Adaptive sequence divergence forged new neurodevelopmental enhancers in humans

Abstract

Searches for the genetic underpinnings of uniquely human traits have focused on human-specific divergence in conserved genomic regions, which reflects adaptive modifications of existing functional elements. However, the study of conserved regions excludes functional elements that descended from previously neutral regions. Here, we demonstrate that the fastest-evolved regions of the human genome, which we term "human ancestor quickly evolved regions" (HAQERs), rapidly diverged in an episodic burst of directional positive selection prior to the human-Neanderthal split, before transitioning to constraint within hominins. HAQERs are enriched for bivalent chromatin states, particularly in gastrointestinal and neurodevelopmental tissues, and genetic variants linked to neurodevelopmental disease. We developed a multiplex, single-cell in vivo enhancer assay to discover that rapid sequence divergence in HAQERs generated hominin-unique enhancers in the developing cerebral cortex. We propose that a lack of pleiotropic constraints and elevated mutation rates poised HAQERs for rapid adaptation and subsequent susceptibility to disease.

Keywords: HAQER; comparative genomics; human accelerated regions; human evolution; human genetics; neurodevelopment.

Figure 1:. HAQERs, the fastest-evolved regions of the human genome.

(A) We display the values of velocity (v), initial velocity (v₀), and acceleration (a) in the phylogenetic context of recent human evolution. (B, C) Mean selection parameter estimates for 500bp genomic regions binned by either acceleration (B) or velocity (C). Error bars display the 95% highest density credible interval. Both acceleration and velocity correlate with signatures of selection in human populations. (D) HAQERs (Human Ancestor Quickly Evolved Regions) are identified as regions containing at least 29 mutations in a 500bp window (p < 10⁻⁶) that separate the inferred human-chimpanzee ancestor sequence from the human genome. We count insertions and deletions as one mutation regardless of their length. (E) Locations in the human genome of the 1581 HAQERs (blue markers). Marker amplitude reflects the maximum divergence density observed in each region. HAQERs are distributed across all human chromosomes and enriched near chromosome ends. (F) Cumulative distribution of velocity, initial velocity, and acceleration observed across HAQERs, Human Accelerated Regions (HARs), and random neutral proxy regions (RAND). Regions are filtered to a minimum element size of 50bp. (Bonferroni-adjusted Wilcoxon, ****; p < 0.0001). See also Figures S1–2.

Figure 2:. HAQER sequence divergence was driven by positive selection prior to the human-Neanderthal split.

(A) Derived allele frequency spectra representing 501 individuals from African populations (1002 alleles) for segregating sites within HAQERs, RAND, HARs, ENCODE candidate cis-regulatory elements (cCREs), missense variants (MISSENSE), or ultraconserved elements (UCEs). (B) Representative MCMC trace for the mean selection parameter acting on segregating sites within each set of regions. (C) Posterior mean and 95% highest density credible intervals describing the mean selection parameters for each set of regions inferred from segregating sites from five independent populations of unrelated African individuals. (D) Enrichment for high derived allele frequency (DAF > 0.99, left), low frequency (DAF < 0.01, center), and rare minor allele (DAF < 0.01 or DAF > 0.99, right) segregating sites relative to RAND (*: p < 0.05, Bonferroni-adjusted Mann-Whitney U). Each point represents the enrichment for one population of individuals partitioned from the set of all African individuals. (E) Distribution of the cluster separation (measured as the Dunn Index) between Ancient Hominins and Chimpanzees (left), Modern Humans and Ancient Hominins (center), or Modern Humans and Chimpanzees (right). Comparisons are presented between HAQERs, RAND, and HARs (Bonferroni-adjusted Mann-Whitney U, *: p < 0.05, **: p < 0.01, ****: p < 0.0001). See also Figures S3.

Figure 3:. HAQERs are enriched in bivalent chromatin states.

(A) Overlap enrichment/depletion matrix between HAQERs (top) or HARs (bottom) for 15 chromatin states (rows) from 127 reference epigenomes (columns). HAQERs are enriched for bivalent chromatin states but not for active enhancer and promoter states. An expanded matrix with individual sample annotations is presented in Figure S5A. (B) Volcano plot displaying significant overlap enrichments for HAQERs and the Bivalent Enhancer chromatin state in various tissues. (C) HAQER overlap enrichment for bivalent chromatin states compared between reference epigenomes derived from cultured cells and those derived from primary tissue (t-test, *: p < 0.05, **: p < 0.01). See also Figure S4.

Figure 4:. Rapid sequence divergence in HAQERs generated hominin-specific neurodevelopmental enhancers.

(A) Experimental design. Candidate HAQERs were prioritized based on overlaps with epigenomic datasets, cloned into a STARR-seq vector, and electroporated into developing mouse brains along with a pCAG-GFP transfection reporter. Single-cell sequencing followed dissection and FACS enrichment of GFP+ cells. (B) UMAP projection of 7,170 single cells from two scSTARR-seq experiments, labeled with metacluster identities. Inserts display cells colored by expression of the GFP transfection reporter and Human HAQER0169. (C) Enhancer activity score, defined as the input-normalized UMI count pooled across all cells per 1000 reporter UMIs, for 13 HAQERs. Nearest gene name is displayed below each HAQER ID. We display significant differences in enhancer activity between Hominin (Human, Neanderthal, Denisova) and non-Hominin (Chimpanzee, human-chimpanzee ancestor (HCA)) sequences (Bonferroni-adjusted t-test, p < 0.05). Faded bars represent sequences where Neanderthal and Denisovan had the same sequence in the 500bp genomic region and these duplicate sequences are not included in the statistical analysis. (D) Cell type enhancer activity score, or the input-normalized reporter UMI count normalized to the pCAG-GFP UMI count for each cell averaged across all cells in each metacluster (RG=Radial Glia, IP=Intermediate Progenitor, EN=Excitatory Neuron) (FDR- corrected t-test, p < 0.05). See also Figure S5–6.

Figure 5:. Rapid divergence of hominin-specific neurodevelopmental enhancers near FOXD4 family genes followed multiple segmental duplications.

(A) Experimental design. We cloned the human or inferred human-chimpanzee ancestral sequence of HAQER0059 into an PGK-EGFP reporter plasmid and delivered plasmids to the developing cerebral cortex via in utero electroporation at E15.5 alongside an mCherry injection reporter. We performed dissection, sectioning, and imaging 24 hours later. (B) Representative images of Hoechst-stained coronal sections imaged for the mCherry injection reporter and EGFP enhancer reporter. Scale bar, 100 μm. (C) Left: Quantification of PGK-EGFP reporter signal normalized to the mCherry injection reporter for HAQER0059. Right: Corresponding in vivo STARR-seq results. (*: p<0.05, ***: p<0.001, FDR t-test. Dotted line: Negative Control Mean + 3SD). (D) Phylogeny of HAQER0059 homologs in humans and other great apes. (E) Genomic context for the paralogous regions near the genes FOXD4L3, FOXD4L1, and FOXD4. We present genomic context for the region near FOXD4, which contains HAQER0059, on the reverse strand. The region near FOXD4L3 does not have a nearby HAQER and shares synteny with the great ape ortholog. (F) A model of recent FOXD4 evolution. The great ape ortholog of the human gene FOXD4L3 generated the paralog FOXD4 in the chromosome 9 subtelomere through paired inversion and duplication. Subsequent duplication produced the paralog FOXD4L1 at the fusion site between the ancestral chromosomes 2a/2b, which formed the modern human chromosome 2. See also Figure S6.

Figure 6:. HAQERs are enriched near genetic variants linked to human disease.

(A) Experimental design. For each GWAS catalog variant, we identified all linked variants (R² > 0.7). We calculated overlap enrichments between HAQERs and the set of all linked variation for all SNPs associated with a GWAS trait. (B) FDR-corrected p-values for overlap enrichments between HAQERs and GWAS Catalog traits. The dotted line marks pAdj = 0.05. (C) Representative overlaps between HAQERs (red) and 1000 Genomes variants (purple) linked to GWAS catalog variants (black labels). See also Figure S6.