The analysis of APOL1 genetic variation and haplotype diversity provided by 1000 Genomes project

We analyzed all the SNPs data of APOL1 derived from 1000 Genomes Phase 3 Pipeline, included 2504 individuals from 26 human populations as described in NCBI Variation Glossary (http://​www.​ncbi.​nlm.​nih.​gov/​variation/​docs/​glossary). Through the comprehensive consideration, we classified those populations as four race included Asia, Africa, Admixed and Europe (Additional file 1: Table S2) [30, 32, 33]. First, we downloaded the VCF files containing the all SNP for APOL1 gene region (between positions 36,649,117 and 36,663,577 at chromosome 22) directly from the 1000 Genomes server (ftp.1000 genomes.​ebi.​ac.​uk//vol1/ftp/). A total of 2028 SNPs across APOL1 was identified using the dbSNP database (http://​www.​ncbi.​nlm.​nih.​gov/​SNP/​), while the number of APOL1 SNPs officially released by 1000 Genomes Project is 612. The two sequences defined by NCBI (DNA: NG_023228.1, and mRNA: NM_145343.2) were used as references in the study.

Generally, a SNP is considered as a true polymorphic site if its minor allele frequency (MAF) presents at least 1%. In this matter, we get 100 SNPs filter by MAF ≥ 1% in APOL1 gene region. As the conventional nomenclature, we consider the first base in APOL1 gene sequence as nucleotide +1, and classify the functional effect of each SNP as intronic, coding synonymous mutations, coding missense mutations, 5′ untranslated region (UTR) and 3’UTR (according the NCBI SNP annotation and Variation Viewer) (Additional file 1: Table S3). Only the SNPs located in the regulatory region or exon region were included for further analysis. Considering the annotation of APOL1 in NCBI (NM_145343.2), the first base of APOL1 initial codon ATG located at the nucleotide 893. The previous study indicated that there were many regulatory elements in the APOL1 gene upstream [14]. In this scenario, we defined nucleotides between −1200 and +892 (included APOL1 gene upstream and 5′ untranslated region) as CDS upstream regulatory region (URR) to include all regulatory elements. The 3’UTR covered 1497 nucleotides from nucleotide 12,964 to 14,461.

Then we checked the 100 SNPs reference allele, alter allele and MAF in NCBI and Ensemble database compared with 1000 Genomes Browser to ensure the integrity and correctness of the information. For example, rs74904227, rs136162, rs136163, rs80424, rs136174 and rs189436505, the six SNP sites were the triallelic alleles according NCBI and Ensemble, which be properly corrected as per 1000 Genome data (Although a triallelic SNP is described at NCBI, we did not find the third allele in 1000 Genome data). In the Allele2 column of Additional file 1: Table S3, the underline bases indicate the standard alter allele in 1000 Genome data. However, NCBI, Ensemble and 1000 Genomes data indicated that rs136162 was trialleic, so this polymorphism site was discarded to keep the consistency of the data format.

Haploview 4.2 software [34] was used to analyze linkage disequilibrium (LD) among the 99 variation sites. It showed that the LD was stronger within a block, and the number of major haplotypes was limited [35]. Hardy–Weinberg equilibrium p value (HWp) was calculated for each SNP by Haploview 4.2 software. We used a set of tagging SNPs selected from LD data and tagger option of Haploview to predict values of remaining SNPs in the dataset [36], combined the block-based and block-free approach [37] to improve the representativeness and accuracy of tag SNP, and assess the accuracy of those tag SNP by evaluating the value of r² and D’ in Haploview analysis. We chose Tag SNP for APOL1 URR, coding region and 3’UTR region respectively, which presented relative higher MAF and meet the HW equilibrium tested by Haploview software. Finally, the selected tag SNPs were used to evaluate the haplotypes in the four populations by HaploStat software (version 1.7.7) [38]. Chi-square statistical method was used to test the differences of each haplotype among four populations.

APOL1 gene was surrounded by some of the most polymorphic genes in the human genome (Fig. 1) such as APOL2 (13,117 bp upstream), APOL3 (86,894 bp upstream), and APOL4 (48,238 bp upstream), and other non-APOL family loci such as MYH9 (13,746 bp downstream). APOL1 gene had four LD blocks (Additional file 1: Figure S1) and 43 variants conformed to HWp > 0.05. A total of 12 polymorphism loci distribution (bold SNP) in APOL1 whole segment were selected as tag SNP (Additional file 1: Figure S2).

There were 8 haplotypes with a global frequency higher than 1% (Table 1). The frequency of each haplotype in Admixed, Asian, African and European population was listed in Table 2. Out of the eight haplotypes, the frequencies of seven haplotypes were significantly different among the four populations (P > 0.01). The frequency of haplotype S-3 is extremely higher in Africa than in other three populations (Fig. 2).

There were 79 SNPs in the URR of APOL1 and 92 SNPs in 3’UTR. Total 12 SNPs in URR (Additional file 1: Table S4) and 24 SNPs in 3’UTR (Additional file 1: Table S5) were considered as common variants with MAF ≥ 1%.

Eight SNPs in URR were selected as Tag SNP for haplotype analysis. There were 18 haplotypes emerged, but only 4 haplotypes had a global frequency higher than ≥1% (Table 3). We named the 4 haplotypes as URR-1, URR-2, URR-3, and URR-4 briefly. Considering the global frequency of each haplotype, it was worthy mentioned that the four haplotypes could account for more than 96% of all haplotypes (Table 3). The frequency was significant different among the four populations for each of the 4 haplotypes (Table 4). The frequency of haplotype URR-1 was much lower in European than in other three populations. The haplotype URR-2 was very rare in African populations (Table 4, Fig. 3).

Four SNPs in 3′ UTR region were selected as Tag SNP and produced 9 haplotypes. Five haplotypes reached a global frequency higher than 1% (Table 5). The haplotypes distributed very differently in African than in other three groups: UTR-1 and UTR-5 were much higher, and UTR-2, UTR-3, as well as UTR-4 were much lower in African population (Table 6).

Only 52 SNPs in APOL1 coding region were recorded in the 1000 Genome database. The MAFs of 11 variant presents at least 1% and 7 of them presented more than 5% (Additional file 1: Table S6). The 11 SNPs were either coding synonymous mutations or missense variants, most of them located in exon 7 (only the first one located in exon 6) (Additional file 1: Figure S3). Seven SNPs were selected as Tag SNPs for haplotype analysis. The three haplotypes with a global frequency higher than 1% were listed in Table 7, named as H-1, H-2 and H-3 haplotype. The frequency of H-2 haplotype was much lower in African population and H-3 was much lower in Asian population (Table 8).

Both MAFs the two SNPs belonging to G1 in Africans were 26%, but both were absent in Asians. In the Hardy–Weinberg equilibrium analysis, HWp < 0.05 indicate that they’re not confirm to Hardy–Weinberg equilibrium and should be discard. Considering the importance of G1 in APOL1 function, we extended our haplotype analysis to include the two SNP. Total 26 haplotypes were observed for the 9 Tag SNPs and 3 haplotypes reached a global frequency higher than 1% (Additional file 1: Table S7). The distribution of the three haplotypes was significantly different among the four populations and was similar to the three haplotypes without G1 variants (Additional file 1: Table S8, Fig. 3). But the frequency of haplotype H-1 decreased in African population when we added the two SNP of G1 for haplotype analysis (Fig. 3).

The 11 selected SNPs in coding region included either synonymous mutations or missense variants, most of them located in exon 7 (only the first one located in exon 6). They jointly participated in the polymorphism of APOL1 four different transcripts and impacted three proteins isoform structure. The limited variants of APOL1 coding region mostly distributed in the PFD, MAD and SRA-interacting domain (Additional file 1: Figure S3) indicated that these SNP has some significant impact for the function of APOL1 protein. The previous study indicated that SP was dispensable for its toxicity, PFD was required but not sufficient for APOL1 mediated toxicity, and integrity of MAD and SRA was critical requirement for the cell injury activity of APOL1 protein [10]. It indicated that those SNPs could contribute a significant effect on the function of APOL1 protein. For example, G1 (rs60910145) played an important role in trypanosome lysis [40] and susceptibility to kidney diseases [26, 27, 41, 42] or schizophrenia [43].

An earlier study reported that about 38% African carried with G1 risk allele [40], in 1000 Genome Project, the MAF of G1 was 26% in African population but absent in Asian and European. Two SNP of G1 didn’t conform to HWp > 0.05, the frequency of G2 also absent in the initial released 1000 Genomes VCF files (Additional file 1: Table S6). Considering the importance of G1 and G2 in APOL1 function [44], we extended our haplotype analysis to include the two G1 SNPs.

The distributions of haplotypes with or without G1 variants were significantly different among the four populations (Fig. 3), G1 two SNP that were previously presented a higher frequency actually pull down the haplotypes frequency when considering all populations pooled together (global frequency). We suspected that the main reason was the two variation sites didn’t conform to the HW equilibrium. The conclusion could safely draw from the two results at different conditions that the influence of two SNPs of APOL1 risk allele G1 in coding region haplotypes seemingly not prominent, it may influence the progress of the relevant diseases independently.

We also analyzed variants in the URR and 3′ UTR of APOL1. The previous study indicated that there were many regulatory elements in the APOL1 gene upstream, for example, activating protein-1 (AP1) at −1034, sterol regulatory element binding protein (SREBP) sites at −1185, and large number of zinc finger binding sites (MZF1) distribution at APOL1 gene promoter region [14]. In this study, we included the nucleotides between −1200 and +892 to cover all upstream regulatory elements as URR and 1497 nucleotides for 3′ UTR for further analysis.

The URR haplotype URR-1 present similar frequency among Admixed populations, Africans and Asians could indicate it play the same role in the three populations. While an opposite pattern is observed for other three haplotypes, their frequency in European populations is higher than other three populations. In addition, URR-2 is very rare in African populations. The results suggest that URR-1 and URR-2 could play different roles in African and European populations respectively.

The 4 haplotypes of 3’UTR haplotype was significant different among the four populations, UTR-1 appear a highest frequency in Africa populations, this finding suggests a stronger association between UTR-1 and APOL1-related disease in African populations. Haplotype UTR-2 and UTR-3 are absent or rare in populations of African ancestry, and haplotype UTR-5 is absent or rare in Asia and present relative higher frequency in Africa. The result shows the diversity of haplotypes effects on susceptibility of APOL1 gene associated diseases.

Considering we have to construct haplotypes on the basis of the accurate genotypes at huge polymorphic sites, HW equilibrium was always checked using Haploview software, and data deviated strongly from the equilibrium were submitted to retyping or discarded. However, there still some shortcomings in this study: (1) The population composition of Admixed group is relatively complex, it may affect the analysis results, data from Africa, Asia, and Europe are more reliable. (2) A little we know about the effect of intron mutations in APOL1 gene, this study did not analyze the mutations in the intron of APOL1 gene.