ChIP identifies potential POU3F2 target genes in melanoma and neural cells
In this study, we have used ChIP-chip to identify POU3F2 occupied promoters and target genes in 501Mel cells. This approach has identified a large set of natural POU3F2 binding sites. Bioinformatic analysis allowed us to identify numerous OCT, PORE, MORE and NORE-like motifs at the occupied loci. However, only few consensus PORE, MORE and NORE motifs were identified. The vast majority could only be identfied by allowing multiple mismatches. A more detailed analysis by DNaseI footprinting and EMSA revealed the existence of many sites that apart from being AT-rich, showed little similarity to the known binding motifs. POU3F2 can therefore bind a more diverse set of sites than those previously described.
At most of the promoters we tested, several closely spaced sites were observed. This indicates that although we identified 2108 distinct occupied loci, each locus may comprise more than one binding site. At some sites, we observed efficient binding in DNaseI footprinting experiments, but the same sites appeared to be poor competitors in EMSA analysis. It has previously been shown that POU3F2 can homodimerise (Smit et al., 2000). One possible explanation for the discrepancy between the footprinting and EMSA results is therefore that binding of POU3F2 to the high affinity sites facilitates or stabilises binding to adjacent lower affinity sites. Such homodimeric interactions could take place on the intact DNA fragment, but not when individual sites are used in EMSA.
Our results also demonstrate that POU3F2 can activate transcription via the POU3F2 binding sites of the Wnt16, Kitl, and Cd36 promoters in cell-specific manner. The combination of these different approaches has therefore allowed us to identify and characterise functional POU3F2-occupied promoter elements. Nevertheless, although the POU3F2 binding sites of the Wnt16 promoter can mediate activation in the context of a reporter plasmid, unlike Kitl, the endogenous Wnt16 gene is not activated by POU3F2-overexpression. Other elements of this promoter must therefore modulate the activity of POU3F2. The mechanisms underlying the promoter selective effects of POU3F2 remain to be determined.
We identify 1700 potential target genes with one or more POU3F2 binding sites in their proximal regulatory regions. Ontology analysis of these genes revealed that many are involved in neuron function. For example, typical neural genes like Gabrb1 (gamma-aminobutyric acid, GABA, A receptor, beta 1) Grin3a, (glutamate receptor, ionotropic, N-methyl-D-aspartate 3A), or Glra3 (glycine receptor, alpha 3) are all occupied by POU3F2. The occupancy of numerous neural genes by POU3F2 indicates that, although they are either not or weakly expressed in 501Mel melanoma cells, their regulatory regions remain accessible to transcription factor binding. Melanocytes are derived from the neural crest that also gives rise to several structures including neurons of the peripheral nervous system (Le Douarin et al., 2008). The accessibility of neural promoters to POU3F2 occupancy highlights the fact that melanoma cells retain a memory of their embryonic origins and can be considered as ‘neural’ derivatives of the neural crest.
Previous studies have shown that POU3F2 and MASH1 synergise to regulate proliferation, migration and differentiation of neural progenitors (Castro et al., 2006). Genetic studies and bioinformatic analysis identified target genes that contained a composite MASH1/POU3F2 binding motif and were occupied by both proteins in ChIP assays from embryonic telencephalon. The composite regulatory elements for a number of these genes are present on the arrays used here, but are not occupied in 501Mel cells that do not express MASH1 or MASH2. Thus although many neural genes are occupied by POU3F2 in 501Mel cells, this particular subset that requires cooperative binding with MASH1 are not. In this respect it is also worth noting that we do not observe POU3F2 occupancy or regulation of the Pax3 promoter in 501Mel cells. This promoter has been shown to comprise a POU3F2 binding site and to be regulated by POU3F2 and HOXA1 in neuronal EH3 cells (Pruitt et al., 2004).
POU3F2 occupies several genes of the Wnt/β-catenin signalling pathway such as β-catenin (Ctnnb) itself, the Wnt16 and Wnt8b ligands, the Dkk1 antagonist and the key Apc and Gsk3b components of the intracellular signalling cascade. Wnt signalling plays a critical role in melanocyte development and melanoma (Delmas et al., 2007; Larue and Delmas, 2009) (Larue and Delmas, 2009), however it is also a key player in neural development (Salinas and Zou, 2008). The role of a potential POU3F2 regulation of Wnt signalling in either or both of these processes remains to be determined. In addition, as POU3F2 expression is controlled by Wnt signalling (Goodall et al., 2004a), its occupancy of the above promoters may reflect an autoregulatory loop. Similarly, POU3F2 occupies the Rarb promoter suggesting a cross-talk between POU3F2 and retinoid signalling. Again this may be related to neural functions of POU3F2 where for example, it is required for retinoic acid induced neuronal differentiation of P19 cells (Fujii and Hamada, 1993), rather than its function in melanoma.
POU3F2 occupies the promoters of several genes that are relevant for melanocytes and melanoma. In addition to MITF, genes involved in pigmentation such as Tyr and Tyrp1, the cell cycle regulators Rb1 and Cdkn2a and the Met oncogene that is regulated by MITF to control migration and metastasis (Mcgill et al., 2006) are all occupied by POU3F2. It is important however to stress that we have so far no evidence that POU3F2 regulates any of these genes at least in the context of 501Mel cells in vitro. Nevertheless, our results suggest that, in the appropriate cell context and/or growth conditions, POU3F2 has the potential to control many cellular processes and signalling pathways relevant for melanoma.
POU3F2-regulated genes in melanoma
ChIP-chip and DNaseI footprinting show occupancy of multiple sites on the MITF promoter. In addition to the site previously identified upstream of the TATA element (Goodall et al., 2008), we identified a second site further upstream in the promoter as well as two distinct downstream loci. The multiplicity of sites may reflect the complex regulation of MITF expression by POU3F2. Goodall et al. (Goodall et al., 2008) have previously reported that overexpressed POU3F2 represses MITF expression. Here we also show that POU3F2 overexpression represses endogenous Mitf-m and, consistent with the results of Goodall et al., we have observed repression of a luciferase reporter under the control of the Mitf-m promoter (our unpublished data). Furthermore, we have used RT-PCR to verify the expression of more than 20 POU3F2 target genes and Mitf-m is the only one where we see significant repression. In contrast, Wellbrock et al., (Wellbrock et al., 2008) have shown that in some melanoma cell types POU3F2 mediates the regulatory effect of the BRAFV600E mutation that leads to constitutive activation of the ERK pathway and induces MITF expression. In accordance with this, we clearly show that POU3F2 is a transcriptional activator in melanoma cells that can potently stimulate expression via binding sites in several of its target genes.
As previously proposed (Goodall et al., 2008), POU3F2 may be an activator at some genes and a repressor at others depending on promoter context. Alternatively, to reconcile what may appear to be contradictory observations, we suggest that the POU3F2 binding to the site close to the TATA element may repress Mitf-m transcription through steric hindrance of the basal transcription machinery (for example TFIID that contacts larger regions of promoter DNA than TBP alone), while it may activate transcription of this gene through the other sites in the promoter and introns. The ability to activate or repress may then be modulated by POU3F2 concentration, where it can efficiently compete with TFIID binding and repress MITF expression only at high concentrations. This model would reconcile the observation that POU3F2 is a transcription activator with the observation that cells expressing high POU3F2 show little or no MITF in melanoma tumours (Goodall et al., 2008; Pinner et al., 2009).
POU3F2 overexpression and siRNA-mediated knockdown indicate that only a very small subset of the potential target genes is regulated in 501Mel cells. Most of these genes are direct targets observed in the ChIP-chip assay, but others like the immediate early genes Egr1, Fos and Fosb may be induced due to activation of KITL signalling. Amongst the genes that are directly regulated by POU3F2 in 501Mel cells several are of particular interest in melanoma. The relevance of Serpini1, Napl2, Uts2d and Csn1s1 is unclear, but Itga4, Abcb1, Cd36, and Kitl are all of potential interest in tumourogenesis.
It has previously been shown that expression of Itga4 in B16 melanoma cells and subsequent formation of the α4β1 heterodimer led to a significant reduction in Matrigel invasion and lung metastasis suggesting that Itga4 could play a role in controlling melanoma cell metastasis at the invasive stage (Qian et al., 1994). This observation contrasts with the observation that high POU3F2 expressing cells show enhanced invasiveness.
Abcb1 belongs to a class of multi-drug (MDR) resistance molecules and is widely expressed in many types of cancers. It is expressed only at low levels in 501Mel cells and in many other melanoma cell lines and in melanoma sections, but is strongly expressed in some non-cutaneous melanomas (Chen et al., 2009). Nevertheless, transfection of Abcb1 shows that it can confer MDR in melanoma cells (Lincke et al., 1990). It is also worth noting that the Abcb5 member of this family plays a particularly important role as a marker for malignant-melanoma-initiating cells (Schatton et al., 2008).
CD36 is a multiligand receptor associated with a broad array of physiological processes and involved in a diverse series of disorders, including atherosclerosis, insulin resistance and diabetes, dyslipidemia, tumor angiogenesis, and host defense against Plasmodium falciparum (Isenberg et al., 2009; Sid et al., 2004; Silverstein and Febbraio, 2007, 2009). One function of CD36 that is relevant for tumourogenesis is as a receptor for thrombospondins. Thrombospondin 1 and 2 are potent inhibitors of angiogenesis in vivo. Cd36 acts as a receptor for the thrombospondins on endothelial cells and is necessary for their anti-angiogenic activity. The relevance of the POU3F2 regulation of this gene in 501Mel cells and in more generally in melanoma remains to be determined, although its ability to act as a receptor for thrombospondins may influence their signalling and/or migratory properties.
The results presented here indicate that POU3F2 can activate Kitl expression through a cluster of binding sites in the proximal promoter. Both the endogenous gene and a reporter containing the POU3F2 binding sites are activated by overexpression. Somewhat contradictory results have been reported on the importance of the KITL/KIT pathway in melanoma. It has been shown that exposing KIT expressing melanoma cells to KITL leads to enhanced apoptosis in vitro and enforced expression of KIT reduces tumour growth and metastasis in vivo (Huang et al., 1996, 1998; Zakut et al., 1993). POU3F2 activation of Kitl expression is therefore surprising given that the high expressing POU3F2 cell population in melanoma lesions exhibit a high invasive/metastatic phenotype. However as some types of metastatic melanoma cells do not express the KIT receptor, the KITL expression induced by POU3F2 is not likely to act in an autocrine fashion, but may act on other cells in the tumour microenvironment to facilitate invasion.
On the other hand, amplification and/or activating mutations of the KIT gene have been frequently found in mucosal, acral, and chronic sun-damaged melanomas (Garrido and Bastian, 2009). In these melanomas, the KITL/KIT pathway therefore appears to promote melanoma development. It is also interesting to note that the KITL/KIT pathway is critical for promoting melanocyte migration during embryogenesis and hence may also be important in regulating the migratory/invasive properties of melanoma cells (Wehrle-Haller, 2003). Furthermore, it has previously been shown that activation of the MAPK pathway by KITL signalling in melanoma cells leads to phosphorylation of MITF that induces its degradation via the proteasome (Wu et al., 2000). Our present results suggest that POU3F2 may modulate the properties of melanoma cells via autocrine KITL signalling. As a consequence, POU3F2 may act to repress MITF transcription by binding to its promoter and induce its degradation by activation of the MAPK pathway.
It is significant that many POU3F2 target genes are associated with membrane and interactions with the extracellular matrix and regulation of these genes could significantly alter the invasive and motile properties of melanoma cells. Although these genes are not regulated by POU3F2 in vitro in 501Mel cells, it is possible that in vivo in melanoma tumours, activation of additional pathways leading to modifications of POU3F2 (such as phosphorylation, (Nieto et al., 2007) or changes in expression of coactivators/corepressors could modify the transcriptional properties of POU3F2 to regulate a larger proportion of its target genes and modulate metastatic capacity.