Pitx3 potentiates Nurr1 in dopamine neuron terminal differentiation through release of SMRT-mediated repression

The dopaminergic neurons of the meso-diencephalic dopaminergic system (mdDA system) have been extensively studied in relation to Parkinson's disease, and pioneering studies have explored the possibility of using cell replacement therapy with stem cells as a future treatment(Freed et al., 2001; Chung et al., 2002; Kim et al., 2002; Olanow et al., 2003; Andersson et al., 2006; Hedlund et al., 2008). Ultimately, stem cells could be exploited as an unlimited source of transplantable DA neurons. However, in order to engineer stem cells with mdDA characteristics, the need to obtain the appropriate dopaminergic phenotype through correct molecular coding has been underlined(Chung et al., 2005; Martinat et al., 2006; Smidt et al., 2007). Therefore, much effort has been put into unravelling the multi-step process that produces a genuine mdDA neuronal population in vivo,as this is believed to hold the key to the successful engineering of stem cells in vitro.

One of the key regulators for development of mdDA neurons in vivo is the orphan nuclear receptor Nurr1 (Nr4a2), which is crucial for expression of a set of genes involved in DA metabolism, such as tyrosine hydroxylase(Th), vesicular monoamine transporter (Vmat2), dopamine transporter (Dat) and aromatic L-amino acid decarboxylase(Aadc) (Zetterström et al.,1997; Saucedo-Cardenas et al.,1998; Baffi et al.,1999; Le et al.,1999; Wallén et al.,1999; Smits et al.,2003). Another key factor for the development of mdDA neurons is the paired-like homeobox gene Pitx3. Because of its highly exclusive expression in mdDA neurons within the brain, Pitx3 is a selective marker for the generation of transplantable ES cells(Zhao et al., 2004; Chung et al., 2005; Hedlund et al., 2008). Pitx3 serves a unique function in mdDA neurons, highlighted by the dramatic loss of neurons of the substantia nigra in Pitx3-deficient mice(Hwang et al., 2003; Nunes et al., 2003; Van der Munckhof et al., 2003; Smidt et al., 2004), which is preceded by deficient Th expression during embryonic development(Smidt et al., 2004; Maxwell et al., 2005; Jacobs et al., 2007). Strikingly, although Pitx3 is expressed in all mdDA neurons, only mdDA neurons of the substantia nigra are affected by the loss of Pitx3, emphasizing the complicated role of Pitx3 in mdDA neurons and the diverse properties of subsets within the mdDA neuronal population.

Despite the fact that both Pitx3 and Nurr1 are crucial for proper expression of Th in mdDA neurons, it has long been assumed that Pitx3 and Nurr1 are components of independent developmental pathways(Smidt et al., 2000; Chung et al., 2005). However,there is growing evidence to suggest that these pathways are interconnected at a functional level. Studies aimed at constructing functional DA neurons from ES cells, have shown that combined transduction of Nurr1 and Pitx3 promotes the maturation of ES cells into a dopaminergic phenotype (Martinat et al.,2006). Importantly, the recent discovery of aldehyde dehydrogenase 2 (Ahd2; Aldh1a1) as transcriptional target of Pitx3 in mdDA neurons has highlighted a significant similarity between the phenotype of Pitx3- and Nurr1-null mice (Jacobs et al.,2007). Similar to what was observed in Pitx3^-/- embryos, Ahd2 expression is not detected in the midbrain area of Nurr1-deficient embryos during final differentiation of mdDA neurons (Wallén et al.,1999). These observations strongly suggest a functional relationship between the homeodomain transcription factor Pitx3 and the orphan nuclear receptor Nurr1 in the development of mdDA neurons. In this study, we deduce the molecular mechanism of this functional interplay between Pitx3 and Nurr1 in vivo, with extensive consequences for mdDA development. We establish Pitx3 as an essential regulator of Nurr1-mediated transcription in mdDA neurons, which positions Pitx3 as a crucial factor for the specification of the dopaminergic phenotype both in vivo and in stem cell engineering for future treatment of Parkinson's disease.

MN9D-Nurr1^{Tet On}13N (MN9D) cells were cultured and transfected as described previously(Jacobs et al., 2007).

MN9D cells were transfected with 0.5 μg Pitx3-pcDNA3.1(-)myc-His or an equivalent molar amount of pcDNA3.1(-)myc-His expression vector and His-tagged proteins were purified using Ni-NTA magnetic agarose beads (Qiagen) according to manufacturer's protocol. Purified proteins were separated by SDS PAGE and visualized by protein gel silver-staining. Subsequently, the gel was fixed in 50% methanol(2×15 minutes) and 5% methanol (10 minutes), rinsed with water(3×10 seconds), soaked in 10 μM DTT (20 minutes), incubated in 0.1%AgNO₃ (20 minutes), rinsed with water and incubated in developer solution (3% NaCO₃, 0.05% formaldehyde) until protein bands were visible. The reaction was stopped by adding citric acid and the gel was washed in water. Protein bands of interest were excised from the gel and subjected to nanoLC-ESI-MS Mass Spectrometry analysis (Proteome Factory).

MN9D cells or tissue was homogenized in lysis buffer [50 mM Tris HCl (pH 8), 150 mM NaCl, 5 mM MgCl₂, 0.5 mM EDTA, 0.2% NP-40, 5% glycerol and 0.5 mM DTT + Complete Protease Inhibitor Cocktail (Roche)], and subjected to IP. Two dishes (10 cm in diameter) of MN9D cells or five ventral midbrains of E14.5 C57Bl6-Jico embryos were used for each IP. Magnetic dynal beads(Invitrogen) were blocked in 0.5% BSA and incubated with antibodies overnight. Lysates were incubated with antibody-bound beads for 3 hours at 4°C. Beads were washed in lysis buffer (five times) and captured using a magnetic separator. Elution was performed in elution buffer [50 mM Tris (pH 8.0), 10 mM EDTA, 1% (v/v) SDS and Complete protease inhibitors] at 65°C with frequent vortexing. Proteins were separated on Nupage 4-12% gradient gels (Invitrogen),transferred to Hybond C extra (Amersham) and after overnight blocking in 5%milk powder in PBS at 4°C the membranes were incubated with antibodies in PBS-T [PBS + 0.05% (v/v) Tween-20]. Antibodies used: anti-Nurr1 [E-20; Santa Cruz (SC)], 1:100; anti-Pitx3 (Smidt et al., 2004), 1:10.000; anti-PSF (B92; Sigma), 1:2500; anti-Sin3a(K-20; SC), 1:1000; anti-Ncor (C-20; SC), 1:1000; anti-SMRT (H300; SC), 1:200;anti-RXR (sc-774; SC), 1:500; anti-HDAC1 (sc7872; SC), 1:300. Blots were incubated with SuperSignal West Dura Extended Duration Substrate (Pierce) and exposed to ECL films (Pierce). When applicable, films were digitized on a FLA5000 multi imaging system (Fuji), and the amount of co-immunoprecipitated proteins was determined by densitometry. Ratios were calculated as relative increase in amount of co-immunoprecipitated proteins in Pitx3^-/-embryos compared with C57BL6 embryos.

Chromatin immunoprecipitation was performed according to the protocol supplied by Nimblegen. For the in vitro ChIP-on-chip analysis,MN9D-Nurr1^{Tet On}13N cells were transfected with 0.5 μg Pitx3-pcDNA3.1(-). Nurr1 expression was induced by 3 μg/ml doxycyclin and cells were cultured for 48 hours. Two 10 cm plates were used for each antibody. For the in vivo ChIP-on-chip analysis, five ventral midbrains from C57Bl6-Jico E14.5 embryos were used for each ChIP. Chromatin was sonicated to an average length of 500 bp on a Soniprep 150 sonicator through 15 pulses of 10 seconds at one-third of maximum power. ChIP was performed with either Nurr1 antibodies, Pitx3 antibodies or pre-immune serum(Ctrl). ChIP-DNA was amplified using a Whole Genome Amplification kit (Sigma),according to the manufacturer's protocol and shipped for ChIP-on-chip analysis(Nimblegen) where Cy5-labelled Pitx3- and Nurr1-ChIP samples were hybridized to Cy3-labelled input chromatin on a mouse promoter array set (MM8). Microarray data was analysed using Signalmap software (v1.9). PCR-based validation was performed on 10 ng of amplified ChIP-DNA using the following primers: Vip, 5′-AGCGACTGAGTTGGAGATTC-3′ and 5′-GTAAGCCATGGCACTAGCAC-3′; Vmat2,5′-GTGTCTACTGTCTATCACAG-3′ and 5′-GTCAAAGTGTCCATGAAGCC-3′; control,5′-GACCCGTGTCACTGACCTAC-3′ and 5′-GCCTGCTAGGAGCAGCCTTG-3′. The amount of amplified DNA was determined by densitometry on a FLA5000 multi imaging system (Fuji) and relative enrichment by Pitx3- and Nurr1-ChIP over pre-immune serum (Ctrl)-ChIP was calculated. Statistical analysis was performed using Student's t-test, comparing the relative values of enrichment of promoters of Vip and Vmat2 to the relative values of a non-enriched region (control) in the Vmat2 promoter, as determined by ChIP-on-chip.

Pitx3^-/- and Nurr1^-/- embryos were obtained as described previously (Smits et al., 2003; Jacobs et al.,2007). Pitx3-deficient Aphakia (Pitx3^-/-) or C57Bl6-Jico wild-type mice were crossed with Pitx3^gfp/gfp mice to obtain Pitx3^gfp/+ and Pitx3^gfp/- embryos. Pitx3^gfp/- embryos contain both the classical Ak allele and an allele in which green fluorescent protein (GFP) is knocked-in the Pitx3 locus (Zhao et al.,2004; Maxwell et al.,2005) and are therefore Pitx3 deficient. Pitx3^gfp/+embryos are heterozygous for wild-type Pitx3 and GFP, and are described to have a normal development of the mdDA system (Maxwell et al.,2005).

Genotypes were determined by PCR analysis as described previously(Saucedo-Cardenas et al.,1998; Jacobs et al.,2007).

In situ hybridization was performed as described previously(Smits et al., 2003; Smidt et al., 2004). The following digoxigenin (DIG)-labelled probes were used: Aadc; bp 22-488 of the mouse coding sequence (cds)(Smits et al., 2003), Vmat2; bp 290-799 of mouse cds(Smits et al., 2003), Dat; bp 789-1153 of rat cds, Nurr1; 3′ region of rat Nurr1, En1; 5′ region of mouse transcript, D2R; bp 345-1263 of mouse cds.

Ventral midbrains of Pitx3^gfp/+ and Pitx3^gfp/- embryos at stage E13.5 were dissected in L15 medium (Gibco) and cultured in Neurobasal Medium (Gibco)supplemented with: 2% (v/v) B-27 supplement (Gibco), 18 mM HEPES-KOH (pH 7.5),0.5 mM L-glutamine, 26 μM beta-mercapto-ethanol and 100 units/ml penicillin/streptomycin. Midbrains were cultured and treated with either 0 mM,0.3 mM or 0.6 mM of sodium butyrate (Sigma) for 48 hours.

Cultures were dissociated using a Papain dissociation system (Worthington)and cells were sorted on a Cytopeia Influx Cell sorter. Sort gates were set on forward scatter versus side scatter (life cell gate), on forward scatter versus pulse width (elimination of clumps) and on forward scatter versus fluorescence channel 1 (528/38) filter; GFP fluorescence). Cells were sorted using a 100 μm nozzle at a pressure of 15 PSI with an average speed of 7000 cells/second.

RNA was isolated in Trizol (Invitrogen), according to manufacturer's protocol. Semi-quantitative RT-PCR was performed on equal amounts of RNA (0.1 ng) using a one-step RT-PCR kit (Qiagen), according to the supplied protocol. The following primers were used: Vmat2,5′-GCAGTCACACAAGGCTACCA-3′ and 5′-TGAATAGCCCCATCCAAGAG-3′; D2R,5′-GCCGAGTTACTGTCATGATC-3′ and 5′-ACGGTGCAGAGTTTCATGTC-3′; Th,5′-GCCACGTGGAATACACAAAG-3′ and 5′-GAGGCATGACGGATGTACTG-3′; Tbp,5′-GAGAATAAGAGAGCCACGGAC-3′ and 5′-TCACATCACAGCTCCCCAC-3′. The amount of amplified DNA was determined by densitometry as described above. Statistical analysis was performed by two-way unpaired Student's t-test, comparing the relative transcript levels of sodium butyrate-treated cultures with transcript levels of untreated cultures.

The quantified results represent the average values of experiments performed in triplicate and presented data indicate means with standard errors(s.e.m.). Statistical analysis was performed using Student's t-test(two-way unpaired). P≤0.05 is considered significant and are indicated by asterisks, P<0.01 is indicated with double asterisks.

Based on the discussed similarities between Nurr1- and Pitx3-null mice, we hypothesised that Pitx3 and Nurr1 are interconnected at the level of transcriptional regulation. Notably, functional interactions between homeobox transcription factors and nuclear receptors have been proposed previously(Tremblay et al., 1999; Shan et al., 2005). In the pituitary gland, Pitx1 interacts with the orphan nuclear receptor`steroidogenic factor 1' (SF1; Nr5a1) and multiple co-regulators to activate SF1-mediated transcription (Tremblay et al., 1999; Quirk et al.,2001). For mdDA neurons, the identification of proteins that physically interact with Pitx3 might provide clues to the relationship between Nurr1 and Pitx3 at the level of gene transcription.

Through affinity purification of Pitx3-His proteins followed by mass spectrometry, we identified the transcriptional co-repressors PSF (SFPQ;PTB-associated splicing factor) and Nono [non-POU-domain-containing, octamer binding protein/p54(nrb)] as proteins that specifically interact with Pitx3 in dopaminergic MN9D cells (Choi et al.,1991) (Fig. 1A). Interestingly, PSF and Nono are reported to form a heterodimeric complex involved in repression of nuclear receptor-mediated transcription(Mathur et al., 2001; Shav-Tal and Zipori, 2002; Sewer et al., 2002; Dong et al., 2007). PSF has been most extensively studied in relation to nuclear receptor repression and therefore we focussed on PSF. The interaction of Pitx3 with endogenous PSF in MN9D cells was confirmed by means of co-immunoprecipitation (Co-IP)(Fig. 1B). To investigate whether the interaction of Pitx3 with PSF also exists in vivo, we performed immunoprecipitation (IP) on E14.5 mdDA neurons. Indeed, endogenous PSF clearly co-immunoprecipitated with Pitx3 (Fig. 1C), demonstrating the interaction of Pitx3 and the nuclear receptor co-repressor PSF in developing mdDA neurons.

Most compelling, PSF has been shown to interact with the retinoic acid receptor RXR (Mathur et al.,2001), a well-described heterodimerization partner for Nurr1 in mdDA neurons (Perlmann and Jansson,1995; Wallen-Mackenzie et al.,2003). Therefore, we hypothesised that, like Pitx3, Nurr1 might also interact with PSF. Immunoprecipitation with Nurr1 antibodies was performed in MN9D cells, and endogenous PSF was indeed detected in Nurr1 immunoprecipitates (Fig. 1D). Moreover, PSF was also detected in Nurr1 and RXR-immunoprecipitates of E14.5 mdDA neurons, suggesting that besides Pitx3, the nuclear receptors Nurr1 and RXR also interact with PSF in vivo (Fig. 1E). These observations, together with the previously described involvement of PSF in repression of nuclear receptors, makes PSF an appealing candidate for bridging Pitx3 to Nurr1 in mdDA neurons.

The identification of PSF as interacting protein of both Pitx3 and Nurr1 provided novel leads to the molecular background of the functional interplay between Pitx3 and Nurr1. In their role as developmental regulators, both Pitx3 and Nurr1 are crucial for the induction of Th in mdDA neurons of the substantia nigra (Smidt et al.,2004; Maxwell et al.,2005; Jacobs et al.,2007). In addition, PSF has been shown to repress the TH-promoter in human CHP-212 cells(Zhong et al., 2006). Although the involvement of Pitx3 or Nurr1 has not been investigated in this context,other studies have independently shown that both Nurr1 and Pitx3 can bind and activate the Th promoter (Cazorla et al., 2000; Iwawaki et al.,2000; Lebel et al.,2001; Kim et al.,2003). The finding that both Pitx3 and Nurr1 interact with PSF could indicate that Nurr1 and Pitx3 are part of the same transcriptional complex on the Th promoter and possibly other target genes as part of a general mechanism for transcription regulation of mdDA-associated genes.

In order to investigate the occupancy of Pitx3 and Nurr1 on promoters of a large and unbiased set of genes, we performed ChIP-on-chip analysis. We chose two different approaches to obtain DA neurons. For the first approach, we expressed Pitx3 and Nurr1 in dopaminergic MN9D cells and for the second approach we used in vivo mdDA neurons derived from E14.5 ventral midbrain. Chromatin immunoprecipitation (ChIP) was performed with either Pitx3, Nurr1 or control (pre-immune serum) antibodies, and the IP-efficiency was determined by western blot analysis (Fig. 2A,B). Amplified ChIP-DNA fragments were labelled with Cy3 and hybridized to Cy5-labelled input chromatin on a MM8 mouse promoter array(Nimblegen). The ChIP-on-chip analysis in MN9D cells revealed that 1652 and 541 promoter regions were enriched by ChIP for Nurr1 or Pitx3, respectively(Fig. 2C). In the in vivo ChIP,a smaller number of promoter regions was enriched: 208 by Nurr1 and 140 by Pitx3 (Fig. 2D). This lower number may reflect the more stringent conditions of promoter occupancy in the in vivo situation compared with in a cell line. Strikingly, a substantial amount of promoter regions that were enriched by Nurr1, were also enriched by Pitx3 (see Tables S1 and S2 in the supplementary material). This was observed for ∼22% and 28% of the Nurr1-enriched promoter regions in MN9D cells and in vivo, respectively. Of special interest was the enrichment of promoter regions near genes that have previously been described to be regulated by Nurr1. The ChIP-on-chip analysis of MN9D cells revealed that the promoter of the Nurr1 target gene Vip (vasoactive intestinal peptide)(Luo et al., 2007) was highly enriched by Nurr1-ChIP (Fig. 2E). Remarkably, the same Vip promoter region was clearly also enriched by Pitx3-ChIP, which suggests that Pitx3 and Nurr1 bind to the same promoter region of the Nurr1 transcriptional target Vip. The enrichment of the Vip promoter was validated by PCR on ChIP-DNA,confirming a significant enrichment of the Vip promoter region by both Pitx3- and Nurr1-ChIP compared with Control-ChIP in MN9D cells[Fig. 2K,L; n=3; P<0.01 (Pitx3-ChIP)/P<0.05 (Nurr1-ChIP)].

The promoter of Vmat2, a well described target of Nurr1 in vivo,was clearly also enriched by ChIP for both Nurr1 and Pitx3(Fig. 2F). The enrichment was observed in both MN9D cells and E14.5 mdDA neurons, which is in line with the described regulatory effect of Nurr1 on Vmat2 expression in MN9D cells and in vivo (Hermanson et al.,2003; Smits et al.,2003). PCR-based validation confirmed the significant enrichment of the selected Vmat2 promoter region by ChIP for both Pitx3 and Nurr1 (Fig. 2K,L; n=3; P<0.01). Within the tiled regions of the promoters of Ahd2,Dat, Th and D2R, which are established as Nurr1-regulated genes in mdDA neurons (Zetterström et al.,1997; Le et al.,1999; Castillo et al.,1998; Wallén et al.,1999; Smits et al.,2003), we did not observe mutually enriched regions(Fig. 2G-J). However, because on average 5 kb of each promoter is represented on the array, we cannot exclude the possibility that Nurr1 and Pitx3 targeted sites are outside the tiled regions. Altogether, these data demonstrate the concerted occupancy of Pitx3 and Nurr1 on a set of promoters, including those of the Nurr1-regulated genes Vip and Vmat2.

Thus far, we have shown that Pitx3 and Nurr1 form a complex with a common co-repressor, and bind to regions within the promoters of Nurr1 target genes. These observations make it tempting to hypothesize that these physical associations have functional consequences in mdDA neurons. Two lines of evidence are in agreement with this hypothesis. First, combined transduction of Pitx3 and Nurr1 in ES cells leads to increased transcript levels of DA genes, such as Vmat2, Dat, Th and Aadc(Martinat et al., 2006). Second, the Pitx3 transcriptional target gene Ahd2 is not detected in the midbrain area of Nurr1-deficient embryos at E15.5(Wallén et al., 1999)or E14.5 (F.M.J.J. and M.P.S., unpublished). In order to elucidate the gene-regulatory function of Pitx3 in relation to Nurr1 in the development of mdDA neurons, we performed in situ hybridization on sagittal sections of E14.5 Pitx3^-/- embryos, and compared them with E14.5 Nurr1^-/- embryos (Fig. 3A). The normal expression of En1 and Nurr1 in Pitx3^-/- embryos and the conserved expression of En1 and truncated Nurr1 transcripts in Nurr1^-/- embryos is indicative of a normal presence of mdDA neurons in medial sagittal sections(Fig. 3B,C,H,I). As described previously, the embryonic expression of Dat and Vmat2 was completely lost in Nurr1^-/- mdDA neurons(Fig. 3J,K), whereas the expression of D2R and Aadc was strongly decreased(Fig. 3L,M). Interestingly, Dat was not detected in Pitx3^-/- mdDA neurons,whereas it was clearly expressed in Pitx3^+/+ littermate controls (Fig. 3D). Furthermore, the expression of Vmat2, which is completely absent in Nurr1^-/- mdDA neurons, was drastically reduced in Pitx3^-/- mdDA neurons(Fig. 3E). Notably, Vmat2 was abundantly expressed in the hindbrain of Nurr1- and Pitx3-deficient embryos, clearly demonstrating a mdDA neuron-specific deficiency of Vmat2 expression. Similarly, D2R was expressed at barely detectable levels in Nurr1^-/- mdDA neurons(Fig. 3L) and was strongly reduced in Pitx3^-/- mdDA neurons compared with littermate controls (Fig. 3F). Whereas Aadc expression was clearly decreased in Nurr1^-/-mdDA neurons (Fig. 3M), we observed no difference of Aadc expression between Pitx3^-/- and Pitx3^+/+ embryos in these medial sections (Fig. 3G). Altogether, besides the previously described effects on the expression of Th and Ahd2, loss of Pitx3 clearly affects the expression of Vmat2, Dat and D2R in mdDA neurons during development. Interestingly, next to Nurr1 as essential factor for the dopaminergic phenotype, Pitx3 is crucial for canonical expression of genes previously described as Nurr1 downstream targets, strengthening our hypothesis that Pitx3 and Nurr1 constitute a physical and functional complex in transcription regulation.

Our data demonstrate the concurrent involvement of Pitx3 and Nurr1 in the regulation of the same set of genes within mdDA neurons. However, the relative severity of the analyzed molecular phenotypes suggests that, whereas Nurr1 can be considered a master switch on DA-gene transcription, Pitx3 appears to regulate the transcriptional activity of Nurr1. Generally, the activity of nuclear receptors is highly regulated by co-regulating proteins. Upon activation of nuclear receptors, co-repressors are released and co-activators are recruited (Fowler and Alarid,2004; Wolf et al.,2008). Possibly, recruitment of Pitx3 affects the composition of the Nurr1 transcriptional complex in relation to the interaction with co-repressor proteins such as PSF. As we have demonstrated that the co-repressor PSF interacts with both Pitx3 and Nurr1, the interaction of Nurr1 with PSF might depend on the presence of Pitx3. Another plausible candidate in this respect is the co-repressor Sin3a, which is recruited by PSF to mediate nuclear receptor repression (Mathur et al., 2001). In order to test any alterations of the composition of the Nurr1 transcriptional complex upon Pitx3 ablation in vivo, we performed IP for Nurr1 from E14.5 wild-type and Pitx3^-/- mdDA neurons. Importantly, in accordance with what was observed by in situ hybridization,similar amounts of Nurr1 were immunoprecipitated from wild-type and Pitx3^-/- mdDA neurons(Fig. 4E). Besides PSF, the PSF-interacting co-repressor Sin3a also interacted with Nurr1 in vivo. However, the level of interaction of Nurr1 with either PSF or Sin3a was not affected by the presence of Pitx3 (Fig. 4A,B; n=3). These data indicate that the physical interaction of PSF and Sin3a with Nurr1 is unrelated to the potency of Nurr1 to activate its target genes. This is in line with other studies, reporting that activation of nuclear receptors by ligand binding does not result in dissociation of Sin3a or PSF (Mathur et al., 2001). Although no endogenous ligand has been identified for Nurr1, its ligand-binding domain interacts with the co-repressors Ncor(nuclear receptor co-repressor 1) and SMRT (silencing mediator of retinoic acid and thyroid hormone receptor) (Codina et al., 2004; Lammi et al.,2008). In general, Ncor and SMRT bind to unliganded nuclear receptors, and are believed to keep these complexes in a repressed state(Nishihara et al., 2004). Ligand binding to the nuclear receptor complex results in dissociation of Ncor and SMRT, allowing activation of gene transcription. As a possible regulatory mechanism of Nurr1-mediated transcription in mdDA neurons, the recruitment of Pitx3 to the Nurr1 transcriptional complex might mimic the effect of ligand activation of Nurr1, by inducing the dissociation of SMRT or Ncor. To test this, we examined whether the level of interaction between Nurr1 and the co-repressors Ncor and SMRT was altered upon ablation of Pitx3 in mdDA neurons. Although we did not detect an interaction between Ncor and Nurr1(Fig. 4C), the co-repressor SMRT was clearly co-immunoprecipitated with Nurr1. Strikingly, we observed a significant increase in SMRT interaction with Nurr1 in the absence of Pitx3(Fig. 4D). The SMRT interaction with Nurr1 was increased approximately threefold in Pitx3^-/- mdDA neurons compared with wild type, as determined by densitometry (Fig. 4F; n=4; P<0.01). This suggests that in the absence of Pitx3, the transcriptional activity of Nurr1 is kept in a repressed state by SMRT, and recruitment of Pitx3 leads to activation of Nurr1, through release of the SMRT-mediated repression.

SMRT exerts its repressive actions through recruitment of a set of class I and class II histone deacetylases (HDACs)(Huang et al., 2000; Guenther et al., 2001; Fischle et al., 2002). The de-acetylated state of histones near the transcription start site is responsible for the repression of gene transcription. The potential involvement of HDACs in the regulation of Nurr1 was further strengthened by the detection of endogenous HDAC1 in Nurr1 immunoprecipitates from E14.5 dissected ventral midbrains (see Fig. S1 in the supplementary material). Our data suggest that Nurr1 occupies the promoters of target genes in Pitx3^-/- embryos but is kept in a repressed state owing to the interaction with SMRT/HDAC complexes. Hypothetically, interference with HDAC activity in Pitx3^-/- embryos would result in re-activation of Nurr1-mediated transcription, bypassing the necessity for Pitx3. To test this hypothesis, we focussed on the expression of Vmat2 and D2R because of the reduced transcript level throughout the mdDA area in Pitx3^-/- embryos(Fig. 5B,C) and on Thbecause of the mdDA subset-specific loss of Th expression in Pitx3^-/- embryos (Fig. 5D). In order to isolate mdDA neurons for gene expression analysis by means of FACS sorting, we set up a breeding scheme to obtain Pitx3-deficient Pitx3^gfp/- and heterozygous Pitx3^gfp/+ embryos(Zhao et al., 2004; Maxwell et al., 2005). Importantly, the expression of GFP was highly similar in Pitx3^gfp/- and Pitx3^gfp/+ embryos at stage E14.5(Fig. 5E), showing the normal presence of mdDA neurons in Pitx3-deficient mdDA neurons during embryonic development. Ventral midbrains of Pitx3^gfp/+and Pitx3^gfp/- embryos were dissected at stage E13.5 and cultured for 2 additional days. To interfere with HDAC activity, we treated the ventral midbrains with different concentrations of the a-specific HDAC inhibitor sodium butyrate, after which GFP-positive mdDA neurons were selected by FACS sorting (Fig. 5F-H). Importantly, similar numbers of GFP-positive neurons were collected from all treated and untreated Pitx3^gfp/- and Pitx3^gfp/+ cultures. Equal amounts of RNA from GFP-positive mdDA neurons were subjected to semi-quantitative RT-PCR to determine the relative transcript levels of Vmat2, D2R and Th. Tbp (TATA-box binding protein) was taken along as control, which showed no difference in transcript levels in mdDA neurons of both genotypes and upon treatment with sodium butyrate (Fig. 5I,M; n=3). In heterozygous Pitx3^gfp/+ mdDA neurons, no significant increase in relative amounts of Vmat2, D2R and Thtranscripts was observed upon sodium butyrate treatment(Fig. 5I-L; n=3). In agreement with what was observed for Vmat2 expression by in situ hybridization, a significant lower level of Vmat2 transcripts was observed in untreated Pitx3^gfp/- mdDA neurons,at ∼7% compared with the level of Vmat2 in untreated Pitx3^gfp/+ mdDA neurons(Fig. 5I,J; n=3). Strikingly, treatment with 0.3 mM sodium butyrate resulted in a significant increase in Vmat2 transcripts, restoring levels to 56% (n=3; P<0.01). Vmat2 levels were even restored to 84% in cultures treated with 0.6 mM sodium butyrate (n=3; P<0.01). Similarly, D2R levels were restored from 53% in untreated Pitx3^gfp/- mdDA neurons to 98% in 0.6 mM-treated cultures (Fig. 5I,K; n=4; P<0.01) and levels of Th were increased from 60% in untreated Pitx3^gfp/- mdDA neurons to 94% in cultures treated with 0.6 mM sodium butyrate(Fig. 5I,L; n=3; P=0.05). These data clearly demonstrate that inhibition of HDAC-mediated repression in Pitx3^-/- embryos efficiently restores the transcriptional activation of a set of Nurr1 target genes,bypassing the necessity for Pitx3. These observations signify our proposed mechanism that in mdDA neurons, full activation of Nurr1 target genes depends on the release of the SMRT/HDAC-mediated repression of the Nurr1 transcriptional complex through the regulatory action of Pitx3.

The regulatory effect of Pitx3 on the association of Nurr1 with SMRT and the finding that Nurr1-mediated transcription of Vmat2, D2R and Th is repressed in a HDAC-dependent manner in the absence of Pitx3,enabled us to propose a model in which Pitx3 is a key regulator of Nurr1-mediated transcription (Fig. 6). Accordingly, the recruitment of Pitx3 to PSF within the Nurr1 transcriptional complex leads to (full) activation of Nurr1 target genes, by inducing the release of SMRT/HDAC-mediated repression from Nurr1. Ablation of Nurr1, acting as a master switch on gene transcription, results in complete loss of target gene expression. However, ablation of Pitx3 mainly affects Nurr1 transcriptional activity, which would still allow Nurr1-mediated transcription at a lower level (Vmat2/D2R), or in a certain mdDA subpopulation (Th). This model also explains why expression of the Nurr1 target genes Dat, Vmat2 and D2R is not detected until after Pitx3 is expressed at E11.5, although Nurr1 is already expressed at E10 (Simon et al.,2003; Wallen-Mackenzie et al.,2003; Smits and Smidt,2006; Alavian et al.,2008).

Interestingly, another member of the paired like homeobox transcription factors has been described to functionally and physically interact with nuclear receptors in the pituitary. Pitx1 enhances the transcriptional activity of the orphan nuclear receptor SF1(Tremblay et al., 1999; Quirk et al., 2001; Melamed et al., 2002), and Pitx1, together with the estrogen receptor, modify the repressive androgen receptor/SF1 complex on the LHβ promoter(Jorgensen and Nilson, 2001). Moreover, SF1 has also been shown to interact with PSF(Sewer et al., 2002). The apparent analogy between the regulation of an orphan nuclear receptor by Pitx1 and our data are indicative of a more general mechanism involving PSF, in which homeodomain transcription factors modulate nuclear receptor-mediated transcription. Most likely, more components are involved in the proposed mechanism. The release of the co-repressor SMRT from Nurr1 could be accompanied by the recruitment of activating proteins, although co-activators of Nurr1 have yet to be identified (Castro et al., 1999; Wang et al.,2003; Volakakis et al.,2006; Lammi et al.,2008). Notably, β-catenin has been reported to affect the transcriptional activity of both Pitx factors and SF1(Vadlamudi et al., 2005; Hossain and Saunders., 2003)which makes β-catenin an interesting potential component of the Nurr1/PSF/Pitx3 complex.

In zebrafish, PSF is involved in neuronal differentiation, and loss of PSF results in developmental defects in the mid- and hindbrain(Lowery et al., 2007). Although PSF is expressed in mdDA neurons (data not shown) and forms a complex with Pitx3 and Nurr1, the role of PSF in the Nurr1 transcriptional complex is unclear. The interaction of Nurr1 with PSF is unchanged in Pitx3^-/- embryos, which makes it unlikely that Pitx3 is involved in the recruitment or dissociation of PSF to the Nurr1 transcriptional complex. Importantly, besides PSF, we showed that Nurr1 also interacts with Sin3a, which is involved in transcriptional repression by recruiting SMRT and HDACs (Wolffe,1997). This may indicate that the effect of Pitx3 on dissociation of SMRT from the Nurr1 transcriptional complex also involves PSF and Sin3a.

The shared interaction of Pitx3 and Nurr1 with PSF suggests that Pitx3 and Nurr1 are part of the same transcriptional complex. The results from the ChIP-on-chip analysis further strengthen this hypothesis, as we identified a number of promoter regions that were enriched by ChIP for both Pitx3 and Nurr1. Most intriguingly, both Pitx3 and Nurr1 interact with the promoter of Vmat2, which is crucial for DA metabolism and is regulated by Nurr1 in vivo (Smits et al., 2003). These observations strongly suggest that Pitx3 and Nurr1 are recruited in concert with a number of promoters, including some that have been previously associated with Nurr1.

The shared physical interaction of Pitx3 and Nurr1 with the promoter of Vmat2 is translated to function in vivo, as expression analysis of E14.5 Pitx3^-/- embryos revealed an essential role for Pitx3 for the proper expression of a set of described Nurr1 target genes,including Vmat2. These observations shed new light on the complicated role of Pitx3 in mdDA neurons, as it has been puzzling for a long time that although Pitx3 is expressed in all mdDA neurons, only the subpopulation of the substantia nigra is dependent on Pitx3 for induction of Th and subsequential maintenance into adulthood. This discrepancy indeed holds true for the expression of Th, underlining the existence of mdDA subsets with differential reliance on Pitx3. However, based on analysis of the expression patterns of Vmat2, Dat and D2R, the entire mdDA population is affected by Pitx3 deficiency. Our observations clearly indicate that in addition to the previously described subset-specific role for Pitx3 in the substantia nigra, Pitx3 is crucial for the correct molecular phenotype of mdDA neurons in general. This finding could be an explanation for the aberrant electrophysiological properties of surviving Th-positive neurons in Pitx3-deficient mice, which has remained unexplained until now(Smits et al., 2005).

Interestingly, in vivo expression analysis revealed a striking difference between the relative severity of the Nurr1 target gene deficit between Nurr^-/- and Pitx3^-/- embryos, which is most evident for the expression of Vmat2. It appears that although Nurr1 is crucial for the expression of its target genes, Pitx3 is essential for the level of Nurr1-mediated transcription. The increased interaction of Nurr1 with the co-repressor SMRT in the absence of Pitx3 is a suitable molecular explanation for this observation. The regulatory effect of Pitx3 on the physical interaction of Nurr1 with the co-repressor SMRT strongly suggests that Pitx3 has the potency to induce the release of SMRT-mediated repression from the Nurr1 transcriptional complex. Indeed, interference with the repressive activity of HDACs, which are recruited to unliganded nuclear receptors by SMRT, efficiently restores the transcript levels of the Nurr1 target genes Vmat2, D2R and Th in cultured Pitx3^gfp/- midbrains. This clearly indicates that, in the absence of Pitx3, the Nurr1 transcriptional complex is kept in a repressed state by SMRT/HDAC, which signifies the regulatory role of Pitx3 on the expression of Nurr1 target genes as we proposed in our model.

Altogether, as part of a proposed general modulatory effect of Pitx factors on nuclear receptor activity, our data provide evidence for the existence of a transcriptional complex involving Pitx3 and Nurr1, positioning Pitx3 as an essential activator of the potency of Nurr1 to drive expression of genes essential for mdDA neurons. In conclusion, our data significantly contribute to a better understanding of mdDA neuron development, as we have shown that Nurr1 and Pitx3 act in concert to induce the mdDA characteristics of neurons,which has large consequences for stem-cell engineering as a future treatment for Parkinson's disease.