Volume 45, Issue 8 p. 872-877
Free Access

Multiple Drug Resistance in Cancer Revisited: The Cancer Stem Cell Hypothesis

Dr Vera S. Donnenberg PhD

Corresponding Author

Dr Vera S. Donnenberg PhD

Department of Surgery, Division of Thoracic Surgery, University of Pittsburgh Cancer Institute, University of Pittsburgh School of Medicine, Pennsylvania

Address for reprints: Vera S. Donnenberg, PhD, Hillman Cancer Research Center, 5117 Centre Avenue, Suite 2.42, Pittsburgh, PA 15213Search for more papers by this author
Dr Albert D. Donnenberg PhD

Dr Albert D. Donnenberg PhD

Department of Medicine, Division of Hematology Oncology, University of Pittsburgh Cancer Institute, University of Pittsburgh School of Medicine, Pennsylvania

Search for more papers by this author
First published: 19 March 2013
Citations: 348


The failure to eradicate cancer may be as fundamental as a misidentification of the target. Current therapies succeed at eliminating bulky disease but often miss a tumor reservoir that is the source of disease recurrence and metastasis. Recent advances in the understanding of tissue development and repair cause us to revisit the process of drug resistance as it applies to oncogenesis and tumor heterogeneity. The cancer stem cell hypothesis states that the cancer-initiating cell is a transformed tissue stem cell, which retains the essential property of self-protection through the activity of multiple drug resistance (MDR) transporters. This resting constitutively drug-resistant cell remains at low frequency among a heterogeneous tumor mass. In the context of this hypothesis, the authors review the discovery of MDR transporters in cancer and normal stem cells and the failure of MDR reversal agents to increase the therapeutic index of substrate antineoplastic agents.

The failure to eradicate most cancers may be as fundamental as a misidentification of the target. Our current therapies succeed at eliminating bulky disease and rapidly proliferating cells but often miss a tumor reservoir that is the source of disease recurrence and metastasis. Recent advances in the understanding of normal tissue development and repair provide a basis for revisiting the process of oncogenesis, tumor heterogeneity, and drug resistance.

From a pharmacological perspective, where emphasis is placed on developing therapies and understanding treatment failure, the tumor stem cell hypothesis provides several new insights that may help us rethink strategies for cancer treatment. Understanding the central role played by multiple drug resistance (MDR) transporters in the protection and self-renewal of normal and cancer stem cells may allow us to identify differences that can be exploited therapeutically. Recognizing that normal stem cells in individual tissues differ with respect to damage tolerance and degree of multipotentiality may translate into differential drug susceptibilities and metastatic potentials of cancer stem cells, depending on the tissue of origin.

This review will attempt to elaborate on the tumor stem cell hypothesis by focusing on the discovery of MDR transporters in neoplastic cells and on the growing body of evidence that these transporters are also an essential feature that enables tumor stem cells to circumvent therapy.


Resistance to chemotherapy was recognized as an impediment to efficacious cancer treatment in the earliest stages of anticancer drug development.1 Surprisingly, cancer cell lines selected for resistance to specific compounds frequently demonstrated cross-resistance to a broad spectrum of structurally unrelated agents.2 In a first attempt to account for the mechanism of cross-resistance, Ling and Thompson showed that resistant cells displayed reduced plasma membrane permeability to cytotoxic compounds.3

The advent of molecular approaches led to the isolation of the first candidate genes for MDR. Roninson and colleagues hypothesized that resistance in drug-selected cancer cell lines arose from amplification of a gene product or gene products capable of altering the plasma membrane. Noting cytogenetic abnormalities common to resistant hamster cell lines, they cloned these amplified fragments and showed that the gene or genes encoded in these fragments were amplified in resistant but not susceptible cells. Removal of selective pressure led to reversion to the drug-sensitive phenotype and coincided with the loss of the amplified DNA.4 Further experiments revealed 2 genes, now recognized as the hamster homologs of the human MDR1 and MDR2.5,6 In 1986, Gros and colleagues transfected the gene now known as MDR1 into drug-sensitive hamster cells. Importantly, they showed that gene duplication or mutations were not required for the acquisition of the multidrug-resistant phenotype.7 The MDR1 gene product is now known as P-glycoprotein (ABCB1). Twentyman and colleagues demonstrated that addition of verapamil (now recognized to be a competitive inhibitor of several MDR transporters) significantly increased the susceptibility of drug-resistant human lung cancer cell lines,8 at once providing a means of verifying MDR activity in vitro and a potential therapeutic avenue for increasing the sensitivity of cancer cells to MDR substrate drugs. ABCB1 did not account for all forms of multiple drug resistance, and additional transporters were identified, among them ABGG2, first described as mitoxantrone resistance9 and later as breast cancer resistance protein (BCRP).10 Today, ABCB1 and ABCG2 are recognized as belonging to a family of at least 48 human ABC transporters involved in a variety of essential cellular transport processes.


The fluorescent dyes Hoechst 33342 and rhodamine 123 (R123) are now known to be substrates for the MDR transporters ABCG2 and ABCB1, respectively. This explains several important earlier findings: 1) Hoechst 33342 dim murine bone marrow cells are greatly enriched for high proliferative potential colony-forming cells;11 R3 dim bone marrow cells protected mice after lethal irradiation;12 and provided a 240-fold enrichment in long-term hematopoietic colony-initiating cells as compared to unfractionated bone marrow.13 However, the critical proof of constitutive upregulation of MDR transporters in primitive hematopoietic stem cells came from Goodell et al, who showed that 10% of Sca-1+, lineage-negative murine bone marrow cells (a phenotype used to define early hematopoietic stem cells) were also Hoechst 33342 dim.14 They termed this small cell subset the side population (SP), after their distinctive flow cytometric profile. Compared to whole bone marrow, SP cells, which composed only 0.1% of bone marrow cells, were 1000-fold enriched with respect to cells able to repopulate lethally irradiated mice. Goodell et al showed that the SP phenotype was abrogated by culturing cells with verapamil, an MDR inhibitor, thus demonstrating that a constitutively active MDR transporter was responsible for exclusion of the Hoechst dye. Sorrentino and colleagues elegantly worked out the details, showing that transfection of ABCB1 (P-glycoprotein) into normal murine marrow increased the SP phenotype by 2 orders of magnitude,15 ABCG2 knockout abrogated the side population,16 and both ABCB1 and ABCG2 are constitutively active in SP cells.17

Although hematopoiesis remains the leading paradigm for tissue differentiation and replacement, the study of adult tissue stem cells has gained momentum with the emergence of the field of regenerative medicine. The potentiality and plasticity of tissue stem cells that mediate tissue repair and maintenance constitute an area of intense study. MDR transporter activity, in the form of the SP, has provided the principal means to recognize and purify such tissue stem cells.18 Although little is known about the role of MDR transporters in adult tissue stem cells, we hypothesize that they follow the hematopoietic paradigm, affording resting stem cells a means of protection that allows them to survive toxic insults that destroy cycling progenitor cells and mature tissue.


The discovery of the molecular mechanism of cross-resistance led immediately to attempts to block MDR transporters with putative reversal agents. Reversal of MDR in vitro was easily attained with a variety of inhibitors. However, in vivo, MDR reversal in the clinical setting has proven to be much more difficult. Thus, intracellular concentrations of adriamycin, vincristine, and etoposide were all significantly increased in 7 human non-small-cell lung cancer cell lines co-cultured in the presence of verapamil (2.2–6.6 μm).19 Similarly, verapamil increased drug sensitivity of ovarian cancer cells lines rendered resistant by culture in the presence of doxorubicin. However, an early indication that in vitro drug selection provides a less than ideal model for in vivo cross-resistance was the observation that primary tumor cell lines isolated from patients with doxorubicin-refractory disease failed to demonstrate drug resistance in vitro.20 This observation will be discussed further in the context of innate versus acquired resistance. Furthermore, in ovarian cancer patients receiving verapamil to plasma levels sufficient to cause inhibition of MDR-mediated adriamycin resistance in vitro (720–2767 ng/mL, = 1.5–5.6 μm), there was no evidence of enhanced response or toxicity of adriamycin coadministered at 50 mg/m2.20 Similarly, in a murine model of adriamycin resistance, continuous infusion of verapamil at the maximally tolerated dose failed to increase the sensitivity of resistant P388 lymphoma cells to adriamycin, despite a strong in vitro effect.21

Although there is still no clear-cut explanation for the failure of verapamil to act as a reversal agent in vivo, substantially different results were obtained with cyclosporine, an agent with a 10-fold higher affinity for the MDR on- and off-sites than the chemotherapeutic agent vinblastine.22 When cyclosporine was given to patients with a variety of refractory cancers in combination with etoposide, cyclosporine levels ranging from 297 to 5073 ng/mL (0.25–4.2 μm) were obtained. Cyclosporine administration had a marked effect on the pharmacokinetics of etoposide, with a doubling of the area under the plasma concentration-time curve. As a result of both decreased renal and nonrenal clearance, a 50% dose reduction was required in patients with normal renal and hepatic function. Toxicities were tolerable but consistent with down-modulation of MDR function in the blood-brain barrier, bone marrow, and biliary tract. Unfortunately, the most critical parameter, intratumor etoposide levels, could not be determined by these studies. Clinical results were modest in this phase I trial of refractory patients, with demonstrable tumor regression in 4 of the 25 patients who attained cyclosporine plasma concentrations in excess of 2000 ng/mL.23

Convincing evidence that administration of an MDR reversal agent could increase the intratumor concentration of a chemotherapeutic agent was provided by Bates and colleagues,24 who used the imaging agent Tc-99m sestamibi, an MDR1 substrate, to measure MDR activity in vivo. Using this technique, they demonstrated the efficacy of the nonimmunosuppressive cyclosporine analog PSC 833 (Valspodar) to reverse MDR activity in vivo. Ten patients with metastatic renal or adrenocortical carcinoma were imaged prior to therapy, 1 day after completing a course of vinblastine and on coadministration of vinblastine and PSC 833. Time activity curves and areas under the curve were obtained for tumor, liver, lung, and myocardium. Myocardium was used as a reference tissue to measure sestamibi uptake in the absence of MDR activity. During the coadministration of PSC 833, tumor visualization was markedly enhanced due to inhibition of MDR-mediated sestamibi efflux, suggesting that intratumor vinblastine concentrations were likewise increased.

Targeting MDR substrate drug directly to the tumor has been modeled using immunoliposomes loaded with doxorubicin and KG-1a leukemia cells. The liposomes were targeted to CD34 expressed on the leukemia cells using an anti-CD34 monoclonal antibody. Immunoliposomal doxorubicin showed a higher cytotoxicity against KG-1a cells than did nontargeted liposomal doxorubicin but failed to overcome doxorubicin resistance. Analysis of liposome-target cell interactions revealed that bound liposomes were not internalized. Thus, the increased cytotoxic effect may have been due to drug release proximal to the cells but not to a breach of membrane-associated MDR transporters.25

Further trials of first-generation modulators such as verapamil, quinidine, and cyclosporine proved them to be either inefficacious or associated with unacceptable toxicities. The limited clinical utility of the second- and third-generation MDR inhibitors such as PSC 833, GF120918, VX-710 (Biricodar), and LY335979 for potentiating antineoplastic agents may also be explained in part by multiple and redundant cellular mechanisms of resistance, unfavorable alterations in the pharmacokinetics of cytotoxic agents, and attendant toxicities associated with the systemic inhibition of MDR function. Since MDR transporters are themselves redundant with overlapping activities, specific inhibition of 1 transporter may leave drug resistance essentially intact.17 The take-home message of these studies is that MDR reversal agents can be used to increase the plasma concentration of a variety of antineoplastic agents but not to increase their therapeutic index (reviewed by Tan et al26). The hypothesis that cancer arises uniquely from the mutation of tissue stem cells provides a theoretical framework for understanding this important observation.


The cancer stem cell was first proposed by Fiala in 1968.27 Although modern concepts of stem cell biology were absent, the cancer-initiating cell was clearly hypothesized to be a “stem cell unable to differentiate.” The cancer colony assay proposed by Hamburger and Salmon in the late 1970s introduced the concept that only a small proportion of cancer cells, cancer stem cells, are tumorigenic, and the authors identified these cells as the essential target of therapy.28 However, recent advances in regenerative biology have allowed tumor growth to be understood in the context of the dysregulation of normal tissue replacement and wound healing.

The modern revival of the tumor stem cell paradigm originated in the laboratory of Dr Irving Weissman,29 who first isolated the multipotential hematopoietic progenitor cell.30 Knowledge of the central role that MDR transporters play in protecting normal stem cells has allowed us to further refine this hypothesis and add new insights that may prove relevant to explaining treatment failure, late recurrence, metastasis, and tissue-specific differences in cancer incidence. Such knowledge may also guide us to design rational therapies that take into account similarities and differences between cancer and normal stem cells.

Given the central role of MDR transporters in protecting normal and neoplastic cells, the cancer stem cell hypothesis provides a unified explanation for the successes and failures of cytotoxic antineoplastic therapy (detailed in Figure 1). Namely, the most important target, the resting cancer stem cell, is spared along with its normal tissue stem cell counterparts. On a populational level, different malignancies may appear to be heterogeneous with respect to drug responsiveness. Cancers that respond to therapy initially may appear to acquire drug resistance during the course of treatment. Other cancers may appear to be intrinsically resistant. The cancer stem cell hypothesis posits that in both instances, the resting cancer stem cell, which is both the cancer-initiating cell and its source of replenishment under selective pressure, has innate drug resistance by virtue of its resting stem cell phenotype. Acquired drug resistance in more differentiated cancer cells, through gene amplification or rearrangement, may contribute to an aggressive phenotype, but it is not the primary reason for cancer recurrence or spread after therapy.

Details are in the caption following the image

. Constitutive and acquired multiple drug resistance (MDR) and the cancer stem cell hypothesis. The cancer stem cell hypothesis posits that the cancer-initiating cell is a tissue stem cell, where a stem cell is understood to be a cell that can self-renew, self-protect, and give rise to progenitors of high proliferative activity (amplifying or progenitor cells), which in turn produce mature progeny. The key features of a stem cell, as distinguished from a progenitor cell, are self-renewal, self-protection, damage tolerance, and a developmental state earlier than the highly proliferative progenitor cell that gives rise to fully differentiated daughters. The multistep process of neoplastic transformation begins with a mutation (yellow circle, hit 1) to a cycling tissue stem cell. Normally, tissue stem cells have constitutive MDR activity, but upon entry into the cell cycle, MDR is transiently downregulated, resulting in a window of vulnerability to DNA damage. In a process known as asymmetrical division, cycling stem cells self-replicate, giving rise to an MDR-protected resting stem cell and a drug-sensitive amplifying progenitor cell of high proliferative capacity. The damaged daughter stem cell reverts to an MDR-protected state when it exits the cell cycle, while the damaged progenitor proliferates and gives rise to drug-sensitive dysplastic cells. Additional mutations (hit N) and genetic instability accumulated by the damaged stem cell result in further growth dysregulation and the emergence of frank neoplasia. Throughout this process, the cancer stem cell remains protected and rarely enters the cell cycle. The bulk of the tumor mass is generated by mitotically active, drug-sensitive amplifying progenitor cells. Exposure to MDR substrate antineoplastic agents results in the elimination of drug-sensitive tumor but not resting tumor stem cells. It also imposes selective pressure for mutations that may result in overexpression of MDR transporters (red circle) and the acquired phenotype of drug resistance in tumor progenitor cells and their progeny. The most important concepts to emerge from his model are (1) MDR is constitutively expressed at high levels in tissue stem cells and therefore in the nascent cancer cell, (2) cancer results from accumulated mutations at the stem cell level, and (3) the cancer stem cell hypothesis posits that the clinically relevant target of therapy is a resting cell with drug resistance that is not dependent on therapy-induced gene duplication or translocation.

As detailed above, one of the defining characteristics of adult tissue stem cells is their constitutive resistance to environmental toxins, including most chemotherapeutic agents. In fact, dose-limiting toxicities of many antineoplastic agents occur precisely at drug concentrations that damage normal tissue stem cells. The constitutive drug resistance of normal tissue stem cells is mediated by MDR transporters and detoxifying enzymes. DNA repair mechanisms, tolerance to damage (ie, resistance to apoptosis), and telomerase activity also contribute to the stability of normal tissue stem cells.


In light of the cancer stem cell hypothesis, it is worthwhile to reexamine issues of drug resistance, cross-resistance, and the failure of MDR reversal strategies. The compounds that have been the most studied clinically as reversal agents are verapamil and cyclosporine and its analogs. Despite some promising results in hematological malignancies, the outcomes achieved when MDR blockers were coadministered with substrate drugs have been disappointing. This failure may be partially explained by the redundancy of the individual transporters within the MDR phenotype together with several other resistance-related proteins expressed in solid tumors (eg, glutathione S-transferase, metallothionin, O6-alkylguanine-DNA-alkyltransferase, thymidylate synthase, dihydrofolate reductase, heat shock proteins) or other factors contributing indirectly to resistance such as vascularization.31 However, it follows from the cancer stem cell hypothesis that systemic administration of an efficacious reversal agent would render tumor and normal tissue stem cells equally susceptible to the chemotherapeutic agents, offering no net gain in therapeutic index.

As discussed above, the discovery of the first MDR transporter began with the observation of gene amplification in hamster cells selected in vitro for drug resistance.4 Removal of the drug used for selection resulted in the outgrowth of cells without amplified MDR genes and loss of the multiple-resistant phenotype. However, in vivo drug resistance is not dependent on prior drug exposure, as was demonstrated using the tumor cell culture assay to culture lung cancer cells.32 Current knowledge of regulation of MDR activity in stem cells and their progeny allows reconciliation of these findings. Drug resistance is an innate characteristic of the resting tumor stem cell but must be acquired by more differentiated tumor cells through gene amplification or rearrangement. The idea that transforming events in cancer lead to the juxtaposition of MDR and active genes through gene rearrangement is consistent with Roninson's findings in cell lines but is essentially an epiphenomenon according to the stem cell hypothesis. The cancer stem cell expresses constitutive MDR activity, which is independent of drug exposure, and is downregulated in more differentiated tumor progeny. It has been proposed that selective pressure imposed by chemotherapy leads to both mutation and secondary genetic changes, including MDR upregulation in the bulky tumor.33 However, unless these changes occur in the self-renewing tumor stem cell compartment, the limited proliferative capacity of the bulky tumor ensures that they are self-limiting. Thus, the major barrier to therapy is the quiescent tumor stem cell with constitutive MDR.


Hematologic malignancies stand out among the cancers that are sometimes susceptible to cure. Allogeneic hematopoietic stem cell transplantation has been particularly successful34,35 because stem cell rescue with donor hematopoietic stem cells obviates the need to spare normal hematopoiesis. Although skin, gut, and other tissues with rapid turnover are acutely affected by high-dose therapy, their constitutively drug-resistant stem cells quickly replenish the damaged tissue. Even if the transformed leukemia stem cells have evolved mechanisms that render them more protected from toxic insults than normal hematopoietic stem cells, immune recognition of minor histocompatibility antigens expressed on leukemia stem cells, but not on repopulating donor hematopoietic stem cells, provides the coup de grace in a process now recognized as the graft versus leukemia effect.36

Unfortunately, no analogous ability now exists to rescue nonhematopoietic stem cells following stem cell ablative therapy. If the proposed relationships between normal and neoplastic stem cells prove correct, the inescapable conclusion is that systemic cytotoxic therapies are doomed to failure because regimens that spare resting normal stem cells will also likely spare resting tumor stem cells. Successful therapy awaits the discernment of biological and immunological differences between the tumor and normal stem cells and the exploitation of the hypothesized window of vulnerability (Figure 1) that exists when the cancer stem cell is transiently recruited into the cell cycle.