Balancing selection on a regulatory region exhibiting ancient variation that predates human-neandertal divergence

Abstract

Ancient population structure shaping contemporary genetic variation has been recently appreciated and has important implications regarding our understanding of the structure of modern human genomes. We identified a ∼36-kb DNA segment in the human genome that displays an ancient substructure. The variation at this locus exists primarily as two highly divergent haplogroups. One of these haplogroups (the NE1 haplogroup) aligns with the Neandertal haplotype and contains a 4.6-kb deletion polymorphism in perfect linkage disequilibrium with 12 single nucleotide polymorphisms (SNPs) across diverse populations. The other haplogroup, which does not contain the 4.6-kb deletion, aligns with the chimpanzee haplotype and is likely ancestral. Africans have higher overall pairwise differences with the Neandertal haplotype than Eurasians do for this NE1 locus (p<10⁻¹⁵). Moreover, the nucleotide diversity at this locus is higher in Eurasians than in Africans. These results mimic signatures of recent Neandertal admixture contributing to this locus. However, an in-depth assessment of the variation in this region across multiple populations reveals that African NE1 haplotypes, albeit rare, harbor more sequence variation than NE1 haplotypes found in Europeans, indicating an ancient African origin of this haplogroup and refuting recent Neandertal admixture. Population genetic analyses of the SNPs within each of these haplogroups, along with genome-wide comparisons revealed significant FST (p = 0.00003) and positive Tajima's D (p = 0.00285) statistics, pointing to non-neutral evolution of this locus. The NE1 locus harbors no protein-coding genes, but contains transcribed sequences as well as sequences with putative regulatory function based on bioinformatic predictions and in vitro experiments. We postulate that the variation observed at this locus predates Human-Neandertal divergence and is evolving under balancing selection, especially among European populations.

Figure 1. Map of the NE1 locus.

The Linkage Disequilibrium (LD) block is determined using SNP data of the CEU population from the 1000 Genomes phase 1 data set using the HaploView program . Red represents regions with a high degree of LD and a high log odds score (LOD; D' = 1, LOD>2). The blue box indicates the regions flanking the CNVR8163.1 deletion where statistics for genome-wide comparisons were calculated. The sequence alignments show the breakpoints of the CNVR8163.1 deletion as determined by PCR amplification followed by Sanger sequencing. The first sequence represents human reference (NCBI36/hg18), and the following 9 DNA sequences are from individuals with the CNVR8163.1 deletion. Coordinates 37,624,055 and 37,628,634 mark the breakpoints of this deletion on chromosome 22. “-” in the alignment identifies missing nucleotides in these individuals, and “.” depicts nucleotides which were not shown in that interval for illustrative clarity purposes.

Figure 2. NE1 haplotypes share ancestry with Neandertals.

(A) The principal component analysis (PCA) of the SNP haplotypes from CEU, CHB/JPT and YRI populations indicates population substructure between African and Eurasian populations, as well as within Eurasians. Please note the separation of NE1 and NonNE1 haplogroup. (B) Nucleotide diversity (π) within populations is depicted. For most of the genomic segments, π is higher within the YRI population (blue line) when compared to Eurasian populations (red line). However, there is an increase of π for Eurasian populations at the NE1 locus, surrounding the CNVR 8163.1 deletion. (C) The evaluation of 12 SNPs (that separate NE1 and nonNE1 haplogroups) in the CEU sample, NA12891, as well as Denisovan, Neandertal and chimpanzee consensus haplotypes. (D) Normalized read-depth of the Neandertal and Denisovan sequences across the NE1 locus. Please note the drop of the sequence read-depth to 0 for the region corresponding to the human CNVR8163.1 deletion.

Figure 3. Ancient African origins of the NE1 haplogroup.

(A) Models of scenarios that could lead to NE1 haplotypes observed in humans and Neandertals. The frequency of the NE1 haplogroup is depicted in red and the frequency of the nonNE1 haplogroup in yellow. The red corresponds to higher frequencies, whereas yellow corresponds to lower frequencies of the NE1 haplotypes in the population. The arrows represent the direction of possible admixture events. The left panel represents a model, under which the NE1 haplotypes admixed into Eurasian populations (Asn and Eur) after Human-Neandertal divergence. The second model, which is depicted in the central panel, is similar to the first model, except with the addition of more recent back migration of Eurasian NE1 haplotypes into Africa (Afr). The right panel shows the third model, under which the NE1 haplotypes among humans are explained by persistence of ancient African substructure. All these scenarios were based on the assumption that the NE1 haplotype occurs at high frequency or is fixed in the Neandertal population given that the available Neandertal sequences align well to the NE1 haplotype. (B) Geographical distribution of the NE1 haplogroup. We estimated the proportion of chromosomes that carry the CNVR8163.1 deletion from various sources described in Materials and Methods . The dark red portion of each circle represents the frequency of the homozygous nonNE1 genotypes, the white represents the homozygous NE1 genotypes and the light red represents the frequency of heterozygote genotypes. Note the existence of the NE1 haplotypes (i.e., as heterozygotes, light red) among sub-Saharan African populations (e.g., LWK and ASW) as well as the high frequency of heterozygotes (light red) in the European populations. (C) The pairwise distances between the African (Afr) NE1 haplotypes, the Asian (Asn) NE1 haplotypes, and the European (Eur) NE1 haplotypes, calculated using phase 1 data from the 1000 Genomes Project. p-values were calculated by the Mann-Whitney test.

Figure 4. Selection acting on the NE1 locus.

(A) Maximum likelihood tree based on select NE1 (red) and nonNE1 (blue) haplotypes, with the chimpanzee haplotype as an outgroup. The gray-box indicates the estimated interval for the Human-Neandertal divergence between 400,000–800,000 years ago . Note that the coalescence at this locus is extremely long and very unlikely to have evolved under neutral conditions as modeled here. (B) Comparison of F_ST and Tajima's D values of 10 kb intervals across the human genome. The red to dark blue gradient indicates decreased density of observed events at a given location in the graph. The NE1 locus, and other loci with similar profiles, are highlighted in white.

Figure 5. The regulatory functions of NE1 locus.

(A) The LTR region was determined using the Repeat Masker Track version 3.2.7 . Dark red boxes indicate the location of amplicons for ChIP qPCR. TFBS-1 and TFBS-2 refers to the two predicted transcription factor binding sites that were predicted by the latest ENCODE project data release, available in UCSC Genome Browser for the hg19 assembly. The CNVR8163.1 deletion polymorphism is flanked by black vertical lines. The blue vertical lines indicate the approximate locations of SNPs that differentiate between NE1 and nonNE1 haplogroups and overlap the TFBS-1 and TFBS-2 transcription factor binding sites. The 8 SNPs that overlap with TFBS-2 are from 5′ to 3′, rs132525, rs4306795, rs4434085, rs5750701, rs35853418, rs6001308, rs6001309, and rs9622868) (B) Chromatin immunoprecipitation quantitative PCR (ChIP-qPCR) results across the NE1 locus for representative samples NA12155 and NA10851 that belong to NE1 and nonNE1 haplogroups, respectively. The locations of the amplified segments (P1–P6) are shown in dark red rectangles in (A). The positive control primers amplify a segment within BCL6 gene on chromosome 3 that is known to have high H3K4me2 occupancy. The blue stars indicate significant differences in qPCR amplification between NE1 and nonNE1 haplotypes (p<0.01). The brown and blue arrows indicate qPCR primers that are closest to the predicted transcription binding sites (P1, P2, P3 for TFBS-1 and P7 for TFBS-2). Overall, our results demonstrate that H3K4me2 is enriched in NA12155 cells, which harbor the NE1 deletion as compared to NA10851 cells which do not have the deletion (data plotted represents the average of four replicate experiments ± Std. Error).