MYC Regulates the HIF2α Stemness Pathway via Nanog and Sox2 to Maintain Self-Renewal in Cancer Stem Cells versus Non-Stem Cancer Cells

A hallmark of many tumors is the capacity to maintain a stable population of cancer stem cells (CSC) during multiple generations (1). This is attributed to CSC's ability to undergo asymmetric cellular division, where one daughter cell retains self-renewal ability, whereas the other daughter cell differentiates into non-CSCs, composing the bulk of the tumor (2). Numerous studies demonstrate that CSCs retain this ability of selective or exclusive self-renewal through asymmetric cellular division even after numerous serial transplantations and maintain a stable proportion of CSCs (3, 4). Hence, this maintenance of a stable proportion of CSCs via asymmetric division suggests a revision in the notion of the clonal evolution in cancer (2, 3, 5, 6).

Various mechanisms have been proposed by which CSCs maintain asymmetric self-renewal, including cell polarity, fate determinants, microenvironment modulation (4, 7, 8), phenotypic equilibrium (9), and activation of developmental pathways such as Notch and Wnt (1, 3, 4, 10). In addition, gene products that can confer self-renewal in cancer have been identified including the iPS gene products MYC, Nanog, Sox2, Oct-4, as well as hypoxia-inducible factors (HIF; refs. 11–23). However, it is not clear how MYC and other iPS genes cooperate with HIFs to maintain self-renewal in CSCs versus non-CSCs.

The MYC oncogene plays an important role in the self-renewal of normal stem cells and CSCs (22, 24, 25). MYC is a transcription factor that regulates gene expression. When overexpressed, MYC commonly contributes to human cancer (11, 14). MYC induces an embryonic stem cell signature in CSCs (26). While in cooperation with other iPS genes such as Sox2, Nanog, and Oct-4, MYC elicits reprograming of differentiated cells enabling self-renewal (27) and thereby modulating the iPS genes (19, 28). MYC cooperate with HIF2α (29, 30), a stemness-associated transcription factor that increases self-renewal of embryonic stem cells through coordinated upregulation of Oct-4 and Nanog (31, 32), and the negative regulation of p53 (33). Hence, MYC through interaction with HIF2α and iPS genes could regulate exclusive self-renewal of CSCs.

We investigated self-renewal of CSCs in a transgenic mouse model of MYC-induced T-cell acute lymphocytic lymphoma (T-ALL; refs. 34, 35) and human lymphoma. In MYC-induced T-ALL, we identified Sca-1⁺ CSCs that exhibit dependency on HIF2α for self-renewal. In CSCs but not non-CSCs, MYC preferentially binds to the promoter and activates transcription of HIF2α that is facilitated by Nanog and Sox-2. Finally, MYC mediated activation of HIF2α in ABCG2⁺ but not ABCG2⁻ human lymphoma CSCs. Our observations thereby suggest that MYC maintains exclusive self-renewal of CSCs by preferential activation of HIF2α in CSC versus non-CSCs.

Details of Materials and Methods are provided in the Supplementary Method section.

All the necessary experimental procedures were approved and undertaken in accordance with guidelines of Stanford University (Stanford, CA), Forsyth Institute (Cambridge, MA), Gauhati University (Guwahati, Assam, India), and KaviKrishna Laboratory (Guwahati, Assam, India) institutional animal ethics committee. Seven such transgenic mice were selected for the study and genotype confirmed (Supplementary Table S1). The generation and genotyping of Eu-tTA/tetO-MYC system transgenic lines for conditional MYC-driven lymphoma has been used as described previously (34). The thymus obtained from moribund animals was dissociated to flow cytometry or immunomagnetic sort Sca-1⁺ cells (36), and these cells were expanded in serum-free media containing IL7 and stem cell factors, and then subjected to phenotypic analysis. Multi-color flow cytometry for HIF2α (#NB100-132, Novus Biologicals) and Nanog (#ab184609, Abcam) was done as described previously (33).

Glutathione (GSH) and reactive oxygen species (ROS) levels were measured as described previously (37), whereas apoptosis was measured by calorimetric assay as per the manufacturer's instructions. Analysis of the relative cell number was performed by using Alamar blue assay as described previously (37).

The fluorescence method using 5-dodecanoylaminofluorescein di-β-d-galactopyranoside (C₁₂FDG; Thermo Fisher Scientific, #D2893) was used as described previously (38). MYC-inactivated lymphoma cells served as positive control for senescence (39). Details are given in Supplementary Materials and Methods.

It was performed in methylcellulose medium (Methocult M3134, Stem Cell Technologies) as described previously (15, 37).

IHC was done using a Vector Mouse on Mouse Elite Peroxidase Immunodetection Kit (Vector Laboratories, Inc.) according to the manufacturer's instructions (40). Mouse HIF2α (Novus Biologicals) antibody was used in 1:500 dilution. Western blot analysis was done as previously described by using mouse HIF2α (#NB-132; Novus Biologicals) and mouse β-actin (Cell Signaling Technology; ref. 40).

The real-time PCR (qPCR) was performed using TaqMan gene expression assay as described previously (33). The in vitro inhibition of mouse HIF2α, HIF1α, p53, Sox2, and Nanog were achieved by Accell siRNA obtained from GE Healthcare Dharmacon Inc. as described previously (33). The in vivo inhibitions of HIFs and p53 were achieved by FM19G11 (Millipore; 5 mg/kg i.p.) and pifithrin α (Sigma-Aldrich; 4 mg/kg i.p.) after dissolving in 1.5 % DMSO, and given three times weekly for 2 weeks (41). Inhibition of ROS was accomplished by N-acetylcysteine (NAC; 100 mg/kg i.p.) 30 minutes before the injection of FM19G11.

Chromatin immunoprecipitation (ChIP) was performed using the Magnify chromatin IP Kit (#1626969, Invitrogen) according to the manufacturer's instructions. Briefly, Sca-1⁺ cells were fixed with 1% formaldehyde, sonicated to produce DNA fragments of approximately 100–600 bp and then subjected to immunoprecipitation (19) with c-MYC (2 μg, N-262, #sc-764, Santa Cruz Biotechnology), as well as isotype-matched control for rabbit IgG (NB810-56910; Novus Biologicals) as described previously (42). The mouse and human DNA products were subjected to SYBR green PCR using EpiTect ChIP qPCR Primers (SABiosciences, Qiagen). Results were analyzed using percent input method, where 1% of starting chromatin was used as input (42). For re-ChIP analysis, chromatin products from first immunoprecipitation were treated with 10 mmol/L DTT for 30 minutes at 37°C to prevent the majority of the first antibody from participating in the second immunoprecipitation reaction. Eluates were diluted in dilution buffer and then used for second immunoprecipitation by Nanog (#5232, Cell Signaling Technology) rabbit antibody, as described previously (42).

Peripheral blood and lymph node samples from patients with leukemia (Supplementary Table S2) were obtained after written informed consent and with the approval from the institutional research ethics committee of respective institutions and Declaration of Helsinki. The approved institutions are Stanford University School of Medicine (Stanford, CA), KaviKrishna Telemedicine Care, a branch of Kavi Krishna Laboratory, Dr. B. Borooah Cancer Institute, and Gauhati University (Guwahati, Assam, India). The collection and expansion of T-cell lymphoblastic lymphoma (T-LBL) samples (n = 6, Supplementary Table S2), immunomagnetic sorting of ABCG2⁺, and in vivo transplantation in NOD/SCID mice were performed as described previously (20, 43, 44). Mice were prior treated with 22 mg/kg i.p. busulfan (6 mg/mL injection, Taj Pharma) daily over 3 days (20). The human cell engraftment was confirmed by flow cytometry staining with a human CD45 antibody (BioLegend).

ABCG2⁺ cells were transfected using the JetPEI Reagent (Polyplus transfection) with plasmids encoding constitutively active HIF2α mutant, the HA-HIF-2α-P405A/P531A (gift from Professor William Kaelin, Harvard Medical School, Cambridge, MA; Addgene plasmid #18956; ref. 45) and confirmed by Western blot analysis. The control circular plasmid encoding HA with no HIFα insert was generated as described previously (46).

The statistical calculations were performed with GraphPad Prism 4.0 (Hearne Scientific Software) using Student t test and One-Way ANOVA with Dunnett post-hoc test. The data on the in vivo limiting dilution assay was analyzed by the extreme limiting dilution analysis software, available online (http://bioinf.wehi.edu.au/software/elda/). Statistical comparison of Kaplan–Meier curves was based on the log-rank test.

In a transgenic mouse model of Tet system regulated MYC-induced T-ALL, we have examined whether there was an identifiable CSC population (34) that activates HIF2α pathway (Fig. 1A). We used multiple transplanted T-ALL tumors (D9476, D9482, E0366, E2824, E2825, E6550, and D9479). We confirmed that all tumors exhibited clonal expression of the T-cell receptor Vβ 2 or 4 (Supplementary Table S1), as demonstrated previously (34).

To identify potential CSC markers, we first examined the expression of multiple thymic lymphocyte lineage markers. T-cell development in the thymus involves migration of bone marrow cells that differentiate into early thymocyte progenitors (ETP) and express Sca-1, c-kit, and CD44 (36). We found that a small number of tumor cells expressed Sca-1 (Fig. 1B, left), c-kit, and CD44 (Supplementary Fig. S1A; Supplementary Table S3A and S3B). The Sca-1⁺ cells were positive for both CD4 and CD8, but did not express B-cell maker, CD19 (Fig. 1B, right; Supplementary Table S3A; ref. 34). Hence, a small number of tumor cells expressing Sca-1 could be identified as potential CSC population.

Next, we examined whether Sca-1⁺ versus Sca-1⁻ cells differentially expressed other stem cell genes including: HIF2α, and the pluripotency factors Nanog, Sox2, and Oct4 (33). Sca-1⁺ cells exhibited 6- to 10-fold higher expression of all of these genes except Oct4, which was not significantly expressed (Fig. 1C; Supplementary Fig. S1A and S1B; Supplementary Table S3B). HIF1α and p53 were downregulated by 4- to 6-fold, whereas the level of endogenous MYC remained consistent in Sca-1⁺ versus Sca-1⁻ cells (Fig. 1C; Supplementary Fig. S1C). Flow cytometry confirmed the qPCR results (Supplementary Fig. S1A and S1B; Supplementary Table S3A and S3B). Thus, Sca-1⁺ versus Sca-1⁻ cells exhibited differential expression of stemness associated gene products.

Next, we evaluated the self-renewal capacity of Sca-1⁺ cells by performing in vivo limiting dilution assay. Flow cytometry–sorted Sca-1⁺ and Sca-1⁻ cell population from parental tumors were transplanted into immunocompetent syngeneic mice. Intravenous injection of 10 Sca-1⁺ cells but 1 × 10⁶ Sca-1⁻ cells resulted in engraftment of lymphoma in recipient syngeneic mice (Fig. 1D, left; Supplementary Fig. S2A–S2C; Supplementary Table S4). The maximum likelihood analysis revealed that Sca-1⁺ cells exhibit 5,724-fold higher tumorigenic capacity in comparison with Sca-1⁻ cells (P < 0.00001; Supplementary Table S4). These Sca-1⁺ cells–derived primary tumors reestablished parental heterogeneity. The expression of ETPs and stemness markers CD44, C-kit, HIF-2α, Nanog, and Sox2 were consistently high in Sca-1⁺ versus Sca-1⁻ cells. Whereas the expression of p53 and HIF1α remained low, similar to parental tumors (Supplementary Table S3B). Thus, we have identified Sca-1⁺ cell population enriched with CSCs in MYC-induced lymphoma.

To evaluate whether Sca-1⁺ cells maintain stable frequency and self-renewal state during serial transplantation, a monoclonal transplantation assay was performed where tumors were derived from a single CSC (47). For this purpose, Sca-1⁺ cells were isolated from multiple primary tumors: D9476, E6550, D9479, and E0366 (Fig. 1D left; Supplementary Fig. S2A–S2C). A single Sca-1⁺ cell from these primary tumors was then injected to syngeneic mice to obtain secondary tumors or further reinjected to obtain tertiary tumors. We found secondary and tertiary tumors derived from each of the four parental tumors maintained stable proportion of Sca-1⁺ cells, and their phenotype including the high HIF2α protein expression in comparison with Sca-1⁻ cells (Fig. 1D and E; Supplementary Fig. S2D; Supplementary Table S3A and S3B). In addition, Sca-1⁻ cells from these tumors expressed TCR V β expression as parental Sca-1⁺ cells (Supplementary Tables S1 and S3C). Then, Sca-1⁺ and Sca-1⁻ cells from these primary, secondary, and tertiary tumors were subjected to in vivo limiting dilution assay to obtain CSCs proportions. We found that the ratio of CSC frequency between Sca-1⁺ versus Sca-1⁻ cells remained stable. (Fig. 1F; Supplementary Table S4). Our results suggest that Sca-1⁺ CSCs maintain stable proportion during serial transplantation.

A stable proportion of CSC may be maintained either via asymmetric self-renewal or by conversion of non-CSC to CSCs by the process of phenotypic equilibrium (7, 9). To address this possibility, we subjected Sca-1⁻ cells from primary, secondary, and tertiary tumors (Fig. 1D–F) to in vivo limiting dilution assay. Interestingly, CSC frequency did not become elevated during serial transplantation (Supplementary Table S4), suggesting the unlikely contribution of Sca-1⁻ cells to CSC frequency.

Inactivation of MYC by doxycycline treatment downregulates HIF2α without significantly changing HIF1α in Sca-1⁺ cells (Supplementary Fig. S3A), suggesting that HIF2α pathway may be required to maintain the stemness state of Sca-1⁺ cells. By Flow cytometry, the expression of HIF2α, Nanog, and Sox2 was found to be consistently higher in Sca-1⁺ cells of serially transplanted tumors, while the expression of HIF1α and p53 remained low in the Sca-1⁺ cells (Fig. 2A; Supplementary Table S3B). Thus, Sca-1⁺ cells maintained high HIF2α during serial transplantation. In contrast, the Sca-1⁻ cells maintained low HIF2α expression (Supplementary Fig. S2D). Thus, serially transplanted Sca-1⁺ cells but not Sca-1⁻ cells maintain high HIF2α, suggesting the potential role of HIF2α in the exclusive self-renewal of Sca-1⁺ versus Sca-1⁻ cells.

Hence, we evaluated the role of HIF2α in the self-renewal of Sca-1⁺ cells. Suppression of HIF2α but not HIF1α expression by siRNA resulted in decreased Nanog and Sox2, but increased p53 expression (Fig. 2B; Supplementary Fig. S3A and S3B). These changes were associated with decreased HIF2α protein levels, in vitro cell growth, but no change in cell death (Fig. 2B and C; Supplementary Fig. S3C and S3D). Suppression of HIF2α but not HIF1α was associated with changes in senescence and differentiation markers including reduced BrdU incorporation and increased β-galactosidase staining (Fig. 2D) p21, and p16 (39) and decreased expression of CD44 and C-kit (Supplementary Fig. S3E). Apoptosis measured by caspase-3 activity was not altered (Fig. 2D). Therefore, HIF2α is required to maintain the stemness state of Sca-1⁺ cells.

In embryonic stem cells, HIF2α negatively regulated p53 and ROS to maintain an undifferentiated state of high GSH redox state (33). We explored whether HIF2α employs similar mechanisms of maintaining low p53 and ROS levels in Sca-1⁺ cells. Flow cytometry–based evaluation of p53, ROS, and GSH were performed in Sca-1⁺ cells subjected to HIF2α silencing. Interestingly, suppression of HIF2α indicated decreased GSH, whereas increased p53 and ROS levels in Sca-1⁺ cells. Furthermore, these changes were accompanied by loss of Sca-1 expression (Fig. 2E and F; Supplementary Table S5A). Therefore, HIF2α appears to negatively regulate p53 and ROS for maintaining the stemness state of Sca-1⁺ cells. Indeed, siRNA p53 gene silencing blocked HIF2α silencing from inducing proliferative arrest or senescence in Sca-1⁺ cells (Fig. 2G and H; Supplementary Fig. S3F). Similarly, inhibition of ROS by cotreatment with NAC blocked p53 induction following HIF2α silencing (Supplementary Fig. S3F) and proliferative arrest or senescence (Fig. 2H). Hence, HIF2α suppresses both p53 and ROS to maintain proliferation of Sca-1⁺ cells.

Next, we examined whether HIF2α was required for the self-renewal of Sca-1⁺ cells by injecting these cells to immunocompetent mice. We measured in vivo tumorigenic growth and self-renewal of HIF2α-silenced Sca-1⁺ cells with or without p53 silencing or NAC treatment. HIF2α silencing in Sca1⁺ cells decreased tumorigenic growth by 115-fold (Fig. 3A and B; Supplementary Table S5B) and self-renewal by 11-fold (Fig. 3C–E; Supplementary Table S5C and S5D). However, concomitant NAC treatment or p53 silencing restored both tumorigenic growth and self-renewal (Fig. 3B–E; Supplementary Table S5C and S5D). Importantly, HIF2α silencing did not prevent Sca-1⁺ cells from entering the thymus (Supplementary Fig. S4A–S4C). Thus, HIF2α appears to maintain the self-renewal of Sca-1⁺ cells by negative regulation of ROS and p53.

We examined whether the in vivo inactivation of HIF2α through the small-molecule FM19G11 influenced tumor growth. Tumors in mice treated with FM19G11 showed increased survival comparable with vehicle-treated group (Fig. 4A). Importantly, Sca-1⁺ cells from FM19G11-treated tumors indicated increased p53 but decreased HIF2α, as well as self-renewal capacity of Sca-1⁺ cells (Fig. 4B and C; Supplementary Table S6Aand S6B). FM19G11 combined with NAC reduced the p53 and ROS levels, increased GSH and rescued the self-renewal capacity of Sca-1⁺ cells (Fig. 4 D–F; Supplementary Table S6A and S6B). This further supports the conclusion that HIF2α is required for self-renewal of Sca-1⁺ cells.

However, we noted that 4 of 10 tumor-bearing mice treated with FM19G11 exhibited morbidity within 20–30 weeks (Fig. 4A), suggesting continuing growth of thymic tumors despite treatment. Sca-1⁺ versus Sca-1⁻ cells recovered from these tumors maintained the expression of high HIF2α and low p53, whereas MYC level remained equal (Fig. 4G). Furthermore, Sca-1⁺ versus Sca-1⁻ cells of these post-FM19G11–treated tumors maintained high tumorigenic capacity (Supplementary Table S6C). This suggests that Sca-1⁺ cells that escape sensitivity to FM19G11 did not evolve to a different phenotype but maintained the high HIF2α phenotype.

We examined MYC's role in the exclusive regulation of HIF2α in Sca-1⁺ versus Sca-1⁻ cells. MYC was found to be equally expressed in Sca-1⁺ and Sca-1⁻ cells, obtained from a single Sca-1⁺ cell–derived tumor #D9476 (Figs. 1D and 5A) and expressing TCR V β 4 (Supplementary Table S3C). We infer that other transcription factors may be required for differential regulation of HIF2α in Sca-1⁺ versus Sca-1⁻ cells. One possibility was that the iPS genes Nanog and Sox2 might be involved in this MYC-mediated regulation of HIF2α, because they both were found to be highly expressed in Sca-1⁺ but not Sca-1⁻ cells (Fig. 1C; Supplementary Fig. S1A–S1C; Supplementary Table S3B). Indeed, we found that MYC extensively binds to the promoters of HIF2α, Nanog, and Sox2 but not Oct4 in Sca-1⁺ versus Sca-1⁻ cells (Fig. 5B). Notably, MYC binds preferentially to the HIF1α promoters in Sca-1⁻ versus Sca-1⁺ cells (Fig. 5B). Suppression of MYC expression decreased HIF2α, Nanog, and Sox-2, but upregulated HIF1α expressions in Sca-1⁺ versus Sca-1⁻ cells (Fig. 5C). Thus, MYC preferentially bound and regulated the expression of HIF2α, Sox-2, and Nanog in Sca-1⁺ versus Sca-1⁻ cells.

Next, we examined whether MYC regulation of HIF2α expression was mediated by Nanog and/or Sox-2. We found that co-siRNA silencing of Nanog and Sox2 decreased MYC binding to HIF2α, reduced HIF2α protein levels by 4-fold, and decreased the self-renewal capacity without altering MYC protein level (Fig. 5D–G; Supplementary Fig. S5A–S5C). Thus, MYC cooperates with Nanog and Sox2 to increase HIF2α expression and regulate self-renewal in Sca-1⁺ versus Sca-1⁻ cells.

To confirm our findings from transgenic mouse model system in a human equivalent tumor, we first identified the equivalent ABCG2⁺/HIF2α CSC subpopulation in relapsed cases of human adult acute lymphoblastic lymphoma (T-LBL). ABCG2, a stem cell marker regulated by HIF2α (33), was highly expressed in patients with T-LBL treated with chemotherapy/radiation (48). Our results indicated that T cells recovered from cervical lymph nodes and peripheral blood of the subjects (n = 6; Supplementary Table S2) were highly enriched in ABCG2⁺/HIF2α⁺ cells (Supplementary Fig. S6A–S6C). These ABCG2⁺ cells were CD7-positive but CD8- and CD1a-negative (Fig. 6A; Supplementary Fig. S6D), a phenotype associated with T-ALL CSCs (49).

Serial transplantation assay in NOD/SCID mice showed a 2,000-fold higher self-renewal capacity of ABCG2⁺ versus ABCG2⁻ cells (Fig. 6B and C; Supplementary Fig. S7A–S7D). Phenotypically, ABCG2⁺ cells represented undifferentiated lymphoma gene signature with high expression of MYC, HIF2α, Nanog, Sox2, and CD44. p53 expression was lower in ABCG2⁺ versus ABCG2⁻ cells (Fig. 6D), thereby indicating similar CSC property and stemness phenotype to Sca-1⁺ cells. Importantly, silencing of HIF2α but not HIF1α in ABCG2⁺ cells exhibited similar results that we observed in Sca-1⁺ CSCs (Fig. 6E–G; Supplementary Fig. S8A and S8B). The results suggest that HIF2α maintains the self-renewal of ABCG2⁺ cells in human equivalent cancer.

We furthered our findings and revealed that MYC differentially binds to promoter of HIF2α in ABCG2⁺ versus ABCG2⁻ cells while directly regulates the Nanog and Sox2 binding. Inhibition of MYC led to marked decrease of HIF2α, Nanog, and Sox2 expression in ABCG2⁺ cells (Fig. 7A and B; Supplementary Fig. S8C), indicating that MYC and HIF2α may mediate self-renewal of ABCG2⁺ cells.

To confirm the role of Nanog and Sox2 in the MYC-mediated regulation of HIF2α in ABCG2⁺ versus ABCG2⁻ cells, we performed ChIP assay. Inhibition of Nanog but not Sox2 led to marked decrease of MYC binding to HIF2α promoter (Fig. 7C; Supplementary Fig. S8C). Importantly, Nanog binding to the HIF2α promoter was observed in re-ChIP samples of MYC (Fig. 7D), suggesting that MYC and Nanog may cooperate to regulate HIF2α expression. Furthermore, in Nanog-silenced ABCG2⁺ cells, while MYC protein levels remained unaltered, the HIF2α protein and GSH levels were markedly reduced, p53 and its target genes p21, MDM2, and BAX were induced, and the self-renewal was negatively affected as measured by serial transplantation assay (Fig. 7E–G; Supplementary Fig. S8D). Interestingly, Nanog's role in ABCG2⁺ cell self-renewal was not independent of HIF2α, because Nanog knockdown ABCG2⁺ cells could be rescued by the constitutive expression of a HIF2α transgene (Fig. 7F–G; Supplementary Fig. S8E; ref. 45). These results therefore suggest that Nanog participated in MYC regulation of HIF2α in ABCG2⁺ versus ABCG2⁻ cells to confer exclusive self-renewal capacity and tumor stemness.

CSCs contribute to chemotherapy failure and tumor relapse (2, 3). How CSCs maintain exclusive self-renewal is not clear. Here, we report that Sca-1⁺ CSCs in transgenic mouse model of MYC-driven lymphoma and ABCG2⁺ CSCs from lymphoma subjects demonstrate exclusive self-renewal via specific MYC binding to HIF2α promoter regions specifically in CSCs versus non-CSCs. Furthermore, stemness factors Nanog and Sox2 cooperates with MYC to regulate HIF2α that in turn decreases p53 expression and reduces ROS levels in CSCs. Thus, MYC-HIF2α stemness pathway (Fig. 7H) may contribute to the exclusive self-renewal mechanism of CSCs and regulate their frequency during multiple generations.

HIFs signaling have a complex role in cancer self-renewal, where both HIF1α and HIF2α either promote or inhibit the self-renewal of leukemia (10, 20, 21, 50). MYC is a key regulator of pluripotency and differentiation that acts downstream of NOTCH (51), a developmental pathway that cooperates with HIFs to maintain self-renewal of T-ALL (43, 49, 50, 52, 53) Also, MYC maintains self-renewal in stem cells and CSCs through effects on cellular metabolism (14, 24, 29). How MYC cooperates with HIFs to regulate CSC self-renewal, as well as frequency is less understood.

To investigate the role of HIF2α in the self-renewal of MYC-dependent CSCs, we characterized Sca-1⁺ CSCs in mouse, and human T-lymphoblastic lymphoma patient–derived ABCG2⁺ CSCs. In these CSCs, HIF2α a transcription factor known to maintain stem cell robustness (15), negatively regulates p53 expression and ROS levels to confer self- renewal capacity. Importantly, our findings reveal that high HIF2α expression in Sca-1⁺ CSCs is enabled by distinct MYC binding to HIF2α promoter regions. Pluripotency factors such as Nanog and Sox2 together facilitated this binding, thereby conferring selective or exclusive self-renewal capacity to CSCs versus non-CSCs. Nanog and Sox2 are broadly expressed in human cancers (15, 54, 55). Nanog is known to maintain self-renewing T-cell leukemia cells by suppressing p53 (13), however the molecular mechanism was not clear. Now, our findings in human ABCG2⁺ CSCs suggest that Nanog-mediated p53 suppression was HIF2α dependent. Furthermore, in Sca-1⁺ cells, we found that Nanog cooperates with Sox2 to facilitate MYC-mediated selective upregulation of HIF2α in CSCs versus non-CSCs. Selective or exclusive MYC binding to HIF2α promoters in human ABCG2⁺ CSCs versus non-CSCs was facilitated by Nanog alone. Further experiments are required to understand the independent role of Nanog in self-renewal of human CSCs.

Involvement of the iPS factors Oct4, Nanog, and Sox2 in CSC self-renewal may be tumor type specific (6, 15). Only Nanog was required to regulate human ABCG2⁺ CSCs self-renewal, whereas both Nanog and Sox2 regulated the mouse CSCs self-renewal. Also, MYC protein levels were higher in human ABCG2⁺ CSCs as compared with mouse CSCs. Interestingly, Oct4, a downstream mediator of HIF2α pathway (31, 33) was neither involved in human or mouse T-ALL–derived CSCs. These observations indicate that the role of these stemness effectors might be tumor specific (6, 11, 12, 15, 17).

Our study is suggestive, but does not confirm the possibility of CSCs asymmetric self-renewal at single-cell level. Indirect evidences indicate that single Sca-1⁺ cell may give rise to tumors, representing a clonal population of both Sca-1⁺ and Sca-1⁻ cells with distinct phenotype. Stemness factors, Nanog and Sox2, were involved in regulating HIF2α expression in Sca-1⁺ versus Sca-1⁻ cells, suggesting the possibility of asymmetric self-renewal and maintenance of CSC frequency. Furthermore, previous observations illustrate that tumors maintain a stable population of CSCs (1) that may be attributed to asymmetric cellular division (2). Similarly, in a MYC-induced T-ALL model in zebrafish and in human CD7⁺ lymphoma stem cells, CSCs maintained their frequency during serial transplantation (47, 49). Therefore, CSCs may maintain their frequency by undergoing asymmetric self-renewal. Our T-ALL CSCs model of Sca-1⁺ cell self-renewal may serve as a model to understand the underlying mechanisms that maintain asymmetric self-renewal in cancer.

We identified that MYC and HIF2α negatively regulate p53 and ROS in CSCs and this appears to be required to maintain stemness. Nanog and Sox2 interacted with MYC and HIF2α to maintain balance between stemness and differentiation (Fig. 7H). Thus, targeting this MYC-HIF2α stemness pathway could be a targeted therapy against CSCs.

No potential conflicts of interest were disclosed.

Conception and design: B. Das, B. Pal, D.W. Felsher

Development of methodology: B. Das, B. Pal, D.W. Felsher

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Das, B. Pal, R. Bhuyan, H. Li, A. Sarma, S. Gayan, J. Talukdar, S. Sandhya, S. Bhuyan, G. Gogoi, D. Baishya, J.R. Gotlib

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Das, B. Pal, S. Gayan, A.M. Gouw, A.C. Kataki, D.W. Felsher

Writing, review, and/or revision of the manuscript: B. Das, B. Pal, R. Bhuyan, H. Li, S. Gayan, A.M. Gouw, A.C. Kataki, D.W. Felsher

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Das, A. Sarma, J. Talukdar, S. Sandhya, S. Bhuyan, G. Gogoi, D. Baishya, J.R. Gotlib, D.W. Felsher

Study supervision: B. Das, D.W. Felsher

We thank members of Felsher laboratory, the Stanford flow cytometry facility, Animal Facility, and Laboratory of Immunology and Vascular Biology at the Palo Alto VA Health Care System, CA, Forsyth Institute, Cambridge, MA, and KaviKrishna laboratory, Guwahati Biotech Park, Indian Institute of Technology, Guwahati, India. This research project was funded by grants from the Canadian Cancer Society (to B. Das), Laurel Foundation (to B. Das), and KaviKrishna Foundation, Assam, India (to B. Das). Additional funding was obtained from the Bill & Melinda Gates Foundation through the “Grand Challenges Exploration Initiatives” (to B. Das), Stem Cell Altruism Fund, Thoreau Laboratory for Global Health, University of Massachusetts, Lowell (to B. Das), NIH grants R01CA105102, CA89305-0351, and CA112973 (to D.W. Felsher), Department of Defense grant PR080163 (to D.W. Felsher), Emerson Collective Foundation, KaviKrishna Foundation Fellowship award (to J. Talukdar and S. Bhuyan), KaviKrishna USA award (to R. Bhuyan and B. Pal), and Department of Biotechnology-India grant (to A. Sarma, D. Baishya, and A.C. Kataki).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.