target proteins.

Small molecules that specifically target new proteins can also serve as tools for basic cancer biology research. To understand a protein's function in vivo, we must be able to either inhibit it or activate it, preferably both, and preferably at controlled times. Until now, the most successful approach has been to mutate the gene encoding the protein of interest. But it is increasingly clear that cell-permeable, small molecules that bind to the target protein can complement, and possibly even replace, genetics.

Like mutations, small molecule ligands can either inactivate the protein (examples include staurosporine binding to protein kinase C, or trapoxin binding to the chromatin remodeling enzyme histone deacetylase) or activate it. Activation can result from an allosteric change (e.g., small molecule ligands binding to nuclear hormone receptors), the formation of a new binding site leading to the gain of a new function (e.g., when cyclosporin A binds to cyclophilin, a binding site for calcineurin is formed), or a ligand-induced change in protein localization (e.g., the use of FK1012 to co-localize fusion proteins). Two different routes will likely obtain ligands for all of these functions: biosynthesis or chemical synthesis.

MOLECULAR DIVERSITY FROM BIOSYNTHETIC PATHWAYS

The traditional source of anticancer agents has been natural products that were obtained from the producing organism, animal or plant, and identified through biological assays. NCI has traditionally played a strong role in this type of research, and there are still persuasive arguments for continuing these efforts either directly or through collaborations. Because biological diversity generates chemical diversity, these efforts should be broadly based.

Now is our last best chance to preserve at least a portion of the earth's biological diversity. The biological diversity that can be directly accessed is diminishing, and the habitats with the greatest diversity, such as tropical rain forests and coral reefs, are diminishing the most rapidly. Along with vanishing habitats, traditional knowledge about the uses of those habitats is also disappearing. In addition, the rate of species extinction is likely greater than is generally appreciated. This is because the disappearance of one species, such as a tropical plant, also leads to the extinction of mutualistic species such as insects and specialized microorganisms such as viruses and endophytic fungi. Forces that include increased development, population pressure, and, possibly, global climate change is likely to accelerate the rate of species extinction.

Ironically, the impediments to collecting have increased along with the urgency of doing so. Individual investigators face enormous difficulties in ensuring the intellectual property rights of all participating parties. NCI, however, through a tradition of pioneering mutually acceptable arrangements and its long experience in collecting, can facilitate the collection process.

Advances in our ability to understand and control biosynthetic pathways in the laboratory have opened exciting new prospects for chemical diversity from biosynthesis. Entire biosynthetic operons have been moved from one producing organism to another, and the modular nature of many biosynthetic pathways has been analyzed in detail. We can already modify the final products of biosynthesis by manipulating genetic information. For example, we can produce a novel variant of erythromycin by disabling some biosynthetic steps and providing an "unnatural" starting material. Plans to construct combinatorial biosynthetic pathways that exploit the modular nature of known pathways are well advanced and could potentially lead to large libraries of biosynthetic products ("unnatural natural products").

Another exciting prospect for using the genetic information of biosynthetic pathways to provide molecular diversity is direct cloning of soil DNA. Several groups have investigated the DNA content of soil and concluded that traditional culturing methods capture roughly 0.1 percent of the available genetic pool. While some groups have focused on alternative culturing methods, others have begun to directly extract soil DNA, chop it into fragments likely to catch entire biosynthetic pathways, and introduce these new genes into more readily cultured organisms. The prospect of large clone banks representing the "metagenome" available from soil microorganisms is a potential outcome.

MOLECULAR DIVERSITY THROUGH CHEMICAL SYNTHESIS

It seems likely that synthetic molecules that mimic natural products are most likely to target proteins successfully and that attempts to design small molecules that target proteins will be most successful if the lessons of natural product chemistry are applied. This is because known natural products epitomize the properties that define optimal ligands: high affinity, selective binding to proteins, and promotion or disruption of specific protein-protein interactions.

Today, modern asymmetric synthesis can be used to synthesize virtually any naturally occurring substance, albeit in multi-step processes which often require years of focused effort to develop. In the future, large libraries of natural product-like compounds with rigid frameworks and diverse substituent groups will likely be synthesized. This will be accomplished by applying the new synthetic techniques in the context of new methods to combine synthetic building blocks combinatorially (see below).

A general method for synthesizing small-molecule ligands that modulate protein functions would be extremely valuable both in cancer research and in cancer drug design. We believe that it is now possible to develop a method to make the discovery of such ligands routine and apply that method to the discovery of cell-permeable small molecules that manipulate protein functions relevant to cancer.

LEARNING FROM NATURAL PRODUCTS

It should be possible to apply the lessons of natural product chemistry to the development of new anti-cancer agents. This could be done by employing the insights derived from a study of natural products known to bind to proteins to the design and construction of libraries of synthetic molecules. Although we do not fully understand the principles that govern the interaction between small molecules and proteins, several observations can be made that should be used to guide library design. It is an important priority to continue to try to more fully understand the processes and mechanisms and to explore additional principles that govern interactions between small molecules and proteins.

A major proportion of the known biologically active natural products is relatively large structures with a high degree of stereochemical complexity. A common theme among the molecules is a spatially well defined presentation of functionality, usually over large distances. Many natural products are therefore relatively rigid structures (e.g., Taxol, which interacts with and binds to microtubules, has a very rigid structure). Natural products bearing large, acyclic arrays, such as the polyketides, also have relatively rigid structures due to minimizing syn-pentane interactions and allylic strain. It is easy to rationalize the idea that rigid molecules should be better at binding to proteins; more rigid molecules will have fewer degrees of freedom, so the loss of entropy upon binding to a large molecule such as a protein will be reduced. It has also been learned from nature that, to bind large and often shallow protein surfaces, large chemical structures will be required.

Charge is often used sparingly in natural products (e.g., for alkaloid molecules such as aspidospermine, often one or two nitrogen atoms provide most of the charge in the form of an ammonium ion at physiological pH). If a natural product does possess a large number of charged functionalities, it is often balanced with a large proportion of hydrophobicity. Such is the case for amphotericin B, which displays polyhydroxylated and polyeneyl surfaces. This balancing of charge with hydrophobicity may be important for cell permeability, and it may be particularly important for targeting small hydrophobic pockets on protein surfaces

The known examples of natural product-protein interactions provide some insight into the requirement of size, as well as functionality, for the design of libraries that will interact with diverse protein binding surfaces. Most, if not all of the libraries that have been constructed have been relatively small, focused structures. In the future, it will be important to construct libraries of unprecedented size and complexity, using the power of modern asymmetric synthesis. Natural-product-like libraries will be far more likely to contain protein ligands than the libraries built to date.

Organic synthesis has allowed us to access complex molecular structures on demand. Until recently, these molecules were assembled in a "one reaction per vessel" process, followed by further optimization of the structure. The new approach of synthesizing collections, or libraries, of compounds simultaneously has revolutionized our ability to construct large numbers of related molecules rapidly.

There are currently two methods for the construction of small molecule libraries. The first is parallel synthesis, a miniaturized version of the traditional "one reaction per vessel" method using a multi-vessel apparatus and robotics. Most libraries in the pharmaceutical industry have been constructed with this technique and have been limited in size and diversity. The library members are typically heavily biased to be similar to a well-known "pharmacophore." These libraries often yield large quantities (over 100 mg) of small heterocyclic molecules devoid of stereochemistry.

The second method has similarities, in principle, to methods seen in nature. The synthesis of polyketides such as rapamycin and FK506 involves an iterative sequence that includes sequential Claisen condensations, ketone reductions, dehydrations, and enone reductions. The enzyme modules that perform these functions appear to have been shuffled throughout evolution by genetic recombination. Over time, this shuffling can be viewed as having produced a library of related molecules. The molecules that confer a growth advantage on their host organism are favored, and the organisms that express them tend to form the basis for the next round of gene shuffling. The "split-pool" synthesis method, which involves simply the mechanical intervention of pooling and then splitting flasks during key coupling reactions of multi-step syntheses, does not generate diversity by gene shuffling. It does, however, subject diverse monomers to iterative chemical reactions in an order defined by the user. Split-pool synthesis allows for the convenient synthesis of large libraries of compounds (>one million) in a small number of chemical steps.

The split-pool method is promising, but it is not yet capable of generating compounds of complex natural-product-like structures. It typically yields only minute quantities for use in the early stage of biological analysis, thus making it necessary to develop miniaturized screening methods. (After finding an interesting compound, however, large quantities of the compound are generally available via re-synthesis on the solid phase.)

Although the total syntheses of complex natural products such as rapamycin, Taxol, and calicheamicin have been accomplished, there remains a significant gap between the synthetic strategies that have been used in these projects and the current state of the art in library synthesis, especially split-pool synthesis. However, some strides toward the synthesis of such libraries have been made, and through these efforts, it has become apparent that the library synthesis strategies will have to diverge from those traditionally used to synthesize natural products. Solid phase reactions will play a key role during the coupling steps (where pooling and splitting of reaction flasks occurs), yet traditional solution phase reactions will be required to efficiently prepare the key building blocks of such syntheses. Linkers will be required that are compatible with modern reaction processes, yet that will allow the controlled release of synthetic molecules into miniaturized assay systems. Finally, improved public domain encoding strategies will be required.

ASSAYING THE NEW MOLECULES

The Need for Nanoscale, High-Throughput Assays

Conventional screening in the pharmaceutical industry is generally performed using automated systems that conduct analyses in a 96-well plate format. Small molecules are added in the form of stock solutions, which are made using relatively large amounts of compound stored in vials. This approach has been very useful for screening natural product libraries and synthetic libraries of limited complexity. It is expensive, however, and the high costs (in both money and space) for consumables and robotics tend to preclude its application in academic laboratories. Also, the method is neither rapid nor sensitive enough to screen the million-member libraries that will be constructed using solid phase split-pool synthesis.

Screening Methods Based on Small-Molecule-Dependent Genetic Selections and on the Use of Other Engineered Cell Lines

As previously mentioned, targeting signaling pathways involved in cancer cell growth might make it possible to treat cancer with far fewer side effects than conventional cytotoxic therapies. Finding drugs that block specific protein-protein interactions is a particularly important goal. Signaling pathways function by virtue of a series of such interactions, and disrupting them would be a powerful new approach to cancer therapy. Improved screening methods are needed to identify lead compounds.

Traditionally, the main approach to finding inhibitors of a specific protein, or of a specific protein-protein interaction, has been to purify or express the relevant proteins, and then screen for small molecule inhibitors using pure proteins in vitro. This proven approach has generated many useful drugs. However, pure protein assays have the disadvantage that they do not require the small molecule to manifest its effect in the environment of the living cell, with all of its defense systems and alternative targets.

In contrast, cell-based assays do require the small molecule to act in an environment that is relatively similar to the environment that will be experienced by the small molecule during its experimental or therapeutic use. These assays are thus more likely to generate useful drugs. In the past, however, cell-based assays have tended to be restricted to specific phenotypic effects inherent to a given cell line. There is an increasing interest in the pharmaceutical industry in the genetic engineering of cells to allow the assaying of specific protein function in live cells. But no successful, general methods for screening for compounds that either bind to a specific protein inside a cell, or disrupt a specific protein-protein interaction have yet been found.

It is easy to list the features one would desire from such assays: the target proteins should be introduced into the cell as cDNAs; the assays should be general for any protein or protein interaction; and the assays should be robust and suited to high throughput screening. For synthetic libraries, the assays should be performed in tiny volumes to allow screening of our large split-pool synthesized libraries.

A more subtle, desirable feature is for the assay to take the form of a genetic selection, so that cells where the small molecule has its specific effect can be detected by some optical means or by a growth advantage. This would allow the small molecule to select out a single cDNA-the cDNA for the protein with which it interacts-from a library. This feature would facilitate identification of the protein targets of small molecules selected from cytotoxicity or cytology-based screens. In the future, it would also allow screens of libraries of compounds against libraries of cDNAs, which would lead to accumulation of large amounts of information about protein-ligand interactions, information that could be significant in cancer research.

USING MOLECULES TO ADVANCE OUR UNDERSTANDING OF CANCER BIOLOGY

Target Identification

When anti-cancer drug leads are discovered by their phenotypic effect on cells, for example in cytotoxicity or cytological assays, it is important to identify their targets. Target identification allows us to preview the likely utility of a new agent. For example, a small molecule that targeted DNA would engender little excitement, as we already have many such compounds. However, a small molecule that targeted topoisomerase would be more interesting, as this is a proven therapeutic target for which fewer drugs exist. By far the most exciting compounds would be those that target new proteins such as proteins involved in signal transduction pathways or proteins other than a/ß-tubulin that are involved in mitosis. Target identification also sets the stage for optimizing the affinity of the small-molecule-protein interaction and understanding how the small molecule works structurally and functionally.

Currently, methods for target identification are relatively slow and unreliable. An exception is the NCI multiple cell line screen, which, when employed in conjunction with the program COMPARE, is very useful as an empirical method for identifying molecules that are likely to target DNA or any protein that is the target of a known anti-cancer agent. It can also provide an indication that the effect of a small molecule may be novel. But for small molecules with novel activities, this screen cannot suggest a likely target. Researchers have tended to rely on methods in which the small molecule is used to purify its target from a cell extract, usually in an affinity chromatography format. This method is often effective.

Affinity chromatography suffers from some disadvantages, however. If the protein target is present at low abundance in the cell extract, it can be difficult to detect in the face of more abundant nonspecific binders. Furthermore, protein targets are identified only as gel bands. Identifying the cDNA that encodes the protein found in a particular gel band is much easier than it used to be, but microsequencing is still time-consuming and expensive, and cloning from microsequence can be difficult if the protein is not already known.

A technique that leads directly from small molecule to cDNA would be preferable. Several techniques in the literature have this capacity in principle. For example, phage display allows the construction of bacteriophage that express a specific protein from a cDNA library on their outer surface. Phages that bind to a small molecule could be selected and amplified. Phage display is an attractive technology, and as it improves from work in other laboratories, it may be adopted for target identification. Currently phage display is limited because the protein of interest may not express well or fold properly on the phage surface, and efforts to express cDNA libraries on the surface of phage have failed. Successful application of the technique has therefore been mostly limited to small proteins and peptides.

Similar problems exist with other target identification systems that rely on protein expression in bacteria. Strategies for identifying small molecule targets directly from cDNA libraries may also be developed based on in vitro expression of cDNA pools in reticulocyte lysate, a technique that has been used successfully to identify kinase substrates, specific protease substrates, and proteins degraded during mitosis.

FUNCTIONAL GENOMICS AND HYBRIDIZATION ARRAY TECHNOLOGY

A more radical approach to target identification is now becoming available from progress made in the area of functional genomics, in particular hybridization array technology. Functional genomics deals with whole genome experiments. Using these new tools, we can study the influence of perturbations on a whole genome, not just an isolated portion. The NCI has invested heavily and wisely in genomics and related technology. The Review Group focused on mechanisms for promoting the growing interface between these technologies and small molecule diversity.

Hybridization arrays allow the measurement of the levels of a large panel of mRNAs in a cell. In yeast, the whole genome can now be used as the panel, and this may become possible in human cells over the next few years. For example, fingerprints of the expression levels of tens of thousands of mRNA species from cell and tissue samples will be routinely available. Panels of such mRNAs from normal cells, drug-resistant cells, and from tissues at different pathologic stages or with differing proliferative stages can be set up for probing and high throughput screening. Through hybridization arrays, we can determine how the levels of specific mRNAs respond to the addition of a small molecule whose target is unknown and compare this to its response to the effects of other small molecules and mutations.

Hybridization array technology is a rapidly developing field and two main strategies appear to be emerging: 1) chemical synthesis of relatively small DNA fragments that probe the entire genome; and 2) spotting technologies. Both have advantages and disadvantages and it is too early to prescribe the best approach for any given problem.

This technology provides a kind of "digital fingerprint" of the phenotypic effect of a small molecule, and it is expected that small molecules with similar mechanisms of action will provide related fingerprints. For example, all inhibitors of a given signaling pathway should provide a related fingerprint, and all tubulin binders should provide a different fingerprint. This technology will greatly facilitate the process of determining whether the effect of a small molecule is similar to that of a previously known small molecule.

Similarly, the fingerprint of a small molecule that inhibits a certain signaling pathway will be similar to that of a mutation that disrupts the same pathway. By mapping small molecule effects onto mutations, it should be possible to directly identify candidate gene targets. Banks of mutations that systematically cover the genome are already available in yeast and will eventually become available for other organisms, including the mouse and humans.

RECOMMENDATIONS

STRUCTURAL INVENTORY OF POTENTIAL DRUG TARGETS

INTRODUCTION

The astonishing speed with which HIV protease inhibitors were developed is a clear testament to the value of this approach, and a number of the most promising anticancer drug candidates currently in clinical trials-including the mechanistically novel inhibitors of matrix metalloproteases-were developed via iterative structure-based drug discovery. In addition, many other ongoing preclinical development programs throughout the industry, aimed at inhibiting such diverse targets as cell cycle-dependent protein kinases (CDK2,4), the papillomavirus E2 protein, and vasoendothelial growth factor (VEGF), are making extensive use of structural analysis.

Structural analysis of target proteins in the unliganded state or bound to other macromolecules can provide an impetus for the initiation of drug discovery programs. One recent discovery involves the interaction of the tumor suppressor protein p53 with a modulator protein, MDM2. Because MDM2 inactivates p53 by directly binding its transcriptional activation domain, and because MDM2 overexpression is associated with a loss of p53 function in certain tumors, it is widely thought that an MDM2-binding small molecule that out-competes the p53 activation domain (p53-AD) would be a valuable therapeutic agent. However, experience has proven that protein-protein interactions are generally difficult to antagonize with small molecules because the binding interfaces typically comprise flat hydrophobic surfaces. Unexpectedly, however, the co-crystal structure of p53-AD bound to MDM2 revealed that the activation domain binds into a deep hydrophobic cleft on MDM2. This finding has greatly increased the prospect of small-molecule inhibition, and several companies have since launched efforts aimed at discovering small-molecule MDM2 inhibitors.

Thus there are numerous examples of how structural analysis can facilitate the process of ligand and drug discovery. One exciting direction for the future will entail the use of combinatorial chemistry to drive the initial discovery of ligands to target proteins and to optimize these ligands in close conjunction with high-resolution structural analysis.

The Review Group believes that NCI, operating through the DTP, should focus on removing the obstacles to the process of ligand discovery. In this way, NCI would contribute to the war against cancer by facilitating discoveries that would drive drug development efforts in universities and the biotechnology and pharmaceutical industries.

IMPEDIMENTS TO THE EFFECTIVE DEPLOYMENT OF STRUCTURAL ANALYSIS IN THE WAR AGAINST CANCER

Although structure-aided ligand discovery is an extremely promising strategy, several serious obstacles, discussed below, have prevented it from reaching its potential.

Limited Access to Specialized Instrumentation and Training Required for Structural Analysis

The current system of federal research support does not provide an adequate mechanism for investigators not already expert in structural analysis to gain access to the required equipment and training. The high cost of X-ray and NMR equipment and the specialized expertise required to run experiments on these instruments exclude large numbers of scientists who are interested in pursuing structural studies.

Many of the most interesting structural problems will involve complex multicomponent systems, such as those involved in intracellular signaling, transcriptional activation, and protein trafficking. Multicomponent complexes are notoriously difficult to crystallize, and, frequently, usable diffraction data can be obtained only by using a synchrotron radiation source. But because there is insufficient synchrotron time for current users, a significant expansion of users is not practical without a corresponding increase in the availability of synchrotrons. Moreover, there is no mechanism in place for noncrystallographers to obtain training and assistance at these facilities.

Difficulties in Protein Expression

The difficulty involved in expression of proteins is one of the most serious practical factors limiting structural analysis today. There is a need for increased emphasis on the development of novel protein expression systems. For example, even though biosynthetic deuterium labeling of proteins has extended the range of NMR spectroscopy into the 30-50 kDa range, efficient expression systems using D2O are available only for E. coli. Even in normal media, many proteins fail to fold properly when overexpressed or are toxic to the overexpressing organism. The small volume of the outer-cell membrane to which these proteins are usually targeted seriously limits the overexpression of membrane proteins.

Lack of Structural Representation

High-resolution structures of more than 3,000 unique proteins and protein domains are now available, and new structures are being deposited into public-domain databases at a rate of roughly two per day. However, these structures still comprise an almost minuscule fraction of the protein structures that are potentially relevant to cancer. Fewer than 1 percent of the proteins encoded in the human genome have been structurally characterized, and fewer yet of these "characterized" examples involve proteins that have all of their functional domains intact.

Need for Novel Approaches in Rational Ligand/Drug Design

Much of the burden on high-resolution structural analysis and synthetic chemistry would be relieved if computational modeling could be used to predict the absolute or even relative binding energies of ligand-macromolecule complexes. For this to become possible, fundamental advances are needed in our understanding of the attractive and repulsive forces that control binding interactions in water.

STRUCTURE-BASED APPROACHES AND THE DTP

At present, the capabilities for determination of more explicit high-resolution structures in the DTP consist of a small and skilled crystallography group at Frederick, Maryland. This group operates in a manner that is disconnected from other small-molecule drug discovery and development efforts. The lack of a close connection between synthesis, screening, and structure is a serious deficiency that renders inadequate and outdated the overall DTP effort in small-molecule anticancer drug discovery. The DTP does provide some support to academic laboratories involved in ligand-target interactions relevant to cancer through investigator-initiated research grants, and especially noteworthy contributions have been made in the area of cyclin-dependent kinases and their complexes with inhibitors. However, groundbreaking research in high-resolution structural analysis is not a readily identifiable strength of the extramural NCI program.

RECOMMENDATIONS

The Review Group recognizes that some of the following recommendations can and should be undertaken within the constraints of the current NCI budget for developmental therapeutics. Other recommendations are highly meritorious, but their scope extends both scientifically and financially beyond that of the current NCI developmental therapeutics program.

ANIMAL MODELS

INTRODUCTION

PHARMACOLOGY, TOXICOLOGY, AND FORMULATION

INTRODUCTION

RECOMMENDATIONS

DEVELOPMENT OF BIOLOGICS BY NCI

INTRODUCTION

RECOMMENDATIONS