Progenitor cells undergo a series of stable identity transitions on their way to becoming fully differentiated cells with unique identities. Each cellular transition requires that new sets of genes are expressed, while alternative genetic programs are concurrently repressed. Here, we investigated how the proneural gene Neurog2 simultaneously activates and represses alternative gene expression programs in the developing neocortex. By comparing the activities of transcriptional activator (Neurog2-VP16) and repressor (Neurog2-EnR) fusions to wild-type Neurog2, we first demonstrate that Neurog2 functions as an activator to both extinguish Pax6 expression in radial glial cells and initiate Tbr2 expression in intermediate neuronal progenitors. Similarly, we show that Neurog2 functions as an activator to promote the differentiation of neurons with a dorsal telencephalic (i.e., neocortical) identity and to block a ventral fate, identifying 2 Neurog2-regulated transcriptional programs involved in the latter. First, we show that the Neurog2-transcriptional target Tbr2 is a direct transcriptional repressor of the ventral gene Ebf1. Secondly, we demonstrate that Neurog2 indirectly turns off Etv1 expression, which in turn indirectly regulates the expression of the ventral proneural gene Ascl1. Neurog2 thus activates several genetic off-switches, each with distinct transcriptional targets, revealing an unappreciated level of specificity for how Neurog2 prevents inappropriate gene expression during neocortical development.
The neocortex is comprised of 6 layers of glutamatergic projection neurons that are sequentially generated from dorsal telencephalic progenitors (also designated cortical progenitor cells) between embryonic day (E) 10 to E17 in mouse (Takahashi et al. 1999). Cortical progenitors undergo a series of stable identity transitions over this period. They begin as neuroepithelial cells that span the apicobasal axis of the ventricular zone (VZ), where they undergo symmetric proliferative divisions to expand the progenitor pool (Kriegstein and Noctor 2004). Then, at ∼E10.5, cortical neuroepithelial cells differentiate into radial glial cell (RGC) progenitors that retain their apical and basal contacts while initiating asymmetric neurogenic mitoses at the apical surface of the VZ (Malatesta et al. 2003; Gotz and Huttner 2005). Next, at ∼E11.5, a subset of RGCs give rise to intermediate neuronal progenitor cells (INPs) that initially reside in the VZ, but later lose their apical contacts and migrate outward to form the subventricular zone (SVZ), a more basally-located germinal zone (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004). SVZ INPs have a limited proliferative capacity and undergo symmetric, neurogenic divisions after 1–2 cell cycles (Noctor et al. 2004; Farkas et al. 2008).
Cortical progenitor cell transitions depend on the coordinated activation and repression of distinct genetic programs. RGCs initially express the homeodomain transcription factor Pax6 and later, as they differentiate into VZ INPs, they turn on the T-box transcription factor Tbr2, before finally extinguishing Pax6 expression as they transform into SVZ INPs (Englund et al. 2005; Kowalczyk et al. 2009). Pax6 is required to maintain a RGC identity and to initiate the RGC to INP transition (e.g., Estivill-Torrus et al. 2002; Quinn et al. 2007). Likewise, Tbr2 is required to specify an INP identity (Arnold et al. 2008; Sessa et al. 2008). To promote the RGC to INP transition, Pax6 functions as an activator, initiating transcription of Tbr2 (Holm et al. 2007; Sansom et al. 2009) as well as AP2γ, which is required for INP differentiation in the caudal cortex (Pinto et al. 2009). In contrast, Tbr2 may have transcriptional repressor activity, as it cell autonomously blocks Pax6 expression (Sessa et al. 2008), although it is not yet known if this is due to direct transcriptional repression.
Progression through the different phases of cortical progenitor cell maturation is coordinated with the initiation of neuronal differentiation, which occurs in a restricted subset of progenitors at any given time (Shimojo et al. 2008). In the embryonic neocortex, neurogenesis is induced by the proneural basic helix-loop-helix (bHLH) transcription factors Neurog1 and Neurog2 (Fode et al. 2000; Schuurmans et al. 2004; Mattar et al. 2008). Neurog1/2 regulates several binary cell fate decisions during neocortical development, all of which require the simultaneous activation and repression of alternative genetic programs. For instance, Neurog2 promotes the RGC to INP transition, likely by directly initiating Tbr2 transcription in INPs and repressing Pax6 in RGCs, the latter through unknown mechanisms (Miyata et al. 2004; Britz et al. 2006; Ochiai et al. 2009). Neurog2 also dictates the type of neuron that differentiates, specifying a dorsal, glutamatergic projection neuron identity while simultaneously repressing an alternative ventral, GABAergic interneuronal identity (Fode et al. 2000; Schuurmans et al. 2004; Mattar et al. 2008). Thus, Neurog2 must activate dorsal genetic pathways and repress ventral genes, such as Ascl1, a proneural bHLH protein that specifies a ventral telencephalic identity (Fode et al. 2000; Schuurmans et al. 2004; Mattar et al. 2008).
To date, the Neurog2-regulated cell fate decisions for which a mechanism of transcriptional repression has been identified include the neuronal versus glial fate choice (Sun et al. 2001), and the control of neuronal migration (Ge et al. 2006; Heng et al. 2008; Pacary et al. 2011). Mechanistically, Neurog2 represses astrocytic genes such as GFAP, as well as a negative regulator of migration, RhoA, in an indirect, non-DNA–binding–dependent fashion, sequestering CREB-binding protein (CBP) and other activator proteins from the regulatory regions of these genes (Sun et al. 2001; Ge et al. 2006; Heng et al. 2008; Pacary et al. 2011). In contrast, it is not yet known how Neurog2 represses Pax6, Ascl1, or other ventral genes. We set out to determine if Neurog2 acts as a direct transcriptional repressor to turn off alternative genetic programs during neocortical development, or if instead, Neurog2 functions as a transcriptional activator, switching off inappropriate gene expression by sequestering co-activators, or more directly, by initiating the expression of downstream transcriptional repressors. For this purpose, we generated Neurog2 fusion proteins that function as obligate transcriptional repressor (Neurog2-EnR) or activator (Neurog2-VP16) proteins. We predicted that if Neurog2 directly represses the expression of genes in alternate pathways, Neurog2-EnR would phenocopy wild-type Neurog2 (Neurog2-WT), recruiting co-repressor proteins to shut off the expression of inappropriate genes. Conversely, we predicted that if Neurog2 functions indirectly, Neurog2-VP16 would mimic Neurog2-WT, perhaps by initiating the expression of downstream transcriptional repressors, or by sequestering transcriptional activators away from target promoters in alternate pathways. Here, we demonstrate that Neurog2-VP16 phenocopies Neurog2-WT in gain-of-function assays in vitro and in vivo, suggesting that Neurog2 functions as a transcriptional activator to switch off alternative gene expression. Furthermore, we identify a Neurog2-Tbr2-Pax6 transcriptional cascade that turns off Pax6 expression during cortical progenitor maturation, and Neurog2-Tbr2-Ebf1 and Neurog2-Etv1-Hes5-Ascl1 transcriptional cascades that turn off ventral gene expression in cortical cells. Combined, our data demonstrate that Neurog2 functions as a transcriptional activator to initiate the expression of several genetic off-switches, each of which repress distinct gene expression programs during neocortical development.
Materials and Methods
Animal Breeding and Maintenance
All animal procedures were compliant with the Guidelines of the Canadian Council of Animal Care and were approved by the University of Calgary Animal Care Committee under animal protocol M08003. CD1 (outbred) mice were intercrossed to generate timed pregnancies for in utero electroporation experiments. Embryos were staged by taking the date of the vaginal plug as embryonic day (E) 0.5. The Neurog2kigfp null allele was maintained on a CD1 background and PCR genotyping was performed as previously described (Britz et al. 2006).
Generation of Expression and Reporter Constructs
For electroporation, cDNA was cloned into pCIG2, a bicistronic expression vector that includes an internal ribosome entry site (IRES) green fluorescent protein (GFP) cassette as described (Mattar et al. 2008). A pNeuroD1kbluciferase reporter was previously generated (Huang et al. 2000). To generate a pCMV-NeuroD1 kbluciferase reporter, the 1 kb 5′ NeuroD promoter/enhancer elements were amplified by PCR, and cloned into pMIR-Report, which contains a CMV-luciferase cassette. To generate a Pax6-5 kb–P1 telencephalic-specific reporter, PCR was used to amplify 5130 bp between the P1 and P0 promoters of Pax6, a region encompassing regulatory elements that drive reporter expression in the dorsal telencephalon (Kammandel et al. 1999). The 5130 bp Pax6 fragment was cloned into pGL3 Basic (Promega).
Protein Half-Life Assay
Proteins were transcribed and translated in vitro using a TNT rabbit reticulocyte lysate kit (Promega) and radiolabeled using 35S-methionine according to the manufacturer's instructions. Protein degradation assays were performed as described (Nguyen et al. 2006). Additional details are described in the Supplementary Material.
Cell Culture and Luciferase/β-Galactosidase Assays
The P19 embryonic carcinoma cell line (ATCC# CRL-1825) was obtained from the American Type Culture Collection (ATCC, Manassas, VA, United States of America). Cells were maintained in complete medium, containing Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum (FBS), 5% horse serum (HS), 10 units/mL penicillin, 10 μg/mL streptomycin, and 29.2 mg/mL l-glutamine. DNA for transfection was prepared using Qiagen plasmid plus maxi kit as described by the manufacturer. Details of the luciferase and β-galactosidase assays are provided in the Supplementary Material.
In Utero Electroporation
Endotoxin-free plasmid DNA for electroporation was generated using a column-based purification system (Qiagen, Mississauga, ON, Canada). In utero electroporation was performed as previously described (Dixit et al. 2011). Briefly, DNA (1–3 μg/μL) mixed with Fast Green FCF dye (1:200) was injected into the telencephalic (i.e., lateral) ventricles at defined embryonic stages using pulled borosilicate needles and a Femtojet microinjector. Next, 7 pulses of 45–55 mV were applied within a 7-s interval to the uterus surrounding the head of the embryo using a BTX electroporator. The uterus was replaced in the body cavity, the peritoneum was sutured and skin stapled and embryos were allowed to develop until the designated stage of analysis.
Tissue Processing and RNA In Situ Hybridization
Whole brains were dissected in cold phosphate buffered saline (PBS) and then fixed for 24 h at 4 °C in 4% paraformaldehyde/1× PBS. Tissues were washed in 1× PBS and then cryoprotected overnight (O/N) at 4 °C in 20% sucrose/1× PBS. For electroporated brains, fixation and cryoprotection were performed in the dark to avoid GFP degradation. Tissues were then embedded in optimal cutting temperature (Tissue-TEK) and stored at −80 °C. Coronal sections (10 μm) were collected on Superfrost/Plus slides (Fisher Scientific, Ottawa, ON, Canada) using a Leica cryostat (Richmond Hill, ON, Canada). RNA in situ hybridization using digoxygenin-labeled probes was performed as described previously (Alam et al. 2005). Probes are described in the Supplementary Material.
Sections were blocked for 1 h in 10% HS/1× PBS with 0.1% Triton X-100 (PBST) at room temperature (RT). Briefly, primary antibodies were diluted in blocking solution and applied to sections O/N at 4 °C. Sections were then washed 3× in PBST, and then secondary antibodies were appropriately diluted in blocking solution and applied to the sections for 2 h at RT in the dark. Sections were washed 3× in PBST, incubated 5 min with 4′,6-diamidino-2-phenylindole (DAPI; Sigma), diluted 1/10 000 in PBS, washed 3× in PBS, and mounted using Aqua Polymount (Polysciences Inc., Woodbridge, ON, Canada). Primary and secondary antibodies are described in the Supplementary Material.
Chromatin immunoprecipitation (ChIP) experiments were performed as previously described with some modifications (Demers et al. 2007). Cells from E14.5 mouse telencephalons were dissociated by incubating the tissue in 0.05% trypsin–ethylenediaminetetraacetic acid (Invitrogen# 25300-054) for 10 min at 37 °C. The cells were crosslinked at RT for 10 min with 1% (w/w, final) formaldehyde. Sonication was performed with a Bioruptor (Diagenode) at high power for 30 cycles (30 s on/30 s off). Immunoprecipitations with rabbit anti-Tbr2 (Abcam) or control antibodies (FBS) were performed with 500 μg of chromatin per assay. qPCR was performed with Bio-Rad DNA Engine Opticon 2 real-time PCR detection systems and a QuantiFast SYBR Green Kit for quantification (Qiagen). PCR products were also loaded on a 2% agarose gel to verify the size of the amplification product. Primer design is described in the Supplementary Material.
Quantitative and Statistical Analysis
For the analysis of luciferase/β-galactosidase assays, luciferase data were normalized by dividing raw light readings by the corresponding A420β-gal or Renilla values. Reported n-values correspond to the number of individual experiments performed, each comprised of 3 replicates per sample. For in vivo experiments, brains from at least 3 independent experiments, and at least 3 section for each brain, were processed (n-values refer to # of brains analyzed). Comparisons between control and experimental conditions were performed using a 2-tailed Student's t-test (to compare 2 values), while comparisons between multiple samples were performed by applying a 1-way analysis of variance and Tukey's multiple comparison test using the Graphpad Prizm software (Graphpad Software Inc., La Jolla, CA, United States of America). Statistical variation was determined using the standard error of the mean (SEM).
Generation of Obligate Activator and Repressor Forms of Neurog2
Neurog2 both positively and negatively regulates the expression of downstream genes to carry out its myriad of neocortical functions (Fig. 1A). Despite considerable understanding of how Neurog2 can function as a transcriptional activator, its ability to repress alternative gene expression programs is less well understood. To date, Neurog2 has been shown to repress RhoA (Ge et al. 2006) and GFAP (Sun et al. 2001) by sequestering transcriptional activators or co-activators away from target promoters (Fig. 1A). Here, we asked if Neurog2 employs a similar indirect strategy to suppress Pax6 and Ascl1 transcription, or if instead, Neurog2 functions as a transcriptional repressor. For this purpose, we generated Neurog2 fusion proteins that bind to DNA, but which have obligate activator (Neurog2-VP16) or repressor (Neurog2-EnR) activity. We reasoned that if Neurog2 represses transcription of target genes directly, by binding DNA and recruiting transcriptional co-repressors, then the Neurog2-repressor fusion would have the same effects as Neurog2-WT (Fig. 1B). If, on the other hand, Neurog2 represses transcription of target genes indirectly, either by sequestering co-activators or by inducing the expression of downstream transcriptional repressors, we expected that the Neurog2-activator fusion would function like Neurog2-WT (Fig. 1C).
To generate an activator fusion, we linked the Neurog2 C-terminus to the VP16 transactivator domain (Fig. 1C), which interacts with basal transcriptional machinery and histone acetylases to transactivate target genes (Hall and Struhl 2002). To generate an obligate repressor, the Neurog2 C-terminus was fused to the engrailed repressor domain (EnR; Fig. 1B), which interacts with Grg/transducin-like enhancer of split (TLE) co-repressors that in turn recruit histone deacetylases to target promoters (Buscarlet and Stifani 2007). To ensure that the functions of the Neurog2 fusion proteins could be directly compared, we measured protein half-lives and demonstrated that Neurog2-WT protein had a half-life of ∼0.5 h, consistent with our previous results (Nguyen et al. 2006), while Neurog2-VP16 and Neurog2-EnR proteins had half-lives of ∼0.7 h and ∼1.2 h, respectively (Fig. 1D). Thus, while the EnR and VP16 domains had modest stabilizing effects, Neurog2-VP16 and Neurog2-EnR retained relatively short intracellular half-lives, allowing us to compare their transcriptional and biological activities to Neurog2-WT.
To assess the transcriptional activities of the Neurog2 fusion proteins, reporter assays were performed in P19 embryonic carcinoma cells, which can differentiate into neural lineages, notably in response to Neurog2 (Farah et al. 2000). Importantly, P19 cells express the co-repressors TLE1, TLE2, and TLE4 (Yao et al. 1998), indicating that necessary co-repressors for Neurog2-EnR were available to be recruited, and that these cells are an appropriate system to test Neurog2-EnR function. To monitor Neurog2 transcriptional activity, we used a transcriptional reporter for NeuroD1 (pNeuroD1kb; Fig. 1E), a direct target of Neurog2 (Huang et al. 2000). Twenty-four hours post-transfection, Neurog2-WT promoted a 17-fold increase (n = 10; P < 0.0001) in NeuroD1 reporter activity (n = 9; normalized to 1), while Neurog2-VP16 stimulated a 32-fold increase (n = 6; P < 0.0001), significantly higher than Neurog2-WT (P < 0.0001; Fig. 1E). In contrast, Neurog2-EnR did not transactivate the NeuroD reporter (n = 9; P = 0.16), nor did the VP16 (n = 6; P = 0.72) or EnR (n = 4; P = 0.44) domains expressed alone (Fig. 1E). The inactivity of Neurog2-EnR was not due to a failure of the encoded fusion protein to translocate to the nucleus, as immunostaining of cortical cells transfected with IRES-enhanced green fluorescent protein (EGFP) expression vectors for Neurog2-EnR and Neurog2-VP16 revealed that both fusion proteins localized to the nucleus (Fig. 1F–G″), as previously shown for Neurog2-WT (Mattar et al. 2008). Finally, transactivation of the NeuroD promoter was dependent on DNA binding as mutation of 2 critical amino acids in the DNA-binding domain of Neurog2-WT (N2NR > AQ; Sun et al. 2001; n = 4; P = 0.08; Fig. 1D) and Neurog2-VP16 (Neurog2-VP16NR > AQ; n = 5; P = 0.16; data not shown) ablated transactivation. Neurog2-WT and Neurog2-VP16 thus both function as transcriptional activators, with Neurog2-VP16 displaying enhanced transcriptional activity, either because of the modest increase in protein stability or because of the enhanced transactivation strength of the VP16 domain.
The low basal levels of pNeuroD1kb luciferase activity made it difficult to assess Neurog2-EnR repressor function, so we implemented 2 additional assays. First, in co-transfection assays, we showed that Neurog2-EnR was able to act in a dominant negative fashion to reduce the transcriptional activity of Neurog2-WT on the NeuroD1 promoter (1.2-fold decrease; n = 5; P < 0.01; Fig. 1H). Secondly, we generated a luciferase reporter with high basal transcriptional activity by cloning a Neurog2-responsive enhancer (i.e., E-box cluster from NeuroD1 promoter) (Huang et al. 2000) downstream of a CMV-luciferase cassette (pCMV-NeuroD11 kb; Fig. 1I). Neurog2-EnR suppressed the basal transcriptional activity of pCMV-NeuroD11 kb (2.3-fold decrease; n = 5; P < 0.05), while both Neurog2-WT (1.6-fold increase; n = 8, P < 0.05) and Neurog2-VP16 (1.8-fold increase; n = 8; P < 0.05) modestly increased the constitutive levels of transactivation from the CMV promoter (Fig. 1I). Neurog2-EnR therefore acts as a transcriptional repressor, and furthermore, functions as a dominant negative inhibitor of Neurog2-WT, likely through its ability to dimerize with either endogenous Neurog2 protein itself, or with E-proteins, which are essential Neurog2 cofactors.
Neurog2 Acts as a Transcriptional Activator to Suppress a Radial Glial Cell Identity and Promotes the Transition to an Intermediate Neuronal Progenitor
Previous loss- and gain-of-function studies have suggested that Neurog2 is required and sufficient to promote the RGC to INP transition (Miyata et al. 2004; Britz et al. 2006; Ochiai et al. 2009). Accordingly, expression of the INP marker Tbr2 is reduced in Neurog2 mutant cortices (Schuurmans et al. 2004), while the RGC marker Pax6 is ectopically expressed (Britz et al. 2006). Moreover, Tbr2 has been identified as a direct transcriptional target of Neurog2 (Ochiai et al. 2009), suggesting that Neurog2 likely functions as a transcriptional activator to initiate an INP fate. In contrast, the mechanism by which Neurog2 represses Pax6 expression and an RGC identity is unknown, both in the neocortex, and in the spinal cord, where a similar negative regulatory interaction between Neurog2 and Pax6 is observed (Bel-Vialar et al. 2007).
To determine how Neurog2 promotes the Pax6+ RGC to Tbr2+ INP fate transition, we introduced expression constructs for the Neurog2 activator and repressor fusion proteins into E12.5 neocortical progenitors via in utero electroporation (Dixit et al. 2011). The pCIG2 expression construct had an IRES-GFP cassette, allowing transfected cells to be visualized via GFP epifluorescence or RNA in situ hybridization using a GFP riboprobe (Fig. 1J). We first confirmed that Neurog2-EnR would be able to recruit co-repressors in vivo by examining the expression of TLE1–4 in the E13.5 telencephalon. TLE1 and TLE3 were expressed in dorsal (i.e., neocortical) and ventral (lateral ganglionic eminence) telencephalic germinal zones, where progenitor cells are located (Fig. 1K–N). TLE2, TLE3, and TLE4 were expressed in the neocortical preplate and TLE4 was expressed in the ventral mantle zone, where differentiated neurons reside (Fig. 1K–N). Therefore, Neurog2-EnR has the potential to associate with TLE proteins and functions as a transcriptional repressor in dorsal and ventral telencephalic progenitors and post-mitotic neurons.
RGCs undergo interkinetic nuclear migration during the cell cycle such that G2/M-phase nuclei are positioned at the ventricular surface, while INPs lose their apical contacts and divide in non-surface or basal positions (Englund et al. 2005; Kowalczyk et al. 2009). We first asked how the different Neurog2 fusion proteins influenced cortical progenitor cell maturation by examining the ratio of apical to basal cell divisions following the misexpression of these proteins. For this purpose, we quantitated the number of GFP+-transfected cells that expressed phospho-histone H3 (pHH3), a late-G2/M-phase marker. In E12.5 → E13.5 control electroporations of pCIG2, most GFP+/pHH3+ cells were apically located (Fig. 2A,M,N; Supplementary Table S1), consistent with the inherent bias of this technique to transfect cortical cells with apical contacts (Britz et al. 2006; Kowalczyk et al. 2009). Unexpectedly, however, all 3 forms of Neurog2 (-WT, -VP16, and -EnR) promoted non-surface cell divisions as evidenced by the increased basal-to-apical ratio of mitotic figures (Fig. 2B–D,M,N; Supplementary Table S1). Notably, the reduction in surface-dividing apical progenitors was not due to an increase in apoptosis since overall numbers of electroporated EGFP+ cells expressing activated caspase 3, a marker of cells committed to the cell death pathway, was similar following electroporation with all constructs (Supplementary Fig. S1A–E). We interpreted the basal bias of Neurog2-VP16– and Neurog2-WT–transfected cells as a direct role for Neurog2 in promoting an INP fate by functioning as a transcriptional activator (Miyata et al. 2004). In contrast, we suggest that Neurog2-EnR may increase basal cell divisions because of its ability to induce Ascl1 expression (see below). Indeed, we have previously demonstrated that the increase in Ascl1 expression in Neurog2 mutant cortices is a likely cause of the increase in basal cell divisions (Britz et al. 2006).
To further explore how Neurog2 regulates the RGC to INP transition, we used molecular markers to phenotype the cortical cells misexpressing the different forms of Neurog2 within 24–72 h post-electroporation. Both Neurog2-WT and Neurog2-VP16 rapidly induced the expression of Tbr2, a marker of INPs committed to a neuronal fate (Englund et al. 2005; Arnold et al. 2008; Kowalczyk et al. 2009), while significantly fewer Neurog2-EnR–transfected cells expressed Tbr2 compared with controls (Fig. 2I–L,I′–L′,P,R; Supplementary Table S2). This is consistent with previous reports demonstrating that Neurog2 directly induces Tbr2 transcription and the acquisition of an INP fate (Ochiai et al. 2009). Conversely, the number of GFP+/Pax6+ cells observed in Neurog2-WT and Neurog2-VP16 transfected cortices was reduced after 72 h compared with controls, while significantly more GFP+/Pax6+ cells were evident in Neurog2-EnR–transfected cortices within 24 h (Fig. 2E–G,E′–G′,O,Q; Supplementary Table S2). Notably, Neurog2-WT had a similar capacity to induce Tbr2 and block Pax6 expression whether it was electroporated at a concentration of 3 μg/μL or <2 μg/μL (Supplementary Table S4), indicating that the observed effects were not due to non-specific toxicity of higher DNA concentrations.
Taken together, our data support the model that Neurog2 functions as a transcriptional activator to promote the RGC to INP transition. Furthermore, our data suggest that Neurog2 represses Pax6 transcription and an RGC identity through indirect means, a model we tested further using transcriptional assays.
Neurog2 Acts as a Transcriptional Activator to Repress Pax6 Transcription
There are at least 2 indirect ways by which Neurog2 may repress Pax6: 1) By inducing the expression of a downstream transcriptional repressor or 2) by sequestering co-activators away from Pax6 regulatory elements. To understand how Neurog2 functions, we implemented a transcriptional reporter assay in P19 cells, using previously validated promoter and enhancer sequences for Pax6 that drive reporter expression in the dorsal telencephalon (Kammandel et al. 1999). As expected, both Neurog2-WT (5.7-fold decrease; n = 3; P < 0.0001) and Neurog2-VP16 (20.8-fold decrease; n = 3; P < 0.0001) repressed Pax6 transactivation below control levels (n = 3; Fig. 2S). These data confirm that Neurog2 functions as a transcriptional activator to repress Pax6, indicating that the repression must be at least in part indirect (i.e., Neurog2 does not directly bind to Pax6 regulatory elements). Given that Tbr2 is a direct transcriptional target of Neurog2, and that Tbr2 is able to repress Pax6 cell autonomously (Sessa et al. 2008), the most parsimonious model is that Neurog2 indirectly suppresses Pax6 transcription by initiating Tbr2 transcription, which then feeds forward to repress Pax6. Consistent with this model, Tbr2 also repressed the Pax6 reporter (16.7-fold; n = 3; P < 0.0001; Fig. 2S).
Notably, the repressor form of Neurog2 (-EnR; 3.1-fold decrease; n = 3; P < 0.0001), which can bind DNA and should repress transcription, and a mutant form of Neurog2 that does not bind DNA (Neurog2-AQ; 1.5-fold decrease; n = 3; P < 0.0001), was also able to inhibit Pax6 transactivation, albeit to a much lesser extent (Fig. 2S). This data suggested that Neurog2 may also suppress Pax6 expression at least in part through indirect methods, such as the CBP sequestration model shown for RhoA (Ge et al. 2006) and GFAP (Sun et al. 2001) reporters. However, CBP was not able to rescue the inhibition of Pax6 transactivation by Neurog2-AQ, suggesting that other co-activator or activator proteins must be involved in this indirect mode of repression, at least in vitro (Fig. 2S).
Taken together, our in vivo and in vitro data support the idea that Neurog2 regulates the RGC to INP transition of cortical progenitors through a simple repressor loop, directly initiating Tbr2 transcription, which in turn acts as a direct transcriptional repressor of Pax6. However, other indirect mechanisms may also contribute to the repressive effects of Neurog2 on Pax6 expression and an RGC identity.
Neurog2 Functions as a Transcriptional Activator to Initiate a Cortical-Specific Neuronal Differentiation Program
Neurog2 and the related gene Neurog1 promote precocious neuronal differentiation when misexpressed in cortical progenitors, directly initiating the transcription of several neuronal differentiation genes both in vitro and in vivo (Schuurmans et al. 2004; Mattar et al. 2008). To confirm that Neurog2 functions as a transcriptional activator to induce cortical neurogenesis, we electroporated E12.5 neocortices with expression constructs for each Neurog2 fusion protein, and then quantitated the number of GFP+ cells that co-expressed the panneuronal marker NeuN or the cortical-specific neuronal markers Bhlhb5 (Beta3/Bhlhe22) and Tbr1 (Hevner et al. 2001; Joshi et al. 2008) at 24 h and 72 h post-electroporation. As expected, both Neurog2-WT and Neurog2-VP16 induced the differentiation of more GFP+/NeuN+ neurons compared with pCIG2 controls (Fig. 3A–H,U,U′; Supplementary Table S2). These ectopic neurons expressed cortical-specific markers, as evidenced by the similar increases in GFP+/Bhlhb5+ and GFP+/Tbr1+ double-positive cells in Neurog2-WT and Neurog2-VP16 transfections compared with empty vector controls (Fig. 3I–P,V,V′,W,W′, Supplementary Table S2). Conversely, Neurog2-EnR suppressed neurogenesis, reducing the numbers of GFP+/NeuN+, GFP+/Bhlhb5+, and GFP+/Tbr1+ neurons that differentiated (Fig. 3D,H,L,P,T,U–W,U′–W′, Supplementary Table S2). Furthermore, Neurog2-WT and Neurog2-VP16 reduced the number of GFP+ cells expressing the pan-proliferative marker Ki67+, while Neurog2-EnR did not influence the overall number of cycling cells (Fig. 3Q–T,X, Supplementary Table S2). Neurog2-WT and Neurog2-VP16 thus have similar capacities to promote cell cycle exit and neuronal differentiation in the dorsal telencephalon, acting in an opposing fashion to Neurog2-EnR, which inhibits neurogenesis.
We next asked if Neurog2 similarly functions as a transcriptional activator to induce a cortical identity when misexpressed in an ectopic site. In E12.5 → E15.5 electroporations of the ventral telencephalon, both Neurog2-VP16 and Neurog2-WT (as shown in Mattar et al. 2008) had a similar capacity to promote the differentiation of NeuN+ neurons (Fig. 4A′–D′,E) that expressed cortical-specific markers, including Tbr1 (Fig. 4A″–D″,F), Bhlhb5 (Fig. 4A′′′–D′′′,G), NeuroD6 (Supplementary Fig. S2M–P), Sox5 (Supplementary Fig. S2Q–T), and Nhlh2 (Supplementary Fig. S2U–X; Supplementary Table S3). Notably, following Neurog2-WT and Neurog2-VP16 overexpression, GFP+ cells formed distinct cellular aggregates or heterotopias in the ventral telencephalon, consistent with an alteration in the neuronal identity and hence, adhesive properties (Fig. 4B–B′′′,C–C′′′; Supplementary Fig. S2F,G,J,K). In contrast, pCIG2 and Neurog2-EnR-transfected GFP+ cells were sparsely distributed in the ventral telencephalon, suggesting that neither construct altered the regional identity or adhesive properties of the transfected cells (Fig. 4A–A′′′,D–D′′′; Supplementary Fig. S2E,H,L). Furthermore, Neurog2-EnR did not alter the number of GFP+/NeuN+, GFP+/Bhlhb5+, or GFP+/Tbr1+ neurons that differentiated in the ventral telencephalon compared with pCIG2 controls (Fig. 4D–D′′′, Supplementary Table S3).
Taken together, these data indicate that Neurog2 does indeed function as a transactivator to initiate cortical-specific neuronal differentiation, both in its normal territory and in ectopic sites in the telencephalon.
Neurog2 Functions as a Transcriptional Activator to Suppress Ventral Differentiation Programs, in Part Via a Neurog2-Tbr2-Ebf1 Transcriptional Cascade
Neurog2 is not only required to specify a dorsal, glutamatergic neuronal identity, but also to repress an alternative ventral, GABAergic fate in neocortical progenitors (Fode et al. 2000; Schuurmans et al. 2004). We thus asked if Neurog2 functions as a transcriptional activator or repressor to suppress ventral identities in telencephalic progenitors by performing E12.5 → E15.5 ventral electroporations. Both Neurog2-WT and Neurog2-VP16 efficiently suppressed the expression of several ventral neuronal genes, including Ebf1, Six3, and GAD1, while pCIG2 and Neurog2-EnR had no effect on the expression of these ventral markers (Fig. 4H–S). Conversely, in E12.5 → E15.5 electroporations of the dorsal telencephalon, neither Neurog2-WT, Neurog2-VP16, nor Neurog2-EnR influenced the expression of “ventral” markers, including those expressed by GABAergic interneurons (Dlx1, Ebf1, Six3, and GAD1) and oligodendrocytes (Olig2; Supplementary Fig. S3), the exception being Ascl1, which was induced by Neurog2-EnR (Fig. 6, Supplementary Fig. S3L, see below).
Taken together, these data indicate that Neurog2 functions as a transcriptional activator to suppress a ventral telencephalic identity, indicative of an indirect mode of repression. To explore the indirect repression mechanism further, we first examined Tbr2, given our demonstration above that this gene functions as a downstream effector of Neurog2-mediated repression. We first asked if Tbr2 had any capacity to suppress ventral differentiation programs in the neocortex by electroporating Tbr2 into E12.5 cortical progenitors. As previously reported, misexpression of Tbr2 does not initiate cortical neuronal differentiation (Sessa et al. 2008), as exemplified by the lack of induction of Tbr1 expression (Fig. 5A,A′,B–B″). However, in E12.5 → E15.5 electroporations of the ventral telencephalon, Tbr2-overexpressing cells formed tight cellular aggregates (Fig. 5C), similar to the heterotopias evident following the misexpression of Neurog2 in ventral domains (Fig. 4). The clustering of Tbr2-overexpressing cells was suggestive of a change in the regional identity and adhesive properties of these cells. Strikingly, we observed a clear repression of the ventral marker Ebf1 in the Tbr2-transfected patch (Fig. 5D vs. control side in D′), but surprisingly, other components of the ventral differentiation program, such as Six3 and GAD1 (Fig. 5E,F), were expressed normally in Tbr2-overexpressing cells.
To further investigate regulatory interactions between Tbr2 and Ebf1, we examined Ebf1 expression in Tbr2 mutant cortices. In E14.5 wild-type embryos, Ebf1 expression was restricted to the mantle zone of the ventral telencephalon, with no expression detected in dorsal telencephalic domains, not even in migrating cortical interneurons (Fig. 5H,H′). Strikingly, in E14.5 Tbr2 mutants, Ebf1 was ectopically expressed in the dorsal telencephalon, specifically within the SVZ (Fig. 5I,I′), where Tbr2 is normally expressed (Fig. 5G,G′). We reasoned that the ectopic expression of Ebf1 in Tbr2 mutants was most likely due to a cell autonomous requirement for Tbr2 to repress Ebf1, rather than the ectopic migration of Ebf1-expressing cells from ventral to dorsal domains, as interneuron migration is strongly reduced in Tbr2 mutants (Sessa et al. 2010). To test if Tbr2 was indeed a direct repressor of Ebf1 transcription, we used 2 approaches. First, we performed luciferase assays using previously generated reporters containing regulatory sequences from the proximal and distal promoters of Ebf1 (Roessler et al. 2007). We found that Tbr2 was able to suppress transactivation of the Ebf1 proximal promoter (1.8-fold decrease; n = 3; P < 0.0001; Fig. 5K) and not the distal promoter (n = 3; P > 0.05; Fig. 5J). Accordingly, multiple conserved Tbr2-binding sites were identified in the Ebf1 proximal promoter region (Fig. 5L). To test if Tbr2 could directly bind the Ebf1 proximal promoter, we performed anti-Tbr2 ChIP experiments. Tbr2 was specifically enriched on Ebf1 proximal promoter sequences in chromatin isolated from E14.5 cortices (Fig. 5M,N).
Taken together, we have identified a novel Neurog2-Tbr2-Ebf1 repressor loop in the embryonic neocortex that involves direct transcriptional interactions, with Neurog2 turning on the expression of Tbr2 (Ochiai et al. 2009), and Tbr2 in turn binding to and repressing the transcription of Ebf1.
Neurog2 Indirectly Suppresses Ascl1 Expression Through DNA-Binding–Dependent and –Independent Mechanisms
We were struck by the inability of Tbr2 to repress all ventral genes, such as Six3 and GAD1, which are also negatively regulated by Neurog2 (Fode et al. 2000; Schuurmans et al. 2004). This suggested that Tbr2 was not the only downstream effector in Neurog2-mediated repressor cascades. To further investigate how Neurog2 represses ventral fates, we focused on Ascl1, a proneural bHLH gene that is negatively regulated by Neurog2 in cortical progenitors (Fode et al. 2000; Schuurmans et al. 2004; Britz et al. 2006; Mattar et al. 2008). Ascl1 is necessary and sufficient to promote the differentiation of oligodendrocytes (Parras et al. 2007) and neurons with a ventral regional identity and GABAergic neurotransmitter phenotype (Casarosa et al. 1999; Fode et al. 2000; Schuurmans et al. 2004; Britz et al. 2006; Mattar et al. 2008). To first test if Neurog2 regulates Ascl1 cell autonomously, we examined Ascl1 expression in heterozygous and homozygous E15.5 Neurog2GFP knock-in (KI) embryos. While very few GFP+/Ascl1+ double-positive progenitors were observed in Neurog2GFPKI/+ cortices (Fig. 6A,B,B′), many double-positive cells were observed in Neurog2GFPKI/GFPKI cortices (Fig. 6C,D,D′), indicating that Ascl1 is upregulated within Neurog2 mutant cortical cells. We next asked if Neurog2 functions as a transcriptional activator or repressor to regulate Ascl1. In E12.5 → E13.5 electroporations of the dorsal telencephalon, both Neurog2-WT and Neurog2-VP16 reduced the number of GFP+/Ascl1+ progenitors compared with control transfections (Fig. 6E–G,E′–G′,M; Supplementary Table S2). Conversely, Neurog2-EnR induced the ectopic expression of Ascl1 at both the transcript (Fig. 6K,L vs. 6I,J control transfection; Supplementary Fig. S3L) and the protein level (Fig. 7H,H′,M; Supplementary Table S2), promoting a 3.9-fold increase in the percentage of GFP+/Ascl1+ cells. Neurog2 thus acts as a transcriptional activator to cell autonomously suppress Ascl1 expression.
To understand the mechanism(s) by which Neurog2 indirectly suppresses Ascl1 expression, we subcloned 5 kb of the Ascl1 upstream regulatory and promoter sequences into a luciferase reporter. Consistent with an indirect mode of transcriptional repression, both Neurog2-WT (3.9-fold decrease; n = 3; P < 0.0001) and Neurog2-VP16 (7.5-fold decrease; n = 3; P < 0.0001) repressed Ascl1 transactivation below control levels within 24 h post-transfection of P19 cells, whereas Neurog2-EnR (n = 3; P > 0.05) had no effect on the Ascl1 reporter (Fig. 6N). To further probe how Neurog2 represses Ascl1, we also examined the activities of Neurog2 fusion proteins carrying a mutation in the DNA-binding domain (AQ mutants; Sun et al. 2001). While Neurog2-WT-AQ (1.5-fold decrease; n = 3; P < 0.01) and Neurog2-VP16-AQ (2.9-fold decrease; n = 3; P < 0.0001) had some ability to repress the Ascl1 reporter, their repressive activities were less than the proteins that retained their DNA-binding capacity (Fig. 6N). Furthermore, CBP had no capacity to rescue Neurog2-mediated repression of the Ascl1 reporter, suggesting that this sequestration model does not apply to Ascl1 regulation (Fig. 6O). Thus, Neurog2 represses Ascl1 transcription through novel DNA-binding–dependent and –independent mechanisms. We focused on identifying the novel DNA-binding–dependent mechanisms of Neurog2-mediated repression.
Etv1 is a Neurog2-Regulated Gene That Regulates Expression of the Ventral Identity Determinant Ascl1
Our data suggested that Neurog2 functions as a transcriptional activator to repress Ascl1 expression, but the intermediaries of this indirect regulation are unknown. To identify such molecules, we examined the expression of several candidate transcriptional regulators in E13.5 Neurog2 mutant neocortices, at a stage when Ascl1 expression is upregulated (Fode et al. 2000; Schuurmans et al. 2004). Of the genes tested, Tbr2 (Fig. 7A,A′,B,B′) and Sox5 (Fig. 7C,C′,D,D′) were expressed at reduced levels in Neurog2 mutant neocortices, while Pax6 (Fig. 7E,E′,F,F′), Etv1 (Fig. 7G,G′,H,H′), and Sox6 (not shown) were upregulated (Schuurmans et al. 2004; Britz et al. 2006; Azim et al. 2009). Conversely, in E12.5 electroporations, Neurog2 and Neurog2-VP16 increased Tbr2 (Fig. 2J,J′,K,K′,P,R) and Sox5 (Supplementary Fig. S2R,S) expression in the dorsal and ventral telencephalon, respectively, while Pax6 (Fig. 2F,F′,G,G′,Q) and Etv1 (Fig. 7I,I′,J,K) expression were repressed by Neurog2 in cortical progenitors.
We predicted that those transcription factors positively regulated by Neurog2 (i.e., Tbr2, Sox5) might repress Ascl1 expression, while those negatively regulated by Neurog2 (i.e., Etv1, Pax6, Sox6) may induce Ascl1 transcription. To directly test this, we used the 5 kb Ascl1 transcriptional reporter assay in P19 cells. Introduction of Tbr2, Sox5, Sox6, and Pax6 did not have significant effects on Ascl1 reporter activity, suggesting that these transcriptional regulators do not directly repress Ascl1 transcription—at least not through the 5 kb of Ascl1 regulatory sequence tested (Fig. 7L). Because of the inherent limitations of this in vitro assay, we did further tests to determine if Tbr2 repressed Ascl1 expression, even though it did not repress the Ascl1 reporter in vitro (n = 3; P > 0.05; Fig. 7L). In E12.5 electroporations of the ventral telencephalon, Tbr2 did not reduce the number of progenitors that expressed Ascl1 (pCIG2 control: 57.45 ± 2.11% GFP+Ascl1+/GFP+, n = 5 vs. Tbr2: 50.64 ± 5.52% GFP+Ascl1+/GFP+, n = 3; P = 0.21; 16h post-electroporation). Moreover, Ascl1 expression was not altered in Tbr2 mutant cortices (Robert F. Hevner, unpublished observation). We thus conclude that Tbr2 is not a transcriptional repressor for Ascl1, validating the results of our in vitro reporter assay.
Of the transcription factors repressed by Neurog2, Etv1 was the only factor to increase Ascl1 reporter activity (2.0-fold increase; n = 3; P < 0.0001; Fig. 7L). Accordingly, in E12.5 → E15.5 Etv1 cortical electroporations, ectopic Ascl1 expression was detected in the transfected patch (Fig. 8A–E). We thus conclude that Etv1 is able to initiate ectopic Ascl1 expression, and suggest that Neurog2 may be required to repress Etv1 so that Ascl1 is not ectopically expressed in cortical cells.
Etv1 Represses Hes5 Expression to Suppress Ascl1 Transcription
Etv1 expression is initiated downstream of fibroblast growth factor signaling, which has recently been shown to regulate Notch signaling (Akazawa et al. 1992). We thus speculated that Etv1 may regulate Ascl1 expression via regulatory interactions with the Notch pathway. Indeed, the Notch effector proteins Hes1 and Hes5 are well known suppressors of proneural gene expression in neural progenitors (e.g., Arber et al. 2000; Nieto et al. 2001; Hansen et al. 2010), with Hes1 directly recruiting TLE co-repressors to Ascl1 regulatory elements (Ju et al. 2004). We first examined if like Etv1, Hes5 expression levels were regulated by Neurog2. Hes5 transcript levels were reduced in the dorsal telencephalon of E13.5 Neurog2 mutants compared with wild-type controls (Fig. 8F,G). Moreover, Neurog2 induced ectopic Hes5 expression in E12.5 → E15.5 cortical electroporations (Fig. 8H–K). Finally, Neurog2 induced a 10-fold increase (n = 3; P < 0.0001) in the activity of a 3 kb Hes5-luciferase reporter (Nieto et al. 2001). Neurog2 is thus required and sufficient to promote Hes5 expression in cortical progenitors.
To determine if Etv1 also regulates Hes5 transcription, we generated obligate activator (Etv1-VP16) and repressor (Etv1-EnR) fusions and tested their activities in vivo and in vitro. In E12.5 → E13.5 cortical electroporations, Etv1-VP16 repressed Hes5 expression (Fig. 8L–N), while conversely, Etv1-EnR led to the ectopic expression of Hes5 (Fig. 8O–Q). A similar trend was observed in reporter assays in vitro, with Etv1 (n = 3; P = 0.02) and Etv1-VP16 (n = 3; P < 0.0001) repressing while Etv1-EnR (n = 3; P < 0.0001) activated a Hes5 3 kb transcriptional reporter (Fig. 8S). Furthermore, we confirmed that Hes5 was able to repress transactivation of the Ascl1 reporter in vitro (Fig. 8T), while Etv1 (n = 3; P < 0.0001) and Etv1-VP16 (n = 3; P = 0.03) activated this reporter (Fig. 8U). Unexpectedly, Etv1-EnR was also a strong activator of the Ascl1 reporter activity (2.1-fold increase; n = 3; P < 0.0001; Fig. 8U), despite the strong induction of Hes5 expression by this construct. Moreover, both Etv1-VP16 and Etv1-EnR were able to induce ectopic Ascl1 expression in E12.5 → E15.5 electroporations (Supplementary Fig. S4). While Etv1-EnR's ability to induce ectopic Ascl1 expression was unexpected, a previous report has demonstrated that Hes5 is converted to an activator of Ascl1 transcription with the onset of neuronal differentiation (Ju et al. 2004).
In summary, we have uncovered a novel repressor cascade in neocortical progenitors, demonstrating that Neurog2 indirectly suppresses Etv1 expression, which in turn indirectly represses Ascl1 transcription, possibly through interactions with Notch signaling and the Hes genes (Fig. 9A).
Neurog2 encodes a proneural bHLH transcription factor that is required in the embryonic neocortex to promote progenitor cell maturation, neuronal fate specification, neuronal differentiation, and migration (Fode et al. 2000; Schuurmans et al. 2004; Hand et al. 2005; Britz et al. 2006; Ge et al. 2006; Heng et al. 2008; Pacary et al. 2011). To control these events, Neurog2 initiates the expression of several downstream genes and genetic programs, such as NeuroD1 and a dorsal glutamatergic fate, Tbr2 and an INP fate, and Rnd2 to promote radial migration (Huang et al. 2000; Seo et al. 2007; Heng et al. 2008; Ochiai et al. 2009; Pacary et al. 2011). Here, we investigated the regulatory logic underlying Neurog2′s ability to simultaneously repress alternative gene expression programs in neocortical lineages, focusing on the repression of Pax6 and a RGC fate, and Ascl1 and a ventral GABAergic neuronal identity (Fode et al. 2000; Schuurmans et al. 2004; Britz et al. 2006; Ge et al. 2006). Our studies confirm that Neurog2 functions as a transcriptional activator to carry out its neocortical functions, including its ability to repress alternative cell fates. Furthermore, we identify Tbr2, which is a direct Neurog2 transcriptional target (Ochiai et al. 2009), as a key effector of Neurog2-mediated transcriptional repression. Specifically, we show that Tbr2 represses both the transcription of Pax6 and an RGC fate, and Ebf1 and a ventral neuronal identity. Finally, we identify Etv1 as a Neurog2-repressed gene that can indirectly promote Ascl1 expression, both in vitro and in vivo. Mechanistically, we suggest that Etv1 may indirectly promote Ascl1 expression through its ability to influence Hes5 transcription, which in turn regulates Ascl1 (Ju et al. 2004). Our studies highlight the importance of transcriptional repression in controlling cell fate decisions in the neocortex, and provide evidence for a derepression mode of cell fate specification, whereby cortical cells acquire their identities at least in part by repressing alternative cell fates.
Tbr2 is a Downstream Effector in the Neurog2-Mediated Repression of a RGC Identity
The progression from RGC to INP involves the activation and repression of distinct genetic pathways. Here, we found that Neurog2 functions as a transcriptional activator to promote Tbr2 transcription in INPs and to repress Pax6 expression in RGCs, thereby promoting the RGC to INP transition. The ability of Neurog2-WT and the activator fusion (Neurog2-VP16) to induce Tbr2 expression was expected as Tbr2 is a direct transcriptional target of Neurog2 (Ochiai et al. 2009). In contrast, the mechanism by which Neurog2 represses Pax6 expression was previously unknown. We found that Neurog2 functions through indirect means to repress Pax6 transcription. The most parsimonious model that fits our data is that Neurog2 induces the expression of Tbr2, which is known to be a direct Neurog2 transcriptional target (Ochiai et al. 2009), and that Tbr2 in turn functions as a transcriptional repressor of Pax6. Indeed, in our transcriptional reporter assay, Tbr2 could repress Pax6 transcription. Furthermore, misexpression of Tbr2 in cortical progenitors suppresses Pax6 expression in a cell autonomous fashion (Sessa et al. 2008).
In previous gain-of-function studies, Tbr2 misexpression was shown to sustain INP proliferation, delaying cell cycle exit and partially blocking the expression of neuronal markers (Sessa et al. 2008). This is consistent with our data, where we found that Tbr2 did not induce the expression of the cortical neuronal marker Tbr1 in the ventral telencephalon. Furthermore, we found that misexpression of Neurog2-VP16, a strong transcriptional activator, led to a sustained upregulation of Tbr2 expression after 72 h post-electroporation, which was accompanied by a delay in the onset of Tbr1 expression. The zinc finger transcriptional repressor Insm1 is another Neurog2-regulated gene that is sufficient to promote an INP fate, but does not induce neuronal differentiation when misexpressed in cortical progenitors (Farkas et al. 2008). Tbr2 and Insm1 thus behave very differently than Neurog2 in gain-of-function assays, as Neurog2 induces neuronal differentiation and cell cycle exit. If Neurog2′s sole function is to induce neuronal differentiation, the rationale for inducing the transcription of downstream genes (e.g., Tbr2, Insm1) that block this process is at first glance contradictory. However, our study supports the idea that Tbr2 (and possibly Insm1 by extension) plays an essential role in stabilizing the INP progenitor state, and that they achieve this in part by repressing alternative gene expression programs (e.g., Pax6 and an RGC identity). This allows for the orderly progression through different progenitor states and stages of differentiation.
Given that Tbr2 and Insm1 do not promote neuronal differentiation, additional Neurog2-regulated genes must promote the conversion of INPs to differentiated neurons. Possibilities include the Neurog2-regulated genes NeuroD1 and NeuroD4 (Mattar et al. 2008), which may be required to bias Tbr2+ INPs to undergo differentiative, neurogenic divisions. In such a model, Neurog2 might rapidly induce the expression of Tbr2 to suppress an RGC fate and promote an INP identity, followed by the activation of NeuroD1/D4 (and other genes) to promote neuronal differentiation once the INP fate is stabilized. Consistent with such a sequential model, we previously showed that it takes 72 h for Neurog2 to induce Neurod1 expression in the ventral telencephalon (Mattar et al. 2008), whereas in this study, we found that Neurog2 can induce Tbr2 transcription in ventral domains within 24 h post-electroporation.
Tbr2 is a Downstream Effector in the Neurog2-Mediated Repression of an Ebf1+ Ventral Neuronal Identity
To promote the differentiation of glutamatergic projection neurons, Neurog2 must simultaneously activate cortical-specific neuronal genes while suppressing the acquisition of a ventral, GABAergic neuronal fate. A failure to suppress competing ventral developmental programs leads to the misspecification of cortical lineages in several mutants, including Neurog2, Emx2, and Pax6 loss-of-function mice (Fode et al. 2000; Muzio et al. 2002; von Frowein et al. 2006; Mangale et al. 2008). Here, we demonstrated that Neurog2 functions as a transcriptional activator to induce the expression of several cortical markers (e.g., NeuroD6, Sox5, Nhlh2, Tbr1, Bhlhb5). Furthermore, we showed that Neurog2 represses a ventral identity through indirect means given that both Neurog2 and Neurog2-VP16 can block the expression of Ascl1 and markers expressed in ventral interneuron lineages (Ebf1, Six3, GAD1). How does Neurog2 repress ventral marker expression by functioning as an activator? Our study demonstrates that Neurog2 regulates the expression of several transcriptional regulators that may contribute to the repression of ventral marker expression. Included is Tbr2, which we found plays a critical yet specific role in repressing a ventral neuronal identity, suppressing the expression of Ebf1 and not other genes expressed in ventrally-derived neurons. The specific requirement for Tbr2 was verified by examining Tbr2 mutant cortices, demonstrating that Ebf1 was ectopically expressed. This ectopic expression is unlikely due to the aberrant migration of ventrally-derived, Ebf1+ interneurons into the Tbr2 mutant neocortex as interneuron migration is reduced rather than elevated in Tbr2 mutants (Sessa et al. 2010).
Ascl1 is upregulated in Neurog2 mutant cortices, where it is an essential component of the pathways underlying the ventral misspecification of cortical lineages (Fode et al. 2000; Schuurmans et al. 2004). Interestingly, misexpression of Neurog2-EnR in the cortex, which blocks Neurog2 function, similarly promotes the ectopic expression of Ascl1 and not other ventral markers, such as Dlx and GAD1. This is reminiscent of Sox6 mutant cortices, in which Ascl1 is upregulated, while downstream genes such as Dlx1 and GAD1 are not (Azim et al. 2009). One possibility is that the ectopic expression of Ascl1, which is required and sufficient to promote the expression of ventral differentiation programs (Casarosa et al. 1999; Britz et al. 2006), does not reach sufficiently high levels in Neurog2-EnR transfections (or in Sox6 mutants). Alternatively, in our E12.5 cortical transfections, Ascl1 may not be induced by Neurog2-EnR sufficiently early, as the competence window for cortical progenitors to be respecified in response to Ascl1 is short, with only early-born neocortical neurons (i.e., <E14.5) responding to ectopic Ascl1 in transient (i.e., in utero electroporation) and chronic (Neurog2KIAscl1 allele) gain-of-function experiments (Parras et al. 2002; Britz et al. 2006).
Using a candidate approach, we identified Tbr2 and Sox5 as 2 potential transcriptional repressors that were downregulated in Neurog2 mutant cortices, and could thus be involved in suppressing a ventral telencephalic identity in cortical neurons. Surprisingly, however, Tbr2 was not able to suppress the expression of Ascl1, either in our transcriptional assay in vitro, or in cortical progenitors in vivo, suggesting that other Neurog2-regulated genes carry out this function. Furthermore, Sox5 was not able to repress our Ascl1 transcriptional reporter, and while this is an artificial system, both Sox5 and Ascl1 are upregulated in Sox6 mutant cortices (Azim et al. 2009), making it unlikely that Sox5 represses Ascl1 directly, at least in the absence of Sox6. Interestingly, Sox6 was upregulated in Neurog2 mutant cortices, along with Pax6 and Etv1, but of these factors, only Etv1 was able to transactivate the Ascl1 reporter. In an effort to better understand how Etv1 regulates Ascl1 transcription, we examined potential indirect interactions with Hes5. Our data suggest that Etv1 may contribute to the repression of Ascl1 within cortical progenitors by indirectly downregulating Hes5, which is itself a direct transcriptional repressor of Ascl1, at least in dividing neural progenitors (Ju et al. 2004). Thus, in a Neurog2 mutant cortex in which Etv1 is ectopically expressed, Hes5 expression is reduced, thereby reducing the transcriptional repression of Ascl1. While our data support a model, whereby Etv1 regulates Hes5 transcript levels, we cannot rule out the possibility that Neurog2 regulates Hes5 expression more directly, through its ability to initiate transcription of the Dll1 Notch ligand. However, there are precedents for other transcription factors directly regulating Hes5 gene expression independent of Notch signaling, most notably the zinc finger transcription factors Fezf1 and Fezf2 (Nieto et al. 2001).
In other developmental systems such as the spinal cord, transcriptional repressors specify cell fates by repressing the expression of other transcriptional repressors. This implies that cells acquire their identities by repressing alternative cell fates—a paradigm that has been termed the derepression mode of cell fate specification (Muhr et al. 2001; Bylund et al. 2003). In this model, inappropriate cell fates are inhibited by transcriptional repressors. Here, we found some parallels in the neocortex, demonstrating that Neurog2 functions as a transcriptional activator to ensure that alternative differentiation pathways (i.e., ventral programs) are not derepressed, and a cortical identity is specified.
This work was supported by a CIHR (MOP-44094) Operating Grant to C.S. C.S. is an Alberta Heritage Foundation for Medical Research (AHFMR) Senior Scholar. C.K., P.M., S.L., and G.W. were supported by a CIHR Training Grant in Genetics, Child Development and Health. P.M. was supported by a Heart and Stroke Foundation Studentship, an AHFMR Studentship, a CIHR Canada Graduate Scholarship, a Killam Trust Award, a Lionel E. McLeod Health Research Scholarship (AHFMR). C.K. also held a QEII Scholarship. R.D. was supported by a CIHR Canada HOPE Fellowship.
We thank Jennifer Chan, Olivier Britz, Sarah Childs, Erica Watson, and Dawn Zinyk for critical reading of the manuscript. We would also like to thank Lisa Marie Langevin, Natasha Klenin, and Dawn Zinyk for technical support. Conflict of Interest: None declared.