Hippocampal pyramidal neurons are important for encoding and retrieval of spatial maps and episodic memories. While previous work has shown that Zbtb20 is a cell fate determinant for CA1 pyramidal neurons, the regulatory mechanisms governing this process are not known. In this study, we demonstrate that Zbtb20 binds to genes that control neuronal subtype specification in the developing isocortex, including Cux1, Cux2, Fezf2, Foxp2, Mef2c, Rorb, Satb2, Sox5, Tbr1, Tle4, and Zfpm2. We show that Zbtb20 represses these genes during ectopic CA1 pyramidal neuron development in transgenic mice. These data reveal a novel regulatory mechanism by which Zbtb20 suppresses the acquisition of an isocortical fate during archicortical neurogenesis to ensure commitment to a CA1 pyramidal neuron fate. We further show that the expression pattern of Zbtb20 is evolutionary conserved in the fetal human hippocampus, where it is complementary to the expression pattern of the Zbtb20 target gene Tbr1. Therefore, the disclosed Zbtb20-mediated transcriptional repressor mechanism may be involved in development of the human archicortex.
Hippocampal pyramidal neurons are key components of an intricate neuronal network that encodes and retrieves spatial maps and episodic memories including contextual fear. They are the principal neurons of the CA1–CA3 subfields of the hippocampus proper (also referred to as Cornu Ammonis or Ammon's horn), which together with the dentate gyrus (DG) and the subiculum constitute the hippocampal formation (Lopes da Silva et al. 1990). The hippocampal formation is a 3-layered archicortex devoid of upper-layer neurons. Whereas subicular projection neurons acquire molecular identities similar to those of cortical deep layers V and VI (Ishizuka 2001), pyramidal neurons of the hippocampus proper develop alternative fates and are organized in a compact stratum pyramidale.
The zinc finger and broad complex, tramtrack, bric-a-brac (BTB) domain-containing protein 20 (Zbtb20), which is also known as DPZF, HOF, Oda-8, Zfp288, and Znf288, is expressed by the developing hippocampal projection neurons, and the developmental specification of CA1 pyramidal neurons has been shown to be dependent on Zbtb20 (Mitchelmore et al. 2002; Nielsen et al. 2007; Nielsen et al. 2010; Xie et al. 2010). Ectopic expression of Zbtb20 in the developing subicular and transitional retrosplenial areas is sufficient to commit neuronal precursor cells to a CA1 fate (Nielsen et al. 2007, 2010). CA1 pyramidal neurons are not specified in Zbtb20-deficient mice, but are replaced by projection neurons with both deep and upper-layer molecular identities (Xie et al. 2010; Rosenthal et al. 2012). Despite the instructive role of Zbtb20 in determination of the CA1 pyramidal neuron fate, the molecular mechanisms governing this are unknown.
Cortical projection neuron subtypes can be defined by morphology, laminar position, gene expression, electrophysiological properties, and axonal projection patterns (Guillemot et al. 2006; Molyneaux et al. 2007). In terms of axonal projections, 3 main types of projection neurons are present in the isocortex: Associative projection neurons that extend axons within a single cortical hemisphere, callosal projection neurons that send axons via the corpus callosum to the contralateral hemisphere, and corticofugal projection neurons that send axons out of the cortex to subcortical or subcerebral targets (Molyneaux et al. 2007).The specification of projection neurons in deep layers of the isocortex requires an intricate network of transcription factors (TFs). These TFs in part function to repress the acquisition of alternative deep-layer fates during neurogenesis (Molyneaux et al. 2007; Leone et al. 2008; Kwan et al. 2012). For example, Fezf2 and Bcl11b (also known as Ctip2) control the specification of subcerebral projection neurons, where Bcl11b acts downstream of Fezf2 (Arlotta et al. 2005; Chen, Schaevitz, et al. 2005; Chen, Rasin, et al. 2005; Molyneaux et al. 2005; Chen et al. 2008). The specification of callosal projection neurons requires Satb2, which directly represses Bcl11b (Alcamo et al. 2008; Britanova et al. 2008). In turn, Fezf2 suppresses Satb2 expression and the acquisition of a callosal fate (Chen et al. 2008). Both the TFs Sox5 and Tbr1 are required for the specification of corticothalamic projection neurons in layer VI, where they directly repress Fezf2 (Hevner et al. 2001; Kwan et al. 2008; Lai et al. 2008; Bedogni et al. 2010; Han et al. 2011; McKenna et al. 2011).
In adult rodents, CA1 pyramidal neurons can be distinguished from CA3 neurons based on their expression of the TFs Pou3f1 (also known as Oct6 and Scip) and Bcl11b. In contrast to subcerebral projection neurons, CA1 pyramidal neurons express Bcl11b independent of Fezf2 (Chen, Schaevitz, et al. 2005; Molyneaux et al. 2005). They further coexpress Bcl11b with moderate levels of Satb2, which is in contrast to the mutually exclusive expression patterns of these TFs in the isocortex (Alcamo et al. 2008; Britanova et al. 2008; Nielsen et al. 2010). Another characteristic feature of CA1 pyramidal neurons is that they receive input from CA3 pyramidal neurons through Schaffer collaterals that target their basal dendrites in stratum oriens and apical dendrites in stratum radiatum. Both projections to the thalamus and projection neurons with a corticothalamic molecular identity are rare in CA1 (Cenquizca and Swanson 2006; Nielsen et al. 2010). Likewise, callosal projection neurons are not present in the CA1 field since CA1 pyramidal neurons extend collaterals to the contralateral hippocampus via the hippocampal commissure (Cenquizca and Swanson 2007). It is therefore conceivable that Zbtb20 represses genes that control the development of these projection neuron subtypes in the isocortex. To test this hypothesis, we reveal the Zbtb20 targetome in the developing hippocampus. By combining chromatin immunoprecipitation (ChIP) with multiparallel sequencing (ChIP-Seq), we show that Zbtb20 binds to TF genes that control the development of callosal projection neurons, deep-layer projection neurons, and upper-layer projection neurons. Using a Zbtb20 gain-of-function mouse model, we show that these TFs are repressed during the Zbtb20-mediated specification of CA1 pyramidal neurons. We further show that expression of Zbtb20 is evolutionary conserved in the developing hippocampus of human and mouse, where Zbtb20 and the Zbtb20 target gene Tbr1 display complementary expression patterns. Taken together, the study discloses a novel transcriptional regulatory mechanism, whereby Zbtb20 ensures commitment to a CA1 pyramidal neuron fate.
Materials and Methods
Human Brain Specimens
Human brain tissue was obtained from extra uterine pregnancies and legal abortions. Informed consent was obtained according to the guidelines of the Helsinki Declaration II. All participants were given oral and written information concerning the study and have given their informed consent to participate according to and approved by “The Research Ethics Committee of the Capital Region” (KF—V.100.1735/90). This material has been used earlier to describe general features of the developing human hippocampus (Stagaard Janas, Nowakowski, Mollgard, 1991; Stagaard Janas, Nowakowski, Terkelsen, et al. 1991).
The D6/Zbtb20S transgenic mouse strain has been previously described (Nielsen et al. 2007). All animal experiments were in accordance with Danish and European animal welfare regulations and were licensed by the Danish Animal Experimentation Inspectorate.
For staining of mouse brain sections, the following primary antibodies were used: Rat anti-Ctip2/Bcl11b (1:500; Abcam), rabbit anti-CDP1/Cux1 (1:100; Santa Cruz Biotechnology), rabbit anti-Fez1/Fezf2 (1:100; IBL-International, Hamburg, Germany), rabbit anti-Fog-2/Zfpm2 (1:200; Santa Cruz Biotechnology), goat anti-Foxp2 (1:200; Santa Cruz Biotechnology), goat anti-Mef2c (1:100; Santa Cruz Biotechnology), mouse anti-Satb2 (1:100; Abcam), rabbit anti-Satb2 (1:1000; Britanova et al. 2005), goat anti-Sox5 (1:200; Santa Cruz Biotechnology), rabbit anti-Tbr1 (1:500; Abcam), rabbit anti-Tle4 (1:100, Santa Cruz Biotechnology) rabbit anti-Y2R (1:500; Neuromics), and rabbit anti-Zbtb20 (1:500; Mitchelmore et al. 2002). For staining of the human fetal brain sections, we used the rabbit anti-Zbtb20 (1:1000) and rabbit anti-Tbr1 (1:200) antibodies.
Immunohistochemistry of Mouse Brain Sections
Animals were deeply anesthetized with 2,2,2-tribromoethanol (Avertin) and perfused transcardially with 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS), pH 7.4. The brains were removed and post-fixed in the same fixative overnight at 4°C. Whole heads from embryos and newborn pups were fixed overnight at 4°C in 4% PFA. Fixed brains were incubated over night at 4°C in 20% sucrose in PBS, embedded in Tissue-Tek O.C.T. compound, frozen in liquid nitrogen, sectioned at 20 μm using a cryostat, and collected on Superfrost Plus slides (Thermo Fisher Scientific). Antigens were retrieved by boiling of the sections in 10 mM sodium citrate (pH 6.0) in a microwave oven. Next, the sections were incubated with primary antibodies overnight at 4°C in PBS containing 0.1% Triton X-100 (PBS-T). For immunofluorescence, appropriate Alexa 488- or 594-conjugated secondary antibodies raised in donkeys were used at 1:500 dilution (Molecular Probes), and sections were mounted in Vectashield mounting medium with 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories). For immunohistochemistry, appropiate biotinylated secondary antibodies were used (1:100; Dako), followed by a strepavidin–biotin–peroxidase complex (Dako) for 30 min in PBS-T. The sections were developed by incubation in 3,3′ diaminobenzidine tetrahydrochloride (DAB), dehydrated, and mounted with Depex. For immunohistochemistry (IHC) with the rabbit antibody to Y2R, adult (P21) mice were deeply anesthetized with Avertin and killed by cervical dislocation. The brains were removed and snap frozen in liquid nitrogen, and sectioned at 20 μm using a cryostat. Sections were fixed for 10 min with 4% PFA and incubated with the antibody to Y2R overnight at 4°C. The next day the sections were incubated with EnVision™ (Dako) according to the manufactorer's instruction and developed by incubation with DAB. Coimmunofluorescent stainings for Bcl11b and Satb2 were done as previously described (Nielsen et al. 2010).
Immunohistochemistry of Human Brain sections
A total of 14 human fetuses were examined. They ranged from 30 to 200 mm crown-rump length corresponding to 8–20 weeks post-conception (WPC). The fetal brain material was processed as described earlier (Stagaard Janas, Nowakowski, Mollgard, 1991; Stagaard Janas, Nowakowski, Terkelsen, et al. 1991). Sections were deparaffinized and rehydrated in xylene followed by a series of graded alcohols in accordance with established procedures. The sections were treated with a fresh 0.5% solution of hydrogen peroxide in methanol for 15 min for quenching of endogenous peroxidase and were then rinsed in Tris-buffered saline (TBS, 5 mM Tris–HCl, 146 mM NaCl, pH 7.6). Nonspecific binding was inhibited by incubation for 30 min with blocking buffer (ChemMate antibody diluent, DakoCytomation) at room temperature. Antigen retrieval was performed with microwave oven using Tris-EGTA (TEG) buffer, pH 9, and a boiling time of 10 min. After heat treatment, the sections rested for 20 min at room temperature. The sections were then incubated overnight at 4°C with the polyclonal rabbit antibodies (anti-Zbtb20 and anti-Tbr1) in blocking buffer (ChemMate antibody diluent, DakoCytomation). The sections were washed with TBS and detected by a REAL EnVision Detection System, peroxidase/DAB+, rabbit/mouse (Dako). The sections were dehydrated in graded alcohols followed by xylene and coverslipped with Depex mounting media.
DiI Tracing of Neuronal Fibers
Brains from 7-day-old mice were fixed in 4% PFA over night. Single 1,1′,di-octadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate (DiI) crystals (Molecular Probes) were placed in the medial entorhinal cortex (MEC). Brains were kept in the dark in 4% PFA for 6 weeks at 37°C for DiI diffusion, then washed in PBS, embedded in agar, and sectioned at 100 μm on a vibratome (VT1000S, Leica). The sections were mounted in Vectashield with DAPI and immediately photographed with a digital camera (Nikon Ds Ri1) mounted on an upright fluorescence microscope (Nikon E800).
Hippocampi of newborn wild-type C57BL/6 mice were dissected, immediately frozen on dry ice, and stored at −80°C. All samples used for ChIP-Seq or ChIP-quantitative PCR (qPCR) consisted of hippocampi from at least 10 animals. Chromatin was precipitated from hippocampal tissue with an affinity-purified rabbit antibody to Zbtb20 (Mitchelmore et al. 2002) using Active Motif's FactorPath method (Active Motif) as previously described (Labhart et al. 2005). In brief, protein/chromatin cross-linking was obtained with freshly prepared formaldehyde (1% w/v) for 15 min at room temperature. Cross-linking was stopped by addition of 0.125 M glycine (final concentration), and the tissue was disrupted with a Dounce homogenizer. Lysates were sonicated to shear the DNA. Genomic DNA (input DNA) was prepared by treating aliquots of chromatin with RNase and proteinase K. Crosslinks were reversed by incubation overnight at 65°C, and DNA was purified by phenol–chloroform extraction and ethanol precipitation. Pellets were resuspended and the resulting DNA was quantified on a NanoDrop spectrophotometer (Thermo Fisher Scientific). Extrapolation to the original chromatin volume allowed for the determination of the total chromatin yield. For ChIP, an aliquot of chromatin (30 μg) was precleared with protein G agarose beads (Invitrogen). Genomic DNA regions bound by Zbtb20 were isolated using 4 μg of the affinity-purified rabbit antibody to Zbtb20. Protein/chromatin complexes were washed, eluted from the beads with sodium dodecyl sulfate buffer, and subjected to RNase and proteinase K treatment. The ChiPed DNA was de-crosslinked and extracted as described above.
ChIP-Seq Library Construction and Data Analysis
ChIPed DNA and input DNA (control) were prepared for amplification by converting overhangs into phosphorylated blunt ends and by adding an adenine to the 3′-ends. Illumina adaptors were added, and the library was size selected (175–225 bp) on an agarose gel. The adaptor-ligated libraries were amplified for 18 cycles, before the amplified DNAs (DNA libraries) were sequenced on the Illumina Genome Analyzer II.
Approximately, 11 million quality-filtered Illumina sequence tags (35 bp long) per sample (Zbtb20 ChIP and input DNA) were aligned to the mouse genome (NCBI Build 37.1) using the ELAND algorithm. Aligns were extended in silico (using Active Motif software) at their 3′ ends to a length of 140 bp, which is the average genomic fragment length in the size-selected libraries, and assigned to 32-nt bins along the genome. Peak locations were determined using the MACS algorithm (Zhang et al. 2008) comparing ChIP against input using a moderate cutoff of P = 10−7, which corresponds to a fragment density threshold of 10 (5 consecutive bins containing at least 10 aligns). Graphs of genomic regions with peak enrichment values were generated using the Affymetrix Integrated Genome Browser (IGB). Information about evolutionary conservation across placental mammals was obtained from the UCSC Genome Browser (Pollard et al. 2010; Rhead et al. 2010). Information about ultra-conserved enhancer elements was obtained from the VISTA Enhancer Browser (Visel et al. 2007).
Chromatin was precipitated from 3 samples of hippocampal tissue as described above using the Zbtb20 antibody and rabbit IgG, respectively. The latter was used as a control. qPCR reactions were performed in triplicate on specific genomic regions of interest using SYBR Green Supermix (Bio-Rad), site-specific primer pairs (primer sequences are provided in Supplementary Table S2), and an aliquot of ChIPed DNA. The resulting signals were normalized for primer efficiency by carrying out qPCR for each primer pair using input DNA. Final results were scaled as copies of DNA detected per 1000 genome equivalents of input DNA. Data are depicted as boxplots, showing median as well as the 25 and 75 percentiles. Comparison of medians was done using the Mann–Whitney rank sum test, and a P-value <0.05 was considered significant.
Affymetrix Array Profiling
One-day-old (P1) D6/Zbtb20S and wild-type control mice were killed by cervical dislocation. The brains were removed and cut in 500 μm coronal sections by a McIlwain tissue chopper. Next, the dorso-medial cortex (subiculum and retrosplenial cortex) was dissected from 2 or 3 sections, immediately frozen in liquid nitrogen, and stored at −80°C. Total RNA was extracted using RNAeasy Lipid Tissue Mini kit (Qiagen) from tissues of 21 wild-type and transgenic mice, respectively. RNA from tissues of 7 animals was pooled in 3 independent tubes, hereby generating 3 pools of RNA from the dorso-medial cortex of wild-type and transgenic mice, respectively. The quantity of RNA was determined with a NanoDrop spectrophotometer (Thermo Scientific), and the RNA quality was assessed using an Agilent 2100 Bioanalyser (Agilent Technologies). Approximately, 500 ng of RNA from each pool was amplified (aRNA) and fragmented using the MessageAmpTM III RNA amplification kit (Ambion). Fragmented aRNA was hybridized to Affymetrix Mouse Genome 430 2.0 arrays containing 45 101 probe sets recognizing >39 000 transcripts. The affy and affyQCReport packages implemented in the statistical programming language R (www.bioconductor.org) were applied for initial data analysis. Data were normalized using the loess method, and expression index calculation was performed by robust multiarray average (RMA) expression measure.
To assess the impact of Zbtb20 binding in proximity to annotated gene loci, the 3807 genes that had Zbtb20 ChIP-Seq peaks within a region spanning 10 kb upstream of the annotated transcriptional start site and 10 kb downstream of the polyadenylation signal were tested for differential expression between the dorso-medial cortex of wild-type and D6/Zbtb20S transgenic mice. Differentially expressed genes were identified using Significance Analysis of Microarray (Tusher et al. 2001), in Multiexpression Viewer (http://www.tm4.org/mev/). Genes with a false discovery rate <0.05 were considered significant.
Possible biological functions of the significant genes were assessed by analysis of over-represented gene ontology (GO) terms using Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/; Release 08.11.2011). TF genes were identified as genes included in the GO category: “Regulation of Transcription” (GO_0045449; Release 06.11.2010). The gene set was downloaded from DAVID, and an overlap with significant genes was identified using Microsoft Access. To validate the findings in an independent setting, a previously published data set comparing gene expression in hippocampi of 2-day-old Zbtb20-knockout mice and wild-type littermates (Xie et al. 2010) was downloaded. The 4 raw CEL files (GEO accession numbers: GSM506899–GSM506902) were processed using loess and RMA as described above. Using Multiexpression Viewer, heat maps for the Zbtb20 target genes were constructed for the data set in this study and for the data that were published by Xie and colleagues. Data were row standardized for each gene by subtracting the mean expression from all samples from the expression value and dividing by the standard deviation.
Identification of Dysregulated Mef2c Target Genes
To identify Mef2c target genes, whose expression is dysregulated as a consequence of Zbtb20-mediated repression of Mef2c in D6/Zbtb20S mice, the following strategy was used. To exclude the possibility that the expression level of Mef2c target genes was directly regulated by Zbtb20, we first removed 7101 probe sets corresponding to the 3807 potential Zbtb20 target genes from the Affymetrix gene expression data set. When a gene was recognized by more than one probe set, the probe sets were collapsed to the probe set with highest expression using the gene symbol as identifier. Next, a consensus gene set of 191 gene symbols (Mef2_03; systematic name: M17361) corresponding to genes with a potential Mef2c binding site (WKCTAWAAATAGM) located within 2 kb up- or downstream of the transcription start site (TSS) was downloaded from the Molecular Signatures Database (MSigDB; Subramanian et al. 2005). Gene set enrichment analysis (GSEA; Subramanian et al. 2005) was used to test for overall differential expression pattern of the Mef2c target genes in the CA1-transformed areas of D6/Zbtb20S mice when compared with corresponding areas of wild-type mice. GSEA ranks the genes according to a signal-to-noise value: (XA − XB)/(sA + sB), where X is the mean and s is the standard deviation for the 2 classes A and B [CA1-transformed subiculum and retrosplenial cortex of D6/Zbtb20S mice (A) and subiculum and retrosplenial cortex of wild-type mice (B), respectively]. The output from GSEA is an enrichment score, describing the imbalance in the distribution of ranks of gene expression in each class between the CA1-transformed subiculum and retrosplenial cortex of D6/Zbtb20S mice (A) and the subiculum and retrosplenial cortex of wild-type mice (B). The significance of the observed enrichment score is estimated by comparing with a null distribution of enrichment scores generated from 1000 randomly permutated gene lists. The genes contributing to the enrichment score are termed “leading edge” genes. These genes that were considered Mef2c target genes are dysregulated by Zbtb20-mediated repression of the Mef2c gene.
Identification of Over-represented DNA Sequence Motifs
To identify possible Zbtb20 consensus-binding sequences, we searched the DNA sequences of Zbtb20-bound regions with peak values ≥20 (3205 peaks in total) for over-represented sequence motifs using the online RSAT peak-motifs identification tool (Thomas-Chollier et al. 2012). We searched for over-represented oligonucleotides with oligomer lengths above 7 using a fifth order Markov background model.
The gene expression data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number: GSE38837.
Identification of the Zbtb20 Targetome in the Developing Murine Hippocampus
To reveal the Zbtb20 targetome, we performed ChIP with an affinity-purified Zbtb20-specific antibody and chromatin prepared from the mouse hippocampus at postnatal day 0 (P0), when expression of Zbtb20 is pronounced in the developing hippocampal projection neurons (Mitchelmore et al. 2002; Xie et al. 2010; Fig. 1C and Supplementary Fig. S1). Precipitated DNA, as well as control input DNA, was analyzed by multiparallel sequencing (ChIP-Seq). This generated approximately 11 million sequence reads for Zbtb20-precipitated DNA and control input DNA, respectively, which were aligned to the mouse genome. By use of the MACS peak finding algorithm (Zhang et al. 2008), we identified a total of 11 184 genomic regions that were enriched in Zbtb20-precipitated DNA. These regions (also referred to as peaks) are considered to be potential Zbtb20-binding sites and they could be mapped to a total of 3807 gene loci, that is, within 10 kb upstream of the TSS or 10 kb downstream of the polyadenylation signal (Fig. 1E).
Since Zbtb20 functions as a transcriptional repressor (Xie et al. 2008), we next tested which of the 3807 candidate genes that are significantly downregulated during CA1 pyramidal neuron development. We dissected the CA1-transformed subicular and retrosplenial cortex from Zbtb20-misexpressing D6/Zbtb20S transgenic mice (Nielsen et al. 2007, 2010) and the corresponding areas from wild-type mice, and compared their gene expression profiles using microarrays (Fig. 1A–F). We hereby identified a total of 346 candidate genes that appear to be directly repressed by Zbtb20 (Fig. 1G). According to the GO database (http://www.geneontology.org), there is a significant enrichment for genes with roles in neurodevelopment, including neurogenesis, neuronal migration (i.e. cell motion), axonogenesis, dendritogenesis, and neuronal circuit formation (e.g. axon guidance, synapse organization, synaptic transmission), implying that Zbtb20 regulates various aspects of pyramidal neuron development in CA1 (Supplementary Table S1).
We next tested the hypothesis that Zbtb20 directly represses TF-encoding genes that are important for neurodevelopment in the isocortex. We used the GO term “Regulation of transcription” to identify 42 potential TF genes that are significantly downregulated during ectopic CA1 pyramidal neuron development in D6/Zbtb20S transgenic mice (Fig. 1H). These genes include Cux1, Cux2, Fezf2, Foxp2, Mef2c, Rorb, Satb2, Tbr1, Tle4, and Zfpm2 that are all expressed by projection neuron subtypes in the isocortex (Molyneaux et al. 2007; Leone et al. 2008; Kwan et al. 2012). Using a previously published microarray data set that compares the gene expression profiles in hippocampi of 2-day- old wild-type and Zbtb20-deficient mice (Xie et al. 2010), we find that Fezf2, Rorb, Satb2, Tbr1, Tle4, and Zfpm2 are upregulated in the Zbtb20-deficient hippocampus (Fig. 1H), supporting the notion that they are direct targets of Zbtb20-mediated repression. The data further suggest that there is little functional redundancy for Zbtb20 in repression of these genes during hippocampal neurogenesis.
We next used the RSAT Peak-Motifs identification tool (Thomas-Chollier et al. 2012) to identify DNA sequence motifs that are over-represented in the Zbtb20-bound regions and therefore might be recognized by Zbtb20. We found 3 over-represented sequence motifs that are present in 33.7%, 21.7%, and 22.9% of the analyzed Zbtb20 peaks, respectively (Supplementary Fig. S2A–C). Hence, the identified motifs may account for up to 78.3% of Zbtb20-binding events in the analyzed peaks, suggesting that Zbtb20 also binds to more degenerative sequence motifs that were not identified. Notably, the second and third motifs are very similar (Supplementary Fig. S2B,C). We next compared the motifs with previously published Zbtb20-binding sequences. Interestingly, sequence elements that have a high degree of similarity with motifs 2 and 3 are found in an oligonucleotide sequence that binds the zinc fingers of Zbtb20 in an in vitro electrophoretic mobility shift assay (Mitchelmore et al. 2002; Supplementary Fig. S2D) and in an in vivo Zbtb20-binding sequence, which is located in the putative promoter of the murine Afp gene (Xie et al. 2008; Supplementary Fig. S2E). These data suggest that the identified motifs are likely to be recognized by Zbtb20. We further find that one or more of the motifs are present in 15 of the 22 peaks (68.2%) that are associated with the above-mentioned TF-encoding genes that are repressed by Zbtb20 in CA1 pyramidal neurons (Supplementary Table S3). We hypothesized that Zbtb20 represses these genes by binding to enhancer elements that would normally promote expression in the developing forebrain. Since enhancers are often evolutionary conserved in mammals and can be identified based on binding of the transcriptional coactivator p300, we obtained a list of genomic regions that are bound by p300 in the forebrain of E11.5 mice (Visel et al. 2009) and investigated whether these regions overlap with Zbtb20-bound regions in the hippocampus (see below).
Zbtb20 Directly Represses Genes Controlling Corticofugal Projection Neuron Specification
The ChIP-Seq data indicate that Zbtb20 binds approximately 8 kb downstream of the TSS of the Fezf2 gene in the evolutionary conserved and well-characterized hs434 enhancer (Fig. 2A,B). This enhancer is essential for isocortical expression of the Fezf2 gene, and it is sufficient to drive transgenic expression in a pattern that mimics endogenous Fezf2 expression (Visel et al. 2007; Kwan et al. 2008; Shim et al. 2012). It is also bound by the enhancer-associated coactivator p300 in the forebrain of E11.5 mice and is therefore functional at this early developmental stage (Visel et al. 2009). To confirm the binding of Zbtb20 to the hs434 enhancer, chromatin was precipitated from 3 samples of wild-type hippocampi dissected from P0 brains. Analysis of ChIPed DNA by qPCR revealed a significant enrichment of this region in DNA precipitated with the Zbtb20 antibody compared with an IgG control, confirming that Zbtb20 binds to this enhancer element in hippocampal neurons in vivo (Fig. 2C). This finding correlates with reduced levels of Fezf2 expression in the Zbtb20-transformed areas of D6/Zbtb20S transgenic mice (Fig. 1H), where Fezf2 protein is only expressed by a few cells in the deep part of the cortical plate (CP), in contrast to the pronounced expression of the TF in corresponding areas of the wild-type brain (Fig. 2D,E and Supplementary Fig. S3A–C). Zbtb20 also binds to 2 evolutionary conserved regions, located approximately 9.5 and 8.9 kb upstream of the TTS of Tbr1, in an intron of the neighboring Psmd14 gene (Fig. 2F–H). Expression of the Tbr1 gene is significantly lower in the Zbtb20-transformed areas of newborn D6/Zbtb20S mice (Fig. 1H), where only few cells express Tbr1 protein when compared with the widespread expression of Tbr1 in the corresponding areas of wild-type mice (Fig. 2I,J and Supplementary Fig. S3D–F). In contrast, expression of the Psmd14 gene is unaltered implying that binding of Zbtb20 does not cause repression of this gene (data not shown). Moreover, Tbr1 is upregulated 2-fold in hippocampi of Zbtb20-knockout mice (Xie et al. 2010), which further supports the notion that Tbr1 is repressed by Zbtb20.
Sox5 was not significantly downregulated in our microarray data. However, Zbtb20 and Sox5 display mutually exclusive expression patterns in the developing CA1 field, and expression of Sox5 is upregulated in the hippocampus of Zbtb20-deficient mice (Nielsen et al. 2010; Xie et al. 2010), suggesting that Zbtb20 represses Sox5. In line with this, the ChIP-Seq analysis reveals several Zbtb20-binding sites in evolutionary conserved regions of the Sox5 locus (Fig. 2K–M) including 2 that are also bound by the enhancer-associated coactivator p300 in the E11.5 forebrain (Visel et al. 2009). One of these regions is a validated enhancer (hs895, VISTA Enhancer Browser; Visel et al. 2007), which is located approximately 103 kb upstream of the Sox5 TSS. The binding of Zbtb20 to these genomic regions correlates with a pronounced downregulation of Sox5 expression in the Zbtb20-transformed areas of the D6/Zbtb20S cortex (Fig. 2N,O and Supplementary Fig. S3G–I). Corticothalamic projection neurons also express Foxp2, Tle4, and Zfpm2 (Kwan et al. 2008; Bedogni et al. 2010; Hisaoka et al. 2010). We find that Zbtb20 also binds to evolutionary conserved sequences in introns of these genes and that transgenic expression of Zbtb20 correlates with reduced expression of the Foxp2, Tle4, and Zfpm2 proteins in D6/Zbtb20S mice (Supplementary Fig. S4A–O). Like Tbr1 and Sox5, both Zfpm2 and Tle4 are upregulated in the hippocampus of Zbtb20-deficient mice (Fig. 1H). These results strongly imply that Zbtb20 directly suppresses the acquisition of a corticofugal projection neuron fate in the developing CA1 pyramidal neurons.
Zbtb20 Prevents the Specification of a Callosal Projection Neuron Identity
We find that Zbtb20 binds to an evolutionary conserved region that is located in an intron of the Satb2 gene (Fig. 3A–C) and is likely to function as an enhancer since it is also bound by p300 in the E11.5 forebrain (Visel et al. 2009). In addition, Zbtb20 binds to the conserved AS021 Satb2 enhancer, which is located in a short interspersed repetitive element approximately 390 kb upstream of the Satb2 gene (Fig. 3A–C). The latter enhancer can drive expression of a transgene in a pattern that recapitulates the expression of Satb2 in deep-layer neurons (Tashiro et al. 2011). Hence, the binding of Zbtb20 to these genomic regions may explain why CA1 pyramidal neurons do not acquire a callosal fate during neurogenesis. This finding is further supported by the findings that the expression of Satb2 is upregulated in the hippocampus of Zbtb20-deficient mice (Xie et al. 2010; Rosenthal et al. 2012) and that Satb2 expression is repressed in neurons of the Zbtb20-transformed transitional cortex of newborn D6/Zbtb20S mice (Fig. 3D,E and Supplementary Fig. S5A–C).
Zbtb20 Represses Genes Controlling Upper-Layer Projection Neuron Development
Both the hippocampus and subiculum lack projection neurons with upper-layer identities. In Zbtb20-deficient mice, upper-layer neurons are generated in presumptive CA1 and subicular areas (Xie et al. 2010; Rosenthal et al. 2012), indicating that Zbtb20 suppresses upper-layer neurogenesis in the archicortex. We find that Zbtb20 directly binds to a number of TF genes that are normally expressed by upper-layer neurons, such as Rorb (Supplementary Fig. S6G–I), which is important for the cytoarchitecture of layer IV in the somatosensory cortex (Jabaudon et al. 2012). The binding of Zbtb20 is also significant in conserved regions of the Cux1 and Cux2 genes (Supplementary Fig. S6A–F), which are expressed by upper-layer neurons and by intermediate neuronal progenitors in the subventricular zone (SVZ) that give rise to upper-layer projection neurons (Nieto et al. 2004; Zimmer et al. 2004). The binding of Zbtb20 also occurs in 4 evolutionary conserved regions of the Mef2c gene (Fig. 3F–H) that is expressed by subsets of upper-layer projection neurons and is essential for the correct positioning of these neurons in the CP (Leifer et al. 1993; Molyneaux et al. 2007; Li et al. 2008). Expression of Mef2c was not revealed in the Zbtb20-transformed retrosplenial cortex of D6/Zbtb20S mice (Fig. 3J), whereas Mef2c expression is widespread in the retrosplenial cortex of wild-type littermates, but not in the subiculum and CA1 (Fig. 3I and Supplementary Fig. S5D–F). Moreover, Mef2c expression is pronounced in the hippocampus of Zbtb20-deficient mice (Xie et al. 2010). Mef2c has previously been shown to function as a DNA-binding transcriptional activator that binds to conserved promoter sequences (Potthoff and Olson 2007). We therefore hypothesized that repression of Mef2c would results in dysregulation of Mef2c target genes. To test this hypothesis, we obtained a list of Mef2c targets from the MSigDB (Subramanian et al. 2005) and tested whether these genes were over-represented among the genes that are repressed in the CA1-transformed areas of D6/Zbtb20S brains. To avoid that changes in gene expression occurred as a consequence of direct binding of Zbtb20 to Mef2c target genes, we first excluded the 3807 Zbtb20 target genes from the Affymetrix gene expression data set. Indeed, expression of the Mef2c targetome was significantly altered in the CA1-transformed cortex of D6/Zbtb20S mice (P = 0.004; Supplementary Fig. S7). Notably, subsets of these Mef2c responsive genes have been linked to the development of upper-layer projection neurons including Sema7a, which is expressed by projection neurons in layer IV (Fukunishi et al. 2011), and Hspb3, which is a marker of callosal projection neurons (Molyneaux et al. 2009). Taken together, our data imply that Zbtb20 represses a number of TF genes that are important for the development of upper-layer projection neurons and that the Zbtb20-mediated repression has a significant effect on a downstream transcriptional network.
Ectopic Zbtb20 Represses TF Genes and Induces CA1 Pyramidal Neuron Development in the Ventral Subiculum and Parahippocampal Areas
Based on differences in gene expression and functions, the hippocampus can be divided into dorsal (septal) and ventral (temporal) parts (Dong et al. 2009; Fanselow and Dong 2010). So far, most analyses of the functional role of Zbtb20 in neurodevelopment have focused on the dorsal part of the hippocampus and adjacent cortical areas. We therefore next tested whether ectopic Zbtb20 also represses its TF-gene targets in the ventral subiculum and neighboring parahippocampal areas of D6/Zbtb20S transgenic mice. Ectopic expression of Zbtb20 is found in the VZ, SVZ, intermediate zone (IZ), and CP of the developing subiculum, pre- and parasubiculum, as well as in the MEC, whereas it is less pronounced in the lateral part of the entorhinal cortex of E15 D6/Zbtb20S transgenic mice (Fig. 4A,B). Cresyl-stained horizontal sections of P21 D6/Zbtb20S brains indicate that the ventral subiculum is transformed to a compact CA1-like stratum pyramidale (Fig. 4C,D) similar to that of the dorsal subiculum in these mice (Nielsen et al. 2010; Fig. 1B,D). Moreover, the molecular layer I is expanded, and both upper-layer neurons and the lamina dissecans are absent in pre- and parasubiculum and MEC (Fig. 4C,D). In line with this, the Zbtb20 target genes, Cux1 and Mef2c, are severely downregulated in these areas of D6/Zbtb20S mice (Fig. 4E–H), confirming that neurons with upper-layer molecular identities are not specified. These upper-layer neurons have been replaced by neurons expressing the deep-layer marker Bcl11b (Fig. 4I–J). Projection neurons in the upper layers of the entorhinal cortex normally give rise to the major afferent projection to the hippocampus, referred to as the perforant path (PP) that terminates in the dentate molecular layer and in the stratum lacunosum-moleculare of CA1 and CA3 (Lopes da Silva et al. 1990). Consistent with the lack of upper-layer neurons, there is a pronounced deficiency in PP projections from MEC to the hippocampus in D6/Zbtb20S brains as revealed by carbocyanine dye DiI tracing (Fig. 4K,L).
To reveal the identity of Zbtb20-expressing neurons in deep layers in the pre- and parasubiculum and MEC of D6/Zbtb20S mice, we next did IHC for Tbr1 and Fezf2 that marks corticothalamic and subcerebral projection neurons, respectively. Expression of these Zbtb20 target genes are severely reduced in pre- and parasubiculum and MEC of D6/Zbtb20S mice when compared with corresponding areas of wild-type mice (Fig. 5A–D), suggesting that the Bcl11b-positive neurons in these areas are not corticofugal projection neurons. Since several of these neurons seem to express Bcl11b independent of Fezf2 and further coexpress Bcl11b with moderate levels of Satb2 (Fig. 5E–H), they appear to harbor a molecular identity of CA1 pyramidal neurons (Molyneaux et al. 2005; Nielsen et al. 2010). Neuropeptide Y2 Receptor (Y2R) marks mossy fibers from DG to CA3 and Schaffer collaterals from CA3 to stratum oriens and stratum radiatum of CA1 (Fig. 5I; Stanic et al. 2006). In D6/Zbtb20S mice, Y2R IHC reveals ectopic Schaffer collaterals in the CA1-transformed subiculum and in the pre- and parasubiculum, where coexpression of Satb2 and Bcl11b is also observed (Fig. 5H,J). These data show that neuronal precursors in the ventral subiculum and adjacent parahippocampal areas are competent for CA1 development (Fig. 5K,L). The data further support the notion that Zbtb20 instructs CA1 fate determination through the disclosed transcriptional repression mechanism.
Expression of Zbtb20 and Tbr1 in the Fetal Human Hippocampus
We next tested whether Zbtb20 is expressed by the developing projection neurons in the fetal human hippocampus. Expression of Zbtb20 protein is detectable in the human hippocampal primordium at the eleventh WPC, where a distinct nuclear staining is revealed in the dentate anlage (DA). There are also low to moderate levels of Zbtb20 in progenitor cells in the fimbrial glioepithelium and the primary dentate neuroepithelium (Fig. 6A). Since we identified Tbr1 as a Zbtb20 target gene in the developing murine hippocampus, we also analyzed the expression pattern of the Tbr1 protein in the developing human hippocampus. We find that Tbr1 is present in many cell nuclei in the ammonic plate (AP), but is absent from the Zbtb20-expressing DA (Fig. 6B). At 13 WPC, the developing hippocampus has grown considerably, and expression of Zbtb20 is now pronounced in all germinal zones of the hippocampus. Nuclear localization of Zbtb20 is also found in immature projection neurons in the dentate migratory stream, in the granule cell layer of the DA, and in the IZ and AP of the CA1–CA3 subfields (Fig. 6C). At this developmental stage, expression of Tbr1 is absent from the hippocampus with the exception of the CA1 field, where the protein is expressed in cells in the deepest part of the AP (Fig. 6D), which is in line with the expression pattern of Tbr1 in the developing murine hippocampus (Fig. 2I and Supplementary Fig. S3D). At 20 WPC, nuclear localization of Zbtb20 is pronounced in immature neurons in the dentate migratory stream, in the granule cell layer and hilus of the dentate primordium, and in developing pyramidal neurons in the CA1–CA3 fields (Fig. 6E). These data show that expression of Zbtb20 is evolutionary conserved in the developing hippocampus of human and mouse (Fig. 6E and Supplementary Fig. S1). Consistent with its murine expression pattern, Zbtb20 is also undetectable in the CP of the human subiculum (Fig. 6E and Supplementary Fig. S8A). Nuclear expression of Tbr1 is widespread in deep layers of the developing subiculum, whereas only a weak cytoplasmic reactivity for Tbr1 is present in the hippocampus (Fig. 6F and Supplementary Fig. S8B). Hence, nuclear expression of Zbtb20 and Tbr1 appear to be mutually exclusive in the developing pyramidal neurons in the AP and CP of the hippocampal formation. It is therefore conceivable that Zbtb20 represses the expression of Tbr1 during neurogenesis of the human hippocampus.
Here, we show that Zbtb20 during the development of CA1 pyramidal neurons directly represses genes, which control the subtype specification of projection neurons in the isocortex. Our data suggest that this novel gene regulatory mechanism consolidates the commitment of neuronal progenitors in the midline cortex to a CA1 pyramidal neuron fate.
Among the Zbtb20-repressed genes are Tbr1 and Sox5, which are required for the specification of corticothalamic projection neurons in the isocortex (Hevner et al. 2001; Kwan et al. 2008; Lai et al. 2008; Bedogni et al. 2010; Han et al. 2011; McKenna et al. 2011). Repression of these genes correlates with the rare occurrence of projections to the thalamus from the rodent CA1 field (Cenquizca and Swanson 2006). We further find that Zbtb20 directly represses Satb2, which is essential for the specification of callosal projection neurons (Alcamo et al. 2008; Britanova et al. 2008). This finding helps to explain why CA1 pyramidal neurons extend collaterals to the contralateral hippocampus via the hippocampal commissure as opposed to the corpus callosum (Cenquizca and Swanson 2007). The developmental specification of subcerebral projection neurons is controlled by Fezf2 and Bcl11b, where Bcl11b acts downstream of Fezf2 (Arlotta et al. 2005; Chen, Schaevitz, et al. 2005; Chen, Rasin, et al. 2005; Molyneaux et al. 2005; Chen et al. 2008). In Fezf2-deficient mice, isocortical expression of Bcl11b is downregulated and Satb2 expression is increased, implying that Fezf2 promotes Bcl11b expression through suppression of Satb2 that, in turn, directly represses Bcl11b (Alcamo et al. 2008; Britanova et al. 2008; Chen et al. 2008). Bcl11b is also expressed by developing CA1 pyramidal neurons (Nielsen et al. 2010), which correlates with the Zbtb20-mediated repression of Satb2 in these cells. We find that Zbtb20 directly represses the Fezf2 gene, which is in agreement with previous analyses of Fezf2-deficient mice, which showed that CA1 pyramidal neurons in contrast to isocortical and subicular neurons express Bcl11b independent of Fezf2 (Chen, Schaevitz, et al. 2005; Molyneaux et al. 2005). We further find that Zbtb20 represses the TF genes Cux1, Cux2, Mef2c, and Rorb, which are expressed by neuronal progenitor cells that give rise to upper-layer projection neurons and/or by upper-layer projection neurons themselves (Leifer et al. 1993; Nieto et al. 2004; Zimmer et al. 2004; Franco et al. 2012; Jabaudon et al. 2012). Together, our data suggest that Zbtb20 represses fate determinants that control the developmental specification of major subtypes of deep and upper-layer neurons in the isocortex. This is supported by the presence of ectopic deep and upper-layer neurons in the hippocampus and subiculum of Zbtb20-deficient mice (Xie et al. 2010; Rosenthal et al. 2012). These results are summarized in Figure 7.
The hippocampal formation develops adjacent to the cortical hem in the ventral part of the medial pallium. The cortical hem is a signaling center that imparts hippocampal competence to adjacent cortical progenitors during invagination of the dorsal midline between E10 and E12 (Grove et al. 1998; Lee et al. 2000; Mangale et al. 2008). After E12, the medial wall of the telencephalon has regionalized into the choroid plexus at the ventral tip, followed dorsally by the cortical hem and the hippocampal primordium. Through secretion of Wnt morphogens, the cortical hem regulates the proliferation of hippocampal progenitor cells and presumably stimulates these cells to express specific cell fate determinants that are important for hippocampal arealization (Lee et al. 2000; Machon et al. 2007; Mangale et al. 2008; Solberg et al. 2008; Karalay et al. 2011). For example, progenitors for DG granule neurons are located in the ventral part of the hippocampal neuroepithelium adjacent to the hem, and specification of these progenitors appear to depend on higher doses of Wnts when compared with lower doses for progenitors of CA3 and CA1 that are located progressively further away from the hem (Solberg et al. 2008). In line with this, canonical Wnt-signaling induces expression of the TF Prox1 that specifies the dentate granule neuron identity over a CA3 pyramidal neuron fate (Karalay et al. 2011; Iwano et al. 2012). Both CA1 and subicular pyramidal neurons are present in small patches at caudal levels of the Wnt3a-deficient cerebral cortex (Lee et al. 2000), which indicates that although the expansion of their progenitors are compromised, neurons with CA1 and subicular identities can be specified in the absence of Wnt3a signaling. From this, we infer that hem-derived Wnt signaling mainly affects the fate of progenitor cells in the DG and CA3 areas, but less for CA1 and subicular areas. It is therefore unclear how the 2 latter areas of the archicortex are specified next to the transitional isocortex. Our data on cortical misexpression of Zbtb20 indicate that the neuronal progenitors in the subiculum and in adjacent transitional cortical areas harbor similar developmental competencies to those in the CA1 field, since they only require the instructive role of Zbtb20 to become CA1 pyramidal neurons. This notion is further supported by the observations in Zbtb20-deficient mice that progenitor cells of both CA1 and the subiculum are able to generate upper-layer neurons in the absence of Zbtb20 (Xie et al. 2010; Rosenthal et al. 2012). The subiculum is composed of projection neurons that harbor identities, which are similar to deep-layer neurons in layers V and VI of the adjacent transitional isocortex (Ishizuka 2001). The majority of these neurons are generated prior to E15 (Angevine 1965) when Zbtb20 is not expressed in this area (Supplementary Fig. S1B). Following E15, Zbtb20 expression is pronounced in cells in the VZ and IZ, but not in the CP of the developing subiculum (Supplementary Fig. S1C), suggesting that Zbtb20 at this developmental stage represses genes that would otherwise endow the subiculum with upper-layer neurons (Fig. 7), as it was observed in Zbtb20-deficient mice. In the D6/Zbtb20S transgenic mice, where CA1 pyramidal neurons develop from subicular precursors, the misexpression of Zbtb20 occurs prior to E15 in both the IZ and the CP. Since expression of Zbtb20 in the wild-type subicular IZ is not sufficient to instruct CA1 identity in these cells, we hypothesize that the full commitment to a CA1 fate is a progressive process that requires early expression of Zbtb20 in both the IZ and the CP to prevent the specification of projection neurons with alternative deep-layer fates.
Projection neurons of the neocortex and the midline cortex develop from distinct pools of progenitor cells in the dorsolateral and medial pallium, respectively. Corticogenesis of the neocortex involves a pronounced expansion of layers III and IV compared with the transitional midline cortex (Millerand and Maitra 2002). This expansion of upper-layer neurons correlates with the presence of a population of Cux2-positive progenitor cells in the VZ and SVZ of the dorsolateral pallium that mainly generates upper-layer neurons (Franco et al. 2012). Contrary to the dorsolateral pallium, there is a lack of Cux2 expression in the VZ and SVZ of the medial pallium (Zimmer et al. 2004). Hence, the Cux2-positive pool of progenitors for upper-layer neurons does not appear to be present in the developing midline cortex. This notion is supported by our finding that the misexpression of Zbtb20 in the midline cortex leads to the generation of a highly homogenous population of CA1 pyramidal neurons that are organized in a compact stratum pyramidale. In contrast, ectopic expression of Zbtb20 in the neocortex results in 2 major neuronal subsets that settle in distinct lamina (Nielsen and Jensen, unpublished data). This phenotype most likely results from transformation by Zbtb20 of both Cux2-positive and Cux2-negative progenitor cells.
Although it has not been a focus of this study, it should be noted that Zbtb20 is also expressed by cells in the VZ, SVZ, and IZ of the retrosplenial cortex at E17 (Supplementary Fig. S1C). These cells are likely to give rise to the late specified layer II neurons of the retrosplenial cortex (Zgraggen et al. 2012) since Zbtb20 is expressed by these neurons postnatally (Rosenthal et al. 2012). It is therefore possible that Zbtb20 in addition to its roles in the archicortex may be involved in the subtype specification of developing upper-layer neurons in the retrosplenial cortex. Although this is intriguing, our transgenic mouse model does not allow us to address this question, since misexpression of Zbtb20 in this area occurs so early that Zbtb20 consolidates CA1 fate specification before differentiation of upper-layer neurons begins.
Finally, we find that Zbtb20 and the Zbtb20 target gene Tbr1 are expressed in complementary patterns in the developing human hippocampus, implying that Zbtb20 regulates human hippocampal neurogenesis by a similar transcriptional repressor mechanism. Moreover, the evolutionary conserved expression of Zbtb20 in the developing human hippocampus highlights the importance of continued characterization of Zbtb20 target genes. In particular, since the Zbtb20 gene is linked to human disorders, such as microdeletion syndromes, that are associated with developmental abnormalities and autism (Molin et al. 2012).
This work was supported by grants from the Augustinus Foundation, Aase and Ejnar Danielsen's Foundation, and the Lundbeck Foundation.
We thank M.R. Hansen, C.S. Landré, S. Forchhammer, H. Hadberg, P.S. Froh, and H. Nguyen for excellent technical assistance, Dr T.A. Kruse for advice on design of the microarray experiment, and the Epigenetic Services team at Active Motif for help with the ChIP-based assays. Conflict of Interest: None declared.