The cerebral cortex is the most complex structure of the mammalian brain. It is the site of numerous higher-order sensory, motor, and cognitive functions. Cortical function relies on the proper formation of specialized cortical areas as well as on their sophisticated interconnections (Glasser et al., 2016).

During development, regionalization of the embryonic brain is achieved through multi-step processes. Sources of diffusible signaling molecules act as organizing centers and pattern neighboring domains through regulation of specific transcription factors expression, thereby creating molecular compartments that lead to the generation of distinct cortical fields (Rubenstein et al., 1998; Sur and Rubenstein, 2005; O'Leary et al., 2007).

Cortical projection neurons are generated in the germinal zones (GZ) of the dorsal telencephalon and, following cell-cycle exit, migrate radially to the cortical plate. Previous work has shown that regional differences in the proliferative programs in the GZ have far reaching consequences for histogenesis of cortical areas (Dehay et al., 1993; Polleux et al., 1997; Lukaszewicz et al., 2005).

Neuron number and types specific of each cortical layer and area are defined by the fine-tuned balance between proliferation and differentiation of cortical progenitor cells. While cell biology mechanisms underlying the switch from proliferative to differentiative divisions have been identified (Fish et al., 2006; Delaunay et al., 2014, 2017; Mora-Bermudez et al., 2014; Paridaen and Huttner, 2014; Matsuzaki and Shitamukai, 2015), it has been shown that the increasing fraction of progenitor cells that quit the cell cycle to embark on neuronal differentiation correlate with a lengthening of the G1 phase of the cell cycle (Takahashi et al., 1995; Calegari et al., 2005; Salomoni and Calegari, 2010; Arai et al., 2011). G1 phase is considered as a time window of susceptibility to differentiation signals (Mummery et al., 1987) and G1 phase lengthening increases the competence of a proliferating cell to withdraw from the cell cycle and to differentiate (Zetterberg et al., 1995).

During corticogenesis, proliferative and differentiative divisions are characterized by short and long G1 phases respectively (Dehay et al., 2001; Lukaszewicz et al., 2002, 2005; Calegari and Huttner, 2003; Dehay and Kennedy, 2007; Pilaz et al., 2009). Progression through G1 phase is regulated mainly by the kinase activity of Cyclin D/CDK4 and Cyclin E/CDK2 (Sherr and Roberts, 2004), both of which have been shown to play a key role in determining neuron number during mouse mid-corticogenesis (Lange et al., 2009; Pilaz et al., 2009). In particular, Cyclin D1 (Ccnd1) dynamic expression levels have been shown to be at the heart of a regulatory network that control the balance between cortical progenitor proliferation and differentiation (Ghosh et al., 2014).

Here we have explored the transcriptional regulation of Ccnd1 expression during early corticogenesis. Numerous transcriptional factors binding to the Ccnd1 promoter have been identified (Klein and Assoian, 2008). It is targeted by TCF/βcatenin complex (Shtutman et al., 1999; Tetsu and McCormick, 1999), effector of the Wnt pathway, which plays a key role in regulating cortical expansion (Chenn and Walsh, 2003). More recently, it has been reported that the transcription factor PAX6, known to regulate proliferation and differentiation of cortical progenitors (Warren et al., 1999; Estivill-Torrus et al., 2002; Quinn et al., 2007; Sansom et al., 2009; Mi et al., 2013) binds to the Ccnd1 locus (Sun et al., 2015).

PAX6 plays a key role in forebrain patterning and cortex arealization (Stoykova et al., 1997; Bishop et al., 2000, 2002; Muzio et al., 2002; Englund et al., 2005). Interestingly, Pax6 shows a complementary expression pattern to the transcription factor Sp8 in the developing pallium with a rostro-ventral^High gradient (Sahara et al., 2007; Borello et al., 2014). SP8 is a zinc finger transcription factor belonging to the Sp-family of transcription factors (Zhao and Meng, 2005). SP8 acts downstream of FGF8 signaling (Storm et al., 2006), regulates forebrain patterning and cortical arealization (Sahara et al., 2007; Zembrzycki et al., 2007; Borello et al., 2014), and regulates cortical progenitor cell differentiation (Borello et al., 2014). Interestingly, SP5/SP8 have been shown to act as co-activators of the Wnt pathway in mouse embryos and differentiating embryonic stem (ES) cells (Kennedy et al., 2016).

We have therefore sought to analyze the putative role of SP8, together with PAX6 and βcatenin, on the transcriptional control of Ccnd1. Our ChipSeq and mouse genetics analysis reveal that Ccnd1 is a target gene of SP8. We show that SP8 is a critical player in the regulation of Ccnd1 expression during in vivo mouse corticogenesis. SP8 is able to modulate the moderate repressive transcriptional activity exerted by PAX6 on the Ccnd1 exon 1 region in vitro. By contrast, we did not observe cooperation between SP8 and βcatenin on Ccnd1 activation from the promoter/5′end of the gene. Finally, we demonstrate that SP8 is able to specifically activate gene expression from the Ccnd1 exon 5 fragment, containing part of the 3′UTR, suggesting that the 3′-end of the Ccnd1 gene may be target of gene regulation at multiple levels, including transcription.

Cyclin D1 is a major cell-cycle regulator (Ekholm and Reed, 2000; Sherr and Roberts, 2004) and has been shown to be at the heart of a regulatory network controlling the balance between proliferation and differentiation in the cerebral cortex (Ghosh et al., 2014).

βcatenin is one of the main regulators of Ccnd1 expression (Shtutman et al., 1999; Tetsu and McCormick, 1999; Klein and Assoian, 2008). Our observations that Ccnd1 expression signal does not necessarily correlate to regions of high Wnt/βcatenin activity during early in vivo corticogenesis is consistent with the idea that activation of the Ccnd1 gene might be modulated by cooperation with other transcription factors. Indeed, Axin2, a direct target and recognized proxy of the Wnt/βcatenin pathway activity (Yan et al., 2001; Jho et al., 2002; Lustig et al., 2002; Kim et al., 2007; Al Alam et al., 2011; van Amerongen et al., 2012; Bowman et al., 2013), is strongly localized in the medial pallium where Ccnd1 expression is low or absent and weakly expressed in the dorsal pallium where Ccnd1 is highly expressed (Figures 1A,B and Figures S3C,D). Altogether these data suggest that while βcatenin regulates Ccnd1 expression during corticogenesis in vivo, other transcription factors are also at work to produce the observed Ccnd1 expression pattern in the dorso-medial pallium.

Sp8 and Ccnd1 expression patterns in the early mouse corticogenesis in vivo are consistent with a potential role of SP8 on Ccnd1 gene regulation. The present data confirm this hypothesis and show the identification of Ccnd1 as the first SP8 target gene.

SP8 binds on the Ccnd1 locus on regions of active chromatin, as indicated by the H3K27ac ChipSeq results. Interestingly, peaks with higher intensity were positioned at the promoter/exon1 region, and in exon 5, containing also the first 299 bp of the 3′UTR. Our findings were further confirmed by results from SP8 ChipSeq experiments using a second SP8 antibody (Table S1 and data not shown). When we tested the responsiveness of these regions to SP8 we found that SP8 was able to activate gene expression from the Ccnd1 Ex5 but not from the Ex1 fragment, containing the last 293 nucleotides of the Ccnd1 promoter.

These results are unexpected. The regulatory regions of the genome are generally considered to localize outside of the coding sequences to keep the regulatory and the coding codes separated. However, a theoretical study predicts that the human genome, compared to a synthetic string of DNA letters, could accommodate short functional regulatory motifs in the protein coding regions (Itzkovitz and Alon, 2007). In addition, different studies aimed at identifying regulatory regions in the genome found that a small percentage of these regulatory domains are located in the coding sequences (Cawley et al., 2004; Visel et al., 2009) and that they are functional (Ritter et al., 2012). Recently, a comprehensive study mapping transcription factor binding on human genome exons in many cells lines found that ~15% of human codons specify both amino acids and transcription factor binding sites (Stergachis et al., 2013). Stergachis and colleagues suggest the fascinating hypothesis that the transcription factors binding to conserved sequences inside a gene exons have a role in codon choice and protein evolution. Numerous studies report that intergenic regulative regions like enhancers are sites of active transcription (De Santa et al., 2010; Kim et al., 2010; Natoli and Andrau, 2012; Shlyueva et al., 2014; Kim and Shiekhattar, 2015; Li et al., 2016), blurring the distinction between transcribed gene regions and regulative domains.

The Sp-family transcription factors bind preferentially GC and/or GT-reach regions in TATA-containing and TATA-less promoters and stimulate transcription by associating with the basal transcription complex and other transcription factors (Lania et al., 1997; Philipsen and Suske, 1999; Zhao and Meng, 2005). Consistently, the SP8 ChipSeq experiments showed that 78% of SP8 peaks correspond to gene promoters while genome wide SP8 binds only ~2% of gene exons and UTRs (see Figure S5 for details on the genome-wide SP8 binding localization). Interestingly, while our bioinformatics analysis identified several SP8 binding sites in the Ccnd1 promoter contained in the Ex1 fragment (Table S4 and Figure S1), the ChipSeq experiment indicates that the SP8 summit is located in the coding region. We hypothesize that SP8 binding on Ccnd1 exons is related to the fine-tuned regulation of Ccnd1 transcription, probably through a precise chromatin 3D structure, as well as to Ccnd1 mRNA maturation. There is also the possibility that SP8 is part of an epigenetic complex regulation of Ccnd1 locus replication and transcription. These questions will require further investigations.

Of interest, PAX6, which generally colocalizes with enhancers, binds Ccnd1 exon 1 (Sun et al., 2015). The predicted PAX6 binding site starts at position + 392 from the TSS (166 nucleotides downstream to the ATG), (Figures 2, 5A–C and Figure S1). Interestingly, while SP8 failed to directly regulate Ccnd1 Ex1 fragment expression, we show that SP8 was able to counteract the repressive activity exerted by PAX6 on the Ccnd1 Ex1 fragment in vitro. Moreover, the repressive activity we observed with PAX6 is consistent with the moderate increase of Ccnd1 mRNA observed in the Pax6 LOF E12.5 mutant forebrain (Mi et al., 2013; Sun et al., 2015).

SP8 could, therefore, interfere with PAX6 effect on Ccnd1 expression. It is possible that, due to the close proximity of the SP8 and PAX6 consensus, the two transcription factors compete for the binding on Ccnd1 exon 1. As mentioned above, Sp8 and Pax6 show opposite gradient of expression during early corticogenesis. At mid-gestation when Pax6 expression becomes homogeneous in the pallium, Sp8 expression is expressed at low levels. These data suggest that Ccnd1 is activated differentially by Sp8 and Pax6 in opposite domains of the pallium and is modulated by PAX6 and SP8 dosages along the neurogenic gradient.

SOX2 activates Ccnd1 in a dose-dependent manner during corticogenesis (Hagey and Muhr, 2014). SOX2, binding on different sites on the Ccnd1 locus and interacting with the TCF/βcatenin complex, regulates Ccnd1 expression and cortical progenitor cell mode of division and rate of differentiation (Hagey and Muhr, 2014). Considering the Sp8 graded expression in the pallial VZ and the strong Sp8 expression in the subpallial SVZ (Waclaw et al., 2006; Borello et al., 2014), one can hypothesize that a dose-dependent differential transcriptional regulation is also operant for SP8 (Figure 7).

In contrast to PAX6, βcatenin was able to activate transcription from the Ccnd1 5′UTR, and this activation was not dependent or modulated by SP8. The mouse Ccnd1 Ex1 fragment described here contains a TCF/LEF consensus that is conserved among different species, including human (Klein and Assoian, 2008), suggesting a critical and fundamental role for this site in Ccnd1 regulation. In addition, Tetsu and colleagues showed that activation of the CCND1 human minimal promoter,−962CD1 (Albanese et al., 1995), by βcatenin depends on the presence of TCF binding sites but not of other transcription factors (Tetsu and McCormick, 1999). These observations are in agreement with our results showing that the exon 1 fragment, containing only the last 293 nucleotides of the mouse Ccnd1 promoter, was sufficient to support βcatenin activity. The fact that βcatenin activity was independent of SP8 indicates that these two transcription factors do not cooperate by binding the Ccnd1 exon 1 region. However, a potential cooperation between βcatenin and SP8 binding to different Ccnd1 exon fragments (i.e., exon 5) needs further investigation.

Our results show that SP8 is able to specifically activate gene expression from the Ccnd1 Ex5 fragment. Consistently with our luciferase results, we found a cluster of putative SP8 binding sites at the end of the ORF in Ex5 fragment; this cluster overlapped with the SP8 summit identified in our ChipSeq experiments (Table S1 and Figure S2). These findings are very interesting as they rise the possibility that SP8 might control gene expression from binding to regions located at the 3′ end of the Ccnd1 gene in addition to the classical enhancer/promoter regulative domains located upstream of the target genes.

Human CCND1 3′UTR region has been shown to act as a critical regulatory element. Different miRNAs are predicted to bind human and mouse Ccnd1 3′UTR and regulate the level of Cyclin D1 expression (Deshpande et al., 2009; Ghosh et al., 2014); truncation or mutation of human CCND1 3′UTR alter the stability of the CCND1 transcript activating its oncogenic potential (Lebwohl et al., 1994; Molenaar et al., 2003; Wiestner et al., 2007; Deshpande et al., 2009; Ghosh et al., 2014). In addition, different Snps are present in the 3′UTR of mouse and human CCND1: Snp rs7178, localized on CCND1 3′UTR, is involved in neuroblastoma (Wang et al., 2011), and Snp rs7177, localized on CCND1 3′UTR, is involved in cognitive behavior (Rietveld et al., 2013). Considering that there is a 78.1% identity between human and mouse Ccnd1 3′UTR (as revealed using the ECR Browser Ovcharenko et al., 2004), these observations suggest a similar role in gene regulation and neurogenesis for the mouse Ccnd1 3′UTR.

Our data, showing that SP8 binds and specifically regulates Ccnd1 transcription from a region located at the end of the ORF in exon 5 and close to the 3′UTR, suggest that the 3′-end of the Ccnd1 gene may be a target of gene regulation at multiple levels, including the transcriptional one. The in vitro validation of the activity of SP8, as well as the interaction with PAX6 and βcatenin on the Ccnd1 locus, is based on an assay commonly used to screen the activity of genomic regulative regions. In addition, we provide further evidence based on manipulation of SP8 levels of expression in vivo in GOF and LOF transgenic mice as well as on RNAseq data that both clearly show a role for SP8 for Ccnd1 expression regulation at early stages of pallium development.

In summary, multiple signals regulate Ccnd1 transcription in mouse pallium during corticogenesis, resulting in a complex pattern of Ccnd1 expression. SP8 appears as a major player in this regulation, uncovering a potential novel role of the Sp-transcription factor family in transcription regulation, which awaits further analysis.

UB: Conceived the work; UB, BB, and ED: Collected and analyzed the data; UB, BB, DP, and CD: Wrote the paper.