The Timing of Upper-Layer Neurogenesis Is Conferred by Sequential Derepression and Negative Feedback from Deep-Layer Neurons

The sequence of DL and UL neurogenesis is fixed regardless of the timing of projection neuron induction in vivo

To investigate the emergence of UL neurons, we first assessed the temporal window in which UL projection neurons can be induced in vivo. We established previously a transgenic mouse line in which the onset of neocortical projection neuron production can be synchronously manipulated using the expression of a forkhead transcription factor (TF), Foxg1 (Kumamoto et al., 2013). In this mouse line, endogenous Foxg1 sequences are replaced with a lacZ coding sequence and a tet-transactivator (tTA) coding sequence (Foxg1^tTA/lacZ knock-in null mice), which drives Foxg1 expression under the regulation of the tetO minimal promoter of an independent tetOFoxg1 transgenic allele (Foxg1^tetOFoxg1 mice; Fig. 1A). Foxg1 expression is suppressed in the presence of Dox, whereas Dox removal results in Foxg1 induction (Tet-off system). In Foxg1^tetOFoxg1 mice, we demonstrated previously that the delayed Foxg1 induction by Dox withdrawal at E14.5 is sufficient to switch to DL neurogenesis after a prolonged period of earliest-born Reelin⁺ Cajal–Retzius (CR) neuron production (Kumamoto et al., 2013). Therefore, we induced Foxg1 at progressively later stages of corticogenesis, at E14.5, E15.5 and E16.5, to assess the temporal window in which UL projection neurons can be induced in vivo (Fig. 1B). To circumvent the early developmental defects that result from the loss of Foxg1, Dox was administered starting at E9.5 (Fig. 1B), when the size differences between the Foxg1 heterozygote and homozygote cortices are minimal (Xuan et al., 1995). To label the temporal cohorts of neurons born after Foxg1 induction, we injected pulses of BrdU and EdU at a 24 h interval and monitored layer-specific subtype production at E18.5 (Fig. 1B). To assess the relative proportions of neurons that were produced at each time point, BrdU- or EdU-labeled cells that were colabeled with Ctip2 (DL neuron marker) or Satb2 or Brn2 (UL neuron markers) of the total BrdU- or EdU-labeled cells were quantified per 100 μm strip of dorsolateral cortices through the VZ to the marginal zone (MZ; Fig. 1C–AM). To obtain accurate birthdating, the nuclei that contained full BrdU/EdU content were quantified to eliminate cells that underwent more than one cell division after BrdU/EdU incorporation (see Materials and Methods). Consistent with previous results (Kumamoto et al., 2013), when Foxg1 was induced at E14.5, we observed a shift in the production of Ctip2⁺ DL neurons that were colabeled primarily with BrdU (Foxg1^tetOFoxg1 cortex, 72.6 ± 2.6% Ctip2⁺/E14.5 BrdU⁺ cells; control cortex, 5.9 ± 0.7% Ctip2⁺/E14.5 BrdU⁺ cells; Fig. 1C–D′,I–J′,AM). Additional analyses that used sequential E14.5 BrdU and E15.5 EdU labeling indicated that Ctip2⁺ neuron production preceded Satb2⁺ or Brn2⁺ UL neuron generation (Foxg1^tetOFoxg1 cortex, 22.1 ± 4.5% Brn2⁺/E14.5 BrdU⁺ cells, 75.2 ± 6.8% Brn2⁺/E15.5 EdU⁺ cells, 15.1 ± 2.5% Satb2⁺/E14.5 BrdU⁺ cells, and 54.6 ± 12.6% Satb2⁺/E15.5 EdU⁺ cells; control cortex, 88.1 ± 8.1% Brn2⁺/E14.5 BrdU⁺ cells, 18.8 ± 7.1% Brn2⁺/E15.5 EdU⁺ cells, 32.0 ± 5.3% Satb2⁺/E14.5 BrdU⁺ cells, and 16.2 ± 3.2% Satb2⁺/E15.5 EdU⁺ cells; Fig. 1E–H′,K–N′,AM). Notably, although both BrdU- and EdU-labeled cells in the control cortex migrated into the CP (Fig. 1C–H′), the majority of BrdU⁺ and EdU⁺ cells in the Foxg1^tetOFoxg1 cortex failed to integrate into the preceding neuronal layer and accumulated beneath the overlying CP. This overlying CP in the Foxg1^tetOFoxg1 cortex consists of supernumerary CR cells that were generated during the period of Foxg1 suppression (data not shown and Kumamoto et al., 2013) and thus resemble the migration termination of projection neurons beneath the CR cell containing MZ during normal development (Fig. 1I–N′).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

The sequence of DL and UL neurogenesis is fixed regardless of the timing of projection neuron induction in vivo. A, Schematic diagram of the Foxg1^tetOFoxg1 mice. Foxg1 transgene expression was initially repressed by Dox administration, and Foxg1 expression was induced by Dox withdrawal. B, Experimental design. After Dox administration at E9.5, Foxg1 was induced by Dox withdrawal at the indicated embryonic stages, and the dams were subsequently administered pulses of BrdU (red arrowheads) and EdU (blue arrowheads) at the specified time points. All embryos were analyzed at E18.5 (green arrowheads). C–Z, AA–AL, Double immunohistochemistry for the indicated markers in coronal sections from E18.5 Foxg1^tTA/+ control and Foxg1^tetOFoxg1 mice. C′–Z′, AA′–AL′, Enlarged views of the boxed regions shown in C–Z and AA–AL. The arrowheads indicate cells that were double labeled with respective markers (Ctip2, Satb2, or Brn2 with BrdU or EdU). Asterisks indicate nonspecific background staining of blood vessels by mouse monoclonal antibody. AM, Quantitative analysis of the percentage of BrdU⁺ and EdU⁺ cells that expressed Ctip2, Satb2, or Brn2 in the indicated genotypes. Cells in the entire 100 μm radial strips (as represented in C–Z, AA–AL) from the dorsolateral cortices, including VZ cells, were counted. The results are expressed as the mean ± SEM. **p < 0.01. Scale bars: C′–Z′, AA′–AL′, 20 μm.

Next, we performed the same analysis in a 1-d-later shifted experiment, in which we induced Foxg1 at E15.5 by Dox removal (Fig. 1B, middle row). We determined that E15.5 Foxg1 induction was also sufficient to induce both DL and UL subtypes (Fig. 1U–Z′,AM). E15.5 BrdU and E16.5 EdU injections revealed that Satb2⁺ and Brn2⁺ neurons were also generated later compared with Ctip2⁺ neurons in these mutants; Ctip2⁺ DL neurons were predominately generated at E15.5 (37.3 ± 6.8% Ctip2⁺/E15.5 BrdU⁺ cells and 0% Ctip2⁺/E16.5 EdU⁺ cells; Fig. 1U–V′,AM), and Satb2⁺ and Brn2⁺ cells were generated at E16.5 (16.7 ± 5.0% Satb2⁺/E15.5 BrdU⁺ cells, 34.6 ± 0.6% Satb2⁺/E16.5 EdU⁺ cells, 27.7 ± 3.6% Brn2⁺/E15.5 BrdU⁺ cells and 35.4 ± 2.9% Brn2⁺/E16.5 EdU⁺ cells, respectively; Fig. 1W–Z′,AM). Together, these results suggested that UL progenitors could not bypass DL competence for their production even at the latest period of corticogenesis. This finding was consistent with the results obtained when we further induced Foxg1 1 d later at E16.5 by Dox removal (Fig. 1B, bottom row). We observed that Ctip2⁺ cells were not detectable in the cortex (Fig. 1AG–AH′), and Satb2⁺ and Brn2⁺ UL neurons were absent (Fig. 1AI–AL′,AM). The chase period after BrdU and EdU injections (at E16.5 and E17.5, respectively) may not have provided sufficient time for these markers to be turned on in these mice; however, we identified a low proportion of BrdU- or EdU-labeled cells that expressed Satb2 and Brn2 in the control cortex but not in the Foxg1^tetOFoxg1 cortex (Fig. 1AC′–AF′, arrowheads). These data indicated that these birthdates correspond to the latest period of neurogenesis (Hevner et al., 2003) and to the onset of gliogenesis (Seuntjens et al., 2009) during normal development. The results also suggested that Foxg1 induction is sufficient to initiate sequential DL–UL competence but cannot override gliogenic competence. Consistent with these observations, the contribution of E16.5-born progenitors to Olig2⁺ glial cells was comparable between the Foxg1^tetOFoxg1 and control cortices (30.1 ± 1.5 and 19.5 ± 5.9%, respectively; data not shown), which was similar to the result of a previous study (Seuntjens et al., 2009). Collectively, these data demonstrate that cortical progenitors have a restricted competence window in vivo, such that UL competence strictly follows DL competence and is terminated by gliogenic competence.

UL progenitors arise from the Foxg1⁺ cell lineage

The conserved sequence of DL and UL competence throughout the corticogenesis period indicates the following two potential, non-mutually exclusive, mechanisms of UL neurogenesis: neocortical progenitors use intrinsic cues that induce bias toward DL competence over UL competence and/or feedback signaling from DL neurons is necessary to trigger UL neurogenesis during cortical development. These mechanisms are supported by the observations that DL neurons are generated more efficiently compared with UL neurons in vitro (Espuny-Camacho et al., 2013) and that UL neurons favor high-density aggregate conditions compared with low-density cultures for their induction (Shen et al., 2006; Eiraku et al., 2008; Gaspard et al., 2008; Kadoshima et al., 2013). To test the former mechanistic hypotheses, we induced neocortical projection neurons in a restricted number of cells by electroporating pCAGGS–Foxg1 and pCAGGS–GFP constructs into E14.5 Foxg1 constitutive knock-out (Foxg1^−/−) cortices (Fig. 2A–D″), which lack DL and UL neuron production and exhibit CR cell overproduction (Kumamoto et al., 2013). Consistent with the role of Foxg1 in switching from early CR cells to DL neurogenesis (Kumamoto et al., 2013), Ctip2⁺ DL neurons were induced in the GFP⁺ cells in the pCAGGS–Foxg1 and pCAGGS–GFP-introduced cortex (Fig. 2I). Notably, within this cluster, we also detected Satb2⁺ cells (Fig. 2J) and Brn2⁺ cells (Fig. 2R), which were mostly complementary to Ctip2⁺ cells (Fig. 2K,S). However, within Satb2⁺ neurons, GFP was either not expressed in these cells or only expressed at low levels (Fig. 2L–L‴, open arrowheads). We further assessed the expression of Foxg1 protein in Satb2⁺ cells by double immunohistochemistry of Foxg1 and Satb2 and found that Foxg1 expression was either undetectable in Satb2⁺ cells (Fig. 2AB–AB″, open arrowheads) or expressed at very low levels (Fig. 2AB–AB″, filled arrowheads). This was in contrast to Ctip2⁺ cells, which generally expressed high levels of Foxg1 protein (Fig. 2X–X″, filled arrowheads). This result suggested that, during Foxg1 induction, Satb2⁺ cells may be generated later within the Foxg1⁺ lineage; thus, the GFP and Foxg1 plasmid is diluted through multiple rounds of cell division.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

UL neurons are induced in a restricted number of cortical cells. A, Experimental design. I.U.E.P., In utero electroporation. Foxg1^−/− constitutive knock-out mice were used. B, Lower-magnification views of the cortex, which indicate representative GFP electroporated areas. Control pCAGGS–GFP electroporated cells are dispersed randomly in the CP, whereas Foxg1-introduced GFP⁺ cells are clustered. C, D, pCAGGS–GFP alone or pCAGGS–GFP and pCAGGS–Foxg1 were introduced in a restricted number of Foxg1^−/− cells. E–G, I–K, GFP, Ctip2, and Satb2 immunohistochemistry. E–G and I–K are each from the same section showing different channels. H–H‴ and L–L‴ indicate enlarged views of the sections shown in E–G and I–K. M–O, Q–S, GFP, Ctip2, and Brn2 immunohistochemistry. M–O and Q–S are from the same section showing different channels. P–P‴ and T–T‴ indicate enlarged views of the sections shown in M–O and Q–S. U–W, Y–AA, Ctip2, Satb2, and Foxg1 immunohistochemistry. U–W and Y–AA are each from the same section showing different channels. X–X″ and AB–AB″ indicate enlarged views of the sections shown in U–W and Y–AA. Filled arrowheads indicate layer marker⁺ cells that express GFP or Foxg1, and open arrowheads indicate layer marker⁺ cells that are negative for GFP or Foxg1. Scale bars: B, 100 μm; C–D″ E–G, I–K, M–O, Q–S, U–W, Y–AA, 50 μm; H-H‴, L–L‴, P–P‴, T–T‴, X–X″, AB–AB″, 10 μm.

To directly test this hypothesis, we coelectroporated pCAGGS–Foxg1 and pCAGGS–Cre constructs into Foxg1^−/−;Rosa26^{loxp–stop–loxp–YFP/+} mice to label all progeny of E14.5 Foxg1-introduced progenitors and examined the yellow fluorescent protein (YFP)⁺ cells at E18.5 (Fig. 3A–G). We identified Satb2⁺ (Fig. 3F–F″,G, arrowheads) and Ctip2⁺ (Fig. 3D–D″,E, arrowheads) expression in YFP⁺ cells but not in YFP-negative (YFP⁻) cells (Fig. 3D–D′,F–F′). Therefore, we defined the emergence of Satb2⁺ cells as originating from cells with a Foxg1 lineage after the onset of Foxg1 expression. Cell birthdating studies that used consecutive E14.5 BrdU and E15.5 EdU injections (Fig. 3H) confirmed that Satb2⁺ cell generation followed Ctip2⁺ cell generation in the same cluster (Ctip2⁺/BrdU⁺ cells, 39.4 ± 1.2%; Ctip2⁺/EdU⁺ cells, 14.7 ± 1.8%; Satb2⁺/BrdU⁺ cells, 16.0 ± 1.7%; and Satb2⁺/EdU⁺ cells, 36.5 ± 2.1%; Fig. 3I–L″,M). Notably, within these restricted neocortical neuron clones, the number of Satb2⁺ and Brn2⁺ cells did not exceed the number of Ctip2⁺ cells generated in the same cluster (Fig. 2K,S and data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

UL progenitors arise from the Foxg1⁺ cell lineage. A, Experimental design. I.U.E.P., In utero electroporation. B–G, Foxg1, YFP, Ctip2, and Satb2 immunohistochemistry images acquired by confocal microscopy. C, E, G, Higher-magnification views of the boxed regions shown in B″, D″, and F″, respectively. Arrowheads indicate neurons that are double positive for the two indicated markers. In the Rosa26^{loxp–stop–loxp–YFP/+} reporter, YFP predominantly localizes to the cytoplasm. H, Experimental design. I–L″, Double detection of BrdU or EdU with the following neuronal subtype markers: DL neurons, Ctip2; UL neurons, Satb2. Arrowheads indicate Ctip2⁺ or Satb2⁺ cells double labeled with BrdU, and open arrowheads indicate Ctip2⁺ or Satb2⁺ cells double labeled with EdU. M, Quantitative analysis of the percentage of BrdU⁺ and EdU⁺ cells that express Ctip2 or Satb2. The results are expressed as the mean ± SEM. **p < 0.01. Scale bars: B–B″, D–D″, F–F″, I–L″, 50 μm; C, E, G, 10 μm.

Foxg1 confers the sequence of DL and UL competence through Tbr1 repression within a closed transcriptional cascade

The Foxg1 lineage studies indicated that the cascade downstream of Foxg1 triggers the sequential competence of DL and UL neuron production. Next, we identified the transcriptional cascade that determines this sequence of layer neurogenesis. Previous genetic studies have indicated that the segregation between the principal layer subtypes of the cerebral cortex is established on a closed transcriptional network, in which cross-repression between the four TFs, namely, Fezf2, Ctip2, Satb2, and Tbr1, is sufficient to establish the subcerebral, intracortical, and corticothalamic projection identities within the postmitotic neurons (Alcamo et al., 2008; Britanova et al., 2008; Chen et al., 2008; Han et al., 2011; McKenna et al., 2011; Srinivasan et al., 2012). We hypothesized that triggering a sequence of neurogenesis requires a break in this equilibrium, which occurs through a derepression of one of the genes in the transcriptional loop (Fig. 4A). Thus, we examined which of these TFs respond in vivo to Foxg1 induction. We assessed previous gene expression data in which E14.5, E15.0, E15.5, and E16.5 Foxg1^tetOFoxg1 neocortical progenitors were isolated after Foxg1 induction by E14.5 Dox removal (Kumamoto et al., 2013; Fig. 4B). Using these datasets, one-way ANOVAs with post hoc tests were performed to identify differentially expressed transcripts (Fig. 4C; see Materials and Methods). According to these unbiased criteria, among the four TFs, only Tbr1 exhibited a significant response to Foxg1 induction [E16.5 Foxg1 induction/E14.5 control = −0.9445 (log₂), corresponding to −1.9245 (log₁₀)-fold change; Fig. 4C)]. These results suggest that, within the aforementioned transcription network (Fig. 4A), Foxg1 may determine the sequence of layer neurogenesis primarily through Tbr1 repression. Notably, Tbr1 has been reported to be expressed in the majority of early-born glutamatergic neurons of the cerebral cortex, including preplate CR cells and subplate (SP) neurons (Hevner et al., 2001). Therefore, we assessed Tbr1 expression regulation during normal cortical development. Immunohistochemistry was performed using Tbr1 antibodies (Abcam rabbit anti-Tbr1 in Fig. 4D–H and Millipore chicken anti-Tbr1, data not shown) and Foxg1 antibodies (Santa Cruz Biotechnology goat anti-Foxg1 in Fig. 4D–H). At E11.5, Tbr1 protein expression in preplate neurons preceded the onset of Foxg1 expression in the postmitotic neurons (Fig. 4D), which is consistent with the absence of Foxg1 expression in CR neurons (Hanashima et al., 2004). Instead, the Foxg1 protein was weakly detected in progenitor cells at this stage, in which the detection levels slightly varied between the different antibodies that were tested (Santa Cruz Biotechnology goat anti-Foxg1, Abcam rabbit anti-Foxg1, StemCulture rabbit anti-Foxg1, and rabbit anti-Foxg1 from Y. Sasai; see Materials and Methods; Fig. 4D and data not shown). At E13.5, Foxg1 expression increased in progenitor cells and postmitotic neurons, in which its expression partially overlapped with Tbr1⁺ cells in the CP (Fig. 4E). The expression levels of the Foxg1 protein were higher in postmitotic neurons compared with progenitor cells in all stages that were examined; however, we observed downregulation of the Foxg1 protein during migration (Fig. 4E,F), which is consistent with a previous report (Miyoshi and Fishell, 2012). At E15.5, Tbr1 expression was downregulated in the cells of the upper part of the CP that were Foxg1⁺, which consisted primarily of postmigratory layer V neurons (Fig. 4F). By postnatal day 4 (P4), Tbr1 and Foxg1 exhibited complementary expression patterns; Tbr1 was detected in L6 and SP within cells that were low or undetectable for the Foxg1 protein (Fig. 4G,H).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Foxg1 regulates Tbr1 transcription in vitro. A, Schematic model of identified genetic interactions between the layer-subtype TFs. B, Experimental scheme of the transcriptome analysis performed by Kumamoto et al., 2013. Temporal cohort of CD133⁺ cells isolated from Foxg1^tetOFoxg1 mice, in which Foxg1 was induced from E14.5, was used as samples. C, Colored values represent relative fold changes in gene expression between the E16.5 Foxg1^tetOFoxg1 cortex/E14.5 Foxg1^tetOFoxg1 cortex of all TFs that significantly responded to Foxg1 based on a five-dataset (E14.5, E15.0, E15.5, E16.5, and E15.5 Dox⁺ negative control) ANOVA cluster analysis. Tbr1 probe: E16.5/E14.5-fold change (log₂) = −0.9445 (log₁₀ = −1.9245). D–H, Developmental expression of Tbr1 (red) and Foxg1 (green) in E11.5, E13.5, E15.5, E18.5, and P4 wild-type cortices. Rabbit anti-Tbr1 (Abcam) and goat anti-Foxg1 (Santa Cruz Biotechnology) antibodies were used. Tissues from different stages were processed at identical conditions, with the exception that E18.5 (G) and P4 (H) brains were perfused before 1 h fixation. I, View of the mouse Tbr1 gene locus (top), placental mammal basewise conservation by PhyloP (middle) and predicted Foxg1 binding sites of the Tbr1 promoter region (bottom). J, Luciferase reporter assay that used Tbr1 promoter regions. Expression of pCAGGS–Foxg1 repressed the luciferase activity of the Tbr1 promoter constructs by 31.0 ± 2.9%. Error bars represent the SEM obtained from four replicates. Statistical analysis was performed using Student's t test, **p < 0.01. Scale bars: D–H, 50 μm. iz, Intermediate zone; ppl, preplate.

Next, we examined whether Foxg1 can regulate Tbr1 transcription by searching for Foxg1 consensus binding sites (Tao and Lai, 1992) within 10 kb of the transcription start site of the Tbr1 gene (Fig. 4I). We identified a 4 kb region of the mouse Tbr1 promoter that consists of multiple conserved Foxg1 binding sequences (Fig. 4I), which were cloned upstream of the luciferase reporter gene to assess transcription. Cotransfection of pCAGGS–Foxg1 reduced the luciferase reporter activity in Neuro2a cells by 31.0 ± 2.9% (p < 0.01; Fig. 4J). These results indicated that Foxg1 can repress Tbr1 promoter activity in vitro.

We next assessed whether Foxg1 can repress Tbr1 expression in vivo. We used Foxg1^−/− constitutive knock-out brains in which a loss of Foxg1 results in excessive production of the earliest-born CR cells in the cortex (Kumamoto et al., 2013). As expected in these cortices, we observed Tbr1 upregulation in the CP because of its expression in CR cells (Fig. 5B′; Hevner et al., 2001). Therefore, we coelectroporated pCAGGS–Foxg1 with pCAGGS–GFP constructs into E14.5 Foxg1^−/− cortices and examined Tbr1 expression at E18.5 (Fig. 5A). Foxg1 induction cell-autonomously repressed Tbr1 expression within this GFP⁺ cell cluster (Fig. 5C–C″, compared with control pCAGGS–GFP-only transfected cells in Fig. 5B–B″); Ctip2 (Fig. 5C‴) and Fezf2 (Fig. 5D′) expression were only induced in this Tbr1⁻ GFP⁺ population (Fig. 5C‴,D–D‴).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Foxg1 represses Tbr1 expression before Fezf2 and Ctip2 expression onset. A, Experimental scheme. Foxg1^−/− constitutive knock-out mice were used. I.U.E.P., In utero electroporation. B–D‴, GFP, Tbr1, Ctip2, and Fezf2 immunohistochemistry of E18.5 coronal sections from electroporated Foxg1^−/− cortices. B–B″, C–C‴, and D–D‴ indicate a combination of different channels taken from the same sections. E–L‴, GFP, Tbr1, Ctip2, and Fezf2 immunohistochemistry of E15.5 coronal sections from electroporated Foxg1^−/− cortices. F–F‴, H–H‴, J–J‴, and L–L‴ indicate enlarged views of the boxed regions indicated in E, G, I, and K, respectively. Filled arrowheads indicate GFP⁺ cells that are Tbr1⁻ and Ctip2⁻. Open arrowheads indicate GFP⁺ cells that are Tbr1⁺ in controls. Scale bars: B–D‴, E–E′, G–G′, I–I′, K–K′, 50 μm; F–F″, H–H‴, J–J‴, L–L‴, 10 μm.

To further assess whether Foxg1 primarily represses Tbr1 expression to derepress Fezf2/Ctip2 rather than indirectly induces Fezf2/Ctip2 expression through the repression of other TFs and thereby represses Tbr1 as a downstream target gene within the transcription loop (Fig. 4A), we harvested pCAGGS–Foxg1 electroporated brains at an earlier time point (24 h after electroporation); we then examined Tbr1 and Ctip2 expression within the GFP⁺ cells (Fig. 5A,E–L‴). Notably, 24 h was sufficient to repress Tbr1 in Foxg1 electroporated GFP⁺ cells (Fig. 5G–H,H″). Furthermore, this downregulation of Tbr1 preceded the onset of Ctip2 and Fezf2 protein induction (Fig. 5H′,K–K′,L–L‴). Moreover, in contrast to the control pCAGGS–GFP-only electroporated cortices, in which GFP⁺ cells migrated into the Tbr1⁺ CR cell layer (Fig. 5B–B″,E,F), the Foxg1-introduced GFP⁺ neurons failed to integrate into the overlaying supernumerary CR cell layer and accumulated beneath this layer (Fig. 5C–C‴,G,H). These distribution patterns again resembled the migration termination of projection neurons beneath the CR cell containing MZ during normal development.

Next, we assessed whether conditional removal of Foxg1 expression during normal development resulted in Tbr1 upregulation in vivo. We generated Foxg1 conditional knock-out mice (cKO), in which Foxg1 was inactivated from E12.5 by Dox administration in the Foxg1^tetOFoxg1 mice and examined cortices at E14.5 (E12.5 Foxg1 cKO harvested at E14.5; Fig. 6A,B). Notably, although Tbr1 expression was primarily restricted to postmigratory neurons within the CP in control cortices (Fig. 6C–C‴), the Tbr1 protein was highly upregulated in early differentiating neurons that were Tuj1⁺ in the Foxg1 E12.5 cKO cortices (Fig. 6D–D‴). This upregulation was accompanied by the loss of Fezf2 in the CP (Fig. 6F,F′,F″″). Furthermore, in the Foxg1 cKO cortices, we identified a loss of Ctip2 high-expressing cells in the upper part of the CP; however, low Ctip2 expression persisted throughout the CP, and Tbr1 and Ctip2 were coexpressed in these neurons (Fig. 6F″,F‴,F″‴). These results further indicated that Tbr1 upregulation in these mutants is attributed to the loss of Foxg1 expression rather than to the loss of Ctip2 expression within the transcriptional network (Fig. 4A), thus repressing Fezf2 independent of Ctip2 expression regulation (Fig. 4A). Consistent with this view, the segregation between Fezf2⁺ and Tbr1⁺ cells in the CP was detectable during normal development at E14.5 (control cortex; Fig. 6E,E′) and primarily complementary at E15.5 (wild-type cortex; Fig. 6G,G′,G″,G″‴), at which point Ctip2 expression still broadly overlapped with Fezf2⁺ and Tbr1⁺ cells (Fig. 6E–E″‴,G–G‴‴). Together, these data show that continued repression of Tbr1 by Foxg1 is required for the acquisition of Fezf2⁺ DL neuron identity in the neocortex. Collectively, these results demonstrate that Tbr1 repression by Foxg1 confers the sequence of DL and UL competence by establishing the DL (Fezf2^ON/Satb2^OFF/Ctip2^HI) identity (Fig. 4A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Conditional inactivation of Foxg1 at E12.5 results in Tbr1 upregulation and loss of Fezf2 expression. A, Experimental scheme. Foxg1^tetOFoxg1 mice were used. Dox was administered at 200 μg/g for constitutive suppression of Foxg1. B, Dox was administered from E12.5 (Foxg1 OFF) onward, and the embryos were harvested at E14.5. C–C‴, D–D‴, Tbr1, Tuj1 immunohistochemistry, and DAPI staining. E–E″‴, F–F″‴, Tbr1, Fezf2, and Ctip2 immunohistochemistry. G–G‴‴, Tbr1, Fezf2, and Ctip2 immunohistochemistry in wild-type cortex at E15.5. C–C″, D–D‴, E–E″‴, F–F″‴, and G–G‴‴ indicate a combination of different channels taken from the same sections. Scale bars: C–C‴, D–D‴, E–E″‴, F–F″‴, 50 μm; G–G‴‴, 20 μm. iz, Intermediate zone.

Ablation of early postmitotic DL neurons prolongs DL competence at the expense of UL neurogenesis

Tbr1 repression by Foxg1 favors the acquisition of a DL fate over an UL identity within the transcriptional network (Fig. 4A). Because of its nature, the subsequent transition from DL to UL competence requires the repression of DL determinants to terminate DL competence. Shortly after Foxg1 induction in the Foxg1^−/− cortices (24 h after electroporation; Fig. 7A), Foxg1-introduced GFP⁺ cells established apical and basal processes that contacted the ventricular surface and the CP (Fig. 7C, filled arrowheads and open arrowheads, respectively). Based on previous observations that UL neurons favor aggregate culture conditions (Shen et al., 2006; Eiraku et al., 2008; Gaspard et al., 2008; Kadoshima et al., 2013), we hypothesized that repressing the DL identity may require short-range negative feedback signals provided directly by the DL neurons. To directly assess whether postmitotic neurons propagate negative feedback signals that regulate DL competence, we ablated postmitotic DL neurons using genetic approaches. To selectively manipulate DL neurons, we first assessed the termination of DL competence during normal development. E14.5 wild-type mice were labeled with a pulse of BrdU (50 mg/kg body weight) and were subsequently administered BrdU in their drinking water (1.5 mg/ml); then, DL neurogenesis was examined at E18.5 (Fig. 7D). We used Ctip2 and Zfpm2 (merged in green channel) immunohistochemistry to detect all DL neurons (layer V neurons, Ctip2⁺; layer VI and SP neurons, Zfpm2⁺; Kwan et al., 2008) and determined that almost no DL neurons were generated from E14.5 onward during normal development (1.1 ± 0.4% Ctip2⁺Zfpm2⁺/BrdU⁺ cells; Fig. 7E–G′).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Neurog2^CreER/+ mice recombine early postmitotic neurons in the neocortex. A, Experimental scheme. Foxg1^−/− constitutive null mice were used. I.U.E.P., In utero electroporation. B, C, GFP immunohistochemistry of 50 μm cortical slices. Foxg1-introduced GFP⁺ cells established apical processes contacting the ventricular surface (filled arrowheads) and extended basal processes (open arrowheads). D, Wild-type mice were used. Pregnant dams were administered a single-pulse BrdU injection at E14.5, and BrdU was subsequently administered in the drinking water. E–G, Ctip2 and Zfpm2 (merged in green) and BrdU immunohistochemistry in coronal sections (12 μm) of E18.5 cortex. G′, Enlarged view of the boxed regions shown in G. H, K, Schematic diagram of mouse lines crossed to assess Cre recombination. Tamoxifen was administered at the indicated dates to induce recombination, and neuronal labeling was analyzed at E13.5 and E18.5. I, J, β-gal histochemistry (blue) of Neurog2^CreER/+;Rosa26^{loxp–stop–loxp–LacZ/+} mice. L, M, β-gal histochemistry (blue) of Nestin^CreER/+;Rosa26^{loxp–stop–loxp–LacZ/+} mice. When tamoxifen was administered at E12.5 and examined at E13.5, a few labeled cells were observed outside the VZ (I). These labeled neurons had acquired deep laminar positions at E18.5 (J). This result was in contrast to that derived from a Nestin–CreER transgenic mouse line crossed with the same Rosa26^{loxp–stop–loxp–lacZ/+} mice, in which recombination of the reporter allele occurred within the progenitors and labeled all progenitors and postmitotic neurons from E12.5 onward (L, M). N, Schematic diagram of the mouse lines. Neurog2^CreER/+ mice were crossed with Rosa26^{loxp–stop–loxp–YFP/+} reporter mice. Tamoxifen was administered at E13.5 to induce recombination, and the embryos were analyzed 18 h after tamoxifen administration. O–Q, Double immunohistochemistry of GFP and Unc5d, an SVZ marker. These data confirmed labeled cells within the SVZ but not within the VZ. R, Schematic diagram of the mouse lines that were crossed to label early postmitotic neurons using a distinct reporter line, Rosa26^{CAG--loxp–stop–loxp–tdTomato/+}, which expresses high levels of reporter fluorescent protein. Tamoxifen was administered on the indicated dates to induce recombination, and the embryos were analyzed at E14.5. S–AA, Double immunohistochemistry of Sox2 (radial glial progenitor marker; S–U), Tbr2 (basal progenitor marker; V–X), and Ki67 (proliferative cell marker; Y–AA) with RFP. These data suggested that the Neurog2^CreER/+ line is capable of labeling postmitotic neurons, but not progenitor cell populations, during serial tamoxifen administration at E11.5–E13.5. Scale bars: B–C, E–G, S–AA, 50 μm; G′, 20 μm. iz, Intermediate zone.

To ablate postmitotic neurons, we used Neurog2 gene expression. A previous study reported dynamic Neurog2 protein expression during the cortical progenitor cell cycle; Neurog2 expression is low in the Ki67⁺ cycling progenitors, but it is elevated in neuronally committed (Ki67⁻) cells (Miyata et al., 2001). We hypothesized that Neurog2 expression onset will allow the manipulation of early postmitotic neurons before the maturation of the DL identity necessary for negative feedback signal propagation. Therefore, we used the Neurog2 gene locus to drive the ablation of this population (Neurog2^CreER/+ mice) (Zirlinger et al., 2002). First, we crossed the Neurog2^CreER/+ mice with three distinct reporter lines to assess recombination within the cortical cells (Fig. 7H,N,R). We crossed Neurog2^CreER/+ mice with Rosa26^{loxp–stop–loxp–LacZ/+} reporter mice (Soriano 1999). When tamoxifen was administered at E12.5 and the cortices were examined at E13.5 (Fig. 7H), we identified a few labeled cells outside the VZ (Fig. 7I), and these labeled neurons occupied deeper laminar positions at E18.5 (Fig. 7J). This result was in contrast to the findings derived from a Nestin–CreER transgenic mouse line (Balordi and Fishell, 2007) crossed with the same Rosa26^{loxp–stop–loxp–lacZ/+} mice, in which recombination of the reporter allele occurred within the progenitors and labeled all progenitors and postmitotic neurons from E12.5 onward (Fig. 7K–M). We further crossed Neurog2^CreER/+ mice with Rosa26^{loxp–stop–loxp–YFP/+} reporter mice (Srinivas et al., 2001) and administered tamoxifen at E13.5 to induce recombination; the embryos were analyzed 18 h later (Fig. 7N–Q). Double immunohistochemistry of GFP and Unc5d, which is a subventricular zone (SVZ) marker, confirmed labeled cells within the SVZ but not the VZ. We further used a Rosa26^{CAG–loxp–stop–loxp–tdTomato/+} reporter mouse line that expresses high levels of reporter fluorescent protein during recombination (Madisen et al., 2010). Tamoxifen was administered on 3 consecutive days (noon of E11.5, E12.5 and E13.5) to induce recombination, and the embryos were analyzed at E14.5 (Fig. 7R). Double immunohistochemistry of Sox2 (a radial glial progenitor marker), Tbr2 (a basal progenitor marker), and Ki67 (a proliferating cell marker that is expressed in all mitotic cells) with RFP confirmed Neurog2^CreER/+ recombination in postmitotic neurons during tamoxifen administration at E11.5–E13.5 but not the Sox2⁺ (Fig. 7S–U), Tbr2⁺ (Fig. 7V–X), or Ki67⁺ (Fig. 7Y–AA) progenitor cell populations (Fig. 7S–AA).

To ablate early postmitotic DL neurons, we used diphtheria toxin A chain (DTA) expression driven from the same Rosa26 locus (Rosa26^{loxp–stop–loxp–dta/+} line; Ivanova et al., 2005). Neurog2^CreER/+ mice crossed with Rosa26^{loxp–stop–loxp–dta/+} mice were administered tamoxifen at E11.5, E12.5, and E13.5 to induce cell death in the DL cells (Fig. 8A). Consistent with the ablation of DL neurons, we confirmed a significant decrease in Ctip2/Zfpm2⁺ neurons at E14.5 (merged in red, 150.0 ± 23.2 cells in DL-ablated cortices and 408.0 ± 14.3 cells in controls, p < 0.01; Fig. 8B–C′,F). In contrast, Tbr2⁺ cell number was maintained (276.3 ± 27.6 cells in DL-ablated cortices and 258.3 ± 6.9 cells in controls, p = 0.63, not significant; Fig. 8D–E′,G), indicating that progenitor cells were not eliminated by this ablation, consistent with the absence of Neurog2^CreER/+-mediated recombination in this population (Fig. 7S–AA). Therefore, we assessed the effects of DL ablation on neocortical neurogenesis by quantifying the number of DL and UL neurons at E18.5 (Fig. 8H). We determined that the number of Ctip2⁺ neurons also decreased at this stage (501 ± 45 cells in DL-ablated cortices vs 713 ± 8.6 cells in controls, p < 0.01; Fig. 8I,J,O). Notably, at E18.5, the production of Satb2⁺ neurons in these animals also slightly decreased (677 ± 85 cells in DL-ablated cortices vs 1079 ± 61 cells in controls, p < 0.05; Fig. 8K,L,O), and the proportion of Ctip2⁺ and Satb2⁺ cells was comparable with the controls (Ctip2⁺ cells/all marker⁺ cells; 40.8 ± 1.8% in DL-ablated cortices vs 39.4 ± 1.4% in controls, p = 0.63, not significant, Satb2⁺ cells/all marker⁺ cells; 54.2 ± 1.6% in DL-ablated cortices vs 59.2 ± 1.6% in controls, p = 0.1, not significant; Fig. 8M–N′,P). To confirm these results, we used an alternative marker for UL neurons, Cux1, which is expressed in most UL neurons (Nieto et al., 2004). The ablation of DL neurons also reduced the number of Cux1⁺ cells at E18.5 (1206 ± 74 cells in DL-ablated cortices vs 1393 ± 75 cells in controls, p < 0.01; Fig. 8S,T,W). This ablation resulted in a similar ratio of Ctip2⁺/Cux1⁺ neurons compared with controls at E18.5, in which the proportions of Ctip2⁺ or Cux1⁺ cells were not significantly different (Ctip2⁺ cells/all marker⁺ cells; 36.2 ± 0.8% in DL-ablated cortices vs 37.0 ± 1.1% in controls, p = 0.58, not significant, Cux1⁺ cells/all marker⁺ cells; 63.6 ± 0.8% in DL-ablated cortices vs 62.7 ± 1.0% in controls, p = 0.48, not significant; Fig. 8U–V′,X).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Neuronal subtype ratio is maintained despite genetic DL neuron ablation. A, Schematic diagram of the mouse lines that were crossed to ablate DL neurons. Tamoxifen was administered on the indicated dates to induce neuronal ablation. The embryos were analyzed at E14.5. B–E′, Immunohistochemistry of Pan-DL markers (Ctip2 and Zfpm2, merged in red; B–C′) and Tbr2 (D–E′) in Neurog2^CreER/+;Rosa26^{loxp–stop–loxp–dta/+} mice (DL-ablated) and tamoxifen-administered wild-type littermate controls. F, G, Quantitative analysis. The total number of cells that expressed Ctip2 and Zfpm2 (F) or Tbr2 (G) per 100 μm cortical strip from the dorsolateral cortex was counted. The data represent the mean ± SEM values. **p < 0.01. N.S., Not significant. H, Schematic diagram of the mouse lines that were crossed to ablate DL neurons. Tamoxifen was administered on the indicated dates to induce neuronal ablation. The embryos were analyzed at E18.5. I–N, Ctip2 and Satb2 immunohistochemistry in Neurog2^CreER/+;Rosa26^{loxp–stop–loxp–dta/+} mice (DL-ablated) and Rosa26^{loxp–stop–loxp–dta/+} mice (controls). M′, N′, Enlarged view of the boxed regions shown in M and N. O, Quantitative analysis of absolute marker⁺ cell number per unit area: Satb2⁺ only (red), Ctip2⁺ only (green), and Ctip2⁺ and Satb2⁺ (yellow) cells. The total number of cells per 260-μm-wide cortical strips from the dorsolateral cortices was counted. P, Quantitative analysis of the proportion of marker⁺ cells in same radial strips used in O: Satb2⁺ only (red), Ctip2⁺ only (green), and Ctip2⁺ and Satb2⁺ (yellow) cells of the total cells that expressed either of the two (Ctip2 or Satb2) markers. Q–V, Ctip2 and Cux1 immunohistochemistry in Neurog2^CreER/+;Rosa26^{loxp–stop–loxp–dta/+} mice (DL-ablated) and Rosa26^{loxp–stop–loxp–dta/+} mice (controls). U′, V′, Enlarged view of the boxed regions shown in U and V. W, Quantitative analysis of absolute marker⁺ cell number per unit area: Cux1⁺ only (red), Ctip2⁺ only (green), and Ctip2⁺ and Cux1⁺ (yellow) cells. The total number of cells per 260-μm-wide cortical strips from the dorsolateral cortices was counted. X, Quantitative analysis of the proportion of marker⁺ cells in same radial strips used in W: Cux1⁺ only (red), Ctip2⁺ only (green), and Ctip2⁺ and Cux1⁺ (yellow) cells of the total cells that expressed either of the two (Ctip2 or Cux1) markers. The data represent the mean ± SEM values. *p < 0.05, **p < 0.01. Scale bars: B–E′, I–N′, Q–V′, 50 μm. N.S., Not significant.

Because the relative DL neuron/UL neuron ratio was maintained despite the ablation of a significant number of DL neurons (Fig. 8B–C′,F), this finding prompted us to investigate whether the elimination of DL neurons affects the temporal window for DL competence. We injected EdU at E14.5 to monitor the neurons that were born from progenitors in post-ablated cortices (Fig. 9A) using pan-DL neuronal markers. Consistent with the previous results (Fig. 7D–G′), we found almost no Ctip2/Zfpm2⁺ cells that were colabeled with EdU in the control (Rosa26^{loxp–stop–loxp–dta/+}) cortices (3.1 ± 1.1%; Fig. 9B–B‴,D). Remarkably, in Neurog2^CreER/+;Rosa26^{loxp–stop–loxp–dta/+} (DL-ablated) cortices, 23.7 ± 2.2% (p < 0.01) of EdU⁺ cells were colabeled with Ctip2 or Zfpm2 (Fig. 9C–C‴,D), which indicates that the ablation of DL neurons prolonged the production period of DL neurons. These results also suggest that the observed total number of DL neurons at E18.5 (Fig. 8O) included neurons that were generated after ablation, and therefore, the number of ablated neurons appears under-represented in this population. Next, we assessed whether feedback signaling propagated from DL neurons is required only to terminate the DL competence or this signaling also drives UL competence. In the latter case, the number of UL neurons born at E14.5 was expected to decrease in the DL-ablated cortices compared with the controls. Consistent with the general decrease in the Satb2⁺ neuron population (Fig. 8O), we determined that Satb2⁺ cells born at E14.5 (EdU⁺ cells) also decreased at the expense of increased DL production (77.5 ± 8.2% in control cortices and 43.5 ± 7.8% in DL-ablated cortices, Satb2⁺/EdU⁺ cells, p < 0.05; Fig. 9F–F‴,G). The Cux1 analysis confirmed these findings; the number of Cux1⁺ neurons generated at E14.5 was also significantly decreased (40.3 ± 3.0% in control cortices and 16.1 ± 5.0% in DL-ablated cortices, Cux1⁺/EdU⁺ cells; p < 0.05; Fig. 9I–I‴,J). Collectively, these results demonstrate that the onset of UL neuron generation is controlled by the termination of DL competence, which is propagated through postmitotic DL neurons. To our knowledge, this study provides the first evidence that progenitor cells use negative feedback cues to regulate the neocortical subtype transition.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Ablation of postmitotic DL neurons prolongs DL competence at the expense of UL neurogenesis. A, Experimental scheme. B–I, Immunohistochemistry of EdU and pan-DL markers (Ctip2 and Zfpm2, merged in green) or UL markers (Satb2 and Cux1). B‴–I‴, Filled arrowheads indicate Ctip2⁺/Zfpm2⁺, Satb2⁺, or Cux1⁺ cells double labeled with EdU, and open arrowheads indicate Satb2⁻ or Cux1⁻ EdU⁺ cells. D, G, J, Quantitative analysis of the percentage of EdU⁺ cells that expressed pan-DL markers (Ctip2 and Zfpm2) or UL markers (Satb2 and Cux1). The results are expressed as the mean ± SEM. *p < 0.05, **p < 0.01. Scale bars, 50 μm. K, A model for the switch in layer-specific subtypes. Before Foxg1 expression, preplate cells express Tbr1 by default, most likely through one or more transcription activator(s) binding to the Tbr1 promoter region. The mechanism of Satb2 repression in preplate cells is unclear; Satb2 expression may require appropriate transcription activator(s) downstream of Foxg1. After its induction, Foxg1 initiates general projection neuron competence through global gene suppression (Kumamoto et al., 2013) and also suppresses Tbr1, which causes progenitor cells to acquire the DL (Fezf2^ON/Satb2^OFF/Ctip2^ON) identity. Finally, the transition from DL to UL neurogenesis requires a feedback signal from postmitotic DL cells that represses Fezf2/Ctip2 and, thus, confers the UL (Fezf2^OFF/Satb2^ON/Ctip2^OFF) identity.