As part of this research, Dr. Morrison’s lab decided to optimize the off-the-shelf assay to see whether changes could detect a broader range of cancer cells. “In addition to going longer and detecting a 10-fold increase in tumorigenic cells, if we use more highly immunocompromised mice we get almost a 200-fold increase in the detectable frequency of tumorigenic cells. Co-injection with Matrigel™ (which generally improved cell engraftment and thus increases cell survival without conferring tumorigenicity) yielded up to a 20-fold increase.
“Modifications in xenograft assays combine to dramatically increase the detectable frequency of tumorigenic cells, so that they were actually quite common,” Dr. Morrison said. In a larger study, initial data indicated that one in one million cells has the potential to form a tumor but, by slightly modifying the assay, the ratio changed to one in four.
“Therefore, for the first time in cancer biology, we can study tumorigenesis from single human cancer cells in vivo.” Rather than proving that metastatic melanoma doesn’t follow the cancer stem cell model, it proves that the tumorigenic cells are frequent.
“A separate question is whether the 28 percent of cells in which we can detect tumorigenic capacity are intrinsically different from the nontumorigenic cells that we cannot detect,” Dr. Morrison continued. In his experiments, involving 50 markers, in which 17 that are frequently heterogeneously expressed, the positive and negative fractions resulting in tumorigenesis. “We are unable to detect a hierarchical organization in melanoma, having looked very hard.” If such a marker is eventually found, he suggests it will have a very shallow hierarchy.
The second portion of the workshop, led by Franziska Michor, Ph.D., computational biologist at Memorial Sloan-Kettering, addressed therapeutic resistance and the origin of cancer stem cells.
“If you have a drug that kills tumor stem cells, then over time, the tumor stem cell population will die out. If you have drugs that kill tumor cells but leave the cancer stem cells intact, then the tumor cell population will grow back,” she hypothesized.
In investigating the molecular response to imatinib among 68 patients, Dr. Michor found two distinct response slopes—one at 20 days life span and another at 125 days life span during therapy—with two distinct decay rates. The main body of cancer cells was depleted at a rate of 5% per day, while the other cells were depleted by 0.8% per day. This dramatic difference indicates that there are distinct subpopulations of cancer cells.
The idea that distinct subpopulations of cancer cells exists is further supported by a German study of patients who discontinued imatinib therapy after three years. Within one week, the cancer cells rapidly rebounded in 60% of the patients, peaking at levels higher than baseline. Therefore, it appears that cancer stem cells were not depleted by the drug and are, in fact, driving the disease. “There’s something going on that leads to resistance to imatinib,” Dr. Michor said.
Because imatinib is so specific as to be useless if a single base changes, it may be possible to answer that question while exploring the origin of tumorigenic cells. To do so, she turned to a mathematical analysis of the evolution of cancer stem cells, focusing upon identifying the mutation that triggered their drug resistance and considered several theories: that tissue-specific stem cells accumulated all the mutations needed to transform into tumor cells, that progenitors also accumulated the necessary mutations, and that the mutation conferring self-renewal to progenitors first arises in the stem cells without changing their phenotype.
Based upon her work with JAK2V617F mutations, she determined that progenitor cells are the most likely cell of origin for those mutations, which lead to cancer. This finding also may be relevant to other tumor types in tissues organized with a differentiation hierarchy.
Her next step was to identify the predicted first mutation. Using primary glioblastomas, Dr. Michor investigated cells in the subventricular zone of the brain that targets for transfection. Using mathematical modeling of a small number of cell divisions, she found that, “self-renewing transit amplifying cells are the most likely cell of origin for gliomas.”
In general discussions, Dr. Michor broached the idea that “some cancers really seem to follow the cancer stem cell model but are reversibly organized.” The role of microenvironments also was touched upon and should be the subject of additional research. “Tumor microenvironments are very important, but we don’t know how,” Dr. Morrison said. Microenvironments appear to kick off morphogenetic processes, so the markers that often are used become moving targets that, ultimately, are of little value, Dr. Michor added.
There are a lot of fundamental questions that remain about cancer stem cells, Dr. Morrison emphasized. “The way we think about cancer stem cells will change dramatically during the next five years. Cancer is so endlessly resourceful.”