Maternal SDF1 3′A Polymorphism Is Associated with Increased Perinatal Human Immunodeficiency Virus Type 1 Transmission

Abstract

Genetic polymorphisms in chemokine and chemokine receptor genes influence susceptibility to human immunodeficiency virus type 1 (HIV-1) infection and disease progression, but little is known regarding the association between these allelic variations and the ability of the host to transmit virus. In this study, we show that the maternal heterozygous SDF1 genotype (SDF1 3′A/wt) is associated with perinatal transmission of HIV-1 (risk ratio [RR], 1.8; 95% confidence interval [CI], 1.0 to 3.3) and particularly postnatal breastmilk transmission (RR, 3.1; 95% CI, 1.1 to 8.6). In contrast, the infant SDF1 genotype had no effect on mother-to-infant transmission. These data suggest that SDF1, which is a ligand for the T-tropic HIV-1 coreceptor CXCR4, may affect the ability of a mother to transmit the virus to her infant. This suggests that a genetic polymorphism in a gene encoding a chemokine receptor ligand may be associated with increased infectivity of the index case and highlights the importance of considering transmission as well as clinical outcome in designing chemokine-based therapies for HIV-1.

Chemokine receptors act as coreceptors for human immunodeficiency virus type 1 (HIV-1) entry into cells and play a major role in determining cell tropism. Macrophage-tropic HIV-1 variants typically recognize the CCR5 chemokine receptor and are frequently found at the earliest stages of infection, at least in the American and European cohorts examined to date (1, 5). Studies of a 32-bp deletion in CCR5 (CCR5Δ32) demonstrate that individuals with defective CCR5 receptors are less susceptible to infection, further supporting the model that CCR5 viruses (R5 viruses) are more readily transmitted (3, 6, 7, 11, 17). CCR5Δ32 is less common in Africa and less is known about genetic cofactors for transmission in African populations (9). In contrast to R5 virus, T-tropic HIV-1 variants are more common later in infection, and these viruses frequently recognize the CXCR4 coreceptor (X4 viruses) (1, 5). A genetic polymorphism in the untranslated region (UTR) of the gene coding for a CXCR4 ligand, SDF1 (SDF1 3′A/3′A), has been associated with delayed progression to AIDS in homozygous individuals from some cohorts, whereas in others, the same homozygous genotype has been associated with rapid progression to AIDS, but with prolonged survival after diagnosis of AIDS (4, 8, 12, 20, 21). Thus, the role of SDF1 in HIV-1 disease is unclear, and even less is known regarding how the SDF1 3′A allele affects virus replication and transmission.

Previous studies of the SDF1 allele have focused only on the association between mutations and disease progression, yet it is important to also consider how changes may affect the probability that the host will transmit the virus. The relationship between chemokine and chemokine coreceptor allelic variation in infected individuals and viral transmission may provide important clues regarding the mechanism of HIV-1 infection. Moreover, this information may be useful in developing strategies to limit the spread of HIV-1. Since the index case for infection can be most clearly identified in the setting of mother-to-child transmission and the timing of infection can be defined with some reliability, we examined the association between SDF1 3′A, maternal virus burden, and vertical transmission in a cohort of mother-infant pairs in Nairobi, Kenya. The cohort was part of a randomized clinical trial of breast and formula feeding among HIV-1 seropositive mothers (12a). Pregnant women attending Nairobi City Council antenatal clinics underwent voluntary counseling and testing for HIV-1, and HIV-1 seropositive women were invited to participate in the clinical trial. Women who were resident in Nairobi, who planned to remain in Nairobi after delivery, who agreed to randomized assignment of infant feeding, and who planned to follow-up in the study for a 2-year period were eligible to participate in the study. In this cohort, the frequency and timing of vertical HIV-1 transmission had been defined (12a). Infants in the cohort were monitored at birth, at 6 and 14 weeks, and at 3-month intervals thereafter with serial PCR assays for HIV-1 in DNA from peripheral blood mononuclear cells to determine infant infection status (12a, 16). Mother-infant pairs were monitored for 2 years after delivery. Infants were defined as HIV-1 infected if the last two consecutive PCR assays were positive, if the last PCR assay was positive and it was the last sample obtained from the infant, or if enzyme-linked immunosorbent assay testing at ≥15 months was positive and PCR data were unavailable. Infants were defined to be uninfected if none of the criteria for infection were met and the last PCR carried out was negative or enzyme-linked immunosorbent assay at ≥15 months was negative and PCR data were unavailable.

Genotyping of maternal and infant peripheral blood mononuclear cells was conducted by using PCR-restriction fragment length polymorphism techniques previously described (21). All heterozygote and homozygote samples were repeated on two separate days to ensure accuracy of assay. Furthermore, three heterozygote and six wild-type samples were amplified and directly sequenced in both the forward and reverse directions in order to confirm the G-to-A sequence change. Of 318 women typed for the SDF1 3′A allele, 89% were wild type, 10% were heterozygous, and 1% were homozygous for the mutation, with a mutant allele frequency of 0.060. This allele frequency was lower than that observed among Caucasians (0.211) but somewhat higher than that observed among African Americans (0.057) in cohorts described by Winkler et al. (21). Of 331 infants, 89% were wild type and 11% were heterozygous for mutation with a mutant allele frequency of 0.054.

In order to determine the risk ratio (RR) of perinatal HIV-1 transmission for the SDF1 3′A mutant allele, Kaplan-Meier survival analysis and Cox regression were conducted to determine the time of infant infection. The time of infant infection was estimated by taking the midpoint between the infant's age at the time of the last negative HIV-1 test and the infant's age at the first positive test. Comparisons were made between women with wild-type SDF1 and those heterozygous for the SDF1 3′A mutant allele. Chi-square tests were also used to determine the association between infant HIV-1 infection and the maternal SDF1 3′A mutant allele. Five women homozygous for the mutant allele were identified. In this cohort, there was not sufficient statistical power to evaluate the effect of a homozygous 3′A mutation on transmission, as only two of the five women who were homozygous for the allele had adequate infant HIV-1 follow-up data. Thus, all five were excluded from the analysis. In the analysis there were 306 infants, of whom 75 were HIV-1 infected. Among the 275 infants born to mothers with wild-type SDF1, 63 (23%) were HIV-1 infected, while 12 (39%) of the 31 infants heterozygous for the SDF1 3′A mutation were infected. Maternal heterozygosity for the SDF1 3′A mutation was significantly associated with increased risk of infant infection in the cohort (odds ratio [OR], 2.1; 95% confidence interval [CI], 1.0 to 4.6; P = 0.05). In survival analysis, among the women heterozygous for the SDF1 3′A mutation there was a trend for increased risk of perinatal HIV-1 transmission (RR, 1.8; 95% CI, 1.0 to 3.3) (Table 1). The incidence rate of perinatal HIV-1 infection was 17.2 infections per 100 person years among infants of mothers with wild-type SDF1 3′A, compared to an incidence rate of 33.1 infections per 100 person years among infants of mothers heterozygous for the SDF1 3′A mutation (P = 0.06).

TABLE 1.

Maternal SDF1 3′A mutant allele and viral markers and perinatal HIV-1 transmission

Characteristic tested	RR (95% CI)	Adjusted RR (95% CI)d
Plasma medium viral load >44,000 copies/ml	0.8a (0.4–1.7) (P = 0.6)
Overall infant HIV-1 infection	1.8 (1.0–3.3) (P = 0.06)	1.9 (1.0–3.7) (P = 0.05)
Early infant infectionb	1.8 (0.8–4.1) (P = 0.2)	1.9 (0.8–4.3) (P = 0.1)
Late infant infectionc	3.1 (1.1–8.6) (P = 0.03)	3.4 (1.1–10.5) (P = 0.04)

^aOR. 

^bAge less than 2 months. 

^cNegative infant HIV-1 DNA PCR at ≥2 months of age followed by positive infant HIV-1 test. 

^dAdjusted for plasma viral load. 

In this cohort, infants were monitored frequently, enabling determination of whether HIV-1 infection occurred early (within the first 2 months of life) or late (at or after 2 months of age). Infants were defined to have acquired infection early if the first positive PCR sample was collected before 2 months of life. Infants who had a negative PCR at ≥2 months followed by a positive PCR were defined as having acquired late infection. Early infant infection could have been acquired in utero, at delivery, or through early breastmilk ingestion, while late infant infection was likely acquired through breastmilk ingestion (Nduati et al., submitted). Among 252 infants with serial PCR assays and SDF1 genotype data, 16 (7%) of 228 with maternal wild-type SDF1 had late postnatal infection versus 5 (21%) of 24 with the maternal SDF1 3′A heterozygous mutation. Thus, the SDF1 3′A mutation was significantly associated with late postnatal transmission (OR, 3.5; 95% CI, 1.2 to 10; P = 0.02). This association was also significant using survival analysis, in which we observed that the SDF1 3′A/wt heterozygous genotype was associated with increased risk of infant infections occurring late (RR, 3.1; 95% CI, 1.1 to 8.6), but not with early infections (RR, 1.8; 95% CI, 0.7 to 4.8) (Table 1 and Fig. 1). The incidence rate of breastmilk HIV-1 infection was 4.4 infections per 100 person years among infants of mothers with wild-type SDF1 3′A, compared to an incidence rate of 13.9 infections per 100 person years among infants of mothers heterozygous for the SDF1 3′A mutation (P = 0.03).

FIG. 1.

Time to late postnatal infant infection versus maternal SDF1 genotype. The graph represents a Kaplan-Meier survival curve of time to first PCR detection of HIV-1 in infants infected at or after 2 months of age, stratified by maternal genotype. Solid line, SDF1 heterozygous; broken line, SDF1 wild type.

Because plasma HIV-1 RNA levels provide a marker for both HIV-1 disease progression (10, 14, 15) and perinatal HIV-1 transmission (2), we determined the effect of the SDF1 3′A mutation on plasma viral RNA levels in order to determine whether the effect of the maternal SDF1 3′A mutation was via an effect on the maternal plasma viral load. Plasma viral RNA was determined for maternal specimens obtained at 32 weeks of gestational age by using the GenProbe quantitative HIV-1 assay (13). The median CD4 count in the maternal cohort was 415 cells/mm³, and the median maternal plasma HIV-1 RNA level was 46,000 copies/ml. None of the mothers received antiretroviral therapy. Logistic regression was used to compare plasma viral loads between women with and without the SDF1 3′A mutation by dichotomizing plasma viral load at the median value for the women in the cohort (44,000 copies/ml). In this antiretroviral drug-naive cohort, plasma viral RNA levels did not differ significantly in women with and without the SDF1 3′A mutation (Table 1). Plasma HIV-1 RNA levels were significantly associated with transmission in the cohort. The median viral load among transmitters was 88,105 copies/ml versus 32,548 copies/ml in nontransmitters (P < 0.001, Mann-Whitney U test). Therefore, multivariate Cox regression was conducted to determine the effect of the SDF1 3′A mutation on perinatal HIV-1 transmission controlling for maternal plasma viral load because it is associated with perinatal HIV-1 transmission in this cohort (John et al., submitted). Using this model, the maternal SDF1 3′A heterozygous allele had an effect on transmission overall (adjusted RR, 1.9; 95% CI, 1.0 to 3.7) as well as on postnatal breastmilk HIV-1 transmission (adjusted RR, 3.4; 95% CI, 1.8 to 27.4) independent of maternal plasma viremia. The observation that the effect of the maternal SDF1 3′A mutation on perinatal HIV-1 transmission was independent of plasma viral RNA levels suggests that the association between the SDF1 3′A UTR mutation and transmission may reflect a qualitative effect, such as a change in viral phenotype or cell tropism, rather than a quantitative effect on the virus population.

Because it is difficult to distinguish whether the maternal genetic association is due to an effect within the mother or within the genetically related infant, we also determined the effect of the infant SDF1 3′A mutant allele on perinatal HIV-1 transmission. As would be expected, infant SDF1 3′A heterozygous status was significantly associated with maternal SDF1 3′A heterozygosity (P < 0.001). Infants heterozygous for the mutation were not at increased risk of HIV-1 infection (RR, 1.1; 95% CI, 0.6 to 2.2). This was also true for early and late transmission and was independent of maternal plasma viremia (data not shown). This suggests that the association between maternal SDF1 3′A heterozygocity and transmission reflects an effect on the likelihood that the mother will transmit the virus and does not reflect an effect on the susceptibility of the infant.

It is unclear how a G-to-A mutation in the 3′ UTR of SDF1 may affect protein function, and, in turn, how this may affect the biology of HIV-1 infection in individuals harboring this genetic polymorphism. Changes in the 3′ UTR could influence RNA processing or stability, leading to changes in protein expression. Because X4 viruses are more frequently detected in individuals with advanced clinical disease and individuals homozygous for SDF1 3′A had delayed disease progression, Winkler et al. proposed that the SDF1 3′A mutation may enhance expression of SDF1, thereby inhibiting the emergence of X4 variants (21). This model is supported by the observation by van Rij et al. that the frequency of syncytium-inducing strains is highest for individuals with SDF1 wt/wt, intermediate for those with SDF1 wt/3′A, and lowest for those with SDF1 3′A/3′A (20). The model in which SDF1 3′A mutation favors replication of R5 viruses could also be invoked to explain our observation, because M-tropic R5 viruses are more commonly transmitted from mother to infant than T-tropic X4 viruses (18, 22). This model can be expanded to explain the increased infectivity of breastmilk in women carrying the mutation because cellular breastmilk has a substantially higher proportion of macrophages than peripheral blood (19). Thus, it will be of interest to examine the levels of breastmilk virus in relation to SDF1 genotype. Among women in the cohort in whom we have evaluated breastmilk for the presence of infected cells, those with the SDF1 3′A genotype had a higher prevalence of breastmilk HIV-1 DNA in the first 3 months postpartum (seven of nine [78%]) than women with wild-type SDF1 (33 of 55 [60%]). The low number of women both heterozygous for SDF1 3′A and with breastmilk HIV-1 PCR results limited our ability to determine whether the mutation increases the likelihood of HIV-1 proviral shedding in breastmilk. Therefore, further studies of both breastmilk HIV-1 RNA and DNA in relation to this mutation are needed to clarify this issue. An alternative explanation for our epidemiologic observation of increased perinatal HIV-1 transmission from women with the SDF1 3′A mutation is the possibility that there is an allelic association between SDF1 3′A and another maternal cofactor responsible for increased infectivity. If this is the case, then the SDF1 3′A allele may serve as a useful marker to identify this other determinant of transmission.

Previous analyses of allelic variation of HIV-1 chemokine coreceptors and their ligands have focused on their role in disease progression or protection from infection. This study represents, to our knowledge, the first report that examines effects of chemokine genetic polymorphisms on infectivity of index cases. In these HIV-1-infected African women, we did not observe any effect of the heterozygous SDF1 3′A mutation on survival (data not shown), which is consistent with previous studies of the heterozygous allele in predominantly Caucasian male cohorts (4, 8, 12, 20, 21). However, we were able to detect an association between the heterozygous SDF1 3′A mutation and vertical transmission, suggesting that the biological phenotype linked to the SDF1 3′A mutation may have a more profound effect on infectivity of the index case than on their own disease progression. The ability to detect correlates of the heterozygous SDF1 3′A mutation suggests that perinatal transmission may be a more sensitive marker for this mutation than disease progression, in which effects have only been observed among homozygotes. If we consider the most straightforward interpretation of these data, namely, that the 3′A mutation had some direct effect on SDF1 expression, these data suggest that altering the amount of chemokine that binds to CXCR4 affects the virus pool capable of transmission. In this way, either the quantity of transmissible virus or viruses with specific phenotypic properties may be increased. Our findings suggest the need for further studies of large cohorts to determine the effect of the SDF1 3′A mutation on infectivity in order to better understand the biologic relevance of this mutation. Certainly, our data emphasize the importance of considering both clinical outcomes and transmission effects when developing HIV-1 antiviral therapies designed to competitively inhibit amplification of HIV-1 variants with particular biological phenotypes.