Retinoic acid counteracts developmental defects in the substantia nigra caused by Pitx3 deficiency

Parkinson's disease (PD) is a common progressive neurodegenerative movement disorder attributed to selective loss of pigmented dopaminergic (DA) neurons in the substantia nigra pars compacta (SNc). Interestingly, adjacent DA neurons located in the ventral tegmental area (VTA) are largely spared(Hirsch et al., 1988). Analysis of animal models for PD demonstrates a similar susceptibility to degeneration of SNc DA neurons, indicating a conserved phenomenon among species (Blum et al., 2001). Although several genes have now been associated with PD(Moore et al., 2005; Abou-Sleiman, 2006), the exact mechanism underlying the specific degeneration of DA neurons located in the SNc is not fully understood.

The homeodomain transcription factor Pitx3, exclusively expressed in meso-diencephalic DA (mdDA) neurons in the brain(Smidt et al., 1997), is essential for the proper development of the mdDA system. Analysis of Pitx3-deficient mice has revealed that the loss of Pitx3 expression leads to the selective loss of a neuronal subset located primarily in the SNc(Hwang et al., 2003; Nunes et al., 2003; van den Munckhof et al., 2003; Smidt et al., 2004a; Smidt et al., 2004b). The mechanism by which Pitx3 influences the development of this specific mdDA subpopulation is still unknown. Although Pitx3 deficiency results in early developmental defects of mdDA neurons, whereas PD is a progressive neurodegenerative disease with late onset, both affect the same mdDA neuronal population. Therefore, knowledge of molecular pathways controlled by Pitx3 might provide new insights into mdDA pathology, as in PD. However, until now,no target genes of Pitx3 have been identified, although some genes are associated with Pitx3 function (Smits et al., 2006). One of these genes encodes the enzyme aldehyde dehydrogenase family 1, subfamily A1 (Aldh1a1; also known as Raldh1 or Ahd2). In embryonic stem (ES) cells, transgenic expression of Pitx3 leads to an increased ratio of ES cells that are positive for both Ahd2 and tyrosine hydroxylase (Ahd2+/TH+) (Chung et al.,2005). Although this suggests a relationship between Pitx3 and Ahd2 expression, a relatively large portion of cells (approximately 70%) that were positive for both Pitx3 and TH still showed no induction of Ahd2.

In mice, Ahd2 is first detected in the mesencephalic flexure as early as embryonic day (E)9 (Haselbeck et al.,1999; Westerlund et al.,2005). There is, however, a marked change in Ahd2 distribution during different stages of development within this area. From E9 until E11.5,Ahd2 is found in a broad area ranging from the ventricular zone to the most ventral neuronal population within the mantle layer(Wallen et al., 1999). By contrast, from E12.5 onwards, Ahd2 distribution is largely confined to a selective subpopulation of mdDA neurons in a rostro-ventral portion of the mdDA area (McCaffery and Drager,1994; Wallen et al.,1999; Chung et al.,2005). These data suggest that Ahd2 has a dual role in the presumptive mdDA area throughout development, functioning both in the proliferative and migratory stages, and during the final differentiation and maintenance of a selective mdDA neuronal subpopulation. Ahd2 is involved in the production of retinoic acid (RA) from retinol, which is crucial for neuronal patterning and differentiation(Duester, 2000; McCaffery et al., 2003). Via binding to the RA receptor (RAR) and retinoid X receptor (RXR), RA has been shown to induce tissue-specific gene transcription leading to cellular differentiation (Chambon,1996). RA is detected in the midbrain area during early embryonic stages (Niederreither et al.,2002a), and in the postnatal and adult brain (McGaffery and Drager, 1994). Interestingly, Ahd2 is the only aldehyde dehydrogenase present in the mdDA area during the developmental and adult stage(Niederreither et al., 2002b),suggesting an important role for Ahd2-mediated RA production in the mdDA system.

Based on the restricted expression pattern of Ahd2 within the mdDA neuronal population, together with the implicated relationship between Pitx3 and Ahd2 in ES cells, we studied the relationship between Pitx3 and Ahd2 in more detail in mdDA neurons. Our results show that Pitx3 regulates Ahd2expression both in vivo and in vitro. Furthermore, Pitx3 interacts with a highly conserved region in the proximity of the transcriptional start site of the Ahd2 gene, suggesting that Pitx3 activates Ahd2 directly at the transcriptional level. Most intriguingly, we show that maternal supplementation of RA counteracts the developmental defects caused by Pitx3 deficiency in a specific mdDA neuronal subpopulation. In this study, we provide evidence for the existence of a developmental cascade in which Pitx3 is positioned centrally in RA-dependent final differentiation of mdDA neurons.

Experiments were carried out in Pitx3-deficient mice from two different breeding strategies. Adult C57Bl/6-Jico wild-type mice (Charles River Laboratories, France) and Aphakia (ak) mice, which have been described previously (Semina et al.,2000; Rieger et al.,2001; Smidt et al.,2004a) were used in homozygous breeding. In heterozygous breeding,male ak mice were crossed to female C57BL/6-Jola mice to generate heterozygous F1 hybrids. F1 hybrids were intercrossed to generate Pitx3^+/+, Pitx3^+/- and Pitx3^-/- progeny. Genotypes were determined by PCR analysis of tail DNA. Pregnant mice were decapitated or euthanized by CO₂ asphyxiation and embryos were collected at E13.5, E14.5 or E18.5 (the day on which the copulatory plug was detected was considered E0). Adult mice were euthanized by CO₂ asphyxiation and brains were collected. Mice were bred in our laboratory under standard conditions(21.0±1.0°C and 50% humidity) with ad libitum access to standard food (Hope Farms) and tap water, on a 12:12 hour light:dark cycle. All experimental procedures were approved by the ethical committee for animal experimentation of the Utrecht University, the Netherlands.

Embryos and adult brains were collected and immediately frozen on dry ice. Sections (16 μm) were cut and collected on SuperFrost Plus slides (Menzel Gläser). In situ hybridization (ISH) with digoxigenin (DIG)-labeled and[³³P]-labeled cRNA probes was performed as described previously(Smidt et al., 2004a; Smits et al., 2005).[³³P]-labeled sections were dehydrated, air-dried and exposed to BAS-MS 2340 imaging plates (Fuji) for 3-5 days or to Kodak Biomax MR films(Kodak) for 3 weeks. Autoradiographic BAS-MS 2340 imaging plates containing the hybridization signals were scanned using the FLA-5000 imaging system(Fuji), and quantitative analysis was performed using the AIDA Image Analyzer Software (Raytest). Gene expression was analyzed in both the left and right SNc (Pitx3^+/+: n=8; Pitx3^+/-: n=5). Per animal, two adjacent sections in the SNc were analyzed. To evaluate statistical significance, data were subjected to the Student's t-test (two tailed). The following probes were used: a 1142 bp fragment of the rat Th cDNA(Grima et al., 1985), an Aadc (also known as Ddc - Mouse Genome Informatics) fragment containing bp 22-488 of the mouse coding sequence(Smits et al., 2003), a fragment containing bp 290-799 of the coding region from the mouse Vmat2 (also known as Slc18a2 - Mouse Genome Informatics)gene (Smits et al., 2003), an alpha-synuclein (a-synuclein; Snca) fragment containing bp 20-420 of the coding region from the mouse cDNA(Smidt et al., 2004a), and an Ahd2 fragment containing bp 568-1392 of the mouse coding sequence.

Embryos and adult brains were collected, incubated overnight in 4%paraformaldehyde (PFA) at 4°C, and embedded in paraffin. Sections (7μm) were cut on a microtome, collected on SuperFrost Plus slides (Menzel Gläser), de-paraffinated through xylene, rehydrated through an ethanol series and incubated in 0.3% H₂O₂ in Tris-buffered saline (TBS) for 30 minutes. Next, sections were boiled in citrate buffer (pH 6) for antigen retrieval, blocked with 4% fetal calf serum in TBS for 30 minutes and incubated overnight with rabbit anti-TH antibody (Pel-Freez,Arkansas, USA; 1:1000) in TBS/0.5% Triton at 4°C. The next day, sections were incubated for 1 hour with biotinylated goat anti-rabbit immunoglobulin in TBS (1:1000), followed by incubation with avidin-biotin-peroxidase reagents(ABC elite kit, Vector Laboratories, 1:1000) for 1 hour in TBS. The slides were stained with DAB (3,3′-diamino-benzidine) for a maximum of 10 minutes, until the background was lightly stained. Slides were dehydrated with ethanol and mounted using Entellan. Quantification of neurons was performed using a microscope (Zeiss Axiovert 405M) attached to a camera system(MicroPublisher 5.0 RTV) and imaging software (Openlab 5.0.1, Improvision). The number of TH-immunoreactive (TH-IR) neurons was counted unilaterally in anatomically matched adjacent coronal sections. For each E14.5 embryo, four sections were counted in the caudal and rostral domain of the mdDA system. For each E18.5 embryo, sections containing the SNc were analyzed (27-40 sections)and the average number of TH-IR neurons per section was calculated. Only neurons in which cell nuclei could be recognized were counted. Quantification was performed by an independent observer in a blind design. To evaluate statistical significance, data were subjected to the Student's t-test(two tailed).

For double-immunofluorescence staining, sections were incubated overnight with rabbit anti-Ahd2 (Abcam, diluted 1:100) in combination with sheep anti-TH(Chemicon, diluted 1:500) in PBS/0.5% Triton at 4°C. The next day,sections were incubated with fluorophore-conjugated secondary antibodies in PBS (Alexa-Fluor-488-conjugated goat anti-rabbit and Alexa-Fluor-555-conjugated donkey anti-sheep, diluted 1:400; Molecular Probes)for 1 hour at room temperature and embedded with 90% glycerol.

TH-IR sections were counterstained for 5 minutes in 0.5% cresyl violet and briefly rinsed in an acetate buffer (pH 4). The sections were then differentiated in 96% ethanol for 60 seconds, dehydrated in 100% ethanol,cleared in xylene and mounted with Entellan. Neuronal number per section in the rostral mdDA area was determined in cresyl violet-stained sections,delineated according to TH-IR area. Quantification (n=3) was performed by an independent observer in a blind design. To evaluate statistical significance, data were subjected to the Student's t-test(two tailed).

Adult brains were collected, incubated overnight in 4% PFA at 4°C, and embedded in paraffin. Sections (7 μm) were cut and collected on SuperFrost Plus slides (Menzel Gläser). ISH on paraffin wax sections was performed as described for frozen sections with the following modifications: sections were first deparaffinated through xylene, rehydrated through an ethanol series, boiled in citrate buffer (pH 6), and incubated in 0.2 M HCl for 15 minutes. Sections were further treated as described above for DIG-ISH on frozen sections, except that, after termination of the alkaline phosphatase reaction, sections were immunostained for TH with the avidin-biotin-peroxidase complex (ABC) method, as described above.

RNA from E18 mouse whole brain or MN9D cells was isolated using Trizol(Invitrogen) according the manufacturer's guidelines. A sample of Pitx3^+/GFP FACS-sorted mdDA cells from E16 embryos was isolated via the guanidine thiocyanate method.

Full-length Pitx3 cDNA was amplified from cDNA originating from E18 whole-brain RNA, with the following primers: forward,5′-CCCTGCCTGCGCTCCAGAAC-3′; reversed:5′-CCCTGTTCCTGGCCTTAGTC-3′. Pitx3 was ligated in pGEM-T easy vector (Promega) for sequence analysis, and subsequently cloned into pcDNA3.1(-) vector (Invitrogen) using the ApaI and BstxI restriction sites.

To distinguish between the Ahd2 and Aldh1a7 genes,primers with 100% homology to both genes were designed: forward,5′-GACTGATGAGATGCGCATTG-3′; reversed 5′-GTCTTGAGCTCAGTGTATTC-3′. For in vivo determination, RNA originating from E16 ventral midbrain Pitx3^+/GFP cells was subjected to OneStep reverse transcriptase (RT)-PCR (Qiagen). For in vitro determination, RNA from MN9D cells transfected with Pitx3-pcDNA3.1 was used. The obtained PCR fragments were cloned into pGEM-T easy vectors and eight different clones of each cloning were sequence-analyzed (Baseclear, The Netherlands).

MN9D (MN9D-13N) cells were cultured in Dulbecco's Modified Eagle Medium(DMEM) supplemented with 10% (v/v) hiFCS, 100 units/ml penicillin, 100 units/ml streptomycin and 2 mM L-glutamine in a humidified atmosphere with 5%CO₂ at 37°C. Cells were grown on 10-cm-diameter dishes coated with poly L-lysine. At 2 hours prior to transfection, culture medium was replaced with medium without antibiotics, and transfection was performed using Lipofectamine 2000 (Invitrogen), according to the manufacturer's guidelines. Cells were transfected with 5 μg of Pitx3-pcDNA3.1 or an equal molar amount of empty pcDNA3.1 vector together with carrier DNA and 1 μg of EGFP-N1 vector (Clontech). At 6 hours after transfection, cells were divided and allowed to grow for an additional 30 hours. Transfection efficiency was determined based on GFP fluorescence using a fluorescent microscope, and plates were matched based on their relative transfection efficiency. For MN9D cells, a typical transfection efficiency with Lipofectamine is approximately 80-90%. Finally, cells were harvested for RNA isolation or used for a chromatin immunoprecipitation assay.

Relative gene expression levels were determined using a OneStep RT-PCR kit(Qiagen). We used 50 ng total RNA per reaction, and for each transcript the linear phase was determined and reactions were stopped during that phase for analysis (for primer sequences, see Table 1). Samples were loaded on 2% agarose gels and, after gel electroforesis, gels were scanned using a FLA-5000 imaging system (Fuji) and relative amounts of DNA were measured with a densitometer. Each analysis was replicated at least three times with independent RNA samples to confirm the obtained results. To determine whether RA regulates the expression of TH, we cultured MN9D cells in the presence of 1 μM all-trans RA for 36 hours as described previously (Castro et al.,2001).

MN9D cells were transfected with Pitx3-pcDNA3.1 or empty pcDNA3.1 expression vector at 90% confluence. At 6 hours after transfection, cells were divided and allowed to grow for an additional 30 hours. Formaldehyde was added directly to the medium to a final concentration of 1%, and cells were incubated at room temperature for 10 minutes. Glycine was added to a final concentration of 0.125 M, and cells were incubated at room temperature for an additional 5 minutes. Cells were washed twice with PBS, harvested and pelleted by centrifugation. ChIP was performed as described previously(Matthews et al., 2005), with some adjustments. Cells were resuspended in lysis buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Na-deoxycholate]containing protease inhibitors (Complete Protease Inhibitor Cocktail Tablets,Roche) and were incubated on ice for 20 minutes. Subsequently, the lysate was put through a 27/3-4 G syringe three times. Chromatin was sonicated using a Soniprep 150 sonicator to obtain DNA of an average length of 400-500 bp. Chromatin was pre-cleared with 60 μl of a 50% slurry of Protein A agarose/Salmon Sperm DNA (Upstate) for 1.5 hours at 4°C. Pre-cleared chromatin was incubated overnight with approximately 1 μg of Pitx3 antibody(Smidt et al., 2000), or with 1 μg of rabbit whole serum immunoglobulins (Sigma) as a negative control. Antibody-DNA complexes were captured by adding 20 μl of a 50% slurry of Protein A agarose/Salmon Sperm DNA for 1 hour at 4°C. Immunoprecipitated complexes were washed for 10 minutes with the following wash buffers: twice with 1.5 ml of buffer 1 [20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1%Triton X-100, 0.1% sodium dodecyl sulfate (SDS)], once with 1.5 ml of buffer 2[20 mM Tris-HCl (pH 8.0), 500 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS],once with 1.5 ml of LiCl buffer [20 mM Tris-HCl (pH 8.0), 250 mM LiCl, 1 mM EDTA, 1% NP-40, 1% Na-deoxycholate], and twice with 1.5 ml of TE [10 mM Tris-HCl (pH 8.0), 1 mM EDTA]. For elution of the immunocomplexes, 150 μl of elution buffer (TE, 1% SDS) was added and beads were shaken on a vortex for 30 minutes at 4°C. This was performed twice and samples were pooled. Crosslinks were reversed overnight at 67°C and complexes were precipitated and dissolved in TE. Proteinase K was added and samples were incubated for 2 hours at 45°C. Samples were extracted once with phenol/chloroform (1:1)and twice with chloroform. DNA was precipitated and dissolved in 100 μl of PCR grade MilliQ. For each PCR reaction, 2 μl of this DNA sample was used(for primer sequences, see Table 1).

Following timed matings of Pitx3^+/- parents, pregnant mice were supplemented with all-trans-RA (Sigma), which will be referred to as RA in this article, twice daily from E10.75 to E13.75. RA was dissolved in corn-oil and mixed with powdered food to a final concentration of 0.25 mg/g food, which was supplied ad libitum as previously described(Niederreither et al., 2002c; Mic et al., 2003). RA was dissolved freshly each time and supplemented food was supplied at 12-hour intervals. The control Pitx3^+/- animals were treated according to the same protocol, but were supplemented with corn oil without RA. Embryos were isolated at E14.5 and E18.5, and weight-matched Pitx3^+/+ and Pitx3^-/- littermates were analyzed by immunohistochemistry.

The Ahd2 gene displays a high degree of sequence homology to Aldh1a7, which is also known as Ahd2-like or aldh-pb (86% on mRNA and 96% on protein level). In contrast to Ahd2,Aldh1a7 cannot catalyze the conversion of all-trans-retinal to RA(Hsu et al., 1999; Montplaisir et al., 2002). Most previous analyses did not clearly distinguish between these two transcripts. To determine which of these genes is actually expressed in mdDA neurons, reverse transcriptase (RT)-PCR was performed on RNA of FACS-sorted Pitx3^+/GFP mdDA neurons from E16 embryos(Zhao et al., 2004). Sequence analysis of cloned PCR products identified all clones (100%) as Ahd2,indicating that Ahd2 is the predominantly expressed gene in mdDA neurons (data not shown).

Previous work has indicated that Ahd2 expression is relatively enriched in SNc mdDA neurons, whereas its expression in the VTA is relatively scarce(McCaffery and Drager, 1994; Chung et al., 2005). To study the expression pattern of Ahd2 in more detail, the distribution of Ahd2 was examined in adult mdDA neurons of wild-type mice(Fig. 1). High levels of Ahd2 mRNA were observed in cells located in the SNc, as revealed by the expression pattern of Th on adjacent sections(Fig. 1Aa-Af). In addition, a large number of Ahd2+ cells was observed in the VTA. These cells were predominantly located in the ventral and lateral part of the VTA, with some scattered Ahd2+ cells located more dorsally(Fig. 1Ae-Al). Although Ahd2 is expressed in a large number of cells in the SNc and VTA, the expression patterns of Ahd2 and Th did not overlap completely. In particular, Th+ cells in the medial part of the SNc and dorso-medial part of the VTA appeared to lack Ahd2 expression. Ahd2 in situ hybridization (ISH) combined with an immunohistochemical staining for TH clearly confirmed that not all TH+ neurons in the adult SNc and VTA co-expressed Ahd2 (Fig. 1B). In the lateral tip of the SNc, TH+ cells that did not express Ahd2 were intermingled with TH+ cells that did express Ahd2(Fig. 1Bb). In the more medial part of the SNc, TH+ cells that lacked Ahd2 expression were predominantly located dorsally (Fig. 1Bc). In addition, in the VTA, TH+ cells that lacked Ahd2 expression were mainly observed in the dorsal VTA(Fig. 1Bd-Bf). Finally, antibody double-labeling experiments assured the subpopulation-specific distribution of Ahd2 protein in the SNc (Fig. 1Ca-Cc,Cg-Ci) and VTA(Fig. 1Cd-Cf,Cj-Cl). Not all TH+ cells in the SNc and VTA expressed Ahd2 protein, but those cells that did express Ahd2 protein were all TH+, indicating that Ahd2 and TH are co-expressed in a specific mdDA subpopulation. Taken together, based on both transcript and protein distribution, Ahd2 marks a specific mdDA subpopulation in the adult brain.

It has been reported previously that Ahd2 expression is a marker for proliferating and differentiating mesencephalic DA cells that becomes progressively more restricted during late differentiation and the postnatal stages (McCaffery and Drager,1994; Wallen et al.,1999). To study the developmental expression of Ahd2during late differentiation in more detail, the distribution of Ahd2mRNA was examined in E13.5 wild-type embryos(Fig. 2). Ahd2expression was confined to neurons located in the more lateral parts of the mdDA area, as identified by the expression of Th mRNA on adjacent sections (Fig. 2). Aadc-positive young migrating neurons, which are located more dorsally(Smidt et al., 2004a), and proliferating cells in the ventricular zone did not express Ahd2 at this time point. These data indicate that, at E13.5, Ahd2 expression is restricted to a specific subpopulation of mdDA neurons located in the lateral part of the mdDA area. By contrast, the medial part of the mdDA area was devoid of Ahd2 expression. Clearly, Ahd2 shows mdDA subpopulation-specific expression, indicating that mdDA subpopulations are already specified during development.

The expression pattern of Ahd2 described above, appears to identify the neuronal subset that is lost in Pitx3-deficient animals. To study that possibility in detail, we analyzed the expression of Ahd2 in adult brain and during the late differentiation phase in E13.5 mdDA neurons of Pitx3-deficient mice (Fig. 3). The expression pattern of Th revealed the previously reported malformation of the adult mdDA system in Pitx3-deficient mice (Smidt et al., 2004b),characterized by neuronal loss predominantly in the SNc(Fig. 3A). As expected, Ahd2 expression was completely absent in the rostral part of the mdDA area of adult Pitx3-deficient mice, which includes the SNc and lateral part of the VTA, as identified by the expression of Th mRNA on adjacent sections (Fig. 3Aa-Ad). By contrast, in the caudal part of the mdDA area, some Ahd2+ neurons could still be observed in the VTA(Fig. 3Ae,Af). Antibody double-labeling experiments demonstrated that the Ahd2+ neurons co-expressed TH protein,indicating that these cells are DA neurons (see Fig. S1 in the supplementary material). At E13.5, Ahd2 expression was completely lost in the lateral part of the Pitx3-deficient mdDA area, which is normally positive for Ahd2. By contrast, Th expression in adjacent sections was still observed, indicating that there were still mdDA neurons present in this area (Fig. 3B). Taken together, these data clearly demonstrate that Ahd2 expression is highly affected in Pitx3-deficient mice, already early in development.

The apparent loss of Ahd2 expression in Pitx3-deficient mdDA neurons during development suggests that Pitx3 is involved in Ahd2 gene activation. To study this in more detail, we analyzed the Ahd2 expression level in Pitx3^+/- mice, which display reduced Pitx3 expression(Rieger et al., 2001). We performed quantitative radio-active ISH studies on coronal adult mdDA sections of Pitx3^+/- mice containing the SNc(Fig. 4). Ahd2 mRNA levels in the SNc of Pitx3^+/- mice(Fig. 4, black bar; n=5) were significantly decreased compared with Pitx3^+/+ mice (Fig. 4, white bar; n=8; P<0.05). Expression levels of other mdDA markers, including Th, Vmat2 and alpha-synuclein were not altered in Pitx3^+/- mice. The fact that Ahd2expression was specifically decreased in Pitx3^+/- mice strongly indicates a dose-dependent effect of Pitx3 on the expression level of Ahd2.

To investigate further whether Pitx3 regulates Ahd2 gene transcription, we analyzed whether Ahd2 expression is influenced by Pitx3 in the DA cell line MN9D. This cell line expresses a multitude of DA markers, and Nurr1 (also known as Nr4a2 - Mouse Genome Informatics)overexpression results in increased levels of DA-related genes, such as Th, Vmat2 and Aadc(Hermanson et al., 2003). In addition, Nurr1 induces morphological differentiation and cell cycle arrest(Castro et al., 2001). Therefore, we argued that this cell line is an appropriate cell model in which to analyze Pitx3 function. Under normal conditions, Pitx3 and Ahd2 were both expressed endogenously in MN9D cells, albeit at a very low level (Fig. 5A). However Pitx3 overexpression led to a drastic increase in endogenous Ahd2transcript levels, whereas the control gene, TATA box binding protein(Tbp), showed no alteration in expression levels(Fig. 5A,B). To confirm that the upregulated gene was Ahd2 and not Aldh1a7, as was the case in vivo (see above), we performed sequence analysis on cloned PCR fragments, revealing that the majority of transcripts (88%) were encoded by the Ahd2 gene (data not shown). In addition, Pitx3 overexpression showed no effect on the transcript levels of other DA markers, including Th, Aadc, Vmat2 and alpha-synuclein(Fig. 5A,B). In total, three independent RNA samples were analyzed, which all gave similar results. These data demonstrate that endogenous Ahd2 expression largely relies on Pitx3 function, whereas Pitx3 has no effect on the expression of genes involved in neurotransmitter phenotype in these cells.

The drastic increase in Ahd2 transcript levels by Pitx3 overexpression may be mediated via direct transcriptional regulation. To investigate whether Pitx3 can bind to enhancer elements within the Ahd2 promoter, we performed a ChIP assay using specific Pitx3 antibodies. First, we compared rat and mouse Ahd2 promoter regions ranging from 10 kb upstream to 10 kb downstream of the transcriptional start site (TSS). Because the exact position of the TSS of the mouse Ahd2gene is not specified, we compared several expressed sequence tag (EST)sequences, picked the sequence that was found at the 5′ end of many of the transcripts and regarded that sequence as the +1 position. We identified multiple conserved regions containing consensus and non-consensus putative Pitx3-binding sites, and designed primers to amplify those regions that contained fully conserved putative Pitx3-binding sites in both rat and mice. For the analysis, we selected the following mouse genomic regions either upstream (-) or downstream (+) of the TSS: region 1 (-8455 to -8283 bp),region 2 (-4374 to -4252 bp), region 3 (-2981 to -2838 bp), region 4 (+1829 to+1989 bp), region 5 (+2080 to +2181 bp), region 6 (+4204 to +4330 bp), region 7 (+6478 to +6615 bp), region 8 (+6785 to +6941 bp) and region 9 (+7640 to+7765). Next, MN9D cells were transfected with Pitx3 expression vector to be analyzed in a ChIP-assay. Chromatin was subjected to immunoprecipitation with either αPitx3 antibodies or non-specific rabbit immunoglobulins (IgG). The immunoprecipitated DNA fragments were subjected to PCR, including IgG immunoprecipitated DNA as a negative control and input chromatin as a positive control. Region 2(Fig. 6A) contains a putative Pitx3-binding site that has been shown to be bound by Pitx3 in an electro mobility shift assay (EMSA) (Chung et al.,2005). In our assay, however, we observed no amplification of this region, indicating that, under the conditions we used, Pitx3 is not bound to that region. This difference can be explained by the use of different approaches to determine functional interactions of a transcription factor to a presumed binding site. It should be noted that, in contrast to a ChIP assay,the chromatin structure in living cells, which can strongly reduce accessibility for transcription factors, is disregarded in an EMSA approach. Interestingly, we successfully amplified a highly conserved genomic region(region 5) specifically from αPitx3 immunoprecipitated DNA(Fig. 6A,B), whereas no amplification of any of the other selected conserved regions was observed(data not shown). The genomic region 4, which lies further upstream than region 5 (Fig. 6A) and shows moderate sequence conservation between rat and mouse, but contains the consensus Pitx3-binding site TAATCC(Wilson et al., 1996), was not amplified (Fig. 6B). This implies selective binding of Pitx3 to the highly conserved region 5. Confirmation of the observed results in a second independent ChIP assay assured selective binding of Pitx3 to region 5. Furthermore, closer analysis of this highly conserved region revealed that, although sequence conservation between mouse/rat and human is generally poor in introns of Ahd2(approximately 60%), the homology in this specific intronic region is 91% over a stretch of 60 bp (Fig. 6C). In addition, a non-consensus putative Pitx3-binding site, AAATCT(Wilson et al., 1996; Yuan et al., 1999), which is fully conserved between mouse, rat and human, is found in this region. Altogether, these data indicate that Pitx3 interacts with a highly conserved region 2.1 kb downstream of the TSS of the Ahd2 gene, which may account for the observed regulation of Ahd2 gene transcription in MN9D cells.

RA generated by aldehyde dehydrogenases is crucial for morphogenesis and cellular differentiation in early eye development. Developmental defects of the eye caused by Raldh2 deficiency can be effectively rescued by maternal supplementation of synthetic RA during the period when endogenous RA is crucial for proper optic cup formation(Mic et al., 2002; Mic et al., 2004; Molotkov et al., 2006). Ahd2 is a potent generator of RA and is present in the mdDA area throughout embryonic development (Hsu et al.,1999; Haselbeck et al.,1999; Wallen et al.,1999). This suggests a possible role for Ahd2 at this stage of mdDA neuronal development in the synthesis of RA. With Ahd2 under the transcriptional control of Pitx3, it is appealing to hypothesize that the developmental defect in Pitx3-deficient mice is due to a loss of RA synthesis. If RA is crucial for proper development of Ahd2+ neurons in the mdDA system, we could bypass the necessity of Pitx3-mediated Ahd2 expression in these neurons by maternal dietary RA administration. Because Ahd2 is expressed in a subset of mdDA neurons, we expected a compensating effect of RA exclusively in that subpopulation. To identify the Ahd2+ mdDA subpopulation at E14.5, when the mdDA phenotype is clearly visible in Pitx3-deficient mice (Smidt et al., 2004a; Smidt et al.,2004b; Maxwell et al.,2005), we analyzed Ahd2 expression in wild-type brain at this stage. Ahd2 expression was not detected in caudal sections of the mdDA area,where, ultimately, the VTA is formed (Fig. 7Ba-Bc). In rostral sections, Ahd2 fully co-localized with TH in a specific subset of TH-immunoreactive (TH-IR) neurons(Fig. 7Bd-Bf). Ahd2 was predominantly present in mdDA neurons in lateral and ventral positions. The majority of TH-IR neurons in dorso-medial positions did not co-express Ahd2. In addition, TH-IR neurons in a distinct population in most lateral positions were largely negative for Ahd2.

To determine whether RA synthesis by Ahd2, mediated via Pitx3, has a role in the development of these Ahd2+ neurons, we aimed to compensate for Pitx3 deficiency by maternal dietary RA administration. Pregnant Pitx3^+/- mice were supplemented with RA from E10.75 to E13.75, and Pitx3^+/+ and Pitx3^-/-littermates were analyzed at E14.5. Detailed analysis showed that, in caudal regions of the mdDA area, no effect of RA treatment is observed in either the Pitx3^+/+ or Pitx3^-/- embryos(Fig. 7Ca-Cd; Fig. 7D, left panel). In Pitx3^+/+ embryos, RA treatment had no effect on the occurrence of mdDA neurons in the rostral regions(Fig. 7Cg,Ch; Fig. 7D, right panel). As previously reported (Smidt et al.,2004a), untreated Pitx3^-/- embryos displayed a significant loss (61%; n=3, P<0.01) of TH-IR neurons in the ventro-lateral regions (Fig. 7Ce,Cg; Fig. 7D, right panel). Strikingly, in rostral sections of Pitx3^-/- embryos, RA treatment led to a significant increase (70%; n=3, P<0.01) in number of TH-IR neurons(Fig. 7Ce,Cf; Fig. 7D, right panel). As a result, the loss in number of TH-IR neurons caused by Pitx3deficiency was drastically reduced after RA treatment. This demonstrates that RA treatment has an exclusive effect in Pitx3^-/- embryos,affecting specifically the rostral mdDA subpopulation, which is most affected by Pitx3 deficiency.

In order to understand the nature of the observed rescue effect, we performed Nissl staining to quantify the average amount of neurons per section in the rostral E14.5 mdDA area (n=3; Pitx3^-/-,untreated: 394±18; Pitx3^-/-, RA-treated:391±36; Pitx3^+/+, untreated: 404±52; Pitx3^+/+, RA-treated: 380±45). No statistical difference in neuronal number was found between all groups, which indicates that, at E14.5, no neurons are lost in either RA-treated or untreated Pitx3^-/- embryos, compared with Pitx3^+/+ embryos. Furthermore, RA is unable to induce TH expression in MN9D cells (data not shown), which suggests that the observed increase in TH-IR neurons is not due to a general induction of TH expression.

To determine whether the effect of temporary RA treatment is maintained at later developmental stages, we treated pregnant Pitx3^+/-mice with RA from E10.75 until E13.75 and analyzed Pitx3^+/+ and Pitx3^-/- littermate embryos at E18.5 (Fig. 8). RA-treated Pitx3^-/- embryos showed a significant increase in number of TH-IR neurons (59%; n=3, P=0.01) within the SNc. These data indicate that temporary RA treatment has a sustained effect on the mdDA neuronal population. In agreement with the E14.5 data, the effect was highly restricted to Pitx3^-/- embryos(Fig. 8A,B).

Pitx3^-/- mice are characterized by impaired innervation of the dorsal striatum. To determine whether the increased number of TH-IR neurons in RA-treated E18.5 Pitx3^-/- embryos was accompanied by increased innervation, we analyzed TH-immunoreactivity in dorsal regions of the striatum at multiple levels(Fig. 8C,D). Interestingly,increased TH-immunoreactivity was observed in dorsal regions of the striatum in RA treated Pitx3^-/- embryos, whereas this area was largely devoid of TH-immunoreactivity in untreated Pitx3^-/- embryos. These observations indicate that the TH-IR neurons that are maintained in RA-treated Pitx3^-/-embryos are those neurons that normally innervate striatal regions located more dorsally. Altogether, these data strongly suggest that RA-dependent signaling is essential for the proper development of a specific subset of mdDA neurons initially affected in Pitx3-deficient mice.

During early developmental stages, the mdDA neuronal population is not homogeneous, but consists of multiple neuronal subpopulations with different temporal and topographical expression patterns(Smits et al., 2006). An early determinant of subset specification is Pitx3 and, based on its temporal expression pattern, two distinct mdDA neuronal populations exist during early development. A ventro-lateral population, which expresses Pitx3 before TH, can be identified, as well as a dorso-medial population, which expresses TH before Pitx3 (Maxwell et al., 2005). The existence of mdDA subpopulations is highlighted by the expression pattern of Ahd2. In the adult brain, we observed Ahd2 expression in a ventrally located subset of TH-IR neurons in the SNc and VTA. During embryonic development, this mdDA subpopulation-specific expression was observed as early as E13.5, showing Ahd2 expression exclusively in the lateral parts of the mdDA area. The restricted expression of Ahd2 is also clearly visible in the major projection area of the mdDA system, where Ahd2 protein is detected in dorso-lateral parts of the adult striatum (McGaffery and Drager, 1994). Interestingly, innervation to these specific regions of the striatum is lost in Pitx3-deficient mice, caused by a selective loss of mdDA neurons predominantly in the SNc. In order to investigate the fate of Ahd2-expressing mdDA neurons in Pitx3-deficient mice, we determined the distribution of Ahd2 transcripts in both the adult and the embryonic brain. Although we observed a small population of mdDA neurons in the ventro-medial part of the VTA that still expressed Ahd2, most of the expression of Ahd2 was completely lost. This indicates that the majority of Ahd2+ mdDA neurons,except for a small distinct population, are fully dependent on Pitx3 expression for proper development. In E13.5 Pitx3-deficient embryos,no expression of Ahd2 was observed in the lateral region of the mdDA area, where the gene is normally expressed. However, Th is still expressed in that area, which indicates that a considerable amount of mdDA neurons is still present. Considering the amount of TH+/Ahd2+ cells in wild-type E13.5 embryos, and the amount of remaining TH+ cells in the Pitx3-deficient brain, this suggests that loss of Ahd2 expression is due to the ablation of Pitx3 itself rather than to neuronal loss.

The fact that Pitx3 expression is reduced in Pitx3^+/- animals(Rieger et al., 2001) provided a tool by which to analyze the in vivo consequence of Pitx3 gene dosage. Quantitative analysis of Ahd2 expression in the SNc of Pitx3^+/- mice revealed a significant reduction in the expression level of Ahd2. Expression of other DA marker genes was unaffected in the SNc of Pitx3^+/- animals, which points to a gene-dosage effect of Pitx3 on Ahd2 expression specifically,unrelated to neuronal loss. In perfect agreement with these data, we observed a drastic increase in endogenous Ahd2 transcript levels when Pitx3 was overexpressed in the DA cell line MN9D. This effect was specific for Ahd2, because the levels of Th, Aadc, Vmat2, Tbp and alpha-synuclein were unaffected. Further evidence for the Pitx3-mediated regulation of Ahd2 transcription was provided via a ChIP assay, which was performed to analyze the binding of Pitx3 to a selection of highly conserved regions in the proximity of the Ahd2 TSS. We showed that Pitx3 immunoprecipitation specifically selected for a genomic region in intron 1 of the Ahd2 gene. This region displays a surprisingly high sequence conservation in mouse, rat and human, in contrast to the surrounding intronic sequence of the Ahd2 gene. A non-consensus putative Pitx3-binding site, AAATCT, is contained within this region, which was also found to be fully conserved.

Because different lines of evidence suggest that Ahd2 is under the transcriptional control of Pitx3, we hypothesized that Ahd2, mediated through Pitx3, might have a role in the development of mdDA neurons. Ahd2 is a potent generator of RA (Hsu et al.,1999; Fan et al.,2003), which is essential for the proper development of many structures in the embryo (Duester et al.,2003), and is involved in neuronal patterning, survival and neurite outgrowth (Clagett-Dame et al.,2006). Although Ahd2-deficient animals are viable and exhibit no gross abnormalities, the brain, and, in particular, the mdDA system, have not been analyzed yet (Fan et al., 2003). Although the first step in RA synthesis by alcohol dehydrogenases is an ubiquitous process, tissue specificity of RA synthesis is achieved by restricted expression of aldehyde dehydrogenases, such as Ahd2(Molotkov et al., 2002). In agreement with the Ahd2 expression pattern, RA is synthesized in the mdDA area during early development, and during the postnatal and adult stages (McGaffery and Drager, 1994; Niederreither et al.,2002a). In this study, we show that maternal supplementation of RA counteracts the developmental defect in the mdDA area in Pitx3^-/- mice. After RA treatment, a significant increase in the number of TH-IR mdDA neurons was observed in the rostral part of the mdDA system of E14.5 Pitx3^-/- embryos. This effect was specific to the rostral part of the mdDA system which is the region that is most affected by Pitx3 deficiency. Importantly, no effect of RA-treatment was observed in Pitx3^+/+ embryos, which rules out the possibility that the observed increase in number of TH-IR neurons is due to increased proliferation. In addition, RA is unable to induce TH expression in MN9D cells, which suggests that the observed increase in TH-IR neurons is not due to a general induction of TH expression. Rather, a possible explanation for the increase in TH-IR neurons is that RA induces the establishment of the proper mdDA identity of immature Pitx3^-/- neurons. Indeed, previous studies show that, at E14.5, a large population of immature Pitx3^-/- neurons do not achieve their proper mdDA identity, but are still present in the area as TH-negative neurons (Maxwell et al.,2005). In agreement with these findings, our data indicate that the total number of neurons in the rostral region of the mdDA system is unaffected by either Pitx3 deficiency or RA treatment. This strongly suggests that the observed decrease of TH-immunoreactivity in E14.5 Pitx3^-/- embryos is caused by the fact that immature neurons have not established their proper mdDA identity, rather than being lost. Therefore, the increase in the number of TH-IR neurons in RA-treated Pitx3^-/- embryos appears to be the result of restored mdDA neuronal differentiation.

The effect of the temporary RA treatment was preserved to a comparable level at E18.5, as the number of TH-IR neurons was still significantly increased in RA-treated Pitx3^-/- embryos. This implies that, for a specific mdDA subset, RA is crucial between E10.75 and E13.75 to induce proper mdDA neuronal identity. Once the correct neuronal identity is established, these neurons appear to be maintained without further RA treatment. Strikingly, the increase in TH-IR number in the mdDA area is accompanied by increased dorsal innervation of the striatum. This strengthens the idea that RA induces the occurrence of a distinct functional mdDA subset in Pitx3^-/- embryos, which is normally severely affected by Pitx3-deficiency.

Altogether, these data indicate that RA can compensate for the loss of Ahd2 expression as a consequence of Pitx3 deficiency. This positions Ahd2 and its RA-generating potency central in a pathway for mdDA neuronal development, which might be linked to the final differentiation of a specific mdDA neuronal subpopulation. Interestingly, Ahd2 is downregulated in the Parkinsonian SNc, and other components of the RA-synthesis pathway have been correlated to PD (Buervenich et al.,2000; Buervenich et al.,2005; Galter et al.,2003; Grünblatt et al.,2004). Most appealing, by linking local RA synthesis to mdDA neuronal development and maintenance, a novel mechanism is proposed, with essential implications for clinical pathology as seen in PD.

We would like to thank Thomas Perlmann for kindly providing the MN9D cells,Meng Li for providing the Pitx3^+/GFP FACS-sorted cells, and Horst Simon for his kind gift of the Ahd2 fragment.