Genetic Restriction of AIDS Pathogenesis by an SDF-1 Chemokine Gene Variant

See allHide authors and affiliations

Science  16 Jan 1998:
Vol. 279, Issue 5349, pp. 389-393
DOI: 10.1126/science.279.5349.389


Stromal-derived factor (SDF-1) is the principal ligand for CXCR4, a coreceptor with CD4 for T lymphocyte cell line–tropic human immunodeficiency virus–type 1 (HIV-1). A common polymorphism,SDF1-3′A, was identified in an evolutionarily conserved segment of the 3′ untranslated region of the SDF-1 structural gene transcript. In the homozygous state, SDF1-3′A/3′Adelays the onset of acquired immunodeficiency syndrome (AIDS), according to a genetic association analysis of 2857 patients enrolled in five AIDS cohort studies. The recessive protective effect ofSDF1-3′A was increasingly pronounced in individuals infected with HIV-1 for longer periods, was twice as strong as the dominant genetic restriction of AIDS conferred by CCR5 andCCR2 chemokine receptor variants in these populations, and was complementary with these mutations in delaying the onset of AIDS.

HIV-1 strains isolated from recently infected individuals are predominantly macrophage-tropic (M-tropic) and non–syncytium-inducing, and they co-opt CC-chemokine receptor proteins as entry ports in combination with CD4 molecules (1, 2). Over the course of HIV-1 infection, viral phenotype and coreceptor use broaden to include the appearance of T lymphocyte cell line–tropic (T-tropic) variants near the time when AIDS symptoms are first observed (1, 3, 4). T-tropic strains induce the formation of syncytia in CD4+cell lines in vitro, infect peripheral blood mononuclear cells (PBMCs) faster, and replicate more aggressively than do the early M-tropic isolates (1, 4). The occurrence of T-tropic isolates usually precedes a precipitous drop in CD4 T cells, which suggests that these viruses may contribute to T cell depletion. T-tropic HIV-1 enters target cells by means of CD4 and CXCR4 as a coreceptor complex, although T-tropic strains can also use CCR5 (4). Stromal-derived factor (SDF-1, also called pre–B cell growth stimulating factor), a powerful chemoattractant cytokine, is the natural ligand for CXCR4. Recent experiments have shown that SDF-1α (one of two transcriptional splice variants of the SDF1gene) is capable of down-regulating CXCR4 on cells by induction of endocytosis, effectively blocking infection by T-tropic but not M-tropic HIV-1 strains (5, 6).

The use of available CXCR4 coreceptors by viral strains that emerge during late stage HIV-1 infection, together with the demonstration that SDF-1 effectively inhibits HIV-1 replication, prompted a polymorphism search for SDF-1 structural gene variants that might influence HIV-1 transmission or pathogenesis. We screened 1354 of the 3526 base pairs (bp) represented in human SDF-1β transcripts with a series of polymerase chain reaction (PCR) primers and single-strand conformation polymorphism (SSCP) heteroduplex assays (7) in a subgroup of 144 patients enrolled in five epidemiologic cohorts assembled to monitor HIV-1 infection and AIDS (8-11). Sequence analysis of a common variant revealed a G → A transition at position 801 (counting from the ATG start codon) in the 3′ untranslated region (3′UTR) of the reference sequence (GenBank accession numberL36033). The polymorphism (designated SDF1-3′UTR-801G-A and abbreviated SDF1-3′A below) is represented in the SDF-1β transcript but not in the SDF-1α transcript. Because this variant eliminated an Msp I restriction site, a PCR–restriction fragment length polymorphism (RFLP) assay was used for rapid detection of genotypes (7). The allele and genotype frequencies ofSDF1-3′A were determined in 2857 individuals from five AIDS cohorts (8-11). The following SDF1-3′A allele frequencies were found: Caucasians, 0.211 (n = 1835); Hispanics, 0.160 (n = 131); African Americans, 0.057 (n = 859); and Asians, 0.257 (n = 37) (12).

A role for SDF1-3′A in HIV-1 infection was investigated by genotyping 2419 HIV-1–infected patients and 435 HIV-1–exposed uninfected individuals. No significant differences in SDF1allele or genotype frequencies were observed in initial comparisons of exposed (or at risk) uninfected versus infected individuals in separate or combined cohort analyses [Fisher's exact test (FET),P = 0.16 to 1.0] (12). However, a group of 79 high-risk exposed uninfected individuals from MACS (those with extremely high-risk sexual practices) (10, 13) showed a highly significant elevation in SDF1-+/3′Aheterozygotes [50.6% among high-risk uninfected individuals compared to 31.1% among infected patients; FET, P = 0.002], suggesting a protective effect against HIV-1 infection for this genotype.

The influence of SDF1 genotypes (+/+, +/3′A, and 3′A/3′A) on disease progression among HIV-1–infected individuals was analyzed by means of Cox proportional hazards models (14) (Fig.1 and Table1). A subgroup of 639 seroconverters from four cohorts was included in the analysis (8-11). These participants had well-characterized dates of seroconversion, with a first positive HIV test no more than 3 years after the last negative test, or, in the case of some SFCC participants, before the end of 1980 (8, 9, 15). Three AIDS endpoints reflecting advancing morbidity were evaluated: (i) AIDS-1993, as defined by the U.S. Centers for Disease Control (16) (that is, HIV-1 infection plus AIDS-defining illness or decline of CD4 T lymphocytes to <200 cells/mm3) or death; (ii) the more stringent AIDS-1987 definition (16) (HIV-1 infection plus AIDS-defining illness) or death; and (iii) death during follow-up for an HIV-1–infected patient (97% of these had AIDS-1993).

Figure 1

Kaplan-Meier survival curves of seroconverters, showing relation ofSDF1-3′A/3′A recessive protection to AIDS endpoints (15). Left panels:SDF1-3′A/3′A genotype survival (red) is compared with survival of SDF1-+/3′A (yellow) and SDF1-+/+ (blue) genotypes. Caucasians in the combined (ALIVE, MACS, MHCS, and SFCC) cohorts (8-11) [n = number of patients, RH = relative hazard,P = log likelihood P value based on the Cox proportional hazards model (14) forSDF1-3′A/3′A and +/3′A survival compared to SDF1-+/+ survival]. The value of nfor AIDS-1993 is smaller than for AIDS-1987 or death because several subjects had CD4 T lymphocyte counts below 200 before HIV infection; for these subjects AIDS-1993 was impossible to define. Right panels: Survival curves for protective genotypes for SDF1,CCR2, and CCR5 versus +/+ at the three loci. The protective genotypes are SDF1-3′A/3′A,CCR2-+/64I or 64I/64I, and CCR5-+/Δ32 and Δ3232. The four curves represent the following genotypes: +/+ at SDF1, CCR2, andCCR5 (blue); one or more CCR2/CCR5protective genotypes and SDF1-+/+ (green);SDF1-3′A/3′A and +/+ atCCR2/CCR5 (orange); andSDF1-3′A/3′A and protection by one or moreCCR2/CCR5 protective genotypes (pink).× indicates single events; • indicates patient censoring. Summary statistics for the combined cohort analyses are shown in Table 1 (12). Log-log survival time versus log time plots were examined for proportionality with the combined cohort analysis. The plots were parallel and did not intersect, as assumed in the Cox proportional hazards model (14).

Table 1

Survival analysis of protection from progression to AIDS outcomes by SDF1-3′A/3′A variant,CCR5, or CCR2 protective polymorphisms.Analysis 1: Survival analysis for progression to three AIDS endpoints among HIV-1–infected seroconverters forSDF1-3′A/3′A versus SDF1-+/+ orSDF-+/3′A genotypes, as in Fig. 1. Seroconverters in combined cohorts including only Caucasians and for all ethnic groups were analyzed using the Cox proportional hazards model (13, 14). A log likelihood test (1 df) (LL),P value, relative hazard (RH), and 95% confidence interval (CI) were calculated for each variable in the analysis of AIDS outcomes (14). Times to AIDS-1993, AIDS-1987, and death (see text) were calculated from the midpoint of the last HIV-1–negative test date and the first HIV-1–positive test date. The value of n for AIDS-1993 is smaller than for AIDS-1987 or death because several subjects had CD4 T lymphocyte counts below 200 before HIV infection; for these subjects AIDS-1993 was impossible to define. Seroconverters with an interval greater than 3 years between last negative and first positive were excluded from the analysis. Analyses were adjusted for age, where age is a categorical variable with three categories: <30, 30 to 40, or >40 years old. Analysis 2: SDF1-3′A/3′A versus SDF1-+/+ orSDF1-+/3′A controlling for the protective genotypes of CCR2 and CCR5 (21).Analysis 3: CCR2-64I/64I orCCR2-+/64I orCCR5-+/Δ32 versus CCR5-+/+ andCCR2-+/+ (normal at two loci) controlling for the protective genotype of SDF1. The LL P values use a Bonferroni correction for multiple independent tests performed in each of the three analyses. An LL calculation for χ2 was performed because of the small numbers of patients and few failures inSDF1-3′A/3′A individuals (14). [+/+] for SDF1 includes SDF1-+/+ andSDF1-+/3′A.

View this table:

For the combined and separate cohort analyses, theSDF1-+/+ and SDF1-+/3′A individuals were indistinguishable in the pattern of progression to the three AIDS endpoints (Fig. 1). However, there was a marked slowing in progression to AIDS for individuals with the SDF1-3′A/3′Agenotype [that is, relative hazards (RH) < 0.65; Table 1]. The delay was statistically significant with MACS and combined cohorts for AIDS-1987 and death, and with SFCC for death, for Caucasians or all ethnic groups (12, 13, 17). The extent of observed protection from AIDS progression associated with theSDF1-3′A/3′A genotype follows a gradation in combined and SFCC cohorts across increasingly severe AIDS endpoints (Fig. 1 and Table 1). The RH value for the combined Caucasian cohort sample was 0.65 for AIDS-1993, 0.36 for AIDS-1987, and 0.24 for death (lower values indicate increased protection; Table 1). The tendency to display increased protection in later stages of HIV-1 infection was also seen in MACS (RH = 0.59 > 0.22 > 0.1, respectively) and SFCC (RH = 0.83 > 0.30 > 0.00) cohorts (12). This gradation was extended when time to CD4 <200 cells/mm3 (alone without AIDS disease or death) was used as an endpoint (Caucasians: RH = 0.67, P = 0.18; all races: RH = 0.63, P = 0.13) (18). The gradation suggests that SDF1-3′A/3′A protection is more pronounced in later stages of HIV-1 infection and is possibly related to interference with the appearance of T cell–tropic HIV-1 strains.

The protective effects of SDF1-3′A/3′Ahomozygotes were also apparent in defined disease category analyses, which allow the inclusion of seroprevalent patients (those whose seroconversion date is unknown because they were HIV-1 antibody–positive at the time of enrollment) in the slow/nonprogressor category (9). A significant elevation in the frequency ofSDF1-3′A/3′A was observed among slow/nonprogressors within the combined cohorts for every AIDS outcome (Fig. 2A). The relative risk for AIDS avoidance, estimated by a case/control odds ratio, ranged from 2.4 to 4.1 for the three AIDS endpoints. The results of both the survival (Fig. 1 and Table 1) and disease category analyses (Fig. 2A) reveal a strong recessive SDF1-3′A association with delayed clinical outcomes of HIV-1 infection.

Figure 2

(A) Defined disease category analysis of SDF1-3′A/3′A genotype frequencies for each cohort and combined cohorts for the three endpoints: AIDS-1993, AIDS-1987, and death (see text). Seroconverters who progressed to the designated outcomes before the cutoff time (defined below) were compared to seroconverters plus seroprevalents who survived outcome-free for at least that length of time. Imputed seroconversion dates for the seroprevalent subgroup for MHCS, HGDS, and ALIVE were provided by the cohort investigators (10, 11). For MACS, date of enrollment was used as the starting date. Cutoffs (in years) were chosen as the time when approximately half of all seroconverters had progressed to the outcomes. The following times for the outcomes and cutoffs were used. For AIDS-1993: MACS, 6 years; MHCS, 9 years; SFCC, 12.5 years; and combined cohorts, 8 years. For AIDS-1987: MACS, 7.5 years; MHCS, 11.5 years; SFCC, 14.5 years; and combined, 10 years. For death: MACS, 8 years; MHCS, 11.5 years; SFCC, 14 years; and combined, 12 years (9). The number of individuals in each disease category is at the top of the bars. *P < 0.05, **P < 0.01 (FET for the null hypothesis of noSDF1-3′A/3′A protection compared withSDF1-+/+ plus SDF1-+/3′A). Bars are for Caucasians; triangles indicate SDF1-3′A/3′Afrequencies for all ethnic groups. The relative risk of rapid progression for unprotected (SDF1-+/+ orSDF1-+/3′A) patients (relative toSDF1-3′A/3′A patients) was estimated for Caucasians in the combined cohorts by calculating the case/control odds ratios with slow progressors as controls; that is, the risk for each category is the ratio of the percentage of rapid progressors to the percentage of slow progressors. Estimate relative risks (with 95% confidence intervals and FETP values in parentheses) were as follows: for AIDS-1993, 2.4 (1.0 to 7.2, P = 0.02); for AIDS-1987, 3.4 (1.2 to 13.4, P = 0.007); for death, 4.1 (1.2 to 21.5,P = 0.007). (B) Frequencies of the protectiveSDF1-3′A/3′A genotype alone (black bars) or in combination with at least one CCR2/CCR5protective genotype (CCR5-+/Δ32,CCR2-+/64I, orCCR2-64I/64I; shaded bars) in six intervals of increasing survivorship from midpoint (seroconverters) or imputed (seroprevalents) seroconversion dates in Caucasians. Genotypic frequencies were determined separately for time to AIDS-1993, to AIDS-1987, and to death (see text) using seroconverters progressing to the outcomes in less than 3.5 years, 3.5 to <7 years, and 7 to <10 years, and including seroconverters and seroprevalents progressing to the outcomes in 10 to <13 years, 13 to <16 years, and ≥16 years. The number of individuals observed in each category is shown above the bars. The average frequency of the protective genotype for Caucasians is shown as an arrow. There is a statistically significant trend (Mantel-Haenzel χ2) toward enrichment of SDF1-3′A/3′A genotypes at increasing survival intervals for AIDS-1993 (P = 0.04) and AIDS-1987 (P = 0.01), and toward enrichment of compositeSDF1-3′A/3′A plus CCR5 genotypes at increasing survival intervals for AIDS-1993 (P = 0.005), AIDS-1987 (P = 0.01), and death (P = 0.05).

Variant alleles within the coding regions of the chemokine receptors CCR5 and CCR2, which are coreceptors for M-tropic HIV, have been shown to delay the rate of progression to AIDS (8, 9, 19, 20). The mutant alleles CCR5-Δ32 and CCR2-64Iare dominant, genetically independent, and equally protective (8,9, 21). An estimated 25 to 30% of long-term survivors who remain AIDS-free for >16 years can be attributed to a protective genotype for either CCR5-Δ32 or CCR2-64I(9). A survival analysis of the relative contributions ofCCR5-Δ32, CCR2-64I, andSDF1-3′A genotypes (Fig. 1 and Table 1) reaffirms the protective effects of CCR2, CCR5, andSDF1 variant genotypes on progression to AIDS when the influence of the other protective loci are considered as covariables (14, 21). For AIDS-1987 and death endpoints, the extent ofSDF1-3′A/3′A protection in combined cohorts (Caucasian and all ethnic groups) is approximately twice that seen withCCR2 or CCR5 protection (that is, RH forSDF1 versus CCR from Table 1 equals 0.37:0.64 for AIDS-1987 and 0.24:0.56 for death;P = 0.03) (22). In addition, CCRand SDF1 protection are additive in AIDS cohorts, because patients with both SDF1 and CCR protective genotypes avoid AIDS outcomes longer than do patients with only single-gene protection (P = 0.05 for AIDS-1993,P < 0.01 for AIDS-1987 and for death; Cox model log likelihood test) (14, 23). For example, only 1 of the 10 seroconverter patients who were genotypicallySDF1-3′A/3′A and either CCR2- orCCR5-protected has progressed to AIDS-defining conditions (AIDS-1987), while 5 of 13 SDF1-3′A/3′A,CCR2-+/+; CCR5-+/+ patients did. However, 11 of 23 dual-protected seroprevalent patients ultimately succumbed to AIDS, although their time interval to AIDS was unknown.

The cumulative effects of the SDF1-3′A/3′Aprotective genotype, separately or in combination with CCRprotective genotypes, were assessed over six intervals after HIV-1 seroconversion (Fig. 2B). The results reveal a significant increase ofSDF1-3′A/3′A genotypes among patients who avoid AIDS for longer periods (P = 0.02). There was a complete absence of a dual CCR/SDF1 composite variant genotype among patients who developed AIDS-1987 or died within the first 10 years after HIV-1 infection, and only a single individual with SDF1 plus CCR protective genotypes developed AIDS-1993 during this interval. Combined with the survival analyses, these data emphasize the protective effect of theSDF1-3′A/3′A genotype and suggest that its effect is more than additive with the protection provided by CCR2and CCR5 variant alleles (23).

The SDF1-3′A variant is located in a segment of the 3′UTR of the SDF-1β transcript (6) that is highly conserved in sequence (69% sequence between human and mouse SDF-1β 3′UTR sequence with no gaps). This extent of conservation within the segment suggests that it may serve as a target for cis-acting factors influencing transcript abundance, synthesis, transport, stability, or splice product abundance (24). The simplest hypothesis forSDF1-3′A/3′A action involves up-regulation of the quantity of SDF-1 protein available to bind CXCR4 and stem the appearance of late stage T-tropic HIV-1 strains in infected patients (25). This mechanism would be consistent with the gradation in survival outcome whereby SDF1 protection is more pronounced in late stage AIDS outcomes than for earlier stages (Fig.1).

The frequency of SDF1-3′A homozygous recessive individuals is low (<5%) in the study populations, but the effect is quite strong. Although SDF1 protection is more apparent in later stages of HIV-1 infection, the principal effect may actually involve a strong protection against rapid progression to AIDS (Fig.2B). The extent of observed SDF1-3′A/3′Aprotection is twice as strong overall as the influence of eitherCCR5-Δ32 or CCR2-64I. This extent of protection mediated by a potential regulatory region ofSDF1, effective in a test population and without obvious clinical cost, offers an attractive opportunity for therapeutics that could mimic the action of the variation.


Stay Connected to Science

Navigate This Article