Ventral Mesencephalon-Enriched Genes That Regulate the Development of Dopaminergic Neurons In Vivo

Microarray data processing and gene expression profiling of the developing VM

First, the microarray data were analyzed for reproducibility. The correlation coefficient for five consecutive time points (E10–E14) in biological triplicates ranged from 0.96 to 0.99 with a mean correlation coefficient of 0.97 between three replicates, demonstrating a high degree of reproducibility. A series of genes was detected that are known to be critical for the specification and differentiation of mDA neurons as well as for DA biosynthesis and release, suggesting high reliability of the data (data not shown). This ensured a high likelihood of identifying differentially expressed genes involved in mDA neuron development from the microarray data.

To identify those differentially expressed genes reliably, two different but complementary statistical methodologies, limmar and SAM, were used to analyze the microarray data further. We assessed the significance of differences in gene expression by pairwise comparisons at each embryonic age. Those genes, whose levels were altered at least twofold in each pairwise comparison (10 comparisons in total for five time points), and intersected by the two statistical methodologies, were selected for further analysis. Up to 1285 such genes were selected (limmar p = 0.01, SAM Δ = 0.4). Among them, expression levels of 979 genes were upregulated, whereas 306 genes were downregulated during E10–E14.

Quantitative analysis showed that the most remarkable change in the number of gene transcripts (215) between two adjacent embryonic ages occurred between E11 and E12. Gene expression variation was much less common between E10 and E11 as well as between E13 and E14 (supplemental Fig. S1A,B, available at www.jneurosci.org as supplemental material). No genes were detected whose levels were altered more than twofold between E12 and E13 or between E13 and E14. These data suggest that VM cells between E10 and E14 share common genetic determinants at different stages of development. E12 is a turning point for gene expression variation during VM development with greater fluctuation in gene expression during E10–E12 and less variation from E12–E14.

VM cells of different embryonic age share clusters of developmentally regulated genes

It has been shown that multiple genetic cascades control the development of mDA neurons in an mDA neuron-specific or -nonspecific manner. To obtain a global view of the first set of microarray data, we first performed hierarchical clustering analysis to reveal changes in gene expression while mDA neurons undergo differentiation. We found that the majority of genes were expressed in the VM at each embryonic age examined, but with different profiles, given that their levels were specifically upregulated or downregulated during development (supplemental Fig. S1B–D, available at www.jneurosci.org as supplemental material). This suggests that mDA neuron development is regulated by a developmental-specific genetic network. In this context, combinations of a small set of genes, whose expression is embryonic age dependent, may play critical roles at a specific developmental stage. Together with the data shown in supplemental Figure S1A (available at www.jneurosci.org as supplemental material), this analysis also shows that the VMs at each embryonic age were molecularly similar to each other. In particular, gene expression profiles between adjacent ages (E10 vs E11, E13 vs E14) were extremely highly similar. This further emphasizes the significance of a small set of differentially expressed genes in VM development.

Identification of VM-enriched genes

To identify VM-enriched genes, 162 upregulated and 24 downregulated genes were selected for further analysis from 1285 differentially expressed genes identified in microarray data. Their trends of expression were analyzed individually. As a result, a set of VM-enriched genes was identified, and expression trends of some of the potentially important genes are shown in Figure 1. These genes fall into one of three categories based on expression profiles, which is suggestive of a specific role in a distinct aspects of VM development: (1) genes whose levels were significantly upregulated or downregulated in early development of mDA neuron (E10–E11) (Fig. 1A), and might be important for the maintenance of DA precursor cells status; (2) genes whose levels were significantly altered in intermediate development of mDA neurons (E11–E12) (Fig. 1B) and might be required for withdrawal of mDA precursor cells from mitosis and initiation of postmitotic differentiation; or (3) genes whose levels were significantly changed in late development of mDA neurons (E12–E14) (Fig. 1C) and might be important for late mDA neuron differentiation, including migration and neurite outgrowth.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Expression profiles of a subset of VM-enriched genes identified by microarray analysis. A subset of biologically interesting genes that were selected from a larger group of differentially expressed genes is shown. They were classified into three groups (A–C), based on expression profiles suggestive of their potential roles during mDA development. The prototypical expression profile of each gene is shown in the right panel. The genes shown in bold are those selected for further analysis. Their graphic gene expression profiles are shown in Figure 3 (except for Ebf1). Arrows indicate upregulated (↑) or downregulated (↓) expression levels of the corresponding genes during mDA neuron development. Values represent means ± SEM; *p < 0.01 (ANOVA followed by Newman–Keuls test, n = 3), compared with controls.

Laminar structure of the VM

To better describe the characteristics of the spatial expression pattern of VM-enriched genes, we first defined the morphological organization of the E12 VM using ISH and immunostaining with the aid of established markers for mDA neurons that also showed altered expression in the microarray data. At this point, the VM cells undergo withdrawal from the cell cycle and switch to postmitotic differentiation as described in supplemental Figure S1E–H (available at www.jneurosci.org as supplemental material). The gene expression pattern demonstrated that the VM in sagittal section may be divided into four layers. Intense signals for both Sox2 and Ki67, which are known to be present in mitotic neuroepithelial cells of the ventricular zone (VZ) (Kele et al., 2006), were detected in the outer stripe of the dorsal layer of the VM (designated layer I) (Fig. 2A,B). In support of this, double-labeled staining showed that Ki67⁺ cells were devoid of βIII-tubulin (Fig. 2B) and TH (data not shown). In contrast, Ngn2 hybridization signals were primarily concentrated in the stripe (layer II) ventral to the VZ (layer I) (Fig. 2C), suggesting that the cell fate of neuroepithelial cells is restricted to mDA progenitor cells. Moreover, immunoreactivities for Nurr1, a marker for immature mDA neurons, were more intense in the stripe dorsal to the subregion enriched with βIII-tubulin and TH signals (Fig. 2D,E). The stripe enriched with Nurr1⁺ cells is designated layer III. Subsequently, cells with prominent βIII-tubulin and TH signals were considered layer IV (Fig. 2J).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Laminar organization of ventral mesencephalon. In a sagittal section of mouse embryonic brain (E12), the VM may be divided into four layers along the dorsoventral axis based on the restricted expression of various known cell markers. A, C, In situ hybridization shows that Sox2 is expressed primarily in layers I and II, while Ngn2 mRNA is more abundant in layer II than layer I. B, Ki67 and βIII-Tul protein are expressed predominantly in layers I and IV, respectively. D–P, Nurr1 and Ebf are expressed in layer III and Pitx3, TH, and βIII-Tul are predominantly expressed in layer IV. M–P are enlarged views of the corresponding stippled boxes indicated in J–L. Double immunostaining revealed that βIII-Tul is expressed at lower levels in layer III but at higher levels in layer IV. TH and VMAT2 have expression patterns similar to βIII-Tul. L is a merged photo of J and K; O is a merged photo of M and N. Arrows in J indicate layers III/IV and layer IV/mesencephalic marginal strip boundaries. Scale bars, 100 μm.

We found that early B-cell factor (Ebf) protein (revealed by pan-Ebf antibody that recognizes Ebf1–3) was mainly expressed in layers III and IV, with higher level in layer III (Fig. 2F,G). However, Ebf1–3 mRNA was mainly expressed in layer III (Fig. 3 K,L; see Fig. 5A), in accordance with the findings that Ebf protein is transiently expressed in early postmitotic mDA neurons (Thuret et al., 2004). This suggests that Ebf⁺ cells are mainly immature mDA neurons. These layer III cells showed lower levels of TH, Pitx3, and βIII-tubulin, although these proteins are dominantly expressed in mature mDA neurons (layer IV) (Fig. 2J,M). Thus, we delineated a four-stage process for mDA neuron development in the VM. These data support the hypothesis that mesencephalic cells undergo continuous differentiation while they migrate ventrally toward the marginal zone. The position of migrating mDA neurons corresponds to the developmental stage of mDA cells that express a specific set of genes. An illustration of this process is shown in Figure 3R.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Regional specific expression of genes identified from the microarray analysis of the VM (A–Q). In situ hybridization in sagittal sections of the mouse embryonic brain (E12) shows restricted expression of 17 genes selected in the distinct layers of VM. Dashed lines indicate the position of the VM. Scales indicate the four-layer laminar structure in the VM. A–E, Ttyh1 and Fabp7 are expressed in layer I, while Hmgb2, Id3, and Ccnb2 are predominantly expressed in layers I and II. F, G, Ccnd2 and Cdc2a show restricted expression in layer II. H–J, Stmn2, Lrrn3, and Dll1 are expressed in layer II/III cells. K–N, Ebf2, Ebf3, Klhl1, and Cdk5r1 are specifically expressed in layer III. O–Q, Nurr1, Chl1, and Foxp2 mRNA are detected in layers III and IV. The prototypical expression profile of each gene is shown in the right panel. The data represent means ± SEM of three independent experiments. Asterisks indicate statistical significance between two adjacent time points (p < 0.05). Scale bar, 100 μm. R, Schematic representation of a potential model for the generation of mDA neurons from neuroepithelial cells. The model shows a four-layer laminar organization in the VM (left), which corresponds to a four-stage mDA neuron development (right). Some genes that are expressed by developing mDA cells, identified by microarray analysis, are indicated at right.

Characterization of the spatial expression pattern of selected VM-enriched genes

Having defined the “migration–differentiation” model for mDA neuron differentiation in the highly hierarchical structure of the VM, we performed large-scale ISH to validate the spatial expression of 186 genes (162 upregulated and 24 downregulated genes) selected from 1285 differentially expressed genes from the microarray data. The ISH results of 37 representative genes including those described in Figure 1 (bold) are presented in Figures 3 and 5 and supplemental Figure S2 (available at www.jneurosci.org as supplemental material). The VM-enriched genes identified included transcription activators and repressors (e.g., Ebf1–3, Foxp2), cell adhesion molecule close homolog of L1 (Chl1), intracellular signaling molecule Stmn2, transmembrane receptors (e.g., Lrrn3, Dll1), cytoskeleton protein Kelch-like protein 1 (Klhl1), and cyclin-dependent kinase 5 regulatory subunit 1 (Cdk5r1). These have not all been studied in the VM development.

All of these genes were highly expressed in the VM (Fig. 3; see Fig. 5; supplemental Fig. S2, available at www.jneurosci.org as supplemental material), confirming and extending the microarray results. Their hybridization signals were restrictively distributed in the VM with distinct patterns. Fatty acid binding protein 7 (Fabp7) and Tweety homologue 1 (Ttyh1) hybridization signals were detected specifically in the layer I (Fig. 3A,B). In contrast, Ccnd2 and Cdc2a hybridization signals were highly enriched in layer II, compared with Ngn2 (Figs. 2C, 3F,G). Some gene transcripts, such as Hmgb2, Id3, and Ccnb2, were evenly distributed in both layers I and II (Fig. 3C–E). Moreover, the mRNA expression of Ebf2, Ebf3, Klhl1, and Cdk5r1 was primarily restricted to layer III cells (Fig. 3K–N). The best-described mDA neuron marker, Nurr1, was also localized in this subregion (Fig. 3O). Similar to those genes expressed in adjacent layers, expression of Stmn2, Lrrn3, and DII1 was detected in both layers II and III (Fig. 3H–J), while Chl1 and Foxp2 were expressed in layers III and IV (Fig. 3P,Q). Thus the present study identified a set of VM-enriched genes expressed in distinct pattern, as summarized in Figure 3R (right). These data support the notion that the stepwise DA neuron developmental process is controlled by the combination of a large group of less-restricted genes with a small set of VM-enriched genes. Region-specific expression of the VM-enriched genes identified implies their specific roles in distinct stage of VM development.

Determination of the genetic pathways of identified VM-enriched genes

To further investigate whether the VM restricted genes identified were truly correlated with mDA neuron development, we used two gene knock-out mouse models: Ngn2^−/− and Lmx1b^−/−. Both genes have been reported to be vital for mDA neuron development in different manners (Smidt and Burbach, 2007). Consistent with previous reports (Kele et al., 2006), we found that in addition to abrogated production of Nurr1⁺ and TH⁺ mDA neurons in the VM of E12 Ngn2^−/− mutants (Fig. 4A,A′; supplemental Fig. S3a,a′, available at www.jneurosci.org as supplemental material), large number of cells accumulated in layers I–III rather than migrating to layer IV. This was evidenced by the fact that the density of Lmx1a⁺ or Lmx1b⁺ cells, indicating a laminar structure in the VM, was increased remarkably in layers I–III, in particular layer III, with an absence in layer IV compared with wild-type mice (Fig. 4B–C′; supplemental Fig. S3c,c′, available at www.jneurosci.org as supplemental material). These data are in accordance with a previous report (Andersson et al., 2006) and support the hypothesis that Ngn2 controls the differentiation and migration of mDA neurons, indicating that Ngn2^−/− mutants are thus very useful for determining potential genetic associations between Ngn2 and selected genes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

VM-enriched genes Ebf, Foxp2, Pbx3, and Npas3 are regulated by Ngn2. Expression of these genes in the VM was determined in E12 Ngn2 and Lmx1b null mutant mice. Established mDA neuron markers Nurr1, Lmx1a, Lmx1b, and AADC served as positive controls. A, A′, Deficiency of Ngn2 results in depletion of Nurr1. B–C′, Double immunofluorescence shows that Lmx1a⁺ or Lmx1b⁺ cells accumulate in layers I–III, suggesting that their ventral migration is arrested in the absence of Ngn2. D–G′, Immunostaining (D, D′) or in situ hybridization (E–G′) shows that Ebf, Foxp2, Pbx3, and Npas3 are not expressed in the absence of Ngn2, whereas their expression is normal in Lmx1b^−/− mutants (J–K′). H–N′, Lmx1b depletion does not alter Nurr1, Ngn2, Ebf, Foxp2, and Npas3 expression, whereas AADC expression disappears. Note that in Lmx1b^−/− null mutants, the mesodiencephalic structure is shrunken and severely distorted (right). The sagittal views of the brain sections are not always well matched between mutants and controls. Scale bars, 200 μm. Data represent at least three independent experiments. Dotted lines indicate diencephalon/mesencephalon and mesencephalon/metencephalon boundaries. O, Downstream molecules that are regulated by Ngn2 and Lmx1b are shown.

Using Ngn2^−/− mutant mice, we examined the expression of several transcription activators and repressors (Ebf, Foxp2, Pbx3, Npas3, Fabp7, Hmgb2, Id3, and Klf7) that we identified as VM-enriched genes in the present study. We found that Ebf immunosignals were absent in Ngn2^−/− mutants at E12 (Fig. 4D,D′). ISH analysis showed that Ebf1 mRNA was also undetectable in these mutants (data not shown). Conversely, expression of Ebf protein (Fig. 4K,K′) and Ebf1–3 mRNA (supplemental Fig. S3B, available at www.jneurosci.org as supplemental material) remained unchanged in Lmx1b^−/− mutants. Additionally, Foxp2, Pbx3, and Npas3 expression was absent in the VM of Ngn2^−/− mutants (Fig. 4E–G′). In contrast, expression of Fabp7, Hmgb2 (supplemental Fig. S3A, available at www.jneurosci.org as supplemental material), Id3, and Klf7 (data not shown) in the VM was not markedly altered in the absence of Ngn2 activity. Moreover, Lmx1b depletion did not significantly affect expression of any of these transcription activators and repressors in E12 VM (Fig. 4, right; supplemental Fig. S3B, available at www.jneurosci.org as supplemental material), despite the fact that aromatic l-amino acid decarboxylase (AADC) (Fig. 4I,I′) and Pitx3 (supplemental Fig. S3A, available at www.jneurosci.org as supplemental material) expression was diminished. Together, these results suggest that Ebf, Foxp2, Pbx3, and Npas3 are likely to be involved in mDA neuron development under the control of Ngn2. A sketch of genetic pathways is shown in Figure 4O.

Together, these data demonstrated that a set of VM-enriched genes with specific spatial and temporal expression characteristics was identified. It is likely that these genes play important roles in the respective phases of the differentiation and migration of mDA neurons.

Ebf1 is expressed primarily in immature mDA neurons

To begin to understand the roles of identified VM-enriched genes, we selected Ebf1 for further study, because Ebf1 was highly expressed in the VM with a peak level at E12 (Fig. 1), when mDA neural progenitors withdraw from mitosis and start to differentiate/migrate to establish nigrostriatal projections. These results suggest that Ebf1 may have an important and novel function in these processes. To test this hypothesis, we characterized more precisely the expression pattern of Ebf1 in the VM. ISH analysis showed that Ebf1 mRNA was predominantly expressed in layer III of E12 VM, with much low level in layer IV (Fig. 5A–D). In contrast, TH and vesicle monoamine transporter 2 (VMAT2) were expressed in both layers III and IV with higher levels in layer IV (Fig. 5B,C). Combined Ebf1 ISH with TH immunostaining revealed that some Ebf1⁺ neurons in layer III expressed TH at E13 (Fig. 5D,E). A few Ngn2⁺ cells also coexpressed Ebf (data not shown). These results indicate that Ebf1 is highly expressed in immature mDA neurons, compatible with the four-layer laminar model shown in Figure 3R. At E14 and E15, Ebf immunosignals were colocalized with TH in the large number of mDA neurons in the SN primordium (Fig. 5F,G,I). Indeed, Ebf1 hybridization signals were also detected in the SN primordium at E15 (Fig. 5H) but not at postnatal day 0 (P0) (data not shown). Thus, the period of transient Ebf1 expression in embryonic mDA neurons coincided with the early period of SN formation. Thus, Ebf1 is not required for the control of cell proliferation and mDA neuron generation in the VM.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Expression of Ebf1 in the VM. A–E, In situ hybridization reveals that Ebf1 is expressed specifically in layer III (A), whereas TH and VMAT2 mRNA are predominantly expressed in the layer IV of E12 VM in a parasagittal section (B, C). D, E, Prominent Ebf1 hybridization signals (purple) are detected primarily in layer III at E13. Combined in situ hybridization and immunostaining show that Ebf1 hybridization signals (purple) are colocalized with TH immunoreactivity (brown) in some of layer III cells. III, Layer III; IV, upper part of layer IV. E is a magnified image of the stippled box shown in D. Arrows indicate double-labeled mDA neurons in layer III. F, G, Double-labeled immunofluorescence photomicrographs showing coexpression of Ebf and TH (indicated with arrows) in E14 embryos in a coronal section. H, I, In a coronal section of an E15 mouse embryo, Ebf1 hybridization signals are detected in both the SN and VTA (H). Double-labeled immunofluorescence shows that at E15 when the SN is being formed, a few TH⁺ neurons migrating toward the SN express Ebf protein (I). Aq, Aqueduct. Scale bars, 100 μm.

Impaired formation of the substantia nigra pars compacta in Ebf1^−/− mutants

Surprisingly, we found that Ebf1^−/− mice exhibited severe impairment in the formation of the substantia nigra pars compacta (SNc) during development. At E15, the number of TH⁺ cells migrating diagonally upward was remarkably reduced in the Ebf1^−/− mutants. The deficiency was persistent even at later stages E16 and P0 (Fig. 6A–H). In contrast, the SNc primordium appeared normal at E13 and E14 in Ebf1^−/− mutants (data not shown). At P0, the number of TH⁺ cells in the SNc was only 54% of the control (mutants: 499.3 ± 11.4 vs control: 919.7 ± 40.1, n = 3 for each, p < 0.05). Remarkably, only 13% of TH⁺ cells migrated to the lateral part of the SN (SNL) compared with the control (mutants: 11.5 ± 3.1 vs control: 121.0 ± 8.9, n = 3 for each, p < 0.05) (Fig. 6I–L), although the size and shape of the P0 Ebf1^−/− mutant brain were not markedly altered compared with wild-type mice, as illustrated in Figure 7, A and B. The disorganization in the SN remained in the adult; the distribution of mDA neurons was severely perturbed with a complete absence of TH⁺ neurons in the SNL and a scrambled structure in the substantia nigra pars reticulata (SNr) (Fig. 6M–P). Interestingly, the deficiency was observed in both the rostral and intermediate parts of the SNc (Fig. 6), from which the DA neurons send their projections to the striatum constituting the mesostriatal pathway. In contrast, TH expression in the caudal part of the SNc was not markedly altered in either embryos or neonates (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Deficiency in the formation of the SNc in Ebf1^−/− mutants. The distribution patterns of mesencephalic DA neurons between Ebf1^−/− mutants and wild-type mice are compared at various stages during development. Immunohistochemistry shows that TH expression pattern is altered in the SNc of embryonic and postnatal Ebf1^−/− brain (A–L). In situ hybridization analysis shows that Ebf1 depletion results in a dramatic loss of TH⁺ neurons in the SNc of adult Ebf1^−/− mice compared with wild-type mice (M–P). Arrows indicate the location of the SN. Arrowheads indicate the medial terminal nucleus of the accessory optic tract (MTN), which separates the SN and VTA. Q and R show the distribution pattern of Pitx3⁺ cells in the Ebf1^−/− mutants and wild-type mice at P0. Scale bars, 200 μm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

A, B, Schematic representations of coronal views of the basal forebrain in wild-type and Ebf1^−/− mutants, respectively, showing the location of DA neurons in the SNc (black dots in brown colored areas) and their axonal projections (mesostriatal pathway in red) toward the striatal patches (dark brown) in the caudate–putamen nucleus (CPu, blue). The DA neurons in the VTA send their projections to the nucleus accumbens (Acb), constituting the mesolimbic pathway (yellow). A and B were adopted with permission from Yue et al. (1999) with modifications. C, D, Calbindin immunostaining shows no marked changes in calbindin⁺ cells in the VTA. Mispositioned mDA neurons express Girk2 at P30 in the absence of Ebf1. E, F, TH immunostaining reveals that the density of TH⁺ terminals in the CPu, but not in the Acb, was markedly reduced in Ebf1^−/− mutants at E16 compared with wild-type mice. G and H are magnified fields indicated by stippled boxes in E and F. Scale bars, 200 μm.

To determine whether Ebf1 might act as a postmitotic selector of SN neuronal identity and therefore SN neurons might change their identity and acquire that of VTA neurons in the absence of Ebf1, we analyzed the expression of markers preferentially expressed in the SN or VTA of young DA neurons, such as Girk2 (for SN) and calbindin (for VTA), to assess whether there is a change in the proportion of DA neurons immunoreactive for these markers between mutant and control mice. The number of calbindin⁺ neurons in the VTA was not significantly changed in Ebf1-null mice at P0 (control: 1026 ± 30.56 cells vs Ebf1^−/− mutants 1050 ± 53.69 cells, 4 wild-type and 4 mutants, p > 0.05) (Fig. 7C,D). Moreover, those DA neurons that failed to migrate to the SNc were Girk2 positive (data not shown). These data suggest that Ebf1 is specifically required for the development of nigral but not VTA DA neurons.

To determine whether the disorganized SN in Ebf1^−/− mutants is a secondary effect of a distorted brain following Ebf1 depletion, we compared the VM architecture of Ebf1^−/− mice with that of wild-type littermates. The entire VM laminar structure labeled with Lmx1a antibody in E12 Ebf1^−/− mice was normal (Fig. 8A). Similarly, at P0, the general structural arrangement of nuclei in the midbrain, revealed by Nissl staining, was not markedly altered between mutants and controls (data not shown). Oculomotor nuclei that are located laterally to the SN, representing a non-DA brain structure adjacent to the SN, were largely identical between mutant and wild-type mice, as revealed by Isl1 immunostaining in P0 Ebf1^−/− mutants (data not shown). Together, these data suggest that Ebf1 depletion in embryos and neonates does not lead to substantial disorganization of neuronal lamination in the VM or the neighboring brain structure.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Both radial glial fibers and tangential fibers are well conserved in the absence of Ebf1. A, Photomicrographs showing that the radial glial fibers stained with nestin antibody and laminar organization revealed by Lmx1a immunofluorescence are well formed in E12 Ebf1^−/− mutants (n = 2). c and d are magnified images of the stippled boxes in a and b. I–IV indicate layers I–IV. B, The distribution profile of tangential fibers stained with L1 antibody in the medial (a–d) and lateral (e–h) parts of the VM of E15 Ebf1^−/− mutant embryos is very similar to that in wild-type mice (n = 3). b, d, f, and h are magnified views of the stippled boxes indicated in a, c, e, and g, respectively. Arrows indicate tangential fibers in green. Arrowheads indicate TH⁺ neurons in red. Scale bars, 150 μm.

Lack of mDA neurons in the SNc of Ebf1^−/− mutants results from impaired diagonal upward migration from the VTA

We then asked whether reduction of mDA neurons in the Ebf1^−/− mutants was due to loss of TH expression or apoptosis of mDA neurons in the SNc. Pitx3 immunostaining, which specifically labels mDA neurons (Nunes et al., 2003), was used to visualize these cells. The advantage of this staining is that it allows visualization of highly compact mDA neurons individually, thus facilitating their quantification. The number of Pitx3⁺ cells was reduced to 53.7% in the SNc of Ebf1^−/− mutants compared with controls at P0 (mutants, 618.24 ± 64.2 vs control: 1151.0 ± 55.1, n = 4 for each, p < 0.05). In contrast, the number of Pitx3⁺ cells in the VTA was slightly increased (mutant: 4052 ± 279.0 vs control: 3656.75 ± 296.0, n = 4), although there was no change in the total number of Pitx3⁺ cells in the VM (including the SN, VTA, and those along the midline) (mutant: 4669.0 ± 274.5 vs control: 4745.0 ± 409.4, n = 4). The terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick-end labeling (TUNEL) staining showed no alteration in apoptotic cells in either Ebf1^−/− mutants or wild-type embryos at E14, E15, E16, and P0 (data not shown). These data suggest that in Ebf1^−/− mutants, the lack of mDA neurons that are destined for the SNc does not result from a deficiency in their neurogenesis or survival.

Interestingly, the effect of Ebf1 depletion on mDA neuron development appeared to be specific to nigral DA neurons. Only the projection terminals of mDA neurons in the striatum were dramatically reduced in Ebf1^−/− mutants, whereas those in the accumbens nucleus were not altered (Fig. 7E,H). This is consistent with the observation that the number of Pitx3⁺ neurons in homozygous mutant mice was modified in the SNc but not the VTA.

Both radial and tangential fibers are unaffected by the absence of Ebf1

To gain insight into the defective migration of mDA neurons in Ebf1-deficient mice, we examined the structure of both radial and tangential fibers in the VM. It was observed that Nestin⁺ radial glial fibers extended from the VZ to the pial surface of the VM at E12 (Shults et al., 1990) and were arrayed in a similar pattern between Ebf1^−/− mutants and wild-type embryos (Fig. 8A). Moreover, the tangential fibers revealed by L1 antibody (Ohyama et al., 1998) extended normally in the medial part of the VM through the middle toward the lateral part in Ebf1^−/− mutants as those in wild-type mice. In the middle and lateral parts of the VM, L1-bearing fibers appeared as short and clustered structures, and the migrating TH⁺ neurons were alongside these fibers (Fig. 8B).