Conditional inactivation of Ezh2 in the developing midbrain affects progenitor cell expansion
To address the role of Ezh2-mediated H3K27me3 in the developing midbrain, we conditionally deleted Ezh2 in mice homozygous for the floxed allele of Ezh2 using the Wnt1-Cre allele (Fig. 1a) [10, 12]. Wnt1-Cre
+/Ezh2
[SET]
lox/lox conditional knock-out (Ezh2 cko) mice survive to late developmental stages, but die around E18, displaying craniofacial abnormalities and heart malformations caused by concomitant activity of Wnt1-Cre in the neural crest [12]. In the midbrain, ablation of Ezh2 was evident from E10.5 onwards (Additional file 1: Figure S1). Of note, Ezh1 expression was very low in the embryonic midbrain and, importantly, was not affected upon conditional Ezh2 inactivation (Additional file 1: Figure S1). Ezh2 loss was associated with widespread loss of H3K27me3, as shown by immunohistochemistry at E12.5 (Fig. 1b, c). Using the ROSA26 Cre reporter allele driving β-galactosidase expression, we could also show full Wnt1-Cre-mediated recombination in the caudal midbrain [13]. Histological analyses revealed a marked reduction of the neuroepithelial thickness in the midbrain of Ezh2 cko embryos at E12.5 as compared to normal embryos, which was even more pronounced at E14.5 (Fig. 1d). Furthermore, horizontal expansion of the neuroepithelium was decreased in mutant midbrains, as was most apparent in the isthmal and inferior tectal region at E12.5 (Fig. 1d).
These data are consistent with altered cell cycling of mutant neuroepithelial progenitor cells [14]. Indeed, the number of proliferative cells incorporating the thymidine analogue EdU during a 1-hour EdU pulse was significantly reduced in the developing midbrain of Ezh2 cko embryos at E12.5 as compared to control littermates (Fig. 2a). The decrease of proliferative cells in the mutant midbrain could be associated with mutant neuroepithelial cells preferentially choosing to exit rather than to remain in the cell cycle. To address this possibility, we determined the fraction of Ki67-positive dividing cells after a BrdU pulse of 24 hours. Cells that had left the cell cycle were BrdU-positive but Ki67-negative, while cells that were still in the cell cycle at the time point of analysis were both BrdU- and Ki67-positive. At E12.5, a highly significant increase of cells exiting the cell cycle was detectable in the mutant as compared to the control (Fig. 2b). Immunohistochemistry for the NPC marker Sox2 and the differentiation marker Dcx further demonstrated that decreased proliferation in the midbrain of Ezh2 cko embryos at E12.5 was accompanied by a reduction in the number of progenitor cells and a concomitant increase in differentiation (Fig. 2c). The increased neurogenesis in the Ezh2 cko midbrain was also confirmed at E14.5 (Additional file 1: Figure S2). Cell survival was impaired in the dorsal rostral midbrain of mutant embryos but unchanged in the area used for quantification of mitotic cells (Fig. 2d). Thus, Ezh2 is essential for proper midbrain formation by controlling the pool size of NPCs.
Ezh2 controls proliferation of neural progenitor cells by repressing cell cycle regulators and inhibitors of Wnt signaling
To identify the molecular mechanisms mediating Ezh2-dependent midbrain development, we used microarray analysis to compare the global gene expression patterns of control versus Ezh2 cko cells isolated from the dorsal midbrain of E10.5 embryos. Cluster analysis of the transcriptome data indicated that the vast majority of differentially expressed genes were transcriptionally upregulated upon loss of Ezh2 (Fig. 3a). This is consistent with the role of Ezh2 as a transcriptional repressor [15]. Gene ontology analysis of process networks revealed that differentially expressed genes were involved, among others, in negative regulation of proliferation, Wnt signaling, and cell cycle regulation (Additional file 1: Figure S3). Since misregulation of those processes very likely contributes to the described mutant phenotype, we focused our analysis on the aforementioned process networks. Among the genes derepressed in Ezh2 cko cells, were the cyclin-dependent kinase inhibitors (Cdkn) 2a and 2c, which negatively regulate cellular proliferation [16–19]. Increased expression of Cdkn2a and Cdkn2c was also demonstrated by quantitative RT-PCR performed on midbrain cells isolated from E11.5 embryos (Fig. 3b). Moreover, in situ hybridization on sagittal sections of E12.5 control and mutant midbrains revealed the specific increase in expression of the cell cycle inhibitor Cdkn2a in Ezh2 cko embryos (Fig. 3d). Finally, we performed an H3K27me3 ChIP assay on wildtype E11.5 midbrain cells and revealed that the promoters of Cdkn2a and Cdkn2c were occupied by H3K27me3. Thus, these cell cycle inhibitors appear to be direct targets of Ezh2-mediated epigenetic repression (Fig. 3c).
In addition, we also found other potentially relevant genes to be differentially expressed upon loss of Ezh2. In particular, inhibitors of the Wnt signaling pathway, such as Wif1 and Dkk2 [20, 21], were also significantly upregulated in Ezh2 cko cells (Fig. 3a,b,e). H3K27me3 ChIP analysis confirmed that these Wnt signaling inhibitors appear also to be direct targets of Ezh2 activity (Fig. 3c). Canonical Wnt signaling has been demonstrated to control maintenance of midbrain neuroepithelial cells [22, 23]. Therefore, we investigated whether Wnt signal activity is indeed affected by loss of Ezh2. To this end, we made use of the BAT-gal Wnt signaling reporter allele, which monitors β-catenin activity by driving β-galactosidase expression in Wnt signaling-active cells [24]. In mice harboring this reporter allele, we observed a prominent reduction in the number of β-galactosidase-positive neural cells in the Ezh2 cko midbrain at E12.5, as compared to the control (Fig. 3f). Accordingly, targets of Wnt signaling, such as CyclinD1 and Lef1, were downregulated in the mutant midbrain, as shown by immunohistochemistry and quantitative RT-PCR, respectively (Fig. 3g,h).
Our data indicate that reduced canonical Wnt signaling might also contribute to the phenotype of Ezh2 cko mice. To better understand the role of canonical Wnt signaling in regulating midbrain size, we took advantage of a mutant allele of the Wnt signaling component β-catenin (Ctnnb1
dm/flox referred to as Ctnnb1 sign mt) that disrupts Wnt/β-catenin-mediated transcriptional output but not cell-cell adhesion [22, 25]. While total loss of β-catenin leads to disintegration of the midbrain [22], loss of β-catenin signaling function did not affect the integrity of the neuroepithelium. However, very similar to Ezh2 cko embryos (Fig. 1), the thickness and overall size of the midbrain was drastically reduced in Wnt1-Cre/Ctnnb1 sign mt cko mice at E12.5 (Fig. 3i). Therefore, Ezh2 appears to regulate the size of the developing midbrain both by direct repression of cell cycle inhibitors and, indirectly, by sustaining β-catenin signaling.
Ezh2 represses forebrain identity in the developing midbrain
Intriguingly, the microarray analysis of control and Ezh2 cko midbrain pointed to an additional set of Ezh2-regulated genes that are known to exhibit brain area-specific, rather than general cellular functions in the developing neuroepithelium. Notably, several forebrain specification genes were derepressed in Ezh2 cko midbrains (Fig. 3a). In situ hybridization experiments, immunohistochemistry, and quantitative RT-PCR experiments were used to corroborate this finding. While normally Foxg1 is strongly expressed in the forebrain but absent in the midbrain, it was upregulated in the midbrain of Ezh2 cko embryos at E12.5 (Fig. 4a). A quantitative analysis at E11.5 revealed a more than 75-fold induction of Foxg1 expression in the mutant midbrain (Fig. 4e). Likewise, the midbrain is normally devoid of Pax6 expression, whereas upon loss of Ezh2, Pax6 became broadly expressed in the midbrain, displaying a 22-fold induction at E11.5 (Fig. 4b,e). Furthermore, the forebrain markers Dlx2 and Emx1 were ectopically expressed in the midbrain of Ezh2 cko embryos at E11.5 (Fig. 4d,e). However, when comparing mRNA levels of Ezh2, Pax6, Foxg1 and Emx1 in wildtype forebrain, wildtype midbrain, and Ezh2 cko midbrain of E12.5 embryos it became apparent that expression levels of ectopic forebrain markers in the mutant midbrain did not reach those of the forebrain (Fig. 4a; Additional file 1: Figure S4).
In most rostral regions of the dorsal midbrain, Ezh2-dependent H3K27me3 was only partially depleted in Ezh2 cko embryos (Additional file 1: Figure S5). Incomplete Wnt1-Cre-mediated recombination was shown by tracking of recombined cells using the aforementioned ROSA26 Cre reporter allele (R26R) [13]. Therefore, non-recombined cells were intermingled with clusters of Ezh2-deficient cells in the rostral midbrain of Ezh2 cko embryos at E12.5 (Additional file 1: Figure S5). Strikingly, in this area, Ezh2 exhibited a perfectly inverse relationship with Foxg1 and Pax6 expression patterns, respectively, pointing to cell-autonomous mechanisms underlying the gain of forebrain markers in Ezh2 cko midbrain cells (Fig. 4a,b). In support of this, H3K27me3 ChIP experiments performed with midbrain cells from control embryos at E11.5 demonstrated that the forebrain specification genes Foxg1 and Pax6 appear to be direct targets of Ezh2-mediated repression (Fig. 4c).
In the developing forebrain, Pax6 acts upstream of the transcription factor Dmrta1, which itself regulates the expression of the proneural gene Neurog2 [26]. Strikingly, Pax6 upregulation in the Ezh2 cko midbrain was associated with significant upregulation of both Dmrta1 and Neurog2 (Fig. 4f,g). Thus, although forebrain neuronal layer-specific markers could not be analyzed at later stages due to the substantial mass reduction and disturbed morphology of the mutant midbrain (Fig. 1d; data not shown), our data reveal the ectopic upregulation of a forebrain transcriptional program in the midbrain of Ezh2 cko embryos.
Ezh2 regulates midbrain identity by indirect mechanisms
Comparable to Pax6, Foxg1, Dlx2, and Emx1 in the developing forebrain, the transcription factors Pax3 and Pax7 have been shown to establish midbrain identity during vertebrate brain development [27, 28]. To address whether expression of these midbrain specification factors was also affected by loss of Ezh2, we performed quantitative RT-PCR and immunohistochemistry. While expression of Pax3 and Pax7 was unchanged at E11.5 as shown by qPCR, it was significantly downregulated at E12.5 (Fig. 5a). Consistent with these results, immunohistochemistry confirmed the presence of Pax3 at E11.5 (Additional file 1: Figure S6C) and the highly reduced expression of both transcription factors at E12.5 in Ezh2 cko. Indeed, whereas Pax3 and Pax7 were detected in the entire dorsal midbrain neuroepithelium in control embryos, many cells in the mutant dorsal midbrain were devoid of Pax3 and Pax7 or showed reduced staining intensity (Fig. 5b,c). Thus, Ezh2-mediated H3K27me3 is required for proper expression of midbrain specification genes.
Our study identified the forebrain specification genes Foxg1 and Pax6 as targets of Ezh2 activity, which is in agreement with their increased expression in the Ezh2 cko midbrain (Fig. 4). In contrast, the loss of midbrain identity markers in Ezh2 cko embryos cannot be explained by direct Ezh2-mediated repression. In chicken embryos, overexpression of Pax6 has been reported to indirectly repress Pax3 and Pax7 expression in the trigeminal placode and at the forebrain-midbrain boundary, respectively [29, 30]. However, while we found Pax6 to be strongly upregulated in the Ezh2 cko midbrain already at E11.5 (Fig. 4e), Pax3 and Pax7 were downregulated at E12.5 only (Fig. 5a), rather arguing against control of the midbrain specification factors by Pax6. To directly address this hypothesis, we performed in utero electroporation of a Pax6-overexpression vector together with a GFP expression vector. In parallel we electroporated the GFP expressing vector alone as a control. Monitoring GFP expression two days after in utero electroporation revealed the efficient targeting of the murine dorsal midbrain by this method (Additional file 1: Figure S7A). Coronal sections of electroporated midbrains were then used to quantify the number of Pax3- and Pax7-expressing cells per GFP-positive cells by immunofluorescence. For each condition, the midbrains of three embryos were electroporated and more than 800 cells were analyzed (Additional file 1: Figure S7B). However, as shown in Fig. 6a and Additional file 1: Figure S7C, ectopic expression of Pax6 did not influence Pax3 and Pax7 expression in dorsal midbrain cells. Hence, increased Pax6 expression is apparently unable to repress Pax3 and Pax7 in the established murine midbrain and is, therefore, unlikely the cause for downregulated expression of midbrain fate determinants in Ezh2 cko embryos.
Previously, Wnt/β-catenin signaling was shown to activate Pax3 and Pax7 expression in the lateral neural plate and during neural tube closure [31–33]. Therefore, reduced Pax3/Pax7 expression in the Ezh2 cko midbrain might be due to decreased canonical Wnt signaling in mutant brain tissue (Fig. 3). To investigate whether Wnt/β-catenin is required for expression of Pax3 and Pax7 in the dorsal midbrain, we performed immunohistochemistry on sagittal sections of Wnt1-Cre/Ctnnb1 sign mt cko embryos at E12.5. Loss of β-catenin signaling not only affected midbrain size, but also resulted in drastically reduced expression of both Pax3 and Pax7 (Fig. 6b,c). In fact, the midbrains of Wnt1-Cre/Ctnnb1 sign mt cko embryos displayed a phenotype very comparable to the one of Ezh2 cko embryos (Fig. 5), with many mutant cells lacking Pax3 and Pax7 expression. Thus, the loss of midbrain identity markers in the Ezh2 cko midbrain is apparently caused by indirect mechanisms, involving Ezh2-mediated control of canonical Wnt signaling.