Characterization of APOBEC3 variation in a population of HIV-1 infected individuals in northern South Africa

The study population was comprised of a total of 192 HIV-1 positive individuals from several ethnic groups (Venda, Bapedi, Tswana, Tsonga and Swati) who presented for routine care in clinics and hospitals in the Waterberg and Vhembe districts of the Limpopo province in Northern South Africa. There were 116 females and 76 males with an age range from 4 to 98 years and their viral load and CD4+ cell count ranged from < 20 to 623,250 copies/ml and 5 to 1353 cells/μl, respectively (Additional file 1: Table S1). These individuals were recruited from July 2013 to December 2015. DNA was extracted from peripheral blood mononuclear cells (PBMC), using the QIAamp DNA blood mini kit (Qiagen) according to the manufacturer’s instructions.

Primers to amplify the four APOBEC3 genes (A3D, A3F, A3G, and A3H) were designed using Geneious® 8.1.5 software (Biomatters, Inc.). A nested PCR strategy was used to amplify each APOBEC gene. The outer primer set was designed to flank and amplify a long gene fragment in the 1st polymerase chain reaction (PCR), while two sets of primers were designed to amplify two fragments of each gene in a nested PCR using the 1st round PCR product as the template (Table 1). The primer sets were chosen using the information for the A3D, A3F, A3G and A3H genes in the Ensembl Genome Browser (ENSG00000243811, ENSG00000128394, ENSG00000239713 and ENSG00000100298).

The Takara (LA) PCR Kit Ver. 2.1 for long DNA fragments amplification (Clontech) was used to amplify the complete 12.16 kb A3D, 13.31 kb A3F, 10.74 kb of A3G, and 6.8 kb A3H genes in a 1st round PCR reaction using genomic patient DNA. The 1st round primary PCR products were then used as templates in “nested” PCR reactions to generate shorter PCR products/ All the PCR reactions contained: 1X PCR Mg²⁺ plus buffer, 400 μM dNTPs, 0.2 μM of each primer (Table 1) and 1.25 units of LA Taq high fidelity polymerase in a total volume of 20 μl. The following cycling conditions were used for all PCR reactions: Initial denaturation at 94 °C for 1 min, 30 cycles of denaturation at 98 °C for 10s, annealing at temperatures varying from 53 °C to 68 °C for 15 min (depending on primers) and extension at 72 °C for 10 min. Final amplicons were purified using AMpure XP beads (Beckman Coulter) and quantified using a Qubit 3.0 Fluorometer with the dsDNA HS kit (Invitrogen). Equimolar concentrations of the two shorter amplicons generated for each gene were pooled and normalized to 1 ng using 10 mM Tris elution buffer.

Purified Tn5 transposase enzyme was used to fragment about 1-10 ng of DNA amplicons to sizes ranging from 35 bp to 700 bp, tagged with sequencing adaptors, in a manner similar to the protocol used in the Illumina Nextera Kit. The reaction mixture contained: 4 μl tagmentation buffer (5X TAPS-DMF), 1-5 μl Tn5 transposase (1X-5X) and 1-10 ng DNA, with an addition of nuclease free water to add up to a final volume of 20 μl. The reaction was performed at 55 °C for 5 min. The Tn5 transposase enzyme was produced and characterized in the University of Virginia laboratory, using published protocols [28]. Following this step, unique Illumina dual-index barcodes (index1 (i7) and index 2 (i5)) were added to each sample in a short PCR of 12 cycles, followed by a second AMpure XP bead purification, generating 300-500 bp indexed fragments for sequencing. Using the full complement of Nextera XT indices, up to 96 individual samples were pooled for each run.

After purification, libraries were size-verified using a bioanalyzer 2100 with a High Sensitive DNA assay kit (Agilent Genomics), quantified and normalized to a concentration of 4 nM each. The normalized libraries were then pooled, and denatured into single strands. For good cluster generation, 1.8pM of the pooled library spiked with 25–30% PhiX was then loaded into the sequencing cartridge. Biological sample sheets were created in Basespace by labeling each sample with the appropriate index and setting up a sequencing run for the MiSeq. Each run generated approximately 25 million reads/sequences per sample.

Sequences were demultiplexed automatically on the MiSeq as part of the data processing steps and ends pairing. FASTQ files were generated for each sample representing the two paired-end reads. Sequence quality was validated using the Galaxy NGS platform Quality Control tools for sequence manipulation which includes the fastQC program.

Sequencing data quality, including the duplication rate, percent GC, and read quality was assessed by quality control tools for high throughput sequencing data [29, 30]. After filtering low coverage samples, reads were aligned against the human genome with BWA-MEM [31]. Alignments were sorted, marked for duplicates, and indexed using SAMtools [32]. Variants were called using Freebayes, a Haplotype-based tool to detect variants using short-read sequencing data [33]. Variant calls were normalized and decomposed with vt, a unified representation of genetic variants, and functionally annotated using SnpEff, a program for annotating and predicting the effects of single nucleotide polymorphisms [34]. Comprehensive annotation and prioritization was performed using the GEMINI framework for Integrative Exploration of Genetic Variation and Genome Annotations [35]. All further data manipulation and analysis was performed using R, a Language and Environment for Statistical Computing [36].

There is limited availability of APOBEC3 gene sequences from African populations, and when sequencing has been performed, it has often been limited to A3G [39]. In this study, we applied next generation sequencing to determine variation in the coding exons of the APOBEC genes A3D, A3F, A3G and A3H in DNA from 192 HIV-1 positive individuals residing in the Limpopo province of northern South Africa. The proteins expressed from these genes have all been shown to be capable of HIV restriction [5]. APOBEC 3 variation in this region has not been reported previously.

In order to better understand the A3 genetic changes observed in each subject, all clusters of variation within the genes were assigned into haplotypes as described in materials and methods and their frequencies calculated. These haplotypes were classified as either confirmed or unconfirmed based on the number of heterozygous variants. This classification was necessary due to the fact that the NGS reads were short and thus in many cases we could not determine if SNPs occurred on the same chromosome (Table 3). Nonsynonymous variants were considered and their genotypes (homozygous or heterozygous) were indicated. Low frequency variants (MAF < 5%) were excluded from the haplotype assignment. Comparisons were made to the GRCh37 human genome whose combinations are represented as haplotypes in A3D, A3F and A3G (Table 3). We identified four confirmed haplotypes for A3D, four confirmed haplotypes for A3F and four confirmed haplotypes for A3G (Table 3). It is worth noting that only haplotypes for A3G and A3H have been described previously [12, 15, 40, 41]. In the case of A3H, there are seven well characterized and six additional haplotypes that were recognized more recently. The seven well characterized haplotypes of A3H were recently described as having an impact on the genetic diversity of HIV-1 Vifs in the global pandemic [12, 15, 16]. All of the known A3H haplotypes (I-XIII) are combinations of 5 nucleotide changes located in exons 2, 3 and 4. Haplotypes II, V, and VII have been termed stable, because of the observed relatively long half-lives of the encoded proteins, enabling them to restrict HIV-1. Four of the haplotypes (I, III, IV, VI) have been termed unstable, since the encoded protein half-lives have been shown to be short, resulting in complete loss of the ability to restrict HIV [12, 39]. In our subjects, we identified 4 haplotypes for A3H: the stable haplotype II (15 N, 18R, 105R, 121E 178D), haplotype III (15Δ, 18R, 105R, 121E, 178D), haplotype IV (15Δ, 18 L, 105R, 121E, 178D) and haplotype X (15 Δ, 18R, 105R, 121E, 178E) (Table 3) [11, 12, 39]. Haplotypes III, IV and X all have the amino acid 15 deletion, known to make the Apobec 3H protein unstable. From the data in Table 2 and this haplotype analysis we can conclude that 41.4% of our patient population cannot express any stable ApoBec3H proteins and thus lack the ability to restrict HIV using Apobec 3H.

We next compared the nonsynonymous and synonymous variant frequencies in the South African population in our study to previously reported variant frequencies in the following populations: African (AFR), East Asian (EAS), European (EUR), Ad Mixed American (AMR), and South Asian (SAS), as reported in the 1000 Genome Project phase III, the HapMap project (NCBI), the dsSNP database and the Ensembl genome browser. We also compared our allele frequencies to the ExAC consortium database that contains sequences from more than 60,000 individuals (Table 4).

In a previous study by Duggal and colleagues that compared Apobec 3 variation between Africans, Asian and Europeans, nonsynonymous variation in A3D (R97C, R248K); A3F (A108S, V231I, Y307C); A3G (H186R, E275Q (now Q275E) and A3H (15Δ, R18L, R105G (now G105R), E121K/D, E178D) were reported [13]. Our data suggest that several variants occur more frequently in our South African population than in the “African” population they previously studied [13]. These include R97C and T238A in A3D; A108S and Y307C in A3F; Q275E in A3G and N15Δ, R18L, G105R and E178D in A3H (Table 4).

Overall, the EAS, EUR, AMR, SAS populations and the ExAC consortium database showed a higher level of Apobec 3 conservation than our study population (Table 4). For example, the A3D sequences in these populations were more closely related to the reference GRCh37 human genome (98–100%) than in our SA population, resulting in signficant p-values for all the variant comparisons where allele frequencies were available. In the case of A3F and A3G, several variants were also present more frequently than in the other populations (see Table 4). In the case of A3H, the N15Δ variant was clearly present in significantly higher frequency in our population compared to the others. This was also the case for all of the other observed variants, with the exception of R18L and K140E, which as discussed above is likely a sequencing error or an extremely rare variant. R18L was significantly lower in all of the populations, with the exception of the AFR population, where it was not significantly different. This is in contrast to all of the other variants, which were significantly higher in our SA population than in the AFR population. In the case of A3 D, F and G, the frequency for some of the variants were also significantly higher in our population than in the AFR population, whereas others showed more similar allele frequencies (see Table 4).

The term “Africans” has been loosely used to describe datasets generated from different parts of the African continent. To provide a more accurate comparison, we next compared the variants detected in our study to the various components of the AFR data set that consist of more specific African subpopulations or people of African descent (Table 5). These included Americans of African Ancestry in USA (ASW); African Caribbeans in Barbados (ACB); Gambians in the Western Gambia (GWD); Esan in Nigeria (ESN); Luhya in Webuye, Kenya (LWK); Mende in Sierra Leone (MSL) and Yoruba in Ibadan, Nigeria (YRI). We noticed higher levels of single nucleotide changes in our population (with significant p-values) compared to most of the other populations for the following variants: T238A in A3D, S327S in A3F, S60S, Q275E and G363R in A3G and all of the variants in A3H with the exception of R18L (and K140E-see above). (Table 5). Notably, the variant frequency of R97C in A3D is almost the same as in ASW and LWK but higher than in the other populations. The frequency of R48P in A3F and the frequency of R256H in A3G were similar among all Africans.