Epigenetic regulation of cortical neurogenesis; orchestrating fate switches at the right time and place

Introduction

The cerebral cortex is composed of a remarkably diverse population of neural cell types that arise from a heterogeneous group of progenitors with distinct spatial and temporal identities. The establishment of this stratified structure involves the highly orchestrated coordination of neural progenitor specification into distinct cellular subtypes, and their migration and integration into the local and long-range circuitry of the cortex.

The embryonic cerebral cortex arises initially from a pool of neuroepithelial cells (NECs) which undergo mass expansion during the closure of the neural tube. These NECs divide into radial glial cells (RGCs) at the onset of cortical neurogenesis [1]. RGCs serve as the primary neural progenitors of the developing cortex, and as they transition through a series of temporal competence states, give rise to distinct progeny [2,3]. Early on in neurogenesis RGCs divide asymmetrically to directly generate deep layer neurons in a process referred to as direct neurogenesis. During mid-neurogenesis the developmental competence of these RGCs transitions to generate upper layer neurons through the initial generation of intermediate progenitor (IP) cells via indirect neurogenesis [4,5]. These IPs symmetrically divide to form two new neurons that radially migrate past earlier born neurons to settle in the superficial layers of the developing cortex. In this way the cortical layers are generated in an inside-out fashion and corticogenesis as a whole is contingent to the temporally regulated transition of RGCs through a number of competence states.

As cortical development progresses, a proportion of RGCs then transitions to generate glia as opposed to neurons [6,7]. This termination of neurogenesis and subsequent production of astrocytes and oligodendrocytes is crucial for cortical development and determines the total neuronal number of the cortex [2]. Adding to the complexity of corticogenesis, cortical interneurons are generated outside of the cortex, mostly arising from RGCs in the transient germinal zone of the medial ganglionic eminence (MGE) [8,9]. These immature neurons tangentially migrate to the cortex where they laminate the cortex in the same temporally regulated inside-out manner during neurogenesis [10].

Producing the correct number and type of neuron requires precise spatial and temporal regulation of cell type and stage specific transcriptional programs of both progenitor and their progeny. However, the genetic and epigenetic networks that function to coordinate complex dynamic transcriptional programs remain poorly understood. In this review, we discuss the dynamic effects of epigenetic and RNA modifications that guide cortical cell fate specification and their impact on the development of the cortex.

Histone Modifications

The chromatin structure of eukaryotes consists of DNA wrapped around the nucleosome, an octamer of two copies of each H2A, H2B, H3, and H4 histone proteins. N-terminal of the tails of these histone proteins can undergo a variety of post-translational covalent modifications including methylation, acetylation, and phosphorylation, ubiquitination and sumolyation at specific lysine (K) and arginine (R) amino acid residues [11]. These histone modifications function to regulate the accessibility of associated DNA regulatory elements that ultimately lead to activation or repression of gene expression. Histone methylation at lysine residues occur commonly on histones H3 and H4, and include well documented modifications at H3K4, H3K9, and H3K27 positions in the chromatin. Methylation at H3K9 and H3K27 on the chromatin is associated with transcriptional silencing, whilst methylation at H3K4, H3K36 and H3K79, is associated with transcriptional activation[11]. It is therefore unsurprising that the conformational changes to chromatin structure induced by these modifications are necessary for cells to adjust their transcriptional response to intra- and extra-cellular signals as they transition between competence states during development. These modifications are associated with cis-regulatory elements including promoters, enhancers, silencers and insulators. The dynamic patterning of histone marks associated with promoters and distal enhancers have been characterized during progenitor fate transitions in a number of developmental lineages [11–14].

H3K27me3, mediated by Ezh2 and the Polycomb group complex, is dynamically altered in its distribution in the chromatin across developmental lineage progression[15,16**]. Temporal changes in the patterning of H3K27me3 enrichment in NECs and RGCs aid in the transition of these cells through developmentally regulated competence states during corticogenesis. In NECs, H3K27me3, and H3K4me3 are found together across the chromatin[12,16–18],genes marked by these two modifications are considered ‘bivalent’ and are abundant in early stem cells, and reduce during developmental lineage progression. Genes within bivalent domains are generally involved in cell fate specification and are poised for activation during cell differentiation [17,18].During cortical neurogenesis, the promoter region of genes involved in neural development such as the IP marker Tbr2, are dynamically regulated by H3K27me3[16**]. Loss of the histone methyltransferase Ezh2 in RGCs at different times during corticogenesis leads to differential effects on cell fate determination. Deletion of Ezh2 before the onset of neurogenesis accelerates neural lineage progression through a reduction of H3K27me3 and a subsequent increase in overall gene expression. [19]. In contrast deletion of Ezh2 during neurogenesis prolongs the process of neurogenesis and delays the onset of astrogliogenesis [20], establishing the role of H3K27me3 in the transition of radial glia through different cell competence states. The importance of this tight bivalent switch in the establishment of temporal diversity in cell fate choice in the developing cortex requires further research. H3K9me3 is a repressive post-translational histone modification that also plays a role in cell fate specification. Deletion of Setdb1, a histone methyltransferase that catalyzes H3K9me3, causes an increase in upper layer neurons through indirect neurogenesis, at the cost of deep layer neurons and also leads to early onset of astrogliogenesis [21].

Acetylation of lysine 27 histone H3 (H3K27) mediates conformational changes associated with transcriptional activation [11]. Cbp is a histone acetyltransferase that catalyzes a number of acetyl marks including H3K27ac. Knockdown of Cbp in RGCs reduced the number of late born upper layer neurons being generated and a diminished transition to astrogliogenesis [22]. H3K27ac is highly dynamic in enhancer-promoter regulation during neurogenesis, and becomes progressively restricted across neuronal differentiation[23]. Prdm16 has been found to temporally regulate the epigenetic state of these enhancers in RGCs to instruct neuronal fate specification through driving RGCs transition between competence states [24]. These enhancers were found to regulate genes that reflect cellular identity of distinct cell types through neuronal differentiation, indicating the importance of the temporal regulation of histone modification at these enhancer elements in defining cell fate during neurogenesis. However, the mechanisms which Prdm16 regulates these enhancer elements and affects H3K27ac leading to distinct cell fate choices is not understood.

DNA Methylation

DNA methylation, the addition of a methyl group to DNA molecules is a covalent epigenetic modification fundamental to development, which functions by influencing DNA-protein interactions [25,26]. Cytosine nucleotide methylation occurs most prominently at CpG residues, but has been found to occur at other dinucleotides, particularly CpA in the nervous system [27]. 5-methylcytosine (5mC) is generally considered a repressive mark, and cell-type specific dynamic changes in 5mC can help define cellular identity through transcriptional repression [28]. Methyltransferases Dnmt3a and Dnmt3b establish de novo methylation, whilst Dnmt1 identifies hemi-methylated DNA strands and is essential for the re-establishment and maintenance of DNA methylation. DNA methylation shows dynamic global changes across the developmental progression of RGCs through different competence states[29 **]. RGCs undergo to successive waves of demethylation in early and late corticogenesis, firstly to demethylate neurogenic genes initiating neurogenesis and then the demethylation of gliogenic genes marking the onset astrogliogenesis. Finally glial cells undergo mass de novo methylation at the sites of genes conferring neuronal identity to solidify glial fate [29**]. While evidence points towards its dynamic role in the temporal patterning of the cortex, the exact consequence of DNA methylation in cell fate determination is not well established. It is likely to play a lesser supplementary role in fate determination as a part of chromatin modification networks and as a consequence of higher-order histone modifications. In fact, there is evidence of this interplay; H3K9me3 plays a role in setting DNA methylation during early neurogenesis. Uhrf1 mediates Dnmt1 associated 5mc re-establishment and maintenance. It can recognize H3K9me3 marks and recruits Dnmt1 to the genome in an H3K9me3 dependent manner [30–33]. Therefore, H3K9me3 directed transcriptional repression may be a primary force in the role of DNA methylation maintenance and reestablishment for the orderly timing of cortical lineage progression.

Chromatin Remodeling

Gene expression and thus cell fate specification is regulated by large multimeric ATP-dependent chromatin remodeling complexes during development. Two chromatin remodeling complexes highly investigated in neural development are the NuRD complex and the BAF complex. In the developing cortex, cell-type and developmental stage specific expression of distinct sub-units within these complexes functions to enact temporal and cell specific gene expression programs to aid in the normal progression of a cell through developmental lineage. The BAF (SWI/SNF) complex is highly expressed throughout neural development and forms distinct complex compositions in different cell types across development. Brahma and Brg1 are core to the complex and are responsible for the ATP catalysis. RGCs uniquely express a BAF complex that includes BAF45a BAF53a, and BAF55a which are important for neural progenitor maintenance and regulation of the switch between neurogenic and astrogliogenic commitments. Deletion of either BAF45a/53a subunits leads to impaired proliferation of RGCs [34*]. As progenitors transition into neurons the progenitor subunits are replaced with BAF45b/c, BAF53b, and BAF55b [34*]. Deletion of BAF complexes leads to a global shift from active histone modification to repressive H3K27me3 marks. Transcriptional profiling found repressive marks predominately at genes associated with neuronal differentiation [35,36]. In contrast, genes with increased levels of active histone marks were associated with NE cell fate [35]. These studies provide insight on the role of BAF chromatin-remodeling complexes on the cell fate choices in corticogenesis.

The NuRD complex is involved in nucleosome remodeling and maintains a histone deacetylase activity through addition of HDAC1/HDAC2 as core subunits [37]. Other core members of complex include the methyl-CpG-binding domain proteins, Mbd1/2/3 and the histone binding proteins Rbbp4/7. The core subunit responsible for the ATP activity of the complex is provided by the chromodomain-helicase DNA binding proteins Chd3/4/5. Subunit exchange of the Chds are necessary for different aspects of corticogenesis with Chd4 initially promoting proliferation of precursors, Chd5 later facilitating neuronal migration and Chd3 directing proper laminar specification [38**]. The differential regulation of distinct sets of genes by these Chd subunits facilitate the action of the complex in establishing cortical neuron organization. Other members of the NuRD complex have been linked to cell fate determination. Conditional deletion of Mbd3 in the developing mouse brain lead to defects in cortical layer specification coupled with reduced cortical thickness[39]. There is an associated loss of Tbr2⁺ intermediate progenitors. Evidence suggests a role for the NuRD complex in silencing transcriptional machinery to control cell fate determination and differentiation.

High throughput chromosome conformation capture technologies have allowed for the assessment of the 3D spatial organization of the genome. Topologically associated domains (TADs) are subregions within segmented chromatin compartments that act as regulatory subunits and play role in differentiation). TADs are strongly associated with Cohesin and CCCTC-binding factor CTCF; CTCF binds well defined motifs and mediates chromatin looping interactions through dimerization and helps shape and maintain the 3D architecture of chromatin. Sequences flanking CTCF looping anchors are highly enriched for active histone modifications, and transcriptional machinery including RNA polymerase II (RNAPII). These machinery centers of transcription for lineage specific genes[40]. This 3D architecture exhibits dynamic changes across neural development and during the process of differentiation [41] suggesting an important role for the topological arrangements of the chromatin in cell fate determination. However how these conformational changes affect cell fate and developmental gene expression has yet to be completely explored.

Non-coding RNAs

Members of the family of non-coding RNAs, including microRNAs (miRNA) and long non-coding RNAs (lncRNA) are functional regulatory RNA that post-transcriptionally modulate gene expression. LncRNAs are categorized as RNAs longer than 200 nucleotides, whilst miRNAs are short ncRNAs consisting of about 20 nucleotides. Single cell analysis of lncRNAs in the developing human neocortex identified cell type specific expression of LncRNAs despite bulk tissue analysis [42*]. Loss of function studies of a number of lncRNAs indicate the importance of their role in cellular differentiation and fate specification in corticogenesis. The Intergenic lncRNA, Pnky, is a neural-specific lncRNA that regulates neurogenesis in both embryonic and postnatal proliferative niche. Pnky forms a complex with Ptbp1, a splicing factor and together they function to maintain RGC proliferative capacity [43]. A pair of LncRNAs, linc-Brn1b and linc-Brn1a, play an important role in the production of IPs and the specification of late born cortical neurons through their regulation of neuronal migration gene Brn1 [44].

MiRNAs bind mRNA transcripts and induce degradation and repress translation of their target mRNAs. is; miRNAs encoded by the miR-17–92 cluster are upregulated during this time point and are required for maintaining RGC and IP populations through repression of Tbr2 and Pten [45]. Deletion of miR-17–92 suppresses proliferation and expansion of RGCs, and increases the production of IPs [45].; and has been studied extensively during the specification of cortical neurons. Interestingly miR-9 and miR-124 have been implicated in the transition of the BAF complex from containing NPC subunits to neuronal subunits across during differentiation[46], providing evidence of the complex interactions between of different epigenetic modifiers involved in dictating cell fate in the developing cortex. A recently defined class of regulatory non-coding RNA that influence gene expression is enhancer RNAs (eRNAs). These RNAs are transcribed from active enhancer regions and are therefore tied to the patterning of active histone marks such as H3K27ac and H3K4me [47,48][47]. These active enhancers recruit in RNA Polymerase II that transcribes eRNAs in a bidirectional manner[47,49]. These eRNAs have been shown to promote enhancer-promoter looping and gene expression regulation at target promoters, and interact with transcriptional machinery to enact expression changes downstream of enhancer activation [49,50]. The direct function of different eRNAs on their targets during corticogenesis has yet to be elucidated, however due to known dynamic changes of histone modifications at enhancers during corticogenesis, it is likely these eRNAs have a complimentary role in cell fate specification. Indeed, recently Cbp was found to bind eRNAs at enhancer sites and the binding directly regulates the enzyme’s histone acetyltransferase activity at the enhancer and subsequent acetylation and expression of target genes[51**]. eRNAs likely bind and recruit proteins that regulate chromatin architecture and transcriptional activation to promote looping between enhancer and promoter elements in a locus specific manner to drive the expression of lineage specific genes through differentiation.

RNA modifications

Covalent modifications on nascent RNA regulate their interactions with various RNA binding proteins controlling the translation of target proteins. The most abundant and highly characterized of these modifications is N⁶-Methyladenosine (m⁶A), leading to increased rates of translation and decay of mRNA. This mRNA modifier is dynamic and is catalyzed by a Mettl3-containing methyltransferase complex. Knocking down Mettl3 in mouse ESCs impairs differentiation and increases proliferation of these cells [52]. m⁶A modification was found to be necessary for the termination of naïve pluripotency by destabilizing the mRNA transcripts. Conditional knock-out of Mettl14, a member of the methyltransferase complex necessary for m⁶A catalysis [53], leads to m⁶A depletion prolonging the cell cycle in RGCs and extending the neurogenic phase into postnatal development [54**]. This results in a specific decrease in later born upper layer neurons with no changes seen in lower layer neurons, and is coupled with a decrease in astrocyte numbers. m⁶A is enriched in mRNAs related to neuronal differentiation, cell cycle, and transcription factors, with the modification promoting their decay[53**]. Recent evidence of interplay between RNA modifications and histone modifications has emerged. M⁶A marks are present of Ezh2 mRNA transcripts, and depletion of the marks subsequently reduces Ezh2 protein expression and H3K27me3 modification on histones [55]. Therefore, dynamic m⁶A modification is necessary for the timely transition of RGCs between competence states to direct cell fate specification in the developing cortex.

Outlook

Understanding the processes underlying cell fate decisions has been a fundamental area of research in corticogenesis. Remarkable progress has been made in elucidating the complex genetic and epigenetic networks associated with the temporally defined cell differentiation and fate specification machinery in recent years. Rapidly advancing high-throughput sequencing technologies has allowed for the unbiased identification of dynamic epigenomic patterns during cortical development. Cell type specificity can now be defined through the use of single cell sequencing technologies including scRNAseq and scATAC-seq. With ever-improving tools and technologies it has become easier to perform combinatorial integrative analyses of both chromatin and transcriptional changes at single cell resolution. Advances in capturing full RNA profiles in cells using nascent transcriptomic technologies such as PRO-seq has allowed for the analyses of previously inaccessible species of RNA with functional relevance in development. As the field progresses, particular attention should be devoted to how dynamic interactions between the various epigenomic networks of transcriptional regulation, from high-order chromatin architecture to the DNA methylome, instruct the cell fate competence states in NPCs. Elucidating the interplay of the various modifications and their modifiers at a single cell resolution and in a time-resolved manner will shed a light on the determinants of the highly coordinated cell fate specification process in cortical development. Most importantly, as we begin to harness the single cell resolution descriptions of these dynamic changes in epigenetic modifications, we need to shift focus to understanding their significance.

Figure 1.

Cortical progenitors transition through a series of competence states, beginning with the symmetrical divisions of neuroepithelial cells that will eventually transition into radial glia. During early neurogenesis radial glial self-renew and give rise to neurons through direct neurogenesis. The majority of neurons born this way are destined to occupy deep layers of the cortex (blue). As corticogenesis continues, indirect neurogenesis becomes the main form of neuronal production as radial glia begin to give rise to intermediate progenitors that will typically undergo a symmetric terminal division producing a pair of neurons destined for the superficial layers of the cortex (green). As neurogenesis comes to an end radial glial change their competence and become gliogenic, generating cortical astrocytes.

Figure 2.

Illustration of key epigenetic factors and their influence on specific aspects of progenitor proliferation and fate.

Table 1.

Table complementing Figure 2 outlining the molecular function of key epigenetic factors and the specific consequences of their disrupted function.

Epigenetic Regulation	Gene	Biological Role	Consequences of depletion
Histone modifications	Ezh2 Setdb1 Cbp Prdm16	Enzymatic subunit of PRC2. catalyzes H3K27 trimethylation, leading to repression of genes. Catalyzes H3K9 methylation; regulates gene silencing and transcriptional repression. Coactivator with p300; histone acetyltransferase catalyzing H3K27ac leading to gene activation and active chromatin Histone H3 methyltransferase, regulation of developmental enhancer activity	Decreased H3K27me3; increase overall gene expression, increased intermediate progenitors and neurons, accelerated astrogliogenesis [19]; prolonged neurogenesis, delayed astrogliogenesis [20] Decreased H3K9me3; Increased indirect neurogenesis, reduced direct neurogenesis [21] Decreased H3K27ac; Reduced upper layer neurons from indirect neurogenesis, delayed astrogliogenesis [22] Decreased indirect neurogenesis, and defective differentiation of late born upper layer cortical neurons.[24]
Chromatin remodeling	BAF155, BAF170 Chd4 Mbd3	Core members of the BAF complex, remodel nucleosomes in a lineage specific manner, controlling developmental chromatin state. Member of the NuRD Complex, involved in chromatin remodeling. Promotes the proliferation of intermediate progenitors. Core component of NuRD complex, involved in recruitment of PRC2 complex for transcriptional repression.	RGC fate change, overactive proliferation, reduction of indirect neurogenesis and upper layer neurons [35]; global increase in repressive H3K27me3 marks [36] Microcephaly, increased cell cycle exit of progenitor cells, depletion of IPCs [38] Impaired cortical layer specification, reduced cortical thickness, loss of intermediate progenitors and indirect neurogenesis. [39]
Posttranscriptional regulation	Pnky/Ptbp1 Lnc-Brn1b/Lnc-Brn1a miR-17-92 Mettl14	lncRNA; mediates differentiation and involved in RNA splicing with Ptbp1. Complex mediates splicing program suppressing neurogenesis. lncRNA with role in neuronal proliferation. Specific mRNA regulation Along with Mettl3, forms a RNA binding scaffold complex that catalyzes n6 adenosine methylation.	Enhanced neurogenesis, reduced RGC proliferation [43] Loss of intermediate progenitors and late born upper layer neurons from indirect neurogenesis [44] Reduced proliferation of RGCs, increased indirect neurogenesis [45] Prolonged RGC cell cycle, prolonged neurogenesis, isolated loss of upper layer neurons from indirect neurogenesis [54]