Pitx3 Is a Critical Mediator of GDNF-Induced BDNF Expression in Nigrostriatal Dopaminergic Neurons

Pitx3 is a critical homeodomain transcription factor for the proper development and survival of mesodiencephalic dopaminergic (mdDA) neurons in mammals. Several variants of this gene have been associated with human Parkinson's disease (PD), and lack of Pitx3 in mice causes the preferential loss of substantia nigra pars compacta (SNc) mdDA neurons that are most affected in PD. It is currently unclear how Pitx3 activity promotes the survival of SNc mdDA neurons and which factors act upstream and downstream of Pitx3 in this context. Here we show that a transient expression of glial cell line-derived neurotrophic factor (GDNF) in the murine ventral midbrain (VM) induces transcription of Pitx3 via NF-κB-mediated signaling, and that Pitx3 is in turn required for activating the expression of brain-derived neurotrophic factor (BDNF) in a rostrolateral (SNc) mdDA neuron subpopulation during embryogenesis. The loss of BDNF expression correlates with the increased apoptotic cell death of this mdDA neuronal subpopulation in Pitx3−/− mice, whereas treatment of VM cell cultures with BDNF augments the survival of the Pitx3−/− mdDA neurons. Most importantly, only BDNF but not GDNF protects mdDA neurons against 6-hydroxydopamine-induced cell death in the absence of Pitx3. As the feedforward regulation of GDNF, Pitx3, and BDNF expression also persists in the adult rodent brain, our data suggest that the disruption of the regulatory interaction between these three factors contributes to the loss of mdDA neurons in Pitx3−/− mutant mice and perhaps also in human PD.


Introduction
The majority of the brain's dopamine (DA) is synthesized by neurons located in two nuclei of the mammalian ventral midbrain (VM): the substantia nigra pars compacta (SNc), innervating predominantly the dorsolateral striatum, and the ventral tegmental area (VTA) projecting mainly to limbic and cortical areas of the brain (Björklund and Dunnett, 2007). The SNc neu-rons control the execution of voluntary movements, and their selective degeneration in Parkinson's disease (PD) is responsible for the characteristic motor symptoms of this disease (Dauer and Przedborski, 2003).
We have previously shown that Pitx3 activates BDNF expression in primary cell cultures (Li et al., 2009). In the present study, we demonstrate that GDNF, Pitx3, and BDNF are engaged in a feedforward regulatory pathway in the rodent VM during development and in adulthood. This feedforward interaction promotes the survival of a rostrolateral mdDA neuronal subset during embryogenesis and possibly also of adult and aging SNc neurons and protects mdDA neurons against neurotoxic insults in vitro, suggesting that it might also be relevant for the pathogenesis of PD.

Materials and Methods
Animals. C57BL/6 mice were purchased from Charles River. Adult male and pregnant female Sprague Dawley rats were purchased from the Experimental Animal Center (Shanghai, China). Generation and genotyping of Pitx3 ϩ/GFP knock-in mice is described by Zhao et al. (2004). Mice carrying the floxed Ret allele and a transgene encoding the Cre recombinase driven by the dopamine transporter (Dat) promoter (Kramer et al., 2007) often undergo Cre-mediated excision in the germline; their progeny thus carries one Ret-null allele in all cells (Ret ϩ/ Ϫ mice). Ret ϩ/ Ϫ mice were intercrossed and their offspring was genotyped by PCR. Newborn Ret Ϫ/Ϫ pups lacked kidneys and died shortly after birth, thus confirming that these mice were Ret-null mutants. Collection of embryonic stages was done from timed pregnant females; noon of the day of vaginal plug detection was designated as embryonic day 0.5 (E0.5). Animal treatment was conducted in accordance with the Laboratory Animal Care Guidelines approved by Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences (Shanghai, China) or under federal guidelines as approved by the HMGU Institutional Animal Care and Use Committee (Munich, Germany).
Primary VM cultures and neurotrophin treatments. Primary VM cultures were prepared from E14 rat embryos or E11.5 Pitx3 ϩ/GFP and Pitx3 GFP/GFP mouse embryos as described by Du et al. (2005). For GDNF treatments, E14 rat primary VM cells were treated after 7 days in vitro (DIV) with 10 ng/ml recombinant rat GDNF (Sigma) in 0.9% saline or with 0.9% saline alone (control) for 3, 24 or 48 h. In some cases, 10 g/ml NF-B inhibitor (SN50; Calbiochem) were added for 1 h to the culture medium before the GDNF treatment. Cells were harvested for immunostaining and RT-PCR assays after 3 or 24 h and for Western blot analyses after 2 d of GDNF treatment. For BDNF treatments, dissociated E11.5 Pitx3 ϩ/ GFP or Pitx3 GFP/GFP mouse VM cells were plated at a density of 1.4 ϫ 10 5 cells/well and cultured in DMEM/F12 with 2% B27 supplement (Invitrogen) containing 20 ng/ml recombinant human BDNF (R & D Systems) in 0.1% bovine serum albumin (BSA) or 0.1% BSA alone. BDNF or vehicle (0.1% BSA) was added with each medium change every second day. Cells were fixed for immunostaining after 4 d.
6-Hydroxydopamine treatments. Dissociated E11.5 Pitx3 ϩ/GFP or Pitx3 GFP/GFP mouse VM cells were plated at a density of 1.4 ϫ 10 5 cells/well and cultured in DMEM/F12 with 2% B27 supplement for 3 DIV. Cells were treated after 3 DIV with 20 ng/ml BDNF or 20 ng/ml recombinant human GDNF (R & D Systems) or with vehicle (0.1% BSA) alone 2 h before the addition of 10 M 6-hydroxydopamine (6-OHDA; catalog no. H116, Sigma) or vehicle alone (PBS) to the culture medium. Cells were fixed 24 h after the addition of 6-OHDA for immunostaining.
SH-SY5Y cells were transfected with 400 ng/well pGL3-Pitx3 promoter vector and 8 ng/well pRL-SV40 vector using Lipofectamine 2000 (Invitrogen). SN50 (10 g/ml) was added 30 min before 10 ng/ml or 20 ng/ml GDNF was given to the medium. HEK293 cells were cotransfected with 200 ng/well pGL3-BDNF wild-type or mutant promoter vector, 400 ng/ well pcDNA3.1-Pitx3 (Peng et al., 2007) or pcDNA3.1 alone, and 8 ng/ well pRL-SV40 using Lipofectamine LTX and Plus Reagent (Invitrogen). Cells were lysed in Passive Lysis Buffer 24 h (BDNF promoter assays) or 36 h (Pitx3 promoter assays) post-transfection, and Firefly and Renilla Luciferase luminescence were measured in a Centro LB 960 luminometer (Berthold Technologies) using the Dual-Luciferase Reporter Assay system (Promega) according to the manufacturers' instructions. Firefly luminescence was normalized against Renilla luminescence for each well, and relative values (fold induction) were calculated by setting the normalized value of the control transfection as 1.
Semiquantitative RT-PCR and real-time PCR assays. Semiquantitative RT-PCR was performed as reported in (Peng et al., 2007). PCR primers and conditions are provided in Table 1. For quantitative (q) PCR assays, Figure 4. Pitx3 is required for the onset of BDNF transcription in an mdDA neuronal subpopulation at E12.5. a-d, Representative midbrain coronal sections of wild-type mouse embryos at E11.5 (a-c, consecutive sections) and E12.5 (d), hybridized with Pitx3 (b), BDNF antisense (c) and BDNF sense (d) riboprobes. a, Nissl-stained bright-field view of the dark-field picture shown in b. BDNF is not expressed in the murine VM at E11.5, although Pitx3 starts to be expressed in a bilateral VM domain at this stage (red arrows in b). e-j؆, Representative coronal sections at different rostrocaudal levels of the midbrain from E12.5 Pitx3 ϩ/GFP (e-g؆) and Pitx3 GFP/GFP (h-j؆) mice, hybridized with BDNF antisense riboprobe (e-j), or immunostained for GFP (e-j). Merged images are shown in (e؆-j؆). BDNF expression is not induced in a rostrolateral and medial GFP ϩ mdDA domain of the Pitx3 GFP/GFP embryos at E12.5 (h-j؆). Note that the BDNF sense probe does not give a signal at this stage in wild-type embryos (section shown in d), demonstrating the specificity of the BDNF antisense riboprobe. k-m, Two highly conserved Pitx3/Bicoid-like transcription factor BSs in the human and mouse BDNF promoter regions [Site 2 and Site 9, positions relative to the transcription start site (TSS, arrow)] were tested for BDNF promoter activation by Pitx3 (k). Cotransfection of the pGL3-BDNF promoter vector containing these two conserved Pitx3 BSs (WT, wild type) and pcDNA3.1-Pitx3 vector (red bars) into HEK293 cells resulted in a 4.9-fold activation of the BDNF promoter relative to the control (pcDNA3.1 only, green bar) (Pitx3: 5.86 Ϯ 0.24, mean ϮS EM; **p Ͻ 0.005 in the one-way ANOVA for repeated measurements) (m). Site-directed mutagenesis (MT) of Pitx3 BS 2 (MT Site 2) and/or 9 (MT Site 9) (l ) significantly attenuated BDNF promoter activation by Pitx3 (BDNF WT promoter: 5.86 Ϯ 0.24; BDNF promoter with MT Site 2 (2M): 5.11 Ϯ 0.27; BDNF promoter with MT Site 9 (9M): 4.04 Ϯ 0.22; BDNF promoter with MT Site 2 and MT Site 9 (2M/9M): 3.6 Ϯ 0.08, mean Ϯ SEM; *p Ͻ 0.05; **p Ͻ 0.005 in the one-way ANOVA for repeated measurements) (m). Data were derived from three independent experiments. Scale bars: (in d and j؆), 100 m.
1 l of a 1:3 diluted cDNA was amplified in a DNA Engine Opticon (MJ Research) together with 0.5ϫ SYBR Green I (Roche Applied Science) and 5 pmol of each primer (Table 1). A standard curve was constructed using plasmid DNA containing the Pitx3 cDNA (from 10 2 to 10 7 molecules) (Peng et al., 2007). Signals from specific Pitx3 and BDNF PCR products were normalized against ␤-actin, and relative values were calculated by setting the normalized value of controls as 1. Samples from three independent experiments were analyzed in duplicate reactions for each cDNA, and all reactions were repeated more than twice.
Intrastriatal injections of GDNF. Adult male rats (n ϭ 5) were anesthetized with chloral hydrate (400 mg/kg, i.p.). Twenty microliters of a 15 ng/l GDNF solution in 0.9% saline were injected into the right striatum (according to bregma: anterior-posterior, ϩ0.7; medial-lateral, Ϯ3.5; dorsal-ventral, Ϫ4.0) with a Hamilton syringe. Vehicle (0.9% saline) was injected into the contralateral striatum. The injection rate was 1 l/min, and the cannula was left in place for another 10 min before it was retracted. After 48 h, animals were sacrificed by CO 2 asphyxiation. For immunohistochemical analyses, animals were intracardially perfused with PBS followed by 4% paraformaldehyde (PFA), and brains were collected, postfixed in 4% PFA for 24 h, cryosectioned, and processed as described above. For Western blot analyses, brains were immediately removed and ventral midbrain tissue was dissected and processed as described before.
Statistics. All values shown are mean Ϯ SEM. Statistical significance between groups was assessed by paired t tests, independent samples t tests, or one-way ANOVA followed by post hoc S-N-K multiple comparisons using the SPSS 10.0 software (SPSS Inc.). A value of p Ͻ 0.05 was considered significant.

Induction of Pitx3 expression by NF-B-mediated GDNF/Ret signaling
To determine whether GDNF-mediated signaling is involved in the transcriptional activation of the Pitx3 gene during embryogenesis, we first compared the spatiotemporal expression patterns of GDNF and Pitx3 in the embryonic mouse VM. GDNF is expressed in the mantle zone (MZ) of the midbrain basal plate (ventrolateral midbrain) at E10.5, one day before the onset of Pitx3 transcription in this brain region (Fig. 1a,c,e). Pitx3 expression is detected at E11.5 in the MZ of the adjacent floor plate. At this stage, GDNF expression is restricted to a smaller domain in the midbrain basal plate (Fig. 1b,d,f ), and becomes undetectable in the mouse VM at E12.5 (data not shown).
Given the spatiotemporal correlation of GDNF expression and the initiation of Pitx3 transcription in the embryonic VM, we next asked whether GDNF can promote the expression of Pitx3 in primary VM cells in vitro. Treatment of E14 rat primary VM cell cultures (grown in the absence of neurotrophins for 7 DIV) with 10 ng/ml GDNF for 24 h significantly increased the number of Pitx3 ϩ and Pitx3 ϩ /TH ϩ cells within the VM cell population by 1.7-and 1.3-fold, respectively (Fig. 1g-m). Under these experimental conditions, GDNF treatment did not significantly increase the number of TH ϩ neurons (Fig. 1m).
To explore whether GDNF signaling is required for induction of Pitx3 expression in vivo, we counted the number of Pitx3 ϩ cells in GDNF receptor Ret knock-out embryos. We found that the number of Pitx3 ϩ cells in Ret Ϫ/Ϫ embryos was reduced by 17.7% at E12.5 and by 10.4% at E14.5 compared to their heterozygous or wild-type littermates (Fig. 1n-u). Pitx3 expression was particularly lost in ventromedial (Fig. 1nЉ,qЉ, insets) and lateral (data not shown) TH ϩ cells in the Ret Ϫ/Ϫ embryos, suggesting that GDNF signaling might specifically induce Pitx3 expression in a TH ϩ mdDA neuron subpopulation. The partial recovery of Pitx3 ϩ cell numbers in the E14.5 Ret Ϫ/Ϫ embryos might be due to compensation by other GDNF receptors, such as GFR␣1 and NCAM, and could explain the lack of an mdDA phenotype in postnatal DAT-Ret flx//flx mice (Kramer et al., 2007;Paratcha and Ledda, 2008).
We next tested whether GDNF signaling directly activates the Pitx3 promoter in rodent VM cells. GDNF-mediated signaling enhances the nuclear translocation of NF-B protein complexes and subsequent activation of NF-B target genes in mdDA neurons Wang et al., 2008). Analysis of the human, mouse, and rat Pitx3 promoter regions using bioinformatics prediction tools revealed the existence of conserved NF-B BSs in these regions (Fig. 2a), and treatment of E14 rat primary VM cultures after 7 DIV with GDNF for 3 or 24 h increased the nuclear translocation of NF-B, as expected ( Fig. 2b-k). To determine whether this treatment resulted in increased activation of the Pitx3 promoter, we made use of a Pitx3 promoter/reporter construct containing a conserved NF-B BS. GDNF treatment of SH-SY5Y cells transfected with this Pitx3 reporter construct resulted in a dose-dependent activation (80 -110% increase) of the Pitx3 promoter (Fig. 2l ). This activation was prevented when we treated the cells with SN50, an inhibitor of NF-B nuclear translocation . We thus concluded that GDNF/Ret signaling is necessary and sufficient for the activation of the Pitx3 promoter in the embryonic rodent VM, and we identified NF-B as a mediator of this process.  a, b) and Pitx3 GFP/GFP (c, d) embryos at E12.5 revealed an increase of apoptotic Pitx3 ϩ mdDA neurons [cCasp3 ϩ (red) and GFP ϩ (green) double-labeled cells, white arrowheads) in the mutant VM. White arrows point at erythrocytes that were stained unspecifically for cCasp3. e, Quantification of cCasp3 ϩ / GFP ϩ double-labeled cells relative to the total amount of GFP ϩ (Pitx3 ϩ ) cells in these sections showed a significant increase of apoptotic mdDA neurons in the rostral (cCasp3 ϩ /GFP ϩ cells: E12.5 Pitx3 ϩ/GFP , 0.506 Ϯ 0.047%; E12.5 Pitx3 GFP/GFP , 1.676 Ϯ 0.185%; n ϭ 3, mean Ϯ SEM; ***p Ͻ 0.001 in the independent samples t test), but not caudal (cCasp3 ϩ /GFP ϩ cells: E12.5 Pitx3 ϩ/GFP , 0.902 Ϯ 0.068%; Pitx3 GFP/GFP , 1.056 Ϯ 0.042%, n ϭ 3, mean Ϯ SEM), VM of the Pitx3 GFP/GFP embryos. Scale bar: (in d), 100 m.

GDNF-mediated activation of Pitx3 expression induces BDNF transcription in vitro
We have previously shown that overexpression of Pitx3 upregulates the transcription and secretion of BDNF protein from neuroblastoma cells and primary VM as well as astrocyte cultures (Peng et al., 2007;Yang et al., 2008). We therefore hypothesized that GDNF-mediated activation of Pitx3 expression might subsequently induce the transcription of BDNF in these cells. To test this hypothesis, we treated E14 rat primary VM cultures after 7 DIV with GDNF and evaluated the transcription of Pitx3 and BDNF under these conditions. As expected, GDNF treatment increased the transcription of Pitx3 and BDNF in these cultures by 1.3-and 1.4fold, respectively, and this effect of GDNF on Pitx3 and BDNF transcription was blocked by the application of the NF-B inhibitor SN50 (Fig. 3a,c). These results were also confirmed at protein levels ( Fig.  3b). To further establish whether the GDNF-induced activation of BDNF transcription was mediated by Pitx3 and not by other Pitx3-independent pathways, we depleted Pitx3 in these cultures by siRNAmediated knock-down. Quantitative RT-PCR indicated that transfection of a Pitx3 siRNA after 1 h of GDNF treatmentreduced Pitx3 mRNA levels by 56%, concomitant with a 32% reduction of BDNF mRNA levels in these cultures (Fig. 3d,f ). A slight but not significant decrease in Pitx3 and BDNF mRNA levels was also observed after transfection of a control (nonsilencing) siRNA oligonucleotide (Fig. 3d,f ). This reduction was most likely unspecific and due to the lipofection of the oligonucleotides, as a similar decrease in mRNA levels was observed in shamlipofected GDNF-treated cultures as compared to untransfected controls (data not shown). We also used Western blotting to confirm that Pitx3 knock-down leads to a reduction of Pitx3 and BDNF protein levels in GDNF-treated primary VM cultures (Fig. 3e). These results therefore suggest that Pitx3 mediates the transcriptional activation of the BDNF gene following treatment of primary VM cells with GDNF.

Pitx3 is required for BDNF transcription in an mdDA neuronal subset in vivo
Our results raised the possibility that Pitx3 is an essential regulator of BDNF gene expression in the rodent VM. To address this possibility, we analyzed the time course of BDNF expression in wild-type (Pitx3 ϩ/ϩ and Pitx3 ϩ/GFP mice) and Pitx3- Figure 6. Lack of BDNF expression in an mdDA neuronal subpopulation that disappears at later embryonic/postnatal stages in the absence of Pitx3. (a-l؆) Representative coronal midbrain sections from E14.5 (a-f؆, different rostrocaudal levels) and P30 (g-l؆) Pitx3 ϩ/GFP (a-c؆; g-i؆) and Pitx3 GFP/GFP (d-f؆; j-l؆) mice hybridized with a BDNF riboprobe (a-f ) or immunostained for TH (a-f, g-l ) and BDNF (g-l). Merged images are shown in a؆-f؆ and g؆-l؆. BDNF is expressed strongly in a rostrolateral TH ϩ mdDA domain in the wild-type (Pitx3 ϩ/GFP ) embryos (a-c؆; g-i؆), but this BDNF ϩ domain was completely lost in the VM of Pitx3 GFP/GFP embryos at E14.5 (d, d؆, e, e؆,f, f؆), concomitant with a strong reduction at E14.5 (d-f) and followed by the almost complete loss at P30 (j-l؆) of the rostrolateral (SNc) TH ϩ mdDA neurons in these mutants. Scale bars: (in f؆, J؆, and l ؆),100 m.
null mutant (Pitx3 GFG/GFP ) mice, which allow the fate mapping of Pitx3-expressing (GFP ϩ ) cells (Maxwell et al., 2005). In wild-type mouse embryos, Pitx3 expression in the VM was first detected around E11.0 -E11.5 (Fig. 1e,f ) (Smidt et al., 1997); at this or earlier stages, BDNF expression was not detected in the VM (Fig. 4a-c and data not shown). We found that BDNF was widely transcribed in the wild-type and Pitx3 ϩ/GFP VM at E12.5 and exhibited a particularly prominent expression domain in the rostrolateral and medial VM overlapping with GFP ϩ (Pitx3 ϩ ) mdDA neurons (Fig. 4e-gЉ and data not  shown). Remarkably, BDNF expression in this rostrolateral and medial domain was entirely lost in the Pitx3 GFP/GFP embryos at E12.5, although the corresponding GFP ϩ neurons were still present in this domain and despite the persistent expression of BDNF in the adjacent Pitx3-negative (GFP Ϫ ) VM tissue of the null mutants ( Fig. 4 h-jЉ).
To establish whether the loss of BDNF expression in the rostrolateral and medial VM of the Pitx3 GFP/GFP embryos is rapidly followed by the death of GFP ϩ (Pitx3 ϩ ) mdDA neurons in these embryos, we assessed the proportion of apoptotic (cCasp3 ϩ ) and GFP ϩ cells relative to the total number of GFP ϩ mdDA neurons in wild-type (Pitx3 ϩ/GFP ) and mutant (Pitx3 GFP/GFP ) embryos at E12.5 (Fig. 5a-e). This proportion was significantly increased by ϳ3-fold in the rostral but not caudal VM of the Pitx3 GFP/GFP embryos as compared to their wild-type littermates (Fig. 5e), indicating that the lack of BDNF expression indeed correlates with a reduced survival of mdDA neurons in the rostral VM of the Pitx3 GFP/GFP embryos.
We next investigated the effect of Pitx3 inactivation on BDNF expression in late gestational and postnatal mdDA neurons. BDNF expression was still not detected in the rostrolateral and medial VM domains of Pitx3 GFP/GFP embryos at E14.5 (Fig. 6a-fЉ) and E16.5 (data not shown). Loss of Pitx3 resulted in a notable decrease of TH ϩ SNc mdDA neurons in the null mutant VM at postnatal day 30 (P30), concomitant with a strong reduction of BDNF-expressing cells in the same region (Fig. 6 g-lЉ). Notably, BDNF transcription appeared to be very low or almost undetectable in the caudomedial mdDA domain throughout embryonic development , and this mdDA neuronal subset was less affected in the Pitx3 GFP/GFP -null mutants at embryonic and postnatal stages (Figs. 4h-jЉ, 5e, 6d-fЉ,j-lЉ;Maxwell et al., 2005). Collectively, our results indicate that Pitx3 directly activates the BDNF promoter in an mdDA neuron subpopulation located in the rostrolateral and medial VM of the mouse embryo and suggest that the Pitx3-mediated activation of BDNF expression is necessary for the survival of this mdDA neuronal subset at embryonic and postnatal stages.

BDNF augments the survival of mdDA neurons in Pitx3-null mutant primary VM cultures
To further investigate whether the lack of BDNF expression contributes to the loss of mdDA neurons in Pitx3 GFP/GFP mice (Fig. 5 and Maxwell et al., 2005), we tested whether BDNF treatment is sufficient to rescue the numbers of mdDA neurons in primary VM cultures derived from E11.5 Pitx3 GFP/GFP embryos. We found a reduction of GFP ϩ cell numbers (cells that would express Pitx3 in the wild type) by 13%, TH ϩ cells by 37.5%, and GFP/TH double-positive cells by 40% in the untreated Pitx3 GFP/GFP cultures as compared to untreated wild-type (Pitx3 ϩ/GFP ) cultures ( Fig. 7a-f,m). This result is consistent with in vivo data from E12.5 Pitx3 GFP/GFP mice showing a reduction of GFP ϩ cells by 21%, TH ϩ cells by 54%, and GFP/TH double-positive cells by 48% relative to the Pitx3 ϩ/GFP controls (Maxwell et al., 2005), and it confirms a selective loss of GFP ϩ and TH ϩ cells in the absence of Pitx3 also in primary VM cultures. Notably, BDNF treatment of Pitx3 GFP/GFP VM cultures increased the numbers of GFP ϩ cells by 29.6%, TH ϩ cells by 84.7%, and TH/GFP double-positive cells by 66.7% relative to the untreated Pitx3 GFP/GFP cultures, thereby reaching similar numbers as in the BDNF-treated wild-type (Pitx3 ϩ/GFP ) cultures (Fig. 7g-m). To determine whether this was due to a survival-promoting effect of the BDNF treatment on the cultured mdDA neurons, we assessed the proportion of apoptotic (cCasp3 ϩ ) mdDA (GFP ϩ ) neurons in the untreated and BDNFtreated VM cultures from wild-type (Pitx3 ϩ/GFP ) and mutant (Pitx3 GFP/GFP ) embryos. As expected, this proportion was significantly increased in the untreated Pitx3 GFP/GFP VM cultures as  (control, a, c, e, g, i, k) and 15 ng/l GDNF into the contralateral side of the striatum (b, d, f, h, j, l ) 48 h earlier, immunostained for TH (a, b, g, h), Pitx3 (c, d) and BDNF (i, j). Merged images are shown in e, f, k, and l. Insets in k and l are higher magnifications of the boxed areas in k and l, respectively. m, Schematic drawing of the striatal injection site and position of the VM sections shown in a-l. n, Western blots using protein extracts from vehicle-treated control and GDNF-treated contralateral side of the VM showed that unilateral intrastriatal injection of GDNF leads to a 2.8-fold increase of Pitx3 expression (quantification in o) and 6.1-fold increase of BDNF expression (quantification in p) in the ipsilateral mdDA neurons (control: Pitx3, 0.18 Ϯ 0.02; BDNF, 0.33 Ϯ 0.04, mean Ϯ SEM; GDNF-treated (GDNF): Pitx3, 0.68 Ϯ 0.07; BDNF, 2.35 Ϯ 0.18; **p Ͻ 0.01 in the independent samples t test for repeated measurements; n ϭ 3). Scale bar: (in f ), 50 m.
compared to the untreated controls ( Fig. 7n-s,w). Notably, the increased death of GFP ϩ cells in the Pitx3 GFP/GFP VM cultures was rescued back to control levels by the addition of BDNF to these cultures (Fig. 7t-w), indicating that in the absence of Pitx3, exogenous BDNF application prevents the cell death of cultured Pitx3-null (Pitx3 GFP/GFP ) mdDA neurons.

Intrastriatal injection of GDNF upregulates Pitx3 and BDNF expression in the adult SNc
Our previous results strongly suggested the existence of a feedforward mechanism for the initiation and/or maintenance of GDNF, Pitx3, and BDNF expression in the murine VM during embryonic development. We therefore hypothesized that this Figure 9. BDNF, but not GDNF, protects Pitx3-null mutant mdDA neurons against 6-OHDA neurotoxicity. a-x, Primary VM cultures derived from E11.5 Pitx3 ϩ/GFP (a-l ) and Pitx3 GFP/GFP (m-x) embryos were treated after 3 DIV with vehicle (0.1% BSA) (a-f, m-r), 20 ng/ml GDNF (g-i, s-u), or 20 ng/ml BDNF (j-l, v-x); after 2 h, cells were incubated with 10 M 6-OHDA (d-l, p-x) or vehicle (PBS; a-c, m-o) for 24 h and immunostained for TH (red in a, d, g, j, m, p, s, and v) and GFP (green in b, e, h, k, n, q, t, and w); merged images with DAPI stain are shown in c, f, i, l, o, r,  u, and x. y, Quantification of TH-immunopositive cells in these eight experimental groups revealed that in Pitx3 ϩ/GFP VM cultures both GDNF and BDNF protect TH ϩ mdDA neurons against 6-OHDA toxicity, but in Pitx3 GFP/GFP VM cultures only BDNF (and not GDNF) protects these neurons against 6-OHDA toxicity [TH ϩ /DAPI ϩ cells (white arrowheads): Pitx3 ϩ/GFP vehicle-treated, normalized to 1; Pitx3 ϩ/GFP 6-OHDA-treated, 0.615 Ϯ 0.025; Pitx3 ϩ/GFP 6-OHDA ϩ GDNF-treated, 0.816 Ϯ 0.037; Pitx3 ϩ/GFP 6-OHDA ϩ BDNF-treated, 0.768 Ϯ 0.066, mean Ϯ SEM, n ϭ 3; Pitx3 GFP/GFP vehicle-treated, normalized to 1; Pitx3 GFP/GFP 6-OHDA-treated, 0.521 Ϯ 0.096; Pitx3 GFP/GFP 6-OHDA ϩ GDNF-treated, 0.533 Ϯ 0.051; Pitx3 GFP/GFP 6-OHDA ϩ BDNF-treated, 0.946 Ϯ 0.111, mean Ϯ SEM, n ϭ 4]. Paired t test was used for statistical analysis of treatment effects. *p Ͻ 0.05; **p Ͻ 0.01; ns, not significant. Data were derived from at least three independent experiments. Scale bar: (in x), 50 m. feedforward mechanism might also persist in the adult brain, as it is known that GDNF released from striatal target cells is taken up and retrogradely transported to the soma of mdDA neurons, where it promotes their survival in the early postnatal and adult rodent brain (Tomac et al., 1995;Oo et al., 2003;Kholodilov et al., 2004). To test this hypothesis, we injected GDNF protein unilaterally into the striatum of adult rats, and 48 h later we analyzed the endogenous expression levels of Pitx3 and BDNF proteins in the ipsilateral (GDNF-treated) SNc relative to the contralateral (vehicle-treated) control side. We found that Pitx3 and BDNF protein levels were increased by 2.8-and 6.1-fold, respectively, in the ipsilateral (GDNFtreated) SNc, as determined by immunohistochemistry ( Fig. 8a-l ) and on Western blots (Fig. 8n-p). These data provide direct evidence that retrograde GDNF signaling stimulates Pitx3 and BDNF expression in mature SNc mdDA neurons.
BDNF, but not GDNF, protects Pitx3 ؊/؊ mdDA neurons against 6-OHDA neurotoxicity Given the previous result, we wanted to know whether the neuroprotective effect of GDNF against 6-OHDA toxicity on mdDA neurons might be due to Pitx3mediated activation of BDNF expression in these neurons. Therefore, we treated wild-type (Pitx3 ϩ/GFP ) and Pitx3-null mutant (Pitx3 GFP/GFP ) VM cultures after 3 DIV with 6-OHDA and assessed the effect of a previous 2 h incubation with GDNF or BDNF on these 6-OHDA-treated cultures. Treatment with 6-OHDA significantly reduced the numbers of TH ϩ cells in both wild-type and mutant cultures ( Fig. 9a-f,m-r,y) as expected. Prior incubation with GDNF or BDNF significantly rescued the numbers of TH ϩ mdDA neurons in the wild-type (Pitx3 ϩ/GFP ) VM cultures, reaching a similar level as that under normal (untreated) conditions ( Fig. 9 g-l,y). Remarkably, the numbers of TH ϩ mdDA neurons in the 6-OHDA-treated Pitx3 GFP/GFP VMcultureswereonlyrescuedbacktonormal (untreated) levels after a previous incubation of these cells with BDNF, but not with GDNF ( Fig. 9s-y), indicating that GDNF cannot protect TH ϩ mdDA neurons against 6-OHDA-induced neurotoxicity in the absence of Pitx3.
induced apoptotic cell death, we assessed the relative proportion of apoptotic (cCasp3 ϩ ) and GFP ϩ mdDA neurons in the 6-OHDA-treated VM cultures alone and after previous incubation with GDNF or BDNF. As expected, treatment with 6-OHDA alone significantly increased the numbers of apoptotic mdDA neurons in these cultures, regardless of their genotype ( Fig. 10a-f,m-r,y). In line with our previous findings, BDNF significantly decreased the numbers of apoptotic mdDA neurons down to normal (untreated) levels in both Pitx3 ϩ/GFP and Pitx3 GFP/GFP 6-OHDA-treated cultures (Fig. 10jl,y), whereas GDNF only decreased these numbers significantly in Pitx3 ϩ/GFP but not in Pitx3 GFP/GFP 6-OHDA-treated cultures ( Fig. 10 g -i,s-u,y). Altogether, these results strongly suggest that GDNF exerts its neuroprotective and survivalpromoting effect on at least a subpopulation of mdDA neurons through the activation of Pitx3-mediated BDNF expression in these cells. In the absence of Pitx3, GDNF cannot protect the mutant mdDA neurons against 6-OHDA-induced neurotoxicity and apoptotic cell death, and these cells are only rescued by the exogenous application of BDNF acting downstream of GDNF and Pitx3, as suggested by our previous results.

Discussion
The secreted factors GDNF and BDNF and the transcription factor Pitx3 are individually required for the proper development of mdDA neurons or for their survival during postnatal and adult stages (Baquet et al., 2005;Smidt and Burbach, 2007;Pascual et al., 2008;Oo et al., 2009), but it remained unclear whether a regulatory interaction exists between these three factors during embryonic development or in the adult brain. Here we show that NF-B-mediated GDNF/Ret signaling is, at least in part, sufficient for the activation of Pitx3 expression, which is in turn required for the transcription of BDNF in a rostrolateral and medial mdDA neuronal subpopulation, thereby promoting the survival of these neurons during mouse embryonic development (Fig.  11). We also show that this feedforward regulation of GDNF, Pitx3, and BDNF expression is still active in the adult rodent brain and that, in the absence of Pitx3, BDNF but not GDNF protects mdDA neurons against 6-OHDA toxicity in vitro. Altogether, our data indicate that the regulatory interaction between GDNF, Pitx3, and BDNF is necessary for the survival and protection of mdDA neurons against neurotoxic insults during embryogenesis, and that this regulatory interaction might also be relevant for the survival and neuroprotection of adult mdDA neurons (Fig. 11).

GDNF/Ret signaling activates Pitx3 expression in an mdDA neuronal subpopulation
The relevance of GDNF signaling for the normal development of mdDA neurons in vivo has remained controversial until now. Inactivation of the murine GDNF, Ret, and Gfra1 genes apparently does not interfere with prenatal mdDA neuron differentiation or survival (Airaksinen and Saarma, 2002;Kramer et al., 2007;Paratcha and Ledda, 2008). Moreover, persistent overexpression of GDNF driven by the TH promoter leads to a reduction of SNc DA neurons (Chun et al., 2002), whereas intrastriatal or intranigral injections of GDNF protein during early postnatal stages, or the constitutive activation of the Ret receptor, result in increased numbers of TH ϩ SNc neurons and an enhanced DA metabolism (Beck et al., 1996;Mijatovic et al., 2007). The transient but detectable expression of GDNF in the midgestational VM shown here might have been missed by previous in situ hybridization studies due to lower sensitivity or slight differences in the staging of the embryos (Hellmich et al., 1996;Golden et al., 1999), but GDNF expression in the embryonic rat VM was also shown by Choi-Lundberg and Bohn (1995) using a more sensitive RT-PCR-based detection method. Our and previous findings therefore suggest that the transient expression of GDNF in the rodent VM might have two different functions during development: on the one hand, GDNF signaling might activate Pitx3 transcription in an mdDA precursor subpopulation; on the other hand, GDNF might provide transient trophic support to developing progenitors and postmitotic neurons in this region of the brain. The requirement of GDNF signaling for these two processes, however, might be compensated by other factors during embryonic development, leading to the absence of an mdDA phenotype in the GDNF, Ret, and Gfra1-null mutants at birth (Airaksinen and Saarma, 2002;Paratcha and Ledda, 2008). Our and previous results also suggest that the tight regulation of GDNF protein levels is critical for proper mdDA neuron development. The notion that a transient (rather than persistent) expression of GDNF during embryonic development might have an important permissive role is supported by the observation that sustained GDNF expression in differentiating TH ϩ SNc neurons results in a marked reduction of these neurons shortly after birth (Chun et al., 2002).
Pitx3 is necessary for the activation of BDNF transcription in an mdDA neuronal subset Pitx3 is required for the initiation of TH expression in an mdDA neuronal subset during development and for the survival of these Figure 11. Feed forward regulation of GDNF, Pitx3, and BDNF expression in the embryonic and adult rodent brain. a, During development, GDNF (red) is transiently expressed in the midbrain BP of the E10.5/E11.5 mouse embryo, where it is required for the NF-B-mediated induction of Pitx3 expression in an mdDA neuronal subset within the adjacent midbrain floor plate. Pitx3 (green) is in turn necessary for the activation of BDNF expression in a rostrolateral/medial mdDA neuronal subpopulation at E12.5. Neurotrophic support of this mdDA neuronal subset by BDNF (blue) might be required for their survival throughout development and their protection against neurotoxic insults. b, In the adult brain, striatal uptake and retrograde axonal transport of GDNF (red) to the cell soma maintains the proper levels of Pitx3 (green) and BDNF (blue) expression in SNc mdDA neurons and might thereby contribute to the sustained survival and neuroprotection of these neurons throughout adulthood.
neurons during subsequent prenatal and postnatal stages (Maxwell et al., 2005;Smidt and Burbach, 2007). Aldh1a1 was identified as a target gene that could mediate the prosurvival activity of Pitx3, as maternal RA complementation partially rescues the loss of mdDA neurons in Pitx3 mutants (Jacobs et al., 2007). Here we identify BDNF as another target gene of Pitx3 in the rostrolateral and medial mdDA neuronal subpopulation that is most affected in Pitx3 Ϫ/Ϫ mice (Smidt et al., 2004;Maxwell et al., 2005). Because BDNF treatment augmented the survival of mdDA neurons in primary VM cultures prepared from E11.5 Pitx3 GFP/GFP -null mutant embryos, we suggest that the failure to induce BDNF expression in rostrolateral and medial (SNc) mdDA precursors during development contributes to the preferential degeneration of these neurons in the Pitx3 GFP/GFP mice. Interestingly, retinoid signaling via retinoic acid receptors was recently shown to induce BDNF transcription and to prevent the inflammatory degeneration of mdDA neurons in vitro and in vivo (Katsuki et al., 2009). While our data strongly suggest a direct activation of the BDNF gene by Pitx3, we cannot exclude that Pitx3 might indirectly maintain BDNF expression through the induction of Aldh1a1 and subsequent local production of RA in an mdDA neuronal subset.

GDNF protects mdDA neurons against 6-OHDA toxicity only in the presence of Pitx3
Treatment of Pitx3 GFP/GFP -null mutant VM cultures with 6-OHDA resulted in an increased survival of mdDA neurons in these cultures only after preincubation with BDNF, but not with GDNF, suggesting that GDNF acts upstream of Pitx3 and BDNF to protect mdDA neurons against cytotoxic insults and to increase their survival under these conditions. BDNF is in fact more potent than GDNF in promoting the survival of mdDA neurons after unilateral 6-OHDA lesion of the SNc in organotypic cultures of the rat VM (Stahl et al., 2011). Moreover, GDNF appears to promote primarily the survival of calbindin-expressing VTA neurons under basal conditions (Meyer et al., 1999) and of VTA and rostromedial SNc neurons after 6-OHDA lesion (Barroso-Chinea et al., 2005), which are the cell groups that are not affected by the loss of BDNF expression in our Pitx3 GFP/GFP mice. Together, our and previous findings therefore suggest that GDNF acts as a potent survival factor for those medial mdDA neuronal subpopulations projecting to the GDNF-rich ventral striatum (Barroso-Chinea et al., 2005). However, the rostrolateral mdDA neurons projecting to the dorsolateral striatum (where lower levels of GDNF are expressed; Barroso-Chinea et al., 2005) depend on Pitx3-mediated local BDNF synthesis for their survival and neuroprotection.

The feedforward regulation of GDNF, Pitx3, and BDNF expression in the adult SNc might be relevant for the pathogenesis of PD
We also found that the intrastriatal delivery of GDNF protein in the adult rat brain increased the expression of Pitx3 and BDNF in the ipsilateral SNc. This finding raises the possibility that retrograde transport of GDNF or local signaling at axon terminals in the striatum controls the levels of Pitx3 and BDNF proteins in adult mdDA neurons (Fig. 11) and might have implications for the pathogenesis and treatment of PD. Among all neurotrophic factors, GDNF exhibits the most severe decrease in the SNc of PD patients (Chauhan et al., 2001). Reduced intranigral GDNF levels might lead to reduced Pitx3 expression and thus to decreased expression of BDNF in SNc mdDA neurons. This might render these neurons more susceptible to death in the diseased brain. A reduction of BDNF mRNA and protein levels in the SNc of PD patients has been consistently reported by several groups (Mogi et al., 1999;Parain et al., 1999;Howells et al., 2000;Chauhan et al., 2001). Interestingly, some of these studies also noted in the patients the preferential loss of a subset of SNc neurons that normally express high levels of BDNF (Parain et al., 1999;Howells et al., 2000). Moreover, several polymorphisms in noncoding regions of the PITX3 gene (which might alter PITX3 expression) are associated with idiopathic or early onset PD (Fuchs et al., 2009;Bergman et al., 2010;Haubenberger et al., 2011;Le et al., 2011), and one polymorphism in the promoter region of the human BDNF gene was associated with familial PD (Parsian et al., 2004). These findings, together with our results and the observation that conditional ablation of the BDNF gene during mouse development leads to the selective loss of a subset of TH ϩ neurons in the SNc, but not VTA (Baquet et al., 2005), raise the intriguing possibility that the rostrolateral (SNc) mdDA neuron subpopulation expressing high levels of BDNF is most affected by the loss of PITX3 in human PD patients and cannot be protected by the exogenous application of GDNF. We suggest that a developmental failure to induce normal expression of Pitx3 coupled to age-related decreases in GDNF levels (Miyazaki et al., 2003) might lead to a depletion of BDNF in SNc neurons and predispose them to neurodegeneration (Porritt et al., 2005). According to this model, alterations in this neurotrophic factor reinforcement loop might lead to reduced neurotrophic support of SNc mdDA neurons, thus contributing to PD. Further dissection of the molecular underpinnings of this regulatory network might therefore facilitate the development of novel therapeutic approaches for the treatment of PD.