Division of Cancer Biology Home Page Office of the Director Cancer Cell Biology Branch Cancer Etiology Branch Cancer Immunology and Hematology Branch DNA and Chromosome Aberrations Branch Structural Biology and Molecular Applications Branch Tumor Biology and Metastasis Branch Integrative Cancer Biology Program Mouse Models of Human Cancer Consortium Tumor Microenvironment Network

Think Tank Home | Tumor Immunology | The Tumor Microenvironment | Tumor Stem Cells and Self-Renewal Genes | Cell Decisions in Response to DNA Damage | Cancer Etiology | Epigenetic Mechanisms in Cancer | Inflammation and Cancer | Cancer Susceptibility and Resistance

 

Executive Summary of the Tumor Stem Cell & Self-Renewal Genes Think Tank

Executive Summary of the Tumor Stem Cell & Self-Renewal Genes Think Tank

 

 

The Tumor Stem Cell and Self-Renewal Genes Think tank was convened to identify the major scientific issues in the field and provide recommendations to the NCI to help advance research in this area.  The participants came to the following scientific conclusions:

 

 

 

 

 

 

Finally, it was recommended that the National Cancer Institute:

 

 

 

 

Introduction

 

The clonal nature of most malignant tumors is well established.  Experiments spanning several decades have shown, however, that as many as 106 murine or human tumor cells are required to transplant a new tumor from an existing one.  Two theories have been developed to account for the observation that apparently not every tumor cell is a tumor initiating cell (T-IC).   The stochastic theory predicts that every tumor cell can form an entirely new tumor; however, entry into the cell cycle is a stochastic event with low probability.  Alternatively, tumor cells may exist in a hierarchical state in which only a small number of cells possess tumor initiating potential.  If the stochastic model is correct then tumor cells are biologically homogeneous and genetic or epigenetic programs that allow for tumorigenesis are operative in the majority of cells that comprise a tumor.  The hierarchical model, however, predicts the tumor cells possess a functional heterogeneity and that quantitatively the cells capable of tumorigenesis are a relatively minor population among the bulk of tumor cells.

  

 Recent data from both hematologic malignancies and solid tumors have suggested that there are only minor populations of cells in each malignancy that are capable of tumor initiation.  These tumor initiating cells have the functional properties of a tumor stem cell.  They appear to be capable of asymmetric division and self renewal, and are only a minor faction among the bulk of more differentiated cells in the tumor.  These observations have profound implications for tumor biology research as well as successful tumor therapy.  The Tumor Stem cell Think Tank addressed several of the outstanding scientific questions that will define research in this area, as well as the needs of the research community to promote progress in this research.

 

(1) What is the current evidence for tumor initiating stem cells among tumors arising from a variety of tissues?

           

Currently, tumor stem cells have been isolated and characterized in several hematologic malignancies and two solid tumors. The critical experimental design that underlies all these studies is the development and use of a functional assay for tumor establishment and the prospective isolation of the T-IC.  Often normal tissue stem cell markers are used to identify these populations, but a functional assay such as transplantation of human leukemic stem cells into immunodeficient murine models such as the NOD/SCID mouse is most important in identifying tumor initiating cells.  One of the first tumors in which a stem cell was identified was acute myeloid leukemia (AML).  In this disease, the frequency of the leukemic stem cell (LSC) was approximately 1 per million AML blasts, establishing that not every AML cell had LSC capacity.  A CD34+, CD38- cell fraction representing 0.1-1% of the tumor cells possessed all the leukemia initiating activity in the NOD/SCID model.  By contrast, the CD34+, CD38+ cells and the CD34- cells, which comprise most of the cells in the tumor, could not initiate leukemia.  A multiple myeloma stem cell has also been characterized.  Multiple myeloma cell lines and primary patient derived cells express the cell surface marker syndecan -1 (CD138). Expression appears during the course of B-cell differentiation.  A population of cells representing <5% of the cells in the bulk population of multiple myeloma cells were found to be CD138- and possessed in vitro clonogenic potential.  These cells also engrafted successfully into NOD/SCID mice, whereas CD138+ cells did not engraft.   CD138- cells were also CD19+ and CD20+, and they expressed higher levels of KI67 (a cell proliferation antigen) than CD138+ cells.  Recently a mammary carcinoma stem cell has been isolated primarily using three cell surface markers (CD44, CD24, and epithelial specific antigen).  The tumor initiating capacity of the cells was verified in a NOD/SCID engraftment assay, and the T-ICs represented only 2% of the unfractionated cells. 

    

Finally, a putative brain tumor stem cell has also been isolated.  These cells appear to be between 0.3 - 25 % of the cells in the brain tumors examined.  They are positive for the neural stem cell marker CD133 and have a marked capacity for self renewal and differentiation.  Transplantation of these putative neural tumor stem cells into the forebrains of NOD/SCID mice yields tumors phenotypically resembling the tumors from which the stem cells were isolated.

 

The think tank participants arrived at a consensus that:

 

  1. Isolation of stem cells from more hematological malignancies and solid tumors is required to validate the general nature of the presence of stem cells in tumors.
  2. Functional assays, such as NOD/SCID mouse engraftment and single cell tumor initiating capacity assays, are necessary to establish the true stem nature of isolated cells.  Cell surface morphological markers alone are insufficient to accurately characterize stem cells.
  3. Efforts are needed to isolate prospectively and purify the T-IC. Since many tissues do not have as rich a source of cell surface markers as the blood system, effort should be expended to develop the means for T-IC purification.

 

 (2)  Is the tumor stem cell a derivative of a normal tumor stem cell or a later more differentiated progenitor cell?

        

It is unclear whether tumor stem cells arise exclusively from normal tissue stem cells, or from progenitors that have differentiated from the stem cell itself. From a theoretical standpoint, it has been well established that neoplasia arises as a consequence of the acquistion of multiple oncogenic events. Thus, the initial events must occur in cells that persist.  Hierarchical differentiation of blood cells from a stem and progenitor population has been well studied and characterized in the hematopoietic system.  It is thus not surprising that a great deal of evidence on the nature of the tumor stem cells is available from studies of hematopoietic malignancies.   In AML, the stem cell population appears to share many of the markers of the normal hematopoietic cell (e.g. CD34+, CD38-, HLA-DR-); however the leukemic stem cell appears to overexpress the IL-3 Ra subunit (CD123) relative to normal hemapoietic stem cells (HSC).  The similar cell surface properties, self-renewal properties, and complexity of stem cell hierarchies, as determined by clonal stem cell tracking of both LSC and normal stem cells, suggest that they are closely related and have been interpreted to suggest that LSC are derived from HSC.  In the blood system, only the stem cells have the long-life required to accumulate all the initiating mutations and these could be passed to the progeny that are continuously produced from such stem cells.  

 

In experimental murine systems, it has been found that the more committed myeloid progenitor cells, especially those employing the MLL oncogene, are also capable of becoming tumor stem cells.  This gene codes for a transcription factor that regulates Hox gene expression.  The Hox genes have been implicated in stem/progenitor cell expansion and self renewal.  MLL fusion proteins created by chromosomal translocations are frequently associated with acute lymphoid and myeloid leukemias.  Transduction of an MLL fusion protein into purified populations of either hematopoietic stem cells or more committed granulocyte macrophage progenitors gives rise to cells that produce a rapid AML.  Although the disease is similar with both populations, the stem cell fraction is much more potent. Thus, a more differentiated cell can also act as a tumor initiating stem cell if the genetic alteration endows this cell with self-renewal capacity.   In chronic myelogenous leukemia (CML), the t(9;22) translocation joining the BCR and ABL genes is found in hematopoeitic stem cells, but the mRNA and protein for the fusion gene are found only in later progenitrors.  In AML patients carrying the AML-1 Eto translocation in hematopoietic stem cells, the stem cell fraction is still capable of normal differentiation, but the more commited cells possess clonogenic leukemia potential only, suggesting that the AML/ETO translocation created a pre-leukemic stem cell.   Additional alterations occurred in the more committed cells to create a frank leukemia.  Thus, the tumor stem cell can arise in hematologic malignancies from multiple alterations occurring in normal stem cells, or the initial alteration could occur in the stem cell, with additional hits occurring in a more differentiated progenitor cell.  Finally, it is also possible that under some rare circumstances, the initiating event could occur in the committed progenitor to convert it into a self renewing stem cell.  It will be important to determine if the leukemogenic pathways obtained in experimental murine systems are recapitulated in the human disease.

   

Mammary tumor stem cells are CD44+, CD24- , and epithelial stem antigen positive, and this phenotype overlaps with that of epithelial stem cells.  Definitive evidence, however, that these cells arise from the normal mammary tumor cells is thus far lacking. 

  

The tumor stem cell isolated from a variety of brain tumors, including slowly proliferating astrocytomas and highly malignant medulloblastomas and glioblastomas, contain tumor stem cells that express the CD133 antigen and nestin found on normal neural tumor stem cells.  The neural tumor stem cell does not contain any of the markers characteristic of more differentiated neural cells.  Again, these data suggest that the normal neural stem cell is the precursor of the brain tumor stem cell.  This conclusion about neural tumors is supported by molecular genetic studies performed in neural cells.  Simultaneous knockout of p53 and NF-1 in neuronal precursor cells results in the development of brain tumors.  Imaging studies demonstrate that the tumors arise in two areas:  the subventricular zone of the lateral ventricle and the hippocampus.  These areas of the brain have been previously shown to be reservoirs of normal neural stem cells. Studies of the generation of neurofibromas from Schwann cells have also shown that primitive neural crest cells can become tumor stem cells. The transcription factor Krox20 participates in the differentiation of these Schwann cells, but has recently also been shown to become expressed in the primitive neural crest.  Homozygous deletion of the Krox 20 gene using a Krox 20 – Cre system causes some of the deleted cells to become neurofibromas.   It is postulated that the cells must accumulate other stochastic events (e.g. loss of NF-1) to become fibromas, because all the cells do not become transformed. 

  

 The participants concluded the following:

 

  1. Among the solid tumor stem cells identified there is an overlap between the phenotype of the normal tissue stem cell and the tumor stem cell, but definitive evidence as to whether tumor stem cells are actually transformed normal stem cells is not yet available
  2. In the hematologic and neural malignancies there is evidence that tumor stem cells may be heterogeneous.  Some may arise from the normal stem cells population, but others may arise from more differentiated progeny cells that have acquired self renewal capacity
  3. The normal tissue stem cell is a likely target for the initial carcinogenic insult, because it has the long life required to accumulate the multiple genetic or epigenetic changes required for malignant transformation.
  4. Although the events required for malignant transformation may all accumulate in the normal tissue stem cell, this cell does not necessarily become the tumor stem cell.

 

(3)  What genetic pathways may be important in maintaining the tumor stem cell state?

 

The proteins involved in self renewal in normal tissue stem cells appear to be subverted in tumorigenesis to allow the tumor initiating cells to maintain self renewal capacity.  Two families of proteins related to self renewal were considered in detail:  the poycomb gene Bmi-1 and the Wnt signaling pathway proteins.  The polycomb genes have an essential role in embryogenesis, regulation of the cell cycle and lymphopoieisis.  These genes are essential for the silencing of other families of genes.  It has been shown by RT-PCR analysis that knockout of the polycomb gene Bmi-1 in mice results in a progressive loss of all hematopoietic lineages.  This loss results from the inability of the Bmi-1 (-/-) stem cells to self renew.  Bmi-1 (-/-) cells displayed altered expression of the cell cycle inhibitor genes p16 INK4a and p19ARF, and down regulation of a gene coding for an inhibitor of apoptosis.  The p16 and p19 proteins interact with the p53/Rb regulated cell cycle pathways.   Introducing genes known to produce acute myeloid leukemia (AML) into Bmi-1(-/-) hematopoeitic stem cells (fetal liver cells) induced AML with normal kinetics.  However, the Bmi-1(-/-) leukemic stem cells from primary recipients were unable to produce AML in secondary recipients.  These results demonstrate that Bmi-1 is also required for self renewal of leukemic stem cells in AML.

           

Another group of genes involved in self renewal are those involved in the Wnt signal transduction cascade.  The Wnt protein binds to a receptor called Frizzled and activates cell fate decisions during tissue development.  It has been shown that deletion of the TCF-4 gene, a transcription factors at the end of the Wnt signal transduction cascade, causes early neonatal death in mice.  The mice lacking the gene have a single histological defect - the intestinal stem cell lining is absent.  It has also been shown that inhibitors of Wnt signaling leads to inhibition of hematopoietic stem cell growth in vitro and reduced hematopoietic reconstitution in vivo .  Activation of Wnt signaling in hematopoeitic stem cells leads to increased expression of Hox B4 and Notch-1 genes previously replicated in self renewal of hematopoietic stem cells.  The Wnt signaling pathways has been shown to be involved in both hematopoietic malignancy and colon carcinoma. 

   

Although the Wnt ligands themselves are only rarely involved in tumorigenesis, mutations mimicking Wnt receptor (Frizzled) activation induce a set of genes associated with repression of differentiation and potentiation of self renewal.  In general, these mutations involve Wnt signal transduction proteins: activation of ß- catenin and inactivation of the (APC) adenomatosis polyposis coli protein.  In myeloid leukemia, non-phosphorylated ß-catenin accumulates in granulocyte macrophage progenitors as they progress toward leukemia.  These normally more committed progenitors can thus acquire self renewal properties.  A similar accumulation of non-phosphorylated ß-catenin has also been observed in multiple myeloma cells.  In colon cancer, the APC gene is mutated early in the development of 90% of colon carcinomas.  Similarity in gene expression patterns between populations of colon cancer cells and colon epithelial stem cells has also been observed by DNA microarray analysis.  It is possible that mutations in the Wnt signaling pathway maintain the program of stem cell genes in the “on” position.

     

Two other proteins that may play a role in tumor stem cell biology are nucleostemin and the tumor suppressor PTEN.  Nucleostemin is abundant in self renewing cells such as mouse embryonic and neural stem cells as well as several human cancer cells.  Although the exact function of nucleostemin is not yet known, it behaves like a molecular switch to control cell division, perhaps through binding to p53.  Knockout of the PTEN phosphatase in prostate cancer cells allows the expression of genes associated with metastasis.   As metastatic cells are likely prostate tumor stem cells, it is likely this gene may regulate expression of stem cell related genes.

 

In conclusion, it appears that the genetic and biochemical pathways regulating the “stem” phenotype in normal stem cells are subverted to maintain the tumor stem cell phenotype. 

 

(4)  Can the tumor stem cell phenotype be epigenetically programmed or reprogrammed?

           

Human tumor cells often demonstrate abnormal patterns of DNA methylation.  DNA methylation provides an epigenetic mechanism for altering gene expression by silencing genes.  Hypermethylation frequently underlies the silencing of tumor suppressor genes.  The opposite condition, in which DNA is hypomethylated, often in concert with regional hypermethylation, has been observed in a spectrum of human tumors.   That such hypomethylation can have a fundamental effect on tumor cell development has recently become clear.  Transgenic mice that are heterozygously deleted for a DNA methyltransferase show a substantial decrease in total genomic methylation in all tissues.  At 4 to 8 months of age, these mice develop an aggressive T-cell lymphoma with a high frequency of trisomy in Chromosome 15.  There appears to be a link between DNA hypomethylation and chromosomal stability.  Chromosome 15 is frequently duplicated in T-cell lymphomas and the c-myc oncogene is located on this chromosome. c-Myc is overexpressed in many of the hypomethylated tumors and  in T-cell lymphomas as well. 

           

Further evidence that hypomethylation may have a causal role in carcinogenesis was obtained by crossing the DNA methylase heterozygote mice with mice prone to develop soft tissue sarcoma with simultaneous loss of heterozygosity (LOH) of the NF-1 and p53 genes.  The resulting transgenic mice develop sarcomas at an earlier age and demonstrate an increase in LOH in the hypomethylated versus the normally methylated cells.  The increase in the rate of LOH is the result of a specific effect of hypomethylation on the stability of pericentric and centromeric chromosomal regions. 

    

Other experiments with nuclear cloning provide evidence that the tumorigenic phenotype of tumor stem cells can be reprogrammed epigenetically.  Introducing the nucleus of a murine melanoma cell into an enucleated murine ovum with subsequent transfer into a surrogate female resulted in a normal offspring.  The neural crest derived cells including melanocytes were all normal in the resulting mouse. Thus, the micro-environment induced a genetic reprogramming. However, the genetic mechanisms that existed in the tumor nuclei were maintained as these mice had a high incidence of neoplasia.  Thus, both genetic and epigenetic mechanisms may be active in neoplastic development. This result with melanoma was a rare experimental success among a number of tumor nuclei tested, and the reasons for the lack of repeated success with nuclei from other tumors is under investigation. 

           

That the microenvironmental niche that a cell resides in can have profound effects on the cellular phenotype is also demonstrated by studies on the fate of embryonic stem cells derived from the blastocyst inner cell mass.   When they are transplanted back into another embryo, they maintain their normal phenotype. If, however, they are transplanted into a differentiated microenviroment such as kidney or liver, the cells form a teratoma.

  

In conclusion, there is evidence to suggest that epigenetic mechanisms can produce an inheritable tumor stem cell phenotype. Epigenetic control also underlies a great amount of stem cell regulation as exemplified by the polycomb gene Bmi-1, thus providing a link to the control of cancer stem cells.

 

Future Directions:

     At the conclusion of the Think Tank, a number of important future basic research questions were synthesized from the scientific discussion.  They are summarized below:

 

1)      What governs the rate of proliferation of stem cells and in particular tumor stem cells?  A corollary question would be whether the size or quality of the stem cell microenvironment (niche) acts as a constraint on stem cell growth?

2)      What creates the stem cell niche for a tumor stem cell (i.e. do tumor stromal cells constitute the tumor stem cell niche?)

3)      Does the transition from semi-linear to exponential cell growth occur when the tumor cells become independent of the stromal niche?

4)      Can oncogenes and their associated mutations affect asymmetric versus symmetric divisions in stem cells?

5)      Is the object of Darwinian selection in the tumor the T-IC, rather than the more differentiated tumor cells?   It is likely that alleles for malignancy spread in a tissue because they arise in stem cells and provide advantages to the tumor stem cells that carry them.

6)      Are the phenotypes of invasion and metastasis uniquely connected to the tumor stem phenotype? 

7)      Stem cell quiescence versus growth and differentiation must ultimately be understood in terms of progression through the cell cycle.  It will be important to determine whether the retinoblastoma (Rb) gene product is as critical in this process as it appears.  Rb plays a key role in cell cycle progression and differentiation in a number of tissues.  Hypophosphorylation of Rb forces cells to leave the cell cycle and enter Go, therefore regulation of this protein is likely to play a role in regulating true stem cell state.  This is substantiated by studies on the Bmi-1 gene and the genes it regulate that interact with the Rb/p53 cell cycle pathway.

8)      Most human carcinogens are strong tumor promoters and weak initiators of carcinogenesis.  If tumor promoters work by increasing the size of the target population, then they must work by increasing the population of already initiated cells.  Thus, it is important to understand whether tumor promoters work on initiated epithelial stem cells or on stromal cells, as the stromal cells may control the size of the stem cell niche.

9)      Can the roles of mutation and epigenetic mechanisms be distinguished in the generation of the tumor stem cell phenotype?

 

In addition a number of questions related to the future of cancer therapy were also considered.

 

1)      Can the current benchmark for measuring the success of cancer therapy, tumor shrinkage, be changed to something more biologically relevant?   The success of therapy can only really be measured by understanding the effect of the therapy on the tumor stem cell.  Unfortunately this is difficult to obtain because of the lack of tumor stem cell markers.

2)      Can we develop xenograft models that recapitulate the stem cell transition to more differentiated progenitors in a tumor?  Such models might be useful to begin understanding the effects of therapy on tumor stem cells.

3)      Are there methods for treating tumors that might cause a collapse of the stem cell niche?  It is likely that most human tumors depend on stromal cells that define the niche and can control the size of the stem cell niche.

4)      Are the cells that form the tumor stem cell niche different from those forming the normal tissue stem cell niche?  Might any differences present an opportunity for directing selective therapy to tumor stromal cells?

5)      Finally, might tumor stem cells be more or less sensitive to apoptosis inducing stimuli?  Such information is critical in designing therapies that destroy the tumor stem cells responsible for continued expansion of the tumor and subsequent metastasis.

 

 

 

Specific Recommendations for the NCI:

 

Technical Recommendations:

 

1) Develop methods to encourage surgeons involved in clinical research to collect tumor specimens directly into trypsin disaggregation solutions and provide viable freezing of cells in large banks to allow scientists to isolate unselected populations of tumor stem cells.

 

2) Encourage the development of real time PCR technology to examine gene expression in tumor stem cells.  This technique appears to give more valuable information than DNA microarray technology for analysis of stem cells.

 

3) Improve in vivo and in vitro functional assays for tumor stem cells to allow for more accurate identification of these cells in concert with cell surface phenotype identification.

 

4) Improve in vivo organotypic assays to understand symmetric versus asymmetric cell division.

 

Administrative Recommendations:

 

1) Continue National Cancer Institute support for opportunistic basic research into tumor stem cell biology.

 

2) Develop a Research Consortium to facilitate transdisciplinary approaches and to provide specialized research reagents that will advance research in this area.