Evolution of Cancer Modeling Drives Drug Discovery
Beverly A. Teicher
National Cancer Institute, NIH

he models used in anticancer drug discovery are imperfect but have, over the past 40 years, continued to improve and improve again. As the tools for cancer drug discovery continue to improve, the lessons learned from earlier models are merged into the next generation of models, steadily moving toward models that more closely recapitulate the patient and the disease. We are closing in on the cure.

The early cancer models in mice and in cell culture identified potential anticancer agents empirically searching for compounds which killed tumor cells. The mouse leukemias and solid tumors selected compounds targeting proliferating cells. The development of human cancer cell lines refined this approach by testing panels of eight or nine cell lines representing major cancer types, thus, identifying potential anticancer agents selective for major malignant diseases. Promising compounds moved on to human tumor xenograft models of those same diseases, and sometimes tumors grown from those same cells. The advent of genomic analysis of tumors to identify mutations, copy number variations, deletions and insertions in DNA has shone a light on the darkness of cancer. The history of cancer modeling is a history of steady progress, lessons learned, and ground-breaking advances in our fight against cancer.

Mouse Tumor Models

In the early 1960s, the development of mouse models of cancer, often carcinogen-induced tumors, led to an immense cancer drug discovery effort dependent upon mice bearing mouse leukemias or solid tumors such as melanoma and lung carcinoma. The leukemias were chemically induced by painting the skin of mice with the carcinogen methylcholanthrene.

Between 1965 and 1970, these mouse tumor models led to three important principles of cancer therapy that remain applicable today:

  1. Every tumor cell must be killed to achieve a cure, the corollary being that a single malignant cell can produce a tumor that can kill the host.
  2. There is a constant fractional kill (percentage) of tumor cells by a therapeutic rather than an absolute number of tumor cells.
  3. The percentage of tumor cells killed is absolutely dependent on the dose of the therapeutic.

Thousands of compounds were screened through the Cancer Chemotherapy National Service Center (CCNSC). These early efforts at rational anticancer drug discovery were successful, and several drugs discovered through this screening process are still in use today, but screening in tumor-bearing mice often led to the identification of compounds that were too toxic to be useful as drugs in cancer patients. Then, between 1985 and 1995, the development of human tumor cell lines spurred another large effort in cancer drug discovery.

The National Cancer Institute (NCI) 60 Cell Lines Screen

In 1991, the NCI-60 Human Tumor Cell Lines Screen started to provide a model for drug discovery using human tumor cells growing in cell culture that was much faster and less costly than using tumor-bearing mice. Since then, more than 200,000 compounds have been tested in the 60 human cancer cell lines screen as an empiric initial test for anticancer activity. The 60 cell lines screen continues to be very useful to drug hunters today. Medicinal chemists use response patterns in the 60 cell lines screen to guide incremental improvements in selectivity and potency of potential drug molecules.

Success in drug discovery improved with the use of human tumor cells. The 60 cell lines screen was developed at a time when tetrazolium dyes were first used as a cell culture endpoint. These dyes were a breakthrough for cell-based assays because they were fast, replacing colony-formation assays that often took a week or more. However, the limited dynamic range of tetrazolium dyes did not produce a therapeutic benefit and resulted in a focus on 50% kill values where the dyes were most accurate.

Human Tumor Xenografts and Genetically Engineered Mice

The next major development was the human tumor xenograft, an in vivo pre-clinical model of human tumor cells growing in an immunodeficient mouse. These models were first accomplished in the 1970s but didn’t become widely used until the 1990s mainly due to the cost of housing and maintaining the nude (immunodeficient) mice. These in vivo pre-clinical models of human tumor cells, which were developed from human tumor fragments or from a human tumor cell line implanted in an immunodeficient mouse, allowed the examination of human tumor cells growing in a mass. The normal tissue component of the tumor was mouse, as was the compound metabolism and critical normal tissue response.

The mutant rodent hosts’ being immunodeficient prevented rejection of the xenografted human tumor cells. The problem remained that the mouse was a robust host compared with human patients. The bone marrow of the mouse is, in some cases, 100 times less sensitive to cancer chemotherapeutic agents than human bone marrow. Thus, doses of drugs discovered using human tumor xenografts in mice often must be decreased for the drug to be tolerated by patients.

Another major limitation is the inability to test immunotherapy agents. In parallel with patient tumor xenografts in immunodeficient mice, the human tumor cloning assay was an early attempt in the 1980s to grow cells directly from human tumor biopsies in cell culture. The human tumor cloning assay was used as a model to test individual patient tumors for sensitivity to several drugs as a method to select the optimal treatment for the patient. However, the limited growth of the “fresh” human tumor cells in culture restricted the utility of the human tumor cloning assay. Genetically engineered mouse models (GEMM) of cancer are an alternative to transplantable mouse tumor models. First developed and patented in 1988, GEMMs are mice harboring one or more oncogenic mutations from birth. GEMM mice develop tumors slowly during their lifetime, frequently in an internal organ, requiring imaging or sacrifice to measure response to a therapeutic regimen. But because the tumors are produced by only a single or a few genetic alterations/mutations, many GEMM tumor models do not reflect the complexity of the human disease and the tumors are too easy to cure.

“Patient-Derived” Models

More recently, between 2005 and 2018, improvements in growing “fresh” human tumor cells in foreign environments have stimulated the cancer research community to re-evaluate both in vivo and in vitro models. These new models are referred to as patient-derived xenografts or patient-derived cell cultures. Patient-derived tumor cells are more representative of the tumor in the patient than are tumor cells that were passed from mouse to mouse, from culture to mouse, or from culture to culture for many years. In addition, patient-derived tumor cells are often grown in 3D culture, either as spheroids (3D tumor cell clusters in liquid media) or as organoids (3D tumor cell clusters in a jelled media). Advances in drug response detection using luminescence or fluorescence endpoints allow accurate determination of cell viability over logs of cell kill, which in turn achieves fast cell culture endpoints that are as accurate as the gold standard of colony formation. Patient-derived tumor cells are often maintained in media supplemented with multiple kinase inhibitors, cytokines, and growth factors (depending on the tumor’s tissue of origin), potentially selecting drug-resistant tumor subpopulations. Patient-derived models still lack an immune system, although efforts are underway to develop a “humanized” mouse that will recapitulate some human normal tissue properties and can be produced in sufficient numbers to use for drug testing.


The road toward adequate pre-clinical models for cancer drug discovery and drug combination regimens is riddled with potholes; yet drug discovery moves steadily forward, albeit at a relatively slow pace. The National Cancer Institute was involved in the process of cancer drug discovery since the earliest inception. Attacking cancer is one of the greatest humanitarian endeavors ever undertaken globally.

In parallel with the evolution of cancer models, there has been a great effort to identify and elucidate the genetic alterations in malignant diseases. As the cancer research community established cohorts of patient-derived tumors and cultures, patient-derived cells were analyzed genetically as a routine part of their characterization.

The community has learned that targeting tumors based on genetic alterations is more therapeutically beneficial than targeting them by tissue of origin. And this paves the way for further and perhaps more rapid progress in the fight against cancer.