I was very much enjoying reading a new paper, published in Cancer Cell, which clarifies issues from the first report by Morrison’s group (Quintana, et al. Nature, 2008) and explains some assay-based controversies in the field of cancer stem cells. I’ve asked senior first co-author of the paper Mark Shackleton to discuss some of questions via email.
Mark Shackleton is a Group Leader of Melanoma Research Laboratory at the Peter Mac Cancer Center, Melbourne, Australia. He completed his postdoctoral training in Sean Morrison’s lab where studied melanoma cancer stem cells.
1. Dear Dr. Shackleton,
In the recent paper, which was published last week in Cancer Cell, you proposed the new model of heterogeneity in carcinogenesis, so-called “phenotypic plasticity”. This model says that some cancers (melanoma in your case) are driven by reversible phenotypic changes – switching markers on and off all the time as the tumor is progressing. As you noted in the paper this model contrasts with both the cancer stem cell concept and clonal evolution model. What do you think about genotype of the tumor and epigenetics? Could plasticity apply for those and could one tumor combine all 3 mentioned mechanisms of development?
Cellular heterogeneity within tumors could derive from reversible or irreversible mechanisms. One irreversible mechanism is the emergence via clonal evolution of genetic differences between some cells. Another irreversible mechanism is the acquisition of stable, non-genetic differences between cells. This is what happens in the cancer stem cell (CSC) model of cancer progression as this model was originally conceived (Nature 414:105). In the CSC model, non/less-tumorigenic cells in tumors derive from highly/more tumorigenic cells in a non-genetically determined manner that is associated with stable phenotypic differences between more and less tumorigenic cells. Conceptually, this is similar to normal cell ontogeny in many organs, in which downstream progenitors and differentiated cells derive from tissue-specific stem cells by acquiring stable ‘epigenetic’ changes that result in (usually) reduced proliferative potential and increased lineage restriction. The plasticity model refers to reversible differences between cancer cells that would have to be determined by non-genetic factors.
To summarize which mechanisms drive cellular heterogeneity in tumors in which model:
Clonal evolution – genetic’
CSCs – irreversible ‘epigenetic’
Plasticity – reversible ‘epigenetic’
There is general agreement on and indeed evidence to support the notion that these models are not mutually exclusive and that all mechanisms could occur in the one tumor. Our most recent study of melanoma suggests that reversible factors determine much of the phenotypic heterogeneity present in this cancer (Cancer Cell 18:510). However, there is good evidence that clonal evolution also occurs in melanoma. Unlike others, we have found no evidence that melanoma follows a CSC model, although of course it remains possible that in the future markers will be found in optimized assays to distinguish cells with more and less tumorigenic potential in this disease.
2. You noted that even though you used exactly the same antibody, as Boiko et al, in the recent 2010 study, transplanted the same kind of cells (primary from stage III melanoma or xenografted), both groups got different results. It amazes me that using slightly different immunocompromised mouse strains, site of injection and cell suspension preparation we can see 10,000 times difference in sensitivity of assay. Whose assay is the right one and how to solve this controversy? I wonder why different research groups are not willing to run the same protocol in order to find the truth.
It also amazed me that seemingly minor differences in assay conditions could so dramatically influence the underlying disease biology that is revealed. There clearly can be compounding of multiple smallish assay differences that when combined produce large effects.
As for who is right, I don’t think this is for me to say as I will likely be accused of bias! It is up to the melanoma biology field generally to sort out who is right by critically evaluating the data and replicating results. I remain strongly persuaded by the idea that if a tumorigenesis assay seeks to reveal tumorigenic potential in cells, then it should be as sensitive as possible for this purpose. The CSC model has only ever addressed the potential of cancer cells for tumor formation. In the CSC model, non-tumorigenic cells have lost tumorigenic potential and by definition it should not be possible to induce tumor formation from these cells, irrespective of the assay. Therefore, if cells form tumors in one assay, these cells have tumorigenic potential, even if the same cells do not form tumors in another assay.
In my opinion, it is up to proponents of the idea that melanoma follows a hierarchical CSC model to demonstrate this experimentally in a tumorigenesis assay that is at least as efficient as the one that Elsa Quintana and I developed in Sean Morrison’s lab. It is pointless to make comparisons of data from different groups when the groups use assays that are not equally sensitive at revealing tumorigenic potential.
3. What lessons can young scientists who are studying cancer stem cells learn from these controversies in the field? If cancer stem cell testing in vivo is so heavily assay-dependent, what assay should they choose?
I would recommend to a young scientist embarking on a cancer stem cell project – indeed any scientific project – to be keenly critical of the literature and to rely as little as possible on what others have published until results are confirmed in their own work. For example, in many instances I do not think it is wise to rely on published work to define CSC subpopulations based only on marker expression. Tumorigenic cells in the CSC model are primarily defined by functional assays and it is risky to assume functionality from phenotype in every setting. Even in the hematopoiesis field, in which markers of hematopoietic stem cells are well defined, functional studies are almost always required to identify and characterize these cells. Stringent functional criteria should also apply to identifying and characterizing CSCs. Furthermore, there are several instances in which putative CSC markers have been found not to be generally applicable across multiple cancers of a particular tumor type. As a reviewer of numerous cancer stem cell papers, I take a dim view of conclusions that are based on evaluations of CSCs defined only by phenotype.
The question of which assay to use is critical. As above, I believe that the best assay is the one which most sensitively reveals the malignant potential that is present in a cell – without, of course, inducing such potential. It is very likely that different assay conditions will be optimal for different types – and even subtypes – of cancer. For melanoma, highly permissive assay conditions appear to be relatively simple. However, for other cancer cell types multiple additional factors may be required to support the survival and allow the growth of those cells. At the outset of undertaking a CSC project, I would advise young scientists to invest heavily in tumorigenesis assay optimization.
4. You spend a significant amount of time doing xenotransplantation assays of primary tumor cells obtained from patients. What are your tips and advice for those researchers who are testing tumorigeneity in xenotransplant models? What difficulties do they can face in those kind of assays? What are general flaws of the xenotransplantation assay in cancer biology?
There are numerous factors that need to be considered for the optimization of tumorigenesis assays.
First, it is important that fresh tissue arrives from the surgical theatre to the lab as quickly as possible. Arranging this is no small logistical exercise, requiring co-operative surgeons, pathologists and tissue banking staff.
Second, for solid cancer studies, the process of obtaining a single cell suspension from a solid tumor requires optimization in order to derive the highest number of viable cancer cells. Too little tumor digestion may improve cell viability at the cost of losing undigested pieces of tumor; too much digestion may eliminate residual clumps of undigested tissue but render cell non-viable due to over-exposure to the digestion process. Each aspect of the digestion method should be carefully examined to determine whether there are changes that could improve the yield of viable cells, and these changes should be compared side-by-side with whatever is the ‘standard’ method at that time.
Third, effort should be made to reduce if necessary contamination by dead cells and debris in the cell suspension. In general, if a sample contains >30% dead cells or debris (not uncommon in digests of tumors that contain regions of spontaneous necrosis), we spin out the dead cells by by density centrifugation.
Fourth, assuming flow cytometry is used to purify cancer cells (we have found flow cytometry generally superior to bead separation in terms of cell yield and purity), gates should be carefully chosen to exclude debris and non-cancer cells. Gate-setting can be challenging and if there is doubt about which populations to select and which populations to exclude, then all populations should be sorted, visualized, stained for cancer cell markers and tested functionally to see whether they contain cancer cells.
Finally, sorted cells should be counted, aliquoted, prepared and transplanted into the recipient model as quickly as possible after sorting. Ideally the proportion of non-cancer cells in the sorted fraction should be determined by marker analysis.
Depending on the scientific question you are asking, xenotransplantation of cancer cells could be a deeply flawed approach or well-suited to answering the question. There are obviously numerous ways in which the microenvironment of human cancer cells in immunocompromised mice is different to their microenvironment in native tumors in patients. Many of these differences will I doubt ever be able to be overcome, even in mice that are ‘humanized’ to some degree, and may always cloud interpretations of disease biology to different extents in different cancers. However, our data indicate that xenotransplantation is a sensitive way to reveal intrinsic tumorigenic potential in human melanoma cells. As we have been primarily interested in evaluating tumorigenic potential in melanoma cells, xenotransplantation has been suitable method for us to use.
We wish you great success and luck in your new laboratory! Thank you for the interview!