Author contributions: Y.T. and T.Z.: conception and design, collection and assembly of data, data analysis and interpretation, final approval of manuscript; I.W., K.K., J.P., and J.A.: collection and assembly of data, analysis and interpretation of data, final approval of manuscript; H.L.R., J.V., N.J., B.M., and P.S.: collection and assembly of data, final approval of manuscript; L.M.: provision of study material, final approval of manuscript; B.Q. and P.F.: conception and design, data analysis and interpretation, financial and administrative support, final approval of manuscript; and R.P.: conception and design, data analysis and interpretation, manuscript writing, administrative and financial support, final approval of manuscript. Y.T. and T.Z. contributed equally to this work.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS January 25, 2013.
Melanoma is one of the most aggressive and extremely resistant to conventional therapies neoplasms. Recently, cellular resistance was linked to the cancer stem cell phenotype, still controversial and not well-defined. In this study, we used a Rhodamine 123 (Rh123) exclusion assay to functionally identify stem-like cells in metastatic human melanomas and melanoma cell lines. We demonstrate that a small subset of Rh123-low-retention (Rh123low) cells is enriched for stem cell-like activities, including the ability to self-renew and produce nonstem Rh123high progeny and to form melanospheres, recapitulating the phenotypic profile of the parental tumor. Rh123low cells are relatively quiescent and chemoresistant. At the molecular level, we show that melanoma Rh123low cells overexpress HIF1α, pluripotency factor OCT4, and the ABCB5 marker of melanoma stem cells and downregulate the expression of Cyclin D1 and CDK4. Interestingly, a short treatment with LY294002, an inhibitor of the PI3K/AKT pathway, specifically reverts a subset of Rh123high cells to the Rh123low phenotype, whereas treatment with inhibitors of mammalian target of rapamycin, phosphatase and tensin homolog or mitogen-activated protein kinase signaling does not. This phenotypic switching was associated with reduced levels of the HIF1α transcript and an increase in the level of phosphorylated nuclear FOXO3a preferentially in Rh123low cells. Moreover, the Rh123low cells became less quiescent and displayed a significant increase in their melanosphere-forming ability. All the above indicates that the Rh123low melanoma stem cell pool is composed of cycling and quiescent cells and that the PI3K/AKT signaling while maintaining the quiescence of Rh123low G0 cells promotes the exit of cycling cells from the stem cell compartment. STEM CELLS2013;31:641–651
Melanoma is one of the most frequently occurring cancers for which there is no effective treatment after metastasis. The most recent therapies targeting the genetic mutations associated with melanomas, for example, b-raf, were initially promising but have failed to enable complete remission [1, 2]. Therefore, it is imperative to elucidate the mechanisms responsible for the extraordinary resistance of melanoma cells and to identify points in the persistence pathways that might yield new approaches for the effective destruction of all melanoma cells. Accumulating recent evidence suggests that cancer stem cells (CSCs), also termed tumor-initiating cells, play an essential role in the initiation, development, and recurrence of cancers and in their resistance to anticancer treatments [3, 4]. CSCs have been identified in leukemia and several solid tumors (reviewed in [5, 6]). Certain features of melanoma tumors such as cellular heterogeneity and plasticity, the aberrant expression of certain developmental molecules important for self-renewal and pluripotency [7, 8] or the expression of surface markers including CD133 , ABCB5 [10, 11], and the more recently reported CD271  collectively suggest the existence of a stem-like population in melanomas. However, both the selectivity of these markers and the hierarchical organization of melanoma cells have been questioned , challenging the existence of CSCs in melanomas. In fact, several surface markers described as strictly stem cell (SC)-specific may change their expression patterns according to cell cycle status or environmental conditions, as was shown for CD133 expression  or for CD34 [15, 16]. Therefore, the isolation of CSCs cannot be based solely on the variable expression levels of cell surface markers. An alternative method of distinguishing CSCs exploits their normal SC-like functional traits including the ability to form spheres, to self-renew and yield differentiating progeny, and to efflux drugs and vital dyes more efficiently than nonstem cells [4, 6, 17]. The dye efflux assay for identifying a side population of low-dye-retaining cells has proven to be a particularly powerful tool for SC isolation, allowing selection for multidrug-resistant cells overexpressing ATP-binding cassette (ABC) transporters including the Hoechst 33342-excluding breast cancer resistant protein1 (BCRP1) or ABCG2 transporter, and the Rh123-excluding multidrug resistance ABCB1 and ABCB5 transporters (MDR1 and MDR/TAP, respectively) [18, 19].
Interestingly, quiescent long-term hematopoietic stem cells (HSCs) were described as low-Rh123-retaining cells . These cells remain in the metabolically inactive G0 state for the majority of their lifetimes and divide only infrequently to preserve their high self-renewal potential. Presumably, quiescent CSCs have similar properties; therefore, a switch from the quiescent to the active state may be controlled by analogous pathways that are aberrantly regulated in CSCs. In the present study, we identified a rare subpopulation of low-Rh123-retaining cells with SC-like traits in both human melanomas and in human and mouse melanoma cell lines. In addition, we showed that the PI3K/AKT signaling pathway may differentially regulate quiescent and cycling melanoma stem-like cells.
MATERIALS AND METHODS
Rhodamine123, diamidino-2-phenylindole (DAPI), DiI, and Hoechst 33342 were purchased from Molecular Probes (Leiden, The Netherlands http://www.probes.com/resources). Verapamil, camptothecin, propidium iodide (PI), trypsin/EDTA, Pyronin Y, paraformaldehyde (PAF) solution, and poly-2-hydroxyethylmetacrylate were purchased from Sigma-Aldrich, St. Quentin Fallavier France. Fotemustine and cisplatin were gifts from CHRU Lille, France. Phosphatase and tensin homolog (PTEN) inhibitor bpV(phen), Rapamycin, PD98059, and LY294002 were purchased from Calbiochem (France). Epidermal growth factor was obtained from Stem Cell Technologies (Grenoble, France), and recombinant human basic Fibroblast Growth Factor from PromoKine-PromoCell GmbH (Heidelberg, Germany). Collagenase type I, dispase, and B27 supplement were provided by Invitrogen (St. Aubin, France).
Tumor Samples, Spheres, and Culture
Tumor samples (n = 6) were obtained from melanoma patients (Clinique de Dermatologie, CHRU, Lille) who had undergone tumor excision surgery. The storage and use of human biological samples were declared and performed according to the local Person's Protection Committee and the ethical rules approved by the Department of Health, France. Immediately after surgical removal, tumors were rinsed in phosphate buffered saline (PBS), digested with 2 mg/ml of collagenase type I and 3 mg/ml dispase for 1-3 hours followed by 0.5% trypsin/EDTA treatment for 30 minutes and then passed through a 70-μm filter (BD Biosciences, Le Pont de Claix, France). The cells were cultured in 25 ml flasks in RPMI supplemented with 10% fetal calf serum (FCS) (Lonza, Verviers, Belgium). The Me14M-tumor #4-derived culture has been passaged more than 50 times and, therefore, is considered to be an established human melanoma cell line. Supporting information Figure S1 illustrates the karyotype of this new Mel4M cell line, and Supporting information Table S1 presents its CGH profile. The HBL and LND human melanoma cell lines were provided by Pr. Ghanem, Bruxelles, Belgium. Their phenotype was confirmed by morphology and CGH analysis (HBL). The mouse B16 and human A375 melanoma cell lines was purchased from ATCC (B16, CRL-6323e). The A375, HBL, LND, and Mel4M melanoma cell lines and the A549 lung adenocarcinoma  cell lines were cultured at 37°C in a humidified atmosphere with 5% CO2 in RPMI, and B16 in Dulbecco's modified Eagle's medium, both supplemented with 10% FCS. The medium was changed every 3 days. Spheres were generated in 24-well plates as described previously , with the exception that the plates were coated with a solution of 0.5 mg/ml poly-2-hydroxyethylmetacrylate in ethanol to prevent cell attachment. For the label-retaining cell (LRC)-assay, cells were stained with 1.0 μM DiI fluorescent dye for 10 minutes at 37°C in PBS suspension then rinsed and plated as above. Sphere cultures were fed every 5 days by adding a 1:1 volume of fresh medium, and sphere-forming unit (SFU) values (no. of spheres × 100/no. of plated cells) were estimated after 7-14 days. Spheres were counted under the microscope by two independent experimenters. For the self-renewal assay, primary spheres were recovered, dissociated by short trypsinization and replated at a clonal density of 1,000 cells per milliliter.
Rhodamine123 Exclusion Assay
Cells were adjusted to a concentration of 106 cells per milliliter in RPMI with 10% FCS. The Rh123 exclusion assay and fluorescence-activated cell sorting (FACS) were performed as previously described . When indicated, cells were either pretreated with 10 μM LY294002 for 30 minutes at 37°C before Rh123 loading or pretreated and treated during 20 minutes of Rh123 loading (50 minutes exposure). After removing Rh123, a 60-minute exclusion step was performed in the presence or absence of LY294002. Total treatment with LY294002 did not exceed 110 minutes. Rapamycin (100 nM), bpV(phen) (500 nM), and PD98059 (10 μM) were added 50 minutes before the Rh123 exclusion step. After FACS, the collected cells were plated for sphere generation, cell cycle analysis, quantitative real-time polymerase chain reaction (qRT-PCR), western blots, and chemoresistance assays. For confocal microscopy, cells grown on Lab-Tek eight-well chambered glass coverslips (Nalge Nunc Int., Rochester, NY) were treated with either 10 μM LY294002 or vehicle for 50 minutes at 37°C and were then rinsed, cooled on ice, and analyzed by confocal microscopy (using a Zeiss LSM710 Confocal Microscope). Images were obtained to define the fluorescence intensity after Rh123 loading. Subsequently, slides were repositioned on the confocal microscope according to the preregistered XYZ dimensions, and images were again obtained to register the levels of fluorescence intensity after the Rh123 exclusion step. Finally, cells were fixed with PAF and processed for immunohistochemistry according to standard procedures. Primary rabbit anti-human phosphorylated FOXO3a antibody (dilution 1:100; Ozyme, St. Quentin en Yvelines, France) was applied for 2 hours followed by secondary AlexaFluor 568 donkey anti-rabbit (dilution 1:2,000) (Invitrogen, St. Aubin, France) for 1 hour, both at room temperature (RT). To visualize nuclei, DAPI was added 5 minutes before micrographs were taken.
Tumor samples and spheres were embedded in Tissue-Tek OCT (Sakura Finetechnical, Tokyo, Japan) and cryosectioned. Sections (10 μm) placed on glass slides were air-dried and fixed in PAF solution, and immunohistochemistry was performed according to the standard procedures. Monoclonal anti-Mart-1, anti-M-CAM, anti-Nestin antibodies (Abs), and goat polyclonal anti-OCT4 antibody were purchased from Santa Cruz Biotech. (Heidelberg, Germany) and anti-CD133 antibody was purchased from Miltenyi Biotec. (Paris, France); all were used at a dilution of 1:100. Secondary AlexaFluor 488 goat anti-mouse or anti-goat antibodies (Life Technologies, St. Aubin, France) were used at a dilution of 1:1,000. Negative controls were performed by replacing the primary Abs with irrelevant Abs of the same isotype. DAPI or PI (0.5 μg/ml) were used for nuclear counterstaining. All slides were mounted under a coverslip with Vectashield Mounting Medium (Vector Laboratories, Nanterre, France) and were photographed using a Leica fluorescence microscope or an epifluorescence microscope equipped with a laser-scanning confocal imaging system (Leica, Göttingen, Germany).
RNA Extraction, RT-PCR, and qRT-PCR
RNA extraction was performed following the manufacturer's protocol (RNeasy Kit, Qiagen, Courtaboeuf, France) with optional DNase treatment, and samples were stored at −80°C. Primers were designed to amplify cDNA fragments ranging in size from 142 to 389 bp as follows: HIF-1α, 5′CGTTCCTTCGATCACTTGTC3′ and 5′TCAGTGGTGGCAGTGGTAG T3′; OCT-4, 5′GTGGAGGAAGCTGCAAACAATGAAA-3′ and 5′-GACCGAGGAGTTAC AGTGCAGTGAAG-3′; and 18s, 5′-AAACGGCTACCACATCCAAG-3′ and 5′-CGCTCCCAAGATCCAACTAC-3′. For qRT-PCR, RNA was transcribed into cDNA using random hexamers and the High Capacity Reverse Transcription kit from Applied Biosystems (Foster City, CA). All qRT-PCR reactions were performed with TaqMan fluorescent probes provided by Applied Biosystems. The relative expression ratio for each gene was calculated by the ΔΔCt method. The calculated values represent the expression level for each gene relative to the expression of 18S ribosomal RNA, which was used as the endogenous control.
Western Blot Analysis
Western blot analysis was performed using ready-to-use NuPAGE 4%–12% Bis–Tris polyacrylamide gels according to the supplied instructions (Invitrogen, St. Aubin, Paris, France). Blots were probed with the appropriate primary antibodies to Actin (Sigma-Aldrich, St. Quentin Fallavier, France), Cyclin D1 (Santa Cruz Biotech., Heidelberg, Germany) and CDK4 (Abcam, France), Akt, pAkt, Foxo3a and pFoxo3a (Ozyme, St. Quentin en Yvelines, France) followed by a horseradish peroxidase-conjugated secondary antibody (Bio-Rad, Marne-la-Coquette, France). Immunodetection was performed using an ECL+ chemiluminescence kit from Amersham.
ABC Transporter Activity Assay
To determine the activity of the MDR and BCRP transporters, their substrates, dihydrorhodamine123 (DHR, 10 μM), and 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA, 40 μM), respectively, were used as previously described . Assays were performed using melanoma Mel4M and A549 lung adenocarcinoma cell lines either treated or not treated with LY294002 before substrate exclusion. Data are represented as percentages of the relevant control.
Cell Cycle Analysis
The cell suspension was rinsed twice with PBS and fixed with 70% ethanol at −20°C. After rinsing with cold PBS, cells were incubated with 50 μg/ml PI and 5 μg/ml RNase in PBS for 30 minutes at RT. The reaction was stopped by placing cells on ice only before flow cytometry analysis. To determine the percentage of cells in the G0 phase of the cell cycle, cells were preincubated with anti-human FITC-conjugated Ki67 antibody (Abcam, Paris, France) or matching isotype for 30 minutes on ice before adding PI and RNase. The Pyronin Y/Hoechst 33342 assay was performed with 5 × 105 cells preincubated overnight at −20°C with 70% ethanol before the addition of 4 μg/ml of Hoechst 33342 and 2 μg/ml PY and incubation for 20 minutes at RT. Flow cytometry analysis of the total population permitted the gating of cells with low and high PY (FL2) fluorescence intensities. This gating was applied to Rh123low and Rh123high cells to determine the percentage of G0 cells in the G1/G0 fraction.
All results are expressed as the means ± SEM of at least two independent experiments and three repetitions. Comparisons between means were assessed using the unpaired Student's t test. Welch's correction was applied when unequal variance was observed. The statistical analyses were performed using the GraphPad Prism 4.0 software. A p value ≤0.05 was considered to be significant.
Melanoma Tumors Contained Low-Rh123-Retaining Cells with Stem Cell Activity
In the absence of definite cell surface markers for melanoma SCs, we used a functional Rh123 efflux assay that was previously used to differentiate normal  and cancer  SCs within a subpopulation of cells with reduced Rh123 retention ability (Rh123low cells). To determine whether melanomas also contain the Rh123low subset, cells from six human metastatic tumors and four human and mouse melanoma cell lines were assessed for their ability to efflux the Rh123 dye. Rh123 not only is a specific substrate for the ABCB1 and ABCB5 transporters but also binds to active mitochondria, and its retention thus reflects the metabolic state of a cell [23, 19]. Therefore, a low Rh123 fluorescence may result from either a low number of active mitochondria in the cell or a high drug efflux activity. To establish unambiguously the predominant role of ABC transporters, we used verapamil, a well-known pharmacological inhibitor of drug efflux . Figure 1A presents flow cytometry profiles for human Mel4M cells stained with Rh123. Equal dye loading in the presence or absence of verapamil corroborated that ABC transporters did not contribute to Rh123 uptake (Fig. 1A, upper panel). However, interestingly, after 1 hour of incubation in a dye-free medium, a small yet distinct (4.04% of the total population) verapamil-sensitive subpopulation of Rh123low cells appeared (Fig. 1A, lower panel), indicating that Rh123 efflux in these cells was ABCB1/ABCB5-dependent. This finding established a direct link between the Rh123low phenotype and the activity of these transporters in melanoma cells.
Using the human Mel4M cell line as a model, we then investigated in detail the cellular and molecular aspects of the Rh123low cell compartment. The Rh123low cells displayed low forward scatter (FSC) and side scatter (SSC) (Fig. 1B), respectively, signifying the small size and low inner complexity typical of immature cells. To explore their metabolic activity, we used a double labeling (Pyronin Y/Hoechst 33342) technique to stain both RNA and DNA . This analysis revealed that only 0.34% of the Rh123high cells and up to 19.76% of the Rh123low fraction poorly incorporate Pyronin Y (PY) (Fig. 1C), indicating that the Rh123low fraction is enriched for low-RNA-content thus metabolically inactive, quiescent cells . Accordingly, Rh123low cells expressed significantly lower levels of the proliferation markers Cyclin D1 and CDK4 than their Rh123high counterparts (Fig. 1D).
The Rh123low subpopulation represented from 0.56% to 6.42% of total freshly dissociated tumor cells and from 0.71% to 5.77% of established cell lines (Fig. 1E). These proportions remained unchanged even after a long propagation in culture, suggesting that the size of the Rh123low cell compartment may be intrinsically programmed.
The Rh123low Subpopulation Contained Sphere-Forming and Self-Renewing Cells, Recapitulating the Heterogeneity of the Original Tumor
A functional analysis of sorted Rh123low Mel4M tumor cells (Fig. 2A) revealed that this fraction was 9.5-fold enriched for the sphere-forming ability that defines stem cells , in comparison with their Rh123high counterparts. The superior sphere-forming ability of Rh123low cells was confirmed in three other melanoma tumors (data not shown). Melanosphere cells reformed new spheres for a minimum of four consecutive generations (Fig. 2B, upper panel), demonstrating their self-renewal and long-term repopulating abilities. Interestingly, the passage of Mel4M generation one (g1) Rh123low tumor cell-derived spheres to g2 spheres was accompanied by a 2.6-fold increase in the SFU values (Fig. 2B, upper panel) and a 3.5-fold increase in the percentage of Rh123low cells (compare the lower panels of Figs. 1A and 2B), confirming a tight relationship between Rh123low cells and sphere-forming ability and indicating that the Rh123low compartment can expand in response to changing culture conditions. A comparative immunohistofluorescence analysis revealed that the melanospheres resembled their parental tumors. The results shown in Figure 2C demonstrate that both contained similar proportions of cells, positive and negative for the melanoma differentiation marker Mart-1, the invasion and metastasis marker M-CAM, the embryonic marker of pluripotency and self-renewal OCT4, and the stem cell markers Nestin and CD133. Thus, the melanoma Rh123low cells were self-renewing and capable of initiating tumor-like spheres, recapitulating the parental tumor's cellular heterogeneity and gene expression pattern.
Recent data suggest that tumor heterogeneity, analogously to normal tissues, originates from a hierarchical organization of tumor cells [4–6, 26]. Such organization appears to exist also in melanoma, wherein ABCB5-positive cells were reported to have stem cell properties . Using qRT-PCR analysis, we showed that the levels of ABCB5 transcripts expressed specifically by a subset of chemoresistant CSCs  were considerably higher in the Rh123low stem-like subpopulation than in Rh123high cells (Fig. 2D). Accordingly, the Rh123low cells were significantly less sensitive than their Rh123high counterparts to fotemustine, cisplatin and camptothecin, chemotherapeutics currently used against melanoma (supporting information Fig. S2). As expected, the Rh123low cell-enriched melanospheres divided 3.5 times more slowly than adherent cells (Fig. 2E), most likely due to the presence of slow-cycling cells, observable in Figure 2F as DiI LRCs. Simultaneously, 1-week-old monolayer cultures were free of LRCs (data not shown). Consequently, LRC-containing melanospheres injected into immunodeficient SCID mice produced tumors much more slowly than did adherent cells (supporting information Fig. S3). Moreover, the adherent Rh123low cells, the majority of which were cycling, appeared to be relatively more tumorigenic and generated bigger tumors than the main Rh123high subpopulation (supporting information Fig. S4). Collectively, our data demonstrated that the Rh123low, but not the Rh123high phenotype, identifies a small subset of ABCB5-overexpressing, slow-cycling, self-renewing, stem-like melanoma cells. A significant proportion of these stem-like cells remained in G0 metabolically inactive (PYlow) state in adherent cultures and label-retaining in tumor-recapitulating melanospheres.
Inhibition of the PI3K/AKT Signaling Pathway Increased the Pool of Rh123low Melanoma Stem-Like Cells by Enhancing ABCB1/ABCB5 Functionality
Normal and cancer stem cells oscillate between quiescent and active states to persist and to self-renew [26, 28, 29], and the PI3K/AKT signaling pathway controls the passage between these two states [30, 31]. Because this pathway is frequently deregulated in melanoma cells, we questioned whether it controls the size of the Rh123low melanoma stem-like cell compartment. Western blot of sorted Rh123low and Rh123high cells established that cells in this compartment expressed higher levels of pAKT than the Rh123high major population (supporting information Fig. S5). Mel4M cells were treated 50 minutes before the Rh123-efflux stage with LY294002, an inhibitor of PI3K/AKT signaling pathway, bpV(phen) inhibitor of PTEN, a phosphatase inactivating the PI3K/AKT signaling and with rapamycin and PD98059, inhibitors of mTOR and ERK signaling, respectively, downstream AKT effectors both involved in the self-renewal and proliferation of stem cells [32, 33]. As illustrated in Figure 3A, a short LY294002 treatment significantly increased the pool of Rh123low cells whereas inhibitors of the mTOR and ERK signaling pathways had no effect and inhibitor of PTEN slightly (p < .05) decreased the pool of Rh123low cells. LY294002 inhibition of the PI3K/AKT signaling pathway also increased from 6.4 ± 0.06% to 25.2 ± 0.05%, the Rh123low pool in another A375 melanoma cell line. The size of this pool inversely correlated with the levels of phosphorylated AKT (pAKT) present in the LY294002 and PTEN inhibitor-treated cells (Fig. 3B). Moreover, the size of the Rh123low pool was directly proportional to the duration of LY294002 exposure (Fig. 3C). Interestingly, the size of this pool was sustained, but not increased, after 24 hours of pretreatment with LY294002, demonstrating that deactivation of the AKT pathway enlarged the Rh123low cell compartment very rapidly to its maximum (Fig. 3D). No cell death was observed in the treated samples, even after 24 hours (Annexin V/7AAD test, data not shown). However, the size of the pool was diminished by verapamil (Fig. 3C), demonstrating that the effect of LY294002 is not related to mitochondrial dysfunction but is instead ABC transporter-dependent. LY294002 may thus affect either the functionality of these transporters and/or the expression levels of the corresponding genes. The results of qRT-PCR assays of actively growing control and LY294002-treated Mel4M cells (supporting information Fig. S6) rather ruled out the possibility of transcriptional regulation of the ABCB1 and ABCB5 transporters and suggested that LY294002 may increase their functionality. Because LY294002 has been reported to decrease the activity of ABCG2 (BCRP1) and ABCC1 (MRP-1) transporters in several models [34, 35], we compared the effects of LY294002 on the efflux of both Rh123, a substrate of ABCB1/ABCB5, and DCFH-DA, a substrate of ABCG2/ABCC1  in Mel4M cells and in the irrelevant control A549 lung adenocarcinoma cells. The results revealed a significant decrease in the intensity of Rh123 fluorescence in melanoma but an increase in A549 cells. The fluorescence intensities of DCFH-DA were greater than the control values in both cell lines treated with LY294002 (Fig. 3E). Because transporter activity is inversely proportional to substrate fluorescence intensity, we concluded that the short exposure to LY294002 specifically upregulated the activity of MDR transporters only in melanoma cells and this upregulation contributed to the Rh123low phenotype. The observed 2.5-fold increase in the Rh123low subset was very rapid and occurred during the 60 minutes Rh123 exclusion step (Fig. 3A and 3C). At this time point, neither cell death nor cell division could significantly change the size of the total cell population. Therefore, the majority of the new, LY294002-induced Rh123low cells may have been derived from Rh123high cells shifting to the Rh123low phenotype while activating their dye exclusion machinery.
Inhibition of the PI3K/AKT Signaling Pathway Regulated the Transition of G0 Quiescent Cells to the G1 Phase of the Cell Cycle Within the Rh123low Subset
Previously, it was shown that Rh123low hematopoietic stem cells were relatively quiescent when compared to cycling and more mature progenitors [36, 37]. To determine the cycling activity of the Rh123low and Rh123high cells, we performed cell cycle analysis with an anti-Ki67 antibody that recognizes cycling cells preferentially over G0 quiescent cells. As expected, fewer S/G2/M cycling cells (9.5%) and more G0/G1 (85.2%) cells were detected in the Rh123low subset than in the corresponding Rh123high compartment (30.2% and 65.8%, respectively; Table 1). Furthermore, whereas LY294002 short treatment did not significantly modify the proportions of cycling and noncycling cells in either subset, this treatment markedly reduced the proportion of G0 cells and increased the proportion of G1 cells in the Rh123low subset (Table 1 and Fig. 4A, 4B). This effect was accompanied by an increased ability of LY294002-treated Rh123low, but not Rh123high cells to form spheres (Fig. 4C), indicating that the inhibition of the PI3K/AKT signaling pathway increased not only the proportion of Rh123low cells but also their stem cell activity while rendering them less quiescent and metabolically more active as judged by PYhigh phenotype. Consistently, the sphere forming ability was decreased when cells were treated with PTEN inhibitor (Fig. 4D). These results imply that the PI3K/AKT signaling pathway controls the size of the stem-like Rh123low pool and contributes to the maintenance of their quiescence.
Table 1. LY294002 significantly decreased the G0 fraction and increased the G1 in the Rh123low cell compartment
Cell cycle analysis of fluorescence-activated cell sorting-sorted Mel4M control Rhodamine123low (Rh123lo) and Rhodamine123high (Rh123hi) cells and cells treated with 10 μM LY294002 for 50 minutes using the Ki67/PI assay. Numbers indicate the mean percent of cells in the particular phase of the cell cycle obtained from flow cytometry analyses.
p < .05 treated vs. untreated cells.
LY294002 Downregulated the Levels of HIF1α Transcripts and Provoked the Retention of Phosphorylated FOXO3a in the Nucleus
To gain insight into the molecular mechanisms by which PI3K/AKT signaling affected the size of the Rh123low pool and cellular quiescence, we determined the levels of the downstream signaling targets HIF1α and OCT4 in control- and LY294002-treated Mel4M cells. As illustrated in Figure 5A, the transcripts of HIF1α and its downstream target OCT4  were differentially expressed in Rh123low and Rh123high cells. LY294002 treatment lowered HIF1α expression to almost basal levels in Rh123low cells but upregulated it in their Rh123high counterparts without significantly affecting OCT4 expression. This finding excluded OCT4 and established HIF1α as an LY294002 effector. Another target of the PI3K/AKT signaling pathway is FOXO3a, which controls cellular quiescence and is inactivated by AKT . We found that the expected inhibition of AKT phosphorylation (pAKT) by LY294002 was associated with an unexpected increase in the level of phosphorylated FOXO3a (pFOXO3a) (Fig. 5B) normally destined for degradation in the cytoplasm. Interestingly, vital confocal microscopy of LY294002-treated Rh123low and Rh123high Mel4M cells (see the representative image in Fig. 5C) revealed that pFOXO3a accumulated in the majority (63.3%) of the nuclei of Rh123low cells, whereas in 80.9% of Rh123high cells displayed greatly decreased levels. Overall, we demonstrated that the Rh123low phenotype is associated with melanoma SC-like activity and that these cells overexpressed HIF1α in a PI3K/AKT-dependent manner and preferentially retained pFOXO3a in their nuclei independent of pAKT. In contrast, the Rh123high cells had low levels of both HIF1α and nuclear pFOXO3a. Taken together, these findings suggest a functional link between the Rh123low phenotype of melanoma SCs, high levels of HIF1α, and the nuclear localization of pFOXO3a.
Whether melanoma tumors are organized hierarchically and whether they contain stem cells responsible for tumor development and relapse is still debated (see recent review, ). We applied a functional Rhodamine123 (Rh123) efflux assay to demonstrate that metastatic melanoma and melanoma cell lines contain a small subset of low-Rh123-retention (Rh123low) cells with stem cell-like activities including a superior ability to self-renew, generate tumors in vivo and to repopulate tumor-like spheres with progeny recapitulating the phenotypic heterogeneity of the parental tumors. Rh123 exclusion assay has been used successfully to distinguish the most primitive Rh123low HSCs resting in the G0 phase of the cell cycle from the less primitive actively cycling Rh123high subpopulation [41, 24]. In the present study, using this assay, we determined that also the population of melanoma stem cell varies in size and is composed of both quiescent Rh123lowPyronin (PYlow) and slow-cycling Rh123lowPYhigh subpopulations. We also identified a third subpopulation of melanoma stem cell that is capable temporarily switch from the actively cycling Rh123highPYhigh to Rh123lowPYhigh phenotype upon LY294002 treatment. These findings suggest that melanoma stem cells, as HSCs  may coexist in three different states: (a) quiescent, preserving long-term self-renewal abilities, (b) slow-growing, reminiscent of the slow-cycling melanoma cell subpopulation recently identified by Roesch et al.  and (c) actively proliferating state feeding into cellular mass of a tumor or cell cultures. The low expression of proliferation markers and the high expression of pAKT, OCT4, and HIF1α characterize the Rh123low cells. HIF-1α, a downstream target of pAKT controls cellular quiescence through c-Myc inactivation and the repression of oxidative phosphorylation [44–46] and may also be responsible for controlling quiescence of melanoma stem cells whereas OCT4 for their maintenance  and the undifferentiated state . Thus, the lower levels of HIF1α and OCT4 transcription factors in the Rh123high cells may retain cells in cycling state and contribute to cellular heterogeneity in melanoma tumors and in tumor-like melanospheres consisting mainly of Rh123high cells. Accordingly, Rh123high cells proliferate more quickly than Rh123low cells but display limited tumor-like sphere initiation and maintenance abilities. Clearly, these abilities may be intrinsic to the slow-cycling Rh123low subset of melanoma stem-like cells.
Interestingly, the size of the Rh123low compartment, although constant under defined conditions, could be enlarged by the LY294002 inhibition of the PI3K/AKT signaling pathway. We found that a transient deactivation of AKT induces both the shift of a fixed number of flexible Rh123high cells into the Rh123low population and the slight but significant increase in HIF1α expression, implying that the phenotype switching mechanism, previously described to operate in melanomas [3, 43], may contribute to enlargement of the Rh123low cell compartment and may require HIF1α. However, simultaneously, the proportion of Rh123low cells accumulating in the G1 phase of the cell cycle increased in this compartment at the expense of G0 cells, suggesting that the inhibition of the PI3K/AKT signaling pathway renders them less quiescent. Thus, the combination of both the phenotype switching mechanism and the activation of quiescent Rh123low cells appears to increase the proportion of cycling Rh123low stem cells upon the temporal deactivation of AKT. Consequently, the capacity of Rh123low cells to form tumor-like melanospheres is augmented, implying that the cycling fraction of Rh123low cells is also responsible for tumor development in vivo. These cellular changes were associated with molecular modifications leading to the upregulation of ABCB1/ABCB5 transporter activity, the downregulation of pAKT and its target HIF1α, and an intriguing accumulation of nuclear pFOXO3a that awaits further investigation.
Although it is still unknown how this AKT/HIF1α/pFOXO3a-mediated molecular machinery governs a quick reentry of Rh123high cells into the Rh123low compartment and the rapid transition of melanoma stem cells from the quiescent G0 state to the G1 phase of the cell cycle, our preliminary data indicate that these events can be uncoupled upon a long-term exposure to LY294002. While the proportion of the phenotype switching cells could not be increased above the short-term size, the sustained deactivation of PI3K/AKT signaling seems to restrain the Rh123low quiescent cells in the G0-phase of the cell cycle and interestingly, diminishes the pFOXO3a levels below the background (not shown). Is then, pFOXO3a responsible for the rapid reentry of Rh123low quiescent cells into the cell cycle in addition to HIF1α? The answer to this question awaits further studies but recent findings of Sykes et al. [48, 49] determined that nuclear pFOXO's are responsible for maintenance of leukemia initiating cells.
We were surprised that short-term inactivation of PI3K/AKT in melanoma cells was sufficient to stimulate the G0 to G1 transition. However, this is in agreement with the recent findings that cells respond differently to a short- and long-term AKT inhibition  and rapidly, within 2 hours, to Cyclin C activation [51, 52]. The study showing that the activation of quiescent muscle satellite cells is regulated at the RNA post-transcriptional level  may provide some insight into mechanisms responsible for a rapid response to short-term environmental changes.
The swiftness of the LY294002-induced changes argues that a phenotypic switch of a small fraction of Rh123high cells may also allow for the rapid adjustment of melanoma cells to varying environmental cues. These cells are most likely poised to readily upregulate ABCB1/ABCB5 transporter activity in response to the PI3K/AKT signaling pathway inactivation. Thus, this pathway has opposing effects on ABCB1/ABCB5 transporters in melanoma (based on our results) and on the ABCG2 transporter conferring the side-population phenotype in glioma . This contrast implies that the function of ABC transporters may be regulated in a transporter- and cell-specific manner. We do not yet know the mechanisms by which the PI3K/AKT signaling pathway may inactivate ABCB1/ABCB5 in melanoma, but transporter trafficking  and/or phosphorylation may be involved .
We assume that the activation of the PI3K/AKT signaling pathway in the Rh123low stem-like cells inactivates their phenotype-conferring ABCB5 transporters. Consequently, Rh123low stem-like cells lose their Rh123low phenotype and acquire the properties of the Rh123high cells. Recently, it was reported that the fast-growing MCF7 breast SCs acquire the slow-cycling activity of SCs by asymmetric suppression of the AKT pathway . Our data showing that inactivation of the PI3K/AKT signaling pathway promptly generates slow-cycling Rh123low cells by a phenotype switch are consistent with these findings. However, as LY294002 also renders the Rh123low G0 cells transiently less quiescent, it appears that the PI3K/AKT signaling pathway has a dual function in melanoma, both maintaining the quiescence of Rh123low G0 stem-like cells but accelerating the proliferation of their cycling subset.
In conclusion, our results reported here suggest a model of melanoma stem cell compartmentalization with a specific mode of response of each compartment to environmental changes. We determined that melanoma contains a rare subpopulation of Rh123low stem-like cells existing in a quiescent and slow-cycling state and suggest that the PI3K/AKT signaling pathway maintains their quiescence but promotes the reversible exit of Rh123low cells from the stem-like state. Clearly, inactivation of the PI3K/AKT signaling pathway differentially affects quiescent and cycling melanoma cells, suggesting a “no-win” situation for drugs targeting this pathway. The PI3K/AKT downstream target HIF1α [57, 58] and nuclear pFOXO3a [48, 49] may be the main effectors of these opposite PI3K/AKT functions. The dual and reversible activity of this pathway appears to be a crucial factor contributing to cellular plasticity and may be related to the unusual resistance of melanomas to anti-tumor treatments.
This research was supported by the Institut National de la Santé et de la Recherche Médicale (INSERM) and by a grant from the Institut National du Cancer (National Cancer Institute) of France, the Ligue Nationale Contre le Cancer, the Institut Pour la Recherche sur le Cancer de Lille (IRCL) and by SILAB-Jean Paufique Corporate Foundation. This work is a part of the tumor dormancy OncoLille program of SIRIC. We thank Malgorzata Czaradzka for her technical assistance on the project and Meryem Tardivel (BiCell-IFR114) for her expert technical assistance with the confocal microscopy work. K.K. is currently affiliated with the INSERM UMRS 940, Paris, France; Institut Universitaire, Hématologie, Université Paris 7 Denis Diderot, Paris, France; I.W. is currently affiliated with the Laboratory of Genomics and Metabolic Diseases, CNRS UMR8199 Institut Pasteur de Lille, BP245, F-59019, Lille, France; H.L.R. is currently affiliated with the CNRS URM8161, Institut Pasteur de Lille, 1 rue Calmette, BP245, 59019 Lille, France; T.Z. is currently affiliated with the UTCG,RDJ- Institut de Biologie, 44093 Nantes, France.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors declare that no competing financial interests exist.