|Home | About | Journals | Submit | Contact Us | Français|
The co-emergence of a six-layered cerebral neocortex and its corticospinal output system is one of the evolutionary hallmarks of mammals. However, the genetic programs that underlie their development and evolution remain poorly understood. Here we identify a conserved non-exonic element (E4) that acts as a cortex-specific enhancer for the nearby Fezf2, which is required for the specification of corticospinal neuron identity and connectivity. We find that SOX4 and SOX11 functionally compete with the repressor SOX5 in the trans-activation of E4. Cortex-specific double deletion of Sox4 and Sox11 leads to the loss of Fezf2 expression and failed specification of corticospinal neurons and, independent of Fezf2, a reeler-like inversion of layers. We show evidence supporting the emergence of functional SOX binding sites in E4 during tetrapod evolution and their subsequent stabilization in mammals and possibly amniotes. These findings reveal that SOX transcription factors converge onto a cis-acting element of Fezf2 and form critical components of a regulatory network controlling the identity and connectivity of corticospinal neurons.
The emergence and expansion of the neocortex in mammals has been crucial to the evolution of complex perceptual, cognitive, emotional and motor abilities1–3. The neocortex is organized into six layers based largely on the distinct subtypes of excitatory projection (or pyramidal) neurons and their patterns of connectivity4–9. Upper layer (L2-4) projection neurons form synaptic connections solely with other cortical neurons. In contrast, the majority of neurons in the deeper layers (L5 and L6) project to subcortical regions. Studies of laminar inversion in mice lacking the reelin (RELN) protein10–14 have shown that neuron identity and connectivity are determined by birth order rather than laminar position, suggesting that neuronal specification and positioning are separately encoded.
The layer-specific pattern of connectivity is dependent on cortical areas. The long-range projections of L5 neurons in somatosensory-motor areas form the corticospinal (CS) system that directly connects the neocortex (Ncx) with various subcortical regions15–22. A major component of the system, the CS (or pyramidal) tract, descends through the brainstem and into the spinal cord to provide a high degree of direct control over the precise motor functions affected in many clinical conditions21,22. Despite these important functional implications, the genetic programs controlling CS system development and evolution remain unclear.
Phenotypic specification and evolution of neuronal circuits depend on precise regulation of the timing, location, and level of gene expression23,24. Transcriptional control via cis-regulatory elements has emerged as a crucial mechanism25–29. The cis-regulatory mechanisms underlying the specification of distinct neuronal cell types and circuits, however, remain poorly understood. Specification of CS neurons and the formation of the CS tract critically depend on Fezf2 (or Fezl, Zfp312), which encodes a zinc-finger transcription factor (TF) highly enriched in early cortical progenitor cells and their deeplayer neuron progenies30–35. Inactivation of Fezf2 disrupts the molecular specification of deep-layer neurons and the formation of corticofugal projections, including the CS tract, without affecting the inside-out pattern of neurogenesis and lamination33–35. Misexpression of Fezf2 in L2-3 cortico-cortical projection neurons35,36 or striatal interneurons37 alters their molecular profile and induces ectopic subcerebral projections. These findings indicate that the tightly regulated transcription of Fezf2 is likely critical to the proper specification of distinct types of cortical projection neurons.
In this study, we used bacterial artificial chromosome (BAC) engineering and genetic inactivation in mice to identify and characterize a cortex-specific Fezf2 enhancer and its trans-regulators. We show that three SOX TFs converge onto the Fezf2 enhancer to control CS system development via functional binding sites that emerged in tetrapods. We also found Sox4 and Sox11 to be required for Fezf2-independent regulation of cortical RELN expression and laminar organization. Thus, these findings reveal novel developmental genetic programs that control layer formation and CS neuron identity, and the regulatory mechanisms by which they may have evolved.
Previously, it has been shown that the BAC transgenic mouse harboring 200 kb of the mouse Fezf2 locus and the Gfp reporter gene (Fezf2-Gfp38) recapitulates the spatio-temporal expression of endogenous Fezf2 39,40. This indicated that the cis-regulatory elements required for cortical Fezf2 expression are located within the BAC sequence. Based on the remarkable similarity in cortical expression pattern of Fezf2 between mouse and human39,41, we hypothesized that the regulatory elements are also highly evolutionarily conserved. Comparative sequence analysis revealed several conserved non-exonic elements (CNEEs) within the Fezf2-Gfp BAC (Fig. 1a). To test whether the selected CNEEs regulate Fezf2-Gfp expression, we generated multiple lines of BAC transgenic mice in which one of the CNEEs within the Fezf2-Gfp BAC was deleted (ΔE1-4) (Fig. 1b; Methods). We analyzed GFP expression in whole mount brain preparations and tissue sections of transgenic mice at embryonic and postnatal ages (Fig. 1c-g, and Supplementary Fig. 1, 2). We found no change in ΔE1-3 mutants, which expressed GFP in Ncx and pontine CS axons similar to wildtype Fezf2-GFP mice (Fig. 1d-f). In contrast, the deletion of CNEE E4 (ΔE4), located 7.3 kbp downstream of the Fezf2 transcription start site, resulted in a dramatic loss of GFP expression in the cortex, but not in the hypothalamus (Hyp) or olfactory bulb (OB) (Fig. 1g). Moreover, in ΔE4 mice, GFP-positive CS axon fascicles were lost from the ventral surface of the pons (Fig. 1g). Consistent with this finding, E4 was identified as an in vivo binding site of the enhancer-associated protein p300 (or EP300) in mouse embryonic forebrain27. Together, these results demonstrate that E4 is a cis-regulatory element acting as a cortex-specific enhancer of Fezf2.
To directly test the function of the E4 enhancer, we deleted the E4 region, while leaving intact the entire Fezf2 coding and proximal promoter regions, using homologous recombination in embryonic stem cells (Fig. 2a, b). Mice lacking E4 (E4−/−) were viable, fertile, and without overt behavioral or motor phenotypes. Quantitative (q) RT-PCR analysis revealed that Fezf2 expression is dramatically down-regulated in Ncx of these mice (Fig. 2c). Immunostaining of tissue sections of E4-null mice for PRKCG (or PKCγ) and L1CAM (or L1), two proteins expressed by CS neurons and their axons11, revealed that CS axons were absent from the pons (Fig. 2g), similar to mutant mice with a cortex-specific deletion of Fezf2 (Fezf2fl/fl;Emx1-Cre) (Fig. 2f). Furthermore, the expression of Bcl11b (or Ctip2), a gene functioning downstream of Fezf233,36, was also down-regulated in the E4-null Ncx (Supplementary Fig. 3). Taken together, these results indicate that the E4 enhancer is required for neocortical Fezf2 expression, molecular specification of L5/6 neurons, and the formation of CS tract.
We have previously shown that SOX5, a SOXD member of the large family of SOX TFs, binds and represses the transcriptional activity of the E4 enhancer39. Although somewhat surprising, SOX5 itself is also required for CS tract formation independent of its repression of Fezf239. Because different SOX TFs are known to compete for a common motif to mediate both activation and repression of regulatory elements42–44, we hypothesized that other SOX members may activate the E4 enhancer, perhaps acting competitively against SOX5-mediated repression. Consistent with this hypothesis, our sequence analysis using MatInspector (Genomatix) revealed eight putative SOX binding sites in E4. To prioritize which SOX TFs may be good candidates as potential trans-regulators of the E4 enhancer, we searched for those that are most highly correlated in their spatio-temporal expression pattern with FEZF2 using the www.humanbraintranscriptome.org database45. The highest correlated SOX genes were three members of the SOXC group (SOX4, SOX11, and SOX12) and SOX5 (Supplementary Fig. 4). These TFs play crucial roles in regulating cell fate and differentiation42–44, 46, and a de novo deletion of SOX11 was described in a patient with autism and intellectual disability47. Moreover, Sox4 and Sox11 act as transcriptional activators45 and their expression patterns overlap with that of Fezf2 in developing cortex (Supplementary Fig. 5).
To determine if SOX4 or SOX11 binds E4, we performed chromatin immunoprecipitation (ChIP)-PCR assays in Neuro-2a cells transiently expressing V5-tagged SOX4 or SOX11 (Methods). Anti-V5 antibodies precipitated E4, but not E1-3, DNA, confirming binding of SOX4 and SOX11 to the E4 enhancer (Fig. 3a). Moreover, recruitment of RNA polymerase II (Pol II) to E4 occurred in the presence of SOX4 or SOX11, which is consistent with increased transcriptional activity. To test the functional consequence of SoxC expression on E4, we expressed a luciferase reporter under the control of E4 (pGL4-E4) in Neuro-2a cells (Fig. 3b). Luciferase activity driven by the E4 enhancer was significantly increased by co-transfected Sox4 or Sox11, but not Sox12.
To dissect which sequences within the E4 drive Fezf2 expression, we generated four truncated versions of the E4 sequence (E4F1-4) (Fig. 3c). The luciferase activity of the E4F2 fragment was increased by SOX4 and SOX11, but not SOX12 (Fig. 3d). Using MatInspector, six potential SOX binding sites within the E4F2 sequence were predicted. To better define the precise basepairs with which SOXC proteins interact, we first tested whether the zebrafish E4F2 sequence, which exhibits 74.2% sequence identity with the mouse E4F2, is activated by SOXC proteins. Remarkably, the zebrafish E4F2 sequence was not activated by SOX11 (Supplementary Fig. 6), indicating that the SOXC-interacting sequences of mouse E4F2 lie within the sequences that are divergent in zebrafish. Of the predicted SOX binding sites in mouse E4F2, three putative SOX binding sites (SB1-3) are absent from the zebrafish E4F2. To test the function and specificity of these sites, we generated mutant versions of E4F2 by substituting SB1, SB2, or SB3 with the zebrafish sequence (Fig. 3e). Only mutations of the mouse SB2 site significantly attenuated the ability of SOX4 and SOX11 to activate the luciferase reporter gene, indicating that this site is crucial to species differences in SOXC-mediated transactivation of E4. Next, we assessed the ability of SOX4 and SOX11 to bind the SB2 site in vitro using an electrophoretic mobility shift assay (EMSA). We synthesized V5-tagged SOX4 and a higher-affinity, but equally specific, truncated form of SOX11 (SOX11(1-276aa)), since the binding affinity of native SOX11 is weak43. Biotinylated SB2 DNA was shifted in the presence of SOX4 or SOX11(1-276aa) and supershifted by an anti-V5 antibody (Fig. 3f), but not when excess unlabeled or mutated SB2 DNA was used. Therefore, SOX4 and SOX11 directly bind to and activate the transcription activity of E4 via SB2.
To test whether SOX4 and SOX11 functionally compete with the repressor SOX5 in the trans-activation of E4, we used the pGL4-E4 luciferase assay. Luciferase activity was significantly increased with increasing amounts of co-transfected Sox11 and to a lesser extent, Sox4, whereas increased amounts of Sox5 significantly decreased luciferase activity (Fig. 3h, i). Taken together, our sequence analysis and assays, both in vitro and in vivo, demonstrate that SOX4 and SOX11 functionally compete with SOX5 repression to activate Fezf2 transcription via direct binding to sites within E4.
To test the phenotypic consequences of Sox4 and Sox11 inactivation, we generated cortex-specific deletions of Sox4 and Sox11 using the Emx1-Cre line (Methods) to circumvent the prenatal lethality associated with constitutive deletions of the two genes44. Single conditional knockout mice (cKO: Sox4fl/fl;Sox11fl/+;Emx1-cre and Sox4fl/+;Sox11fl/fl;Emx1-cre) were both viable and fertile, whereas the conditional double knockout mice (cdKO: Sox4fl/fl;Sox11 fl/fl;Emx1-cre) died within the first postnatal week (data not shown). Inspection of the cerebrum during the first postnatal week revealed no obvious gross defects in single Sox4 or Sox11 cKO animals (Supplementary Fig. 7). However, in the absence of both Sox4 and Sox11, we observed a reduction in the size of the cerebral hemispheres and OB.
To assess the requirement of Sox4 and Sox11 in E4-mediated activity, we transfected primary cortical neurons cultured from heterozygous littermate control or Sox4; Sox11 cdKO embryos with control or E4-containing luciferase constructs (Fig. 4a). In control neurons, the presence of E4 increased luciferase activity by 2.5±0.3 fold (P=1.9×10−4, one-tailed Student’s t-test). This increase was abolished in neurons culture from Sox4; Sox11 cdKO (1.1±0.1 fold; P=0.866), indicating that the two SOXC TFs are major activators of the enhancer in cortical neurons. Next, using qRT-PCR, we found a significant reduction in neocortical Fezf2 expression in cdKO compared to control or single cKO littermates at postnatal day 0 (P0) (Fig. 4b). In addition, immunostaining for L1CAM and PRKCG revealed a dramatic loss of CS axons in Sox4; Sox11 cdKO mice but not single cKO littermates at P0 and P6 (Fig. 4c, d, Supplementary Fig. 8), whereas the organization of other brainstem tracts was not affected (Fig. 3c, d). Because the absence of L1CAM and PRKCG staining could reflect their down-regulated expression in the cdKO, as opposed to loss of CS axons, we confirmed the absence of the CS tract using the CRE-responsive CAG-Cat-Gfp transgenic line to express GFP in all cortical projection neurons and their axons39,40 (Supplementary Fig. 9). Two additional observations indicate that the loss of Fezf2 and CS axons in these mice was not due to an absence of L5 neurons. First, only a moderate increase in cell death was detected in the motor-sensory areas, from which CS axons originate (Supplementary Fig. 10). Second, many BCL11B-immunopositive L5 neurons were present in the cdKO motor-sensory areas, albeit lightly immunostained (Supplementary Fig. 11), indicating that neurons that would normally give rise to the CS tract were present. Analysis of additional laminar markers further revealed an inversion of cortical layers similar to the reeler mutant mouse10–12 (Supplementary Fig. 11), indicating that Sox4 and Sox11 are also required for proper laminar positioning of neurons. Previous studies have shown that the CS tract is present in reeler mice12, indicating that this laminar phenotype occurs independently of Fezf2 and probably in response to defects in the RELN signaling pathway13,14. In support of this possibility, RELN was absent from cdKO Ncx (Supplementary Fig. 12). Thus, the combined deletion of Sox4 and Sox11 leads to defects in laminar position of neurons and molecular specification and connectivity of CS neurons.
The differences in SOXC-mediated transcription between the mouse and zebrafish E4F2 suggest that species differences in the E4F2 sequence have functional implications. Analysis of SB1-3 sites within the E4F2 sequence in 23 species revealed that SB2 is conserved in all analyzed mammals and the two available non-mammalian amniotes (chicken and lizard) (Supplementary Fig. 13). To investigate the functional consequences, we used luciferase assays in Neuro-2a cells to analyze the activity of E4F2 sequence from different vertebrates (Fig. 5a, b) in response to SOX11, a more powerful trans-activator than SOX4 (Fig. 3). Of the constructs containing an E4F2 sequence of a mammal (human, chimpanzee, macaque, or mouse) or non-mammalian amniote (chicken), which have conserved SB1-3 sequences, the luciferase reporter activity was strongly increased following co-transfection with Sox11. The reporter activity of a non-amniote tetrapod (Xenopous) E4F2 construct was moderately increased compared to the empty or zebrafish luciferase plasmid, but not as high as the increase with mammalian constructs (P = 3.9×10−16, one-tailed Student’s t-test) (Fig. 5b). In contrast, the reporter construct containing the zebrafish E4F2 sequence, which has a highly divergent SB2 sequence from mammals, exhibited only basal level activity similar to the empty control luciferase plasmid. To confirm that this is dependent on sequence variations between species, we mutated zebrafish E4F2 (ZeE4F2-m2) by “murinizing” its SB1, SB2, or SB3 sequence. Reporter activity of the murinized zebrafish SB2 (ZeE4F2-m2), but not SB1 or SB3, was robustly induced by SOX11 (Fig. 5c). ZeE4F2-m2, however, was insufficient to restore the level of expression observed in the wildtype mouse sequence, suggesting the potential contribution of additional sequences. Thus, evolutionary differences in the E4F2 sequence, especially within SB2, are directly related to functional differences in the ability of SOX11 to activate this regulatory element.
Next, to investigate the functional consequences of species differences in the FEZF2 coding sequence, we tested the ability of the zebrafish Fezf2 to rescue the effects of mouse Fezf2 deficiency. We electroporated in utero the neocortical wall of the floxed Fezf2 (Fezf2fl/fl) mice40 with Cre and CRE-responsive Gfp-expressing constructs with or without zebrafish Fezf2 at embryonic day (E) 12.5 (Fig. 5d). Analysis of electroporated mice at P0 revealed that the zebrafish Fezf2 was sufficient to rescue the formation of the CS tract in Fezf2-null mouse neurons (Fig. 5e). Taken together, these results suggest that sequence variations in the E4 enhancer, but not the Fezf2 coding region, imparted an essential role of Fezf2 in CS tract evolution.
Our data provide critical insight into the genes and regulatory components controlling CS system development, centering on the cis regulation of Fezf2. We show that genetic inactivation of E4 results in compartmentalized phenotypic effects largely limited to the loss of cortical Fezf2 expression and CS system. The spatio-temporal dynamics of cortical Fezf2 is further controlled by SOX5 and TBR1, two TFs expressed post-mitotically in projection neurons39,40,46, 48, 49. Upon its initial activation in early cortical progenitors, Fezf2 is highly expressed in deep-layer neurons during early corticogenesis but subsequently repressed in L6 neurons by SOX5 and TBR1 to create a postnatal L5-enriched pattern. The earlier expression of SOX4 and SOX11 is consistent with their role in Fezf2 activation prior to the L6 upregulation of SOX5, a functional competitor. Despite what is known about the function and regulation of Fezf2, a number of key questions have yet to be addressed. First, additional, as yet untested, transactivators and regulatory elements likely contribute to Fezf2 regulation in a context-dependent manner. These include the mediators of Fezf2 repression in L2-4 projection neurons. Second, the distinct, and seemingly incongruous, effects of SOX5 in repressing Fezf2 but being independently required for CS tract formation39 have not been resolved. Third, the direct transcriptional targets of FEZF2 in CS neurons remain largely unknown.
We also show that SOX4 and SOX11 play a crucial role in regulating RELN expression and the inside-out pattern of cortical layer formation, independent of E4 or Fezf2 and likely involving interactions with distinct regulatory elements. Moreover, SOX4 and SOX11 have additional roles, as mice lacking both genes exhibit smaller cortex and OB, and increased cell death. Thus, SOX4 and SOX11 have pleiotropic functions, which are likely mediated by distinct regulatory elements and downstream target genes that are involved in multiple developmental processes.
Our results indicate that, following their emergence in tetrapods, functional SOX binding sites have retained high conservation via purifying selection in mammals and some amniotes, thus directly linking species variations in regulatory sequences to functional outcomes. E4 sequence substitutions in SB2 may constitute an evolutionary turning point for Fezf2 function during forebrain development, possibly facilitating the formation of descending telencephalic pathways including the CS system. Whereas minor projections from the ventral (subpallial) telencephalon to the spinal cord are present in amphibians2, direct dorsal telencephalo-spinal projections resembling the CS tract have only been reported in mammals and some birds50. We propose a scenario in which the concurrent emergence of the described regulatory mechanisms and direct telencephalo-spinal projections in early amniotes, together with subsequent changes in genetic programs driving the patterning and expansion of a six-layered dorsal pallium, made possible the evolution of the CS system in mammals.
Selected CNEE sequences were deleted from a Fezf2-Gfp BAC by recombineering and GFP expression from modified BACs was analyzed by transgenesis. To confirm the requirement of E4 enhancer for Fezf2 expression, a germ-line E4 deletion mutant was generated. Putative E4 trans-activators SOX4 and SOX11, identified by E4 sequence and co-expression analyses, were tested using ChIP and luciferase reporter assays. To analyze the requirement of Sox4 and Sox11 for Fezf2 expression and cortical development, we generated cortex-specific single and double Sox4 and Sox11 null mice. Complete materials and methods are described in supplemental information.
Full Methods and associated references are available in the online version of the paper at www.nature.com/nature.
All experiments were performed in accordance with a protocol approved by Yale University’s Committee on Animal Research. The generation of mice with the floxed Sox4 and Sox11 alleles (Sox4fl/fl and Sox11fl/fl, respectively) was described elsewhere44,51. The CAG-Cat-Gfp and Emx1-Cre P1 artificial chromosome transgenic mice were a generous gift from Melissa Colbert52 and Takuji Iwasato53, respectively. The Fezf2-Gfp BAC transgenic mouse was generated by the GENSAT project 38.
To generate BAC reporter constructs harboring deletions of putative enhancer elements (E1-4) in the mouse Fezf2-Gfp BAC (RP23-141E17-Gfp; the GENSAT project 38), homology arms A and B flanking the sequences targeted for deletion (E1, Chr14: 13204475-13204944, 470bp; E2, Chr14: 13183314-13183991, 678bp; E3, Chr14: 13180797-13181283, 487bp; E4, Chr14: 13170100-13170960, 861bp) were amplified by PCR, and cloned into the PL451 shuttle vector. SW102 E. coli transformed with the Fezf2-Gfp BAC was induced for recombineering at 42 °C and electroporated with the linearized shuttle vector. After neomycin selection, Flpase expression was induced by arabinose to excise the neomycin cassette54. Founder lines were confirmed by PCR using a primers corresponding to sequences from E1-4 (Supplementary Table 1). Fezf2-Gfp lines were maintained on a C57BL6/J background and all animals were housed under identical conditions. The intactness of the integrated BACs was confirmed by PCR and deep sequencing of genomic DNA.
The targeting vector for the deletion of the E4 enhancer was constructed by a recombineering-based method with an unmodified BAC clone (RP23-141E17). A 6627bp fragment containing the E4 enhancer was retrieved from the BAC and inserted into the PL253 vector by recombination in SW102 bacteria. The removal of the E4 sequence was achieved using the strategy described above for the generation of the ΔE4 BAC transgenic mouse. ES cells (C57BL6/J) were electroporated with the resulting targeting construct and expanded under positive (pGK-neo, from PL451 vector) and negative (MC1-TK, from the PL253 vector) selection. Correctly recombined clones were identified by long distance-PCR, confirmed by sequencing, and used for blastocyst injection and subsequent generation of mice with the targeted allele. The neomycin cassette was subsequently removed by flippase-mediated recombination by breeding with Flp mice. The mice were genotyped using the P1/P2 primer set (400bp) for the wildtype allele and the P1/P3 primer set (205bp) the E4 deletion allele (Supplementary Table 1).
Postnatal day 0 (P0) and P6 brains were fixed by immersion in 4% paraformaldehyde (PFA) overnight at 4 °C and sectioned using a vibratome (Leica). Immunostaining was performed as described20. The following primary antibodies were used: anti-L1CAM (rat, 1:300, Millipore), anti-PRKCG (rabbit, 1:300, Santa Cruz), anti-GFP (chicken, 1:3000, Abcam), anti-CUX1 (rabbit, 1:150, Santa Cruz), anti-SATB2 (mouse, 1:200, Genway), anti-BCL11b (rat, 1:250, Santa Cruz), anti-ZFPM2 (rabbit, 1:250, Santa Cruz), anti-RELN (mouse, 1:300, Millipore) and anti-CASP3, active (rabbit, 1:200, Millipore). We tested a total of 12 anti-SOX4 (Abcam, ab52043, ab86809, ab90696; Abgent, AP2045a, AP2045c; LS Bioscience, LS-B3520; Millipore, AB5803, AB10537; Pierce, PA1-38638, PA1-38639; Santa Cruz Biotechnology, sc-17326; Sigma, HPA029901) and 10 anti-SOX11 (Abcam;ab42853; Aviva Systems Biology, ARP33328, ARP38235; LS Bioscience, LS-B1567, LS-C10306; Millipore; AB9090; Pierce, PA5-19728; Santa Cruz Biotechnology, sc-20096, sc-17347; Sigma, HPA000536) commercially available antibodies on western blots and immunostaining. Three antibodies for each SOX4 and SOX11 recognized a single band of the expected size on western blot but none of them were suitable for immunostaining or ChIP assay. In addition, a few antibodies exhibited nuclear immunostaining, but also had strong background staining of tissue sections of the knockout brain, so we concluded that they are not specific.
For expression of SoxCs and Fezf2, full-length cDNAs (mouse Sox4, BC052736; mouse Sox11, BC062095; mouse Sox12, BC067019, zebrafish Fezf2, BC085677) were inserted into pCAGEN. For ChIP and EMSA, PCR amplified products (for mouse Sox4, Sox11 full-length cDNA and Sox11(aa 1-276) ) were inserted into pcDNA3.1/V5-His TOPO vector (Invitrogen). For luciferase reporter plasmids, PCR amplified products (for E1-4, E4F1-4, and E4F2 of six species) and annealed 190bp complementary synthetic oligonucleotides (for MoE4F2-m1-3, ZeE4F2-m1-3) were inserted into pGL4 (Promega). The sequences of the PCR primers and synthetic oligonucleotides are listed in Supplementary Table 1.
To show which genes in the SOX family are highly correlated with human FEZF2, we used the data generated by Kang et al., 2011 (ref. 45) for pairwise Pearson comparisons. This dataset is generated by exon arrays and is available from the www.humanbraintranscriptome.org website or NCBI GEO with the accession number GSE25219. It covers 16 brain and cortical regions over 15 periods, ranging in age from embryonic development to late adulthood. The evaluation of gene expression in each region and each period was detailed in the study 45. In this study, we averaged gene expression values for each of 11 neocortical regions/area and 15 developmental periods represented in the dataset, and then performed pairwise Pearson correlation analysis for FEZF2 and the members of the SOX family. Totally, 18 SOX genes were represented in the database. The correlation analysis was shown by heatmap and the correlation coefficients were indicated in each cell (Supplementary Fig. 4).
After testing available anti-SOX4 and anti-SOX11 antibodies and finding none to be specific and suitable for chromatin immunoprecipitation (ChIP) assay, we performed ChIP with Neuro-2a cells were transfected with pcDNA3-Sox4-V5 or pcDNA3-Sox11-V5 plasmid using Lipofectamine 2000 (Invitrogen). At 36 hrs after transfection, cells were cross-linked in 1% formaldehyde for 10 min at 37 °C, and processed for ChIP assay using the EZ-ChIP kit (Millipore) according to manufacturer’s instructions. Capture of the DNA fragments was tested by PCR using primers E1-F, 5’- TGGAGAGAAGGCCAACAAAC -3’, E1-R, 5’- GCTGGGGATGGAGAAGAATA -3’; E2-F, 5’- TCACCAAAGCGCCTTTTTAT -3’, E2-R, 5’- GTAAGCGGACATGCCATTTT -3’; E3-F, 5’- TGACTTTCCCCAGCCTTCTA -3’, E3-R, 5’- CAACAGCTCACCCACACAAT -3’; E4-F, 5’- ATGCCTAGCCCCAAAGAAAT -3’, and E4-R, 5’- TTAACTCCCCCTTTGGCTCT -3’.
Double strand DNA probes were generated by annealing with complementary single strand oliogonucleotides. Each DNA probe was biotin-end labeled with Klenow enzyme and then purified using the G-50 Sephadex Column (Roche). The labeled and un-labeled (cold probe; 100 fold) probes were incubated with SOX4-V5, SOX11(1-276aa)-V5 which produced using TNT Quick coupled transcription/translation kit (Promega) in a DNA-binding buffer (75mM NaCl, 1mM EDTA, 1mM DTT, 10mM Tris-HCl(pH 7.5), 6% Glycerol, 2ug BSA, 16ng poly (dI-dC), 0.1ug Salmon sperm DNA) at room temperature for 30 min. For supershift assay, anti-V5 antibody (Invitrogen) were added into the binding reaction and incubated for additional 10 min. Shift assay was carried out in 6% polyacrylamide gel in 0.5x TBE buffer. Subsequently, the labeled DNA was transferred on a Nylon membrane (Amersham Biosciences). For detection, we using the Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific) carried out according to manufacturers’ recommendations. All primer sequences were described in the Supplementary Table1.
Neuro-2a cells were transfected using Lipofectamine 2000 (Invitrogen) with one of pCAG-Sox5, pCAG-Sox4, pCAG-Sox11, pCAG-Sox12 or empty pCAGEN, together with one of the pGL4 (Promega) luciferase vectors generated with enhancer sequences as described above. A renilla luciferase plasmid (pRL, Promega) was co-transfected to control for transfection efficiency. The luciferase assays were performed 48hrs after transfection using the dual-luciferase kit (Promega) according to manufacturer’s instructions. Primary cortical neurons were prepared from E14.5 heterozygous littermate control (Sox4fl/+;Sox11fl/+;Emx1-cre and Sox4fl/+;Sox11fl/+) and Sox4;Sox11 cdKO (Sox4fl/fl;Sox11fl/fl;Emx1-cre) cortices and transfected with pGL4 or pGL4-E4 with pRL using the AMAXA Mouse Neuron Nucleofector Kit (VPG-1001; Lonza). At 48 hr after transfection, the luciferase assays were performed as described above.
Plasmid DNA (4 µg/µl) mixture (pCAGEN-Cre; pCALNL-Gfp or pCAGEN-Cre; pCALNL-Gfp; pCAGEN-ZeFezf2) was injected into the lateral ventricles of embryonic mice at embryonic day 12.5 (E12.5) and transferred into the cells of ventricular zone by electroporation (five 50ms pulses of 40V at 950ms intervals) as described elsewhere35, 39. Brains and tissue sections of electroporated animals were analyzed for GFP expression after fixation with 4% PFA at postnatal day 0 (P0).
Neocortex was dissected from P0 brain and RNA was isolated using the RNeasy kit (Qiagen). QRT-PCR was performed using primers listed in Supplementary Table 1 and the SYBR FAST qPCR kit (KAPA Biosystems) according to the manufacturer’s instruction. Gene expression levels were normalized to Gapdh expression.
We thank W. Han, Y. Imamura Kawasawa, D. Liu, and T. Nottoli for technical help;, A. Giraldez and A.M.M. Sousa for reagents; and the Sestan laboratory for discussions. This work was supported by the National Institutes of Health (NS054273, MH081896), March of Dimes, and McDonnell Scholar Award (N.S.).
Supplementary Information is linked to the online version of the paper at www.nature.com/nature.
Author ContributionsS.S., K.Y.K., and N.S. designed research; S.S. performed experiments; S.S., and K.Y.K. performed confocal imaging, M.L. analyzed co-expression and deep sequencing data, S.S., K.Y.K., M.L., and N.S. analyze the data, V.L. generated floxed Sox4 and Sox11 allele mice; N.S. conceived the study, S.S., K.Y.K. and N.S. wrote the manuscript. All authors discussed the results and implications and commented on the manuscript at all stages.
Competing Financial Interests
The authors declare no competing financial interests.