ABSTRACT
During neocortical development, neurons are produced by a diverse pool of neural progenitors. A subset of progenitors express the Cux2 gene and are fate restricted to produce certain neuronal subtypes; however, the upstream pathways that specify these progenitor fates remain unknown. To uncover the transcriptional networks that regulate Cux2 expression in the forebrain, we characterized a conserved Cux2 enhancer that recapitulates Cux2 expression specifically in the cortical hem. Using a bioinformatic approach, we identified putative transcription factor (TF)-binding sites for cortical hem-patterning TFs. We found that the homeobox TF Lmx1a can activate the Cux2 enhancer in vitro. Furthermore, we showed that Lmx1a-binding sites were required for enhancer activity in the cortical hem in vivo. Mis-expression of Lmx1a in hippocampal progenitors caused an increase in Cux2 enhancer activity outside the cortical hem. Finally, we compared several human enhancers with cortical hem-restricted activity and found that recurrent Lmx1a-binding sites are a top shared feature. Uncovering the network of TFs involved in regulating Cux2 expression will increase our understanding of the mechanisms pivotal in establishing Cux2 lineage fates in the developing forebrain.
INTRODUCTION
During forebrain development, neural progenitor cells give rise to many different types of neuronal and glial cells that form the various telencephalic structures and circuits. This vast cellular diversity arises through the interplay between early tissue patterning pathways and gene regulatory networks (GRNs). Early in development, multiple tissue organizers and signaling centers provide morphogenic cues that govern regional identity and size. Within these different regions, complex transcriptional programs further diversify multipotent progenitor cells toward specific cell fates. The transcription factors (TFs) that establish the different GRNs to specify cell fates often work by binding gene regulatory elements, such as enhancers, to boost or suppress expression of target genes. Following transcriptional activation of GRNs, neural progenitors divide and eventually differentiate into specified cells. Identifying the signaling and transcriptional networks that establish regional identity and subtype fate specification during embryonic development will greatly enhance our understanding of forebrain development and function.
The TF Cut-like homeobox 2 (Cux2) is dynamically expressed in complex spatiotemporal patterns in the developing mouse forebrain (Zimmer et al., 2004). During early brain development, a subset of neural progenitors weakly express Cux2 transcripts in a ‘salt and pepper’ pattern (Franco et al., 2012). We previously fate mapped the lineage output of Cux2+ progenitors in the neocortex and found that this subset of neural progenitors are fate restricted to produce late-born corticocortical neurons in upper layers (Franco and Müller, 2013; Franco et al., 2012; Gil-Sanz et al., 2015). Our studies indicated that Cux2+ progenitors in the developing forebrain are committed to this fate even before the onset of neurogenesis. This model has been a matter of controversy, as other studies have suggested that neocortical progenitors are homogeneous and multipotent (Eckler et al., 2015; Gao et al., 2014; Guo et al., 2013). However, recent studies using a variety of fate-mapping methods have independently confirmed the existence of neural progenitors that only generate late-born corticocortical neurons (García-Moreno and Molnár, 2015; Llorca et al., 2018 preprint). Nevertheless, a consensus model has yet to emerge and the underlying mechanisms that restrict these progenitors to specific cell fates remain largely unknown.
Cux2 knockout mice do not display any significant phenotype with respect to progenitor cell fate specification (Cubelos et al., 2008), implying that Cux2, although a useful marker for a fate-committed progenitor population, does not necessarily instruct fate in this context. We reasoned that a deeper understanding of Cux2+ cell fate commitment in forebrain progenitors could be achieved by uncovering the upstream GRNs responsible for the complex patterns of Cux2 expression. Interestingly, neural progenitors in the dorsal telencephalic midline (DTM) strongly express Cux2 in a more complete pattern than progenitors in adjacent regions, suggesting that this forebrain region might contain critical transcriptional regulators of the Cux2 locus.
Previous studies have uncovered enhancers active in the developing mouse telencephalon, including an 856 bp element in intron 2 of the Cux2 genomic locus that could drive strong transgene expression in the DTM (Hasenpusch-Theil et al., 2012; Visel et al., 2008). Here, we have characterized this element as an active enhancer in the developing forebrain and show that it is specifically active in the cortical hem, but not in the adjacent hippocampus or neocortex. We further analyzed this enhancer for possible upstream regulators of Cux2 expression. Among many bioinformatically identified candidates, we tested several transcription factors known to function in or be expressed within the cortical hem. Using an in vitro approach, we demonstrate that Lmx1a is a strong activator of the Cux2 hem-specific enhancer. We further show by chromatin immunoprecipitation from embryonic forebrain that Lmx1a can bind the endogenous Cux2 enhancer region. Additionally, shRNA-mediated knockdown of Lmx1a in vivo caused a decrease in Cux2 enhancer activity in the cortical hem. Conversely, Lmx1a gain of function in the hippocampus, a region normally devoid of Lmx1a expression, increased activity of the Cux2 enhancer. Finally, we analyzed other enhancers that exhibit specific activity in the cortical hem and identify recurrent Lmx1a-binding sites as a common motif shared between these distinct hem-specific enhancers. Our results suggest Lmx1a functions as an upstream regulator of a conserved Cux2 enhancer in the cortical hem, and raise the possibility that Lmx1a is a crucial TF in the GRN that specifies cortical hem fate.
RESULTS
Early forebrain expression of Cux2 begins at the dorsal telencephalic midline
To better understand when and where the earliest transcriptional regulators of Cux2 are active in the developing telencephalon, we sought to define the temporal and spatial patterns of Cux2 gene expression. We crossed Cux2Cre/+ mice to the Ai9 Cre-reporter line and used recombination (tdTomato+) as a readout of the cumulative transcriptional history of the Cux2 genomic locus. Cux2Cre/+;Ai9fl/+ brains were analyzed at E9.5, E10.5, E12.5 and E14.5 (Fig. 1). We found that the earliest consistent pattern of recombined cells in the forebrain first appeared in the dorsal telencephalic midline (DTM) at ∼E9.5 (Fig. 1A). At this age, a few recombined cells also began to appear scattered very sparsely throughout the adjacent hippocampal and neocortical neuroepithelium (Fig. 1A). By E10.5, the entire DTM was recombined, and the number of tdTomato+ neuroepithelial cells was increased in the neocortex (Fig. 1B). At E12.5 and E14.5, the DTM is reorganized into two distinct structures: the cortical hem and choroid plexus epithelium (Grove et al., 1998). Essentially, all cells in the cortical hem and choroid plexus were recombined at E12.5 (Fig. 1C) and E14.5 (Fig. 1D). In contrast, only a fraction of cells in the adjacent hippocampal primordium and neocortex were recombined (Fig. 1C,D). In fact, we observed a strikingly sharp border of complete-to-sparse recombination at the boundary between the cortical hem and the hippocampal primordium. To trace the long-term lineages of these Cux2+ cells in the DTM to later ages, we next used Cux2-CreERT2 mice (Franco et al., 2012) for temporal fate mapping. We crossed Cux2-CreERT2 mice to the Ai9 reporter strain and administered a single dose of tamoxifen at E10.5 (Fig. 1E). At E16.5, many cells in the cortical hem and choroid plexus were recombined (Fig. 1F). By postnatal ages, the cortical hem neuroepithelium transforms in the fimbria. We found many recombined cells in the fimbria at postnatal day 10 after tamoxifen administration at E10.5 (Fig. 1G). These data indicate that forebrain activation of the Cux2 locus occurs earliest and most uniformly in the DTM, including the cortical hem and choroid plexus.
Spatiotemporal development of Cux2 expression in murine telencephalic progenitors. (A-D) Coronal sections of forebrains from Cux2-Cre;Ai9 embryos showing recombination as a cumulative readout of Cux2 expression. Recombined cells express tdTomato (red). Sections were counterstained for nuclei with DAPI (blue). Boxes outline the areas shown in more detail on the right: the dorsal midline (middle panels) or neocortex (right panels). (A) E9.5: recombination is apparent in the dorsal-most region of the telencephalic neural tube and in scattered cells in the telencephalon. (B) E10.5: recombination becomes robust in the nascent cortical hem and choroid plexus, with salt-and-pepper recombination in the neocortex. (C) E12.5: recombination is nearly ubiquitous in the cortical hem, while still mosaic in the neocortex. (D) E14.5: recombination is complete in the cortical hem and much of the choroid plexus, while the neocortex exhibits a still expanding, mosaic pattern. (E) Temporal fate mapping through tamoxifen induction. Cux2-CreERT2 mice allow temporary activation of CreERT2 only in Cux2+ cells for a period up to 24 h. Subsequently, Cre recombines the Ai9 allele for permanent labeling of cells with tdTomato. (F) Coronal section of an E16.5 Cux2-CreERT2;Ai9 forebrain labeled at E10.5 showing Cux2-lineage cells (tdTomato+ and arrows) within the cortical hem (dashed box). (G) Coronal section of a P10.5 Cux2-CreERT2;Ai9 forebrain labeled at E10.5 showing recombined Cux2-lineage (tdTomato+) cells (arrows) residing within the cortical hem-derived fimbria (dashed box). Scale bars: 200 µm (A-D,F,G, left); 50 µm (A-D, middle and right); 100 µm (F,G, right). CH, cortical hem; CP, choroid plexus; F, fimbria; HP, hippocampal primordium.
Spatiotemporal development of Cux2 expression in murine telencephalic progenitors. (A-D) Coronal sections of forebrains from Cux2-Cre;Ai9 embryos showing recombination as a cumulative readout of Cux2 expression. Recombined cells express tdTomato (red). Sections were counterstained for nuclei with DAPI (blue). Boxes outline the areas shown in more detail on the right: the dorsal midline (middle panels) or neocortex (right panels). (A) E9.5: recombination is apparent in the dorsal-most region of the telencephalic neural tube and in scattered cells in the telencephalon. (B) E10.5: recombination becomes robust in the nascent cortical hem and choroid plexus, with salt-and-pepper recombination in the neocortex. (C) E12.5: recombination is nearly ubiquitous in the cortical hem, while still mosaic in the neocortex. (D) E14.5: recombination is complete in the cortical hem and much of the choroid plexus, while the neocortex exhibits a still expanding, mosaic pattern. (E) Temporal fate mapping through tamoxifen induction. Cux2-CreERT2 mice allow temporary activation of CreERT2 only in Cux2+ cells for a period up to 24 h. Subsequently, Cre recombines the Ai9 allele for permanent labeling of cells with tdTomato. (F) Coronal section of an E16.5 Cux2-CreERT2;Ai9 forebrain labeled at E10.5 showing Cux2-lineage cells (tdTomato+ and arrows) within the cortical hem (dashed box). (G) Coronal section of a P10.5 Cux2-CreERT2;Ai9 forebrain labeled at E10.5 showing recombined Cux2-lineage (tdTomato+) cells (arrows) residing within the cortical hem-derived fimbria (dashed box). Scale bars: 200 µm (A-D,F,G, left); 50 µm (A-D, middle and right); 100 µm (F,G, right). CH, cortical hem; CP, choroid plexus; F, fimbria; HP, hippocampal primordium.
Cux2 regulatory element contains characteristics of an active enhancer and recapitulates the endogenous Cux2 expression pattern in the cortical hem
Non-coding gene regulatory elements, such as enhancers, can act as crucial platforms for TFs that drive cell fate decisions (Pattabiraman et al., 2014). To gain insights into some of the transcriptional programs that specify area and subtype fate in the telencephalon, we sought to identify enhancers that could recapitulate Cux2 expression in the developing forebrain. Previous studies identified an 856 base pair (bp) region within the human Cux2 gene (hs611) that exhibits extreme human-rodent sequence conservation, suggesting an important functional role for this non-coding element (Visel et al., 2008). Indeed, both the human element (Pattabiraman et al., 2014; Visel et al., 2008, 2013) and the corresponding murine region (Hasenpusch-Theil et al., 2012) can drive restricted expression of a lacZ reporter gene in transgenic mouse embryos, indicating their role as functional enhancer elements. This enhancer lies within intron 2 of the Cux2 gene (Fig. 2A). Using DNaseI hypersensitivity (ENCODE Project Consortium, 2012) and histone modification (Shen et al., 2012) data from the UCSC Genome Browser database (genome.ucsc.edu/) (Kent et al., 2002), we found that this genomic region has characteristics of transcriptionally active chromatin in E14.5 forebrain tissue, including prominent DNaseI hypersensitivity and histone marks H3K4me1 and H3K27ac (Fig. 2B). Together with previously published studies, these data suggest that this region is an active enhancer for Cux2 in the developing forebrain and led us to characterize it in more detail.
Genomic location and chromatin characteristics of a Cux2 enhancer. (A) Schematic of the murine Cux2 genomic locus, showing the location of an 856 bp cis-regulatory element in the proximal region of intron 2. (B) UCSC Genome Browser data demonstrating key enhancer characteristics for the Cux2 856 bp element, including epigenetic marks H3K4me1 and H3K27ac, a prominent DNaseI hypersensitivity peak, and a high degree of evolutionary conservation across taxa.
Genomic location and chromatin characteristics of a Cux2 enhancer. (A) Schematic of the murine Cux2 genomic locus, showing the location of an 856 bp cis-regulatory element in the proximal region of intron 2. (B) UCSC Genome Browser data demonstrating key enhancer characteristics for the Cux2 856 bp element, including epigenetic marks H3K4me1 and H3K27ac, a prominent DNaseI hypersensitivity peak, and a high degree of evolutionary conservation across taxa.
Interestingly, the human and murine elements both exhibited activity patterns in the developing forebrain similar to that of Cux2, including strong activity in the DTM (Hasenpusch-Theil et al., 2012; Visel et al., 2008). To better characterize the spatial activity of this candidate cis-regulatory element in the developing mouse forebrain, we first cloned the 856 bp murine enhancer into an expression vector (Wilken et al., 2015) with a minimal promoter (TATA box) driving Cre recombinase. We then introduced the plasmid into the developing forebrain of Ai9 Cre-reporter mice at E12.5, using in utero electroporation (Fig. 3A). We co-electroporated a plasmid expressing GFP from the ubiquitously expressed synthetic CAG promoter (Niwa et al., 1991) as a marker of electroporated cells (Fig. 3A). Electroporations were performed to target different regions of the telencephalon, including the DTM, hippocampal primordium and neocortex. We analyzed patterns of GFP expression and Cre-mediated recombination (tdTomato expression) at E14.5. As controls, we compared recombination patterns in the Cux2Enhancer-Cre electroporations with those of the minimal promoter construct alone (MINp-Cre, no enhancer) or with a strong and ubiquitous promoter (CAG-Cre). We found that recombined tdTomato+ cells in the MINp-Cre electroporations were very sparse in the DTM, hippocampus (Fig. 3B) and neocortex (Fig. 3E), consistent with weak expression from the TATA box alone. Conversely, the CAG-Cre construct drove recombination ubiquitously throughout the electroporated regions, including in the DTM, hippocampus and neocortex (Fig. 3C,F). Interestingly, recombination in the Cux2Enhancer-Cre electroporations was almost completely restricted to the DTM, namely the cortical hem and choroid plexus epithelium (Fig. 3D). Very few tdTomato+ cells were present in the hippocampus (Fig. 3D) or the neocortex (Fig. 3G). These data demonstrate that the activity of this Cux2 enhancer is restricted within the telencephalon to the cortical hem and choroid plexus, recapitulating a specific aspect of the complex endogenous Cux2 expression pattern.
The Cux2 enhancer exhibits activity restricted to the cortical hem. (A) Schematic of experimental workflow: E12.5 Ai9 reporter forebrains were electroporated in utero with constructs expressing Cre recombinase driven by either a minimal TATA-box promoter (Minp), a ubiquitous CAG promoter or the Cux2 enhancer. CAG-GFP was co-electroporated to mark electroporated cells. Forebrains were harvested at E14.5 for analysis. (B-D) Coronal sections of electroporated brains showing the dorsal midline region. All electroporated cells express GFP (green) and recombined cells express tdTomato (red). The boundary between the cortical hem and hippocampal primordium is marked by expression of LMX1A protein (yellow). Sections were counterstained for nuclei with DAPI (blue). Electroporation of Minp-Cre causes minimal recombination in the cortical hem and hippocampal primordium (B), whereas CAG-Cre leads to ubiquitous recombination throughout the electroporated regions (C). Cux2Enhancer-Cre drives robust recombination in the cortical hem, but not in the adjacent hippocampal primordium. (E-G) Coronal sections of electroporated brains showing the neocortex. Similar to the dorsal midline, Minp-Cre drives minimal recombination (E) and CAG-Cre drives ubiquitous recombination (F) in the neocortex. The Cux2 enhancer (G) exhibits no activity in the neocortex. Scale bars: 100 µm in B-D; 50 µm in E-G. CH, cortical hem; CP, choroid plexus; HP, hippocampal primordium.
The Cux2 enhancer exhibits activity restricted to the cortical hem. (A) Schematic of experimental workflow: E12.5 Ai9 reporter forebrains were electroporated in utero with constructs expressing Cre recombinase driven by either a minimal TATA-box promoter (Minp), a ubiquitous CAG promoter or the Cux2 enhancer. CAG-GFP was co-electroporated to mark electroporated cells. Forebrains were harvested at E14.5 for analysis. (B-D) Coronal sections of electroporated brains showing the dorsal midline region. All electroporated cells express GFP (green) and recombined cells express tdTomato (red). The boundary between the cortical hem and hippocampal primordium is marked by expression of LMX1A protein (yellow). Sections were counterstained for nuclei with DAPI (blue). Electroporation of Minp-Cre causes minimal recombination in the cortical hem and hippocampal primordium (B), whereas CAG-Cre leads to ubiquitous recombination throughout the electroporated regions (C). Cux2Enhancer-Cre drives robust recombination in the cortical hem, but not in the adjacent hippocampal primordium. (E-G) Coronal sections of electroporated brains showing the neocortex. Similar to the dorsal midline, Minp-Cre drives minimal recombination (E) and CAG-Cre drives ubiquitous recombination (F) in the neocortex. The Cux2 enhancer (G) exhibits no activity in the neocortex. Scale bars: 100 µm in B-D; 50 µm in E-G. CH, cortical hem; CP, choroid plexus; HP, hippocampal primordium.
Developmentally expressed cortical hem transcription factors, Lmx1a and Emx2, activate the Cux2 enhancer in vitro
As pioneering regulators of development, TFs often control gene expression by acting on enhancers. To identify candidate transcriptional regulators of Cux2 expression in the cortical hem, we analyzed the hem-specific enhancer sequence for predicted TF-binding sites using the JASPAR database (Khan et al., 2018) with a ‘predicted’ and ‘consensus’ match threshold of 80% or higher. Our analysis revealed a large list of predicted TF-binding sites, including putative sites for various TFs expressed throughout much of the telencephalon (Table S1). To narrow the list, we focused on a small subset of TFs known to be expressed in the developing cortical hem, including Emx2, Lmx1a and Msx1 (Fig. 4). Consensus binding sequences for these three TFs overlapped with each other in the Cux2 enhancer in 12 predicted sites with high sequence conservation (Fig. 4A,B). Lmx1a and Msx1 are expressed in the DTM at E8.5 and E9.5, respectively, and their expression continues into adulthood (Failli et al., 2002; Furuta et al., 1997). Emx2 is expressed by neural progenitors in the hippocampus and neocortex beginning at E8.5, with the dorsomedial-most expression domain extending into the cortical hem (Simeone et al., 1992a,b; Tole et al., 2000a; Yoshida et al., 1997). Analysis of the Allen Developing Mouse Brain Atlas (2008) in situ hybridization database (developingmouse.brain-map.org) confirmed expression of all three TFs in the cortical hem at E12.5 (Fig. 4C).
The Cux2 enhancer contains multiple predicted binding sites for forebrain-patterning transcription factors expressed in the cortical hem. (A) Schematic of the Cux2 enhancer with putative binding sites for Emx2, Lmx1a and Msx1, predicted from the JASPAR database at >80% threshold. Predicted sites are numbered 1-12. Color-coded sequences show species conservation for human (Homo sapiens), mouse (Mus musculus) and chicken (Gallus gallus): asterisk indicates a single-base mismatch in one species. (B) JASPAR motifs for consensus binding site sequences of the candidate TFs. (C) Sagittal sections from the Allen Developing Mouse Brain Atlas in situ hybridization database showing mRNA expression of candidate TFs in the cortical hem of E11.5 mouse forebrains (Allen Institute for Brain Science. Allen Developing Mouse Brain Atlas: Lmxa1, http://developingmouse.brain-map.org/experiment/show/100058786; Msx1, http://developingmouse.brain-map.org/experiment/show/100092693; Emx2, http://developingmouse.brain-map.org/experiment/show/100047257.) Scale bar: 400 µm. Ctx, neocortex; CH, cortical hem.
The Cux2 enhancer contains multiple predicted binding sites for forebrain-patterning transcription factors expressed in the cortical hem. (A) Schematic of the Cux2 enhancer with putative binding sites for Emx2, Lmx1a and Msx1, predicted from the JASPAR database at >80% threshold. Predicted sites are numbered 1-12. Color-coded sequences show species conservation for human (Homo sapiens), mouse (Mus musculus) and chicken (Gallus gallus): asterisk indicates a single-base mismatch in one species. (B) JASPAR motifs for consensus binding site sequences of the candidate TFs. (C) Sagittal sections from the Allen Developing Mouse Brain Atlas in situ hybridization database showing mRNA expression of candidate TFs in the cortical hem of E11.5 mouse forebrains (Allen Institute for Brain Science. Allen Developing Mouse Brain Atlas: Lmxa1, http://developingmouse.brain-map.org/experiment/show/100058786; Msx1, http://developingmouse.brain-map.org/experiment/show/100092693; Emx2, http://developingmouse.brain-map.org/experiment/show/100047257.) Scale bar: 400 µm. Ctx, neocortex; CH, cortical hem.
As key players in telencephalic patterning, these TFs serve as ideal candidates for regulating Cux2 expression in the cortical hem during early forebrain development. To begin to test whether these TFs can activate the hem-specific Cux2 enhancer, the enhancer element was cloned into a minimal promoter vector driving nuclear mCherry expression (Fig. 5A). cDNAs for the candidate TFs were cloned into the bicistronic expression vector pCIG (Hand et al., 2005), which drives expression of both the TF and GFP from the CAG promoter. Each candidate TF plasmid was co-transfected with the Cux2Enhancer-mCherry plasmid into murine immortalized neuroectodermal (NE-4C) cells (Schlett and Madarász, 1997). Twenty-four h after transfection, mRNA was collected from the cells and analyzed by RT-qPCR for levels of the GFP and mCherry reporter transcripts (Fig. 5A). We also used RT-qPCR to verify that each TF was being overexpressed compared with baseline expression in pCIG control transfected cells (fold change for Emx2, 177,711.74±17,580.78; Lmx1a, 220.77±25.97; and Msx1, 53.96±22.38). Compared with the pCIG empty vector negative control, Emx2 and Lmx1a overexpression significantly upregulated activation of the Cux2 enhancer (Fig. 5B). In contrast, Msx1 did not change Cux2 enhancer-driven mCherry levels compared with control (Fig. 5B).
Lmx1a strongly activates the Cux2 enhancer in vitro. (A) Schematic of experimental workflow. The Cux2Enhancer-mCherry plasmid was transfected into NE-4C cells together with either empty pCIG vector (CAG-IRES-GFP) or pCIG that expresses candidate TFs (CAG-TF-IRES-GFP). The effects of candidate TFs on expression of Cux2Enhancer-mCherry in NE-4C cells was quantified using RT-qPCR of mCherry mRNA. (B,C) RT-qPCR quantification of mCherry transcripts. Data are mean fold change (±s.e.m.) over pCIG vector alone (dotted line), calculated using the ΔΔCt method. (B) Expression from the Cux2 enhancer is activated by expression of Emx2 and Lmx1a, but not Msx1. (C) When putative TF-binding sites were mutated in the Cux2 enhancer, Emx2 still activated the Cux2 enhancer but Lmx1a-mediated activation of the enhancer was greatly attenuated.
Lmx1a strongly activates the Cux2 enhancer in vitro. (A) Schematic of experimental workflow. The Cux2Enhancer-mCherry plasmid was transfected into NE-4C cells together with either empty pCIG vector (CAG-IRES-GFP) or pCIG that expresses candidate TFs (CAG-TF-IRES-GFP). The effects of candidate TFs on expression of Cux2Enhancer-mCherry in NE-4C cells was quantified using RT-qPCR of mCherry mRNA. (B,C) RT-qPCR quantification of mCherry transcripts. Data are mean fold change (±s.e.m.) over pCIG vector alone (dotted line), calculated using the ΔΔCt method. (B) Expression from the Cux2 enhancer is activated by expression of Emx2 and Lmx1a, but not Msx1. (C) When putative TF-binding sites were mutated in the Cux2 enhancer, Emx2 still activated the Cux2 enhancer but Lmx1a-mediated activation of the enhancer was greatly attenuated.
Activation of the Cux2 enhancer in the cortical hem requires Lmx1a-binding sites
Our in silico data predicted multiple binding sites for Emx2 and Lmx1a that overlapped each other (Fig. 4A,B), which correlated well with our in vitro studies demonstrating that these factors can activate the Cux2 enhancer (Fig. 5B). To directly test whether these putative TF-binding sequences were required for enhancer activation, we generated a mutant enhancer construct in which the central eight base pairs of 10 out of the 12 predicted binding sites were mutated (see Materials and Methods for details). The mutated Cux2Enhancer-mCherry plasmid was then tested (Fig. 5A) for activation by Emx2 and Lmx1a. Compared with the wild-type Cux2 enhancer, activation of the mutated enhancer by Lmx1a was greatly attenuated (Fig. 5C). Interestingly, Emx2 was still able to upregulate expression from the mutated enhancer (Fig. 5C), raising the possibility of other more crucial Emx2-binding sites within the enhancer.
We next tested the activity of the mutated enhancer in vivo by in utero electroporation of the mutated enhancer driving Cre recombinase into Ai9 reporter embryos (Fig. 6A). In contrast to the wild-type Cux2Enhancer-Cre that drives recombination specifically and robustly in the cortical hem (Fig. 6B), the TF-binding site mutant Cux2Enhancer-Cre construct was unable to drive any recombination at all in the cortical hem (Fig. 6C). These in vivo data indicate that at least some of the putative binding sites for homeobox TFs are required for activity of the Cux2 enhancer in the cortical hem. Together with our in vitro data, this suggests that Lmx1a and/or Emx2 may be critical regulators of the Cux2 enhancer. Although we did not rule out a possible role for Emx2 in regulating the Cux2 enhancer, we focused on Lmx1a for further functional analysis for two primary reasons: (1) Lmx1a expression is tightly restricted to the cortical hem, in an identical pattern to the Cux2 enhancer; conversely, Emx2 expression extends into the adjacent hippocampal primordium and neocortical neuroepithelium, where the Cux2 enhancer is inactive; and (2) Emx2 over-expression in NE-4C cells could still activate the mutant enhancer, but this mutant enhancer was inactive in the cortical hem.
TFBS-mutated Cux2 enhancer activity abolished in vivo. (A) Schematic of experimental workflow. Ai9 reporter embryos were electroporated in utero at E12.5 with either the wild-type Cux2enhancer-Cre plasmid or the mutated version that is no longer activated by Lmx1a. CAG-GFP was co-electroporated as a marker of the electroporated cells. Forebrains were harvested at E14.5 for analysis of recombination. (B,C) Coronal sections of electroporated brains showing the dorsal midline region, including cortical hem. Boundary between the cortical hem and hippocampal primordium is marked by expression of LMX1A protein (yellow). Sections were counterstained for nuclei with DAPI (blue). Activation of the wild-type Cux2 enhancer in electroporated GFP+ cells led to robust recombination in only the cortical hem (B), whereas the TFBS-mutated enhancer lost all activity in the cortical hem (C). Scale bars: 100 µm. CH, cortical hem.
TFBS-mutated Cux2 enhancer activity abolished in vivo. (A) Schematic of experimental workflow. Ai9 reporter embryos were electroporated in utero at E12.5 with either the wild-type Cux2enhancer-Cre plasmid or the mutated version that is no longer activated by Lmx1a. CAG-GFP was co-electroporated as a marker of the electroporated cells. Forebrains were harvested at E14.5 for analysis of recombination. (B,C) Coronal sections of electroporated brains showing the dorsal midline region, including cortical hem. Boundary between the cortical hem and hippocampal primordium is marked by expression of LMX1A protein (yellow). Sections were counterstained for nuclei with DAPI (blue). Activation of the wild-type Cux2 enhancer in electroporated GFP+ cells led to robust recombination in only the cortical hem (B), whereas the TFBS-mutated enhancer lost all activity in the cortical hem (C). Scale bars: 100 µm. CH, cortical hem.
Lmx1a occupies cortical hem-specific Cux2 enhancer in the developing telencephalon
Our in vitro and in vivo interrogation of the Cux2 enhancer revealed Lmx1a to be a possible regulator of enhancer activity. To determine whether Lmx1a occupies the endogenous Cux2 enhancer locus in the early forebrain, we harvested whole forebrains from E12.5 mice and performed chromatin immunoprecipitation (ChIP) using three different Lmx1a antibodies (Fig. 7A). Following ChIP, we used qPCR to probe for Cux2 enhancer enrichment compared with non-specific IgG control ChIP (Fig. 7B) and with a different nearby enhancer region lacking Lmx1a-binding sites (Fig. 7C). Compared with the negative control IgG, the hem-specific Cux2 enhancer was significantly enriched by ChIP using two out of the three Lmx1a antibodies (Fig. 7B, Aviva and ProSci), and trended toward enrichment in the third (Millipore). None of the three Lmx1a antibodies significantly pulled down a nearby enhancer region that lacked Lmx1a-binding sequences (Fig. 7C), confirming specificity for the Cux2 cortical hem enhancer region. These data indicate that Lmx1a occupies the Cux2 enhancer in vivo, where it may act to initiate Cux2 expression specifically in the cortical hem.
ChIP-qPCR of E12.5 forebrains confirms Lmx1a occupancy of the endogenous Cux2 enhancer locus. (A) Schematic of experimental workflow. E12.5 C57BL/6J forebrains were harvested for chromatin immunoprecipitation (ChIP)-qPCR validation of Lmx1a enrichment at the endogenous Cux2 enhancer genomic locus. (B) qPCR of Cux2 enhancer enrichment using three different anti-Lmx1a antibodies, compared with the anti-rabbit IgG negative control. Data are mean percentage input (±s.e.m.). (C) Lmx1a ChIP-qPCR quantification shows no significant enrichment of a nearby enhancer region that lacks Lmx1a binding sites. Data are mean percentage input (±s.e.m.)
ChIP-qPCR of E12.5 forebrains confirms Lmx1a occupancy of the endogenous Cux2 enhancer locus. (A) Schematic of experimental workflow. E12.5 C57BL/6J forebrains were harvested for chromatin immunoprecipitation (ChIP)-qPCR validation of Lmx1a enrichment at the endogenous Cux2 enhancer genomic locus. (B) qPCR of Cux2 enhancer enrichment using three different anti-Lmx1a antibodies, compared with the anti-rabbit IgG negative control. Data are mean percentage input (±s.e.m.). (C) Lmx1a ChIP-qPCR quantification shows no significant enrichment of a nearby enhancer region that lacks Lmx1a binding sites. Data are mean percentage input (±s.e.m.)
Lmx1a gain of function extends spatial competence for Cux2 enhancer activation
In the developing telencephalon, Lmx1a is expressed strongly throughout the DTM, where it functions to promote cortical hem fate and suppress hippocampal and neocortical fate (Caronia-Brown et al., 2014; Chizhikov et al., 2010; Failli et al., 2002). The previously described sharp border of Lmx1a expression between the cortical hem and hippocampus is very similar to that of Cux2, which is expressed strongly throughout the cortical hem (Franco et al., 2012; Saulnier et al., 2013; Zimmer et al., 2004) but only weakly in a limited number of progenitors and neurons in the hippocampus and neocortex (Franco et al., 2012; Zimmer et al., 2004) (Fig. 1C,D). As Lmx1a is not expressed at all in the developing hippocampus where Cux2 expression is initially weak, this provided us with an opportunity to assess whether mis-expression of Lmx1a is sufficient to ectopically activate the Cux2 enhancer. To test this possibility, we co-electroporated our CAG-Lmx1a-IRES-GFP and Cux2Enhancer-Cre constructs into the medial cortex of Ai9fl/+ Cre-reporter embryos in utero at E12.5. We allowed the embryos to continue developing until E14.5 and analyzed the percentage of electroporated cells (GFP+) that activated the Cux2 cortical hem enhancer (tdTomato+) (Fig. 8A). Lmx1a overexpression in the hippocampus was verified by IHC (Fig. 8C, arrowheads). Compared with the empty vector control (Fig. 8B,B′), Lmx1a mis-expression (Fig. 8C,C′) resulted in a significant increase in ectopic enhancer activation in the hippocampal primordium adjacent to the cortical hem (Fig. 8D). These data indicate that Lmx1a is sufficient to activate the Cux2 enhancer in vivo.
Ectopic activation of the Cux2 cortical hem enhancer by Lmx1a overexpression. (A) Schematic of experimental workflow. Ai9 embryos were electroporated in utero at E12.5 with a constitutive Lmx1a expression plasmid along with Cux2Enhancer-Cre, targeting the DTM. Forebrains were harvested at E14.5 for quantification of the percentage of electroporated cells that were recombined. (B-C′) Coronal sections of electroporated brains showing the dorsal midline region. (B,C) Boundary between the cortical hem (CH) and hippocampal primordium (HP) is marked by expression of LMX1A protein (yellow) and a dotted line. Lmx1a overexpression was confirmed by immunohistochemistry (arrows in C). Sections were counterstained for nuclei using DAPI (blue). (B′,C′) Higher magnification views of the boxed areas in B,C, respectively, showing tdTomato expression in electroporated GFP+ cells; arrows indicate double-positive cells. Compared with controls (B,B′), Lmx1a gain of function in the hippocampal primordium resulted in a significant increase in the number of recombined cells outside the cortical hem. (D) Quantification showing the mean percentage (±s.e.m.) of electroporated cells (GFP+) that are recombined (tdTomato+) outside the Lmx1a+ cortical hem in control versus Lmx1a electroporations. Scale bars: 100 µm in B,C; 50 µm in B′,C′.
Ectopic activation of the Cux2 cortical hem enhancer by Lmx1a overexpression. (A) Schematic of experimental workflow. Ai9 embryos were electroporated in utero at E12.5 with a constitutive Lmx1a expression plasmid along with Cux2Enhancer-Cre, targeting the DTM. Forebrains were harvested at E14.5 for quantification of the percentage of electroporated cells that were recombined. (B-C′) Coronal sections of electroporated brains showing the dorsal midline region. (B,C) Boundary between the cortical hem (CH) and hippocampal primordium (HP) is marked by expression of LMX1A protein (yellow) and a dotted line. Lmx1a overexpression was confirmed by immunohistochemistry (arrows in C). Sections were counterstained for nuclei using DAPI (blue). (B′,C′) Higher magnification views of the boxed areas in B,C, respectively, showing tdTomato expression in electroporated GFP+ cells; arrows indicate double-positive cells. Compared with controls (B,B′), Lmx1a gain of function in the hippocampal primordium resulted in a significant increase in the number of recombined cells outside the cortical hem. (D) Quantification showing the mean percentage (±s.e.m.) of electroporated cells (GFP+) that are recombined (tdTomato+) outside the Lmx1a+ cortical hem in control versus Lmx1a electroporations. Scale bars: 100 µm in B,C; 50 µm in B′,C′.
Lmx1a knockdown abolishes activation of the Cux2 enhancer in the cortical hem
The results of our Lmx1a gain-of-function experiments as well as our in vitro results with the mutated Cux2 enhancer point to a sufficiency and necessity, respectively, for Lmx1a to activate this enhancer in a spatially restricted manner. To further determine whether Lmx1a is necessary to activate the Cux2 enhancer, we designed a loss-of-function strategy to knockdown Lmx1a expression in the cortical hem in vivo by in utero electroporation (Fig. 9A). We co-electroporated constructs expressing either a non-targeting control shRNA or Lmx1a-targeting shRNA along with Cux2Enhancer-mCherry and CAG-IRES-GFP into the DTM of E12.5 embryos. We allowed the embryos to continue developing until E14.5 and quantified the number of mCherry+ cells among the electroporated (GFP+) population (Fig. 9A). Lmx1a knockdown resulted in a ∼50% decrease in Lmx1a protein signal in electroporated cells (Fig. 9B,C). Compared with non-targeting controls, Lmx1a knockdown resulted in a drastic decrease in the number of electroporated cells that activated the Cux2 enhancer (mCherry+GFP+/GFP+) (Fig. 9D-F). Furthermore, in the remaining knockdown cells that did activate the Cux2 enhancer, there was a significant decrease in the levels of mCherry reporter expression compared with non-targeted control cells (Fig. 9D,E,G). Together, these data indicate that Lmx1a is necessary for Cux2 enhancer activation in the cortical hem, thereby providing further support for the role of Lmx1a as a critical regulator of Cux2 expression.
Endogenous knockdown of Lmx1a greatly attenuates Cux2 enhancer activation. (A) Schematic of experimental workflow. Knockdown of endogenous Lmx1a mRNA was induced by in utero electroporation of E12.5 C57BL/6J embryos with constructs expressing either targeting or non-targeting shRNAs for Lmx1a together with Cux2Enhancer-mCherry and CAG-GFP. Forebrains were harvested at E14.5 to quantify the percentage of electroporated (GFP+) that expressed mCherry. (B,C) Lmx1a knockdown resulted in a decrease in relative Lmx1a average signal intensity in electroporated GFP+ cells. Arrows in B indicate electroporated Lmx1a knockdown cells. (D,E) Coronal sections of electroporated brains showing the cortical hem (CH). The boundary between the cortical hem and hippocampal primordium is marked by expression of Lmx1A protein (yellow) and a dotted line. Sections were counterstained for nuclei with DAPI (blue). (F) Quantification showing mean percentage (±s.e.m.) of electroporated cells that expressed mCherry in control versus Lmx1a knockdown cells in the cortical hem. (G) Relative mCherry expression quantified only in electroporated cells that expressed mCherry above background levels. Compared with control, Lmx1a knockdown resulted in a significant decrease in the number of mCherry+/GFP+ cells in the cortical hem, as well as a decrease in mCherry intensity in mCherry+/GFP+ cells. Scale bars: 25 µm in B; 100 µm in D,E.
Endogenous knockdown of Lmx1a greatly attenuates Cux2 enhancer activation. (A) Schematic of experimental workflow. Knockdown of endogenous Lmx1a mRNA was induced by in utero electroporation of E12.5 C57BL/6J embryos with constructs expressing either targeting or non-targeting shRNAs for Lmx1a together with Cux2Enhancer-mCherry and CAG-GFP. Forebrains were harvested at E14.5 to quantify the percentage of electroporated (GFP+) that expressed mCherry. (B,C) Lmx1a knockdown resulted in a decrease in relative Lmx1a average signal intensity in electroporated GFP+ cells. Arrows in B indicate electroporated Lmx1a knockdown cells. (D,E) Coronal sections of electroporated brains showing the cortical hem (CH). The boundary between the cortical hem and hippocampal primordium is marked by expression of Lmx1A protein (yellow) and a dotted line. Sections were counterstained for nuclei with DAPI (blue). (F) Quantification showing mean percentage (±s.e.m.) of electroporated cells that expressed mCherry in control versus Lmx1a knockdown cells in the cortical hem. (G) Relative mCherry expression quantified only in electroporated cells that expressed mCherry above background levels. Compared with control, Lmx1a knockdown resulted in a significant decrease in the number of mCherry+/GFP+ cells in the cortical hem, as well as a decrease in mCherry intensity in mCherry+/GFP+ cells. Scale bars: 25 µm in B; 100 µm in D,E.
Multiple Lmx1a-binding sites are a shared feature among cis-regulatory elements active in the cortical hem
Previous work has revealed a number of forebrain enhancers, including some that appear to have restricted cortical hem activity similar to the murine Cux2 enhancer (Pattabiraman et al., 2014; Visel et al., 2008). Images of transgenic animals from the VISTA Enhancer Browser (Visel et al., 2007) show that human enhancer elements hs411, hs611 and hs643 can drive cortical hem-specific lacZ expression in transgenic mouse embryos (Fig. 10A-C). We reasoned that all three enhancers might share features that regulate their activity through a common mechanism, given their very similar spatial and temporal transcriptional activity. To uncover common features between hs611, hs411 and hs643, the sequences of all three genomic regions, together with the murine Cux2 hem enhancer, were analyzed using Analysis of Motif Enrichment (McLeay and Bailey, 2010) through the MEME Suite web portal (meme-suite.org/tools/ame). Compared with 1004 shuffled control sequences, the most enriched motif shared by all four elements was a TTAATTAA motif (P=1.48e-6 by Fisher's exact test). We then ran this top-enriched motif through the Tomtom tool (Gupta et al., 2007) in MEME Suite to compare it with databases of known motifs. The sequence was identified as an Lmx1a consensus binding motif by JASPAR, Jolma and Uniprobe databases (Fig. 10D), indicating that the Lmx1a consensus binding motif is the top enriched motif found in common between all four enhancer sequences.
Lmx1a-binding sites are a common feature of enhancers active in the developing cortical hem. (A-C) Examples of human enhancer elements driving lacZ expression in transgenic mouse embryos. Whole-mount staining images from the Vista Enhancer Browser show activity of hs611 (A), hs411 (B) and hs643 (C) in the cortical hem region. (D) Analysis of Motif Enrichment followed by Tomtom comparison to known motifs identified Lmx1a consensus binding sites as the most significantly enriched motif in the four hem-expressed enhancers. (E) Using the JASPAR database, all cortical hem enhancers were predicted to contain seven or more high-threshold Lmx1a-binding sites. (F,F′) Coronal sections of brains electroporated with hs643-mCherry and CAG-GFP, showing the spatially restricted activity of enhancer Hs643 within the Lmx1a+(yellow) cortical hem. (G,H) Cortical hems electroporated with either control (G) or Lmx1a-targeting shRNA (H) and hs643-mCherry and CAG-GFP. (I,J) Control and Lmx1a-knockdown cortical hems were quantified for the percentage of electroporated GFP+ cells that expressed mCherry (I) and for relative mCherry levels of mCherry+/GFP+ cells (J). Compared with control, Lmx1a knockdown resulted in a significant decrease in the number of mCherry+ cells in the cortical hem, as well as a decrease in the relative expression of mCherry. Boundary between the cortical hem and hippocampal primordium is marked by expression of Lmx1A protein (yellow) and a dotted line. Sections were counterstained for nuclei with DAPI (blue). Scale bars: 100 µm in F,F′; 50 µm in G,H.
Lmx1a-binding sites are a common feature of enhancers active in the developing cortical hem. (A-C) Examples of human enhancer elements driving lacZ expression in transgenic mouse embryos. Whole-mount staining images from the Vista Enhancer Browser show activity of hs611 (A), hs411 (B) and hs643 (C) in the cortical hem region. (D) Analysis of Motif Enrichment followed by Tomtom comparison to known motifs identified Lmx1a consensus binding sites as the most significantly enriched motif in the four hem-expressed enhancers. (E) Using the JASPAR database, all cortical hem enhancers were predicted to contain seven or more high-threshold Lmx1a-binding sites. (F,F′) Coronal sections of brains electroporated with hs643-mCherry and CAG-GFP, showing the spatially restricted activity of enhancer Hs643 within the Lmx1a+(yellow) cortical hem. (G,H) Cortical hems electroporated with either control (G) or Lmx1a-targeting shRNA (H) and hs643-mCherry and CAG-GFP. (I,J) Control and Lmx1a-knockdown cortical hems were quantified for the percentage of electroporated GFP+ cells that expressed mCherry (I) and for relative mCherry levels of mCherry+/GFP+ cells (J). Compared with control, Lmx1a knockdown resulted in a significant decrease in the number of mCherry+ cells in the cortical hem, as well as a decrease in the relative expression of mCherry. Boundary between the cortical hem and hippocampal primordium is marked by expression of Lmx1A protein (yellow) and a dotted line. Sections were counterstained for nuclei with DAPI (blue). Scale bars: 100 µm in F,F′; 50 µm in G,H.
We next used the JASPAR database to search all three human enhancer elements for putative Lmx1a-binding sites (>85% threshold). Similar to the murine Cux2 cortical hem enhancer, hs611, hs411 and hs643 were all predicated to contain multiple high-threshold Lmx1a-binding sites (Fig. 10E). To determine whether Lmx1a was required for activation of any of these hem enhancers, we generated an hs643-mCherry reporter construct and electroporated it into the DTM at E12.5. Analysis at E14.5 confirmed that hs643 activity was restricted to the cortical hem (Fig. 10F,F′). Next, we knocked down Lmx1a expression, as in Fig. 9A, and tested the effects of knockdown on hs643 enhancer activity compared with non-targeting control shRNA (Fig. 10G,H). Similar to our results with the Cux2 enhancer, we found that Lmx1a knockdown decreased both the percentage of electroporated cells that expressed the mCherry reporter (Fig. 10I) and the levels with which the positive cells expressed the reporter (Fig. 10J). These data support a role for Lmx1a as a genuine regulator of hs643 in the cortical hem. As a common feature among cortical hem enhancers, the presence of multiple Lmx1a-binding sites may indicate that Lmx1a sits near the top of the GRN that is active in the developing cortical hem, perhaps as a pioneering TF.
DISCUSSION
Using a combination of in silico, in vitro and in vivo approaches, we characterized a Cux2 gene regulatory element with the goal of uncovering key TFs involved in the transcriptional regulation of Cux2 in neural progenitors. We showed that this Cux2 enhancer recapitulates a specific aspect of the complex Cux2 expression pattern in the developing forebrain, namely strong and precise activity in the cortical hem. Further analysis uncovered the LIM homeobox transcription factor Lmx1a as a positive regulator of the Cux2 cortical hem enhancer. Comparison of three cortical hem-specific human enhancer elements revealed recurring Lmx1a-binding sites were the most shared motif, raising the possibility that Lmx1a is master transcriptional regulator of the GRN that controls cortical hem identity.
Cux2 expression as a tool to uncover regulators of cell-fate decisions
We previously identified a subset of neural progenitors in the developing forebrain that are fate restricted to produce only corticocortical projection neurons in the neocortex (Franco and Müller, 2013; Franco et al., 2012; Gil-Sanz et al., 2015). These progenitors can be identified by expression of the transcription factor Cux2, and lineage-traced using Cux2-Cre and Cux2-CreERT2 knock-in mice. Although Cux2 itself does not appear to control cell fate decisions in the forebrain (Cubelos et al., 2008, 2010), its restricted expression in defined subsets of neural progenitors may be a useful tool for uncovering transcriptional regulators of cell fate during forebrain patterning. The Cux2 expression pattern in the developing forebrain is complex and dynamic (Franco et al., 2012; Gil-Sanz et al., 2015; Zimmer et al., 2004), suggesting that control of the Cux2 locus may involve multiple transcriptional regulatory mechanisms. Using our Cux2-Cre mice crossed to a Cre-reporter line, we identified the DTM as one of the earliest sites of Cux2 expression in the developing forebrain. In contrast to the salt-and-pepper pattern of Cux2 expression in the adjacent hippocampus and neocortex, we find that essentially all neural progenitors in the cortical hem belong to the Cux2 lineage. This raised the interesting possibility that Cux2 expression in different parts of the developing forebrain are controlled by distinct mechanisms. This may also be the case for other transcription factors in the forebrain. For example, an identified enhancer region for Emx2 recapitulates most aspects of Emx2 expression in the forebrain, except for expression in the cortical hem, which is likely controlled by a different unidentified enhancer (Suda et al., 2010). Identification of the various cis-regulatory elements that drive differential Cux2 expression, and the transcription factors that regulate these elements, may therefore lead to a better understanding of the GRNs that control tissue patterning and subtype fate specification in the developing forebrain.
Identification and characterization of a cortical hem-specific Cux2 enhancer
Previous studies have identified genomic regions within intron 2 of the human and murine Cux2 genes that can recapitulate Cux2 expression in the DTM of transgenic mice (Hasenpusch-Theil et al., 2012; Visel et al., 2008). We found that the murine Cux2 element exhibits features of an active enhancer in the E14.5 forebrain, including DNaseI hypersensitivity and histone modifications associated with transcriptionally active chromatin. This region also displays high levels of conservation from human to chicken, pointing towards an important functional role. Using in utero electroporation to test the in vivo activity of this region, we show that it drives expression specifically and robustly in the cortical hem neuroepithelium, but not in progenitors in the adjacent hippocampus or neocortex. These data indicate that this element is a developmentally active enhancer specific for the cortical hem. It will be interesting in future studies to determine which exact features of this enhancer are required for Cux2 expression in the cortical hem, and whether this regulatory element is active in other Cux2 expression domains that share features with the cortical hem, such as the rhombic lip in the hindbrain (Capaldo and Iulianella, 2016).
Lmx1a is a crucial regulator of the Cux2 hem-specific enhancer
Using a bioinformatics approach, we identified several putative TF-binding sites in the Cux2 enhancer. As known regulators of cortical hem development (Chizhikov et al., 2010; Tole et al., 2000b), Emx2, Lmx1a and Msx1 are a promising group of candidate TFs with the potential to regulate the Cux2 enhancer in the cortical hem. Although we chose to focus here on these three candidates, this short list likely does not represent the full complement of TFs that can regulate Cux2 expression via this enhancer. Indeed, our analysis revealed a much longer list of putative TF-binding sites that we did not test further. For example, our analysis confirmed two putative Tcf-Lef1-binding sites that were identified in a previous study on the Cux2 enhancer (Hasenpusch-Theil et al., 2012). Interestingly, that study showed that, although Cux2 enhancer-lacZ transgenic animals normally expressed the lacZ reporter specifically in the cortical hem, when the Tcf-Lef1-binding sites were mutated the reporter expression was much stronger and expanded beyond the cortical hem. These data raise the intriguing possibility that Wnt signaling through Tcf-Lef1 may actually repress activity of the Cux2 enhancer outside the cortical hem. It will be interesting in future to test whether this pathway or potentially other repressors may function in combination with Lmx1a to restrict the Cux2 enhancer activity to the cortical hem, which would have important implications for understanding regional patterning in the developing forebrain.
With respect to the TFs we did test, we did not identify any effect of Msx1 on activity of the Cux2 enhancer in vitro, indicating it may not be a direct regulator of Cux2 cortical hem expression. On the other hand, both Emx2 and Lmx1a activated transcription from the Cux2 enhancer. The majority of the predicted Emx2- and Lmx1a-binding sites overlapped each other, reflective of the similarity of their consensus binding motifs. Interestingly, mutation of these putative binding sites drastically reduced responsiveness of the enhancer to Lmx1a, but not to Emx2. This may suggest substantial redundancy in the number of Emx2-binding sites in the Cux2 enhancer, or that there are Emx2-binding sites remaining that were not mutated and are more critical. Importantly, the mutated enhancer showed no activity in the cortical hem or adjacent hippocampus, suggesting that activation by Lmx1a is more relevant in vivo than activation by Emx2. This would be in line with our further experiments showing that the Cux2 enhancer is not active in the hippocampus or neocortex, where Emx2 is strongly expressed but Lmx1a is absent. However, we do not rule out the possibility that Emx2 may still play an important role in activating the Cux2 hem enhancer, either directly or indirectly. Future studies using Emx2 loss and gain of function in vivo should shed light on the role that Emx2 may play in regulating Cux2 expression in the cortical hem, or possibly in the hippocampus and neocortex.
In further support of Lmx1a as an activator of Cux2 expression, we showed that Lmx1a protein was significantly enriched at the Cux2 enhancer genomic locus in isolated E12.5 forebrains compared with a second putative Cux2 enhancer found within the same intron but that is devoid of Lmx1a-binding sites. As expression of both Lmx1a and Cux2 commences much earlier than E12.5 in the DTM (Failli et al., 2002; Mangale et al., 2008), we would predict that Lmx1a regulates Cux2 expression in the cortical hem during the earliest stages of forebrain development. We also found that modulating Lmx1a levels in the DTM affects Cux2 enhancer activity. Whereas Lmx1a knockdown in vivo resulted in a substantial reduction in the ability of cortical hem cells to activate the Cux2 enhancer plasmid, overexpression in the hippocampus resulted in ectoptic activation of the Cux2 enhancer outside the cortical hem. Taken together with our mutated Lmx1a-binding site experiments, these results demonstrate that Lmx1a is sufficient and necessary for Cux2 enhancer activation in the DTM. Interestingly, however, Lmx1a mis-expression in the neocortex did not result in activation of the Cux2 enhancer construct (data not shown). We also noticed that the further away the hippocampal electroporated cells were from the cortical hem, the less likely they were to express the reporter plasmid (data not shown). This result may suggest the presence of additional factors that regulate Cux2 expression. For example, Lmx1a may require a transcriptional co-activator for maximum activity that is missing from the neocortex. Alternatively or additionally, there may be unidentified transcriptional repressors of the Cux2 enhancer that are expressed specifically in the hippocampus and neocortex. Further studies will be required to fully elucidate the mechanisms that control the complex expression pattern of Cux2 in the different forebrain regions.
Lmx1a as a common activator of cortical hem GRNs
When we compared the murine and human Cux2 enhancers with two other conserved human elements that show activation in the cortical hem, we found that the most enriched motif in all four enhancers corresponds to the Lmx1a consensus binding site. Together with the fact that Lmx1a is one of the earliest markers of the DTM (Failli et al., 2002; Mangale et al., 2008), these data suggest that Lmx1a may sit near the top of the GRN involved in regulating cortical hem cell fate. In support of this idea, we demonstrated that endogenous Lmx1a knockdown greatly attenuated activation of one of these hem-specific enhancers, hs643. Additionally, previous work has shown that cortical hem identity is lost in dreher mutant mice in which Lmx1a is inactivated by a missense mutation (Chizhikov et al., 2010). The sharp border of Lmx1a expression between the cortical hem and adjacent hippocampal primordium further make it an ideal candidate for establishing precise patterns of gene expression during early patterning of the developing forebrain.
An important unanswered question is what lies upstream of Lmx1a during these early patterning stages? Previous work has reported that Lmx1a expression can be activated by BMP4 in the developing forebrain (Srinivasan et al., 2014; Watanabe et al., 2016). As a morphogenetic pathway that is specifically expressed within the cortical hem and choroid plexus, BMP signaling could potentially initiate the Lmx1a-dependent GRN that leads to specific DTM fates. Interestingly, previous studies have reported that upregulation of BMP signaling both in the developing chick olfactory epithelium (Wittmann et al., 2014) and murine mandibular neural crest cells (Bonilla-Claudio et al., 2012) results in significant upregulation of Cux2 expression. Additionally, Cux2 expression appears coincident with BMP4 within the mesenchyme of the developing mouse limb bud (Iulianella et al., 2003). How BMPs activate Cux2 expression in these contexts has not been determined, but it would be interesting to test whether BMP signaling can drive Cux2 expression in multiple tissues through Lmx1a-mediated activation of the conserved enhancer.
Conclusions
In this study, we identify a conserved enhancer and its transcriptional activator, Lmx1a, as an important mechanism for driving restricted expression of Cux2 in the developing forebrain. We further show that recurrent Lmx1a-binding sites are a common motif shared in multiple enhancers with similarly restricted activities. These studies provide a template for future studies aimed at identifying other Cux2 cis-regulatory elements that control its complex expression during forebrain development, and ultimately the upstream GRNs that specify different cell fates among the forebrain progenitor pool.
MATERIALS AND METHODS
Animals
Mice used for experiments were housed and handled in accordance with protocols approved by the UC Anschutz Medical Campus IACUC committee. The following mouse lines were used in this study: Cux2-Cre and Cux2-CreERT2 (Franco et al., 2012, 2011; Gil-Sanz et al., 2015), Ai9 (Madisen et al., 2010) and C57BL/6J (JAX Stock #000664). Mice of both sexes were analyzed in all experiments. Embryos were produced from timed-pregnant females, with noon on the day of the plug being designated E0.5. Tamoxifen induction of the Cux2-CreERT2 line was performed by intraperitoneal injection of 2 mg tamoxifen (Sigma) dissolved in sunflower oil (Sigma) into pregnant mothers at the indicated age. Progesterone (Sigma) was co-administered at half the concentration of tamoxifen to prevent late abortions caused by the mixed-estrogen effects of tamoxifen. For postnatal analysis of tamoxifen-induced animals, pups were delivered by cesarean section at E19.5 and provided with a foster mother until analysis.
Plasmids and in utero electroporation
The murine Cux2 enhancer was cloned from the endogenous genomic locus (NCBI37/mm9 chr5:122,482,512-122,483,367) using a Gblock (IDT) with 5′ and 3′ arms homologous to the multiple cloning site in the backbone vector. The Gblock was cloned by Gibson assembly into the pMinp vector (Wilken et al., 2015), immediately upstream of the TATA box. mCherry or Cre recombinase with a nuclear localization signal (Lewandoski and Martin, 1997) were cloned immediately downstream of the TATA box. Hs643-mCherry was cloned in a similar manner using a Gblock corresponding to human genomic locus GRCh38/hg38 chr9:23,004,731-23,005,790. To generate the TF-binding site mutant version of the Cux2 enhancer, we synthesized a Gblock in which the central 8 bp of putative binding sites identified (excluding sites #8 and 10, Fig. 4A) were mutated to 5′AAGCGCAA3′. Transcription factor cDNAs were either obtained from Addgene (Lmx1a, 45070; Msx1, 34998) or from IDT as Gene blocks (Emx2) and cloned into the SacI and XmaI sites of the pCIG vector (Hand et al., 2005), between the CAG promoter and the IRES-GFP cassette. In utero electroporation of plasmids (0.5-1 mg/ml) was carried out as previously described (Franco et al., 2012; Gil-Sanz et al., 2013) on E12.5 embryos of timed-pregnant mice. Embryos were harvested for analysis at E14.5.
Immunohistochemistry
Brains from E9.5-14.5 embryos were dissected and fixed for 2 h at room temperature in 4% paraformaldehyde. Forebrains were sectioned on a vibrating microtome (Leica VT1200S) at 100 µm increments, or on a cryostat (Leica CM1520) at 15-30 µm increments. Immunohistochemistry was performed on tissue sections as described previously (Winkler et al., 2018) using the following antibodies: rabbit anti-Lmx1a (1:1000, Millipore, RRID:AB_10805970) and rabbit anti-RFP (1:500, LifeSpan Biosciences, RRID:AB_945213). Donkey secondary antibodies conjugated to Alexa Fluor 488, Rhodamine Red-X or Alexa Fluor 647 were purchased from Jackson ImmunoResearch and used at 1:500. Sections were imaged using a Zeiss LSM 780 confocal microscope.
Cell culture and qRT-PCR
Experiments were performed using the immortalized mouse neuroectodermal NE-4C cells (ATCC, CRL-2925) grown in Dulbecco's minimal essential media (MEM; Corning 10-010-CV) with 4 mM L-glutamine (Invitrogen), 10% fetal bovine serum (FBS) (Invitrogen) and penicillin (0.0637 g/l)-streptomycin (0.1 g/l). We did not re-authenticate or test for contamination after receiving the cells from ATCC. Cells were plated on 12-well plates and grown to ∼70% confluency prior to transfection. Cells were transfected with either CAG-Emx2, Lmx1a or Msx1-IRES-GFP, together with Cux2Enhancer-mCherry or the TF-binding site mutated Cux2Enhancer-mCherry for 4-6 h with Lipofectamine 3000 (Invitrogen), with subsequent media change. At 24 h after transfection, RNA was isolated from cells with an RNeasy Plus Kit (Qiagen) and reverse transcribed into cDNA using an iScript RT Kit (Bio-Rad). Expression of mCherry, GFP and the housekeeping gene cyclophilin A was assessed by qRT-PCR (Bio-Rad CFX Connect R-T System). Fold change was calculated using the delta-CT method for both GFP and mCherry, relative to cyclophilin A. Fold changes of mCherry mRNA were normalized to those of GFP, to account for variations in transfection efficiency. The following primers were used: cyclophilin A forward, GAGCTGTTTGCAGACAAAGTTC; cyclophilin A reverse, CCCTGGCACATGAATCCTGG; eGFP forward, ACGTAAACGGCCACAAGTTC; eGFP reverse. AAGTCGTGCTGCTTCATGTG; mCherry forward, GATAACATGGCCATCATCAAGGA; mCherry reverse, CGTGGCCGTTCACGGAG.
Chromatin immunoprecipitation and qPCR
Chromatin-immunoprecipitation (ChIP) experiments for Lmx1a were performed using the SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling). Eight to ten E12.5 C57BL/6J forebrains per experiment (n=3 biological replicates) were harvested and then fixed with 1% EM-grade paraformaldehyde for 10 min (Electron Microscopy Sciences). The chromatin was sheared to ∼500 bp fragments by sonication (Diagenode Bioruptor ultrasonicator) at 10×30 s cycles ON/OFF at 100% power. ChIP was performed at 4°C overnight on 10 µg of DNA with 2 µg of three different rabbit antibodies against Lmx1a (Aviva, RRID:AB_387461; Millipore, RRID:AB_10805970; ProSci, RRID:AB_2316106), anti-Histone H3 as a positive control for enhancer pulldown and amplification (Cell Signaling), and anti-rabbit IgG as a negative control (Cell Signaling). After DNA purification, RT-qPCR was performed for the cortical hem Cux2 enhancer (Chr5:122,482,512-122,483,367) (forward, CCACACTGTCTGGGGACAGAAAGA; reverse, GCAGCCCTGGAAGCATGTAATTTG) as well as for a negative control region using SsoAdvanced Sybr Green (Bio-Rad). The negative control locus used was a putative enhancer located distally to the cortical hem-specific enhancer in Cux2 intron 2 (Chr5:122,375,401-122,375,550) (forward, CCCCTCTTGAAGCTTCCT; reverse, GGTGATAGATGCAGGAGG).
Quantitative analysis of Lmx1a gain of function
CAG-Lmx1a-GFP or empty vector control along with the Cux2Enhancer-Cre constructs were electroporated into forebrains of E12.5 Ai9fl/+ reporter embryos followed by quantification of tdTomato expression at E14.5. The electroporated and quantified region included the area spanning ∼150 µm directly above the cortical hem (demarcated by Lmx1a immunohistochemistry), corresponding to the hippocampal primordium with normally no Cux2 enhancer activity. Harvested brains were sectioned and stained for Lmx1a, and imaged using a confocal microscope. The percentage of GFP+ cells outside the hem that were also expressing tdTomato was quantified. Single-plane confocal images were used for quantification.
Quantitative analysis of Lmx1a loss-of-function
To knock down endogenous Lmx1a expression and assess Cux2 enhancer activity, an Lmx1a-targeting shRNA construct (Sigma, SHCLNG-NM_033652, TRCN0000433282) was co-electroporated with the Cux2Enhancer-H2B-mCherry (or Enhancer hs643-H2B-mCherry) and pCIG constructs into the DTM of E12.5 C57BL/6J embryos. For control experiments, a pLKO.1 non-mammalian shRNA construct (Sigma, SHC002) was electroporated. Knockdown analysis was performed at E14.5. Specifically, harvested forebrains were sectioned, stained for Lmx1a (Millipore, RRID:AB_10805970) and imaged on a confocal microscope at the exact same acquisition settings for every image. Knockdown of Lmx1a was confirmed by measuring relative Lmx1a signal intensity in GFP+ cells between knockdown and control cells (Zeiss Zen Blue 2.3, average pixel intensity). To quantify the effects of Lmx1a knockdown on enhancer activity, the percentage of GFP+ cells in the cortical hem that were also expressing mCherry above background levels was quantified, as well as the relative intensity of mCherry expression only in the mCherry+ cells, compared with control. Single-plane confocal images were used for quantification.
Statistics
All comparisons made were between two groups using a two-tailed, two-sample equal or unequal variance Student's t-tests to analyze all data. Equality of variance was determined using a Bartlett's test. The standard error of the mean (s.e.m.) is reported on all graphs.
Acknowledgements
We thank Joseph Brzezinski (University of Colorado – Anschutz Medical Campus, USA) for providing the minimal promoter plasmids.
Footnotes
Author contributions
Conceptualization: S.P.F., S.J.F.; Methodology: S.P.F., S.J.F.; Formal analysis: S.P.F., S.J.F.; Investigation: S.P.F., B.E.D., S.J.F.; Resources: B.E.D.; Writing - original draft: S.P.F., S.J.F.; Writing - review & editing: S.P.F., S.J.F.; Supervision: S.J.F.; Project administration: S.J.F.; Funding acquisition: S.J.F.
Funding
This work was supported by the Children's Hospital Colorado Program in Pediatric Stem Cell Biology and The Boettcher Foundation (S.J.F.).
References
Competing interests
The authors declare no competing or financial interests.