Functional validation of human cancer stem cells

by Alexey Bersenev on October 21, 2011 · 0 comments

in cancer, hematopoietic

Post to Twitter Send Gmail Post to LinkedIn

This is re-blog from

The cancer stem cell concept is evolving and causing great debate in the scientific and medical world. People are discussing these three most important questions:

  1. Do cancer stem cells (CSC) exist and is the concept valid?
  2. How can we identify and characterize them?
  3. What is the therapeutic relevance (for prognosis and therapeutic targeting)?

In order to estimate prognostic and therapeutic value, CSC should be isolated at a pure level or enriched. Today, almost all protocols for CSC isolation or enrichment rely on surface markers expression. This approach works nicely for normal stem cells, but seem to be seriously flawed for CSC, due to their intrinsic plasticity.

Despite solid experimental evidence, and questionable validity of cancer stem cell (CSC) surface markers, researchers keep publishing tremendous number of papers based on it. As you may know, I’m advocating a functional approach for cancer stem cell identification for some time. Recently, I was happy to find the study from John Dick’s lab, which re-evaluate prognostic value of leukemic stem cells (LSC), based on functional assessment. I think, this study is a breaking point! I hope it will give a start for the new era of accurate cancer stem cell identification and characterization.

any investigation of CSC properties using phenotypically defined populations is unreliable and requires functional confirmation of CSC activity using well-validated tumor- or leukemia-initiation assays.

Because CSC populations from acute myeloid leukemia (AML) have diverse surface marker profile, the authors validated cancer-initiating population in transplantation assay first. After confirmation of biological properties (leukemia-initiating function), cell populations from the same sample were sorted for following gene expression analysis.

The authors use functional validation for evaluation of clinical prognostic value of CSC gene signatures. For gene expression analysis, they used 25 sorted cell populations enriched for functionally validated CSC (called LSC-R) versus 26 cell populations without detectable CSC. Similar to previous studies, they found that LSC-R or HSC-R (normal hematopoietic stem cell control) gene signatures correlated with poor prognosis in AML. The functional approach makes this study very different technically and advantageous compared with previous published work.

To illustrate the utility of genetic signatures in CSC field I’d quote the commentary:

… quantitative assessment of self-renewal signature via, perhaps, a minimal gene set customized for different cancer subtypes can provide a rapid measure of stem cell activity and a prognostic indicator guiding choice of therapy. A predictive expression signature that bypasses the need for in vivo CSC assay would be very advantageous. However, it will be necessary to first validate the independently predictive value of such assays for different types of cancers…

I hope, in the future, carefully assessed CSC gene expression signatures will allow design of “leukemia prognosis chip” and “leukemia therapeutic chip” and will bring us closer to personalized oncology.

The most intriguing part of this study for me was a comparison of leukemic stem cell gene expression profiles across of all published studies. As we can predict, gene profiles of validated AML stem cells (LSC) did not overlap with “non-validated LSC” studies:

In one latter case the LSC nature of each fraction was not functionally validated and, as we and others have shown, the use of CD34 and CD38 to identify stem cell fractions without concomitant functional analysis can misidentify the stem cell nature of sorted cell fractions.

we found no correlation with the LSC list generated by Gal et al. based upon phenotypically defined populations (AML CD34+/CD38- vs CD34+/CD38+ cells).

In a second study of LSC to non-LSC AML populations, Ishikawa et al. used a cohort of 4 samples with 2 populations each to identify a small number of genes.
…the LSC-R does not positively correlate with their LSC up-regulated gene set nor does HSC-R negatively correlate with their down-regulated LSC gene set. This suggests that while this study was successful in identifying some LSC-related stem cell genes, it was limited by small sample size and the gene expression variability inherent in cancer samples.

Unfortunately, the authors didn’t perform side by side comparison with recent (“non-validated”) Gentles et al. study. Also, we still don’t know, how gene expression profile of validated CSC would be different at time of diagnosis and at relapse.


  1. I think cancer stem cell researchers should finally accept that using only surface markers for CSC isolation and enrichment is unreliable.
  2. Before all “therapeutic studies”, isolated CSC should be validated in functional assays (tumorigenicity in xenotransplant, sphere formation assays…).
  3. Findings from previous studies used “not validated” CSC should be questioned and critically re-evaluated.
  4. Researchers entering the field should realize and appreciate great complexity (plasticity and heterogeneity) of cancer stem cells.

I’d like to finish by a quote:

… the approach we have taken in AML provides a paradigm for assessing both the identity and clinical relevance of LSCs and CSCs from other leukemias and solid tumors, respectively. A well-validated and sensitive xenograft assay is essential because only functionally validated populations showed clinical relevance, whereas signatures derived from phenotypically defined populations did not.

{ 0 comments… add one now }

Leave a Comment

Previous post:

Next post: