HLA and HIV-1: Heterozygote Advantage and B*35-Cw*04 Disadvantage

See allHide authors and affiliations

Science  12 Mar 1999:
Vol. 283, Issue 5408, pp. 1748-1752
DOI: 10.1126/science.283.5408.1748


A selective advantage against infectious disease associated with increased heterozygosity at the human major histocompatibility complex [human leukocyte antigen (HLA) class I and class II] is believed to play a major role in maintaining the extraordinary allelic diversity of these genes. Maximum HLA heterozygosity of class I loci (A, B, and C) delayed acquired immunodeficiency syndrome (AIDS) onset among patients infected with human immunodeficiency virus–type 1 (HIV-1), whereas individuals who were homozygous for one or more loci progressed rapidly to AIDS and death. The HLA class I alleles B*35 andCw*04 were consistently associated with rapid development of AIDS-defining conditions in Caucasians. The extended survival of 28 to 40 percent of HIV-1–infected Caucasian patients who avoided AIDS for ten or more years can be attributed to their being fully heterozygous at HLA class I loci, to their lacking the AIDS-associated alleles B*35 and Cw*04, or to both.

HLA class I and class II loci located within the human major histocompatibility complex (MHC) comprise the most polymorphic set of genes known in humans (1–3). Products of these genes present antigenic peptide to T cells, initiating an immune response and clearance of the foreign material. Evolutionary and population studies have led to the general idea that the great diversity and even distribution of allelic frequencies observed in the class I and class II genes of the MHC (HLA in humans) are maintained through selective forces, such as infectious disease morbidity (4, 5). The hypothesis of overdominant selection (heterozygote advantage) at the MHC proposes that individuals heterozygous at HLA loci are able to present a greater variety of antigenic peptides than are homozygotes, resulting in a more productive immune response to a diverse array of pathogens (6).

Compelling evidence for the selective maintenance of MHC diversity has come from analyses of the population distribution ofHLA allele frequencies (7), the high incidence of nonsynonymous (codon altering) base substitutions among peptide-binding regions of HLA transcripts (8), persistence of numerous polymorphic MHC amino acid motifs for several million years through the emergence of multiple species (9), and a concordant increase in infectious disease sensitivity of species with increased MHC homozygosity (10). Several examples ofHLA influence on human pathogen sensitivity have been described, particularly for interaction with malaria and hepatitis B (11).

The AIDS epidemic is characterized by extreme heterogeneity in the clinical course as well as in the incidence of HIV-1 infection among exposed individuals (12, 13), which is probably a result of genetic variants among HIV-1 strains and of host genetic differences such as variants in chemokine and chemokine receptor structural genes (14, 15). More than 50 reports examining a role for HLA variation in AIDS outcomes have appeared, however the reported associations have been difficult to generalize or to affirm in multiple cohorts (12,16). Potential explanations for this difficulty involve a paucity of patients, limitations in patient clinical descriptions, failure to correct for multiple comparisons, and a reliance on serological typings that can miss allele differences found by molecular typing. Nonetheless, concordant AIDS outcomes in sib pair analyses (17), the recurrent implication of twoHLA haplotypes (A1-Cw7-B8-DR3-DQ2 andCw4-B35-DR1-DQ1) (12, 16,18), plus the quasi-species pattern of HIV-1 change in infected patients (19) are strong indicators of HLAinvolvement in HIV pathogenesis.

We performed survival and genetic association analyses to address two hypotheses: (i) that overall or specific locus heterozygosity at the HLA class I loci confers relative resistance to AIDS progression and (ii) that individual alleles at the class I loci vary in their influence on progression to AIDS. HLA class I loci were molecularly typed with DNA from individuals enrolled in five AIDS cohorts: Multicenter AIDS Cohort Study (MACS), Multicenter Hemophilia Cohort Study (MHCS), Hemophilia Growth and Development Study (HGDS), San Francisco City Clinic Cohort (SFCC), and AIDS Linked to Intravenous Experience (ALIVE) Study (20, 21). Survival analyses incorporated data derived from HIV-1–positive individuals with known dates of infection and tested HLA association with progression to three endpoints (15, 22): (i) AIDS-1993, which includes HIV-1 infection plus an AIDS-defining illness, a decline of CD4+ T lymphocytes to <200 cells/mm3, or death; (ii) AIDS-1987, which includes HIV-1 infection plus an AIDS-defining illness or death; and (iii) death.

An individual who is homozygous at HLA-A,HLA-B, and HLA-C displays a limited variety of class I molecules available for antigen presentation to cytotoxic T lymphocytes (CTLs) relative to an individual heterozygous for each class I locus. If maximum diversity in the repertoire of antigen-presenting molecules is advantageous in prolonging time to AIDS after HIV-1 infection, then homozygosity at class I genes should associate with more rapid progression to AIDS. This idea was tested by a genotype survival analysis of 498 seroconverters (patients who had enrolled in at-risk AIDS cohort studies when their HIV-1 antibody status was negative and subsequently became HIV-1 antibody positive). The analysis used a Cox proportional hazards model (23) to examine the rate of progression to three AIDS endpoints in Caucasian, African American, and combined ethnic groups. The results demonstrate a highly significant association of HLA class I homozygosity with rapid progression to AIDS in Caucasians, African Americans, and the combined ethnic group analyses for all three AIDS endpoints (Fig. 1A and Table 1). The observation is apparent when a single HLA class I locus is homozygous while the other two are heterozygous (RH = 1.5 to 4.4, Table 1; Fig. 1, A to D) and becomes more pronounced when two or three HLA class I loci are homozygous (RH = 2.9 to 4.8, Table 1; Fig. 1, A to D).

Figure 1

Kaplan Meier survival curves for seroconverters from combined cohorts that included all ethnic groups [panels (A) to (D)] for HLA class I genotype association with progression to AIDS-1987. The seroconverters were study participants with a known HIV-1 antibody negative test 36 or fewer months before the first HIV-1 positive test. (A) Association between homozygosity at one or more class I loci and progression to AIDS-1987. Relative hazards, RH, and the corresponding P values calculated by the Cox proportional hazards model (23) are given for the singly and multiply homozygous groups compared with the completely heterozygous group (41). The analyses were stratified for age groups (<30, 30 to 40, and >40 years) and for race in analyses where ethnic groups were combined. Mixed racial groups were analyzed because there is no a priori reason to expect that a heterozygosity effect would differ among ethnic groups. Because the Kaplan Meier curves suggest that the relative hazards vary over time, we also applied nonparametric numerical tests (log-rank and Wilcoxon) to the Kaplan Meier analysis. The P values from this analysis were comparable to those obtained from the Cox model analysis. (B to D) Association between homozygosity for individual class I loci [HLA-A (B), HLA-B (C), and HLA-C (D)] and progression to AIDS-1987. Study participants were divided into three groups: individuals who were heterozygous at all three loci; individuals homozygous at the specified locus, but heterozygous at both of the other two loci; and individuals homozygous at the specified locus and also at one or both of the other two loci. RH and P values for the two groups are given as a comparison of homozygotes with heterozygotes at all three loci. Influence of HLA-B*35 (E) andHLA-Cw*04 (F) on progression to AIDS-1987 among combined Caucasian cohorts. Survival analysis of individuals with one or two copies of the B*35 allele compared with individuals lacking B*35 are shown. RH and corresponding Pvalues are given for the two B*35-bearing genotypes (+/B*35 and B*35/B*35) compared with theB*35-negative genotypes. Analyses for (F) were the same as those for (E) with Cw*04 substituted forB*35.

Table 1

Survival analysis of the effect of homozygosity at the three HLA class I loci -A, -B, and -C on susceptibility to progression to three AIDS endpoints (AIDS-1993, AIDS-1987, and death) for seroconverters of MACS, MHCS, SFCC, and ALIVE (20–22). HLA class I typing is described in (40). Homozygosity at one locus only and homozygosity at two or three loci were considered as covariables in a Cox model analysis (only five individuals were homozygous at HLA-A, -B, and -C loci, precluding a robust statistical treatment of triple homozygotes). Relative hazards (RH) and theirP values are given for AIDS-1993, AIDS-1987, and death. There were not a sufficient number of African Americans homozygous for two or three HLA class I loci to provide a robust analysis. Homozygosity at each of the loci (HLA-A, -B, and -C) was considered as a Cox model explanatory variable. Each analysis was performed on seroconverters in combined cohorts, separately for Caucasians and African Americans, and for all ethnicities. Analyses were performed without considering specific allele protection and also as adjusted for the AIDS accelerating effects of B*35 and Cw*04(41). RH values were similar, and adjusted values are presented here. A survival analysis was also performed after adjusting the RH values for the protective effects of CCR5-Δ32,CCR2-64I, and SDF1-3'A (22). The results (27) revealed no epistatic interaction of these loci with HLA homozygosity because the adjusted RH andP values were largely consistent with those reported here. Dashes indicate no data available.

View this table:

Because HLA class I loci are tightly linked and occur in linkage disequilibrium in certain populations (particularlyHLA-B and -C, which are separated by 100 kb on chromosome 6) (24), the effect of homozygosity may be due to a single class I locus, rather than to all three. However, homozygosity for HLA-A, HLA-B, or HLA-C each predisposed their carriers to more rapid progression to AIDS compared with heterozygous individuals at each respective locus (Fig. 1, B to D, and Table 1). Furthermore, the relatively rapid progression of individuals with any single locus homozygosity (HLA-A,-B, or -C) was enhanced when a second or thirdHLA class I locus was also homozygous (Fig. 1, B to D), as suggested in Fig. 1A. The data suggest that each locus contributes separately to the protective effect associated with class I heterozygosity. Three of four patients in our cohorts who progressed to AIDS-1987 in less than 12 months of HIV-1 infection were homozygous at one or more HLA class I loci.

To rule out the possibility that observed HLA class I homozygosity might be marking overall homozygosity at other AIDS-influencing loci within or outside the HLA region, nine short tandem repeat (STR or microsatellite) loci within a 10-Mb region encompassing HLA (25) were genotyped in these patients and tested in a Cox analysis. In addition, a group of 16 STR loci located on other human chromosomes were similarly tested in the combined cohorts. No statistically significant association [with corrections for multiple tests (26)] was found for any AIDS outcomes (P ≥ 0.2) nor was any significant correlation among HLA homozygosity and STR homozygosity detected in these cohorts (27), indicating that the observed effects of HLA class I homozygosity on AIDS progression are due to the HLA class I genes or other very closely linked genes.

Although HLA class I heterozygosity is strongly associated with protection against AIDS, about 75% of the patients that progressed to AIDS within 4 years of HIV-1 infection were heterozygous at all three class I loci. It was of interest to determine whether individual class I alleles had an effect on progression, some of which may override the protective effect of heterozygosity at class I loci. Analysis of 330 Caucasian and 144 African American seroconverters [analyzed separately for allele and genotype distribution because of the differences in allele frequencies between the two ethnic groups (2, 24)] revealed six alleles (A29, B27, B35,B41, Cw*04, and Cw*12) of 63 detected that showed significant association with disease progression (28). Two of these, B*35 and Cw*04, were highly significant in Caucasians (28) (RH = 2.34,P = 2 × 10−6; and RH = 2.41,P = 2 × 10−7, respectively); this was also true when corrections for multiple tests were used (P = 1 × 10−4 and 1 × 10−5, respectively).

The association of HLA-B*35 and HLA-Cw*04was analyzed for both homozygous and heterozygous influence on three AIDS endpoints among combined and separate Caucasian cohorts (Fig. 1, E and F, and Table 2). The results affirm the strong association of HLA-B*35/+ heterozygotes (+ indicates any HLA-B allele except B*35) with rapid progression (Fig. 1E: RH = 2.2, P = 2 × 10−5), whereas even more rapid AIDS progression was observed among HLA-B*35/B*35 homozygotes (Fig. 1E: RH = 9.1, P = 5 × 10−4). SimilarlyHLA-Cw*04/+ heterozygosity promotes rapid progression (RH = 2.3, P = 2 × 10−6) as does HLA-Cw*04/Cw*04 homozygosity (RH = 5.2,P = 0.002). Associations determined with a dominant genetic model (which combines +/B*35 andB*35/B*35 individuals; Table 2) were highly significant for MACS and combined analyses (P ≤ 0.003 for all endpoints) which included considerably more individuals than the other cohorts. Caucasians having HLA-Cw*04 also progressed more rapidly to both AIDS-1993 and AIDS-1987 (P ≤ 2 × 10−5; Table 2).

Table 2

Caucasian seroconverters were analyzed for MACS, MHCS, and SFCC cohorts and the combination of all cohorts. The alleles B*35 and Cw*04 are each considered as codominant allele variables in a Cox proportional hazards model, that is, the allele variable has a value of 0, 1, or 2 corresponding to the number of copies of the allele carried by the individual. The analysis for B*35/+ and B*35/B*35 considers the effect ofB*35 without taking into account the presence or absence ofCw*04; likewise, the analysis of Cw*04 ignores the effect of B*35. Because these two alleles are in strong, positive linkage disequilibrium in Caucasians (24) (that is, on most chromosomes where one of these alleles is present the other is present also), an additional analysis was made for individuals carrying one of these alleles but not both. The three pairs of columns on the right give the results of analyses in which heterozygosity for B*35but not Cw*04, for Cw*04 but not B*35, and for B*35 and Cw*04 together were considered as explanatory variables in a Cox model. Homozygotes for either or both of these alleles were excluded from this analysis (39). The results of a survival analysis in which the RH was adjusted by considering the protective genotypes of CCR5,CCR2, and SDF1 (15) as additional covariants were virtually identical to the unadjusted RH andP values presented here (27). Failure to observe significant association for either HLA-B*35 orHLA-Cw*04 for MHCS and SFCC may reflect small sample size or a biased depletion of very rapid progressors to AIDS in both of these cohorts (20, 32), diminishing ability to observe HLA B*35 and Cw*04 influence on rapid progression. Dashes indicate no data available.

View this table:

Among Caucasians, progression to AIDS-1987 and to death was accelerated to about the same extent among individuals who were heterozygous forHLA-B*35 alone (that is, without HLA-Cw*04), forHLA-Cw*04 alone (without HLA-B*35), or for both alleles (Table 2). The Cox analyses revealed highly significantP values for HLA +/B*35;+/Cw*04 double heterozygotes (P = 2 × 10−5) and for single HLA +/Cw*04 (without B*35;P = 0.004) individuals to AIDS-1987, but there were too few HLA +/B*35 (without Cw*04) to achieve statistical significance (N = 4 individuals;P > 0.05).

To quantify the effect of locus homozygosity andHLA-B*35 plus HLA-Cw*04 on accelerating disease relative to alternative genotypes, we computed the relative risk (which examines the strength of the effect) and the attributable fraction (29, 30) (which estimates the fraction of individuals who progress rapidly or survive without AIDS progression because of their HLA genotype) for various AIDS endpoints (Table 3). The relative risk for early progression (≤6 years) to AIDS ranged from 1.3 to 2.5 forHLA homozygosity, for B*35 or Cw*04genotype susceptibility, or for both. Conversely, full HLAheterozygosity for all genotypes other than those with B*35or Cw*04 alleles increases the likelihood of longer survival (that is, greater than 10 years before AIDS) by a factor of 1.4 to 2.3 (Table 3).

Table 3

Combined Caucasian cohorts were partitioned into two arbitrarily defined disease categories: (i) seroconverter patients who progress to AIDS outcome in ≤6 years after HIV-1 infection and (ii) seroconverter patients who avoid AIDS outcome for ≥10 years. Relative risk (RR) and attributable fraction (AF) were computed on the basis of established methods (29, 30). Freq. susc. genotype, frequency of susceptible genotype.

View this table:

Because of the high frequency of HLA class I locus homozygosity and of B*35- or Cw*04- bearing individuals (26 and 24% respectively; Table 3), there is an appreciable fraction of rapid progressors and long-term survivors that can be attributed to HLA genotype. More precisely, the attributable fraction for rapid progression (within 6 years of infection) to AIDS is 8 to 11% for class I homozygosity, 7 to 26% for bearers of B*35 or Cw*04 or both genotypes, and 19 to 32% for both. Alternatively between 28 to 40% of the study participants who avoid AIDS for 10 or more years do so because they are fully heterozygous for alleles other than B*35 orCw*04 (31).

Our data reveal a strong association of the alleles B*35 andCw*04 with accelerated AIDS progression, but it is not certain whether one or both of these alleles, or undiscovered loci in the HLA acting on AIDS progression, are responsible. A direct effect of HLA allele products on progression to AIDS is plausible, given their role in antigen presentation to T cells and the evidence indicating that CTLs play an important role in protection against HIV-1 (12, 32). However, the failure ofB*35/Cw*04 to protect against AIDS progression may not reflect a simple failure to present HIV-1 epitopes because significant HIV epitope presentation has been reported for both of these alleles (18, 33).

One mechanism to explain rapid progression to AIDS in individuals with certain HLA genotypes may involve the regulation of natural killer (NK) cell activity. A number of studies have indicated that rapid disease progression after HIV-1 infection is correlated with decreased NK cell activity (34). Like CTLs, NK cells are involved in surveillance and killing of foreign or infected cells through a mechanism involving HLA molecules (35). Furthermore, both HLA homozygosity and theB*35-Cw*04 haplotype have been shown to be associated with reduction in NK cell number and activity (36). Recently, the HIV-1–encoded nef gene product has been shown to down-regulate HLA class I molecules (37) and perhaps protect infected cells from CTL-mediated killing (38). The relations amongHLA genetic effects described here, the immunological consequences of HIV-1 nef, and NK cell activity are areas that merit further investigation.

The average time to AIDS after infection is about 10 years (13), and an effective CTL response is thought to keep the virus in check for many years in most individuals. However, HIV-1 undergoes a high rate of mutation within a single host and escape mutant isolates that evade the host immune system evolve in most individuals, resulting in AIDS (19). The data presented here show that maximal heterozygosity at the HLA class I loci slows progression to AIDS and AIDS-related death. A parsimonious explanation for this finding is that, overall, heterozygotes present a broader range of HIV-1 peptides than do homozygotes and, therefore, it takes longer for escape mutants to arise in heterozygotes than in homozygous individuals. In any case, that heterozygotes have a significantly delayed time to death provides strong support for the hypothesis of overdominant selection at the MHC class I loci and affirms the previous indications that infectious diseases play a role in selection for heterozygosity (6, 39).

  • * To whom correspondence should be addressed. E-mail: obrien{at}


View Abstract

Stay Connected to Science

Navigate This Article