Several parallels between stem cell biology and tumour behaviour have been discovered in recent times. Such commonality is apparent in the unlimited capacity for cell division together with the lack of a differentiated phenotype in embryonic and adult stem cells, traits shared with tumour cells. Differentiation is a tightly regulated process that is mediated by the actions of multiple transcription factor families. The POU domain-containing family of transcription factors contains multiple mammalian members divided into six classes, which can be expressed broadly or in a cell-specific manner, and which are regulators of cell fate decisions of many different lineages. Target gene regulation can occur via a POU factor acting alone, or in combination with other POU proteins, ubiquitous co-activators or co-repressors, or other lineage restricted transcription factors. Aberrant levels of POU proteins have been found in several malignancies, including melanoma, connecting the otherwise developmentally restricted gene regulatory functions of POU transcription factors to the critical determinants of malignant transformation. Here, we focus on the role of the BRN2 (POU3F2/N-Oct-3) transcription factor in the melanocytic lineage where it may co-ordinate normal developmental cues that can be re-activated in melanoma. Recent studies have shown BRN2 to be responsive to MAPK pathway activation and to modulate the levels of MITF so as to suppress the differentiated melanocytic phenotype and to enhance tumour metastasis.
POU domain transcription factors are present in many cell lineages where they perform varying functions, either as ubiquitous regulators of ‘house-keeping’ genes, or as developmental- and lineage-specific coordinators of cell fate decisions. Specific targets of many POU proteins have been identified, and include the promoters of genes involved in DNA replication, development of the numerous specialized cell types, as well as viral genes. Selected examples of such genes will be drawn upon throughout this article to help describe the mechanisms, subdomain conformations adopted and interactions with cofactors and partner proteins that POU proteins use to control cell function and cell fate.
After providing an overview of POU domain structure, and mechanisms regulating DNA binding and transcriptional outcomes, this review will focus on how POU proteins may act during differentiation of stem or progenitor cells, and in the relationship between stem cells and tumour cells, with an emphasis on the melanocytic lineage. In this cell type, two POU domain transcription factors are predominantly detectable by DNA binding analysis of nuclear extracts, OCT1 and BRN2, the latter also known as N-Oct-3 when in complex with its DNA target site (Cox et al., 1988; Schreiber et al., 1990; Sturm et al., 1991). The N-Oct-5 complex also detected in melanocytic cell nuclear extracts was demonstrated to be due to proteolytic cleavage of the BRN2 protein during extract preparation (Atanasoski et al., 1997; Smith et al., 1998). We will concentrate on the BRN2 transcription factor because of its demonstrated role in melanocytic cell growth, differentiation and transformation.
POU domain structure
The POU domain itself is a highly conserved DNA binding structure that has defined this family of transcriptional regulators (Herr et al., 1988; Sturm and Herr, 1988). It consists of an N-terminal POU-specific (POUS) domain of approximately 75 amino acids and a C-terminal POU-homeo (POUH) domain of 60 amino acids which are tethered together by a less conserved linker region of variable length (see Figure 1A; reviewed by Ryan and Rosenfeld, 1997). Both the POUS and POUH domains form helix-turn-helix (H-T-H) motifs, utilizing the second and third helix of the subdomain. The third helix (second of the H-T-H motif) is the DNA recognition helix and is therefore responsible for contacting DNA and conferring target DNA sequence binding specificity. In mammals, six classes of POU proteins (POU1F to POU6F, see Table 1) have been described based upon amino acid sequence of the POU domain as well as the extent of linker region conservation (Ryan and Rosenfeld, 1997). Some POU factors exhibit widespread expression (e.g. OCT1), whereas others are restricted to certain cell types (e.g. BRN2).
Table 1. Mammalian POU domain containing gene family members and expression patterns
dMultiple designations have been used for most proteins encoded by the POU gene family members, in particular OCT4 is also commonly known as OCT3.
Nervous system, kidney
Blastocyst, ES cells, brain
ES cells, Oocytes
POU protein DNA binding
Interaction of the POU domain with DNA was first found to occur in a sequence specific manner by binding to the canonical 5′-ATGCAAAT-3′ octameric sequence, but the flexible nature of the linker region between the POUS and POUH subdomains allows for re-orientation of the subunits such that divergent octamer-related sequences can also be recognized (see Figure 1B; Herr and Cleary, 1995). Protein crystallography studies (Klemm et al., 1994) of the OCT1 protein bound to the octamer sequence have shown that the POUS and POUH domains both bind to the major groove of the DNA strand but on opposite faces of the DNA helix, such that each can be considered as half-sites with the POUS domain binding to the 5′-ATGC and the POUH domain to the 3′-AAAT. The linker region was not visible in the crystal structure, implying that it had no rigid conformation, which is supported by the varying lengths and amino acid sequences between the different classes of POU domain proteins. However in the case of BRN2, the linker has been recognized to contain a short α-helical fold, a property shared with other POU3F proteins (Blaud et al., 2004).
The linker region enables POU proteins to have great flexibility in target sequence recognition, and allows for (i) differences in orientation of the recognition half-site sequence, (ii) orientation of the POUS and POUH domains, (iii) introduction of extra bases into the recognition motif and (iv) differences in positioning of the subdomains relative to each other. Hence, POU proteins can produce different transcriptional outcomes in a manner dependent on the context of their binding site. For example, in lactotropes the differential actions of PIT1 on the growth hormone (GH) and prolactin gene promoters giving repression and activation, respectively, are a direct result of spacing of the POUS and POUH recognition half-sites. This is despite the otherwise almost perfect sequence conservation between the binding sites (Scully et al., 2000). Crystallization studies showed that the DNA contacts made by the subdomains of each monomer were separated by 4 bp on the prolactin promoter elements, whereas this spacing is increased to 6 bp in the GH promoter. This 2 bp insertion was sufficient to alter the orientation of the subdomains on the DNA element, such that for the prolactin promoter, the POUS and POUH of each monomer were on opposite faces of the DNA helix, whereas they are on the same face on the GH promoter.
As would be expected, mutation of the POU domain can affect DNA binding, either via reduced target affinity or altered sequence selectivity (Van Leeuwen et al., 1995). Alternatively, interactions with transcriptional co-factors (see below) can be disrupted, leading to reduced transcriptional activation (Miyata et al., 2006; Vallette-Kasic et al., 2001). Consistent with this, point mutations within the POU domains of three POU proteins have been linked to inherited human disorders (Table 1). Two of these genes, BRN4 (De Kok et al., 1997) and BRN3C (Vahava et al., 1998), have been determined as the cause of the X-linked DFN3 and the autosomal dominantly inherited DFNA15 deafness phenotypes respectively. Several studies have reported that mutation of the PIT1 gene causes combined pituitary hormone deficiency, which can be inherited in either a dominant or recessive fashion (Malvagia et al., 2003; Miyata et al., 2006; Vallette-Kasic et al., 2001). For all three proteins, varying aspects of function such as DNA binding or co-factor recruitment would be expected to be disrupted in the mutant protein, leading to altered transcriptional output. This was recently studied for two missense mutations of the BRN3C gene (Collin et al., 2008), both of which resulted in mis-localization of the mutant protein as well as decreased DNA binding activity.
Mechanisms regulating POU protein DNA binding and transactivational activity
Target site sequence
POU proteins are also capable of dimerization on DNA target binding sites (Figure 1C), and this occurs in a highly cooperative fashion on sequences selected by in vitro binding experiments (Rhee et al., 1998). Later studies (Millevoi et al., 2001) reported that, at least for the BRN2 protein, homodimerization on an octamer-like sequence of the L-amino acid decarboxylase (AADC) gene occurred in a non-cooperative fashion, and this was subsequently extended to include the corticotropin releasing hormone (CRH) and the aldose C gene promoters (Blaud et al., 2004). Notably, for all three of these promoters, the BRN2 dimerization motif overlapped with a binding motif for HNF-3β/FOXA2 which also enabled BRN2-HNF-3β/FOXA2 heterodimerization. Furthermore, the OCT1 POU domain bound only as a monomer to these promoters, even in the presence of BRN2 or HNF-3β/FOXA2 (Blaud et al., 2004).
The PORE (palindromic Oct factor recognition element) motif was identified within the intronic enhancer of the murine osteopontin (Opn) gene, and is considered to be a direct target of OCT4 (Botquin et al., 1998). The consensus MORE (more PORE) sequence was devised by converting the POUS binding sites of the PIT1 dimerization site of the prolactin promoter to perfectly match the consensus octamer motif, with homolgous sequences subsequently found to be present in a number of genes (Tomilin et al., 2000). Notably, these binding sites allow for reorientation of the bound POU domains, such that each POU monomer on the MORE motif binds to one face of the DNA molecule (Figure 1C), and thus does not tolerate insertion of more than two additional base-pairs into the sequence between the segments bound by the POU-subdomains (Tomilin et al., 2000). The N-Oct-3 responsive element (NORE) was identified through molecular modelling of BRN2 dimierzation, and is common to the AADC and CRH gene promoters (Alazard et al., 2005). The NORE differs from the MORE and PORE in that the POUH domains bind to overlapping sites (Figure 1C).
The option of choosing from three POU domain homodimerization conformations, with possibly others yet to be characterized, provides a high degree of complexity and allows for exquisite subtlety in directing gene transcription and interaction with cofactors. As a consequence it goes directly against the paradigm that one consensus DNA motif is recognized by one transcription factor. The PORE mode of binding is such that the recognition sequence is contiguous and POUH dominant with the relative tetramer positions similar to the POU–octamer interaction (Figure 1B). MORE mode binding has a discontiguous target sequence with half-sites on opposite strands and is more dependent upon POUS DNA interactions. The NORE mode is again POUH dominant but with the two half-sites overlapping explaining the non-cooperative nature of this homodimerization. The BRN2 POU-linker α-helix is important for the ability to recognize some NORE binding sites (Alazard et al., 2007).
Several post-translational modifications have been described to occur on POU proteins, including oxidation, phosphorylation, ubiquitinylation, sumoylation and glycosylation. Firstly, cysteine oxidation of the BRN2 protein has been shown to decrease binding to both consensus and divergent octamer sequences through inclusion of copper (II), iron (III) and nickel ions, hydrogen peroxide or diamide in the DNA binding assays, which could be rescued with the reducing agent dithiothreitol (Smith et al., 1998). OCT4 DNA binding has also been documented to be decreased by diamide oxidation which also could be rescued with dithiothreitol or the reducing enzyme thioredoxin. Furthermore activity of a reporter construct regulated by an OCT4 responsive element was increased upon co-transfection of OCT4 and thioredoxin expression vectors compared with OCT4 alone (Guo et al., 2004). DNA binding of the OCT2 protein had earlier also been reported to be reversibly inhibited by oxidation (Rigoni et al., 1993).
Phosphorylation of several POU proteins has been documented to provide additional regulation of DNA binding activity. For PIT1, phosphorylation occurs on threonine 220 (T220, found within the POUH subdomain) in response to cAMP and phorbol esters, and caused either an increase or decrease in DNA binding in a manner dependent on DNA sequences flanking the core recognition site (Kapiloff et al., 1991). For the PIT1 binding element of the GH gene, phosphorylation at T220 caused a decrease of PIT1 binding activity (Kapiloff et al., 1991), which was later shown to occur in a M-phase-restricted manner (Caelles et al., 1995). Notably, T220 may also mediate interactions between PIT1 and ETS1 in the regulating the prolactin promoter (Augustijn et al., 2002). The DNA binding activities of OCT1, OCT2 and BRN5 have also been documented to be regulated by phosphorylation (Grenfell et al., 1996; Kasibhatla et al., 1999; Pevzner et al., 2000).
Eisen (1996) has proposed that BRN2 is phosphorylated by protein kinase A (PKA) in response to cAMP, and that this excludes BRN2 from the nucleus preventing interaction with its target DNA binding sites. Recent data (Nieto et al., 2007) has reported the phosphorylation of the human BRN2 POU domain within a conserved RKKRTSI motif of the POUH domain at Serine360 (BRN2-S360) by purified PKA, and by PKA present within nuclear extracts of human melanoma cells. Furthermore, this phosphorylation inhibited binding to PORE and NORE recognition sites, but had no effect on binding to MORE sites, suggesting that BRN2 phosphorylation within a cellular context may be able to direct promoter selectivity by altering target sequence binding capability. Such a scenario was recently described for the murine Brn2-dependent activation of the nestin gene in embryonic cortical neural progenitor cells (Sunabori et al., 2008). In these experiments, Sunabori et al. were able to determine that the Brn2 transcription factor can be phosphorylated on the RKKRTSI murine equivalent position of Serine362 (Brn2-S362), and that when phosphorylated, DNA binding activity was reduced. Furthermore, this phosphorylation event occurred in a cell-cycle dependent manner such that Brn2-S362 was phosphorylated during G2-M, thereby allowing cell-cycle stage regulation of nestin transcription. Combined mutation of Brn2-S362 and threonine-361 (each to alanine) causes only a slight increase in proliferation when overexpressed in murine melanocytes compared with the increased proliferation observed with overexpressed wild-type Brn2 protein, both in vitro and in vivo (Larue et al., 2007).
Using predictive bioinformatic analysis the potential interplay between phosphorylation and O-GlcNAc modification at conserved Ser/Thr residues of the OCT2 protein has been tested (Ahmad et al., 2006). It has been postulated that alternative modifications of the same amino acid (Yin Yang sites) may produce differential binding behaviour of the modified protein to DNA target motifs. The OCT4 transcription factor has also been show to be subject to ubiquitination and sumoylation. The ubiquitin ligase Wwp2 specifically interacts with OCT4 and promotes its ubiquitination both in vivo and in vitro, moreover, this modification dramatically suppresses its transcriptional activity in embryonic stem cells (Xu et al., 2004). In addition, it has been demonstrated that OCT4 is modified by SUMO-1 and disruption of this process decreased protein stability and self-renewal capacity in ES cells (Zhang et al., 2007), with sumoylation resulting in increased stability, DNA binding, and transactivation capacity (Wei et al., 2007).
Nucleocytoplasmic shuttling has recently been reported for the POU3F class proteins OCT6 and BRN4 (Baranek et al., 2005). This occurred in a CRM1 nuclear export factor dependent fashion via a hydrophobic and leucine rich region that is highly conserved in many POU proteins, including BRN2, and which also forms part of the second helix of the POUH domain. Blockage of nuclear export with the CRM1 inhibitor leptomycin B caused a reduction in OCT6 dependent reporter gene activation, suggesting that nucleocytoplasmic shuttling is required for maintaining the transactivation competency of the protein (Baranek et al., 2005). Cytoplasmic localization has also been reported for BRN2 in the developing mouse spinal cord (Tanaka et al., 2004), although it remains to be demonstrated that this occurs via a mechanism similar to that observed for OCT6. In cultured melanocytic cells, BRN2 appears to be exclusively nuclear (Cook et al., 2003), however, experiments similar to those conducted by Baranek et al. (2005) may reveal if nucleocytoplasmic shuttling does occur and if this is similarly required for the transactivation potential of the protein. The recent report of a cytoplasmic localization for BRN2 in nevi (Goodall et al., 2008; Maier et al., 2006) suggests that nucleocytoplasmic shuttling may occur in melanocytic cells.
POU-partner protein combinatorial gene regulation
Families of transcription factors often co-ordinate their activities via interactions with co-activators, co-repressors or act in a complex with transcription factors from the same or other families (reviewed by Remenyi et al., 2004). This combinatorial recruitment of specific co-regulators for POU proteins is in part determined both by the accessibility of the interaction surface patches mediating complex formation and the DNA sequence of the target site, thus providing a means by which lineage-specific gene expression is achieved.
POU domain transcription factors are able to interact with a variety of other proteins required to provide the nucleation site for generation of a transcription unit. For example, OCT1 and OCT2 both interact with general transcription factors including TATA-binding protein (TBP, Zwilling et al., 1994), and OCT1 also forms ternary complexes with octamer-containing DNA oligonucleotides and TFIIB (Nakshatri et al., 1995). BRN2 has also been shown to interact in mammalian two-hybrid experiments with TBP, and GST-fusion protein pulldown assays revealed direct protein-protein contacts between BRN2 and each of the TBP and TFIIB proteins (Smit et al., 2000).
Several POU proteins have been shown to interact with either of the structurally analogous p300/CREB-binding protein (CBP) histone acetyltransferase transcriptional co-activators. For example, BRN2 has been documented to directly interact with p300 using several experimental systems (Smit et al., 2000). The transcriptional outcome for these interactions may be dependent on the cellular context. Thus, CBP enhances PIT1 activation of a PIT1-responsive element, whereas CBP overcomes the SKN1A-mediated repression of the keratin 14 promoter (Sugihara et al., 2001; Xu et al., 1998).
Interactions with lineage-specific co-activators also occur, as demonstrated by the B-cell-specific co-activator, OCAB (also known as Bob1 and OBF-1, Luo and Roeder, 1995). Crystallography studies using the OCT1 POU domain, the canonical octamer motif and an OCAB peptide corresponding to the regions required for stable complex formation by Chasman et al. (1999) indicate that OCAB binds near the centre of the octamer motif (Figure 1D, left panel), and forms contacts with both the POU domain and the octamer DNA motif. POU protein co-factor recruitment is also dependent on the sequence of the binding motif, as OCAB has a specific requirement for an adenosine at position 5 of the octamer sequence (Cepek et al., 1996; Gstaiger et al., 1996). Furthermore, the alternative POU protein binding conformations on different dimerization motifs regulates co-factor recruitment, as the OCT1 dimer formed on the PORE allows OCAB binding, whereas the MORE sequence-mediated dimer does not (Tomilin et al., 2000).
In addition to co-activators, POU factors can recruit co-repressors to regulate transcriptional responses. Recently, OCAB-dependent activation of reporter genes by OCT1 and OCT2 was shown to be repressed by members of the Groucho/TLE family of co-repressors (Malin et al., 2005). These studies demonstrated an ability of TLE family members to specifically repress the activity of different POU proteins, as well as mediating differential selectivity for the monomeric versus dimeric binding of the same POU factor. PIT1 is required for expression of several endocrine-related genes within different cell types of the pituitary (Scully and Rosenfeld, 2002). One such gene is the GH gene, which is expressed by somatotropes, but not lactotropes, where repression is mediated via recruitment of a nuclear receptor co-repressor (NCoR) by the PIT1 homodimer bound to the promoter (Scully et al., 2000).
POU proteins are capable of mediating transcription of viral immediate early genes, which are transcribed upon viral infection, and whose protein products are required for subsequent stages of infection (reviewed by Latchman, 1999). For example, infection by papovirus JC, which causes progressive multifocal leukoencephalopathy, is specific to glial cells because of postnatal expression of OCT6 being restricted to this cell type allowing subsequent synergistic transcription of genes from the early and late stages of infection with the JC virus large T antigen (Renner et al., 1994). Similarly, OCT1 forms multiprotein complexes with the herpes simplex virus VP16 transactivator protein, which also involves the accessory protein HCF (host cell factor, see Figure 1D right panel; reviewed by Wysocka and Herr, 2003).
Whilst no cell type-specific cofactor has been identified for BRN2, Dugast and Weber (2001) have shown an interaction of BRN2 with the ubiquitous NFY transcription factor in regulation of the AADC gene promoter in neuroblastoma cells. This interaction was independent of BRN2 DNA recognition, as mutations of the divergent octamer motif of the promoter to sites that did not allow N-Oct-3 complex formation in DNA binding experiments still allowed transactivation of reporter genes in neuroblastoma cells when BRN2 was ectopically expressed. This is similar to the DNA binding-independent activity of OCT1 in mediating the CREB responsiveness of the cyclin D1 promoter in MCF-7 breast cancer cells (Boulon et al., 2002). Furthermore, the BRN2-NFY interaction may also involve other proteins thought to stabilize DNA binding (Dugast and Weber, 2001). Consensus binding sites that are bound by NFY are present in the Ras-responsive region of the murine Brn2 gene promoter (Goodall et al., 2004b) although it remains to be determined if this enables recruitment of BRN2 to provide autoregulation as described for other POU factors (Chen et al., 1990; Malik et al., 1996; Mccormick et al., 1990; Trieu et al., 2003).
The POU-SOX code
In addition to recruitment of general transcriptional regulatory proteins to coordinate gene expression, POU domain proteins also interact with other families of lineage-restricted transcription factors. One such family are the SOX proteins which like POU factors are critical regulators of cell fate and specification (Wegner, 2005), and it is notable that the combinatorial actions of POU and SOX proteins constitute part of a genome wide transcriptional regulatory code for many genes expressed in the central nervous system (Bailey et al., 2006). The POU-SOX code stipulates that a specific POU protein will only interact with a specific SOX partner to elicit a transcriptional response and was first proposed based on observations in glial cells (Kuhlbrodt et al., 1998a,b). This cell type expresses multiple POU (BRN1, BRN2 and OCT6) and SOX (SOX4 and SOX11) factors, and thus requires tight co-ordination of POU-SOX interactions to prevent aberrant gene activity. To this end, recombinantly expressed SOX11 was found to form complexes with both OCT6 and BRN1 in DNA binding assays using combined target sites comprised of SOX and octamer DNA binding elements (Kuhlbrodt et al., 1998a) derived from the FGF4 gene enhancer, a well established target of POU and SOX proteins (Ambrosetti et al., 1997; Yuan et al., 1995). However, reporter gene activity under the control of a promoter consisting of multimerized FGF4 enhancer sites was only synergistically activated by SOX11 and either of BRN1 or BRN2, but not when in combination with OCT6. Synergistic activation by SOX10 in combination with either of the BRN2 or OCT6 factors was recently reported for the Krox20 gene promoter (Ghislain and Charnay, 2006).
Several other promoters or enhancers have been identified as being regulated via co-binding of POU and SOX proteins, most of which are targets of OCT4 and SOX2 in embryonic stem (ES) cells (Table 2). Notably, depending on the sequence of the regulatory element, the same two proteins can elicit different transcriptional responses which include synergistic and additive activation, as well as antagonism of OCT4-mediated activation of the Opn gene enhancer through binding of SOX2 (Botquin et al., 1998). Furthermore, the promoters of the Oct4 and Sox2 genes contain POU and SOX sites that allow formation of an autoregulatory loop, such that both genes are regulated by the OCT4 and SOX2 proteins (Chew et al., 2005; Okumura-Nakanishi et al., 2005; Tomioka et al., 2002).
Table 2. Gene regulatory elements bound by POU and SOX proteins
Many of the genes regulated by co-binding of POU and SOX factors encode transcription factors whose expression is also a critical determinant of cell fate. For example, SOX10 and either of OCT6 or BRN2 regulate Krox20 expression (Ghislain and Charnay, 2006), which when inactivated in mice blocked Schwann cell differentiation and thereby caused a failure of peripheral nervous system myelination (Topilko et al., 1994). Similarly, in ES cells OCT4 and SOX2 combine to regulate Nanog expression (Kuroda et al., 2005; Rodda et al., 2005), a homeodomain-containing transcription factor which promotes ES cell self renewal (Chambers et al., 2003). POU-SOX interactions can also repress genes that encode transcription factors that, when expressed, are implicated in differentiation of various tissues. This was extensively studied by Boyer et al. (2005) who employed chromatin immunoprecipitation and DNA microarray analysis to identify promoter regions bound by OCT4, SOX2 and NANOG in human ES cells. Thus, POU-SOX interactions can promote adoption of a particular cell fate (or maintain pluripotentcy in the case of OCT4 and SOX2) while also repressing genes that would otherwise lead to differentiation of a divergent lineage. This is achieved via a specific transcriptional outcome (that can be either transcriptional activation or repression) and which is influenced by the relative position, spacing and orientation of target sites, and on the milieu of POU and SOX family members present in a particular cell type.
The correlation of BRN2, SOX9, SOX10 and nestin gene expression in human melanoma cell lines and tissues has recently been examined (Flammiger et al., 2008). Downregulation of either SOX9 and of SOX10 by siRNA treatment markedly decreased nestin levels in these lines, whereas BRN2 ablation did not. Thus no melanocytic gene promoter/enhancer under regulation by a coordinate POU-SOX site has been identified to date (we exclude the MITF promoter in this instance as the POU and SOX target sites are not juxtaposed), but the co-ordinated downregulation of BRN2 and SOX10 concomitant with increased expression of SOX9 in cultured human melanoblasts after induction of differentiation (Cook et al., 2005) and the ability of BRN2 and SOX10 to physically interact (Smit et al., 2000), suggests that the functionality of the POU-SOX code in melanocytic cells is likely to be influenced by the relative expression levels of all POU and SOX proteins expressed in this cell type.
PAX family transcription factors
Both POU and SOX transcription factors are able to interact with transcription factors of the PAX family. The crystal structure of the OCT1/SOX2/FGF4-enhancer ternary complex has been solved and revealed that a highly conserved region of the C-terminal portion of the SOX protein HMG domain interacts with a similarly highly conserved region of the POU-specific domain (see Figure 1E; Remenyi et al., 2003). Further experiments revealed that this interaction surface of the SOX2 protein was also recognized by the PAX6 protein in DNA binding assays (Remenyi et al., 2003). Thus, when expressed in the same cell type, PAX proteins may serve as likely co-regulators for the interactions between POU and SOX proteins.
The PAX family member PAX3 is expressed in melanocytic cells, mutation of which is responsible for Waardenburg syndrome types I and III (Bennett and Lamoreux, 2003). While the exact regions of interaction were not determined, physical interactions between BRN2 and PAX3, BRN2 and SOX10, and SOX10 and PAX3 have been reported (Smit et al., 2000). Additionally, the ability of SOX10 and PAX3 to mediate synergistic transcriptional activation of the MITF/Mitf gene promoter is well established (Bondurand et al., 2000; Potterf et al., 2000). Clearly, expression of PAX3 in melanocytic cells will also impact on the gene regulatory activities of BRN2 and SOX10, whilst also conducting any POU-SOX factor independent activities. This is exemplified by the ability of PAX3 to reduce SOX10-mediated activation of the Dct promoter in co-transfection studies, under conditions where PAX3 activated the promoter of one of its known melanocytic targets, Mitf (Lang et al., 2005).
Other known interactions
The murine Brn1 and Brn2 proteins have recently been shown to interact with the proneural basic-helix-loop-helix transcription factor Mash1 (Castro et al., 2006). This synergistic transcriptional activation occurs via an evolutionarily conserved octamer-E-box motif (termed the Mash1/Brn motif) first identified within the Mash1-specific enhancer of the Delta1 gene, which codes for one of several ligands for the Notch receptor. Notably, using bioinformatics, the co-ordinate Mash1/Brn motif was able to be identified in the promoters of 21 other genes, and which also included several other members of the Notch signalling pathway (Castro et al., 2006).
The BRN2 protein contains a 21 amino acid polyglutamine tract, which at the genomic level is encoded by CAG trinucleotide repeats (Hara et al., 1992; Schreiber et al., 1993; Thomson et al., 1995). Using this region as bait, Waragai et al. (1999) performed a yeast two-hybrid screen with the aim of identifying proteins that interacted with BRN2 via the polyglutamine tract. This approach identified PQBP-1, which localized to the cell nucleus, and decreased the ability of BRN2 to activate transcription in a reporter gene assays, suggesting interactions between PQBP-1 and BRN2 may act to negatively regulate BRN2 activity in the cell. Notably, a coding mutation in the RPF1 protein encoded by the POU6F2 gene (Table 1) identified in a Wilms tumour patient by Perotti et al. (2004) was predicted to disrupt a 10 residue polyglutamine tract, although the effect of this mutation of RPF1-dependent transcription is yet to be assessed.
POU protein redundancy
A number of studies have demonstrated that POU proteins can have overlapping functions in regulation of target gene expression. For example, a compensatory role for Brn1 in Brn2 gene nullizygous mice was first suggested by Nakai et al. (1995) due to an increase in Brn1 DNA binding activity in extracts of brain tissue, however this was insufficient to fully offset the lack of Brn2 as knockout animals died 10 days after birth. Partial redundancy between the POU3F family proteins in oligodendrocytes has been reported (Schreiber et al., 1997), and more recently, full rescue of Oct6 deficiency by Brn1 during the development and maturation of murine Schwann cells in vivo was described (Friedrich et al., 2005). Similarly, redundancy of Brn1 and Brn2 in cortical neuron migration had been demonstrated (Mcevilly et al., 2002; Sugitani et al., 2002). POU protein redundancy is not only restricted to the POU3F class, as replacement of the murine Brn3b locus with Brn3a gene coding sequences allows normal development of the retina (Pan et al., 2005). Whether POU protein redundancy extends to, or even across, other classes remains to be determined.
BRN1/N-Oct-2 DNA binding activity has been detected in nuclear extracts of both melanoma cell lines and melanoma biopsy specimens (Thomson et al., 1993), but was only present when BRN2/N-Oct-3 was also detectable. A redundant role for BRN1 and BRN2 in melanocytic cells is yet to be documented, and the lack of BRN1 expression in BRN2-ablated melanoma cells and the parental control cell line (Thomson et al., 1995) does not point to a simple inverse relationship between these proteins in the melanocytic lineage.
POU factors in differentiation and malignancy
One of the earliest genes expressed from the zygote genome is the OCT4 POU domain transcription factor, which is required for the generation of pluripotent ES cells (Nichols et al., 1998). Cell fate decisions co-ordinated by this protein are extremely sensitive to changes in OCT4 dosage (Niwa et al., 2000), with a two-fold change in expression level causing an increase or decrease sufficient to cause a loss of the differentiated stem cell phenotype, and cause a differentiation to endodermal and mesodermal fates, or to trophectodermal fates respectively. Tai et al. (2005) have reported that OCT4 expression is retained by adult stem cells from various tissues, but this has recently been disputed (Lengner et al., 2007).
Downregulation of POU proteins concomitant with differentiation of other neural crest derived cell lineages has also been reported, most notably the loss of Brn2 (and also Oct6) during Schwann cell transition from a promyelinating to a myelinating phenotype (Jaegle et al., 2003), and is also downregulated during human melanoblast differentiation to pigmented melanocytes in culture (Cook et al., 2003). Within mouse embryos, Brn2 DNA binding activity has been detected in explanted cultures from E9 neural crest (Cook et al., 2003), however this differs from the lack of Brn2 mRNA in murine E11.5 melanoblasts in vivo (defined by expression of Dct-driven lacZ activity) and the detectable expression observed in newborn mouse hair follicle melanocytes (Goodall et al., 2004a). De-differentiation of human melanocytes through passage in culture medium suitable for growth of cells with a melanoblast phenotype caused a decrease in cell pigmentation and an increase of BRN2 protein and DNA binding levels because of the synergistic actions of stem cell factor, endothelin-3 and fibroblast growth factor-2 (Cook et al., 2003), and similarly, retroviral-mediated BRN2 overexpression causes acquisition of a less differentiated phenotype by cultured mouse melanocytes (Goodall et al., 2004b). BRN2 has also been shown to be downregulated in response to differentiating agents in human melanoma cell lines (Sturm et al., 1991). Thus, there appears to be a reciprocal relationship between the level of BRN2 and the differentiated phenotype of melanocytes such that higher amounts of BRN2 reflect an undifferentiated state, whereas lower levels are permissive for melanogenesis.
Cancer tissues of many different origins express POU proteins (Table 3 and below). Recently, Perotti et al. (2004) reported the finding of two germline mutations of POU6F2 in Wilms tumour patients. Both mutations were in the heterozygous state in the constitutional DNA of the Wilms tumour patients, but were coupled with loss of the wild-type allele in tumour DNA. This prompted speculation that the encoded RPF1 protein may act as a tumour suppressor for at least a subset of this tumour type (Perotti et al., 2004), but as yet no studies showing a causal role for these mutations in Wilms tumour have been performed.
Table 3. POU proteins in human cancer
aIncludes POU proteins that have been ascribed a physiological process in a particular cancer type.
bIncludes POU proteins that have been shown to be differentially expressed between subtypes of the same cancer.
cIncludes POU proteins that have been found to be genetically altered within the tumour DNA. Hence studies showing aberrant expression of POU factors in cancer compared with the non-malignant cell type, or expression in a particular cancer type are not included here.
No overt rearrangements of the intronless BRN2 gene have been observed in any cancer, however, gene amplification has been found in several melanoma cell lines (Thomson et al., 1995), consistent with BRN2 being more highly expressed in melanoma cell lines than in melanocytes (Cox et al., 1988; Eisen et al., 1995; Sturm et al., 1994; Thomson et al., 1993). This is also likely to be mediated via a number of autocrine growth factor mechanisms (Berking et al., 2004; Satyamoorthy et al., 2003) or by mutation of components of key signal transduction mechanisms such as those involving BRAF and β-catenin (Goodall et al., 2004a,b) that can be present in melanoma (Figure 2A). Interestingly, BRN2 expression in retinoblastoma cells is repressed by the wild-type, but not mutant, Rb protein (Cobrinik et al., 2006), and although Rb mutation is rare in melanoma, melanocytes cultured from Rb-null mice display a semi-transformed phenotype (Tonks et al., 2005), which may indicate a role for the Rb protein in regulating BRN2 in the melanocytic lineage. Rb can also cooperate with Mitf to promote cell-cycle exit and melanocytic differentiation (Carreira et al., 2005), and thus Rb may serve as a regulator of melanocyte homeostasis (reviewed by Tonks et al., 2006) in part by co-ordinating the relative expression/activities of both BRN2 and MITF.
It is possible that both autocrine growth stimulation and BRAF mutation are required to increase BRN2 in melanoma cells as the effects of specifically silencing mutant BRAF expression in melanoma cell lines could be overcome by culture supplementation with FGF2 or hepatocyte growth factor, but only when a wild-type allele was present (Christensen and Guldberg, 2005). Furthermore, BRN2 protein and DNA binding activities are lower in more differentiated or pigmented melanoma cell lines than in non-pigmented lines (Goodall et al., 2008; Cook et al., 2003; Thomson et al., 1995), reflecting the relationship seen between BRN2 and pigmentation in non-malignant melanocytes. BRN2 contributes to more than just the differentiated phenotype of melanocytic cells, as antisense-RNA mediated BRN2 ablated melanoma cell lines are not tumourigenic when explanted into nude nice (Thomson et al., 1995), and cell lines with higher levels of BRN2 are more invasive in vitro (Goodall et al., 2008).
POU proteins other than BRN2 have been implicated in malignancies of various tissues (Table 3). For example, an oncogenic function for OCT4 has been described in several experimental systems (Abate-Shen, 2003). In a manner analogous to its role during embryogenesis, OCT4 conferred tumourigenicity to murine ES cells in a dosage-sensitive fashion, such that lesions resulting from higher protein expression produced tumours displaying a histologically more malignant phenotype (Gidekel et al., 2003). Additionally, this study showed that overexpression of OCT4 in mouse fibroblasts was sufficient for tumourigenicity, and that this was abrogated by mutation of either the POU domain or the C-terminal domain of OCT4. A similar approach was undertaken by Hochedlinger et al. (2005) who engineered transgenic mice that enabled ectopic expression of OCT4 in response to dietary doxycycline. When the exogenous protein was induced, dysplasia of much of the gastrointestinal tract, as well as the skin was observed, and attributed to an increase in the number of progenitor cells as a result of OCT4-mediated blockage of differentiation.
The contribution of POU proteins to tumourigenicity is likely to be dependent on either direct or downstream transcriptional targets. For example, expression of BRN3b in breast cancer correlates with, and directly regulates, the expression of heat-shock protein-27, a protein associated with increased proliferation, invasion and chemo-resistance in this tumour type (Lee et al., 2005). In melanoma cell lines, abrogation of BRN2 expression causes a slowing of proliferation rates, matrigel invasion, and a loss of tumourigenesis (Goodall et al., 2008, 2004a,b; Thomson et al., 1995) although the transcriptional targets of BRN2 whose gene products mediate these processes remain to be fully investigated.
The only BRN2 targets identified in melanocytic cells are GADD45 and MITF. The promoter of the GADD45 gene is activated by BRN2 in response to UVB (Lefort et al., 2001). However, the basal level of GADD45 present in the BRN2-ablated melanoma cells generated in this study was not reported, and thus it is not clear if BRN2 homeostatically regulates GADD45, or is only required for high level expression following UVB exposure. Others have reported similar actions mediated by OCT1, including GADD45 expression in response to UV and DNA damage, but in cell types that do not express BRN2 (Jin et al., 2001; Takahashi et al., 2001).
Unlike for GADD45, BRN2 acts as a repressor of MITF expression in melanoma cells via binding directly to a region adjacent to the TATA box (Goodall et al., 2008). Importantly, this report showed that while BRN2 and MITF can be co-expressed in melanoma cell lines, it appears that in melanoma tissue samples, BRN2 and MITF are restricted to different subsets of cells within the same tumour. The recent reporting of melanoma cells with lower levels of MITF being more invasive (Carreira et al., 2006) clearly support previous data implicating BRN2 in melanoma tumourigenicity (Thomson et al., 1995). It is remains to be determined if the contribution of BRN2 to melanoma invasion is limited to repression of MITF, or if other target genes are also involved.
In contrast to these results, it has been reported that in some human melanoma cell lines, BRN2 activates the MITF promoter as a consequence of constitutive activation of the V600EBRAF oncogene (Wellbrock et al., 2008), consistent with anti-sense RNA mediated ablation of BRN2 in a human melanoma cell line causing a loss of many melanocytic markers, including MITF (Thomson et al., 1995). Notably, this MITF-promoting activity of BRN2 may differ from its role in normal human melanocytes, where transfection with V600EBRAF causes a repression of MITF (Wellbrock and Marais, 2005) and an induction of BRN2 protein levels (Wellbrock et al., 2008). In human melanoblast cultures, BRN2 expression is downregulated upon induction of differentiation by growth factor withdrawal, and this is concomitant with an upregulation of MITF and its down-stream targets TYR and TYRP1 (Figure 2B). Similar results are observed when BRN2 protein levels are reduced using siRNA, but when the cells are maintained in conditions permissive for high BRN2 expression suggesting that BRN2 may regulate melanocytic differentiation via repression of MITF. Thus it seems that the actions of BRN2 on the MITF promoter are context specific, and most likely depend on the relative expression and activities of other transcription factors (and co-factors) that co-ordinately regulate MITF (Vance and Goding, 2004). The flexibility of sequence recognition, phosphorylation pattern, dimerization capacity and interactive properties of BRN2 make both activation and repression of different gene sets possible depending on the cell state or cell microenvironment. Notably, this may contribute to the cellular heterogeneity seen in melanoma tumour samples (Goodall et al., 2008; Hoek et al., 2008).
BRN2 has been shown to be central to the proliferation of melanoma cells (Goodall et al., 2004a,b; Thomson et al., 1995). Recently, Ivan et al. (2004) examined the expression of the cell cycle inhibitor p27/Kip1, and its inactivator Jab1, in a series of nevi and melanoma samples. This revealed a correlation between decreased p27 levels and increased Jab1 expression occurred during melanoma progression from melanocytes. This inverse relationship is notable in that Jab1 has been shown to interact with the N-terminal 67 amino acids of BRN2 in yeast two-hybrid experiments (Huang et al., 2005). Furthermore, Jab1 was localized to the nucleus of proliferating cells where it induced translocation of p27 to the cytoplasm, promoting its degradation (Tomoda et al., 1999). Jab1 nuclear localization in melanocytic lesions (Ivan et al., 2004) at least provides the opportunity for interaction with BRN2, although this interaction remains to be tested using melanocytic (or any other) cell extracts. Given the proliferation promoting effects of both BRN2 and Jab1, further examination of the nuclear function of Jab1, including potential interactions with BRN2, in melanocytic cells is warranted. It is also of note that siRNA-mediated downregulation of MITF in melanoma cells causes an induction of p27 leading to cell cycle arrest, and that MITF and p27 appear to be reciprocally expressed in the melanocytic lineage (Carreira et al., 2006).
It is noteworthy that melanoma cells expressing lower levels of MITF have been suggested to have a stem cell-like phenotype (Goodall et al., 2008), and that recent studies have suggested the existence of cells with stem cell-like features in melanoma (Schatton and Frank, 2008; Schatton et al., 2008). Given the proliferation and differentiation-promoting activity associated with higher levels of MITF (Carreira et al., 2006), the reciprocal expression levels of BRN2 and MITF in melanoma suggests that BRN2 may be involved in promoting a de-differentiated phenotype via repression of MITF. This is consistent with the observed higher levels of BRN2 in cultured human melanoblasts compared with melanocytes (Cook et al., 2003), and with higher expression and DNA binding activity in non-pigmented melanoma cell lines compared with pigmented lines (Cook et al., 2005; Thomson et al., 1995). Whether increased BRN2 expression alone is sufficient to mediate these effects has yet to be investigated, however other POU domain proteins, notably OCT4, have been shown to regulate ES cell fate in a dosage sensitive fashion (Niwa et al., 2000), contribute to the reprogramming of somatic cells to form induced pluripotent stem (iPS) cells (Takahashi and Yamanaka, 2006), and to cause dysplasia in murine epithelial tissues by preventing progenitor cell differentiation (Hochedlinger et al., 2005).
BRN2 is but one of several tissue-restricted transcription factors known to be expressed in the melanocytic lineage (Vance and Goding, 2004). As presented in Figure 2A, BRN2 may act as a molecular integrator of several extracellular stimuli that subsequently assists coordination of transcriptional outcomes required for normal melanocytic development, differentiation and homeostasis (Haass and Herlyn, 2005). Notably, aberrant activity of signalling molecules involved in transduction of these stimuli have been documented to occur in human melanoma specimens (denoted by a cross in Figure 2A; Davies et al., 2002; Rimm et al., 1999). Physiological outcomes coordinated by BRN2, or downstream transcription factors such as MITF, include genes required for cell survival and maintenance as well as for melanogenesis. While there is no direct genetic evidence for BRN2 participation in determination of the pigmentation phenotype as the Brn2 nullizygous mouse exhibits no reported pigmentation defects (Nakai et al., 1995; Schonemann et al., 1995), transgenic mice expressing wild-type Brn2 under the control of the tyrosinase promoter display a slight increase in pigmentation, whereas overexpressed mutant Brn2 (S362A and T361A) results in development of a white belly spot, attributed to alteration of melanocytic proliferation during embryogenesis (Larue et al., 2007). This latter phenotype is characteristic of piebaldism and notably, mutation of several other melanocytic transcription factors (MITF, SOX10, PAX3 and SNAI2) have been implicated in Waardenburg syndrome types 1-4 (Bennett and Lamoreux, 2003; Tachibana et al., 2003). A number of studies have documented changes in pigmentation, or changes in expression of melanogenic genes including the MITF differentiation co-ordinator to occur in a BRN2-dependent fashion (Goodall et al., 2008, 2004b; Thomson et al., 1995). Whilst these studies have provided insight to the contribution of BRN2 to the phenotype and growth characteristics of cultured melanocytic cells, the full complement of direct BRN2 transcriptional targets await identification. In turn, BRN2-specific small molecule inhibitors such as reported for Brn3 and Pit-1 (Peixoto et al., 2008) may be invaluable for the development of melanoma intervention strategies if BRN2-mediated transcriptional coordination observed in cultured cells are reflective of melanocytic lesions from varying stages of disease progression.
This work was supported in part by a Queensland Cancer Fund grant to RAS and ALC. RAS is a Senior Research Fellow of the Australian NHMRC. The Institute for Molecular Bioscience incorporates the Centre for Functional and Applied Genomics as a Special Research Centre of the Australian Research Council.