The HLA-B/-C haplotype block contains major determinants for host control of HIV

Abstract

A genome-wide association study of people with incident human immunodeficiency virus (HIV) infection selected from nine different cohorts identified allelic polymorphisms, which associated with either viral set point (HCP5 and 5′ HLA-C) or with HIV disease progression (RNF39 and ZNRD1). To determine the influence of these polymorphisms on host control of HIV, we carried out a population-based association study. The analysis revealed complete linkage disequilibrium between HCP5 and HLA-B*5701/HLA-Cw*06, a modest effect of 5′ HLA-C on viral set point in the absence of HLA-B*5701, and no influence of the RNF39 /ZNRD1 extended haplotype on HIV disease progression. No correlation was found between the infection status and any of these genetic variants (P>0.1, Fisher's exact test). These findings suggest a pattern of strong linkage disequilibrium consistent with an HLA-B/-C haplotype block, making identification of a causal variant difficult, and underscore the importance of validating polymorphisms in putative determinants for host control by association analysis of independent populations.

Introduction

Host proteins have crucial roles in the human immuno-deficiency virus (HIV) life cycle and contribute to the diversity of the host response to infection and disease progression. Genetic association analyses of populations at risk for specific HIV-related outcomes have revealed the influence of host allelic polymorphisms on susceptibility to HIV infection, the time to AIDS and death. At least 15 allelic variants in genes encoding human leukocyte antigen (HLA) antigens, chemokine receptors and their ligands, and cytokines have been shown to influence HIV/AIDS susceptibility.^1,2 Nevertheless, each of the described host factors are relevant to only a small proportion of people who continue to resist infection or AIDS-defying illness decades after infection with HIV.

Genome-wide association studies provide opportunities to uncover host polymorphisms that influence susceptibility and resistance to HIV infection free of mechanistic hypotheses. Of the allelic polymorphisms identified in people with incident HIV infection by genome-wide association studies, significant associations between five host proteins and HIV pathogenesis were reported.³ Three of the identified proteins were implicated in the control of the viral set point (the HLA complex, HCP5 and the 5′ region of the HLA-C (5′ HLA-C) gene) and two in the progression of disease (ring finger protein 39 (RNF39) and zinc ribbon domain-containing 1 (ZNRD1)).³ However, rare variants, population differences in linkage disequilibrium patterns, and epistasis can complicate the interpretation of whole genome association studies. Accordingly, observations of determinants for host control of infection based on single nucleotide polymorphisms (SNPs) need to be replicated in relevant populations, checking for patterns of linkage disequilibrium with possible risk-related genes, followed by a deeper genomic and functional analyses of the associated region to verify putative causal variants.

Results

To test for the influence of genetic variation in viral set point and time to HIV disease progression, we analyzed the SNPs in HCP5 (rs2395029), 5′ HLA-C (rs9264942), ZNRD1 (rs9261129, rs9261174, rs3869068) and RNF39 (rs2301753, rs2074480, rs2074479) using data from matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometer assays. Genomic DNA samples were collected from men enrolled in the Chicago component of the Multicenter AIDS Cohort Study, a natural history study of men who have sex with men. Of the 1351 men enrolled in the study, 998 had peripheral blood samples available for SNP genotyping, 562 were infected with HIV and 436 were at risk for infection. We defined disease outcome either by the level of viral RNA in plasma measured after the initial burst of viral replication during acute infection or by the rate of CD4+ T-cell decline over a minimum period of 2 years, including data from at least four study visits. These measurements were taken before the onset of AIDS and before the start of treatment. High-resolution HLA class I and class II allelic typing data were available for all subjects.⁴ Table 1 lists the frequencies of these eight SNPs in these men.

Table 1.

The frequency of HCP5, 5′ HLA-C, RNF39 and ZNRD1 SNPs in men enrolled in the Chicago component of the MACS

Locus	rs number	SNP	Frequency	HWE P-value	LD (r2) value: B*5701	Subjects
HCP5	rs2395029	TT	0.929
		GT	0.069	0.38	0.86	997
		GG	0.002
5′ HLA-C	rs9264942	TT	0.383
		CT	0.482	0.42	0.06	995
		CC	0.135
RNF39	rs2074479	TT	0.679
		CT	0.283	0.20	0.008	997
		CC	0.038
RNF39	rs2074480	AA	0.681
		CA	0.282	0.23	0.008	997
		CC	0.037
RNF39	rs2301753	CC	0.681
		CA	0.282	0.23	0.008	997
		AA	0.037
ZNRD1	rs3869068	GG	0.680
		AG	0.283	0.28	0.008	998
		AA	0.037
ZNRD1	rs9261174	TT	0.681
		CT	0.282	0.23	0.008	998
		CC	0.037
ZNRD1	rs9261129	TT	0.679
		TC	0.284	0.28	0.008	998
		CC	0.037

Abbreviations: HLA, human leukocyte antigen; HWE, Hardy–Weinberg equilibrium; MACS, Multicenter AIDS Cohort Study; RNF39, ring finger protein 39; SNP, single nucleotide polymorphism; ZNRD1, zinc ribbon domain-containing 1.

SNPs are telomeric (top) to centromeric (bottom). The HWE P-values are for a Fisher's exact test for HWE. The LD r² values give the linkage disequilibrium with the B*5701 assuming HWE distribution of the double heterozygotes and ignore 10 patients who had not been HLA typed.

We first compared our data with the previously identified association of genetic variation in the HCP5 gene with HIV-related disease progression rates (Table 2a). HCP5 is a human endogenous retroviral element with sequence homology to the pol gene that encodes two proteins expressed in lymphocytes. Our results confirm that the polymorphism located in the HCP5 gene (rs2395029) is in very strong linkage disequilibrium with HLA-B*5701 (Table 1). In our data set, 71 men carry HLA-B*5701, 66 of whom (including both the HLA-B*5701 homozygotes) carry the minor G allele of rs2395029 that is found at a frequency of f=0.037 in this sample. The remaining five men who carry HLA-B*5701 are homozygous for the major T allele of rs2395029 that is found at a frequency of f=0.963. The effect of this HCP5 rs2395029/HLA-B*5701 polymorphism on the level of HIV RNA in plasma is large; the viral set point in men with rs2395029 homozygous for the more common T allele is 18 010 copies per ml (interquartile range (IQR) 4456–41 680), whereas in men containing at least one G allele it is 3648 copies per ml (IQR 627-11 460), which is highly significant (Mann–Whitney U-test, P=7 × 10⁻⁷). The HCP5 locus is also in linkage disequilibrium with HLA-C*0602, and of the 180 men who carry HLA-C*0602, 63 men carry the minor G allele.

Table 2.

(a) HCP5 and 5′ HLA-C SNP genotype frequency and association between HIV RNA at set point and HLA-B*5701 and C*0602; (b) 5′ HLA-C SNP association with HIV RNA at set point and HLA-B*5701; (c) 5′ HLA-C SNP association with HIV RNA at set point and HCP5

Locus	rs number	SNP	Frequency		HIV RNA in plasma at set point, median and (IQR)	Mann–Whitney U-test P-value	HLA-B*5701 association of given SNP allele	HLA-C*0602 association of given SNP allele
(a)
HCP5	rs2395029	TTa	0.929		18 010 (4456–41 680)	TT vs GT or GG	5 of 916	117 of 916
		GTa	0.069		3648 (627–11 701)	P=7 × 10−7	64 of 69	61 of 69
		GGb	0.002				2 of 2	2 of 2
		TTa	0.383		21 980 (5104–49 850)	TT vs CT P<0.07	2 of 377	1 of 377
5′ HLA-C	rs9264942	CTa	0.482		15 480 (3558–37 050)	CT vs CC P<0.03	41 of 475	111 of 475
		CCb	0.135		7918 (1060–20 780)	CC vs TT P<0.002	28 of 133	68 of 133
(b)
Locus	rs number	SNP	HLA-B*5701 association (+indicates heterozygosity)c		Frequency of genotype, frequency with set point	HIV RNA in plasma at set point, median and (IQR)	Mann–Whitney U-test P-value against heterozygous
5′ HLA-C	rs9264942	TT	+		2, 1	20320	0.298
		CTa	+		41, 14	4920 (732, 12 430)
		CCb	+		26, 12	773 (300–7443)	0.167
		TT	−		375, 137	21 980 (4975, 49820)	0.328
		CTa	−		434, 182	17 000 (4359, 39700)
		CCb	−		105, 34	11 430 (4464–32 600)	0.341
(c)
Locus	rs number	SNP	HCP5 SNP associationd		Frequency of genotype, frequency with set point	HIV RNA in plasma at set point, median and (IQR)	Mann–Whitney U-test P-value against heterozygous
5′ HLA-C	rs9264942	TT	GT		2, 2	17 110 (15 510–18 710)	0.158
		CTa	GT		41, 22	4920 (720–11 720)
		CCb	GT		26, 11	708 (300–5295)	0.057
		TT	TT		379, 137	23 200 (4975, 49890)	0.258
		CTa	TT		439, 184	17 000 (4346–39 570)
		CCb	TT		106, 36	11 430 (4935–29 490)	0.324

Abbreviations: HIV, human immunodeficiency virus; HLA, human leukocyte antigen; IQR, interquartile range; SNP, single nucleotide polymorphism.

^aMajor SNP alleles.

^bMinor SNP alleles.

^cTwo individuals homozygous for HLA-B*5701 were omitted to avoid confounding.

^dTwo patients with HCP5 with GG homozygosity were omitted to avoid confounding.

Note: analysis was carried out without correction for multiple comparisons suggested by previous studies.

There is a pattern of strong linkage disequilibrium between the minor G allele of HCP5 and HLA-B*5701. As a result, only 5 of 916 men carry both TT and HLA-B*5701 and only 5 of 69 men carry GT and not HLA-B*5701. Consequently, we do not have sufficient data to attribute the effect to the two genes differentially. We also note that the predominant B*57 allele in African-Americans, HLA-B*5703, is not in linkage disequilibrium with the HCP5 minor G allele. In a recent study of 16 elite controllers, 10 African-Americans carried the HLA-B*5703 allele, 1 African-American carried the HLA-B*5702 allele, and all 11 of these men carried the major T allele of HCP5.⁵ These observations are consistent with the hypothesis that protection from HIV is conferred by b*57 alleles and that the association of the HCP5 minor G allele with low levels of plasma HIV RNA is due to linkage disequilibrium with HLA-B*5701. Testing this hypothesis will require the analyses of larger populations, including those of African descent, in whom B*57 is not in linkage disequilibrium with the minor G allele of HCP5.

We examined the association between the polymorphism in 5′ HLA-C (rs9264942) with variation in the level of HIV RNA in plasma (Table 2). This polymorphism is thought to influence the expression levels of HLA-C.⁶ In our data set, the frequency of the major T allele at rs9264942 is f=0.624 and the frequency of the minor C allele is f=0.376. We confirmed that the HLA-C polymorphism is in linkage disequilibrium with HCP5/ HLA-B*5701. In all, 69 of the 608 people carrying the minor C allele at rs9264942 also carry HLA-B*5701 (95% confidence interval on prevalence 8.9–14.1%), whereas only 2 of the remaining 377 people do (95% confidence interval on prevalence 0.06–1.9%), which, not surprisingly, is associated with low levels of HIV RNA in the plasma (Table 2a). The median viral RNA set point for men homozygous for the minor C allele (including both men also homozygous for HLA-B*5701) is 7918 copies per ml (IQR 1060–20 780 copies per ml) against 15 480 copies per ml (IQR 3558–37 050 copies per ml) in men heterozygous at rs9264942, and 21 980 copies per ml (IQR 5104–49 850 copies per ml) in men homozygous for the major T allele. The difference for heterozygotes measured against the minor C homozygotes is significant (Mann–Whitney U-test, P<0.03), and when measured against the major T homozygotes, shows a strong trend in the same direction (Mann–Whitney U-test, P<0.07). The significance of the trend is reduced because of the amount of data (P=0.096–0.34), should we stratify the men by whether they carry the HLA-B*5701 allele (Table 2b) or analyze them with a given genotype at the HPC5 (rs2395029) locus (Table 2c) separately. These findings indicate that the key protective effect of 5′ HLA-C is primarily due to linkage disequilibrium with the polymorphism at HCP5/HLA-B*5701, with possibly a modest effect conferred in the absence of these protective alleles.

Next we compared our data with the identified association of genetic variation in the ZNRD1 (rs9261129, rs9261174 and rs3869068) and RNF39 (rs2301753, rs2074480 and rs2074479) genes with HIV disease progression rates. The ZNRD1, RNF39 and HCP5 genes are in close physical proximity along human chromosome 6 (cytogenetic band 6p21.33) within the highly polymorphic HLA class I gene-rich region. The ZNRD1 gene encodes an RNA polymerase I subunit; the RNF39 gene is poorly characterized. The ZNRD1 and RNF39 polymorphisms are in very strong linkage disequilibrium. In particular, except for one man (who was homozygous for T at rs9261174 and A at rs2074480, heterozygous at the other three loci), the pattern of the six SNPs within ZNRD1 and RNF39 are explainable by the presence of two extended haplotypes; a major (82%) TTGCAT haplotype at the respective SNPs, a minor (17%) CCAACC haplotype and rare SNP variants differing from the major haplotypes at a single position each (the observed variants were CTGCAT, TTGAAT, TTGCAC, CCACCC and TCAACC).

Assuming this resolution for all cases, the TTGCAT and CAACC alleles are consistent (Fisher's exact P=0.24) with Hardy–Weinberg equilibrium (HWE), as are all the minor ones except TTGCAC that appears only in the homozygous state and in one man. Therefore, the constituent SNPs are in high linkage disequilibrium (r²>0.97 assuming HWE) and their effects on disease progression cannot be distinguished. These entire haplotypes are also in linkage disequilibrium with the HCP5 polymorphism (rs2395029). Although all the men homozygous for the minor CCAACC haplotype (including its variations) are homozygous for the major T allele at HCP5 rs2395029, we estimate r²=0.008 assuming HWE, so that the effects of the HCP5 polymorphism can be separately studied from those of the haplotype. We analyzed the effect of the ZNRD1, RNF39 and HCP5 polymorphisms on HIV disease progression rates without taking into account the confounding effects of ethnicity, risk factors or HLA. Owing to the high linkage disequilibrium between the six SNPs within ZNRD1 and RNF39, we selected one (rs3869068) as a representative polymorphism. We found no significant effect of the rs3869068 polymorphism on the RNA set point before therapy initiation (P>0.7) or on the CD4 slope at that time (median −34, IQR (−100, −4) in the men carrying the homozygous minor allele compared with median −53, IQR (−98, −23) in the people carrying the major allele; Mann–Whitney U-test, P=0.2).

Discussion

Significant genotype associations for four genetic loci within a short distance on chromosome 6p21 were identified for HIV/AIDS susceptibility in an earlier study,³ with people carrying the minor allele variant of HCP5 (rs2395029) and 5′ HLA-C (rs9264942) having a lower viral RNA set point and people carrying six minor polymorphisms in ZNRD1 (rs9261129, rs9261174, rs3869068) and RNF39 (rs2301753, rs2074480, rs2074479) having a faster progression to AIDS. In our population-based study, we find an association between the HCP5 (rs2395029) minor G allele and 5′ HLA-C (rs9264942) minor C allele and a lower viral RNA set point; however, the key protective association of the minor allele variants in 5′HLA-C and HCP5 may be largely due to linkage disequilibrium with HLA-B*5701. Further, the pattern of the six polymorphisms within ZNRD1 and RNF39 are explainable by the presence of two extended haplotypes that had no influence on HIV disease progression.

Like the earlier study,³ our data show very strong linkage disequilibrium between HCP5 and HLA-B*5701, a known marker in the control of HIV disease.^4,7 Although others confirm an association between the minor allele variant of HCP5 (rs2395029) and 5′ HLA-C (rs9264942) and a lower viral RNA set point, they did not determine whether this was due to linkage disequilibrium between HCP5 and 5′ HLA-C and a nearby locus, such as the HLA-B/-C region.⁸ Although its limited numbers will require confirmation in larger ethnically specific populations, a small African elite suppressor study,⁵ which found that the protective HCP5 minor G allele is not in linkage disequilibrium with HLA-B*57 in that group, is of particular interest herein, as it gives credence to the hypothesis that B*57 confers the protective allele status in the African controllers. Without strong functional data to support a role for HCP5 in the control of HIV, the pattern of linkage disequilibrium suggests an association between loci in a HLA-B/-C haplotype block with low levels of HIV RNA in plasma. These findings underscore the need to locate the causal variant in the genomic region first identified by genome-wide association studies.

Unlike the earlier study, we find no association between the polymorphisms in ZNRD1 and RNF39 or their extended haplotypes and AIDS susceptibility, a finding consistent with others.⁹ Although the early study focused on the individual polymorphisms in ZNRD1 and RNF39, we took a more comprehensive approach, examining the spatial structure of linkage disequilibrium. The strong degree of linkage disequilibrium that spreads across the HLA-B/-C region and their surroundings leads to the existence of extended haplotypes. Apart from trivial departures, we find distinct major and minor haplotypes that accurately describe the linkage disequilibrium data. We did not, however, find an association between ZNRD1 and the alleles within the HLA-A10 serogroup, such as HLA-A*25 and HLA-A*26 in European-Americans or HLA-A*34 and HLA-A*66 in African-Americans.⁴

Our data do not support claims that the HCP5 gene perse explains viral set point, nor that ZNRD1 and RNF39 explain HIV disease progression. The strong linkage disequilibrium between two genetic loci (HCP5 and 5′ HLA-C) plus HLA-B*5701 and between the RNF39 and ZNRD1 polymorphisms, combined with no established functions for the SNP gene products in AIDS pathogenesis, provides few clues as to the causal variant that can influence HIV disease progression. This finding emphasizes the problem of statistical analyses of loci in regions of extended linkage disequilibrium, such as the HLA-B/-C locus, and the importance of validating polymorphisms that explain differences in disease outcome by association analyses of independent populations.

Materials and methods

Study participants

The participants were men enrolled in the Chicago component of the Multicenter AIDS Cohort Study who were followed at 6-month intervals, queried about risk behaviors, tested for antibodies to HIV, and had their CD4+ and CD8+ T-cell numbers enumerated. Men infected with HIV received antiretroviral therapy and levels of HIV RNA in plasma measured by quantitative reverse transcriptase–PCR (Roche Molecular Diagnostic Systems, Branchburg, NJ, USA). All men had equal access to care. Of the 1351 men originally recruited for the Chicago component of the Multicenter AIDS Cohort Study, 998 had peripheral blood samples available for SNP analysis and 562 of them were infected with HIV. The time from the acquisition of antibodies against HIV to AIDS (CD4+ T cells <200 per mm³) was established for only 64 men. Accordingly, we defined disease outcome either by the level of viral RNA in plasma measured after the initial burst of viral replication during acute infection (the viral level at set point), or by rate of CD4+ T cell decline over a minimum period of 2 –years, including data from at least four interval study visits. These measurements were considered before the onset of AIDS and before the start of therapy. All men provided written informed consent according to the guidelines of the human subjects protection committee of Northwestern University.

HLA and SNP genotyping

High-resolution HLA class I and class II allelic typing was previously carried out for all subjects.⁴ The HCP5, 5′ HLA-C, RNF39, and ZNRD1 gene fragments were analyzed using a novel assay developed using iPLEX Gold chemistry (SEQUENOM, San Diego, CA, USA) and analyzed on a MALDI-TOF mass spectrometer. Briefly, samples were amplified by PCR using primers that captured the region surrounding the SNP. A shrimp alkaline phosphatase treatment was used to neutralize unincorporated dNTPs, and then the reaction mixture containing primer, four mass-modified nucleotides and enzyme was added to the amplicons. During a second thermocycling step, the primer was extended through the SNP by one nucleotide, which terminated the reaction. The resulting products were spotted onto a 384-well microchip using a nanodispenser and spectra acquired on the MassARRAY Compact Analyzer (SEQUENOM). The spectra were viewed and analyzed using TyperAnalyzer 4.0 software from Sequenom. The mass spectrometer can detect 0.2 femtomole of target DNA sequence in the presence of a 10⁴-fold excess of other DNA sequences, with accuracy rates >99%.¹⁰ Assay development, validation and quality control measures can be found in Supplementary materials, available at http://www.nature.com/gene/index.html.

Statistical analysis

Analyses were carried out with the statistical package R (R foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org). We tested the strength of the associations with the nonparametric Mann–Whitney U-test statistic, comparing distributions of levels of HIV RNA in plasma or CD4+ T-cell slope values in the set of men who carried the allele with those who did not. As our testing of associations of the polymorphisms with HIV disease was by individual comparisons to previous findings, none of the tests were corrected for the multiple comparisons suggested by previous studies.