Human brain evolution: Emerging roles for regulatory DNA and RNA

Abstract

Humans diverge from other primates in numerous ways including their neuroanatomy and cognitive capacities. Human-specific features are particularly prominent in the cerebral cortex, which has undergone an expansion in size and acquired unique cellular composition and circuitry. Human-specific gene expression is postulated to explain neocortical anatomical differences across evolution. In particular, non-coding regulatory loci are strongly linked to human traits including progenitor proliferation and cortical size. In this review, we highlight emerging non-coding elements implicated in human cortical evolution, including roles for regulatory DNA and RNA. Further, we discuss the association of human-specific genetic changes with neurodevelopmental diseases.

Introduction

The evolution of the human brain with its cerebral cortex occurred over 300 million years [1–3]. The neocortex is a highly-organized and specialized structure which controls higher-order functions, such as cognition, language, reasoning and emotion. Compared to other primates, the neocortex has expanded tremendously in size and accounts for almost 80% of the total brain mass [3]. Additionally, human cortices contain more neurons and glia, which are themselves posited to be more complex. The cellular mechanism of evolutionary cortical expansion was first postulated in the radial unit hypothesis, which attributed size differences to the proliferative capacity of neural precursors [4]. Indeed, an evolutionary increase in progenitor number, including outer radial glia (oRG, also termed basal radial glia), along with an expansion of germinal zones, including the outer subventricular zone (OSVZ), is thought to underlie increased neuron number [5–7]. Humans also undergo longer gestation resulting in prolonged neurogenesis relative to other species--another factor posited to influence human brain development. Taken together, these anatomical and cellular changes collectively contribute to unique human cognitive capacities (Figure 1).

Figure 1. Evolutionary development of the human cortex.

A. Graph showing a phylogenetic tree indicating the brain morphology of chimpanzee and rodents compared to humans. Primate stem has been highlighted. B. Cartoon representing comparisons of rodent and primate neocortical development. Progenitors are designated by their locations in the cortex. Apical radial glia (aRGs) which localize to the ventricular zone (VZ) can undergo symmetric and asymmetric divisions to self-renew or produce neurons directly or indirectly by producing intermediate progenitors (IPC) in the subventricular zone (SVZ). Newborn neurons will migrate through the intermediate zone (IZ) and arrive to the cortical plate (CP) with first born neurons located deeper and late born neurons located superficially. In the developing primate brain, SVZ expansion leads to a new germinal zone, the outer subventricular zone (OSVZ), which includes proliferative IPCs and basal radial glia cells (bRGs). The expansion of basal progenitors in turn, leads to more upper layer (mainly layer II/III) neurons in the primate neocortex. C. Comparisons of notable cellular differences between human and non-human primate cortical development.

State-of-the art genomic and genetic analyses of humans and other species, including comparisons between modern and archaic humans, have implicated human-specific gene expression in cortical evolution [8]. Changes in non-coding regions have long been proposed to have a crucial role in human brain evolution [9]. This is increasingly supported by empirical data suggesting that gene regulatory regions, such as enhancers and noncoding RNAs play an instrumental role in brain evolution [10]. Indeed, broader and fine-tuned regulation of many genes may be achieved by modifications in gene regulatory mechanisms.

In this review, we describe emerging evidence that modern human brain divergence is impacted by noncoding regulatory elements, including regulatory DNA and RNA (Figure 2). We focus on recent evidence implicating changes in promoters, enhancers, and non-coding RNAs in cortical evolution [11,12]. We further highlight current methods for experimental investigation of evolutionary differences. Finally, we discuss known mutations in human-specific loci linked to neurodevelopmental disorders, including schizophrenia, autism spectrum disorder (ASD), and neurodegenerative diseases.

Figure 2. Non-coding regulatory elements in human brain evolution.

Cartoon showing genomic modifications during human brain evolution. Representation of epigenetic changes, non-coding DNA and RNA alterations which have been implicated in human brain evolution, including the HARs, genomic methylation, and diversity of non-coding RNAs.

REGULATORY DNA

Epigenetic differences and evolution

DNA methylation is a stable epigenomic modification implicated in cell identity across human cortical development [13]. Recent studies comparing DNA methylomes of neurons and oligodendrocytes across primates showed that human-specific differences depend upon both specific cell-type and the cytosine-context [14]. Compare to other primates, humans have significantly higher CH methylation (hypermethylation where H is equivalent to A,C or T) but reduced CG methylation (hypomethylation). CH hypermethylation is especially associated with human neurons and CG methylation discriminates neurons from oligodendrocytes. Further, human neuron-specific CG methylation is associated with neurological disease, and schizophrenia risk. Methylation studies have also detected differences discriminating modern humans, Neanderthals, and Denisovans, including within facial and speech anatomy-associated genes [15,16]. These epigenetic differences have been used to reconstruct the anatomy of Denisovans [17]. By analyzing these epigenetic marks, we can further understand how gene expression may have been controlled in non-extant populations.

These studies implicate methylation in the human genome as a force for modern human divergence and raise fascinating questions. How might epigenetic marks parlay into functional changes and how do methylation changes shape modern human brains? The ability to perform gene editing and evaluate brain development using organoids (see box) offers new opportunities to address these questions [18,19].

Human-accelerated regions

Human Accelerated Regions (HARs) are genomic regions, ranging in size, which are conserved in non-human primates and other euarchontoglires, and present an unexpectedly high rate of mutations in humans [12,20]. These fast-evolving regions were identified through comparative genomics, with some studies taking into consideration human SNPs [21–23]. Various groups have estimated the exact number of HARs in our genome, but differences in datasets and computational definitions have led to only partial consensus on these regions’ identification. At least 2700 HARs are estimated to be present in the human genome [20]. Further, based on epigenetic marks and functional assays, around 30–40% of HARs are predicted to function as regulatory enhancers [21,24]. Predicted functions for non-enhancer HARs are less defined.

Many regulatory HARs reside in proximity to genes involved in brain development [25]. This led to the prediction that HARs influence neuronal development, including proliferation, differentiation, and axogenesis [10,24,26,27]. For example, 14 HARs are found within non-coding elements of NPAS3, which is linked to neural development [27,28]. A recent study using Hi-C, a technique to identify long range genomic interactions, in human fetal brain confirmed the physical interactions of several HARs with neurodevelopmental genes [10]. The authors identified three HARs predicted to drive expression of GLI2, GLI3 or TBR1, which are implicated in diverse aspects of cortical development. Inactivation of their enhancer activity in human neural progenitors using dCas9-VP64 (see box) led to decreased expression of the three target genes. These chromatin architecture analyses and epigenetic manipulations in developing human brain cells suggest HARs can influence diverse developmental stages and cellular processes.

HAR activity is also associated with layer specification and cortical size. This is particularly interesting given that over the course of evolution the human neocortex has undergone exceptional enlargement accompanied by increased neuronal generation [8]. The relationship between HARs and cortical size was evidenced by Boyd et al., who used mouse transgenic models to demonstrate differential enhancer activity by the human and the chimpanzee orthologs of HARE5 (HAR Enhancer 5), which promotes expression of Fzd8 [29]. Expression of human HARE5 in mice accelerated progenitor cell cycle and increased brain size. It remains to be seen whether experimental manipulation of HARE5 has a similar impact in primates.

Beyond cortical size, HARs are also implicated in higher human capacities such as language. Complex spoken language, a uniquely human characteristic, is linked to the FOXP2 locus that includes 12 HARs [30]. When introduced into mouse or zebrafish, these human and chimpanzee HAR sequences exhibit strikingly different spatial and expression patterns during brain development. Notably, these orthologous sequences contain distinct predicted binding sites for transcription factors, suggesting possible mechanisms by which FOXP2 expression may be differentially controlled across species. The direct interaction between these HARs and the FOXP2 promoter has not yet been demonstrated and the actual role of these specific HARs in human language is yet uncharacterized.

HARs have also been associated with characteristics of modern humans. For example, changes in loci discriminating modern humans have been linked to cortical surface area, including HARs which control expression of the transcription factor HEY2 [31]. Further, accelerated regions control the differential expression of 212 neurodevelopmental genes which distinguish modern and ancient humans and are postulated to affect cortical size and chromatin regulation [32]. These new data collectively implicate HARs both in macroscopic phenotypes discriminating human and other primates, as well as in subtle changes in the human lineage.

The literature to date suggests enhancer HARs are central for human brain evolution, yet many fundamental questions remain. In particular, the efforts to characterize and index the ~2700 HARs are not yet comprehensive and there is no common repository of datasets. Additionally, to date, high throughput investigation of HARs has been limited to cell culture models. While cell culture can provide broad insights into functions, these assays measure HARs outside their normal genomic context, and transient interrogation may provide just a snapshot of HAR function. Indeed, some HARs show temporal differences across developmental stages, reinforcing the value of in vivo investigation using stable expression in model systems [24,29]. Mouse models have been invaluable for assaying HARs across different tissues and developmental or adult stages, but their use raises concerns with the trans-environment. Remarkably, a vast majority of HARs apparently preserve differential activity regardless of the species tested, showing that regulatory function in cis is more relevant than the trans environment [33]. This provides rationale that HARs can be characterized in model organisms. Ideally, multi-pronged approaches using model organisms and species-specific iPSCs (see box) can provide complementary information.

NON-CODING RNAS

microRNAs

MicroRNAs (miRNAs) are short endogenous single-stranded RNAs (20–24 nts) which are especially relevant in the nervous system, where about 70% of the 2500 mature human miRNAs are expressed [34,35]. Functionally, miRNAs regulate gene expression by repressing translation or degrading their target mRNAs.

Comparison of human, chimpanzee and macaque neocortical transcriptomes led to the discovery of differential miRNA expression including human-specific changes [7]. For example, in silico analysis across 11 mammalian species identified one miR941, which is implicated in cell differentiation by controlling target genes in the hedgehog- and insulin-signaling pathways [36]. A recent study discovered corticospinal motor neuron (CSMN)-enriched miRNAs in the mammalian brain, and showed that misexpressing miR-409–3p in vivo and in vitro increased CSMN neurons at the expense of deep-layer neuron development [37].

miRNAs have also been implicated in multiple aspects of primate divergence including expansion of the germinal zones and increased neural progenitor complexity [34,38–40]. One of the first studies to investigate this question used deep miRseq of laser dissected macaque visual cortex to discover novel primate-specific miRNAs [41]. These miRNAs were specifically expressed in primate germinal zones, with target genes related to cell cycle and neurogenesis. More recent discoveries have used single cell sequencing to identify cell-specific miRNAs and their targets in human fetal brains across development [42]. Their discovery of some miRNAs which are great ape-specific lays the groundwork for future functional studies of miRNAs in brain evolution. Another recent study discovered two miRNAs enriched in ferret proliferative zones that play roles in regulating basal progenitor expansion and neural differentiation [43]. Notably, the validated targets of these ferret miRNAs are related to cell proliferation and neurogenesis. Likewise, miR934 has been shown to influence expression of genes associated with neurogenesis and early-born subplate neurons [44]. Collectively, these studies implicate miRNAs in the evolutionary expansion and function of the OSVZ and basal progenitors in primates.

By controlling gene expression, miRNAs have the potential to significantly influence species-specific brain modifications. Recent studies catalog miRNAs and raise intriguing questions including how human-specific miRNAs regulate function of specific cell subtypes and what are their downstream targets? It is also interesting to consider exactly how miRNAs functionally diverge, as it was recently shown that most primate-specific miRNAs share highly conserved RNA structures [45]. Finally, it is fascinating to consider how miRNAs may have influenced the modern human divergence from our closest relative archaic humans including Denisovans and Neanderthals.

IncRNAs and circRNAs

In contrast to short miRNAs, long-non coding (lncRNA) transcripts are more than 200 nt, with an average length of ~3000 nt [46]. The structure of IncRNAs is similar to that of mRNAs, with 5’ and 3’ modifications, introns and exons, as well as cytoplasmic localization [11]. To date, the human genome is reported to contain more than 16,000 lncRNA genes, about 40% percent of which are expressed in the brain [47]. Further, about one-third of lncRNAs are primate-specific, with a large fraction found in the brain [48]. LncRNAs are implicated in cortical development, including neural stem cell maintenance, differentiation, and neural maturation [49].

Collectively, these data argue that lncRNA may play important roles in brain evolution. In 2006, Pollard et al. discovered the first HAR, called HAR1, which falls within the lncRNA HAR1F [50]. HAR1F is specifically expressed in human Cajal–Retzius neurons during early stages of cortical development although its function is unknown. More recent discoveries implicate a primate lncRNA, LncND, in cortical expansion, as it is expressed in neural progenitors and regulates Notch signaling [51]. Thus, the specific expression patterns of lncRNAs may provide potent spatial and temporal control of traits relevant for evolution. As many lncRNAs act as enhancers, it will be interesting to evaluate if this is the case for human-specific lncRNAs [47].

Recent studies also highlight an emerging function of circular RNAs (circRNAs) in species evolution. circRNAs are single-strand RNAs molecules composed of 1–5 exons formed into a circle as a result of non-canonical back-splicing events [52]. By comparing circRNA expression systematically across humans, non-human primates, and mice, a majority of circRNAs were found to be human-specific [53,54]. While functions for these in evolution are unknown, circRNAs themselves have been implicated in control of translation and are abundant in the brain [55]. These remarkable observations indicate that the future study of circRNAs can give valuable insights into cortical evolution.

NON-CODING REGULATION AND DISEASE

Given their role in evolutionary features which rely upon development, it is not surprising that mutations in non-coding regions, in particular within HARs, are implicated in neurodevelopmental diseases [56]. Doan et al. found a rare de novo duplication of a HAR located upstream of NR2F2, a gene implicated in ASD risk, with higher frequency in ASD probands than in control siblings [26]. Moreover, they identified HARs predicted to be active in the brain and which interact with genes associated with ASD, brain malformations and cortical development. Notably, rare biallelic mutations in these HARs were estimated to contribute to approximately 5% of consanguineous ASD cases. GWAS studies of schizophrenic individuals also detected an enrichment for HAR-regulated genes involved in signaling related to GABAergic neurons, which is dysregulated in affected individuals [57]. In comparison, Primate Accelerated Regions (PARs) were less enriched, suggesting that changes in human fast-evolving regions may have facilitated functional changes. These data point to the influence of mutations in noncoding HARs upon developmental disorders and for human traits such as social behavior and cognitive capabilities.

While developmental diseases and cognitive dysfunction are the most common phenotypes deriving from disruption of HAR activity, recent studies show HAR mutations are also implicated in older-age onset diseases. Specifically, 93 enhancer HARs active in the nervous system have been studied for their association with cancer and Alzheimer’s [58]. Numerous miRNAs and lncRNAs have also been associated with neurodegenerative disorders, but human-specific changes have not yet been linked to these diseases [59]. Collectively, these studies highlight that the continued discovery of human-specific non-coding regulatory elements can give valuable insights into the etiology and treatment of neurological disease.

Summary

Comparative genomic and chromosome architecture studies have shown that a powerful mediator of human brain evolution resides in the non-coding regulatory elements of our genome. Here, we described some of these known changes and their implicated functional roles in cortical evolution. Given the new tools of single-cell (sc) RNA sequencing, genomic editing, and species-specific cerebral organoids, we will continue to gain a richer resolution of how these changes influence function in specific cell subpopulations and across developmental stages. By interrogating non-coding regulatory loci, together with protein-coding changes, we will move closer to understanding what makes our brains different from our closest living and extinct relatives. As mutations within human-specific non-coding loci continue to be associated with various disease contexts, future studies can provide molecular insights into the basis of neurological diseases.

Box - Methods

High-throughput approaches have been used to identify and evaluate the activity of regulatory DNA elements. While scRNA-seq produces a map of differential gene expression across cell types, scATAC-seq generates a list of the open genomic regions within precise stages and cell types. A synergy of these two methods now enables a clearer understanding of changing DNA regulation and gene expression across time and space [60]. New technological advances offer possibilities for functionally screening non-coding elements in mammalian cells. CRISPR interference (CRISPRi) uses sgRNAs complementary to specific DNA regions to drive nuclease-dead Cas9 (dCas9) and interfere with the gene expression. For example, dCas9-KRAB is widely used to repress enhancers and promoters [61,62]. Furthermore, massively parallel reporter assays (MPRAs) can interrogate enhancer activity and single-base variants in a library of regulatory elements using RNA reporters [21]. These approaches permit analyses of libraries of regulatory regions in different conditions allowing an exhaustive view of their activity. A challenge for studying human-specific traits is the need to evaluate functions in vivo, within a tissue architecture and at the level of circuitry and behavior. However, there are ethical and practical issues with obtaining and using human and non-human primate tissues. Brain organoids generated from various species have helped obviate these limitations by providing in vitro models of brain development [19,63,64]. Recent studies have also used hybrid human-chimpanzee brain organoids derived from tetraploid iPSCs [65,66]. This model enables a direct cross-species comparison and resolves the issues of cis- or trans- regulation of non-coding elements, thus providing a new alternative to existing methods.

Highlights.

• Differences in DNA methylomes discriminate modern and ancestral humans

• Human-accelerated regions transcriptionally control genes involved in cortical development

• Post-transcriptional mechanisms including miRNAs, IncRNAs, and circRNAs influence human brain evolution

• Mutations in human-specific loci are associated with neurological diseases