Genomic analysis of transcriptional networks directing progression of cell states during MGE development

Background Homeodomain (HD) transcription factor (TF) NKX2–1 critical for the regional specification of the medial ganglionic eminence (MGE) as well as promoting the GABAergic and cholinergic neuron fates via the induction of TFs such as LHX6 and LHX8. NKX2–1 defines MGE regional identity in large part through transcriptional repression, while specification and maturation of GABAergic and cholinergic fates is mediated in part by transcriptional activation via TFs such as LHX6 and LHX8. Here we analyze the signaling and TF pathways, downstream of NKX2–1, required for GABAergic and cholinergic neuron fate maturation. Methods Differential ChIP-seq analysis was used to identify regulatory elements (REs) where chromatin state was sensitive to change in the Nkx2–1cKO MGE at embryonic day (E) 13.5. TF motifs in the REs were identified using RSAT. CRISPR-mediated genome editing was used to generate enhancer knockouts. Differential gene expression in these knockouts was analyzed through RT-qPCR and in situ hybridization. Functional analysis of motifs within hs623 was analyzed via site directed mutagenesis and reporter assays in primary MGE cultures. Results We identified 4782 activating REs (aREs) and 6391 repressing REs (rREs) in the Nkx2–1 conditional knockout (Nkx2–1cKO) MGE. aREs are associated with basic-Helix-Loop-Helix (bHLH) TFs. Deletion of hs623, an intragenic Tcf12 aRE, caused a reduction of Tcf12 expression in the sub-ventricular zone (SVZ) and mantle zone (MZ) of the MGE. Mutation of LHX, SOX and octamers, within hs623, caused a reduction of hs623 activity in MGE primary cultures. Conclusions Tcf12 expression in the SVZ of the MGE is mediated through aRE hs623. The activity of hs623 is dependent on LHX6, SOX and octamers. Thus, maintaining the expression of Tcf12 in the SVZ involves on TF pathways parallel and genetically downstream of NKX2–1. Electronic supplementary material The online version of this article (10.1186/s13064-018-0119-4) contains supplementary material, which is available to authorized users.

such as promoters and enhancers. By analyzing epigenetic modifications and transcriptional changes in TF knockouts, we have started to uncover the genomic networks and molecular mechanisms that direct brain development [1]. In-depth understanding of the genetically encoded wiring of the brain is important as perturbation of transcription pathways is implicated in disorders such as autism and intellectual disability [2]. Distantly acting REs have been identified based on conservation and activity [3,4]. Their spatial activity and dynamic genomic contacts can be predicted using a combination of TF binding profiling, genomewide 3D chromosome organization mapping and CRISPR/Cas9 editing [5][6][7][8][9][10] Mouse genetic experiments have elucidated the functions of many TFs in the development of the subpallial telencephalon [11,12]. These studies show that the HD protein NKX2-1 is required for regional specification of the MGE by repressing alternative identities, as well as promoting GABAergic and cholinergic cell fates via the induction of TFs such as LHX6 and LHX8 [13][14][15][16][17]. By integrating genomic data with mouse genetics, we confirmed the repressive function of NKX2-1, however its role in transcriptional activation remains unclear. Moreover, additional data suggests that genes genetically downstream of NKX2-1, such as LHX6 and LHX8, are responsible for the loss of gene expression observed in the Nkx2-1cKO [18,19]. Altogether, the genetic program and molecular mechanisms responsible for promoting GABAergic and cholinergic neuron phenotypes, downstream of NKX2-1 remains largely unexplored.
To investigate the signaling pathways of MGE development downstream of NKX2-1, we extended our earlier analysis of the genomic network directing MGE development that is altered in the Nkx2-1 mutant. First we evaluated all loci that showed an epigenetic change, independent of NKX2-1 binding. Via an epigenomic analysis of the NKX2-1 mutant MGE we characterized a large set REs that are implicated in mediating transcriptional repression and activation.
Using a combination of genomics, de novo motif analysis, CRISPR engineering and primary culture assays we characterize REs and TFs central to patterning of the subpallial telencephalon and promoting MGE characteristics. Gene ontology (GO) analysis showed an enriched association of REs activating transcription (aREs) with E-box binding basic-Helix-Loop-Helix (bHLH) TFs. Using CRISPR engineering we deleted hs623, an intronic aRE of the Tcf12 gene which encodes a bHLH TF. Deletion of hs623 reduced Tcf12 expression in the MGE. De novo motif analysis combined with TF motif mutations, showed that OCT/POU and SOX motifs are required for hs623's ability to promote transcription in the MGE.

Mice
The Nkx2-1cKO was earlier described in Sandberg et al. 2016[18] and generated using mice strains previously reported: Nkx2-1f/f [20], Olig2-tva-Cre [21] and AI14 Cre-reporter [22]. All experiments with animals complied with federal and institutional guidelines and were reviewed and approved by the UCSF Institutional Animal Care and Use Committee.

Histology
Immunofluorescence was performed on 16 µm cryosection as previously described [25]. In situ hybridization was performed as previously described [26].
The following primers were used generate the templates used for the in situ probes: Tcf12_F, TCTCGAATGGAAGACCGC; Tcf12_R, CTCCCTCCTGCCAGGTTT

Dissection of Embryos
RT-qPCR and primary culture experiments were performed on E13.5 microdissected MGE. All MGE dissections were performed as follows; the dorsal boundary was defined by the sulcus separating lateral ganglionic eminence (LGE) and MGE. The caudal end of the sulcus defined the caudal boundary.
Septum was removed.

Gene expression analysis in hs623KO
To assay differential gene expression in the hs623KO RNA was purified using RNEasy Mini (Qiagen) and cDNA was generated using Superscript III ® First-Strand Synthesis System for RT-PCR (Invitrogen). RT-qPCR analysis was performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems) using SYBR GreenER qPCR SuperMix (Invitrogen, Cat. No. 11760-100).
Unpaired t-test was used to test significance in gene expression between hs623WT and hs623KO using SDHA as internal control [27,28].

Analysis of hs623 activity in MGE primary MGE cultures
MGE tissue was dissected from E13.5 embryos, triturated and plated onto 24well plates (1 embryo/2wells). Primary cultures were transfected with a total of 500ng DNA using Lipofectamin 2000 (Thermo Fisher) and cultured in Neurobasal Medium (Thermo Fisher) supplemented with 0.5% Glucose, GlutaMAX (Thermo Fisher Scientific) and B27 (Thermo Fisher Scientific). Luciferase assays were performed 48h after transfection using Dual Luciferase Reporter Assay System (Promega). Unpaired t-test was used to test significance between the variants of hs623.

ChIP-Seq Computational Analysis
Differential ChIP-seq analysis was performed as described in Sandberg et al.

De novo motif analysis
Motif analysis performed using RSAT [29] with default settings and genomic, aREs or rREs as background.

Identification of the genomic regulatory network directing MGE identity
We have previously shown that the combined binding of NKX2-1 and LHX6 is a predictive indicator of REs that mediate transcriptional activation in the subventricular (SVZ) and mantle zone (MZ) of the MGE in the developing subpallial telencephalon [18]. There is evidence that NKX2-1 generally acts as a repressor in MGE progenitors (in the ventricular zone [VZ]), whereas LHX6, and potentially other TFs and signaling pathways, some of which are genetically downstream of NKX2-1, are important for activating transcription in the SVZ and MZ of the MGE [18,30]. By studying aREs, we aimed to further explore the molecular mechanisms underlying the transcriptional network directing differentiation of the secondary progenitors in the SVZ.
First we identified aREs and rREs by assessing the genome-wide changes of the two histone marks H3K27ac and H3K27me3 at H3K4me1 positive REs comparing the WT and Nkx2-1cKO MGE [18]. We defined aREs based on the following two criteria; 1) reduced H3K27ac and, 2) increased H3K27me3 in the Nkx2-1cKO. We defined rREs based on the following two criteria; 1) increased H3K27ac and, 2) reduced H3K27me3 in the Nkx2-1cKO (see Methods).
Based on these criteria we identified 4782 aREs and 6391 rREs in the Nkx2-1cKO. See Additional file 1 for a complete list of aREs and rREs. To analyze the in vivo activity patterns of the aREs and rREs we examined transgenic enhancer activity patterns of E11.5 forebrain enhancer activity patterns available in the VISTA database (see VISTA data base; https://enhancer.lbl.gov/) [7]. The activities of rREs were highest in cortex (62%) and LGE and dorsal MGE (52%) and lowest in the ventral MGE (24%)( Figure 1A and 1B [hs848, hs1172 and hs1187]). Note, hs1187 is active in the, NKX2-1 negative, dorsal most part of the MGE illustrating the repressing activity of NKX2-1 on this type of rREs [31]. In contrast, aREs have the highest activities in the MGE (53%) when compared to their activities in the LGE (50%) and cortex (41%) ( Figure 1A and 1B [hs676, hs957 and hs1041]). We also found a higher activity of MGE positive aREs in the SVZ and MZ compared to the VZ, consistent with our previous results for NKX2-1 bound aREs and rREs ( Figure 1B and 1C) [18]. See Additional file 1 for a full list of aREs and rREs VISTA transgenics.
To identify TFs motifs enriched in the aRE and rREs we performed a de novo motif discovery using RSAT [29]. This analysis showed a number of motifs enriched in both aREs and rREs such as SOX motifs, homedomain binding motifs (HOX and POU6f2) and motifs recognized by zinc finger TFs (e.g. SP1 and ZNF384) ( Figure 1D and 1E). Additional analysis identifying motifs differentially enriched between aREs and rREs showed that aREs have a high frequency of E-boxes ( Figure 1E). Interestingly, we found that rREs are enriched in motifs consistent with the binding site of the TF MEIS2 ( Figure 1D). The Meis2 gene is repressed by NKX2-1, and in turn, its RNA is strongly up-regulated in the MGE of the Nkx2-1cKO [18]. These data suggest that Meis2 is central to activating a genomic network promoting LGE and caudal ganglionic eminence We then examined enrichment of annotation terms among the aREs and rREs candidate target genes using GREAT [32]. Top-ranked GO terms for rREs target genes were associated with WNT signaling (beta-catenin binding and PDZ domain binding), transcriptional regulation (such as RNA polymerase II transcription co-activator activity), and enhancer sequence-specific DNA binding ( Figure 1F). Looking specifically at the associated genes for the rREs containing MEIS2 binding motifs we found several genes (Isl1, Ebf1, Tle4, Zfp503, Efnb1 and Efnb2) with higher expression in the LGE and CGE than the MGE. These findings support the hypothesis that MEIS2 directs LGE and CGE identities. The top-ranked GO terms for aREs target genes were associated with phosphatase activity, E-box binding proteins, L-glutamate transmembrane transporter activity and transmembrane-ephrin receptor activity [32]( Figure 1F). Two E-box binding TFs, Tcf4 and Tcf12, which are in the region of a large number of aREs, have reduced MGE SVZ and MZ expression in the Nkx2-1cKO [18]. In combination with the high frequency of E-boxes in aREs, our data suggests that Tcf4 and Tcf12 are components of the genomic network regulating gene expression in secondary progenitors of the MGE that are genetically downstream of NKX2-1.

In vivo characterization of hs623 in the MGE of the forebrain
To learn more about the Tcf12 expression and the transcriptional pathways integrated in the aRE network downstream of NKX2-1, we examined aRE hs623, a highly evolutionarily conserved 914 base pair (bp) sequence that is in an intron of the Tcf12 locus (Figure 2A and 2B). A previous transgenic study show that hs623 drives LacZ expression at E11.5 [33]. The hs623 transgene is active in the forebrain, hindbrain and the spinal cord ( Figure 2C-2E). A coronal section through the telencephalon shows that hs623 activity is restricted to the SVZ and MZ of the MGE, and perhaps labels cell tangentially migrating into the LGE ( Figure 2E). This pattern of activity is supported by histone ChIP-seq analysis of the MGE showing that this locus has histone modifications that are characteristic of active enhancer elements (Figure 2A [H3K4me1+ and H3K27ac+] and 2B). Of note, ChIP-seq analysis of the MGE Nkx2-1cKO shows reduced H3K27ac, providing evidence that the activity of the locus is dependent on the activity of the NKX2-1 and/or its target TFs, as reported earlier ( Figure 2B) [18].

Motif logic direct region specific transcriptional activity
Hs623 is flanked by two highly conserved regions and the activity of one of the regions (hs357) has been tested in vivo [33]. Similar to hs623, hs357 is active in the spinal cord, but unlike hs623 it is active in the pretectum and it lacks activity in the telencephalon, including the MGE (Figure 2F and 2G). Therefore, despite the close proximity of hs623 and hs357, they show differences in regional activity, suggesting that their regional activities are more likely due to differences in their primary nucleic acid sequence rather then their genomic location.
Consistent with the lack of MGE activity, hs357 lacks LHX6 consensus motifs ( Figure 2B). On the other hand, hs623 has four LHX6 consensus motifs ( Figure   2H). Surprisingly, even though hs623 has NKX2-1 binding, it contains no NKX2-1 consensus motifs. However, when extending the sequence analysis to the regions flanking hs623, we find three NKX2-1 consensus motifs, two within hs357 ( Figure 2B). This could explain why we detect NKX2-1 binding a wide region that covers both hs623 and hs357.

CRISPR/Cas9 mediated deletion of hs623 in vivo
To functionally test the requirement of hs623 in vivo, we deleted hs623 using CRISPR/Cas9 (see VISTA database; http://enhancer.lbl.gov). A pair of sgRNAs were designed to delete the 734bp core sequence of hs623, which has NKX2-1 and LHX6 binding ( Figure 2B and 2H). Micorinjection of sgRNAs and Cas9 generated a total of 22 pups. 23% (5 of 22) of the pups carried the desired hs623 deletions and the induced DNA breaks were distributed within 20bp of the predicted cutting site (5' and 3' of hs623, Figure 2H). To minimize potential off target effects we outcrossed the F0 transgenic founders to wild-type CD1 mice.
Four of five founders were fertile and generated a F1 generation; these animals were intercrossed to generate homozygous F2 hs623KO animals. p=0.7367). Due to the overall similarity of the four fertile founders we decided to focus the following analysis on one of the founders (F0: 2458, Figure 2H).

Deletion of hs623 reduces Tcf12 mRNA levels in the SVZ of MGE
Hs623 is a Tcf12 intragenic RE that in transgenic assays activates transcription in the SVZ of the MGE (Figure 2C-E). As noted above, its activity is partly dependent on NKX2-1 activity and Tcf12 transcription is specifically reduced in the SVZ of the MGE in the Nkx2-1cKO ( Figure 2B) [18]. Together, these data suggest that hs623 could be a RE activating Tcf12 transcription in the MGE. To test this hypothesis, we performed RTqPCR on the MGE from hs623WTs and hs623KOs at E13.5. Primers were designed to target all known mouse proteincoding and non-protein-coding genes in the NCBI RNA reference sequences collection that are found 450 kb up-and downstream of hs623 ( Figure 3A). From the RTqPCR we found no significant difference in the expression of the following genes in this region: Myzap, Cgln1, Zfp280d and Mns1 ( Figure 3B). Tcf12 RNAs include a variety of splice variants. Because of this we designed three separate primer pairs to specifically interrogate the different splice variants of Tcf12 ( Figure 3A). We found a reduction in the expression of the short isoforms of Tcf12 isoform 3 and 4 ( Figure 3B, see Tcf12_v1/3-4 and Tcf12_v1/ [3][4][5]. Notably, we did not find any difference in the expression levels of the longer isoforms 1 and 2 of Tcf12 ( Figure 3B, see Tcf12_v1/2-2). Together, these results show that Tcf12 transcription in the MGE is enhanced by hs623.
To obtain spatial information about the reduction of Tcf12 within the MGE we compared the distribution of Tcf12 RNA between WT and hs623KO telencephalon at E13.5 using in situ RNA hybridization. Normally, Tcf12 is broadly expressed in the VZ in the pallium and subpallium. In the ganglionic eminences Tcf12 is also expressed in the SVZ and MZ, with a markedly higher expression in the MGE compared to the LGE. On the other hand, in the hs623KO we observed a reduction of Tcf12 expression that appeared to be specific to the SVZ and MZ of the MGE ( Figure 3C and 3D). This result is consistent with the spatial activity of hs623, which is restricted to the SVZ of the MGE.

Combined activity of POU and SOX TFs are required to maintain gene expression downstream of NKX2-1 in the MGE
To test the functional requirement of the LHX6 motifs in hs623 we made site directed mutations that removed all four LHX6 motifs (hs623ΔLHX). In MGE primary cultures the activity of hs623ΔLHX was reduced by half when compared to the non-mutated hs623 (hs623WT, Figure 4A and 4B). Together, these experiments provide evidence that hs623 activity, in part, depends on LHX6 and LHX8 and that there are additional TFs and signaling pathways required for the activity of hs623. Our earlier motif analysis of aREs discovered an enrichment of additional motifs such as HD-binding motifs (POU6f2 and HOX), SOX motifs and E-boxes ( Figure 1E). To identify additional TF pathways responsible for the activity of hs623 we looked at the other identified de novo motifs within hs623 ( Figure 1D). Located in the center of hs623 we found two octamers (bound by POU TFs), of which one is adjacent to a SOX motif. Octamers are known to pair with SOX motifs to form central functional units regulating development in various cell types [34][35][36]. Initially, we analyzed the activity of the two individual octamers by generating single mutations of the two motifs ( Figure 4A, hs623ΔOCT1 and   hs623ΔOCT2). Mutating octamer 1 (hs623ΔOCT1) caused a significant reduction of hs623 activity in MGE primary cultures, whereas mutating octamer 2 (hs623ΔOCT2) had no significant effect on hs623 activity ( Figure 4B). Octamer 2 is located 3bp from a SOX consensus motif ( Figure 4A). To assess the requirement of this combined motif for hs623 activity, we generated a compound mutant with a combined mutation of octamer 2 and the paired SOX motif (hs623 ΔOCT2+SOX). Hs623ΔOCT2+SOX showed a significantly reduced activity when compared to hs623WT as well as, the two individual single mutants, hs623 ΔOCT2 and hs623ΔSOX ( Figure 4B).
Altogether, our experiments show that Tcf12 expression in the SVZ of the MGE is mediated, at least in part, through hs623, a RE that is strongly dependent on its OCT and SOX motifs and partially dependent on its LHX6 motifs. We have previously shown that gene expression in the SVZ of the MGE (including Tcf12) largely depends on NKX2-1 activity [18]. Existing mechanistic data show that NKX2-1 acts as a transcriptional repressor. Therefore, our findings suggest that the loss of Tcf12 expression in the SVZ of the MGE Nkx2-1cKO is not due to the direct regulation of Tcf12 by NKX2-1, but is a secondary effect due to the loss of additional TFs expressed genetically downstream of NKX2-1, including LHX6, LHX8, OCT and SOX TFs.

DISCUSSION
Technical advancements in genome wide sequencing, chromosome capture and CRISPR/Cas9 technologies are increasing our understanding of genome organization. These data, combined with data showing RE activities in vivo (https://enhancer.lbl.gov/), TF binding and other epigenetic genomic data, and spatial gene expression data (http://www.brain-map.org/, http://www.eurexpress.org/ee/intro.html), are enabling the field to begin elucidating the genomic networks and the molecular mechanisms that direct brain development. Herein we have used many of these approaches to analyze gene expression in the developing mouse MGE. In the context of the Nkx2-1cKO mouse, our analysis of differential (WT vs. cKO) histone ChIP-seq data, and de novo sequence motif analysis, has provided evidence for additional TFs, REs, and signaling pathways that direct MGE development.
In this study, we showed that Tcf12 expression in the SVZ of the MGE is We have earlier demonstrated that NKX2-1 represses transcription in the MGE, similar to other NKX HD TFs that specify ventral parts of the developing neural tube [37,38]. Even at aREs, identified in the Nkx2-1cKO MGE, the NKX2-1 motifs mediate transcriptional repression, as exemplified by the intragenic Tgfb3 RE in Sandberg et al. 2016[18]. On the other hand, in the case of both the Tgfb3 RE and hs623, LHX6 motifs promote enhancer activity. If NKX2-1 only represses transcription, it is unclear how loci such as LHX6 and LHX8 fail to be activated in the Nkx2-1 mutants [16][17][18]. Furthermore it is unclear why NKX2-1 also binds loci that have reduced activity in the Nkx2-1cKO. These results suggest that, in some contexts, NKX2-1 may have an activating function. NKX2-1 binding to these loci might be required to keep them poised for subsequent activation by TFs and signaling pathways parallel and genetically downstream of NKX2-1, such as LHX, OCT, SOX and bHLH TFs. A similar model was presented in two studies looking at motor neuron development. In these cells, combinations of NEUROG2 (bHLH TF), LHX3, ISL-1, ONECUT1 and EBF direct the progression of the motor neuron fate through distinct sets of REs [8,9]. Similar to these models, we find an enrichment of LHX6 binding and e-boxes at aREs, a group of REs with a preferential activity in the SVZ of the MGE. This combination of TF binding and motif enrichment is not seen at NKX2-1 bound rREs, that have a relatively low MGE activity. These data highlight similarities in the molecular mechanisms that direct MGE and motor neuron development over time. In addition to combinatorial activity with other TFs, the activity of NKX2-1 might be affected by changes to chromatin modifications at specific loci over time.
The seemingly dual activity of NKX2-1 in the MGE is similar to its double-edged characteristics in regulating cancer development and progression. In this context, NKX2-1 has a role as lineage-survival oncogene in developing lung cancer tumors. On the other hand, NKX2-1 expression is also associated with a favorable prognosis in affected patients, due to its capacity to attenuate the invasive capacity of carcinomas [39]. Interestingly, this has been shown to be mediated through an abrogation of cellular response to TGFβ induced EMT, a signaling pathway that is directly repressed by NKX2-1 in the MGE [18,40]. By identifying the mechanisms through which NKX2-1 operate in the subpallial telencephalon we might also learn more about its enigmatic role in tumor biology.
Our data provides evidence that LHX6 activates Tcf12 expression through the hs623 RE based on three observations: 1) LHX6 binding to hs623 in vivo, 2) a requirement of LHX6 motifs for hs623 activity in MGE cell culture assays, and 3) reduced expression of LHX6 in the Nkx2-1cKO.
In addition, our data show a combination of NKX2-1 and LHX6 binding to aREs Here, in our new analysis of the Nkx2-1cKO, we found a large number aREs.
Some of these are near the loci of the Tcf4 and Tcf12 bHLH TF encoding genes.
The Nkx2-1cKO shows a near complete loss of Tcf12 expression in the SVZ and MZ of the MGE. We found an aRE intronic to Tcf12 (hs623) that has activity in the SVZ and MZ of the MGE ( Figure 2E). Deletion of hs623 leads to a reduced Tcf12 expression in the VZ and MZ of the MGE. This result suggests that Tcf12 expression is regulated through several aREs, including hs623, and that there is redundancy between these REs. Enhancer redundancy has been demonstrated in the developing telencephalon and limb where REs sharing a similar spatiotemporal activity provides robustness to gene expression [54,55]. We also find that there are different genetic programs directing Tcf12 expression in various cell types of the MGE. Tcf12 expression is initiated in the VZ of the MGE; this expression is largely unaffected in the Nkx2-1cKO, indicating that Tcf12 expression in this region is not mediated through hs623 and largely NKX2-1 independent.
Altogether, these data provide evidence of transcriptional circuitry that connects the initiation of MGE fate in the VZ by Nkx2-1 and Otx2, to the maturation of cells in the SVZ and MZ by driven through REs such as hs632, whose activity integrates signals from LHX, OCT, SOX and bHLH TFs [16,18,56]. Future studies will investigate how TFs, chromatin-binding, -reading and -remodeling proteins integrate to direct GABAergic and cholinergic development in the subpallial telencephalon.

Consent for publication
Not applicable.

Availability of data and material
The datasets supporting the conclusions of this article are available in the NCBI's GEO repository, GEO Series accession number GSE85705 (https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE85705). Additional material is available from the corresponding author upon request.