EP2265732A2 - Genetics variants associated with hiv disease restriction - Google Patents

Abstract

The present invention relates, in general, to human immunodeficiency virus (HIV) and, in particular, to genetic variants associated with restriction of HIV disease progression.

Description

GENETICS VARIANTS ASSOCIATED WITH HIV DISEASE RESTRICTION

This application claims priority from U.S. Prov. Appln. No. 61/071,141, filed April 15, 2008, the entire content of which is incorporated herein by reference.

This invention was made with government support under Grant No. A 10678501 awarded by the National Institutes of Health. The government has certain rights in the invention.

TECHNICAL FIELD The present invention relates, in general, to human immunodeficiency virus (HIV) and, in particular, to genetic variants associated with restriction of HIV disease progression.

BACKGROUND Humans show remarkable variation in vulnerability to infection by HIV-I and especially in the clinical outcome following infection. One of the most striking differences is the plasma level of virus in the non-symptomatic phase preceding progression to AIDS (the viral set point). Some individuals manage to control virus to low or even undetectable levels, while other individuals have much higher plasma viral loads (VL), with inter-patient variation over 4 to 5 logs (base 10) routinely observed (see Fig. 3 in Telenti et al, Nat. Rev. Microbiol. 4:865 (2006)). A small fraction of this variability can be explained by demographic factors and variants of known genes such as chemokines, chemokine receptors and cytokines (about 15% of the variation in the dataset, see also Telenti et al, Nat. Rev. Microbiol. 4:865 (2006), O'Brien et al, Nat. Genet. 36:565 (2004), Telenti et al, Future Virology 1 :55 (2006), Carrington et al, Annu. Rev. Med. 54:535 (2003), Nelson et al, J. Acquir. Immune Defic. Syndr. 42:347 (2006), Bleiber et al, J. Virol. 79:12674 (2005)).

HIV-I set point is a particularly important phenotype not only because of its dramatic variability among individuals, but also because of its relative stability within individuals over time and its impact on disease progression and on infectiousness (Mellors et al, Science 272:1167 (1996)). A better understanding of the causes of the differences in VL could provide pointers to new vaccines and drugs that control the virus. In order to facilitate the discovery of any gene variants that may influence viral control, however, it is essential to move beyond the targeted candidate gene studies that have characterized work to date (Telenti et al, Nat. Rev. Microbiol. 4:865 (2006)).

A whole-genome association study of variation in the host control of HIV- 1 has been carried out focusing on the determinants of VL set point and secondarily on the progression towards AIDS (measured by the decline of CD4 positive cells). This study resulted in the identification of three genetic variants (or groups of variants) that associate with HIV load and restriction of disease progression. Those variants are described below and in PCT/US2008/003964. The present invention relates, at least in part, to the identification of an association of a deletion at chromosome 19 ql3.42 with HIV progression (p=0.0010). As described below, closer analysis of the region has revealed that there are both deletions and duplications in this same position, and that the locus is that of the KIR genes.

SUMMARY OF THE INVENTION

The present invention relates generally to HIV. More specifically, the invention relates to genetic variants associated with restriction of HIV disease progression and to methods of using such variants as prognostic markers. Objects and advantages of the present invention will be clear from the description that follows.

BRIEF DESCRIPTION OF THE DRAWINGS

Figures IA and IB. HIV-I viral load at set point is highly correlated with

HCP5 rs2395029 genotype (T major allele, G minor allele) (Fig. IA) and with HLA-C 5' region rs9264942 genotype (T major allele, C minor allele) (Fig. IB).

Figure 2. Partial map of the HLA Class I region (chromosome 6 p21.3). Indicated are the p-values [-log(P)] of all genotyped SNPs annotated with the gene structure. The 2 independent SNPs that show genome- wide significant association with HIV-I VL at set point are displayed and marked in red. Graph was drawn from WGA Viewer software (see website: genome.duke.edu/centers/ pg2/index_html/downloads/AnnotationSoftware).

Figures 3A-3C. Fig. 3 A. Non linear effect of HLA-C expression levels on HIV-I VL at set point. In italic are shown the numbers of patients linked to each point in Sanger Genevar database for expression data and in the cohort for set point results, respectively. Figs. 3B and 3C show the distributions of respective data according to genotypes.

Figures 4A-4J. (Fig. 4A) rs9264942; (Fig. 4B) rs6457374; (Fig. 4C) rs2395029; (Fig. 4D) rs9261 174; (Fig. 4E) rs2074480, (Fig. 4F) rs7758512, (Fig. 4G) rs9261129; (Fig. 4H) rs3869068; (Fig. 41) rs2301753; and (Fig. 4J) rs2074479. Figure 5. Examples of deletions and duplications at the KIR locus. Sample names are shown on the left, the genes are shown along the bottom. The CNVs are displayed as red (darker) (deletion) or green (lighter) (duplication) bars.

Figures 6A and 6B. Effect of CNVs at KIR2 locus on HIV progression (Fig. 6A) and setpoint (Fig. 6B). Error bars depict standard error, sample sizes are shown above.

Figures 7A and 7B. Linkage disequilibrium with the KIR CNVs, assessed for 500kb either side. Fig. 7A shows the LD with the deletion, Fig. 7B the LD with the duplication.

Figure 8. Depiction of KIR genes and haplotypes from Martin &

Carrington 2008.

Figure 9. CNV calls for the KIR locus in samples of African origin run on the 1 million SNP chip. The CNVs appear to be longer but this is likely to be a result of poorer regional coverage on the 550 chips.

Figure 10. Identification of gene variants that influence early response to

HIV-I .

DETAILED DESCRIPTION OF THE INVENTION The present invention results, in part, from the identification of three genetic variants (or groups of variants) associated with important differences in HIV load, the most important prognostic marker of HIV disease progression. Together, these variants can explain a substantial part of inter-individual variability in HIV plasma levels. Identification of these variants sheds new light on HIV pathogenesis and on interactions between the immune system and the virus, thereby revealing new therapeutic targets. The invention also results, in part, from the identification of an association between a deletion at chromosome 19 ql3.42 and HIV progression (see Example 2 below).

In a first embodiment, the invention relates to two single nucleotide polymorphisms located in the 5' region of the HLA-C gene in the MHC Class I region on chromosome 6: SNP reference numbers are rs9264942 and rs6457374 (for identity of the polymorphisms referenced herein, see ncbi.nlm.nih.gov/projects/SNP). (Fig. 4.) These genetic variants are associated with differences in HLA-C mRNA expression, and their 5' location is likely to explain this effect (promoter/enhancer). HLA*CwA had been suspected of being associated with rapid HIV progression (Carrington et al, Science 283(5408): 1748- 1752 (1999)) but the effect was more recently attributed to linkage of this type with HLA*B35-Px (Carrington, Annu. Rev. Med. 54:535-551 (2003)). In fact, a partial linkage with rs9264942 is a more likely explanation. This is the first observed quantitative effect of an HLA Class I protein on HIV viral load, independently of HLA type.

In a second embodiment, the invention relates to a single nucleotide polymorphism located in a putative coding region of the HCP5 gene in the MHC Class I region on chromosome 6: SNP reference number is rs2395029. The function of the HCP5 gene was previously unknown. However, based on structural analogy with human endogenous retroviruses, a link to anti-retro viral immunity had been suggested (Kulski et al, Immunogenetics 49(5):404-412 (1999)). The genetic variant identified here is located in a region that shares homology with the pol sequence of certain retroviruses (human endogenous retroviruses). It is a non-synonymous change, that is, it is responsible for a change in the amino acid sequence of the predicted peptide generated from this genome sequence. This SNP is in high linkage disequilibrium with HLA*B5701 , which has the strongest known host genetic influence on HIV disease (Carrington, Annu. Rev. Med. 54:535-551 (2003)). It is not yet clear if it acts only as a tag for this type-related effect, or if it is causal. This is the first documented description of an association of HCP5 with HIV disease and raises the possibility that HCP5 and not *B5701 is responsible for HIV control.

In a third embodiment, the present invention relates to seven single nucleotide polymorphisms distributed between three genes in the MHC Class I region on chromosome 6: HCG9, RNF39 and ZNRDl . The SNP reference numbers are rs9261 174, rs2074480, rs7758512, rs9261129, rs3869068, rs2301753 and rs2074479. These variants are in perfect linkage disequilibrium, and it is thus impossible to distinguish between them regarding causality on the observed phenotype (better or worse control of HIV viral load). They are associated with significant differences in the expression of the zinc ribbon domain-containing 1 (ZNRDl) gene, which is a transcription-associated gene. ZNRDl is known to play a role in multidrug resistance phenotype of gastric cancer cells through upregulation of other genes (Shi et al, Cancer Biol Ther. 3(4):377-381 (2004)). Considering its structure and function, it is expected to have a direct influence on regulation of HIV transcription.

It is anticipated that the differential expression of HLA-C and ZNRDl genes play a key role in HIV infectivity. In addition, if HCP5 is responsible for the observed effect on HIV load, its products (RNA, peptides) can be expected to have a direct functional role in defense against retroviruses.

In yet another embodiment, the present invention relates to a method of assessing the rate of a non-symptomatic HIV patient progressing to AIDS. The method comprises assaying DNA from the patient for the presence of a copy number variant (CNV) at chromosome 19 ql3.42. As shown in Example 2 below (see also Fig. 10), a patient carrying a deletion at chromosome 19 ql 3.42 is likely to undergo a more rapid progression to AIDS than is a non-systematic HIV patient that does not carry the deletion. Conversely, a patient carrying a duplication at chromosome 19 ql 3.42 is likely to undergo a less rapid progression to AIDS than is a non-systematic HIV patient that does not carry that duplication.

It is known that genetic variants often occur in groups along a chromosome (haplotype blocks). Thus, identification of the above-referenced SNPs as being associated with viral load may uniquely identify additional SNP(s) present in such haplotype blocks. Even if a SNP associated with viral load were found not to be the definitive causative SNP producing the effect, it is nonetheless a valid marker, often in linkage disequilibrium, or directly in a haplotype block, with the causative SNP. Resequencing of the genomic DNA region around SNPs specifically identified herein may reveal other SNP(s) associated with viral load/disease progression (i.e., the associated interval) - these could be predicted based upon their presence in the same haplotype or by being in linkage disequilibrium with the specific SNPs disclosed here.

The presence of any of the above-referenced polymorphisms in a sample (e.g., a biological sample such as blood) can be determined using a variety of genotyping techniques known in the art (e.g., a "CHIP" or SNP panel). All SNPs described herein are present on Illumina's HumanHap550 genotyping BeadChip (see illuma.com). Suitable techniques also include the use of polymerase chain reaction and extension primers, RJLP analysis and mass spectrometry (see also Ye et al, Hum. Mutat. 17(4):305 (2001), Chen et al, Genome Res. 10:549 (2000).)

Screening for the genetic variants enables clinicians to better stratify a given patient for therapeutic intervention. Additionally, knowledge of the genetic variants allows patients to choose, in a more informed way in consultation with their physician, from available therapeutic approaches. The invention also relates to kits suitable for use in testing for the presence of the polymorphisms identified herein. Such kits can include, for example, reagents (e.g., probes or primers) necessary to identify the presence of the above- referenced polymorphisms. Certain aspects of the invention are described in greater detail in the non- limiting Examples that follows (see also U.S. Provisional Application No 60/907,271, filed March 27, 2007, U.S. Provisional Application No. 60/929,432, filed June 27, 2007, and PCT/US2008/003964, filed March 27, 2008, the contents of all of which are incorporated herein by reference). (See also Fellay et al, Science 317:944-947 (2007).)

EXAMPLE 1

Experimental Details

Criteria for inclusion Patients were eligible for the study if they had: (i) a valid seroconversion date estimation proven by biological markers:

1. a documented positive test and date and a documented negative test less than two years before the first positive test;

2. alternatively, subjects could be included if they showed one or more biological criteria of primary infection: incomplete western blot and/or positive p24 Ag and/or high viremia (>1 million copies per milliliter of blood) and a consistent dynamic pattern of the biological parameters (completion of western blot, negativization of p24 Ag, decrease of peak viremia) - a compatible clinical syndrome was considered supporting evidence;

3. a subset of individuals had long term spontaneous control of viral load below 1000 RNA copies/ml, and were included irrespective of the actual date of seroconversion; and

(ii) plasma HIV RNA determinations in the absence of antiretroviral treatment between 3 months and 3 years after seroconversion. Set point definition

First step: elimination of outlier VL on the basis of clinical or biological arguments

A. (only for Euro-CHAVI cohort) Visual inspection of individual data and definition of queries addresses to the clinical centers and laboratories.

Problematic points were then excluded if a satisfactory explanation was obtained, like acute disease or acute exacerbation of chronic disease, vaccination, immune-modulating treatment, major trauma or laboratory problems. B. (for both cohorts) Examination of the CD4 cells profile, with progression of HIV-I disease (defined by a decrease in CD4 cells below 350 cells/ml) leading to the elimination of simultaneous and subsequent VL results.

Second step: elimination of VL not reflecting the steady-state, through a computerized algorithm. Three types of outliers were identified, corresponding to the 3-phasic evolution of HIV-I viremia:

A. VL measured before the set point has been reached, part of the initial peak of viremia observed during primary HIV infection: they have to be measured during the first year after seroconversion and have a value >0.25 loglO higher than average of subsequent VL. B. VL measured during the accelerating phase of the disease (late ascending slope), which reflects the fact that rapid progressors can evolve into an advanced disease in a short period of time: for patients with a significantly ascending VL slope, only the first 3 results were kept for calculation of the set point. C. VL measured during the set point period, but conflicting with other available results; possibly linked to unreported interfering conditions, laboratory errors, transcription or data-management errors: defined as VL >0.51oglO higher or lower than average of all remaining points.

Third step: calculation of the set point as the average of all remaining VL results.

Cohorts The Center for HIV-AIDS Vaccine Immunology (CHAVI) is led by

Barton Haynes (Duke University, Durham NC, USA). Its Host Genetics Core is led by David Goldstein (Duke University, Durham NC, USA). CHAVI is founded by the National Institute of Allergy and Infectious Diseases (USA). The Euro- CHAVI consortium is coordinated by A. Telenti (University of Lausanne, Switzerland), with the help of S. Colombo (University of Lausanne, Switzerland) and J. P. A. Ioannidis (University of Ioannina, Ioannina, Greece). Participating Cohorts/Studies (Principal Investigators) are: Swiss HIV Cohort Study, Switzerland (P. Francioli); IrsiCaixa, Barcelona, Spain (B. Clotet); Clinics Hospital, Barcelona, Spain (J. M. Gatell); Danish Cohort, Denmark (N. Obel); Modena Cohort, Modena, Italy (A. Cossarizza); San Raffaele del Monte Tabor Foundation, Milan, Italy (A. Castagna); I.CO.NA Cohort, Rome, Italy (A. De Luca); Royal Perth Hospital, Perth, Australia (S. Mallal); Guy Kings St.Thomas Hospital, United Kingdom (P. Easterbrook). All participating centers provided local institutional review board approval for genetic analysis, and each participant provided genetic informed consent.

Quality control of data flow

The following quality control steps were taken to make sure genotyped were correctly called.

1. Infinium BeadStudio Raw Data Analysis All samples are brought into a single BeadStudio file and using the standard Illumina cluster file. An evaluation of the clustering of all samples is made. If a large number of samples do not fit the cluster for a random set of SNPs, a separate BeadStudio file is made for these samples. Once all files are made, any sample that has very low intensity or a very low call rate using the Illumina cluster (<95%) is deleted. All SNPs that have a call frequency below 100% are then reclustered. Any sample that is below a 98% call rate after the reclustering is deleted. Next, a "1% rule" is applied where all SNPs that have a call frequency below 99% are deleted. Any SNPs where more than 1% of samples are not called or are ambiguously called are deleted. It has been shown

(unpublished data) that SNPs with many samples not called (or potentially miscalled) can lead to false positives in statistical associations.

The reclustering step creates SNP calling errors (even with 1000 samples in the file) but a procedure has been identified to prevent the errant calls from being released in the final report. The SNP data is screened within BeadStudio by looking at two criteria. First, all SNPs with a cluster separation value below 0.3 are manually checked to ensure correct calls. Many of these SNPs can be manually fixed but some have to be deleted. Next, any SNP (excluding X chromosome SNPs) with a Het Excess value between -1.0 to -0.1 and 0.1 to 1.0 are evaluated to determine if the raw and normalized data show a clean call. Any SNP cluster that does not appear normal is deleted. This includes SNPs that appear to show a deletion (hemizygotes and homozygous deletion). This is done because these can be artifacts from either the chemistry or an interfering SNP during hybridization. These procedures resulted in a success rate of genotyping calls ranging from 97.5%-99% (13,709-5355 deleted SNPs). Two percent of samples were selected randomly to be genotyped twice independently for quality purpose. The concordance rate for duplicate genotyping was 99.99%. Also, ten SNPs from different chromosomes were re-genotyped using TaqMan assays. The concordance between the BeadChip and Taqman genotype calling was 100%. A total of 535 samples were run on the whole-genome chips. In total, 49 samples were excluded. A few of these were because of complete genotype failure (i.e. below 98%) but most were because of the high level of calling stringency (1% rule).

2. Minor allele frequency (MAF) check for data handling accuracy

This step performs a basic check of the data accuracy on the data flow pipeline from the output of the Illumina genotyping facility to the analytical process. Using PipeQC software, the MAF report from PLINK was checked against the original locus report generated by genotyping facility. A check was made that the two MAF reports match exactly.

3. Specification of gender

This step performs a check on the gender specification obtained from the phenotype database, using the observed genotypes of SNPs on chromosome X, and Y where available. All individuals who were marked as "male" but with significant amount of heterozygous X genotypes (>=1%), or who were marked as "female" but with high frequency of homozygous X genotypes (>=80%) or Y genotype readings, were individually inspected against original data source. If no satisfactory correction could be obtained, these individuals were excluded from further analyses. Four patients were excluded in this step.

4. Cryptic relatedness

This step performs a check on the cryptic relatedness between study participants. An estimate was made of the sharing of genetic information by estimating identity by descent (IBD) using the PLINK software. All pairs of DNA samples showing π>=0.125 (estimated proportion of alleles IBD) were individually inspected and one sample in each pair was excluded from further analyses. 7 samples were removed in this step.

5 5. Genotype missing

This step performs a check whether the genotype missing is skewed towards high or low phenotype values and hence may give rise to spurious association p-values. PLINK software was used to perform this check on the top SNPs discussed herein. No genotype data violated this check. 0

6. Low MAF

All SNPs with a MAF<0.006 were removed. This criterion ensured that at least 6 individuals of the rare genotype are present in the dataset, to control for error in the estimation of asymptotic p-values. 5

7. Hardy- Weinberg Equilibrium (HWE)

This step performs a check whether the observed genotype data deviate from HWE. This check was performed using PLINK software on the top SNPs. A deviation from HWE was defined with a criterion of P-value less than 0.05. No o genotype data violated this check.

8. Recheck of the genotyping quality

The top SNPs showing significant association were subject to a double check for their genotyping quality. This is an individual recheck on the raw and 5 normalized data to be sure that it is called correctly as described in "Infinium BeadStudio Raw Data Analysis" process. No genotype data violated this check.

Modified EIGENSTRAT method to control for stratification This method derives the principal components of the correlations among gene variants and corrects for those correlations in the association tests. In principle, therefore, the principal components in the analyses should reflect population ancestry. It has been noticed, however, that some of the leading axes appear to depend on other sources of correlation, such as sets of variants near one another that show extended association. The potential for inversions has been documented to create this effect and it may be created by other causes of extended linkage disequilibrium as well. For this reason, an inspection was made of the SNP 'loadings' for each of the leading axes to determine if they depended on many or relatively few SNPs, as would be expected if the given axis reflected population ancestry or a more localized linkage disequilibrium effect respectively. This analysis identified several axes clearly due to inversions and suggested that 17 axes should be retained for ancestry adjustment. Significance was, therefore, assessed using the 17 principal components emerging from the EIGENSTRAT analyses as covariates in a logistic regression model which also incorporated gender as a covariate.

EIGENSTRAT axes were selected for use as covariates to adjust for ancestry in subsequent linear regression analyses as follows:

1. To find EIGENSTRAT axes, a start was made with autosomal SNPs with MAF>0.01.

2. On inspection of SNP loadings for each PC axis (the "gamma" coefficients of Price et al (Nat. Genet 38:904 (2006)), several of the top axes were found to be dominated by a small number of SNPs all mapping to the same region of the genome. For example, one axis was found to be dominated by SNPs mapping to a region of chr8p22-23.1 coinciding with a known inversion polymorphism. 3. To correct for these LD effects, and ensure that EIGENSTRAT axes reflected only effects that applied equally across the whole genome (as ancestry effects should), principal components analysis was re-applied to a reduced SNP set in which (i) certain known high LD regions were excluded (chr8: 8000000..12000000, chr6:25000000..33500000, chrl 1 :45000000.-57000000, chr5:44000000..51500000); (ii) SNPs were thinned using the "--indep-pairwise" option in PLINK, such that all SNPs within a window size of 1500 (step size of 150) were required to have r2<0.2; (iii) Each SNP was regressed on the previous 5 SNPs, and the residual entered into the PCA analysis, as suggested by Patterson et al (Patterson et al, PLoS Genet. 2:el90 (2006)).

4. Inspection of SNP loadings on all axes deemed significant by the Tracy- Widom method of Patterson et al, using Q-Q plots against Normal expectation, now revealed no axes dominated by single high-LD regions of the genome.

5. Tracy Widom tests nominated the first 17 resulting PC axes as significant (p<0.05). The first 17 PC axes were, therefore, adopted as covariates in subsequent analyses.

Results

Success in genetic association studies is dependent on precisely defined and measured phenotypes. In this study, strict selection criteria were applied in the identification of participants, and the database was curated at the level of individual patient files. Patients were eligible for the study if they had (i) a valid seroconversion date estimation proven by biological markers and (ii) plasma HIV RNA determinations in the absence of antiretroviral treatment between 3 months and 3 years after seroconversion. Critically, all longitudinal VL data used in the study were individually inspected by an infectious disease clinician with experience in the management of HIV infected subjects. This inspection made it possible to identify atypical VL measures and to query the relevant clinics and labs to either explain the reading or remove the reading or subject. The pattern of change was also considered to determine whether a set point was actually achieved. (Among Euro-CHAVI sub-cohorts and studies, 676 patients were eligible according to the criteria. After inspection of viral load results, queries were defined for 195 of them. On the basis of the answers from the centers 42 patients were excluded because of imprecise assessment of set point.) An algorithm was developed intended to match the decisions made following inspection and reapplied it to all subjects (see Experimental Details above).

For patients with at least 4 CD4 cell count results, a progression phenotype was defined as the time to drop of CD4 cells below 350 per milliliter of blood or the time to antiretroviral treatment start, whichever came first. For patients that did not progress under this definition during the follow-up period, an evaluation was made as to whether there was evidence of CD4 decline (significantly decreasing CD4 slope determined by a simple regression). For those showing CD4 decline the estimated slope of the decline was used to extrapolate the time that CD4 counts would drop below 350 and this was used as the time to progression (Douek et al, Annu. Rev. Immunol. 21 :265 (2003)).

Those subjects that did not progress and that did not show significant CD4 decline were classified as non progressors and a separate case control comparison was constructed as progressors versus non progressors.

The Euro-CHAVI cohort, specifically created for this study, represents a consortium of 8 European and 1 Australian Cohorts/Studies that agreed to participate in the Host Genetic Core initiative of the Center for HIV/AIDS Vaccine Immunology (CHAVI). The CHAVI is a consortium of universities and academic medical centers established by the National Institute of Allergy and Infectious Diseases, part of the Global HIV Vaccine Enterprise. 676 patients have been selected from those cohorts on the basis of the above-mentioned criteria. For the present analysis, only subjects with self-reported Caucasian ancestry were considered: 486 patients were finally included, after elimination of individuals with insufficient quality phenotype (see above) or genotype (see Experimental Details). 386 of them were eligible for progression analysis (309 progressors and 77 non progressors).

All samples were genotyped using Illumina's HumanHap550 genotyping BeadChip with a total of 555,352 single nucleotide polymorphisms (SNPs). A series of quality control (QC) procedures were carried out which included checks for: (i) Infinium BeadStudio Raw Data Analysis quality; (ii) minor allele frequency consistency to assess data handling accuracy; (iii) mismatches between clinical and genetically inferred specification of gender; (iv) cryptic relatedness; (v) pattern of missing genotypes; (vi) low minor allele frequency (vii) Hardy- Weinberg Equilibrium violation; and (viii) visual inspection of genotyping quality for top SNPs (see Experimental Details). In total, 20,251 SNPs were removed based on these QC procedures. Methods were applied to identify deletions and targeted copy number variations (CNV) in the samples and to assess whether they influence the phenotype (using quality-control techniques designed to exclude failed SNPs, large stretches of failed SNPs were identified as putative deletion polymorphisms and examined for differences according to the phenotype; 42 such regions were checked, ranging in size from 700bp (4 SNPs) to 90kb (16 SNPs)). The core association analyses focused on single-marker allelic tests of the QC- passed SNPs, using linear regression. To control for the possibility of spurious associations resulting from population stratification, a modified EIGENSTRAT method was used (Price et al, Nat. Genet. 38:904 (2006)): principal component axes were integrated as co-variables in the regression model. Other covariates considered included gender, age, mode of HIV infection and year of HIV acquisition: only gender had an independent effect on set point (after inclusion of EIGENSTRAT covariates) and was therefore included in the linear regression model. To assess significance, the associations were compared to a conservative threshold based on a straight Bonferroni correction accounting for all 535,101 polymorphisms considered here (Pcutoff=9.3xlO-8) to declare genome-wide significance. Analyses incorporating both the SNP variation and classical HLA typing were carried out on a subgroup of 156 patients with available 4-digit HLA Class I allelic determination.

These analyses have identified two independently acting groups of polymorphisms, associated with classical HLA loci B and C, that explain 15% of the total variation in HIV-I VL at set point. A third set of polymorphisms upstream of the ZNRDl gene, encoding an RNA polymerase subunit, explains 5.8% of the total variation in clinical progression.

HCP5 - HLA-B*5701

One polymorphism located in the HLA Complex P5 (HCP5) gene explains 9.6% of the total variation in HIV-I set point despite a minor allele frequency of 0.05, testifying to the magnitude of the effect of the haplotype associated with this SNP (dbSNP rs2395029, p=9.36e-12). The median VL at set point is 19344 copies per milliliter of blood (cp/ml) for patients homozygous for the common allele (TT, N=439), whereas it is 592 cp/ml for patients with one (TG, N=45) or two copies (GG, N=2) of the minor allele (Fig. 1), and at p=9.36e-12, the association is comfortably genome-wide significant.

The HCP5 gene is located 100kb centromeric from HLA-B on chromosome 6 (Fig. 2), and the associated variant is known to be in high linkage disequilibrium (LD) with the HLA allele B5701 (de Bakker et al, Nat. Genet. 38: 1166 (2006)) (r²=l in our dataset). This particular HLA-B allele has the strongest described protective impact on HIV-I disease progression (Migueles et al, Proc. Natl. Acad. Sci. USA 97:2709 (2000)) and has been associated with low HIV-I viral load (Altfeld et al, AIDS 17:2581 (2003)). Thus, it is demonstrated that, in the context of whole genome influences, the HLA-B5701 is indeed the strongest host genetic factor restricting HIV-I infection through a direct effect on early viral load (Altfeld et al, PLoS Med. 3 :e403 (2006)).

Given the strong functional data supporting a role for HLA-B5701 in restricting HIV, the first hypothesis must be that the association observed here is due to the effect of HLA-B5701 reflected in its tagging SNP within HCP5 (de Bakker et al, Nat. Genet. 38:1 166 (2006)). However, genetics allows no resolution on whether this effect is exclusively due to B5701 or if HCP5 variation also contributes to the control of HIV-I . In fact, HCP5 itself is also a novel and strong candidate for contributing to HIV-I control. HCP5 is a member of a human endogenous retrovirus family (HERV) with sequence homology to retroviral Pol genes (Kulski et al, Immunogenetics 49:404 (1999)). Since expression of the , gene has been documented in lymphocytes (Vernet et al, Immunogenetics 38:47 (1993)), a potential antisense activity against retroviruses has been suggested (Kulski et al, Immunogenetics 49:404 (1999)). Moreover, HCP5 is predicted to encode two proteins and the associated polymorphism results in an amino acid substitution in one of these. A model in which the newly-associated HCP5 variant and the HLA-B5701 allele have a combined haplotypic effect on HIV-I set point is consistent with the observation that suppression of viremia can be maintained in B5701 elite controllers even after HIV-I undergoes mutations that allow escape from cytotoxic T-lymphocytes (CTL) mediated restriction (Bailey et al, J. Exp. Med. 203:1357 (2006)). Perhaps most interestingly, B5701 patients consistently show a lower peak of viremia and present less frequently with symptoms during acute HIV-I infection (Altfeld et al, AIDS 17:2581 (2003)), suggesting control before the time of a maximal CTL response (McMichael et al, Nature 410:980 (2001 )). It is therefore an urgent priority to assess whether HCP5 contributes to the control observed in B5701 positive patients and cell types.

HLA-C The second most significant result in the association study, rs9264942, is located in the 5' region of the HLA-C gene, 35 kb away from transcription initiation (Fig. 2). This SNP explains 6.5% of the observed variation in VL at set point (Fig. 1) and also shows a genome-wide significant association (p=3.77e-09). Strikingly, this same SNP associates with dramatic differences in HLA-C expression levels in independent populations (p=l .69e-07, The Sanger Institute Genevar expression database, created on HapMap samples (Stranger et al, PLoS Genet. 1 :e78 (2005)): www.sanger.ac.uk/humgen/genevar). The protective allele leads to a lower viral load and is associated with higher expression of the HLA-C gene. This strong and independent association with HLA-C expression levels suggests, for the first time, that genetic control of expression levels of a classical HLA gene influences viral control.

Further support for such a role of HLA-C expression levels in restricting HIV-I comes from a second, nearby SNP with an even higher impact on HLA-C expression (p=4.42e-08) that was identified in the same database (Fig. 2). The second variant, rs6457374, is located 3 kb nearer the HLA-C gene (-32 kb in 5' region) and has an independent effect on HLA-C expression and also associates with HIV-I set point, but not independently of rs9264942.

As noted above, two SNPs in HLA-C 5' region (rs9264942 and rs6457374) were found to have high association with HLA-C expression levels in HapMap CEU samples. Based on linear regression on normalized HLA-C expression levels in 60 CEU parents, the allelic regression coefficient (against genotype coded [0,1,2]) and p-value for each SNP fitted separately are: rs9264942 (beta=0.35, p=3.38e-07), rs6457374 (beta=-0.29, p=4.42e-08). When rs9264942 is in the regression model, rs6457374 significantly improves fit when added to the model (p=0.0008, R-squared improved from 0.36 to 0.46). When rs6457374 is in the regression model, rs9264942 significantly improves fit when added to the model (p=0.005, R-squared improved from 0.39 to 0.46). Based on linear regression on HIV-I set point in Euro-CHAVI cohort, with gender and significant eigenstrat axes as covariates, the allelic regression coefficient and p-value for each SNP fitted separately are: rs9264942 (beta=-39, p=3.77e-09), rs6457374 (beta=-0.32, p=2.68e-04). When rs9264942 is in the regression model, rs6457374 does not improve fit when added to the model (p=0.27, R-squared=0.15 in both cases). When rs6457374 is in the regression model, rs9264942 significantly improves fit when added to the model (p= 1.46e-06, R-squared improved from 0.10 to 0.15).

To better describe the relationship between HLA-C expression levels and viral control, each of the possible diploid genotypes involving rs9264942 and rs6457374 were considered and the average mRNA levels for each genotype observed in the Genevar database (CEU parents, N=60) was compared to the average viral load observed for each genotype in the study. These data strongly suggest a non-linear effect of increasing expression levels on VL. At lower HLA- C expression levels variation in expression does not change viral control. But once a threshold expression level is reached, increasing HLA-C expression sharply reduces viral load at set point.

Although these data make a strong case for a causal role for HLA-C expression levels, extensive LD throughout the HLA region makes it necessary to directly test whether the real causal variants could be elsewhere. Specifically, nested regression models were used to assess whether the observed association could be caused by already described functional alleles at HLA-A, HLA-B, and HLA-C. Strikingly, the HLA-C expression SNP shows significant association with HLA-B5701, B27, B35Px, as well as with the HLA allelic groups Bw4 & Bw6. In each case, however, while the HLA-C expression variant can explain the effect of these alleles on HIV-I set point, the reverse is not true. When a linear regression model includes first the HLA allele (or group of alleles), addition of rs9264942 results in a significant increase in the explained variation (see Table 1). On the other hand, none of the HLA alleles considered, with the exception of HCP5/B5701, add significantly to a model that already incorporates the HLA-C variant (Table 1).

Table 1. The impact of HLA-C 5' expression polymorphism rs9264942 on set point is independent of its association with HLA-B alleles previously implicated in HIV-I control. The addition of rs9264942 to the linear regression model improves fit significantly for all HLA-B alleles or groups of alleles that are supposed to have an influence on HIV disease, as shown in Table Ia. In contrast, only HLA-B5701 has an independent impact after taking into account rs9264942 effect. The independency of HLA-C is also clearly seen in the mean values of HIV-I set point for each rs9264942 genotype (Table Ib): the minor allele C is associated with a decrease in VL independently of all considered alleles and groups of alleles. Numbers refer to a subgroup of 156 patients with available 4-digit HLA Class I allelic results, a.

Addition ofrs9264942 to models Addition of HLA alleles to a with HLA alleles model with rs9264942 change in R2 p-value change in R2 p-value

HLA-B27 0.08 7.20E-05 0 0.22

HLA-B5701 0.06 4.30E-04 0.05 1.20E-03

HLA-B35px 0.09 4.70E-05 0 0.84 all 3 above 0.05 2.00E-03 0.05 1.OOE-03

Bw4 group 0.07 3.30E-04 0 0.33

Bw6 group 0.07 1.90E-04 0 0.44 b. rs9264942 genotype N Mean SD

TT 58 4.34 0.86

All patients TC 72 3.76 1.22 CC 26 3.07 1.35

57 4.37 0.84

Patients without HLA-B27 TC 65 3.80 1.24 CC 22 3.08 1.30

58 4.34 0.86

Patients without HLA-B5701 TC 63 3.88 1.13 CC 21 3 32 1 35

TT 48 4 28 0 89

Patients without HLA-B35Px TC 69 3 79 1 20

CC 26 3 07 1 35

TT 47 4 31 0 87

Patients without any of the above alleles TC 54 3 94 1 14

CC 18 3 26 1 34

TT 36 4 42 0 88

Patients without Bw4 specificities TC 15 4 23 0 96

CC 4 2 29 0 43

TT 6 3 88 0 93

Patients without Bw6 specificities TC 18 3 65 1 32

CC 15 3 08 1 47

This observation is intriguing in the global context of the hitherto proposed role of HLA-C in HIV-I pathogenesis. HIV-I nef selectively down regulates the expression of HLA-A and -B but not of HLA-C on the surface of infected cells (Cohen et al, Immunity 10:661 (1999)). Originally, this strategy was considered advantageous for the virus because HLA-A and -B present foreign (notably viral) epitopes to CD8 T-cells resulting in cell destruction, whereas HLA-C binds self peptides and interacts with natural killer cells (NK) in order to avoid NK attack. However, HLA-C also has the ability to present viral peptides to cytotoxic CD8+ T cells and consequently restrict HIV-I (Goulder et al, AIDS 11 : 1884 (1997)); ten HLA C-restricted CTL epitopes have been described in the LANL database (www.hiv.lanl.gov). These observations suggest that there could be a threshold in expression above which HLA-C mediated viral restriction becomes an effective defense mechanism against HIV-I . In such a scenario, the natural incapacity of HIV-I nef to down regulate HLA-C molecules becomes an important advantage for the immune system.

Other set point variants

No other single marker reached genome significance after correction for multiple testing, and none of the identified copy number variations showed any association with HIV-I VL at set point. However, it is expected that other SNPs among those with very low p-values are also real effects. Therefore, the SNPs with the lower 100 p-values and their respective genes (i.e. the closest gene, as annotated in Illumina's HumanHap550 SNPs list) are reported in Table 2. Furthermore, specific gene variants were examined on the basis of their confirmed or suspected link with HIV-I biology: SNPs and genes of interest are listed in Table 3 (for known or suspected HIV-I restricting genes), Table 4 (for host cellular factors interacting with HIV-I , Ref (Goff, Nat. Rev. Micro. 5:253 (2007)) and Table 5 (for groups of genes according to Gene Ontology categories). This approach reveals some potentially interesting candidates with a lesser association with HIV-I VL at set point.

Table 2. SNPs with the top 100 p-values in set point association study

SNP CHR P-value Coordinate Gene Gene ref number Location Position rs2395029 6 9 36E-12 31539760 HCP5 NM_006674 coding NS [335/63] rs9264942 6 3 78E-09 31382360 HLA-C NM_002117 flankιng_5UTR -34526 rs 13207315 6 3 79E-07 31349107 HLA-C NM_002117 flankιng_5UTR -1273 rs 10484554 6 8 06E-07 31382535 HLA-C NM_002117 flankιng_5UTR -34701 rs9368699 6 1 20E-06 31910521 C6orf48 NM_016947 flankιng_5UTR -152 rs13191519 6 1 51E-06 31373732 HLA-C NM_002117 flankιng_5UTR -25898 rs2894207 6 1 82E-06 31371731 HLA-C NM_002117 flankιng_5UTR -23897 rs13216197 6 3 46E-06 31378998 HLA-C NM_002117 flankιng_5UTR -31164 rs2248462 6 3 61 E-06 31554776 HCP5 NM_006674 flankιng_3UTR -13316 rs2516509 6 3 61E-06 31557974 MICB NM_005931 flankιng_5UTR -15971 rs2516513 6 3 61E-06 31555568 HCP5 NM_006674 flankιng_3UTR -14108 rs2249742 6 5 09E-06 31348701 HLA-C NM_002117 flankιng_5UTR -867 rs13210132 6 5 35E-06 31109123 C6orf205 NM_001010909 flankιng_3UTR -43469 rs2523619 6 6 33E-06 31426124 HLA-B NM_005514 flankιng_3UTR -3506 rs3815087 6 7 09E-06 31201567 PSORS1C1 NM_014068 5UTR [135/28] rs9263715 6 7 09E-06 31203781 PSORS1C1 NM_014068 intron -1544 rs 12207951 6 8 26E-06 31580439 MICB NM_005931 intron -934 rs2534678 6 8 26E-06 31571943 MICB NM_005931 flankιng_5UTR -2002 rs11623538 14 840E-06 61720581 SYT16 NM_031914 flankιng_3UTR -83403 rs4131373 1 9 14E-06 3830546 LOC339448 NM_207356 coding S [256/232] rs481830 6 9 29E-06 167152689 RPS6KA2 NM_001006932 flankιng_3UTR -89410 rs12082157 1 1 01E-05 3834968 LOC339448 NM_207356 intron -2251 rs2441130 5 1 04E-05 65123177 NLN NM_020726 intron -860 rs12745720 1 1 10E-05 13985417 PRDM2 NM_012231 flankιng_3UTR -88536 rs3823418 6 1 11E-05 31208922 PSORS1C1 NM_014068 flankιng_3UTR -3521 rs2473710 6 1 17E-05 23127205 HDGFL1 NM_138574 flankιng_3UTR -447334 rs 10793975 9 1 24E-05 130313865 ASS NM_054012 flankιng_5UTR -35961 rs9523736 13 1 25E-05 92082785 GPC5 NM_004466 intron -233751 rs9468932 6 1 26E-05 31372803 HLA-C NM_002117 flankιng_5UTR -24969 rs13416716 2 1 37E-05 185737524 C2orf10 NMJI 94250 flankιng_3UTR -107806 rs3131003 6 1 38E-05 31201462 PSORS1C1 NM_014068 5UTR [30/133] rs1891060 1 1 41 E-05 210357750 PROX1 NM_002763 flankιng_5UTR -192505 rs7539708 1 1 42E-05 172301131 TNR NM_003285 flankιng_5UTR -193543 rs2524123 6 1 53E-05 31373294 HLA-C NM_002117 flankιng_5UTR -25460 rs7190491 16 1 60E-05 82850137 KCNG4 NM_172347 flankιng_5UTR -19280 rs327263 5 1 73E-05 38689392 LIFR NM_002310 flankιng_5UTR -58139 rs12665286 6 1 88E-05 169426889 THBS2 NM_003247 flankιng_3UTR -6619 rs7728604 5 2 06E-05 168152681 SLIT3 NM_003062 intron -2409 rs2473707 6 2 23E-05 23110312 HDGFL1 NM_138574 flankιng_3UTR -430441 rs10412244 19 2 67E-05 21994007 ZNF208 NM_007153 flankιng_5UTR -30455 rs3103309 1 2 88E-05 151083951 HAX1 NM_001018837 flankιng_3UTR -22529 rs2291490 8 3 06E-05 82829767 CHMP4C NM_152284 intron -393 rs1235162 6 3 07E-05 29645204 UBD NM_006398 flankmg_5UTR -9523 rs 10500052 7 3 23E-05 115050925 TFEC NM_001018058 flankιng_3UTR -118228 rs7300317 12 3 27E-05 50918841 KRT7 NM_005556 intron -15 rs 1003921 11 3 39E-05 17756913 KCNC1 NM_004976 flankιng_3UTR -6160 rs1280102 4 3 59E-05 187880915 FAT NM_005245 flankιng_3UTR -3172 rs9348876 6 4 10E-05 31683256 AIF1 NM_001623 flankιng_5UTR -7756 rs2760650 9 4 18E-05 6978476 JMJD2C NM_015061 intron -1810 rs 10736862 9 4 23E-05 130312810 ASS NM_054012 flankιng_5UTR -37016 rs722410 15 4 32E-05 20475539 CYFIP1 NM_014608 intron -1679 rs17823746 20 4 36E-05 24539029 C20orf39 NM_024893 intron -25400 rs2577335 10 4 39E-05 121365172 TIAL1 NM_003252 flankιng_5UTR -19160 rs 1476445 22 4 57E-05 17995566 SEPT5 NM_002688 flankιng_5UTR -80975 rs9536542 13 4 73E-05 53249531 OLFM4 NM_006418 flankιng_3UTR -725344 rs2410194 8 4 92E-05 14460901 SGCZ NM_139167 flankιng_5UTR -4095 rs11878567 19 4 98E-05 5463812 PLAC2 NM_153375 flankιng_3UTR -46552 rs2284178 6 5 30E-05 31540105 HCP5 NM_006674 3UTR [281/1355] rs12889327 14 5 40E-05 94492967 DICER1 NM_030621 flankιng_3UTR -129352 rs11238646 10 5 81E-05 43527382 ZNF32 NM_006973 flankιng_5UTR -63050 rs9407760 9 5 92E-05 16353461 BNC2 NM_017637 flankιng_3UTR -52941 rs2760660 9 5 96E-05 6983849 JMJD2C NM_015061 intron -3325 rs3737473 18 6 04E-05 31973043 STATI P 1 NM_018255 intron -210 rs3094204 6 6 16E-05 31199972 PSORS1C1 NM_014068 flankιng_5UTR -1460 rs7774154 6 6 26E-05 23105043 HDGFL1 NM_138574 flankιng_3UTR -425172 rs1390786 15 6 40E-05 90933773 LOC400451 NM_207446 flankιng_3UTR -27912 rs4813524 20 6 58E-05 24560809 C20orf39 NM_024893 intron -33173 rs8321 6 6 68E-05 30140502 ZNRD1 NM_014596 3UTR [48/162] rs3132685 6 6 73E-05 30053929 HCG9 NM_005844 intron -45 rs 17797872 8 6 90E-05 85912085 LOC 138046 NM_173848 intron -12685 rs919214 12 7 02E-05 100673246 GNPTAB NM_024312 intron -3152 rs3094212 6 7 09E-05 31193750 CDSN NM_001264 intron -465 rs2158276 4 7 15E-05 7757978 SORCS2 NM_020777 intron -148 rs1062470 6 7 36E-05 31192415 CDSN NM_001264 coding S [634/870] rs995906 4 7 37E-05 94449262 GRID2 NM_001510 intron -36471 rs7894582 10 7 41E-05 115694449 NHLRC2 NM_198514 flankιng_3UTR -36007 rs1379868 19 7 46E-05 5778098 NRTN NM_004558 intron -662 rs9257809 6 7 64E-05 29464311 OR12D2 NM_013936 flankιng_5UTR -8145 rs1791403 18 7 65E-05 41969505 CCDC5 NIVM 38443 flankιng_3UTR -7209 rs10738377 9 7 78E-05 1₄₇477₄5 FREM1 NM_144966 intron -7471 rs9910192 17 7 80E-05 68361159 SLC39A11 NM_139177 intron -3621 rs3094205 6 7 84E-05 31199842 PSORS1C1 NM_014068 flankιng_5UTR -1590 rs4238521 15 7 87E-05 78583184 ARNT2 NM_014862 intron -4368 rs1357011 2 8 01 E-05 67649324 ETAA 16 NM_019002 flankιng_3UTR -99997 rs12037583 1 8 26E-05 172324963 TNR NM_003285 flankιng_5UTR -217375 rs3130380 6 8 27E-05 30387110 TRIM39 NM_021253 flankιng_5UTR -15909 rs3749971 6 8.31 E-05 29450755 OR12D3 NM_030959 coding NS [662/288] rs3822272 4 8 40E-05 111041155 IF NM_000204 intron -703 rs12881250 14 8.77E-05 94495465 DICER1 NM_030621 flankιng_3UTR -126854 rs999229 7 9 04E-05 137858052 ATP6V0A4 NM_130841 intron -171 rs7016388 8 9.13E-05 121171047 COL14A1 NM_021110 flankιng_5UTR -35486 rs2248617 6 9.22E-05 31556513 HCP5 NM_006674 flankιng_3UTR -15053 rs2395488 6 9.22E-05 31553889 HCP5 NM_006674 flankιng_3UTR -12429 rs2516424 6 9.22E-05 31556295 HCP5 NM_006674 flankιng_3UTR -14835 rs 1350341 18 9.23E-05 55993514 MC4R NM_005912 flankιng_3UTR -196050 rs2051311 18 9.37E-05 55987861 MC4R NM_005912 flankιng_3UTR -201703 rs 13386401 2 9.57E-05 185690907 C2orf10 NM_194250 flankιng_3UTR -61189 rs2516400 6 9.74E-05 31589085 MICB NM_005931 flankιng_3UTR -2206 rs9261290 6 0.0001002 30146627 RNF39 NM_170770 coding NS [87/194] rs8088744 18 0.0001027 64685793 C18orf14 NM_024781 intron -7093

Table 3. Analysis of SNPs in genes coding for products that are known to have an influence on HIV-I disease.

SNP with lowest riUfuDBr Lowest Bonferroni

Gene nf ^Λ/Pς raw p-value raw p- corrected

(chromosome) value p-value

CCR5.CCR2 3 rs4513489 (3) 2.14E-01 6.43E-01

TRIM5 13 rs12808061 (11 ) 3.48E-01 1

CXCL12 145 rs12266297 (10) 1.19E-02 1

TSG 101 10 rs2132302 (11 ) 1.96E-01 1

CCL5 1 rs16963927 (17) 4.66E-01 4.66E-01

IL10 14 rs4072227 (1 ) 2.69E-03 3.77E-02

CX3CR1 25 rs1513253 (3) 2.12E-02 5.30E-01

APOBEC3G 3 rs8177832 (22) 3.10E-02 9.29E-02

CUL5 16 rs7118335 (11 ) 8.89E-02 1

CCL3 3 rs9972960 (17) 2.59E-01 7.78E-01

IL4 4 rs2243290 (5) 8.56E-02 3 43E-01

Above all 237 rs4072227 (1 ) 2.69E-03 6.38E-01

Table 4 Analysis of SNPs in genes coding for host cellular factors involved in HIV-I biology (Goff SP Host factors exploited by retroviruses Nat Rev Microbiol 2007 Apr,5(4) 253-63)

Boπferroni

Number of SWP with lowest raw Lowest raw SNPs p-value(chromosome) p-value p-value

CD4 13 rs11064404 (12) 9 21 E-03 1 20E-01

SLC2A1 37 rs16830039 (1 ) 1 12E-01 1

CRAT 2 rs7866897 (9) 5 99E-02 1 20E-01

CXCR4 198 rs10197003 (2) 9 09E-03 1

PPIA 1 rs17134462 (7) 8 10E-01 8 10E-01

DYNEIN (37 genes) 714 rs10497770 (2) 1 38E-03 9 85E-01

ACTIN (13 genes) 200 rs728614 (1 ) 6 13E-03 1

UBE2I 6 rs2268049 (16) 5 02E-01 1

PIAS4 5 rs735842 (19) 2 11 E-01 1

SUMO (4 genes) 12 rs2177086 (2) 2 96E-02 3 56E-01

PSIP1 6 rs7856318 (9) 2 60E-01 1

ATM 10 rs11212570 (11 ) 2 70E-01 1

ATR 13 rs2229032 (3) 2 53E-02 3 28E-01

XRCC5 18 rs1364726 (2) 4 95E-02 8 90E-01

LIG4 239 rs1341368 (13) 2 20E-02 1

CCNT1 2 rs3803025 (12) 6 27E-01 1

CDK9 2 rs7039798 (9) 6 46E-02 1 29E-01

HRB 14 rs6753666 (2) 2 01 E-01 1

RANBP1 1 rs2238798 (22) 9 51 E-01 9 51 E-01

KHDRBS1 5 rs12094507 (1 ) 1 26E-01 6 32E-01

XPO1 29 rs4535028 (2) 1 13E-01 1

FLNB 44 rs1658367 (3) 3 06E-02 1

DHX9 4 rs7533385 (1 ) 3 36E-01 1

ETF 1 4 rs3849046 (5) 3 89E-01 1

ABCE1 1 rs9998052 (4) 4 89E-01 4 89E-01

SMARCB1 9 rs738796 (22) 9 09E-02 8 18E-01

STAUFEN.STAU 1 ,STAU2 59 rs6992006 (8) 2 12E-01 1

NEDD4L 161 rs17235322 (18) 8 06E-03 1

HGS 3 rs6565620 (17) 1 60E-01 4 80E-01

ABI2 6 rs2255047 (2) 4 77E-02 2 86E-01

TFAP2A 63 rs4710929 (6) 5 11E-02 1

VPS4A 6 rs3852689 (16) 2 13E-02 1 28E-01

M6PRBP1 18 rs2602699 (19) 1 12E-02 2 01 E-01

Above all 1905 rs10497770 (2) 1 38E-03 1 Table 5 Analysis of SNPs in groups of genes, according to immune and inflammatory terms of Gene Ontology (GO).

SNPs with lowest raw p- Lowest Bonferroni

Number Number

Term value (chromosome, raw p- corrected of SNPs of genes aene) value p-value

B cell mediated immunity 2 rs1863332 (2, MSH2) 1.25E-01 1 T-helper 1 type immune

320 9 rs694428 (9, TLR4) 3.86E-03 1 response T-helper 2 type immune

8 ^C4^J 4 rs4072227 (1 , IL10) 2.69E-03 2.26E-01 response ^X anti-inflammatory response 79 6 rs7580658 (2, PROC) 6.74E-03 5.32E-01 Cellular defense response 899 62 rs4660126 (1 , LYST) 4.77E-04 4.29E-01 defense response 1494 74 rs2577335 (10, TIAL1) 4.39E-05 6.56E-02 Defense response to Gram-

10 2 rs606119 (19, AZU1 ) 3.19E-01 1 negative bacteria Defense response to Gram-

61 6 rs7223092 (17, SPACA3) 9.63E-02 1 positive bacteria defense response to

720 57 rs3093662 (6, TNF) 1.33E-04 9.61 E-02 bacteria defense response to fungi 121 11 rs903769 (12, DCD) 3.38E-03 4.09E-01 defense response to

22 2 rs16910631 (12, CLEC7A) 5.99E-02 1 pathogenic protozoa defense response to virus 168 5 rs10282829 (8, BNIP3L) 3.99E-03 6.70E-01 formation of immunological

175 1 rs261039 (5, DOCK2) 2.19E-03 3.84E-01 synapse humoral defense mechanism

65 8 rs7356880 (6, HLA-DRA) 9.61 E-04 6.25E-02

(sensu Vertebrata) humoral immune response 980 29 rs3093662 (6, TNF) 1.33E-04 1.31 E-01 immune cell activation 30 3 rs3829223 (11 , TOLLIP) 3.53E-03 1.06E-01 immune cell chemotaxis 53 4 rs4072227 (1 , IL 10) 2.69E-03 1.43E-01 immune cell migration 325 9 rs3093662 (6, TNF) 1.33E-04 4.34E-02 immune response 6627 343 rs13207315 (6, HLA-C) 3.79E-07 2.51 E-03 immunoglobulin secretion 100 4 rs6568686 (6, TRAF3IP2) 5.75E-02 1 immunological synapse 336 8 rs11629129 (14, GZMB) 3.12E-03 1 inflammatory cell apoptosis 110 2 rs7085850 (10, FAS) 3.88E-02 1 inflammatory response 5469 200 rs9348876 (6, AIF 1 ) 4.10E-05 2.24E-01 innate immune response 1010 44 rs2980937 (8, DEFB1 ) 1.52E-03 1 negative regulation of

112 3 rs604045 (1 , TGFB2) 6.79E-03 7.61 E-01 immune response negative regulation of

248 9 rs6602392 (10, IL2RA) 1.22E-02 1 inflammatory response positive regulation of

159 6 rs7356880 (6, HLA-DRA) 9.61 E-04 1.53E-01 immune response positive regulation of

3 1 rs6848139 (4, IL2) 7.35E-01 1 immunoglobulin secretion positive regulation of

184 6 rs9892886 (17, PRKCA) inflammatory response 3.41 E-03 6.27E-01 positive regulation of innate 63 4 rs7674887 (4, EREG) 1.12E-02 7.07E-01 immune response regulation of immune

389 11 rs731874 (15, SMAD3) 8.74E-03 1 response regulation of immunoglobulin

31 4 rs3093662 (6, TNF) 1.33E-04 4.14E-03 secretion regulation of inflammatory

90 4 rs7340621 (3, BCL6) 2.42E-02 1 response All above 16003 697 rs13207315 (6, HLA-C) 3.79E-07 6.07E-03

ZNRDl and progression

The strongest association with clinical progression includes a set of seven polymorphisms located in and near the ring finger protein 39 (RNF39) and the zinc ribbon domain containing 1 (ZNRDl) genes, respectively (rs9261 174, rs3869068, rs2074480, rs7758512, rs9261 129, rs2301753 and rs2074479). This group of polymorphisms explains 5.8% of the variation in clinical progression and is marginally shy of genome-wide significance (p=3.89e-07). It also associates modestly with viral load at set point (p=7.1 le-03). Although again in the MHC region, these variants are >1 MB centromeric from the previous candidate SNPs.

Using the Genevar expression database, ZNRDl expression is observed to be significantly associated with the identified SNPs (p=2.0e-03) and notably 2 of them are located in the putative regulatory 5' region 25 and 32 kb away from the gene (rs3869068 and rs9261 174, respectively). As ZNRDl encodes an RNA polymerase subunit, a possible interaction with HIV-I during transcription is the most plausible causal mechanism if this gene indeed restricts HIV-I . Efficiency in provirus transcription is highly variable among individuals. In one study, differences in transcription efficiency alone accounted for 64 to 83% of the total variance in virus production that was attributable to post-entry cellular factors (Ciuffi et al, J. Virol. 78:10747 (2004)).

Although the second gene, RNF39, is poorly characterized, it cannot be ruled out as a candidate as two of the associated polymorphisms result in synonymous and amino acid changes (rs2301753 and rs2074479, respectively). No other relevant association was observed for the quantitative progression phenotype or for the case control comparison of progressors versus non progressors.

Confirmation in an independent sample

In order to confirm the effect of these polymorphisms on viral control, a replication cohort of 140 Caucasian patients was established. Representative polymorphisms were selected for each of the associates reported above (HCP5 rs2395029; HLA-C 5' region rs9264942; ZNRDl 5' region rs9261174) and these were genotyped in a replication cohort. For each of the reported associations, a replication in the same direction was observed (for HCP5: p=0.014, explains 4% of the variation in set point in this dataset; for HLA-C: p=2.8e-03, explains 5.6% of the variation in set point; for ZNRDl: p=0.048, explains 3.5% of the variation in progression). In summary, provided above are the results of the first comprehensive whole-genome association study of human genetic influence on HIV-I disease. It is shown that HLA-C expression has a major impact on the control of HIV-I. It is possible that a HERV derived gene can contribute to the control that is known to associate with the HLA-B5701 allele. The findings confirm the central role of the MHC region in HIV-I restriction and provide new ways to consider its mode of action: it is probably time to move beyond classical allelic discrimination in HLA region. It is also noteworthy that this genome-wide study of host determinants has two whole-genome significant hits, and succeeded in replicating all three of the strongest associations with viral load and progression. This indicates that determinants of host response can often include major effect gene variants, and suggests a degree of urgency in carrying out similar studies for other infectious diseases. Finally, the results described above suggest two possible directions for therapeutic intervention. First, they indicate the possibility that HCP5 can contribute to the control associated with HLA-B5701 which, if true, would present immediate therapeutic opportunities. Perhaps more importantly, it is shown that HLA-C restriction can constitute an important part of the control of HIV-I at sufficiently high expression levels of HLA-C. The latter result implicates HIV-I nef in determining virulence through differential downregulation of HLA class I molecules, limiting the function of HLA-A and -B alleles but highlighting HLA-C because it is resistant to nef. Future vaccine strategies could target HLA-C restricted T cell responses.

EXAMPLE 2

EXPERIMENTAL DETAILS PennCNV PennCNV (http://www.neurogenome.org/cnv/penncnv/) is a hidden

Markov model (HMM) based approach for detection of CNVs from Illumina high-density SNP genotyping data. To call the CNVs, it uses multiple sources of information, including total signal intensity and allelic intensity ratio at each SNP marker, the distance between neighboring SNPs, the allele frequency of SNPs and the pedigree information where available. The input is the BeadStudio file raw intensity data provided after curation. CHAVI files were run with the 2008 March 1 1 version of PennCNV, using the automatic waviness correction to reduce sample failure by QC.

QC Summary In total, 2276 samples were run. Of these, 1 192 (52%) 'failed' the log R ratio QC, (LRR_SD> 0.20). Between 0.2 and 0.28, the samples are usable with caution, as there is expected to be a high number of false positives. Of the 1192 failed samples, only 196 had an LRR_SD >0.280, which means they cannot be used. A further 4 samples were dropped because their waviness factor was unacceptable, even though all these samples were run with their new wave adjustment procedure. All results, therefore, will be visually checked (for both type I and type II errors) in BeadStudio to confirm CNV status.

Genome-wide Analysis

Using new Goldstein pipeline commands, the rawCNV output was converted to a file usable in plink for genome-wide analysis. Separate files were created for deletions and duplications, representing a homozygous CNV as BB, a heterozygous CNV as AB and a homozygous CNV as AA. The WGA analysis was then run on the white samples using an additive linear regression model including sex, age and 11 Eigenstrat axes as covariates.

RESULTS

KIR Locus

One of the most significant associations with HIV progression was a deletion at chromosome 19 ql3.42 (p=0.0010). Closer analysis of the region revealed that there were both deletions and duplications in this same position, and that the locus was that of the KIR genes (Fig. 5). The deletion was present in 2.8% of the sample, and the duplication in 3.7%, and mapped approximately to chrl9:60,014,787-60,018,551 (about 3.8kb).

Association with HIV Phenotypes

Using an additive model in a linear regression with gender, age, and 11 eigenstrat axes as covariates, the associations with setpoint was p=0.002, and the locus explains 0.7% of the variability. The association with time to progression was p=0.02, and the locus explains 0.8% of the variability. It also associated with elite controller status (logistic regression, same covariates): p=0.028, OR=3.5 (Table 6, Fig. 6). There was found to be no interaction with the previously associated SNPs.

Table 6 Association of CNV status at the KIR2DL locus, using an additive genetic model by copy number

Deletion 4.32 38 1166 21 0

No CNV 4.00 1254 1595 659 19.1 2.3

Duplication 3.71 50 2111 25 26.0 8.0

Linkage Disequilibrium ofCNVwith known SNPs The linkage disequilibrium of the CNV with all HumanHap550 SNPs (and their HapMap proxies) was checked for 500kb either side of the CNV (Fig. 7). There was no SNP in very strong r² with either CNV. For the deletion, the r² peaked at rs3859532, which was in an intergenic region just upstream of the LILRA5 gene. For the duplication, the r² peaked at 0.057 at rs633522, an intronic SNP in KIR3DP1. It then dropped down to 0.017 at rs618835, a non-synonymous coding SNP in KIR2DL4 (Tyr53Cys), and then to below 0.01. It should be noted that the SNP coverage within the KIR locus is very sparse and there could be strongly associated SNPs within this region that are not detectable using the HumanHap550 chips. For both the deletion and the duplications, the D' extended throughout the 1Mb region examined. CNV Endpoints

Figure 8 depicts the KIR gene locus (from Martin & Carrington, Methods MoI. Biol. 415:49-64 (2008)). Because the CNVs were called using SNP data, it is not possible to get the exact endpoints of the events, so it is not clear how the deletions and duplications affect the different genes shown here. Figure 7 illustrates that there is a gap in SNP coverage, which was increased when the HumanHap550 chip was upgraded from version 1 (13 SNPs in this region) to version 3 (6 SNPs in this region). The reason for the lost SNPs is not clear, though it is possible that these relatively common CNVs were causing dropped genotype calls and so they were removed for QC purposes. The CHAVI samples of African origin were run on the 1 million chip, which had an extra 32 SNPs in this region. Figure 9 shows the CNV calls on the 1 million SNP chip. It is not clear whether these are population differences between samples of African and European origin but there is currently no reason to believe that the CNVs in the European samples are shorter than those found in the African samples using the 1 million SNP chip. It is likely, therefore, that all KIR genes downstream of KIR2DL3 could be affected in some way by this CNV.

Relevance in relation to previous work

NK cells play an important role in controlling viral infections, having both effector and regulatory functions. They express an array of inhibitory and activating receptors, notably the KIRs, which due to their extreme variability have a unique role in fine-tuning the balance between self-tolerance and cytotoxicity. KIRs bind to their ligands on the surface of target cells, principally to MHC class I antigens. The degree of inhibition mediated by interactions between coinherited KIR and MHC class I gene products determines the activation threshold for NK cells. Several studies have described epistatic interactions between certain KIRs and MHC class I genes that modulate the host response to HIV. The 3DL1/3DS1 locus, encoding receptors for HLA-B Bw4-80I (with an isoleucine in position 80), appears to be particularly important. The presence of an activating KIR3DS1 allele in combination with HLA-B Bw4-80I has been shown to associate with protection against HIV disease progression (Martin et al, Nat. Genet. 31(4):429- 34 (2002), Epub 2002 JuI 22), as well as against opportunistic infections in HIV+ individuals (Qi et al, Plos Pathog 2(8):e79(2006)). On the other hand, various distinct allelic combinations of inhibitory KIR3DL1 and HLA-B have appeared to significantly influence AIDS progression and HIV viral load in a consistent manner: the genetic data correlated with functional differences that distinguish KIR3DL1 allotypes (Martin et al, Nat Genet 39(6):733-40 (2007), Epub 2007 May 13).

There is some controversy, however. Other groups have found that the KIR3DS1 and HLA-B Bw4 impacts on HIV progression were largely independent from each other (Gaudieri et al, Genes Immun 6(8):683-90 (2005); Barbour et al, Plos Pathog 3(4):e43 (2007)). Two very recent papers shed new light on possible KIR3DS1 effects not depending on its interaction with HLA-B. A study showed an increased proportion of KIR3DS1 homozygotes in HIV- exposed uninfected individuals (Boulet et al, AIDS 22(5):595-9 (2008)); another observed that possessing KIR3DS1 is associated with higher NK cell effector functions and decreased CD8 T-cell activation in early HIV-I disease (Long et al, J Virol, epub, Feb 27 2008). Finally, there is some non-conclusive evidence for a possible interaction between the KIR genes 2DL2, 2DL3, 2DS2 and different HLA-C types (Gaudieri et al, Genes Immun 6(8):683-90 (2005)).

In this dataset, there is preliminary evidence that there is an interaction of this KIR CNV with different HLA-B types, likely to be due to different KIR gene dosage. The direction of the HIV-associated effect of the KIR CNV, and the allele frequency are consistent with published evidence of a beneficiary effect of either of KIR2DS1 and KIR2DL1 when they are coinherited with HLA-B Bw4-

801.

* * *

All documents and other information sources cited above are hereby incorporated in their entirety by reference.

Claims

WHAT IS CLAIMED IS:

1. A method of identifying an individual at an increased risk of a rapid progression to AIDS comprising assaying DNA from a patient for the presence of a copy number variant (CNV) at chromosome 19 ql3.42, wherein, a patient carrying a deletion at chromosome 19 ql 3.42 is at an increased risk of a more rapid progression to AIDS than is a non-systematic HIV patient who does not carry said deletion.

2. The method according to claim 1 wherein said deletion is in a KIR gene.

3. The method according to claim 1 wherein said progression is defined by a decrease in CD4 cells below 350 cells/ml.

4. A method of identifying an individual at a decreased risk of a rapid progression to AIDS comprising assaying DNA from a patient for the presence of a duplication at chromosome 19 q 13.42, wherein a patient carrying a duplication at chromosome 19 ql 3.42 is at a decrease risk of a rapid progression to AIDS than is a non-systematic HIV patient who does not carry that duplication.

5. The method according to claim 4 wherein said duplication is in a KIR gene.

6. The method according to claim 4 wherein said progression is defined by a decrease in CD4 cells below 350 cells/ml.