Hypoxia-induced transcriptional repression of the melanoma-associated oncogene MITF

Abstract

Microphthalmia-associated transcription factor (MITF) regulates normal melanocyte development and is also a lineage-selective oncogene implicated in melanoma and clear-cell sarcoma (i.e., melanoma of soft parts). We have observed that MITF expression is potently reduced under hypoxic conditions in primary melanocytes and melanoma and clear cell sarcoma cells through hypoxia inducible factor 1 (HIF1)-mediated induction of the transcriptional repressor differentially expressed in chondrocytes protein 1 (DEC1) (BHLHE40), which subsequently binds and suppresses the promoter of M-MITF (melanocyte-restricted MITF isoform). Correspondingly, hypoxic conditions or HIF1α stabilization achieved by using small-molecule prolyl-hydroxylase inhibitors reduced M-MITF expression, leading to melanoma cell growth arrest that was rescued by ectopic expression of M-MITF in vitro. Prolyl hydroxylase inhibition also potently suppressed melanoma growth in a mouse xenograft model. These studies illuminate a physiologic hypoxia response in pigment cells leading to M-MITF suppression, one that suggests a potential survival advantage mechanism for MITF amplification in metastatic melanoma and offers a small-molecule strategy for suppression of the MITF oncogene in vivo.

Microphthalmia-associated transcription factor (MITF) is a basic helix–loop–helix (bHLH-Zip) transcription factor that regulates key processes in several cell lineages. Driven by alternative promoters, various MITF isoforms differing in their N-termini are regulated in a tissue-dependent manner (1). The melanocyte-specific isoform of MITF (designated M-MITF) is implicated in survival, proliferation, and differentiation of this neural crest-derived lineage (2). Additional MITF isoforms are important for development and function of hematopoietically derived osteoclasts and natural killer and mast cells, as well as for the retinal pigment epithelium. Mutations in MITF are the cause of Waardenburg syndrome type IIa in humans, characterized by pigmentation defects of the eye and hair forelock along with deafness caused by deficiencies of cochlear melanocytes (3–5). Similarly, mouse Mitf mutants exhibit small eyes, deafness, osteopetrosis, and reduced numbers of natural killer and mast cells, and additionally display pigmentation problems varying from complete loss of the melanocytic lineage to ventral spotting in partial loss-of-function settings (6, 7). MITF is also important for the maintenance of melanocyte stem cells in hair follicles, where the hypomorphic allele Mitf^vit is associated with premature hair graying (8).

Among the functional transcriptional targets of MITF are regulators of cell cycle (CDK2, TBX2, p21, p16/Ink4a), survival and metastasis (BCL2, c-Met), miRNA processing [Dicer (9)], cAMP levels [PDE4D3 (10)], and pigmentation (Tyrosinase, TYRP1/2, OA1), along with others including genes important to the biology of osteoclasts and mast cells (reviewed in ref. 11). Transcriptional activity of MITF requires its DNA binding to the E-box sequence CAC(G/A)TG as a homodimer or as a heterodimer with one of its related bHLH-Zip proteins TFEB, TFEC, and TFE3 (collectively referred to as the MiT family) (12). Importantly, MiT proteins have been implicated in multiple human tumors; TFE3 translocations have been identified in papillary renal-cell carcinomas and alveolar soft-part sarcomas (13–17), and TFEB was found to be translocated in papillary renal-cell carcinomas (18, 19). Integrated SNP array and expression profiling analyses enabled identification of MITF as an amplified oncogene in a subset of melanomas (20). MITF was also identified as a direct transcriptional target of the EWS-ATF1 fusion in human clear cell sarcoma, in which deregulated MITF behaves as an essential oncoprotein (21).

Conditions of low oxygen (i.e., hypoxia) arise physiologically during embryogenesis and in various niches of the adult, influencing proliferation, differentiation, and survival of cells. Hypoxic microenvironments can be found under pathologic conditions such as stroke and atherosclerosis and in solid tumors whose growth may be limited by oxygen supply (22). A master regulator of the cellular and physiological response to hypoxia is hypoxia-inducible factor (HIF) 1, a heterodimeric transcription factor composed of a stably expressed β-subunit (HIFβ/ARNT) and a hypoxia-inducible α-subunit (23, 24). Under normoxic conditions, hypoxia inducible factor 1 (HIF1) or HIF2 α-subunits are subjected to VHL-mediated polyubiquitination, which leads to their rapid degradation. This process is dependent upon the hydroxylation status of two conserved proline residues located in the oxygen-dependent degradation domain of HIF1α/2α, which are modified by the EGLN1-3 prolyl hydroxylases (25–28). EGLN1-3 inhibition, as in the absence of oxygen or cofactors, enables accumulation of HIFα subunits, heterodimerization, binding of hypoxia response elements, and transactivation of numerous target genes involved in energy metabolism, angiogenesis, cell growth, apoptosis, migration, and transcriptional regulation (29–31). Increased HIF1 stability/activity can be pharmacologically achieved by using iron chelators/antagonists [e.g., deferoxamine (DFO) and CoCl₂] or prolyl-hydroxylase inhibitors [e.g., dimethyloxalylglycine (DMOG)] or via stimulation of certain signaling pathways (32).

Clinical classification of melanoma frequently combines histologic analysis and immunohistochemical staining for melanocytic markers, such as MITF and/or its target genes MLANA/MART1 and HMB45 (33–35). Although MITF is a relatively specific marker for melanoma and clear-cell sarcoma, studies have indicated variability and heterogeneity in MITF staining intensity (36, 37) and RNA levels (38) within clinical melanoma specimens. The cause(s) for such variation is obscure, yet tumor microenvironment parameters might contribute to such expression effects. To address this possibility, we analyzed MITF expression in human melanoma specimens and cell lines.

Here, we identify hypoxia signaling as a negative regulator of M-MITF expression in vivo and in vitro. M-MITF down-regulation is mediated by HIF1, through recruitment of the HIF1-inducible transcriptional repressor differentially expressed in chondrocytes protein 1 (DEC1) (39–41) to the M-MITF promoter. M-MITF suppression, which can be mimicked by using small-molecule inhibitors of prolyl hydroxylases, has profound inhibitory effects on melanoma survival/proliferation in vitro and in vivo. This regulatory mechanism of MITF may thus provide a unique approach to MITF-targeted antineoplastic therapy.

Results

MITF Levels Decrease Under Hypoxic Conditions in Vitro and in Vivo.

We examined MITF expression in metastatic melanoma biospecimens and observed instances of MITF protein expression restricted to a rim of cells surrounding blood vessels (Fig. 1A). To check whether the loss of MITF signal in the peripheral tissue represented a physiological response to hypoxia, or perhaps was a nonspecific effect of dying cells, we stained melanoma specimens by immunofluorescence using HIF1α and MITF antibodies to compare the levels of both proteins in the same cell. Surprisingly, we found that the HIF1α and MITF signals were mostly inversely correlated and, in many zones, mutually exclusive (Fig. 1B). These observations prompted us to investigate whether MITF expression might be regulated by oxygen supply.

Fig. 1.

MITF levels are down-regulated by hypoxia. (A) H&E and MITF immunostaining of metastatic melanoma shows that MITF expression is limited to a vascularized zone (serial sections). (B) Immunofluorescent staining of human metastatic melanoma indicates reciprocal expression of MITF (green) and HIF1α (red). (C) Expression analysis of MITF and some of its target genes (CDK2, Silver, and Tyrosinase) in melanoma specimens and cell lines (asterisk) shows inverse Pearson correlation relative to hypoxia-induced genes. (D) MITF expression in three melanoma cell lines grown under normal oxygen tension (21%) or subjected to hypoxia (0.5% O₂) or hypoxia mimicry (with 200μM CoCl₂) for 24 h. HIF1α levels are presented as readout of cellular hypoxic response, and α-tubulin (αTub) levels serve as loading control. MALME-3M and UACC257 are lines with amplified MITF [n = 6–8 and n = 4–5 copies, respectively (20)]. (E) MITF levels in primary human melanocytes decrease following CoCl₂ treatment. M-MITF bands are marked with arrowheads.

To survey whether MITF is affected by hypoxic conditions in primary melanoma and melanoma cells lines, we compared the expression level of MITF and some of its target genes (CDK2, Silver, and Tyrosinase) to those of hypoxia-induced genes in a expression array data set that includes primary melanoma specimens and cell lines (38). Although they varied among samples, MITF levels correlated with those of its targets (as expected), yet inversely correlated with the expression level of the hypoxia markers HIF1α, VEGF-A, and IGFBP3 (Fig. 1C), strengthening the notion that hypoxic environment might lead to reduced MITF expression. To test that hypothesis in vitro, three different melanoma cell lines were subjected to 0.5% oxygen in a sealed chamber, grown in atmospheric oxygen, or alternatively treated with the hypoxia “mimetic” CoCl₂, which suppresses prolyl hydroxylase activity. Interestingly, upon HIF1α induction, a concomitant decrease in the level of MITF protein was observed, including in UACC257 and MALME-3M, melanoma lines in which MITF is genomically amplified (20) (Fig. 1D). Similarly, decreases in MITF levels also occurred in primary human melanocytes treated with CoCl₂ (Fig. 1E), indicating that this phenomenon is not a result of genomic aberrations or unique characteristics of malignant (i.e., melanoma) cells.

Furthermore, MITF levels decreased with increasing CoCl₂ concentrations and in a time-dependent manner, inversely correlating with HIF1α (Fig. S1 A and B ). Similar results were obtained by using DFO, an additional prolyl-hydroxylase inhibitor (Fig. S1C).

To examine whether the observed decreases in MITF are specific to the melanocyte lineage, we tested the MITF-expressing human mast cell line HMC1. Notably, because of activation of alternative promoters, mast cells do not express the M-MITF isoform (of 413/419 aa in size), but rather express splice variants of higher molecular weights (as high as 525 aa). As seen in human melanocytes, HMC1 cells showed hypoxia-induced MITF suppression following treatment with 0.5% oxygen or 200 μM CoCl₂ (Fig. S2A), suggesting that the mechanism for MITF down-regulation is not melanocyte-specific. Moreover, MITF expression was similarly suppressed in human clear-cell sarcoma (Fig. S2 B and C), a tumor in which MITF's dysregulated expression is oncogenic and required for viability (21). Overall, we found that MITF expression is down-regulated under hypoxic conditions in a time- and dose-dependent fashion in different cell lineages, including MITF-amplified melanomas and clear-cell sarcomas.

Down-Regulation of MITF During Hypoxia Is Mediated by HIF1.

As our experiments demonstrated an inverse correlation between HIF1α and MITF levels, we next asked whether the decrease in MITF levels is mediated by HIF1. To this end, primary melanocytes and UACC62 melanoma cells were infected with adenovirus expressing HIF1α-GFP (42) or control (GFP only). MITF levels decreased in a HIF1α titer-dependent fashion under normoxic conditions in primary melanocytes (Fig. 2A) and UACC62 melanoma cells (Fig. 2B), demonstrating that HIF1α accumulation by overexpression or following prolyl hydroxylase inhibition (Figs. S1 and S2) is sufficient to induce MITF suppression. To investigate whether HIF1α is necessary for hypoxia-induced suppression of MITF, UACC62 cells were transfected with siRNAs against HIF1α (or scrambled control) and then subjected to 24 h treatment with CoCl₂. Knockdown of HIF1α levels significantly rescued MITF expression, compared with cells transfected with control siRNA (Fig. 2C). Taken together, these experiments demonstrate that HIF1α induction is sufficient for MITF down-regulation, and suggest that hypoxic suppression of MITF is significantly regulated by HIF1.

Fig. 2.

Hypoxia-induced down-regulation of MITF is mediated by HIF1. Western blot analysis of MITF and HIF1α levels in (A) human primary melanocytes or (B) UACC62 cells 48 h after infection with GFP-control or HIF1α-GFP adenoviruses at increasing MOIs as indicated. (C) UACC62 cells were transfected with control or anti-HIF1α siRNA and then subjected to 200 μM CoCl₂ treatment for 24 h. M-MITF bands are marked with arrowheads.

Down-Regulation of MITF Under Hypoxia Is Caused by Reduced Transcription of MITF Gene.

To examine the mechanism through which MITF protein levels decrease following hypoxia, MITF RNA levels were quantified by quantitative RT-PCR (qRT-PCR). The abundance of MITF transcripts (including all isoforms) in UACC62 was found to diminish by four to five fold within 16 h (Fig. 3A). Similarly, MITF RNA levels decreased in hypoxia-treated and adenoviral HIF1α-infected cells compared with controls (Fig. 3B), suggesting that the down-regulation of MITF occurs at the RNA level. We used several strategies to examine whether the observed decrease in MITF RNA is a result of reduced transcription rate vs. RNA instability (or both). First, we compared the relative abundance of newly transcribed versus processed (i.e., spliced) MITF RNA, before and after CoCl₂ treatment, by qPCR reactions by measuring nascent unspliced M-MITF mRNA (amplicon located in intron 1), or spliced mRNA levels (primers flanking intron 1). VEGF RNA levels served as positive control for HIF1α function. CoCl₂ treatment resulted in the same degree of down-regulation for both unspliced and spliced MITF RNA, under conditions that strongly induced VEGF expression (Fig. 3C), pointing to reduced transcription rate rather than altered RNA instability. As a separate approach to this question, we measured the relative association of RNA polymerase II (RNAPII) with the MITF locus in untreated vs. CoCl₂-treated cells by means of RNAPII ChIP followed by qPCR analysis. To assess transcription elongation rather than initiation, we used primers located several kb downstream of the M-MITF promoter (in introns 2 and 6). As a positive control, we amplified a portion of VEGF exon1, which is known to be transcriptionally induced under hypoxia. Our results revealed potent reduction in the association of the RNAPII with the MITF locus following CoCl₂ treatment, whereas RNAPII recruitment to the VEGF locus was augmented under the same conditions (Fig. 3D). These results suggest that the rate of MITF gene transcription is reduced in hypoxic cells.

Fig. 3.

MITF RNA levels are decreased as a result of reduced transcription rate. (A) MITF RNA levels decrease in a time-dependent manner in UACC62 cells treated with CoCl₂. (B) Decrease in MITF RNA levels following incubation in low oxygen conditions or adeno-HIF1α infection (time 24 h). (C) Nascent and spliced MITF RNA species exhibit the same magnitude of down-regulation under CoCl₂ treatment. VEGF levels are shown as positive control. (D) ChIP showing diminished recruitment of the RNAPII machinery to the MITF locus following CoCl₂ treatment of UACC62 cells. (E) Transfection of pcDNA3-full length human M-MITF mRNA to mouse B16F0 melanoma cells reveals that promoter swap can rescue MITF RNA down-regulation. (F) UACC62 polyclonal stable line expressing HA-MITF off a heterologous promoter and UACC62–empty vector control line were subjected to CoCl₂ treatment. MITF protein level was evaluated by Western blot by using MITF and HA antibodies. M-MITF bands are marked with arrowheads. Note that HA-MITF protein levels are unaffected following CoCl₂ treatment under conditions that reduced the endogenous protein. In all CoCl₂ treatments, a dose of 200 μM was applied. Results are triplicates from one experiment (mean ± SD of fold relative to untreated cells) and representative of three independent experiments. P values were calculated relative to control by using a Student t test.

If MITF suppression by hypoxia occurs at the level of transcription, MITF mRNA expressed off a heterologous promoter should be resistant to hypoxic down-regulation. To test this hypothesis, we cloned the full-length cDNA of human M-MITF (including both UTRs) downstream of the CMV promoter and transiently expressed it via transfection in B16F0 murine melanoma cells, which were subsequently treated (or not) with CoCl₂. We then quantified the RNA levels of the ectopic human MITF and compared it with the endogenous mouse Mitf transcript using species-specific primers. Whereas the endogenous mouse Mitf RNA levels were decreased by four to five fold following CoCl₂ treatment, expression of the CMV-driven human M-MITF was unaffected (Fig. 3E). Moreover, we generated stable derivative polyclonal lines of UACC62 introduced with pWZL-Blast empty vector (designated UACC62-Vec) or pWZL-Blast harboring HA-tagged human-M-MITF (UACC62-MITF), which ectopically expresses MITF driven by the Moloney murine leukemia virus LTR promoter. When these two lines were treated with CoCl₂, endogenous MITF protein levels decreased as expected; nevertheless, HA-MITF expression was resistant to CoCl₂-induced down-regulation (Fig. 3F). Collectively, these data indicate that the decrease in MITF expression under hypoxia results from reduced transcription rate and is not caused by shorter RNA or protein half-lives.

Transcriptional Repressor DEC1 Negatively Regulates M-MITF Promoter.

As HIF1 is mostly described as an inducer of transcription rather than direct repressor, and because HIF1α protein stabilization occurs within minutes whereas potent decreases in MITF RNA levels required several hours, we hypothesized that a HIF1 target gene might serve as an intermediate for MITF repression. The DEC1 (also known as BHLHE40) and DEC2 (BHLHE41) genes are known hypoxia-induced transcriptional repressors found to be up-regulated in cancer (43–45), and whose expression can be detected in melanoma lines (refs. 46–48 and Novartis NCI60 microarray data, biogps.org). We indeed detected rapid and significant induction of DEC1 RNA in CoCl₂-treated UACC62 cells as early as 1 h, which preceded M-MITF suppression seen after 3 h of treatment. Expression levels of melanoma-nonspecific MITF isoforms were unaffected during this time. DEC2 expression levels did not increase by 3 h of CoCl₂ treatment, and only a small, insignificant increase was observed at 6 h (Fig. 4A). Interestingly, when overexpressed, both DEC1 and DEC2 could repress M-MITF expression under normoxic conditions (Fig. 4B). Although a DEC2 effect on the M-MITF promoter is conceivable, the lack (or delay) of DEC2 induction argues against its involvement in early steps of M-MITF suppression. Moreover, DEC1 knockdown was sufficient to rescue M-MITF mRNA from down-regulation by CoCl₂ (Fig. 4C), and to restore M-MITF promoter activity in cells overexpressing HIF1α (Fig. 4D), strengthening the role of DEC1 as the suppressor of MITF expression in UACC62 cells.

Fig. 4.

DEC1 suppresses M-MITF transcription. (A) qPCR analysis indicates that DEC1 expression is rapidly induced in UACC62 cells treated with 200 μM CoCl₂, preceding M-MITF down-regulation. (B) Transient overexpression of DEC1 or DEC2 in UACC62 cells is sufficient for M-MITF protein down-regulation. (C) DEC1 silencing using two different DEC1 shRNA (sequence 1 in gray bars; sequence 2 in dotted bars) rescues M-MITF mRNA from down-regulation by 200 μM CoCl₂ compared with cells introduced with control shRNA (black bars). (D) UACC62 cells introduced with shRNA against DEC1 (sequence 1, gray bars) or control shRNA (black bars) were assayed for M-MITF promoter activity using the pGL3-hMIP-2.3K luciferase reporter plasmid. HIF1α (P420A) overexpression resulted in dose-dependent inhibition of M-MITF promoter activity, which was rescued by DEC1 silencing. (E) Reporter assays demonstrate that the HIF1α-induced suppression of the M-MITF promoter activity is dependent on an upstream DEC1 binding site. (F) DEC1 is recruited to the M-MITF promoter, but not to the β-actin promoter, upon CoCl₂ treatment of MALME-3M cells. Immunoprecipitation with a control IgG is shown as control for precipitation specificity. (G) Knockdown of DEC1 fully rescues MITF protein levels in CoCl₂-treated UACC62 cells. M-MITF bands are marked with arrowheads.

By using reporter assays, we observed that the HIF1-mediated suppression of the M-MITF promoter is dependent on a DEC1/2 consensus binding site located approximately 1.8 kb upstream of M-MITF transcription start site, as site-directed mutagenesis of this site prevented the effect of HIF1α on the promoter activity (Fig. 4E). Consistent with these findings, we observed hypoxia-dependent recruitment of DEC1 to the M-MITF promoter (but not to the β-actin promoter, a negative control) by quantitative ChIP (Fig. 4F). In addition, we found that knockdown of DEC1 in UACC62 cells by using shRNA fully rescued MITF protein from CoCl₂-induced down-regulation, whereas a scrambled control shRNA had no effect (Fig. 4G). Collectively, these data point to DEC1 as the factor responsible for hypoxia-mediated suppression of the MITF gene.

HIF1-Mediated Suppression of Melanoma Growth in Vitro: Rescue by Ectopic MITF.

To study the functional relevance of MITF down-regulation by prolyl-hydroxylase inhibition in melanoma cells in vitro and in vivo, we took advantage of our stable cell lines UACC62-Vec and UACC62-MITF (whose ectopic promoter lacks the HIF1/DEC1 repressible element from the endogenous MITF promoter). Total MITF expression in the UACC62-MITF cells was higher than in UACC62-Vec cells (Fig. 3F), yet this level was still significantly lower than in MALME-3M or 501Mel (Fig. 5A), placing it within a typical range for human melanoma lines.

Fig. 5.

MITF rescues UACC62 melanoma cells from DMOG-induced loss of viability. (A) Stable UACC62 lines carrying HA-MITF or empty vector express moderate MITF levels compared with other human melanoma lines. (B) UACC62-Vec cells show dose-dependent reduction in MITF levels following DMOG treatment, whereas HA-MITF levels are unaffected. M-MITF bands are marked with arrowheads. (C–E) Ectopic MITF expression rescues UACC62 cells from DMOG-induced loss of viability. Images in C represent random fields and corresponding crystal violet-stained wells after 6 d of DMOG treatment. Growth/survival curve of UACC62-Vec (D) or UACC62-MITF (E) cells treated with 0 to 1 mM DMOG. Results are triplicates from one experiment (mean ± SD) and representative of three independent experiments.

Preparing for in vivo experiments, we tested the effect of DMOG, a prolyl-hydroxylase inhibitor that has been successfully used in vivo. Treatment of UACC62 cells with 1 mM DMOG reduced M-MITF RNA levels (Fig. S3) with the same magnitude and kinetics of CoCl₂ (Fig. 3A). Similarly, when UACC62-Vec and UACC62-HA-MITF cells were treated with DMOG, dose-dependent down-regulation of endogenous MITF protein was observed as expected. Conversely, exogenous HA-MITF expression in the UACC62-MITF cells was resistant to DMOG (Fig. 5B), as it was resistant to CoCl₂ treatment (Fig. 3F).

We next assayed growth capacity of these cells upon treatment over a dose–response titration of DMOG. The control UACC62-Vec cells showed dose-dependent loss of viability starting at 2 to 3 d after DMOG addition, reaching almost complete loss after 4 to 6 d in the presence of 0.75 to 1 mM DMOG (Fig. 5 C and D). However, UACC62-MITF cells were notably protected under all DMOG doses tested, and their proliferation was uninterrupted during the assay (Fig. 5 C and E). Importantly, similar results were obtained when UACC62-Vec or UACC62-MITF were grown under hypoxic conditions of 0.5% O₂ (Fig. S4A). UACC62-Vec cells were able to grow for 3 to 4 d under hypoxia, but the lack of MITF expression in these setting (Fig. S4B) resulted in a growth crisis and a significant decrease in cell number relative to hypoxic UACC62-MITF cells or cells grown in normoxia.

In addition, UACC62-Vec cells transduced with HIF1α exhibited significantly reduced growth relative to HIF1α-infected UACC62-MITF cells, or cells infected with a control (i.e., GFP) virus (Fig. S4C). This inhibition of cell growth occurred concomitantly with decreased MITF expression in HIF1α-infected UACC62-Vec but not UACC62-MITF infected cells (Fig. S4D). As expected, DEC1 expression was induced by HIF1 at day 2 of infection and anticorrelated with MITF in UACC62-Vec but not UACC62-MITF cells. Unexpectedly, at longer time points of culture growth (days 4 and 6), DEC1 expression was altered relative to day 2, possibly because of parameters of cell confluence/proliferation that also regulate DEC1 expression levels (49).

Taken together, these significant differences in cell growth and viability among genetically matched melanoma lines, differing in the presence of hypoxia-resistant MITF, point to the critical dependency of melanoma cells on MITF expression.

DMOG Treatment Suppresses Melanoma Growth in Vivo.

To preclinically test the strategy of MITF suppression as a therapeutic approach to melanoma in vivo, nude mice were injected s.c. with UACC62 cell xenografts, and upon detection of measurable tumors, treated with DMOG or vehicle (i.e., PBS solution) control. Whereas vehicle-treated mice developed large tumors and had to be killed by day 21 after injection, DMOG-treated mice exhibited profoundly diminished tumor progression and consequently high survival during this time interval (Fig. 6 A and B), without changes in body weight (Fig. 6C). Furthermore, when UACC62-Vec and UACC62-MITF xenografts were compared in mice treated with DMOG, UACC62-MITF melanoma cells abrogated the growth-suppressive effect of DMOG treatment (Fig. 6D), consistent with the model that growth suppression was a consequence of MITF inhibition via the DMOG–HIF1α–DEC1 axis.

Fig. 6.

DMOG suppresses melanoma xenograft tumor growth. (A–C) Nude mice were implanted with 10⁶ UACC62 melanoma cells per site and, after establishment of tumors, treated with DMOG (20 mg per injection, twice per day; n = 10) or vehicle (PBS solution; n = 8). Measured tumor volume (A) and representative images (B) demonstrate DMOG efficacy without affecting mouse weight (C). Data are represented as mean ± SEM. (D) UACC62-Vec or UACC62-MITF polyclonal melanoma lines were injected in nude mice. After establishment of tumors, each group was treated with DMOG (10 mg/d) or PBS solution. Note that DMOG reduced tumor volume in UACC62-Vec–injected mice but not in mice injected with UACC62-MITF harboring hypoxia-insensitive MITF. Data were combined from two independent experiments on day 11 of treatment, and normalized to the vehicle control for each group. Group size was 11 to 17 as indicated. Statistical significance was calculated by Student t test.

Discussion

This study describes a regulatory mechanism for MITF expression, which can be mimicked by using small-molecule inhibitors and may provide a unique therapeutic strategy. Exploring in vivo expression patterns of MITF in metastatic melanomas, we noticed that hypoxic conditions profoundly suppressed MITF levels. In in vitro models, the mechanism appeared to operate through transcriptional suppression of the MITF gene by DEC1, a HIF1 target that is recruited to the MITF locus under hypoxia.

Interestingly, HIF1α was reported to be an MITF target gene, induced following activation of the cAMP pathway in mouse B16/F10 melanoma cells or adenoviral MITF overexpression in MeWo human melanoma cells (50). Our data, gathered from multiple human cell lines/lineages including primary melanocytes and in vivo immunostaining of human specimens, demonstrated a fairly robust inverse correlation between HIF1α and MITF levels, and physiological MITF overexpression did not appear to alter detectable levels of HIF1α protein (Figs. 3F and 5B). Our findings are also in line with immunohistochemical analysis of primary and metastatic human tumors, which demonstrated that MITF and its target gene Melan-A (34) consistently anticorrelate with the hypoxia marker Glut-1 (51). Nonetheless, it remains possible that this might represent a homeostatic loop in which MITF activation/overexpression, which presumably induces HIF1α expression, triggers HIF1-mediated repression of MITF and restores MITF basal levels. In this regard, it is interesting to speculate that HIF1α might regulate MITF levels under physiological conditions. The skin is likely to represent such a (relatively) hypoxic microenvironment (52).

The MITF locus in human and mouse spans a region of approximately 200 kb and comprises at least nine promoters driving expression of alternative first exons (reviewed in refs. 1, 53). Interestingly, although, in melanocytes/melanoma cells, hypoxia causes selective down-regulation of M-MITF (the most abundant isoform in these cells), it can also affect other MITF isoforms expressed in nonmelanocytic cells (Fig. S2). By using ChIP and reporter assays, we mapped the HIF1–DEC1 effect to an E-box within 2 kb of the M-MITF promoter, in a region which is juxtaposed to exons 2 to 9 shared by all MITF isoforms. Although DEC1 can presumably occupy numerous sites within the approximately 200-kb MITF locus, an effect on the M-MITF promoter might hamper accessibility to the common exons 2 to 9, thereby affecting expression of various MITF isoforms, although the precise impact on nonmelanocytic isoforms expression in different cell types remains to be elucidated.

The discovery of MITF amplifications in melanoma is notable for its increased incidence in metastatic, compared with cutaneous, lesions (20). In addition, no MITF amplification has been detected in benign nevi. The relatively common occurrence of melanomas with low MITF expression may arise from selection against the differentiating activity of MITF (e.g., activation of pigment enzyme machinery). Nonetheless, many melanomas typically remain dependent on MITF for viability (54), mimicking dependency of the melanocyte lineage on this factor. Noteworthy, as a late genomic event during melanoma progression, relative to BRAF(V600E) mutation, for example, it has been uncertain what advantage(s) MITF amplification may provide to the tumor. One possibility is raised by the present findings. MITF is known to be a lineage survival factor for melanocytes, based on the phenotypes of numerous MITF mutations in multiple species including human (55). In metastatic melanomas, MITF amplification may sustain tumor viability under conditions of compromised vascular supply caused by higher basal MITF levels and proportionally more residual MITF. It is thus plausible that elevated MITF expression (via amplification) provides a buffer against diminished expression that may result from HIF1α surges within hypoxic microenvironments. One such MITF-amplified melanoma was examined here (MALME-3M; Fig. 1D). In this line, the basal MITF expression is high and remains high even under in vitro conditions of 0.5% O₂, and may thus confer relative tolerance to HIF1-mediated MITF fluctuations. Nevertheless, upon stronger HIF1α induction in response to prolyl hydroxylase inhibition, MITF level still diminishes in MALME-3M, suggesting that this relative tolerance may be pharmacologically targeted.

The use of prolyl-hydroxylase inhibitors, which stabilize HIF1α, seems counterintuitive in the case of cancer in view of the proangiogenic role of HIF1α. However, our findings unexpectedly imply that, in the particular case of melanoma (or a subset thereof), the dependency on MITF as a critical factor for this lineage might be critical, in a way that its down-regulation by HIF1–DEC1 may offer therapeutic advantage despite other HIF1 actions. The prospect of combining antiangiogenic therapies with prolyl-hydroxylase inhibition is thus of particular interest.

In summary, our data demonstrate that the MITF promoter is repressed by the HIF1–DEC1 axis, which could potentially serve as a therapeutic strategy for MITF-targeted therapy of melanoma, clear-cell sarcoma, osteoporosis (targeting osteoclast MITF), and allergy (targeting mast cell MITF).

Materials and Methods

Cell Culture.

Human melanoma lines UACC62, UACC257, and MALME-3M were obtained from the National Cancer Institute. B16F0 and HEK293 were purchased from American Type Culture Collection. 501mel was a gift from Ruth Halaban (Yale University Medical School, New Haven, CT). HMC1 was a gift from Cliff Takemoto (Johns Hopkins University, Baltimore, MD). Primary human melanocytes were isolated from neonate foreskins as previously described (56, 57). Cells lines were maintained in media (Mediatech) supplemented with 10% FBS (Sigma) and 1% penicillin–streptomycin–glutamine (Invitrogen) as follows: UACC62, UACC257 and HMC1 in RPMI-1640; MALME-3M, B16F0, and HEK293 in DMEM; and 501mel in Ham F-10. Primary melanocytes were maintained in TIVA medium (F-10 with 7.5% FBS, 50 ng/mL TPA, 225 μM IBMX, 1 μM Na₃VO₄, and 1 mM dbcAMP; all from Sigma). Two weeks before hypoxia experiments, primary melanocytes were switched to “low TIVA” medium, lacking IBMX and containing only 10 ng/mL TPA.

Hypoxia and Hypoxia-Mimicking Drugs.

Cells were maintained at 37 °C under normoxic conditions (21% O₂, 5% CO₂) or exposed to hypoxia (0.5% O₂, 5% CO₂) in a gas-sealed workstation (In vivo2; Biotrace) when indicated. Hypoxia mimicry was achieved by using CoCl₂ (Sigma), DFO mesylate (Sigma), or DMOG (Frontier Scientific) as indicated.

Plasmids.

Full length human M-MITF was amplified in one step from 501MEL cDNA by using primers 5′-CGGGATACCTTGTTTATAGTACCTTC-3′ and 5′- GTGAAAAACCAAATGCTTTAATGAGGCTATC-3′. The PCR product was cleaned on a minicolumn (Qiagen) and ligated into pCR-Blunt II–TOPO (Invitrogen). Subsequently, the insert was cut out by using KpnI-NotI digestion, and was ligated to pcDNA3.1 (Invitrogen) using the same restriction enzymes. The pcDNA3/HA-M-MITF and pWZL-Blast-MITF constructs were previously described (20, 58). DEC1 and DEC2 cDNAs were amplified by PCR and cloned into pcDNA3-HAHA using EcoRI/NotI and BamHI/XhoI sites, respectively. pGL3-hMIP-2K (harboring nucleotides −2018 to +121 relative to M-MITF transcription start site) and pGL3-hMIP-1.5K (harboring nucleotides −1518 to +121) were generated by amplification of human M-MITF promoter region using forward primers 5′-CGGTACCGGGCCTGGAGAAGTTAATGAATTGC-3′ (for pGL3-hMIP-2K) or CGGTACCCTTATTACCTTCCTCGTAGCTCATG (for pGL3-hMIP-1.5K; with an added 5′ KpnI site) and 5′-TTCCATGGCCAGCATAACAATGTTTTAGG-3′ (with NcoI site) as reverse primer. PCR products were digested with KpnI-NcoI and cloned into the corresponding sites of pGL3-Basic (Promega). pGL3-hMIP-2.3K (harboring nucleotides −2274 to +121) was generated by using the same cloning procedure. In all constructs, the first ATG of luciferase was positioned exactly in place of the M-MITF start codon. Potential DEC1 binding sites were mutated by using a QuikChange II site-directed mutagenesis kit (Stratagene) from CAGTTG CACATG, or CAGCTG to GAGTTG, GAGATG, and GAGCTG, respectively.

Immunohistochemistry and Immunofluorescence.

Immunostaining studies were performed by using formalin-fixed, paraffin-embedded tissue. Sections were cut at 4 μm, dried at 37 °C, deparaffinized in xylene, hydrated in a graded series of alcohol, and subjected to microwave antigen retrieval. Immunohistochemical staining was done with undiluted mouse monoclonal antibody D5 (tissue culture supernatant). Staining was performed on serial slides by using a DAKO LSAB+ alkaline phosphatase detection system and permanent red as the chromogen, with Mayer hematoxylin used for counterstaining. Images were obtained with a lens with a magnification of 20×. For immunofluorescence, samples were blocked with 5% BSA/0.5% Tween-20/PBS solution, and detection of MITF and HIF1α protein levels was achieved by using undiluted mouse anti-MITF (C5) supernatant and rabbit anti-HIF1α (ab19382; Abcam), followed by incubation with Alexa 488- and Alexa 594-conjugated secondary antibodies (Invitrogen), respectively. Images were obtained with a lens with a magnification of 40×.

Transfections.

Cells were plated 24 h before transfection, and were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.

ChIP.

RNA Pol-II ChIP assay was done as previously described (54) using pol-II antibody CTD4H8 (Covance) or isotype matched control IgG in UACC62 cells. DEC1 antibody (Bethyl Laboratories) or rabbit IgG as control were used for DEC1 ChIP in MALME-3M cells.

Adenovirus Production and Infection.

Adenoviruses encoding GFP-HIF1α fusion protein (42) or vector control encoding a short inert polypeptide were purified as previously described (59). Briefly, viruses were propagated in HEK293 cells, which were collected upon detection of viral cytopathic effect. Cell pellets were resuspended in PBS solution and lysed by freeze/thaw cycles. Concentration and purification of virus was performed by using a cesium chloride gradient, followed by intensive dialysis. Multiplicity of infection (MOI) was estimated by OD_260/280, using a multiplier of 2 × 10⁶ cfu/μL per OD unit. For infection, 5 × 10⁵ cells were plated per well in six-well dish and, on the next day, overlaid with 0.5 mL serum-free media containing 10 mM MgCl₂ and concentrated adenovirus at the indicated titer. Following incubation at 37 °C for 30 min, 2 mL of fresh full media was added and the cells were grown until harvest.

RNA Interference.

HIF1α silencing experiments were carried out using transiently delivered RNA duplexes with a lipid nanoparticle. UACC62 cells were plated per well in six-well plates and, 16 to 24 h later, transfected with control (i.e., scrambled) siRNA (Dharmacon) or HIF1α siRNA (M-004018-05; Dharmacon) using 25 nM of siRNA. Twenty-four hours after transfection, cells were treated (or not) with 200 μM CoCl₂ for an additional 24 h. For DEC1 silencing, pLKO plasmids for shLUC (against luciferase), shDEC1, or scrambled sequence (all from RNAi Consortium shRNA Library, Broad Institute, Massachusetts Institute of Technology/Harvard University) were used. The DEC1 targeting sequences are no. 2 (GCACTAACAAACCTAATTGAT) and no. 5 (CCCTTTAAACTTAGAAACCAA). UACC62 cells were infected and selected with puromycin (2 μg/mL) for 3 d. After selection, cells were treated with 200 μM CoCl₂ for the indicated times.

Real-Time PCR/qPCR.

Purification of RNA was performed by using TRIzol reagent (Invitrogen) or RNeasy kit (Qiagen) from triplicate samples in multiple experiments, followed by treatment with RNase-free DNase (Qiagen). RNA (100 ng) was taken for real-time qPCR analysis, which was performed as previously described (54). Expression levels were normalized to 18s rRNA or β-actin as specified. ChIP precipitated DNA was amplified by using 2× SYBR Green mix (Applied Biosystems). Sequences of primers and probes used in the manuscript are specified in Table S1.

Reporter Assays.

Transcriptional activity of the human MITF-M promoter was assessed in UACC62 cells infected with shSCR or shDEC1 by transient transfection of with pGL3-hMIP-2.3K reporter plasmid driving firefly luciferase (50 ng), along with pRL-CMV (Promega) harboring Renilla luciferase (1 ng), and increasing doses of pcDNA3-HIF1α (P420A; 150–750 ng; gift from Bill Kaelin, Dana-Farber Cancer Institute). To maintain constant DNA quantities, pcDNA3-HIF1α doses were complemented with pcDNA3 vector carrying a short nonfunctional peptide (control, 150–750 ng). For mapping of the HIF1-affected element promoter within M-MITF promoter, UACC62 cells were transfected with pGL3-hMIP-2K or -1.5K plasmids, or variant thereof in which potential DEC1 binding sites (E-boxes) were mutated. Transfections were performed with the previously described plasmid amounts using Lipofectamine 2000 (Invitrogen) in 24-well plates according to the manufacturer's protocol. After 48 h of transfection, cells were harvested in 100 μL of lysis buffer and the lysates were assayed by Dual-Luciferase Reporter Assay System (Promega). Firefly luciferase activity was normalized to Renilla luciferase activity.

Western Analysis.

For total protein extracts, cells were washed twice with PBS solution and scraped in 2× lysis/loading buffer [125 mM Tris (pH 6.8), 4.6% SDS, 20% glycerol, 5% β-mercaptoethanol, 0.04% pyronin Y, and complete mini protease inhibitors (Roche)]. Extracts were resolved on SDS-polyacrylamide gels followed by transfer to nitrocellulose membranes (BioRad). Proteins were detected by using antibodies to HIF1α [Novus Biologicals (Fig. 6) or Santa Cruz], MITF (C5 hybridoma supernatant), p42/44 MAPK (Cell Signaling), α-tubulin (Sigma), or β-actin (Santa Cruz). HRP-conjugated secondary antibodies (Amersham) and renaissance chemiluminescence reagents (Perkin-Elmer) were used for signal detection.

Cell Growth Assay.

For analysis of cell growth/number in presence of DMOG, 1 × 10⁵ UACC62-Vec or UACC62-MITF cells were plated in multiple 12-well dishes on day −1. The following day (day 0), DMOG was added at the indicated concentrations to triplicate wells. A representative plate was fixed each day by using 10% ethanol/10% acetic acid, and stained with 0.2% crystal violet fixative. Following washes with water, the plates were dried and images were taken. For quantification of the crystal violet amounts, linearly representing cell number, the dye was redissolved in fixation solution. Color intensity was measured at OD595 in a 96-well plate reader, and normalized to day 0 as 100%. For cell growth assays under hypoxia/normoxia, 1 × 10³ cells were plated per well in 24-well dishes on day −1 and grown in 0.5% O₂ or normoxia from day 0 to day 6. Cells were counted every day by hemocytometer. For analysis of cell growth following HIF1α overexpression, 1 × 10⁴ UACC62-Vec or UACC62-MITF cells were plated in multiple 12-well dishes on day −1. The following day, cells were infected in triplicate wells at an MOI of 100 with an adenovirus coding for HIF1α or GFP. A representative plate was fixed each day and analyzed with crystal violet staining as described earlier.

Statistical Analysis.

Data are expressed as mean ± SD or mean ± SEM as indicated. Categorical variables (e.g., qPCR data) and continuous variables (e.g., comparison of tumors, weights) were analyzed by using a two-tailed Student t test. One-way ANOVA was used to examine differences in response to between groups in DMOG cell number assay. Hypoxia growth curves were compared using paired Student t test. P values of less than 0.05 were considered significant. GenePattern was used for the clustering analysis of MITF expression, and Pearson correlation coefficient values were calculated by Excel.

In Vivo Experiments.

Nude mice (7–8 wk old) were implanted s.c. with UACC62 melanoma cells or with the polyclonal derivative lines UACC62-Vec or UACC62-MITF. Cells (10⁶) were injected at each site in both flanks. Tumors were allowed to reach a volume of 50 to 100 mm³, a point designated as day 0 for each mouse, before treatment with i.p. injections of DMOG [20 mg in 0.2 mL twice daily (Fig. 6A) or 10 mg/d (Fig. 6D)] or PBS solution as control. Measurements were taken every 2 to 4 d with a caliper. The animal use protocol was approved by the institutional animal care and use committee of Dana-Farber Cancer Institute.