I guess cancer stem cells as those specific cells which cause cancer tumors to grow.
The stem cell theory of cancer proposes that among all cancerous cells, a few act as stem cells that reproduce themselves and sustain the cancer, much like normal stem cells normally renew and sustain our organs and tissues. In this view, cancer cells that are not stem cells can cause problems, but they cannot sustain an attack on our bodies over the long term.
The idea that cancer is primarily driven by a smaller population of stem cells has important implications. For instance, many new anti-cancer therapies are evaluated based on their ability to shrink tumors, but if the therapies are not killing the cancer stem cells, the tumor will soon grow back (often with a vexing resistance to the previously used therapy). An analogy would be a weeding technique that is evaluated based on how low it can chop the weed stalks—but no matter how low the weeks are cut, if the roots aren't taken out, the weeds will just grow back.
Another important implication is that it is the cancer stem cells that give rise to metastases (when cancer travels from one part of the body to another) and can also act as a reservoir of cancer cells that may cause a relapse after surgery, radiation or chemotherapy has eliminated all observable signs of a cancer.
Another implication of the cancer stem cell theory is that cancer stem cells are closely related to normal stem cells and will share many of the behaviors and features of those normal stem cells. The other cancer cells produced by cancer stem cells should follow many of the rules observed by daughter cells in normal tissues. Some researchers say that cancerous cells are like a caricature of normal cells: they display many of the same features as normal tissues, but in a distorted way.
Only functional assays define cancer stem cells. For the most part "stem cell markers" shuck as CD133, CD24 CD44 etc. are relatively use less. The "gold standard" for cancer stem cells is the ability of a particular sub-population of cancer cells to be serially propagated at limiting dilutions in mice. The other cell populations can not be serially grown in mice
Cancer stem cells divide asymmetrically similar to stem cells and propagate its population. These slow cycling cancer stem cells could be identified by adding appropriate dye and after number of divisions the slow cycling cancer stem cells will have a stronger dye compared to rest of the cells.
I agree with Daniel Medina. Mentioned markers are not represented on every type of cancer stem cells, e.g., breast cancer stem cells lack CD133 (CD44+ CD24 low/-). Also, ALDH, ABC-transporters are considered as CSC markers. The tendency is "the more m. arkers the tumor cell express the more it is likely to be CSC". Additionaly to Daniel Medina comment, NOD/SCID mice are usually used for these purposes as hosts for transplantation and tumor recapitulation analysis.