NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Kufe DW, Pollock RE, Weichselbaum RR, et al., editors. Holland-Frei Cancer Medicine. 6th edition. Hamilton (ON): BC Decker; 2003.

  • By agreement with the publisher, this book is accessible by the search feature, but cannot be browsed.
Cover of Holland-Frei Cancer Medicine

Holland-Frei Cancer Medicine. 6th edition.

Show details

Evolution of the Discovery Process

, MD and , MD, PhD.

Clinical Observation in the Discovery Process

As this chapter is being written, the process for discovery is changing rapidly. Initial anticancer drug discovery was based on important observations by clinicians and preclinical scientists. An example of such observations, where many feel the field of chemotherapy of cancer began, was the observation by Adair and Bogg in 1931 3 that troops in World War I exposed to sulfur mustard developed lymphoid and bone marrow hypoplasia. They then used this very toxic agent to treat a number of patients with cancer with sulfur mustard. The agent had a limited effect. Further studies of the related nitrogen mustards (of World War II) demonstrated more substantial antitumor effects, particularly in patients with lymphoma. 4, 5

Another classic example of an astute observation by clinicians is when Farber and colleagues noticed that young patients with leukemia who were treated with liver extracts or folic acid had worsening of their leukemia. 6 They then treated the patients with folic acid antagonists and described the antileukemic activity of these antagonists.

Histocytotoxic Effects and Physiologic Observations

Just like the observation that mustard gases cause hypoplasia of lymph nodes, spleen, and (with subsequent clinical activity of nitrogen mustard and other alkylating agents in patients with lymphoma), there have been a number of other correlations between specific histocytotoxicity (toxicity to a particular type of cell) and clinical activity of an agent. One of the best examples is the finding that the agent streptozocin caused destruction of the islet cells in animals (leading to diabetes). 7 Streptozocin was tried and found to have clinical activity in patients with insulinomas (a histocytotoxic effect). 8

In early animal toxicology studies, cisplatinum caused hypoplasia of the testes and ovaries in animals and was subsequently found to have substantial activity in patients with testicular or ovarian cancer. 9–11

As an example of a physiologic observation leading to the development of a series of new anticancer agents, the physiologic observation that rat hepatomas incorporated more uracil than the surrounding normal liver led to the Heidelberger group synthesizing the 5-fluoropyrimidines. 12, 13 The observations by Elion and Hitchings, who surmised (based on the presumption that nucleic acids were in control of cell growth) that purine and pyrimidine analogs would be useful in treatment of cancers, certainly led to the development of 6-mercaptopurine, 6-thioguanine, and many other useful anticancer agents. 14, 15 Although there is still clearly room for important clinical, toxicologic, and physiologic observations, these observations will probably not dominate future drug discovery. However, readers are encouraged to keep their eyes open to clinical, histocytotoxic, and physiologic observations as described above.

Early and Continuing Role of the National Cancer Institute Drug Discovery Program in the Drug Discovery Process

In the early 1960s, the US National Cancer Institute (NCI) began a major leadership role in the discovery and development of new anticancer agents. This was critical at that time because industry felt that investment in this area would not be feasible financially and that the risks were high. The NCI began its leadership by establishing the National Service Center (NSC), which served as a clearinghouse (and a bank of compounds) for all compounds for all chemists, biologists, and other interested investigators. Compounds could be submitted to the NSC, assigned an NSC number (eg, NSC 3101139 for mitoxantrone), and tested in a series of in vitro and in vivo preclinical models. Agents that passed certain hurdles of in vitro and in vivo antitumor activity were considered by a body at the NCI called the Decision Network for bringing them into toxicology studies and eventually into clinical trials.

The very early efforts of the NCI used the KB squamous cancer cell line to screen a large numbers of compounds. Compounds that were active against that cell line were further evaluated in vivo in the two murine leukemias called P388 and L1210. 16, 17 These in vitro and in vivo screens did lead to compounds with clinical activity against leukemia and lymphoma (eg, vinca alkaloids, nitrosoureas). Finding agents with antitumor activity against solid tumors was much less successful. 17 In 1975, the NCI revamped their evaluation process to include in vivo evaluation of compounds against solid tumors (both autologous animal tumors and human tumors growing as xenografts in nude mice). A very important retrospective analysis of that big experiment was conducted by Staquet and colleagues. 18 The results of their analyses are detailed in Table 45-2.

Table 45-2. The Most Predictive Animal Models: The National Cancer Institute Experience 1978–1982.

Table 45-2

The Most Predictive Animal Models: The National Cancer Institute Experience 1978–1982.

As can be seen in the table, the P388 and L1210 leukemias continued to be predictive for activity of new agents in the clinic, as did the B16 melanoma solid tumor autologous model growing in immunocompetent mice. The only human tumor xenograft that appeared to be predictive for clinical antitumor activity was the MX1 mammary tumor xenograft growing in nude mice. The predictivity of these models was not tumor specific but rather in vivo activity in those systems predicted, in general, for clinical activity. 17

Also of importance in the report by Staquet and colleagues was the finding that actual regression of established tumors in these models and the percentage of animals surviving for ≥ 45 days were even stronger predictors of eventual clinical efficacy of the agents. 18 Thus, although these in vivo evaluation systems are often maligned, when a systematic large-scale evaluation of them was examined, some models appeared to be reasonable predictors (although not tumor-specific predictors) for activity in the clinic. 19

Of additional note in Table 45-2 is the finding that five models were judged as not predictive at all for clinical activity. Some of the models, such as the Lewis lung model, continue to be used to justify bringing many new agents into clinical trials (particularly new agents, such as the angiogenesis inhibitors). It is important that investigators who are contemplating bringing new agents into the clinic based on preclinical activity only in the Lewis lung, CD8, Co38, LX1, or CX1 models be mindful of the extensive prior experience demonstrating them not to be predictive for clinical activity.

To be most successful, however, it is important to remember the parameter Staquet and colleagues used for evidence of activity in the clinic (to make the correlation with the preclinical models): tumor shrinkage. Thus, their findings might not be applicable to the newer generations of more cytostatic agents (rather than the cytotoxic agents being evaluated in the 1960s to 1980s).

The NCI made a further revision in their evaluation strategy in 1985 when they placed a large series of 60 human tumor cell lines in place for screening purposes. 17, 20–24 They have attempted to make this a more tumor type-oriented approval with inclusion of breast, lung, colon, kidney, brain, ovary, prostate, and melanoma human tumor lines. 17, 21, 22 At the time of introduction of this concept into the drug evaluation process, there was a great deal of discussion in the research community on how these cell lines were selected. Many investigators felt that cell lines with too rapid a doubling time were selected and did not represent the more slowly dividing solid tumors. However, this 60-cell line screening panel is the system that is currently in place. More can be learned about the 60-cell line success by visiting the Website of the NCI at <>. Agents that are deemed active in the cell line screens (based on criteria outlined in references 17 and 22 and based on other criteria, such as disease-specific activity) are then evaluated in the traditional in vivo models to determine their therapeutic index. If they are found active, they are moved forward into preclinical toxicology studies (see below). 23, 24

The true predictivity of the NCI 60-cell line success is as yet unknown as there are no published analyses as to the true predictivity of this approach for antitumor activity in the clinic. However, many excellent analyses of the results from the NCI 60 cell lines have led to some very provocative and useful research tools.

As an initial example of a unique use of results from the 60-cell line screening, Paull and colleagues 25 described a method (which they dubbed the COMPARE program) in which a compound evaluated in the 60-cell line screen is used as a seed and the COMPARE program is used to detect the compounds that have similar patterns (or fingerprints) of activity against the 60 cell lines. A correlation coefficient is produced to indicate the closeness of other agents to the pattern of activity of the seed compound. It appears that a significant correlation coefficient between two compounds will predict for similar mechanisms of action of the compounds—even the existence of a similar intracellular target, as described by Weinstein and colleagues. 26 If the correlation coefficient of a new compound is low compared with the compounds evaluated in the NCI screen, it might substantially increase interest in the new compound as it is very likely to have a new (unique) mechanism of action.

Of additional importance and of great interest to the rest of the cancer research community is that the NCI 60 cell lines continue to be characterized for presence or absence of specific targets (kinases, telomerase, mismatch repair proteins, receptors). 27–29 The more precise characterization of these lines is relatively certain to help the evaluation of compounds in the lines.

Other preclinical models explored by the NCI for evaluating compounds that probably deserve mention were models put in place to evaluate agents against tumors taken directly from patients. There were two systems including the in vitro human tumor cloning assay (HTCA) and the in vivo subrenal capsule system (SRC). The NCI began a program using the HTCA to screen compounds for activity against tumors taken directly from patients and growing as colonies in soft agar. 30 Although initial correlations were promising between agents detected as active in the HTCA and having subsequent antitumor activity in the clinic, the HTCA was deemed as too logistically difficult to use as an up-front screen to evaluate the vast array of compounds. However, several groups of investigators continue to use the HTCA to determine whether an agent should be taken into the clinic against a particular histologic type of malignancy (eg, gemcitabine against pancreatic cancer). 31, 32

The SRC assay is a system in which tumors are taken directly from patients and are placed under the renal capsule in mice. New agents are administered to the mice to determine the effects of the agent against tumors growing under the renal capsule. 33 Once again, although initial correlations between activity in the SRC system and the clinical activity of the compounds appeared promising, the logistics of evaluating a large number of compounds in this in vivo system were deemed to be too daunting. Thus, attempts using these two methods (the HTCA and the SRC systems) to evaluate large numbers of new compounds against tumors taken directly from patients (and thus not corrupted by in vitro passage in culture) were not pursued based on logistic problems (basically, the unavailability of tumors taken from patients and the technical challenges of the assays—ie, lots of manpower required). However, most investigators would agree that it would be helpful to have some way to evaluate a large number of compounds against tumors taken directly from patients.

Targeted (or Mechanistically Based) Drug Discovery and Evaluation

With the explosion of molecular techniques and knowledge of cell biology, there is now an incredible array of new methods for discovery of potential targets present in tumor cells versus normal cells.

The most recent spectacular example of a targeted approach to therapy is the development of the new agent Gleevec, which has substantial activity against chronic myelogenous leukemia and against gastrointestinal stromal sarcomas. To put the timeline for development of Gleevec into perspective, the basic biology ground work for the discovery and development of Gleevec actually began with the discovery of the Philadelphia chromosome in chronic myeloid leukemias (CMLs) in 1960. 34 This was actually the first consistent chromosomal abnormality discovered in any type of cancer. The abnormality is a translocation involving chromosomes 9 and 22. 35 Work done in 1990 to 1993 documented that the translocation caused the ABL gene on chromosome 9 (a nonreceptor tyrosine kinase) to move next to the BCR (breakpoint cluster region) gene on chromosome 22. This translocation codes the BCR-ABL protein p210 BCR-ABL, which is a 210 kDa transforming protein tyrosine kinase constitutively expressed and responsible for about 95% of patients with CML. 36, 37 This abnormal tyrosine kinase does not exist in patients' normal cells; therefore, it was an excellent target for development of an agent with activity against CML.

Screening was conducted to find compounds with specific activity against the BCR-ABL tyrosine kinase, and the agent (CGP57148, aka STI517, Gleevec) was found to be a selective inhibitor of the kinase by targeting the adenosine triphosphate binding site (pocket) of the enzyme. 38 The agent was tested in vitro and in vivo and found to be selectively toxic to BCR-ABL-positive cells. 38–40 Phase I and II trials in 1999 to 2000 documented that dramatic activity of Gleevec against CML, particularly in the chronic phase. 41, 42 As noted above, this incredible presence of drug development really spanned nearly 40 years from identification of the target until documentation of clinical activity of a compound against the disease (only 10 years if one starts counting from the time of discovery of the target, the p210 BCR, ABL tyrosine kinase). Of note is that additional work has shown that Gleevec also has substantial activity against gastrointestinal stromal sarcomas, which possess abnormalities in the Kit oncogene. 43

Although not within the scope of this chapter, a variety of other targeted therapeutics have also demonstrated substantial clinical antitumor activity (antiestrogens, antiandrogens, aromatase inhibitors, monoclonal antibodies to CD20, CD52, HER-2/neu, etc). However, at the time of the writing of this chapter, there are also some disappointments in the area of targeting. For example, the farnesyl transferase inhibitors that have been developed to target tumors with mutations in ras have not shown activity against pancreatic cancer, 44 even though they have clear clinical activity against breast cancer and acute leukemias, where ras mutations are not thought to be critical to the development or progression of either disease. 45, 46 Another example of difficulties with targeting is the finding that despite the upregulation of the epidermal growth factor receptor (EGFR) in a number of malignancies, the activity of small molecule inhibitors of the EGFR kinase or monoclonal antibodies to the EGFR do not appear to correlate with the increased expression of EGFR (at least by currently available methods of measuring the receptor/kinase). 47

Clearly, there is a great deal of work to do if targeted therapy is to succeed in major tumor types. In a recent piece of work, Druker and Lydon outlined the lessons learned for the development of Gleevec. 48 These lessons provide guidelines for thinking about more targeted approaches, including the following:


Identification of an appropriate target for drug development. The identification of the BCR-ABL tyrosine kinase as the kinase to cause CML (ie, a single molecular defect as a target) was important. The kinase was a “disease gene.” For success, one probably should have a target that is critical to the disease process.


A need for a validated surrogate end point that can be easily mimicked in patients to ensure that one is hitting the desired target and having the desired effects. For CML, this was the Philadelphia chromosome, which could be monitored in blood cells (and by blood counts).


There must be continued improvement in techniques important in drug designs (eg, crystallography, molecular modeling) to optimize the selectivity (specificity) of compounds.

There is no question that with improvements in our understanding of the wiring of the cell, there will be continued efforts at developing targeted agents. 48–52 However, given the terrific amount of work and substantial resources necessary to clinically develop a new agent, it is very critical that targets are selected wisely. 53

Sources of Compounds to Evaluate Against Targets

There is no question that the most rational way to develop a new agent against a target is to understand the crystal structure of the target and rationally design an agent to interact with that target. 48, 49, 53 However, for a vast majority of these targets, there is not a crystal structure identified. Fortunately, there can usually be a method developed to aid in identifying inhibitors of your targets by using molecular screening. 54 Therefore, it is necessary for a drug development team to have access to major libraries of compounds to use against their target in a high-throughput system.

The NCI has always been a key resource for libraries of compounds, including natural products. 55 Most recently, the NCI has made the NCI Diversity Library ( branches/dscb/diversity_explanation.html) available to investigators. In addition, there are a number of commercial libraries available for investigator use.

One of the most interesting new methods that has become available to find initial leads against a target is a new method developed by Kauvar and colleagues called the TRAP assay. 56, 57 TRAP stands for Target Recorded Affinity Profiling (or affinity fingerprinting). With the TRAP method, one only needs to evaluate ~200 compounds (with a “training set” of 60 compounds) to identify a lead molecule against virtually any target. The theory behind the TRAP assay is that the compounds are characterized by biologic properties (their affinities to a panel of reference proteins) rather than chemical structures. Use of the “training set” of 60 starting compounds against a target describes the target in terms of its partial similarities in protein binding patterns to the protein binding patterns of one or more of the training set compounds (their binding to the various reference proteins). If the early promise of the TRAP method is sustained, it will represent a considerable advance in rapidly finding leads against any new target because only a small “training set” of compounds needs to be evaluated to obtain a chemical lead rather than screening thousands of compounds.

By agreement with the publisher, this book is accessible by the search feature, but cannot be browsed.

Copyright © 2003, BC Decker Inc.
Bookshelf ID: NBK13502


  • Cite this Page

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...