APOBEC3H Haplotypes and HIV-1 Pro-Viral vif DNA Sequence Diversity in Early Untreated HIV-1 Infection

Abstract

We examined single nucleotide polymorphisms (SNP) in the APOBEC3 locus on chromosome 22, paired to population sequences of pro-viral HIV-1 vif of peripheral blood mononuclear cells (PBMC), from 96 recently HIV-1 infected treatment naïve adults. We found evidence for the existence of an APOBEC3H linkage disequilibrium (LD) block associated with variation in GA->AA, or APOBEC3F signature, sequence changes in pro-viral HIV-1 vif sequence (top significant 10 SNPs with a top-significant p=4.8×10⁻³). We identified a common 5 position risk haplotype distal to APOBEC3H (A3Hrh). These markers were in high LD (D′ = 1; r²=0.98) to a previously described A3H ‘RED’ haplotype containing a variant (E121) with enhanced susceptibility to HIV-1 Vif (Zhen et al 2009 [1]). This association is confirmed by a haplotype analysis: Homozygote carriers of the A3Hrh had lower GA->AA (A3F) sequence editing on pro-viral HIV-1 vif sequence (p = 0.01), and lower HIV-1 RNA levels over time during early, untreated HIV-1 infection, (p = 0.015 mixed effects model). This effect may be due to enhanced susceptibility of A3H forms to HIV-1 Vif mediated viral suppression of sequence editing activity, slowing viral diversification and escape from immune responses.

1. Introduction

The APOBEC3 locus on chromosome 22 consists of a family of genes coding for cytidine deaminases that cause cytidine (C) to uridine (U) changes in the DNA minus strand sequence at preferred dinucleotide motifs. For example, APOBEC3G (A3G) acts primarily at ‘CC’ in the minus strand DNA (‘GG’ sense strand dinucleotide motif). This induces a change in the first position G, yielding an ‘AG’ in the resulting genomic sense strand DNA sequence [2–4]. By comparison, the APOBEC3F (A3F) member of the APOBEC3 family, recognizes the ‘GA’ plus strand dinucleotide motif and yields an ‘AA’ sense strand sequence.

The HIV-1 Vif protein binds and traffics A3G and A3F for proteolytic degradation, suppressing APOBEC3 editing activity on the HIV-1 genome. Absent Vif, A3G and A3F act to restrict viral replication by targeting the nascent HIV-1 minus strand DNA [5, 6] during pre-integration steps, inducing extensive hypermutation at preferred proviral DNA dinucleotide sites. The A3G and A3F have been heavily studied for association with HIV-1 infection [4, 7], although evidence exists that other members of the A3 family may edit and restrict HIV-1 [8, 9]. Variation in other A3 gene family members may also influence HIV-1 sequence diversification and clinical outcomes such as APOBEC3H [8, 10, 11], which may have an APOBEC3F (‘GA’) like dinucleotide target preference.

We surveyed genetic variation in the human APOBEC3 cassette (A3A, A3B, A3C, A3D(DE), A3F, A3G, A3H) in a cohort of treatment naïve recently HIV-1 infected adults, paired to measures of APOBEC3 sequence editing in the DNA sequence of pro-viral HIV-1 vif. We assayed pro-viral HIV-1 vif DNA in order to survey the reservoir of archived HIV-1 genomes, including those that may be replication incompetent, or replication impaired due to extensive sequence editing by A3G or A3F, or other A3 members [12–14]. We sought to determine if variation in A3H associated with HIV-1 sequence editing, and clinical markers independent of A3G and A3F associations.

2. Materials and Methods

2.1 Study Population

We selected 96 patients for study from the OPTIONS cohort of adults in early HIV-1 infection, which has been described elsewhere [15]. These patients were chosen based on having at least 1 aliquot of viably preserved, frozen peripheral blood mononuclear cells (PBMCs) from a time-point prior to anti-retroviral treatment from which we extracted host DNA. The OPTIONS study was approved by the University of California San Francisco Institutional Review Board, and all patients gave written, informed consent to participate in this study.

2.2 APOBEC3 Single Nucleotide Polymorphism Panel

We assayed a set of APOBEC3 loci single nucleotide polymorphisms (SNP) reported in the HapMap (http://hapmap.ncbi.nlm.nih.gov/) [16] found across the entire APOBEC3 region including APOBEC3A (A3A), APOBEC3B (A3B), APOBEC3C (A3C), APOBEC3DE (A3DE), APOBEC3F (A3F), APOBEC3G (A3G) and APOBEC3H (A3H). We assayed all coding and non-coding SNPs within a 10KB window up and downstream of the first (A3A) and last member (A3H) of the APOBEC3 locus on chromosome 22. SNPs were genotyped using the GoldenGate SNP array assay system (Illumina, San Diego, CA, USA, Methods Supplement 1). The candidate SNPs selected for assaying variation in the APOBEC3 loci were identified via the University of California Santa Cruz Genome Browser (http://genome.ucsc.edu/) based on the HapMap 128 build [16]. We refer to SNPs by their ‘rs’ value in this report [17].

2.2 Viral Sequencing

Details on HIV-1 proviral vif DNA amplification, sequencing and the identification and counting of A3F or A3G signature sequence changes are described in the Methods Supplement.

2.3 Statistical Analysis

2.3.1 Viral sequence statistical analysis

For each potential A3G or A3F dinucleotide motif target, the number of G->A or G->G/A mixture sequence changes were identified, based on A3G and A3F dinucleotide targets mapped within the HXB2 vif reference sequence (http://www.hiv.lanl.gov/content/sequence/HIV/). The A3G or A3F dinucleotide target changes per individual were divided by the number of potential A3G and A3F targets within HXB2 to derive a proportion, which was the principal outcome measure in this study. (Methods Supplement 1 for more details).

2.3.2 Host sequence statistical analysis

We employed a minor allele frequency threshold of 2 % (values below excluded). We adopted a conservative HWE threshold of alpha = 0.05, and further excluded SNPs where greater than 5 % of the population had failed assay values. We tested the association between SNPs and the A3G and A3F signature fractions variables, using an ANOVA test as implemented in PLINK (http://pngu.mgh.harvard.edu/~purcell/plink/) [18]. Similar results were obtained when implemented in R (http://cran.r-project.org/). Linkage disequilibrium analysis and graphical displays were achieved using D′ and r² metrics and GOLD color scheme in Haploview [19]. Haplotype estimation frequencies were achieved using Estihaplo [20] and Fastphase [21]. All other statistical analyses are performed using Stata v10 (College Station, Texas, USA) and R. To relate APOBEC3 positions or haplotypes to HIV-1 clinical markers we employed longitudinal mixed effects modeling, with random effects specified for the individual and time, is SAS System Version 9.1 for Windows XP.

3. RESULTS

3.1 Patient Population Characteristics

The cohort was a median 34.6 years old (Interquartile Range (IQR) 30.6–40.8), 95 % Male, with a median HIV-1 RNA of 4.67 log₁₀ HIV-1 RNA (copies/mL) (IQR 3.7–5.1), and a median CD4+ T cell count of 560 (IQR 472, 705). The subset of patients studied here did not differ from the larger OPTIONS cohort [22, 23].

3.2 HIV-1 Vif Sequence Editing Phenotype Estimation

We recovered complete proviral vif DNA sequence from 94 patients. The fraction of A3F candidate motifs (‘GA’) in HXB2 that appeared as an ‘AA’ or ‘RA’ was a median 0.06 with an interquartile range of 0.04–0.08. The fraction of A3G candidate motifs (‘GG’) in HXB2 that appeared as ‘AG’ or ‘RG’, in patient specimens was a median of 0.095 with an interquartile range of 0.07, 0.12. Both the AA/GA (A3F signature) and GA/GG (A3G signature) ratios were normally distributed (Shapiro-Wilk W test for normal data p=0.5 and p=0.2 respectively). Table 1 displays a representative portion of the HIV-1 vif DNA alignment for all study participants.

Table 1. Characteristics of HIV-1 Vif DNA Sequences (82–158) from Study Subjects in Recent HIV-1 Infection.

Displaying nucleotide positions 82–158 of HIV-1 VIF DNA sequence for HXB2 viral reference strain, plus all 94 subjects for whom sequence was available for this study. Changes in sequence are shown relative to HXB2 reference strain on the top row. APOBEC3G (GG) and APOBEC3F/H (GA) dinucleotide motif positions are shown underlined in the HXB2 strain. Prepared in BioEdit. Full vif sequence for all participants will be deposited in GenBank and other open access databases.

3.3 APOBEC3 Genetics Genotyping Quality Control

In APOBEC3, 167 SNP markers were genotyped in 94 samples. Two individuals were removed due to high genotyping failure rate (>19 % of SNPs failed), with an average overall sample call rate of 93.7%. Thirteen SNPS were excluded due to high failure rate (9 missing in all subjects and 4 that failed in 26 % of subjects). Seventy-seven SNPs with minor allele frequencies below 2 % were removed from the analysis. Nine SNPs with a p-value below 0.05 were excluded by the exact Hardy-Weinberg Equilibrium test. After these quality control steps, 68 SNPs were analyzed, with a successful genotyping rate in the 92 remaining individuals of 99.6%. Lists of all SNPs considered in the association study analysis are in Supplemental Table 1 and 2.

3.4 APOBEC3 Loci Associations with proviral vif DNA Editing Patterns

We found a series of APOBEC3 SNPs that were each significantly associated with mutated fractions for the GA->AA (A3F) signature mutated fraction but not for GG->AG (A3G) signature mutated fraction. Table 2 displays p-values from the test of association of top-ranked SNPs with A3F- and A3G-associated signature mutation patterns. Ten out of 13 significant SNPs (p<0.05) were found within a 25kb span at the telomeric end of the APOBEC3 cluster (rs139339; rs139336; rs139323; rs139317; rs139316; rs139302 (top significant SNP p=4.8 10⁻³); rs139294; rs139283; rs139271; rs2413570). Three SNPs located nearby had p-values close to 0.05 (rs139284; rs139279; rs6519165). Several of these SNPs have been described by OhAinle (rs139302) [8], and Harari et al [24] (rs139294 and rs139302) and Zhen et al. [1] with the rs139302/D178E position. For these, an association signal is replicated by the present study. The 178 position is part of the Harari et al. ‘RDD’/‘GKE’ haplotype, or the Zhen et al ‘RED’ haplotype. These relationships are detailed in Figure 1.

Table 2. APOEBC3H SNP Associations with APOBEC3F/H (GA->AA) and APOBEC3G (GG->AG) signature mutated fraction.

APOBEC3H region SNPs associated with A3F/H or A3G mutated fraction, ranked by p-value of the association to the A3F/H mutated fraction.

SNP number	SNP	Sample Size	APOBEC3F/H Mutated Fraction (GA->AA)		APOBEC3G Mutated Fraction (GG->AG)
			Uncorrected P-value	Rank	Uncorrected P-value	Rank
23	rs139339	92	0.01006	2*	0.1168	4
22	rs139336	92	0.01006	2*	0.1168	4
21	rs139323	92	0.01006	2*	0.1168	4
20	rs139317	92	0.01177	6*	0.1648	11
19	rs139316	92	0.01006	2*	0.1168	4
18	rs139314	92	0.2082	33	0.2698	22
17	rs139302	87	0.004802	1**	0.2935	27
16	rs139296	91	0.2108	34	0.2563	20
15	rs139294	92	0.01401	7*	0.1105	3
14	rs139285	92	0.09264	22	0.201	16
13	rs139284	92	0.05855	15	0.257	21
12	rs139283	92	0.01616	8*	0.1564	10
11	rs139279	92	0.06322	17	0.3194	28
10	rs139271	92	0.02301	10*	0.1989	15
9	rs11705514	92	0.944	66	0.9954	68
8	rs5995668	92	0.4026	43	0.4089	31
7	rs6001423	92	0.399	42	0.5164	40
6	rs5757472	92	0.1802	30	0.2458	19
5	rs17537581	92	0.247	35	0.4431	33
4	rs2413570	92	0.01828	9*	0.1932	14
3	rs5757465	92	0.1747	29	0.07682	1
2	rs6519165	92	0.0542	14	0.1691	12
1	rs4820365	92	0.08952	20	0.5384	42

Figure 1. Relationship of A3H Core Risk Haplotype (CTGCT) to Previously Described SNPs within APOBEC3H.

The figure displays A3H genetic patterns in the present study, prior studies (Harari et al.), and HapMap, at the DNA sequence level, mRNA codon level, protein amino-acid level. Linkage studies within HapMap reveal that the ‘CTGCT’ A3H core haplotype defined in the current study is in strong linkage to a ‘RED’ haplotype pattern within A3H.

3.5 Haplotype Estimation

In order to refine the observed association, we estimated haplotype frequencies in our dataset using the most associated 10 markers listed above, plus an additional 3 markers found in the telomeric region. Working from this set of SNPs, we identified a risk haplotype within the A3H region of 5 non-coding SNPS (rs139339*C; rs139336*T; rs139323*G; rs139317*C; rs139316*T, also referred to here as ‘CTGCT’ or the A3H “core” risk haplotype (A3Hrh), for convenience). The frequency of this haplotype was estimated to be 35.3 % in our cohort.

3.6 Haplotype Analyses

When a haplotype association test was run after phase inference with the A3Hrh 5 core SNPs (CTGCT), the association remain significant (ANOVA, p=0.01), and suggested a recessive model (p=0.05). Four of five members of the A3Hrh were found in A3H haplotypes found in HapMap (rs139336; rs139323; rs139317; rs139316). We examined our A3H core risk haplotype (4 of 5 members, rs139316/C; rs139317T; rs139323G; rs139336C) for association with previously reported SNPs (rs139302, rs139298, rs139297, and rs139294 [8, 10, 11]). We found a 4 member version (CTGC) of our A3H ‘CTGCT’ core haplotype was in complete linkage to rs139302C, rs139298G, rs139297C and rs139294C, with a D′ of 1.0 in all cases within HapMap. This indicates our core haplotype recovered an extended A3H haplotype coding for previously described A3H positions R105, E121 and D178 (“RED”) [1, 8, 10, 11].

3.7 SNP Typing Methods, HapMap Codon Inference and the A3H 121 Position

The A3H 121 codon, present as the second position our A3H risk haplotype (A3H 105/121/178) contains two SNPs, at positions 1 and 3 of the codon. Each of these nucleotides are coding in position 121 (E, K, N or D are possible). However, closely spaced SNPs cannot be simultaneously detected in modern high-throughput SNP assays due to high overlap (loss of specificity) in the binding of probe or primer to the target DNA region. Hence, HapMap only tracks the first position and not the third. We were able to link our A3H risk haplotype to only one SNP within codon 121, position 1 (rs139298) but not codon position 3 (rs139299, not present in HapMap). Based on available information, we inferred a G (GAG) coding for E, or Glutamine. Alternatively, the K121 is coded by AAG (K or Lysine), whereas an N121 (N or Asparagine) is coded by an AAC. Hence, it remains possible a GAC and not GAG were present in some study subjects. This change would lead to expression of D or Aspartic Acid that could yield an RDD haplotype and not the RED haplotype in some subjects.

3.8 Clinical Associations with APOBEC3H Risk Haplotype

Homozygote carriers of the A3H risk haplotype had lower A3F associated mutational activity, than heterozygote of homozygote nulls (Figure 2). In a longitudinal mixed effects models, homozygote carriers of the A3H risk haplotype ‘CTGCT’ had lower HIV-1 RNA levels during 1 year or more of early untreated HIV-1 infection, compared to heterozygote carriers (+0.17 log₁₀ copies/mL, p = 0.015) or homozygote nulls (+0.20 log₁₀ copies/mL, p = 0.018). We did not observe an association of the A3H risk haplotype with CD4+ T cell counts over time.

Figure 2. APOBEC3H Risk Haplotype and APOBEC3F (GA->AA) Signature Mutation Fraction (Proportion of Motifs Converted Relative to HXB2).

A haplotype association test run after phase inference with the 5 core SNPs reveals a significant association is found between carriage of the core ‘RED’ haplotype and A3F signature upon HIV-1 Vif (ANOVA, p=0.01). The results suggest a recessive model (p= 0.05446).

3.9 Confirmation of Prior Findings

Our aim was confirmation and extension of previous findings on the existence of a strong LD block within A3H that associates with sequence variation in HIV-1 vif DNA [1, 8, 10, 11]. Therefore, we did not perform multiple comparison adjustments.

4. Discussion

We surveyed genetic variation in the APOBEC3 gene family locus with a panel of single nucleotide polymorphisms (SNPs), and paired this data to the sequence of proviral HIV-1 vif from peripheral blood mononuclear cells of recently HIV-1, anti-retroviral naïve, infected adults. We observed significant associations between a set of SNPs in the APOBEC3H region, and a resultant A3H risk haplotype, to reduced GA->AA sequence editing pattern typically associated with A3F activity [10], and to lower HIV-1 RNA levels over time.

We found our A3H risk haplotype was in strong linkage disequilibrium with several coding positions within A3H (105/121/178). The second of these, A3H 121, falls in a region of A3H with high structural homology to the closely related A3G protein. This loop region of A3G is bound by the HIV-1 Vif protein that routes A3G for proteolytic degradation and hence suppression of activity. A3H proteins bearing an E121 variant in this region are susceptible to HIV-1 Vif mediated suppression [1] despite demonstrating robust anti-retroviral activity in vitro. In contrast, the A3H D121 and K121 variants show reduced susceptibility to HIV-1 Vif in vitro, and would be predicted to be expressed and retain anti-HIV-1 sequence editing activity during infection in vivo. In our analysis we found evidence that our A3H risk haplotype is in strong linkage with the previously described ‘RED’ haplotype (R105/E121/D178) [10][1]. Hence, the association of our risk haplotype with reduced APOBEC3 signature upon HIV-1 DNA sequence may be due to the carriage of the A3H E121 variant, rendering that A3H protein susceptible to Vif targeting and proteolytic degradation, blocking A3H activity against HIV-1. That said, it is possible that our A3H haplotype does not contain an E at position 121 in all carriers in this study, with a D (or Aspartic acid) carried instead. Few human A3H genes have been re-sequenced in persons with HIV-1 infection, and the true frequency of an E121, K121, D121 and N121, and their linkages, are not yet known. In future studies, we plan to re-sequence A3H and other A3 members in cohorts of HIV-1 infected to determine frequencies of these distinct genetic forms.

We observed that homozygote carriers of the A3H risk haplotype, linked to the previously described RED or RDD haplotypes, appear to have lower A3F signature (GA->AA) nucleotide sequence editing activity, and also lower HIV-1 RNA levels over time early, untreated HIV-1 infection. Lower A3H activity against HIV-1 may slow viral sequence diversification and escape from immune responses, leading to the lowered viral loads observed here. The A3H risk haplotype may bear the E amino acid at A3H position 121 that has been recently reported to confer susceptibility of A3H proteins to targeting by HIV-1 Vif [1]. Further study will be required to determine the true frequency of the E121 variant in humans, and the expression level of an E121 bearing variant in the CD4+ T-cells which support HIV-1 replication. That A3H, or some of its variants, may cause a detectable mutational signature on HIV-1 genomes in vivo suggests A3H may play an independent role in HIV-1 sequence diversification and evolution [25].

Supplementary Material

01

Table S1 summarizes all original candidate SNPs

02

Table S2 summarizes associates of QC’d SNP set with A3F and A3G signature fractions.