The impact of Bone Marrow Stromal Antigen-2 (BST2) Gene Variants on HIV-1 Control in black South African Individuals

Abstract

Bone marrow stromal cell antigen 2 (BST2 or tetherin) is a host-encoded, interferon-inducible antiviral restriction factor which blocks the release of enveloped viruses. Few studies have assessed the role of BST2 polymorphisms on HIV-1 acquisition or disease progression in sub-Saharan Africa. This study investigated the frequency of four HIV-1-associated BST2 variants rs3217318, rs12609479, rs10415893 and rs113189798 in uninfected and HIV-1 infected black South Africans. Homozygosity for the rs12609479-A minor allele, previously associated with decreased HIV-1 acquisition risk, was underrepresented in HIV-1 uninfected black South Africans (2%) compared to reference African (9%) and in particular European populations (61%) (p=0.047 and p<0.0001, respectively). To determine if any of these gene variants influenced HIV-1 control in the absence of antiretroviral treatment (ART), we compared HIV-1 infected ART-naïve progressors [n=72] and controllers [n=71], the latter includes elite controllers [EC: n=23; VL<50 RNA copies/ml]. Heterozygosity for the rs12609479 SNP (G/A) was enriched in progressors compared to ECs (47.2% vs 21.7%, OR=3.50 [1.16–10.59], p=0.03), while rs113189798 heterozygosity (A/G) showed a strong trend of overrepresentation in ECs compared to progressors (47.8% vs 26.4%, OR=0.39 [0.14–1.04], p=0.07). Heterozygosity for the promoter indel rs3217318 (i19/Δ19) was associated with a faster rate of CD4⁺ T-cell decline in progressors (p=0.0134). Carriage of the rs3217318 (i19/Δ19), rs12609479 (G/G), rs10415893(G/A) and rs113189798 (A/G) combined genotype, denoted as i19Δ19 GG GA AG, was associated with significantly higher CD4⁺ T-cell counts in progressors (p=0.03), a finding predominantly driven by the _GG_AG combination. Our data suggest that the possession of select BST2 genotype combinations may be implicated in HIV-1 disease progression and natural spontaneous control.

1. Introduction

Host cells respond to viral infections by mounting immune responses, most notably the type-1-interferon (IFN) (α/β) response which ultimately results in the induction of antiviral proteins and potent cytokines (Haller et al., 2006). One such IFN-induced antiviral protein is the bone marrow stromal cell antigen 2 (BST2). BST2 (also known as tetherin, CD317 and HM1.24), is a host restriction factor which has been reported to efficiently block the release of a multitude of enveloped viruses including retroviruses (Arnaud et al., 2010; Dietrich et al., 2011), filoviruses, arenaviruses, paramyxoviruses (Jouvenet et al., 2009; Kaletsky et al., 2009; Radoshitzky et al., 2010; Sakuma et al., 2009), gamma-herpesviruses (Mansouri et al., 2009; Pardieu et al., 2010) and rhabdoviruses (Weidner et al., 2010). BST2 is a type II transmembrane protein with an unique “two-anchor” topology as it exhibits a transmembrane domain (TMD) towards the N-terminus and a C-terminal glycophosphatidylinositol (GPI) anchor, thereby allowing for the effective “tethering” of virions to the host cell plasma membrane.

Discoveries with regard to the structure and function of BST2 have largely emanated from studies investigating this protein’s role in the suppression of HIV-1. Not only does BST2 impede the release and dissemination of HIV-1 particles but also reportedly restricts productive cell-to-cell transmission (Casartelli et al., 2010) and inhibits viral replication (Barrett et al., 2012; Liberatore and Bieniasz, 2011). BST2 has also been shown to facilitate HIV-1 infected cell clearance by promoting higher levels of HIV-1 antigen expression, enhancing antibody opsonization and consequently promoting higher levels of natural killer (NK) cell mediated antibody-dependent cellular cytotoxicity (ADCC)(Alvarez et al., 2014; Arias et al., 2012).

To counter host defense, HIV-1 encodes the accessory protein Viral protein U (Vpu). Vpu antagonizes the restriction imposed by BST2 by interfering with the anterograde transport of de novo BST2 proteins to the cell membrane, inducing BST2 internalization from the plasma membrane, promoting the sequestration of BST2 in the endosomal and trans-golgi network (TGN) and ultimately targeting this host restriction factor for proteasomal degradation (Arias et al., 2012). Through these countermeasures Vpu restores the dissemination of progeny virions and successful HIV-1 pathogenesis.

Host variability in susceptibility to HIV-1 infection (Kaslow et al., 2005), mother-to-child transmission (Singh and Spector, 2009), viral load, the rate of disease progression as well as intra-host viral evolution (Telenti and Johnson, 2012) are all influenced by host genetics. Despite sub-Saharan Africa being the most affected region globally, studies examining genetic risk factors associated with HIV-1 have largely been conducted in European populations. A better understanding of the population most affected by the HIV epidemic, including the South African black population, is therefore of paramount importance given that future vaccine and cure therapies will be administered in such populations.

A number of BST2 gene variants have been associated with either the risk of HIV-1 acquisition (Hancock et al., 2015; Singh et al., 2018) or disease progression (Laplana et al., 2013). Hancock et al., assessed 23 single nucleotide polymorphisms (SNPs) across the BST2 coding region and 50 kb flanking regions in Americans of both African and European ancestries who inject drugs and identified two SNPs, rs113189798 (A>G) and rs12609479 (G>A), associated with a risk of acquiring HIV-1 (Hancock et al., 2015). The latter SNP (rs12609479) was previously described in a study investigating genetic variability in 581 individuals across four sub-Saharan African countries (Skelton et al., 2014). Although this study described BST2 promoter polymorphisms in African populations (Skelton et al., 2014), associations with HIV-1 acquisition and disease progression were not investigated. Lapalana et al., described two BST2 gene variants associated with HIV-1 disease progression in a population of Caucasian Spanish injecting drug users, namely a 19 base pair (bp) insertion/deletion (indel) polymorphism (rs3217318) in the promoter, and an G>A (rs10415893) SNP in the 3’ untranslated region (UTR), which were shown to be in strong linkage disequilibrium (Laplana et al., 2013). Singh et al., similarly found that the 19 bp insertion (rs3217318) was associated with an increased risk of acquiring HIV-1 (Singh et al., 2018).

The prevalence of three of the aforementioned BST2 genetic variants (rs3217318, rs10415893 and rs113189798) are not known for our black South African population. Furthermore, few studies have assessed the role of BST2 polymorphisms in HIV-1 acquisition or disease progression, particularly in sub-Saharan Africa. The aim of this study was therefore to firstly determine the frequency of the four BST2 variants (rs3217318, rs12609479, rs10415893 and rs113189798) in the background population (HIV-1 uninfected), and to secondly to investigate whether these gene variants are associated with spontaneous control of viraemia in the absence of anti-retroviral therapy (ART), in a cohort of HIV-1 infected black South Africans.

2. Materials and Methods

2.1. Study participants

The study participants comprised of 96 HIV-1 uninfected adult volunteers [HCs: n=96] derived from the larger ESKOM (Electricity Supply Commission) cohort which represents a cross-section of the ethnic sub-groups present in the black South African population (Paximadis et al., 2012). Patient race and ethnicity was self-reported.

In addition, 143 HIV-1 infected black (self-reported) South African individuals were categorized into two groups based on their inherent ability to control HIV-1 infection in the absence of ART: HIV-1 infected progressors [n=72, viral load (VL) >10 000 RNA copies/ml, CD4⁺ T-lymphocyte count <300 cells/μl and requiring ART initiation] and HIV-1 infected controllers [n=71]. The latter group was comprised of elite controllers [ECs: n=23, VL<50 RNA copies/ml]; viraemic controllers [VCs: n=37, VL<2000 RNA copies/ml] and high viral load long-term non-progressors [HVL LTNPs: n=11, VL>10 000 RNA copies/ml, CD4⁺ T-lymphocyte count >500 cells/μl sustained for >7 years].

The HIV-1 infected progressors were recruited from Chris Hani Baragwanath Hospital (Johannesburg, South Africa) and were a subset of a larger cohort prospectively recruited to compare three treatment regimens to prevent tuberculosis (TB), as previously described (Martinson et al., 2011).

The HIV-1 infected controllers were recruited from four sites in Johannesburg, namely the Chris Hani Baragwanath Hospital, Helen Joseph Hospital, Life Brenthurst Hospital and Toga Laboratories. The characteristics of the study participants are shown in Table 1.

Table 1.

Demographic characteristics of HIV-1 infected South African study participants

Phenotypic Groups*	Total number of participants (n)	Age (years) (Median, Range)	Gender (% females)	CD4+ T- lymphocyte counts (cells/μl) (Median, IQRs^)	Viral Load (HIV-1 RNA copies/ml) (Median, IQRs)	Time since HIV-1 diagnosis (years) (Median, IQRs)
HIV-1 infected controllers	71	44 (21–63)	85.9	663 [565–854]	405 [25–1385]	NA
HIV-1 infected controller sub-groups:
Elite controllers (ECs)	23	46 (27–63)	78.3	672 [548–848]	<20 [<20]	15 [8–16]
Viraemic controllers (VCs)	37	44 (21–60)	91.9	700 [559–883]	718 [320–1 180]	8 [8–15]
High viral load long term non-progressors (HVL LTNPs)	11	49 (39–59)	81.8	660 [606–749]	22 410 [14 263–77 820]	8 [8–11]
HIV-1 infected progressors	72	48(33–76)	83.3	178[148–211]	38 444[19 852–95 816]	6 [1–7]

^*Clinical phenotypic groups were classified based on natural abilities to control HIV-1 infection/pathogenesis

^{^}IQR= Interquartile range

HIV-1 RNA levels (RNA copies/ml of plasma) were quantified using the COBAS® AmpliPrep/COBAS® Taqman® HIV-1 test, v2.0 (Roche Diagnostic Systems, Inc., New Jersey, USA) and the FACSCount System (Becton Dickinson, San Jose, California, USA) was used to determine CD4⁺ T-lymphocyte counts.

Ethical clearance for this study was obtained from the University of the Witwatersrand’s Human Research Ethics Committee (Institutional Review Board approval number: M140926) and informed written consent was obtained from all participants.

2.2. PCR and sequencing of BST2

Genomic DNA was extracted from ethylenediaminetetraacetic acid (EDTA)-anticoagulated blood samples using QIAamp DNA Mini Kit (QIAGEN, Dusseldorf, Germany). Nucleotide sequencing was performed on a selection of HIV-1 uninfected individuals to detect respective genotypes of polymorphisms selected to study, namely rs3217318, rs12609479, rs10415893 and rs113189798. Individuals with known genotypes served as controls for the development of allele-specific real-time PCR and indel genotyping assays (described below). The regions harbouring each SNP were amplified separately using the Expand High Fidelity PCR System (Roche, Manneheim, Germany). In brief, each PCR reaction consisted of: ~20 ng genomic DNA as template, 1.4 U Expand High Fidelity enzyme mix, 2.5 μl of 10x PCR buffer (1.5 mM MgCl₂), 200 μM of each deoxynucleotide, 30 μM of each oligonucleotide primer (Table S1), and molecular grade water to a final volume of 25 μl. PCR cycling conditions were the standard conditions for the Expand High Fidelity PCR System (see manufacturer’s instructions). PCR purification was performed with Agencourt AMPure XP magnetic beads (Beckman Coulter, Brea, California, USA) as per the manufacturer’s instructions.

Sequencing of PCR amplicons was conducted using BigDye Terminator version 3.1 chemistry (Applied Biosystems, Foster City, CA, USA) and all amplicons were sequenced bi-directionally using the automated 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA) and analysed on Sequencher™ version 4.5 (Gene Codes Corporation, Ann Arbor, MI).

2.3. Genotyping of BST2 polymorphisms

2.3.1. Size discrimination agarose gel electrophoresis

To genotype the study population for the 19 bp indel (rs3217318), the PCR amplicons were resolved by 2% super fine resolution (SFR ™) agarose gel electrophoresis (Amresco, Solon, Ohio, USA) at 85V, 4°C. The 19 bp difference in amplicon size was clearly discernible (Fig S1) and the results were consistent with the select individuals that were sequenced.

2.3.2. Development of Real-time PCR assays for SNP genotyping

Select individuals harbouring various genotypes identified through sequencing were used as controls for the development of allele-specific real-time SYBR® green PCR assays (described previously (Paximadis et al., 2013)) to genotype the three SNPs (rs12609479, rs10415893 and rs113189798) in the study population. Briefly, one common forward primer and two allele-specific reverse primers (Table S1) were designed for variant discrimination. Two reactions were set-up for each sample, one with the reverse primer displaying 3’ complementarity to the wild-type (major) allele and the other complementary to the mutant (minor) allele. Each reaction consisted of: ~20 ng DNA, 5 μl of 2x FastStart Universal SYBR green master (Rox), 10 μM of each oligonucleotide primer and molecular grade water to a final volume of 10 μl. PCR reactions were run in an Applied Biosystems 7500 PCR system and used to determine the cycle threshold (Ct) values. Differences in the Ct values were used to genotype samples with variances of ≥ 4 Ct being considered homozygous for the major or minor allele depending on the primer-set resulting in more efficient amplification. In heterozygous individuals, both primer sets would bind with equal affinity and result in comparable amplification efficiency (Fig S1).

2.4. Hardy-Weinberg Equilibrium

All polymorphic loci were tested for deviations from Hardy-Weinberg (HW) equilibrium using the exact tests of Wigginton, Cutler and Abecasis (Wigginton et al., 2005), implemented through the computer programme Haploview v4.2 (Barrett et al., 2005).

2.5. Linkage disequilibrium between BST2 polymorphisms

Haploview v4.2 (Barrett et al., 2005) was implemented to identify potential haplotypes within the data set and to compute pairwise linkage disequilibrium (LD) using expectation-maximization algorithms (Barrett et al., 2005), as quantified by the LD coefficient r², between every two minor allele combination. The study participants were classified into three groups namely: HIV-1 uninfected controls, HIV-1 infected controllers and HIV-1 infected progressors, and were analyzed as a total group and independently.

2.6. Statistical Analyses

Fisher’s exact probability tests were performed to calculate the statistical significance (two-tailed) and 95% confidence interval for the odds ratios (ORs) for the allelic and genotypic frequency differences.

The impact of the individual genotypes as well as select combinations of genotypes on viral load (RNA copies/ml of plasma) and CD4⁺ T-lymphocyte count (cells/μl of whole blood) were assessed by non-parametric Mann-Whitney U-test for non-normally distributed data. CD4⁺ T-lymphocyte decline (cells/μl/month) was calculated, for HIV-1 infected progressors, by dividing the CD4⁺ T-lymphocyte cell count difference between enrolment and the last visit before initiation of ART by the number of months within this period. HIV-1 infected progressors with a single sample collection prior to ART initiation were excluded from these analyses (n=20). In addition, since active TB has been reported to accelerate the rate of CD4⁺ T-lymphocyte decline (Martinson et al., 2014), patients with this co-morbidity (n=3) were excluded from these analyses. Thus total number of individuals assessed was n=49.

All statistical measures were considered significant at p<0.05. Due to the exploratory nature of this research and the use of extreme phenotypes, the aim of which was to identify genetic factors potentially associated with natural HIV-1 control, adjustments for multiple comparisons were not performed in this study as these increase the risk of type II errors (failure to detect an effect that is indeed present).

3. Results

3.1. Hardy-Weinberg Equilibrium (HWE) and Linkage disequilibrium (LD) of BST2 polymorphisms

A schematic representation of the BST2 gene and the position of the variants studied here are shown in Fig S1. All the variants under study conformed to Hardy-Weinberg equilibrium. Linkage disequilibrium between the minor alleles was investigated and revealed that only the 5’ UTR indel rs3217318 and 3’ UTR SNP rs10415893 minor alleles displayed moderate linkage disequilibrium (r²= 0.49) in our population when combining all participants (Fig S2A) as well as when the three study subsets (HIV-1 uninfected, HIV-1 infected controllers, HIV-1 infected progressors) were analysed separately (r²= 0.39, r²= 0.60 and r²= 0.49, respectively) (Fig S2B), in keeping with data previously reported by Laplana (Laplana et al., 2013).

3.2. Comparison of allelic and genotypic frequencies of BST2 polymorphisms between HIV-1 uninfected black South Africans and reference populations

Genotypic data for the 19 bp indel rs3217318 was not available from the 1000 Genomes Project, thereby precluding comparisons between our study population and those of European and African reference populations (Auton et al., 2015).

Comparison of the 5’ UTR SNP rs12609479 genotypes (Fig 1A) revealed that both homozygosity for the major allele (G/G) and homozygosity for the minor allele (A/A) differed significantly between African and European populations. Interestingly, minor allele homozygosity (A/A) was significantly underrepresented in our black South African population compared to reference African populations (2.08% vs 8.77%, p=0.047) as well as populations of European descent (61%) (p<0.001) (Fig 1B). This is the only genotype in which the frequencies observed in our population differed significantly from the reference total African population (which includes data from African populations in Kenya, Nigeria, The Gambia, Sierra Leone, America and Barbados)(Auton et al., 2015). It is interesting to note however that minor allele homozygosity for this SNP is considerably variable within the African populations with The Gambia also having low representation (1.8%) and populations from Nigeria having high representation (10–12%).

Fig 1.

Genotypic frequencies of BST2 polymorphisms in the HIV-1 uninfected black South African and reference populations. The distribution of genotypic frequencies are shown for the 5’UTR SNP rs12609479 (A) in the European and African reference populations and the HIV-1 uninfected black South African population (n=96). (B) Homozygosity for the rs12609479-A minor allele is the only genotype to differ significantly between the South African and African reference populations. The genotype distributions of 3’UTR polymorphic variants rs10415893 (C) and rs113189798 (D) are similarly shown. Reference Populations: total European and total African data from the 1000 Genomes Project.

The rs10415893 (G/A) and rs113189798(A/G) heterozygous genotype frequencies of the HIV-1 uninfected black South Africa population largely mirror those of the African reference population whilst being significantly higher (p<0.0001) than those frequencies detected in the reference European population (Fig 1C and D).

3.3. BST2 polymorphisms in relation to HIV-1 control and disease progression

The allelic and genotypic frequencies of the four BST2 variants are shown in Table S2. Minor allele and genotype comparisons across the study groups are shown in Table 2.

Table 2.

Comparison of HIV-1 infected black South African controllers and progressors with regards to rs3217318, rs12609479, rs10415893 and rs113189798 allele and genotype frequencies

Variant*	Minor Allele/Genotype	Progressors vs Controllers			Progressors vs ECs			Progressors vs VCs			Progressors vs HVL LTN
		OR	CI	p	OR	CI	p	OR	CI	p	OR	CI	p
	i19 allele	0.91	0.56–1.51	0.80	0.71	0.36–1.41	0.37	1.15	0.62–2.12	0.76	0.80	0.31–2.03	0.81
−318	Δ19/i19	0.86	0.43–1.72	0.72	0.75	0.27–2.08	0.61	1.05	0.45–2.42	1.00	0.55	0.14–2.15	0.50
	i19/i19	0.91	0.31–2.72	1.00	0.51	0.13–2.10	0.44	1.45	0.34–6.11	0.74	0.91	0.09–9.32	1.00
	A allele	1.16	0.70–1.94	0.60	2.45	1.01–5.91	0.055	0.81	0.45–1.47	0.54	1.17	0.43–3.2	0.81
−479	G/A	1.79	0.88–3.61	0.11	3.50	1.16–10.59	0.03	1.25	0.53–2.94	0.66	1.55	0.40–5.98	0.74
	A/A	0.76	0.23–2.54	0.77	2.58	0.28–23.84	0.65	0.43	0.11–1.61	0.30	0.91	0.09–9.22	1.00
	A allele	0.85	0.51–1.43	0.60	0.82	0.40–1.70	0.70	0.84	0.46–1.58	0.63	0.77	0.29–2.03	0.61
−893	G/A	0.62	0.31–1.25	0.22	0.88	0.31–2.45	1.00	0.55	0.24–1.27	0.20	0.59	0.15–2.23	0.74
	A/A	1.16	0.34–3.99	1.00	0.68	0.15–3.05	0.69	1.45	0.27–7.71	1.00	0.85	0.09–8.45	1.00
	G allele	0.98	0.54–1.78	1.00	0.59	0.27–1.26	0.21	1.90	0.82–4.43	0.17	0.62	0.22–1.72	0.39
−798	A/G	0.87	0.41–1.82	0.85	0.39	0.14–1.04	0.07	1.94	0.70–5.40	0.24	0.58	0.15–2.29	0.47
	G/G	1.28	0.27–6.02	1.00	0.90	0.09–8.84	1.00	2.45	0.26–22.96	0.65	0.49	0.05–5.13	0.47

Progressors (n=72); ECs: Elite controllers (n=23); VCs: Viraemic controllers (n=37); HVL LTNPs: High viral load long term non-progressors (n=11).

^*SNP ID Numbers provided are the final three digits of the reference SNP ID number in table heading.

The rs3217318 and rs10415893 allelic and genotypic frequencies did not differ significantly between the study subgroups (Table 2). The rs12609479-A minor allele was the only allele that showed a trend of overrepresentation in progressors compared to ECs (30.6% vs 15.2%, OR=2.45 [1.01–5.91], p=0.055). Correspondingly, at the genotypic level, rs12609479 heterozygosity (G/A) was significantly enriched in progressors compared to ECs (47.2% vs 21.7%, OR=3.50 [1.16–10.59], p=0.03). In contrast, heterozygosity for rs113189798 (A/G), showed a trend of underrepresentation in progressors compared to ECs (26.4% vs 47.8%, OR= 0.39 [0.14–1.04], p=0.07).

When analysed individually, none of the BST2 variants notably influenced baseline CD4⁺ T-lymphocyte count in HIV-1 progressors (Fig 2A). Homozygosity for the rs113189798-G minor allele was associated with lower HIV-1 viral loads in progressors compared to those wildtype for the major allele and heterozygotes (Fig 2B), however this genotype was present at very low frequency in the black South African population (Table S2) and this finding would need to be validated in a larger cohort.

Fig 2.

Influence of BST2 variant genotypes on markers of disease progression in HIV-1 infected South African black individuals with advanced disease. The impact of BST2 polymorphic variants, rs3217318, rs12609479, rs10415893 and rs113189798, on CD4⁺ T-cell count (A) and viral load (B) and rate of CD4⁺ T-lymphocyte decline (C) in HIV-1 infected progressors. The effect of select rs3217318 and rs10415893 genotype combinations on the rate of CD4⁺ T-cell decline are also shown. Box-and-whisker plots depict median (bold horizontal line), 25^th and 75^th centiles (box margins) and 5% and 95% centiles (whiskers). Outliers are depicted as (●). p-values were calculated using a two-tailed non-parametric Mann-Whitney U test.

The rate of CD4⁺ T-lymphocyte decline, a measure of HIV-1 disease progression, was significantly faster (p=0.0134) (Fig 2C) in HIV-1 infected progressors who were heterozygous for the indel (Δ19/i19) compared to those who were wildtype homozygous (Δ19/Δ19). When analysed in the dominant mode, it was shown that both the rs3217318-i19 minor allele (Δ19/i19+i19/19, p=0.0131) and rs10415893-A minor allele (GA+AA, p=0.0413) were significantly associated with faster rates of CD4⁺ T-cell decline (data not shown). Owing to the linkage between rs3217318 and rs10415893, we additionally evaluated their combined effect and our data showed that the rate of CD4⁺ T-cell decline in individuals heterozygous at both loci (Δ19i19_GA_) was significantly faster in comparison to progressors carrying the wildtype genotypes (Δ19 Δ 19_GG_) (p=0.0114, Fig 2C).

3.4. The effect of combined BST2 genotypes in HIV-1 control

We next questioned the effect of combined genotypes on HIV-1 control.

The frequencies of predominant genotype combinations are shown in Table S3 and a trend of overrepresentation of rs12609479 heterogygosity (G/A) in a wild type background (Δ19Δ19GAGGAA) was observed in progressors compared to controllers (15.3% vs 5.6%, OR=4.95 [1.0–24.1], p=0.07) (Table 3).

Table 3.

Comparison of predominant genotype combinations of four BST2 SNPs in HIV-1 infected sub-groups

Combined BST2 Genotypes*	Progressors vs Controllers			Progressors vs ECs			Progressors vs VCs			Progressors vs HVL LTNPs
	OR	CI	p	OR	CI	p	OR	CI	p	OR	CI	p
Δ19Δ19 GA GG AA	4.95	1.0–24.1	0.07	4.4	0.3–60.6	0.52	3.7	0.62–21.7	0.20	∞	NaN-∞	0.14
Δ19Δ19 AA GG AA	0.90	0.18–4.56	1.00	1.6	0.10–24.7	1.00	0.67	0.11–3.92	1.00	1.6	0.10–24.7	1.00
Δ19i19 GG GA AA	2.88	0.60–13.7	0.26	∞	NaN-∞	0.20	2.00	0.36–11.2	0.67	3.20	0.23–45.2	0.55
Δ19i19 GG GA AG	0.9	0.18–4.56	1.00	0.4	0.05–3.42	0.61	1.33	0.19–9.31	1.00	1.6	0.10–24.7	1.00
Δ19i19 GA GA AA	1.15	0.27–4.87	1.00	2.8	0.20–40.1	0.57	0.88	0.18–4.34	1.00	1.4	0.14–13.6	1.00

Progressors (n=72); ECs: Elite controllers (n=23); VCs: Viraemic controllers (n=37); HVL LTNPs: High viral load long term non-progressors (n=11).

^*BST2 SNP genotypes are arranged in the following order: rs3217318, rs12609479, rs10415893 and rs113189798

Interestingly, homozygosity for the rs12609479 major allele (G/G) in combination with heterozygosity at the other three loci (Δ19i19GGGAAG) was more prevalent in ECs compared to progressors (17.4% vs 5.6%, Table S3), although this finding failed to reach statistical significance (Table 3). However, carriage of this four-genotype combination was significantly associated with higher CD4⁺ T-cell counts in individuals with progressive disease compared to those harbouring all other combinations (p=0.03, Fig 3).

Fig 3.

The influence of combined BST2 genotypes on CD4⁺ T-lymphocyte count HIV-1 infected progressors. Comparison of CD4⁺ T-cell count (cells/μl) in HIV-1 infected progressors carrying select BST2 genotype combinations comprising of rs3217318, rs12609479, rs10415893, rs113189798 genotypes (in sequence). Δ19/i19_GA_ denotes combinations in which heterozygosity for both rs3217318 and rs10415893 polymorphisms is requisite and the genotypes at the other two loci are unrestricted. Similarly, _GG_AG, denotes genotype combinations in which homozygosity for the major rs12609479-G allele and heterozygosity at the rs113189798 locus is mandatory with no restrictions placed at the rs3217318 and rs10415893 loci. Box-and-whisker plots depict median (bold horizontal line) and 5% and 95% centiles (whiskers). Outliers are depicted as (●). p-values were calculated using a two-tailed non-parametric Mann-Whitney U test.

We next stratified this “advantageous 4-variant combination” into 2-variant combinations. Since rs3217318 and rs10415893 showed moderate LD, these genotypes were considered together as Δ19i19_GA_ and the remaining two variants (rs12609479 and rs11389798) were analysed as a pair (_GG_AG). The CD4⁺ T-cell counts of progressors carrying Δ19i19_GA_ did not differ from those not carrying this combination (Fig 3), however it was revealed that progressors carrying the _GG_AG combination had significantly higher CD4⁺ T-cell counts compared those not carrying this combination (p=0.047, Fig 3).

This suggests that the protective effect of the 4-variant combination (p=0.03, Fig 3), is primarily driven by the rs12609479 (GG) and rs11318798 (AG) combination (_GG_AG) (p=0.047). It is interesting to note that ECs showed a strong trend of higher representation of the _GG_AG compared to progressors (39.1% vs 11.1%, OR=0.27 [0.07–0.97], p=0.06) (Tables S4 and S5)

3.5. In silico analysis of the effect of BST2 3’UTR SNPs on putative miRNA binding sites

We conducted in silico analysis to examine the possible effect of the 3’UTR variants (rs10415893 and rs113189798) on putative miRNA binding sites. We elected to use the online resource, miRDB (Wong and Wang, 2015), as this tool allows for user input of the target mRNA sequence, thereby allowing for the impact of single-nucleotide changes to be evaluated. Investigation into the effect of these polymorphisms revealed that the rs10415893-A minor allele results in the putative establishment of a binding site for an additional miRNA hsa_miR4652–5p (Table S6). The rs113189798-G minor allele resulted in the establishment of two additional miRNA binding sites, hsa_miR-6827–3p and has_miR-340–3p, compared to the sequence harbouring the major allele (Table S6). Although the software used is merely predictive in nature, and experimental validation of the predicted miRNAs is lacking, these data support the hypothesis that the BST2 3’ UTR SNPs may affect miRNA-binding sites and consequently impact gene regulation.

4. Discussion

BST2 serves as a potent HIV-1 restriction factor which has been shown not only to effectively block the egress of HIV-1 progeny virions and induce NF-ĸB anti-viral pro-inflammatory responses upon virion binding (Galao et al., 2012), but has recently been reported to downregulate transient protein expression (Narkpuk et al., 2014) which is suggested to further impede viral replication. As retroviruses have developed strategies to counter BST2, it is clear that antagonism of this factor is vital for successful infection and survival. Host genetic factors, including the host-encoded viral restriction factors, have been demonstrated to influence not only susceptibility but also the clinical course of HIV-1 infection. Although numerous studies have been undertaken to investigate the role of polymorphisms in other restriction factors such as Apolipoprotein B mRNA editing catalytic polypeptide 3G (APOBEC3G)(Compaore et al., 2016; Singh et al., 2013) and tripartite interaction motif 5α (TRIM5α) (Goldschmidt et al., 2006; Javanbakht et al., 2006; van Manen et al., 2008) in HIV-1 disease progression, few genetic association studies have to date assessed the role of genetic polymorphisms in BST2.

Non-synonymous mutations, resulting in alterations to the primary amino acid sequence, do not impair virion restriction by BST2 as long as the general protein structure and conformation is retained (Perez-Caballero et al., 2009). It has been reported that such polymorphisms may alter BST2 functionality either through reducing rates of viral internalization (rs141648094, Y8H) or abrogating NF-ĸβ signaling (rs376557702, R19H) (Sauter et al., 2013). However, these exonic non-synonymous variants are very rare with minor allele frequencies of <0.05%. Thus, it is more probable for polymorphisms in the regulatory regions of the gene to influence BST2 function as these variants may alter transcriptional activity (if present within promoter regions) or induce mRNA degradation or translational repression (if polymorphisms reside in miRNA target sites, typically in the 3’UTR). The importance of BST2 cell-surface protein levels is highlighted by the fact that the HIV-1 viral accessory protein Vpu overcomes viral restriction by reducing the cell-surface expression of BST2 through degradation and mistrafficking of BST2 within endosomal compartments (Mitchell et al., 2009; Van Damme et al., 2008).

The present study is to our knowledge the first to provide a description of these BST2 variants (5’UTR indel rs3217318, 5’UTR rs12609479 SNP, 3’UTR rs10415893 SNP and 3’UTR rs113189798 SNP) in the context of differential spontaneous control of HIV-1 infection, particularly with respect to black South Africans, the population which harbours the greatest HIV-1 burden of infection globally.

Using data available from the 1000 Genomes Project (Auton et al., 2015) we found that the genotypic frequencies of three SNPs (rs12609479, rs10415893 and rs113189798) in the black South African HIV-1 uninfected population are significantly divergent from those of reference European populations, whilst largely mirroring the genetic representation observed in the reference African population (which comprised data from several populations of African descent), with the exception of rs12609479 for which the South African population had a significantly lower frequency of minor allele homozygosity compared to the reference African population. This particular variant was previously associated with a decreased risk of HIV-1 acquisition in injection drug users (Hancock et al., 2015).

The allelic and genotypic data for the 5’UTR indel rs3217318 (located 411 bp upstream of the BST2 transcription start site (TSS)) were unavailable from the 1000 Genomes Project for reference African and Caucasian populations (Auton et al., 2015). It is however, noteworthy to highlight that the heterozygous genotype (Δ19/i19) was markedly more prevalent in our HIV-1 uninfected South African black population (44.8%) compared to the HIV-1 uninfected Caucasian population (28%) in the Laplana (2013) study (Laplana et al., 2013). It will be interesting to explore whether this relationship is unique to the black South African population or shared by other African populations.

This 19 bp insertion has been proposed to repress BST2 promoter expression upon IFN-γ stimulus in vitro (Laplana et al., 2013). Hence, in the context of HIV viraemia, the presence of this insertion and the potential lowered expression BST2 levels, may be detrimental by reducing the cell’s ability to restrict the release of virions. In keeping with the proposed functional consequences of the polymorphism, the insertion has been associated with increased risk of HIV-1 acquisition (Singh et al., 2018) as well as faster rates of disease progression (Laplana et al., 2013).

With respect to the 5’UTR SNP, rs12609479 (located 201 bp upstream of the transcriptional start site), the rs12609479-A allele is the major allele in the Caucasian reference population whilst being the minor allele in populations of African descent. The rs12609479-A allele has been predicted to enhance transcription factor (c-Myb) binding in the BST2 promoter and thereby augmenting BST2 expression levels, although the functional evidence to support this hypothesis is lacking (Skelton et al., 2014).

Little is known concerning the role of these four BST2 variants in natural control of HIV-1 infection, particularly in populations most affected by the HIV-1 epidemic.

Heterozygosity for the rs12609479 SNP was the only genotype shown to differ significantly in its representation with HIV-1 infected progressors having higher representation compared to ECs (p=0.03), and the minor allele was also more highly represented in progressors compared to ECs (p=0.055). Carriage of the rs113189798-A/G genotype showed a trend of underrepresentation in progressors compared to ECs (p=0.07).

When we analysed the BST2 variants individually with respect to markers of disease progression, among progressors, the rs113189798-GG genotype was associated with significantly lower viral loads compared to individuals heterozygous and homozygous for the major allele. Another measure of progressive immunologic deterioration in HIV-1 disease is the rate of CD4⁺ T-cell decline. The rate of CD4⁺ T-cell decline is highly variable and may be influenced by a number of factors including gender, age at seroconversion, viral load set-point as well as route of transmission (Cori et al., 2015) and ethnicity (May et al., 2009). A recent study has revealed that the rate of pre-ART CD4⁺ T-cell decline may serve as predictive marker for immune reconstitution failure, the suboptimal recovery of CD4⁺ T-cell counts despite HIV viral suppression, particularly in individuals initiating ART with CD4⁺ T-cell counts below 200 cells/μl (Darraj et al., 2018). Hence, evaluation of this marker in HIV-1 infected progressors is of particular clinical relevance.

Analysis of our data with respect to CD4⁺ T-cell decline in individuals with progressive disease, revealed that heterozygosity for the rs3217318 indel (Δ19/i19) was associated with significantly faster rates of CD4⁺ T-lymphocyte decline (p=0.0134). These findings are in accordance with those from a Spanish cohort of injecting drug users in which carriage of the same genotype was associated with a faster rate of disease progression as measured by CD4⁺ T-lymphocyte decline to <200 cells/μl and ART initiation (Laplana et al., 2013).

Owing to the LD observed between rs3217318 and rs10415893, we considered the impact of the Δ19/i19_ G/A_ diplotype on disease progression. Progressors carrying the heterozygous diplotype revealed significantly elevated rates of CD4⁺ T-cell decline compared to those carrying the homozygous Δ19/Δ19_G/G_ diplotype (p=0.0114). However, as the level of significance was only nominally enhanced in the diplotype compared to the rs3217318 heterozygous genotype, it is evident that the carriage of the rs3217318- Δ19/i19 genotype and not the rs10415893-G/A genotype, has the more pronounced effect and is likely responsible for the observed faster progression rates.

The rs113189798 SNP was described in a meta-analysis across the BST2 gene and the 5 kb flanking regions with the intention of identifying variants associated with a risk of acquiring HIV-1 in injecting drug users (Hancock et al., 2015). The minor rs113189798-G allele was associated with a significantly increased risk of HIV-1 acquisition in Americans of both African and European ancestries (Hancock et al., 2015). In our study, rs113189798 heterozygosity (A/G) showed a trend of higher representation in ECs compared to progressors and homozygosity for the minor allele was associated with significantly lower HIV-1 viral load in individuals with advanced disease. Hence, our results suggest that in the context of viral control, contrary to the proposed role in HIV-1 acquisition, this polymorphism may confer an advantage.

Comparison of the frequencies of combinations of BST2 variant genotypes amongst our sub-groups indicated that homozygosity for the major rs12609479-G allele and heterozygosity at the other loci (Δ19i19GGGAAG) was more prevalent in controllers (11.3%), especially in ECs (17.4%) compared to progressors (5.6%), although this was not statistically significant. This 4-variant combined genotype combination was shown to be advantageous in the context of high levels of viral replication in the progressors as significantly higher CD4⁺ T-cell counts were detected in those carrying this combination when compared to carriage of other genotype combinations. Upon further stratification, it became apparent that the protective effect was predominantly driven by the rs12609479GG and rs113189798AG combination (_GG_AG). It is noteworthy to highlight that the Δ19i19GGGAAG genotype combination, is comprised of BST2 variants associated with lower BST2 expression levels, the implications of which are discussed below.

The rs12609479-A allele which is proposed to create a c-Myb transcription factor binding site (Hancock et al., 2015), is absent whilst the rs3217318-i19 allele which is associated with repressed BST2 expression, is present.

The proposed effects of promoter variants on gene expression and consequently protein expression levels, are attributable to alterations in transcription factor binding sites. However, gene regulation can be achieved through multiple mechanisms and can additionally be exerted at epigenetic and post-transcriptional levels by non-coding RNAs (ncRNAs), particularly long noncoding RNAs (lncRNAs) and microRNAs (miRNAs). Both ncRNA species have been shown to impact not only HIV-1 viral replication and pathogenesis but also disease progression (Lazar et al., 2016; Su et al., 2018).

Recent studies have identified a BST2 associated lncRNA named BST2 IFN-stimulated positive regulator (BISPR) which originates from a bidirectional BST2 promoter (Barriocanal et al., 2014; Kambara et al., 2014). Post-transcriptional inhibition by small interfering RNAs (siRNAs) of BISPR resulted in downregulation of BST2 expression in both studies, thus suggesting that BISPR may either facilitate BST2 expression or enhance transcription by stabilizing BST2 mRNA (Barriocanal et al., 2014; Kambara et al., 2014). It can be proposed that the promoter polymorphisms rs3217318 and rs12609479 may also influence BISPR expression and may additionally thereby indirectly contribute to reduced BST2 expression. The influence of these BST2 variants on lncRNA mediated regulation of BST2 transcription, warrants further investigation.

Although the impact of the 3’UTR SNPs on microRNA binding and BST2 gene expression is as yet unexplored, in silico analysis predicted that the rs10415893-A allele and rs113189798-G minor alleles, present in the highlighted Δ19i19GGGAAG combination, may introduce one and two additional miRNA binding sites, respectively. The suggested consequence of which would be enhanced miRNA-mediated gene repression and in turn reduced BST2 expression.

As BST2 mediated virion retention has been demonstrated to activate NF-ĸβ and thereby induce pro-inflammatory responses, specifically CXCL10, IL-6 and IFNβ cytokine production (Galao et al., 2012), it can be proposed that repressed BST2 gene expression and consequently reduced cell surface protein levels may result in lower immune activation. Since chronic immune activation is widely considered to be a major contributor to progressive CD4⁺ T-cell loss (Sousa et al., 2002), we hypothesize that the Δ19i19GGGAAG combined genotype, as well as the _GG_AG combined genotypes confers a CD4⁺ T-cell preservation advantage to HIV-1 infected progressors by lowering levels of immune activation.

As Vpu-mediated antagonism of BST2 is dependent on downregulation of the restriction factor on the cell surface (Van Damme et al., 2008), overexpression of BST2, as induced by IFNα, may contribute to suppressed HIV-1 replication and lower viral loads, as has been reported (Pillai et al., 2012). However, 63% of individuals in this study were treated during acute infection (Pillai et al., 2012). The benefit of IFNα treatment during acute infection was also shown in SIV infection (Sandler et al., 2014). Importantly however, this study further revealed that this treatment (and by inference BST2 overexpression) is detrimental during chronic stages of the disease by contributing to CD4⁺ T-cell depletion and enlarged reservoir size (Sandler et al., 2014). Hence, this study further highlights that a pro-inflammatory immune response may be beneficial during acute HIV-1 infection whilst being detrimental during chronic stages of the disease (Sandler et al., 2014). It is plausible that there may be a similar dichotomy in protective or deleterious outcome with respect to BST2.

Recently, CRISPR-mediated gene-editing methods have been used to induce endogenous BST2 expression and are being explored as potential novel therapeutic strategies for HIV-1 (Zhang et al., 2019). However, considering our findings, the timing (acute vs chronic infection) of such interventions and their consequences on the clinical course of HIV-1 disease warrants careful investigation.

It is however important to note that the current study has number of limitations. Since racial and ethnic designations were based on patient self-reporting, the possibility of genetic admixture and the potential influence thereof on genetic associations should not be disregarded. Due to the exploratory nature of this study and the use of extreme phenotypes, adjustment for multiple comparisons was not performed on SNPs showing significant associations. Although this decreases risk of Type II errors it does however increase the risk of Type I errors and thereby warrants validation of these results in different cohorts of similar ethnicity.

In summary, we have determined the frequency of four HIV-1 associated BST2 polymorphisms in the black South African population, of which three (rs3217318, rs10415893 and rs113189798) had not yet been previously described.

We further demonstrated that “protective” and “detrimental” designations assigned to the minor 5’UTR rs12609479-A and 3’UTR rs113189798-G alleles in the context of HIV-1 acquisition are discordant with our findings with respect to disease progression in the black South African population. Our results further reveal that HIV-1 disease progression is influenced by possession of select BST2 genotype combinations and specifically implicates the rs12609479GG and rs113189798AG combined genotype, which is enriched in HIV-1 controllers and particularly in ECs, in the preservation of CD4⁺ T-cells in progressive infection.

Our findings suggest that BST2 may contribute to effective control of HIV-1 in the absence of ART. Further investigations into the mechanisms underlying the role of BST2 in promoting spontaneous antiviral immunity to HIV pathogenesis will be important in the design of novel therapies aimed at attaining a “functional cure” for HIV-1.

Supplementary Material

1

Fig S1. Genotyping of the BST2 polymorphisms. (A) Schematic representation of the BST2 gene indicating the positions of the polymorphisms as well as the respective nucleotide base changes. (B) Genotyping the rs3217318 indel by size discrimination gel electrophoresis. PCR amplicons resolved by 2% super fine high resolution agarose gel electrophoresis. Genotyping using banding pattern as indicated on the figure. (C) Example of genotyping using allele specific real-time PCR shown for single nucleotide polymorphism (SNP) rs113189798. Two primers were designed per SNP, one complementary to the major/wild type allele (purple) and one complementary to the minor/mutant allele (pink). The primer resulting in more efficient amplification and resulting in the lower cycle threshold (Ct) value is indicative of homozygosity for that allele. Heterozygosity is characterised by comparable binding and amplification efficacy between both primer sets. bp: base pair. Mw: molecular weight.

Fig S2. Linkage Disequilibrium for BST2 variants in the South African black population. Linkage disequilibrium (LD) plots showing the linkage disequilibrium between the BST2 variants rs3217318, rs12609479, rs10415893, and rs113189798 in all study participants as well as in the study sub-groups, namely, HIV-1 uninfected, HIV-1 infected controllers and HIV-1 infected progressors. Values for r2 LD coefficient are shown between every two minor allele combinations and ranges from 0 indicating no linkage to 100 representative of complete linkage. LD was computed using Haploview v4.2 (26)

2

Table S1. Primer pairs used for the PCR amplification and real-time PCR genotyping of the BST2 SNPs

Table S2. The allelic and genotypic frequency distribution of four BST2 polymorphic variants in HIV-1 uninfected individuals, HIV-1 infected controllers and HIV-1 infected progressors

Table S3. The number and frequency (%) of predominant genotype combinations of four BST2 SNPs rs3217318, rs12609479, rs10451893 and rs113189798

Table S4. The number and frequency (%) of the most predominant two-variant combinations of the BST2 polymorphisms in the study populations

Table S5. Comparison of predominant two-genotype combinations of BST2 SNPs (excluding major allele homozygosity) between HIV-1 infected sub-groups

Table S6. The effect of 3’UTR BST2 polymorphic variants on putative miRNA target binding sites

Highlights.

• BST2 rs12609479-AA genotype is underrepresented in black South Africans

• Select BST2 genotype combinations may contribute to natural HIV-1 control

• HIV-1 progressors with rs12609479-GG/rs113189798-AG have higher CD4+T-cell counts