|Home | About | Journals | Submit | Contact Us | Français|
The cancer stem cell (CSC) hypothesis suggests that clonogenic growth potential within an individual tumor is restricted to a specific and phenotypically defined cell population. Evidence for CSC in human tumors initially arose from studies of AML, but functionally similar cell populations have been identified in an increasing number of malignancies. Despite these findings, controversy surrounds the CSC hypothesis, especially the generalization that clonogenic tumor cells are rare. Nevertheless, efforts to define the cellular processes regulating self-renewal and resistance to anticancer therapeutics, two of the major properties ascribed to CSC, are likely to provide useful insights into tumor biology as a whole.BMT has been at the forefront of clinically translating basic stem cell concepts starting with the original hypothesis that normal hematopoietic precursors could rescue patients from myeloablative doses of radiation or chemotherapy.Even today, a better understanding of CSC may enhance ongoing efforts to induce specific and effective anti-tumor immune responses in both the allogeneic and autologous setting.It is also likely that new clinical research approaches will be required to accurately evaluate novel CSC-targeting strategies.Owing to the capacity to produce remissions in most diseases, SCT may provide the ideal clinical platform to carry out these investigations by studying the ability of anti-CSC agents to prolong relapse free and overall survival.
Two recurrent observations serve as foundations for the cancer stem cell (CSC) hypothesis. The first is the necessity of large cell numbers to reliably produce ectopic tumor growth either in vitro or in vivo. During studies examining the relationship between in vivo growth kinetics and the activity of early chemotherapeutic agents, only a minority of cells within mouse models of lymphoma (AKR) and myeloma (Adj. PC5) were found to be capable of clonogenic growth.1,2 These data were in stark contrast to the well established L1210 leukemia cell line, in which a single cell could reliably produce lethal tumor growth.3 Similar studies in terminal cancer patients showed that large numbers of cells derived from autologous primary or metastatic lesions were required to form tumors at surgically defined s.c. or i.m. sites.4 These studies suggested that tumor cells were functionally heterogeneous with different clonogenic growth potentials in vivo, and similar results were later described in studies examining in vitro clonogenic growth. In the Adj. PC5 mouse model of myeloma, only a fraction of plated cells could give rise to tumor colonies.5 Furthermore, the development of an in vitro assay capable of supporting colony formation by primary human tumors also found that only a minority of cells were clonogenic.6
The second finding that gave rise to the CSC hypothesis is that human tumors may recapitulate many of the morphologic and phenotypic features of the tissue of origin. In CML, the Ph chromosome was detected as a recurrent chromosomal abnormality and suggested that all tumor cells were derived from a single cell.7 Later studies examining patterns of X chromosome inactivation found that CML cells, regardless of their degree of myeloid differentiation, were clonally related.8 Similarly, it is well recognized that many solid tumors histologically resemble the normal tissue from which they arise. These findings suggested that cancers retain some capacity for cellular differentiation along the normal lineage and that individual tumors are composed of phenotypically heterogeneous cell types. Along with findings of infrequent clonogenic growth potential, these observations suggested that the functional capacity of individual tumor cells (that is, their growth potential) was linked to their phenotype. During this time, the existence of normal hematopoietic stem cells that were able to produce the entire spectrum of normal blood cells was shown.9 As a synthesis of these ideas, the CSC hypothesis was generated and suggested that cancers were also organized in a hierarchical manner with primitive CSC generating progeny with some capacity, albeit abnormal, for cellular differentiation.
Full proof of the CSC hypothesis required the development of reliable assays to measure clonogenic growth capacity. Severely immunodeficient SCID and non-obese diabetic (NOD)/SCID mice were found to be capable of supporting multilineage human hematopoiesis and provided an in vivo platform to study the clonogenic potential of normal hematopoietic stem cells.10,11 Moreover, the ability to examine continued human blood cell formation during serial rounds of transplantation allowed in vivo assessment of self-renewal capacity. These methods allowed the clonogenic potential of AML specimens to be examined and provided the first experimental evidence for human CSC. These studies showed that bulk leukemia specimens could reliably engraft NOD/SCID mice, but only when large cell numbers were injected.12,13 Distinct phenotypic cell populations resembling normal hematopoietic stem cells and progenitors can be detected in most clinical AML specimens regardless of their morphologic FAB or WHO (World Health Organization) sub-classification. Small numbers of these primitive cells expressing CD34, but lacking the differentiation marker CD38, could give rise to human AML. These studies suggested many of the properties that have been proposed to define tumor-initiating cells. For example, although cells with clonogenic growth potential were relatively primitive, the resulting blasts seemed to be relatively differentiated and closely resembled the original tumor. Additionally, clonogenic tumor cells could be subsequently transplanted into secondary recipients with similar engraftment and growth potential. These two properties, namely the capacity to phenotypically recapitulate the original tumor and undergo self-renewal as shown by serial transplantation, have become the defining functional properties of CSC. Using these criteria, clonogenic cell populations have been described in an increasing number of human cancers.14,15
Findings regarding CSC in AML and other human cancers have not been without controversy. A recent report studying genetically engineered mouse models of myeloid, B-cell and T-cell tumors showed that tumor growth in syngeneic recipients could be achieved using low numbers of unsorted (that is, bulk) cells.16 These results suggest that the low clonogenic frequency observed in human cancers is, in part, due to the reliance on immunodeficient mice as the primary in vivo method to study tumor growth and self-renewal, and that xenotransplantation barriers may exist even in NOD/SCID mice, which prevent the engraftment of all cells. In light of the pioneering studies carried out by Skipper and colleagues3 showing that a single L1210 cell can produce lethal leukemia as well as a number of other highly aggressive mouse tumors, in which tumor growth in syngeneic recipients is highly efficient, these findings were not necessarily novel or surprising. Although the relative infrequency of clonogenic tumor cells initially contributed to the development of the CSC hypothesis, much of this data was derived from tumors that displayed a range of phenotypic differentiation. The aforementioned studies did not fully examine the second concept giving rise to the CSC hypothesis, namely restriction of clonogenic tumor growth to specific phenotypic populations, and the tumors studied seemed to be phenotypically homogeneous like L1210 cells. Reports ascribing a specific phenotype to cells capable of producing tumor growth in various immunodeficient models may also be in disagreement. In several tumors, such as ALL and colorectal and pancreatic carcinomas, discordant cell phenotypes have been reported to enrich clonogenic cells.17-24 In AML, the cell surface phenotype originally ascribed to clonogenic leukemic cells (CD34+ CD38neg) has recently been questioned as CD38+ cells may produce disease following specific conditioning protocols.25 Finally, in multiple myeloma, we found that plasma cells expressing CD138+ failed to produce disease in NOD/SCID mice, whereas in vivo tumor growth in SCID-hu mice was restricted to plasma cells.26-29 The specifics of each model used to examine in vivo growth or variation among individual patient specimens may explain some of these disparate findings, although it is also possible that multiple unique and distinct populations of tumor cells are capable of clonogenic growth, as suggested by recent data in breast cancer.30
Another area of controversy has been whether all or some human cancers are derived from normal stem cells. The phenotypic resemblance of clonogenic tumor cells in myeloid leukemias and brain tumors to normal hematopoietic and neural stem cells, respectively, has provided evidence in this regard and suggests that genetic or epigenetic modifications may result in the deregulation of self-renewal that is already inherent to normal stem cells.12,31-34 However, the requirement for malignant transformation of normal stem cells has not been firmly ascribed to the CSC hypothesis. In fact, the generation of human cancers from mature cellular compartments, such as memory B cells in multiple myeloma and B-cell precursors in ALL, provides evidence to the contrary.17,19,27 Recent studies have also shown that clonogenic cells may arise from the acquisition of self-renewal capacity within phenotypically mature compartments normally lacking long-term proliferative potential either during tumor initiation or cancer progression.35,36 It is likely that malignant potential is dictated by the presence of distinct abnormalities within a specific cellular context, but the resulting cell capable of tumor propagation, the CSC is defined by the same functional properties, self-renewal and production of cells that are able to recapitulate the original tumor.
Chronic hematologic malignancies, such as CML, myelodysplastic syndrome, follicular non-Hodgkin’s lymphoma, CLL and multiple myeloma, can be clinically generalized as tumors that are responsive to standard therapeutic approaches especially during the initial phases of the disease. However, disease relapse is largely uniform following even CRs, and CSC have been implicated in tumor regrowth because of their clonogenic growth potential.37,38 This clinical pattern also suggests that CSC are more resistant to anticancer therapeutics than cells constituting the tumor bulk. Differential resistance has not been shown for most human cancers in which CSC have been identified, but notable examples exist. A recent study showed that treatment of mice harboring human colon cancer xenografts with typical chemotherapeutic agents enriches functional CSC as measured by limiting dilution during secondary transplantation.39 Furthermore, the clinical relevance of this phenomenon has recently been suggested by a clinical trial in patients with breast cancer receiving neoadjuvant chemotherapy.40 Here, serial analysis of primary patient specimens before and following chemotherapy showed an enrichment of phenotypic breast CSC, as well as increased clonogenic growth capacity in vitro and in vivo. The mechanisms that promote therapeutic resistance in CSC are beginning to be elucidated. In glioblastoma, CSC have been found to be relatively radioresistant through enhanced activation of DNA repair checkpoints.41 In multiple myeloma, we found that CSC were relatively resistant to clinically utilized antimyeloma drugs that included both standard chemotherapeutic and the recently approved agents bortezomib and lenalidomide.27 Although we are currently investigating the exact mechanisms responsible for pan-drug resistance, we found that myeloma CSC display several cellular processes, such as relatively high expression of membrane-bound drug transporters and intracellular detoxifying enzymes, that protect highly resistant normal stem cells from toxic injury. Therefore, mechanisms of therapeutic resistance by CSC may largely be multifactorial and similar to those utilized by normal stem cells.
The resistance of CSC to standard anticancer strategies has prompted the search for relevant therapeutic targets. Self-renewal is likely to play a major role in tumor maintenance and the continued ability to generate new tumor cells. The precise mechanisms that regulate CSC self-renewal are poorly understood, but processes regulating normal stem cells have emerged as candidates (Table 1). These include signaling pathways, such as Hedgehog, Notch, and Wnt that are required for normal embryonic development, and re-activated during the repair and regeneration of injured adult tissues. Each of these pathways has been implicated in the pathogenesis of a wide variety of cancers and the emerging evidence suggests that they may regulate CSC. For example, inhibition of aberrant Hedgehog signaling can limit clonogenic growth in multiple myeloma, CML, pancreatic cancer and brain tumors.42-45,65 Similarly, Notch and Wnt signaling have been implicated in regulating CSC in T-cell ALL, CML, and medulloblastoma.36,44,48,66 Other cellular pathways involved in the self-renewal of normal stem cells include telomerase, which is responsible for maintaining telomeres at the linear ends of chromosomes. In the absence of telomerase activity, progressive telomere shortening during successive rounds of cellular replication induces cellular senescence.53 The critical role of telomerase in the maintenance of normal stem cells is evident in dyskeratosis congenita, as mutational loss of telomerase activity results in the progressive loss of normal hematopoietic stem cells and aplastic anemia.67 The majority of human cancers display increased telomerase activity, and recent evidence suggests that the inhibition of telomerase activity limits the self-renewal of CSC in multiple myeloma.54
Systemic toxicity is a major concern in the clinical development of inhibitors against regulatory pathways shared between normal and CSC. Therefore, targeting strategies specific to CSC are also under investigation. In AML, small molecules that inhibit NF-κB signaling, such as parthenolide and its derivatives, seem to primarily inhibit CSC rather than normal stem cells.60,61,68 Similarly, MoAb-based strategies targeting aberrant expression of CD123, the α chain of the IL-3 receptor, preferentially inhibit leukemic stem cells.69,70 Several recurrent molecular lesions are thought to play critical roles in the pathogenesis of human cancers, such as the BCR-ABL fusion protein in CML, Flt3 mutations in acute leukemias, and Her2/neu over-expression in breast cancer, and recent evidence suggests that each of these may be expressed by CSC.55,71-73 Agents targeting each of these cancer-specific lesions have been approved or are currently in clinical development, but it is unclear whether they can actually inhibit CSC in the clinical setting. In the case of CML, it seems that the clonogenic CSC are resistant to imatinib, an inhibitor of BCR–ABL tyrosine kinase activity, as drug discontinuation results in disease relapse in the vast majority of patients and studies investigating resistance are on-going.74,75 Recently, the second generation BCR–ABL inhibitors, nilotinib and dasatinib, have been approved for use in CML. Although these agents can produce responses in patients with imatinib-resistant CML through the ability to inhibit BCR–ABL harboring additional point mutations, in vitro studies suggest that they similarly lack the potential to eliminate CML stem cells.76-79
SCT offers two primary advantages over conventional therapy.80,81 Dose intensification of standard cytotoxic agents allowable through hematopoietic rescue may improve long-term disease free and overall survival in patients with traditionally chemosensitive diseases, that is, those in which a portion of cases are curable with standard dose therapies, such as AML, aggressive lymphoma and Hodgkin’s disease. In contrast, the benefits of dose intensification have been less definitive in improving outcomes for patients with most chronic hematologic malignancies that are incurable by standard approaches. In these diseases, the allogeneic graft vs tumor effect may be curative in a subset of individuals, and it has been suggested that the immune-mediated anti-tumor effect may be able to overcome resistance to chemotherapeutic agents.82 Growing understanding of the Ag determinants that mediate the graft vs tumor response has led to the development of immunologic strategies that specifically target malignant cells while reducing the risk of GVHD.83,84 However, these data have been largely generated through the examination of bulk tumor cells and if biologically distinct CSC are ultimately responsible for disease relapse, these approaches are unlikely to result in long-term remissions. The rarity of stem cells in diseases such as AML, CML and myeloma, and the lack of methods that can precisely isolate these cells (and thereby exclude contaminating non-stem cell tumor cells or normal stem cells) have complicated attempts to specificallydefine CSC Ag profiles.
Relapse following high-dose chemotherapy suggests that the resistance of CSC is maintained even during dose intensification, and further efforts to deliver even higher amounts of cytotoxic agents are not feasible because of non-hematologic toxicities. Radiolabeled monoclonal antibodies may provide an effective means of escalating cytotoxicity while limiting systemic effects, and radioimmunotherapy has been studied in conditioning regimens either alone or in combination with chemotherapeutic agents.85-87 It is possible that some of these antibodies may bind to Ags likely to be expressed by CSC, such as CD45, in acute leukemias.88,89 Furthermore, targeting the B-cell phenotype of CSC in multiple myeloma through CD20 is currently under study using I-131-tositumomab.59
Several therapeutic strategies targeting CSC are currently in the early phases of clinical development. However, it may be difficult to detect their clinical efficacy if, in fact, they target only a minority of cells. This problem may be especially relevant in studying novel agents. Current earlyphase trials testing novel agents have been developed over the past two decades and the methodology for these trials has not changed significantly during this time.90 Early phase I trials primarily examine feasibility, whereas phase II trials attempt to provide the first evidence of clinical efficacy in selected tumor types. However, the majority of these single-armed trials rely on standard response criteria to show efficacy. These criteria primarily measure rapid changes in tumor bulk using laboratory or radiological measurements that may not necessarily change if only a minority of the tumor cells is inhibited. As decisions to proceed to larger randomized trials are contingent on these early clinical trials, it is likely that agents actually effective against CSC would be abandoned at this stage. However, if sufficient time elapsed, these measurements might detect activity, as the production of new tumor cells is inhibited and the non-targeted cells undergo spontaneous apoptosis. At the current time, the optimal methods to clinically study novel CSC-targeting therapies are unknown, but novel trial designs will need to be developed.
It is possible that SCT can serve as an effective platform to evaluate novel CSC-targeting strategies. Although dose intensification can induce CRs in most patients, autologous transplantation fails to cure many diseases. As CSCs are believed to be responsible for tumor regrowth, the application of novel therapies following remission induction that prevent relapse or prolong disease-free survival may be indicative of activity. Furthermore, correlative laboratory studies capable of detecting and quantifying CSC during this post transplant period may also allow effects against residual CSC to be determined. Such correlative assays have not been definitively established, but we recently quantified CSC in multiple myeloma patients undergoing treatment with a combination of high dose CY and rituximab and found that the in vitro clonogenic growth characteristics could predict disease progression.91 More recently, we have found that multiple myeloma CSC resembling memory B cells and capable of producing disease in NOD/SCID mice are detectable in the peripheral blood and have begun to examine whether the quantification of these cells is associated with clinical outcomes.27 Therefore, novel biomarker strategies capable of serially quantifying CSC may provide unique and relevant end points in clinical trials studying CSC-targeting agents.
The CSC hypothesis has been refined since its conception several decades ago and has provided an attractive means of explaining several discrepancies encountered in clinical oncology, such as the lack of correlation between tumor response and overall survival. However, the CSC concept has also been the source of ongoing controversy. As with many scientific concepts that have ultimately emerged as standard treatment approaches for cancer, such as the use of chemotherapeutic agents and BMT, proof of CSC will ultimately come from the development of therapeutic strategies that provide clinical benefit.92,93 Furthermore, the emergence of novel laboratory methods capable of serially examining CSC within individual patients may help define the optimal clinical context to evaluate their clinical relevance.