May 1, 2008 (Vol. 28, No. 9)

Many Issues Still Must Be Addressed before Betting the Farm on CSCs

GlaxoSmithKline (GSK) inked a $1.4 billion deal with OncoMed to eliminate cancer stem cells (CSCs) with mAbs, an investment based on the belief that cancer stem cells are the source of all solid tumors. Scott Kern, M.D., an expert on pancreatic cancers from the Johns Hopkins University, however, disputes the very existence of CSCs and argues that a belief in them is more akin to religion than science. In stark contrast, Robert Weinberg, Ph.D., professor of biology at MIT, says that CSCs are “mind blowing” and that “the entire mindset of people must now be refocused onto these stem cells.” He further argues that wiping out cancer stem cells will cure the disease.

George L. Gabor Miklos, Ph.D.

Given such radically divergent opinions, one should ask: What are the clinical facts? Let’s start with, what are CSCs? According to Dr. Weinberg, “A cancer stem cell is defined as a cell that when plucked out of the tumor and introduced into a new host like a mouse is able to spawn an entirely new tumor.” Thus, when human cells are taken from resected cancerous tissue and transplanted into immunocompromised mice, only a small percentage of such cells seed new solid tumors.

Phillip J. Baird, M.D., Ph.D.

One viewpoint is that these tumors are initiated by stem cells that have become abnormal through mutations or derive from cells that have reacquired embryonic gene-expression programs via mutations and returned to a stem-cell state. A critical requirement for true believers is to demonstrate in patients that tumors derive from bona fide stem cells. If this essential condition is not met, then we are left with little more than cancer-initiating cells (CICs) defined by growth in a mouse. They have the aura of stemness but not the substance.

Ian E. Haines, M.D.
Existing data reveals that the majority of samples taken from head and neck tumors of patients do not seed new growths in mice. Therefore some tumors contain neither cancer stem cells nor cancer-initiating cells. Furthermore, the cytogenetic data from transplants reveals cells with abnormal genomes having losses or gains of whole chromosomes or parts of chromosomes (aneuploid and segmentally aneuploid cells) and transplants from different patients exhibit different abnormal genomes; no two genomes are the same.

Unanswered Questions

This raises some questions. Are the cells defined by mouse transplantation the same cells that leave the primary tumor, metastasize, and cause the death of the patient, or are they simply the survivors in immunocompromised mice? Are the dangerous cells that disseminate and metastasize in patients derived from bona fide stem cells embedded in different tissues or derived more simply from differentiated cells within a tumor? Does drug resistance after chemotherapy result from an inherent property of CSCs or from selection on a genomically heterogeneous nonstem-cell population challenged with drugs?

The inferred therapeutic value of CSCs stands or falls on data from patients, not from mice. In this context, four examples of related research are provided.

Over four decades ago, researchers (Southam, C. M. and Brunschwig, A., 1961, Cancer 14, 971) transplanted cancerous cells (derived from melanoma, malignant ascites, and ovarian, rectal, and cervical carcinoma in terminally ill cancer patients) to a subcutaneous location in either the thigh or the forearm of the same patient. Of the 27 patients followed during the study, only five had evidence of tumor growth at the new site. Thus most tumors contained neither CSCs nor CICs.

In the second example, to relieve pain in cancer patients with metastases from the breast, ovaries, and lungs, researchers treated patients with a peritoneovenous shunt, allowing cancer cells to move freely from the peritoneal cavity to the jugular vein, hence providing access to the systemic circulation (Tarin, D. et al., 1984, Cancer Research 44, 3584; Tarin, D., 2006-2007, Breast Disease 26, 13).

Although billions of viable cancer cells per week poured into the circulation of some patients, the subsequent histopathological data garnered from autopsies revealed that most patients had no new metastases. Importantly, the level of variation among individual patients was such that even when tumors were from the same organ and had similar histology, they metastasized in some patients but not in others. Hence, if there are CSCs in the same organ of different patients, they differ enormously in their metastatic potential.

Third, Christoph Klein and colleagues analyzed single disseminated cells in bone marrow and lymph nodes from the same patient after curative resection of the primary breast cancer (Klein, C. et al., 2002, The Lancet 360, 683). They not only found that these disseminated cells contained abnormal genomes but also that the genomes in the cells that had lodged in the bone marrow were different from those lodged in the lymph nodes. No two disseminated cells shared the same chromosomal aberrations. If these cells are CSCs, then CSCs exhibit extreme genomic heterogeneity within the same patient.

Cervical Cancer

The fourth example involves cervical cancer data, which is especially informative because early precancerous abnormalities are available for examination from hundreds of millions of PAP smears. No other cancer can be detected so readily at its inception, since most cancers such as brain, breast, and prostate remain inaccessible to viewing in the earliest stages.

In the cervix, the first microscopically visible abnormalities are tetraploid cells, presumably arising from frequent failures of cytokinesis, spontaneous cell fusion, or the well-described HPV-induced tetraploidy (Olaharski, A. et al., 2006, Carcinogenesis 27, 337; Spriggs, A. et al., 1962, The Lancet 1, 1383). The importance of tetraploidy as a generator of diversity cannot be emphasized enough (Fujiwara, T., 2005 Nature 437, 1043; Shi, Q. and King, R., 2005, Nature 437, 1038), since irregular chromosomal segregation in tetraploid cells and their progeny generates cell populations with scrambled, unbalanced, and somatically mutated genomes with enormous numbers of genomic and epigenomic combinations. This variation is a substrate for dissemination, metastasis, and drug resistance, an avenue experimentally pioneered by Peter Duesberg, Ph.D., at the University of California, Berkeley (Duesberg, P., 2007, Scientific American 296, 52).

Genomic variation can also derive from processes as commonplace as cell fusion. Macrophages can fuse with normal or cancer cells, and some hybrids and their derivatives may thus retain the macrophage’s innate property of migrating throughout the body and infiltrating all tissues, an avenue evaluated by numerous researchers (Pawelek, J., 2005, Lancet Oncology 6, 988; Vignery, A., 2005, Trends in Cell Biology 15, 188; Duelli, D. and Lazebnik, Y., 2003, Cancer Cell 3, 445). Derivatives of cell fusion can yield varied transcriptomes that may upregulate some stem-cell markers by default, providing the appearance of derivation from a stem cell.


How then does such clinical and experimental data benefit cancer patients? Unfortunately, despite the hype and the aggressive marketing, there are no therapeutic benefits in the short term. The oncologist’s immediate concern is to help patients live longer and feel as good as possible during treatment. Thus, the primary tumor is resected or treated with radiation, and if cells have not disseminated, then the patient is fully cured. Since the extent of dissemination is generally unknown at the time of surgery, conventional radiotherapy, chemotherapy, and new drug treatments inexorably follow.

In the long term, attempting to eliminate CSCs with mAbs as GSK and OncoMed propose requires initially describing exactly which cells to target in patients. From a clinical perspective, it needs to be established whether the dangerous cells that disseminate in patients equate to the CSCs defined by transplantation in the mouse, or whether disseminating cells simply derive from a heterogeneous population of cells that are the common outcomes of failures in cytokinesis, cell fusion, or virally induced processes.

Until this is resolved, neither CSC researchers nor GSK will know whether they are pursuing clinically relevant cells. Since disseminated cells have very different genomes in each cancer patient, standard drug target specificity is still in the realm of voodoo and handwaving, and needs new perspectives.

As demonstrated experimentally by Dr. Duesberg, when aneuploid cell populations were subjected to various drugs, they rapidly produced drug-resistant cells, whereas diploid cell populations could not achieve this rapid response because they lacked the requisite genomic diversity. The question that then arises is: Does the drug resistance that develops in patients after chemotherapy or radiotherapy derive from CSCs defined by mouse transplantation or from selection for resistance on genomically heterogeneous cell populations in the primary and metastatic growths of the patient?

Until the molecular properties of cancer stem cells and cancer-initiating cells are rigorously evaluated in patients, instead of in mice and tissue culture systems, early detection remains the best hope for patients. Early detection has been successful in identifying precancerous lesions of the cervix, for example. When detection happens later, informed decisions about drugs and their dosages must be based on sound clinical data.

Previous articleInvestigators Generate Sequence-Based Map of Structural Variations
Next articleTorreyPines Licenses Glutamate-Receptor IP from Johns Hopkins