Preferential in-utero transmission of HIV-1 subtype C as compared to HIV-1 subtype A or D

Objective: To determine whether different HIV-1 genotypes present in a single cohort, in Dar es Salaam, Tanzania, showed differences in timing for transmission from mothers to their infants.

Methods: We determined the maternal viral load, transmission time, and the HIV-1 envelope (env) subtype of 253 HIV-1-infected infants enrolled in a randomized double-blind placebo-controlled trial to examine the efficacy of vitamins in decreasing mother-to-child transmission in Tanzania. Classification of HIV-1 positivity in utero was based on PCR results at birth. Infants were classified as intrapartum infected if they scored negative for the sample collected at birth and positive for the sample collected at 6 weeks of age.

Results: We found significant differences in the distribution of transmission time according to subtype. A higher proportion of HIV-1 with subtype C env (C-env) was transmitted in utero than HIV-1 with subtype A env (A-env), subtype D env (D-env), or both combined.

Conclusions: The identification of patterns of mother-to-child transmission times among HIV-1 genotypes may be useful in the selection of drug regimens for chemoprophylaxis. Based on our results, the efficacy of regimens administered only at labor may not protect as large a fraction of infants born in geographical regions with subtype C-env epidemics as compared to epidemics in regions where subtypes A-env and D-env predominate in the population.

Introduction

According to UNAIDS, 800 000 children were infected with HIV-1 in 2001, and the majority of these children lived in sub-Saharan Africa [¹]. Mother-to-child transmission (MTCT) occurs during three main periods: in utero, at the time of birth (intrapartum) and during breastfeeding. Estimates of MTCT range from 16% to 50%, with the highest rates reported in African countries [²]. Results from Dar es Salaam, Tanzania, indicated that 15% of women in antenatal care clinics were HIV-1 infected, and rates of more than 40% MTCT were found [³]. Advanced HIV-1 disease, progression to AIDS during pregnancy, mode of delivery, prolonged rupture of membranes, low levels of serum vitamin A, high levels of viral RNA load in plasma, and low CD4 cell counts have been identified as risk factors in MTCT [^4,5].

Viruses involved in the HIV/AIDS epidemic show extensive genetic variability. The widest range of genetic difference is seen between HIV-1 and HIV-2. The rates of MTCT are much lower for HIV-2-infected women than for HIV-1-infected women [^6,7]. HIV-1 genomes fall into three separate phylogenetic groups: the main group (M); the outlier group (O); and the non-M, non-O group (N) [⁸]. Viruses in the M group cause most of the infections, and are subdivided into subtypes A-D, F1, F2, G, H, J, K, and several circulating recombinant forms (CRF). Although subtypes A, C, D, and the CRF02_AG are all responsible for epidemics in Africa, subtype C has become the most prevalent subtype in Africa and in the world, accounting for about half of all infections [^9,10]. Differences in the time of AIDS progression, MTCT, degree of immune suppression, viral RNA load, and coreceptor use for viral entry have been reported between HIV-1 subtypes [^11-14].

Because the spread of subtype C infections to locations that already have epidemics due to other circulating subtypes has been quite recent [^15,16], comparisons of HIV-1 subtype C to others, such as HIV-1 subtype A and/or HIV-1 subtype D, have usually not been possible. In earlier studies in Dar es Salaam, Tanzania, we reported that mothers infected with subtype C or intersubtype recombinants (IR) transmitted HIV-1 to their infants significantly more frequently than did mothers infected with subtype D [^14,17]. A study in Zimbabwe reported that a large proportion of MTCT occurred in utero [¹⁸]. However, because the epidemic there is caused almost exclusively by HIV-1 subtype C, a comparison to other subtypes within the same population was not possible [¹⁸].

Regardless of advances in the prevention of MTCT, perinatal exposure to HIV-1 continues to be a major problem for infants in Africa. For example, the highly effective AIDS Clinical Trial Group Study 076 [¹⁹] protocol proved to be difficult to implement in most developing countries because various factors impeded the initiation of treatment of the mother long before delivery. The HIVNET 012 clinical trial undertaken in Uganda showed that a single dose of nevirapine given to the mother at the onset of labor and a single dose given to the infant within the first 72 h after birth was able to reduce intrapartum transmission by about 50% at 14-16 weeks of life [²⁰]. Unfortunately, it has been shown that continued breastfeeding undermines the protective effect achieved by in-utero or intrapartum therapies [²¹].

We wanted to establish whether rates of in-utero, intrapartum, or breastfeeding transmission were similar among subtypes, as this information might be valuable to establish optimal preventive strategies in epidemics driven by different subtypes.

Materials and methods

Study and determination of HIV-1 infection

Samples were obtained from infants born to HIV-1-infected women in Dar es Salaam, Tanzania. These mothers belonged to a cohort of 1078 women that were enrolled between 12 and 27 weeks of pregnancy and randomly assigned in a two-by-two factorial design to receive a daily dose of vitamin A alone, multivitamins excluding vitamin A, multivitamins including vitamin A, or a placebo. None of the mothers received any antiretroviral treatment, and after counseling mothers were free to choose to breastfeed their children. Baseline demographic, clinical, and laboratory information was obtained for each pregnant woman at recruitment. According to the World Health Organization stage disease guidelines, the majority of mothers who transmitted HIV-1 to their infants were asymptomatic at stage 1 (77.0%), followed by stage 2 (21.0%) and stage 3 (2.0%) [²²]. None of the mothers was at disease stage 4. The project was reviewed and approved by the Ethical Review Committee of the Harvard School of Public Health, the Research and Publications Committee of Muhimbili University College of Health Sciences, and the Ethical Committee of the National AIDS Control Program of the Tanzanian Ministry of Health. The overall project design, risk factors for transmission, transmission rates, and the effect of multivitamin and vitamin A supplementation on transmission have been reported elsewhere [^3,5,23]. Briefly, neither vitamin A nor multivitamin supplements taken by the mothers had any effect on HIV-1 perinatal transmission in utero or intrapartum. However, vitamin A supplementation was associated with higher rates of transmission by breastfeeding. In this cohort, the number of CD4 cells, plasma RNA viral load, and clinical stage were significant predictors for transmission [⁵].

Samples from the infants were collected at birth, 6 weeks, and subsequently at 3-month intervals until 18 months of age or until interruption of breastfeeding, whichever occurred last. The infants' HIV-1 status was determined by PCR using the Amplicor HIV-1 detection kit (Roche Diagnostic Systems, Inc., Nutley, New Jersey, USA). The protocol was modified to increase primer sensitivity to non-B HIV-1 subtypes. These modifications included changes in the amount of cell lysate and decreasing the annealing temperature during the PCR [^24,25].

Newborns who were PCR-positive within 24 h of birth were considered to have been infected in utero; those who were PCR-negative within 24 h of birth but positive at 6 weeks were considered to have been infected intrapartum, and infants who were PCR- negative at 6 weeks but positive later in life were considered to have been infected by breastfeeding [²⁶]. A β-Globin PCR amplification was included to control for quality of DNA on all samples tested at birth, and an ELISA test was carried out on samples from all infants > 18 months of age who were PCR negative, to confirm their HIV-1-negative status. Physical examinations and questionnaires for potential risk factors were given to all mothers. Blood samples from the mothers were obtained at delivery for absolute CD4 cell counts and for HIV-1 viral load determination using the HIV-1 Amplicor Monitor version 1.5 assay according to manufacturer's instructions. All samples were collected between 1995 and 2000.

Subtype classification

A nested PCR was performed to amplify the C2-C5 regions of the HIV-1 envelope glycoprotein 120 (env gp120). PCR primers, annealing conditions, cloning, sequencing, phylogenetic classification, and identification of IR have been reported previously [¹⁵].

Statistical analysis

Kruskal-Wallis tests were used to assess possible env subtype differences in the frequencies of transmission time for the in-utero, intrapartum, and breastfeeding periods. Logistic regression models were used to assess possible env subtype differences in the odds of transmission time dichotomized into two time periods. The logistic regression models were also used to compute odds ratios of transmission time for two subtypes with 95% confidence intervals (CI), both with and without adjustment for risk factors of transmission.

Results

Characteristics of the mothers

The analysis of baseline characteristics for mothers infected with subtype A-env, C-env, D-env or IR included age, gestational age, time of rupture of membranes, disease stage, CD4 cell counts at delivery, and viral load at delivery (Table 1). The median maternal age was 24 years, the median duration of the pregnancy was 38.9 weeks, and the median interval between the rupture of the membranes and delivery was 1.4 h. The majority of mothers were asymptomatic [²²]. The median log₁₀ RNA viral load at delivery was 5.1, and the median CD4 cell count at delivery was 428 × 10⁶ cells/l. No statistical differences were found in risk factors between mothers infected with subtype C-env versus subtype A-env, subtype C-env versus subtype D-env, or subtype C-env versus subtype A-env and subtype D-env (Table 1). Because no association was found between transmission time and vitamin regimen, and as the trial is ongoing, the identification of the vitamin regimens is still coded.

Table 1

Transmission patterns

The number and distribution of in-utero, intrapartum, and breastfeeding infections according to env subtype are shown in Table 2 and Fig. 1. Of 253 HIV-1-positive infants, 101 infections were caused by HIV-1 subtype A-env (39%), 73 by subtype C-env (28.8%), 53 by subtype D-env (20.9%), 22 by IR-env (8.6%), and four samples could not be typed (1.58%). Fifty-nine infants were infected in utero, 64 intrapartum, and 70 by breastfeeding. Eight infants were categorized as infected early (in utero or intrapartum but not breastfeeding) and 38 as infected late (intrapartum or breastfeeding but not in utero). In 14 infants, the earliest available sample was obtained when the infant was older than 6 weeks of age, making it impossible to determine the precise time of infection. The distribution of infections during each period for subtype A, C, and D fit into different patterns, characterized by the proportions of in-utero infections versus those acquired around the time of birth (Fig. 1). The first pattern was associated with subtype C-env infections, where the largest component corresponded to in-utero infections, and the smallest component corresponded to intrapartum infections. The second pattern included subtype A-env and subtype D-env, in which the number of in-utero infections was smaller than intrapartum infections.

Fig. 1

Table 2

The frequency distribution of transmission time in the in-utero, intrapartum and breastfeeding periods differed between subtype C-env and subtype D-env infections (Kruskal-Wallis test, P = 0.05). To determine which transmission period was most responsible for the observed differences, rates of in-utero transmission for subtype C-env and subtype D-env were each compared to rates of intrapartum transmission, breastfeeding transmission, and intrapartum and breastfeeding transmission grouped together. In the crude analysis, the probability of in-utero transmission divided by the probability of intrapartum transmission [odds ratio (OR)] was significantly higher for subtype C-env than subtype A-env (OR, 2.86; 95% CI, 1.17-6.95; Table 3). Compared to subtype D-env, infections by subtype C-env were significantly more likely to be transmitted in utero than intrapartum (OR, 3.75; 95% CI, 1.25-11.29), breastfeeding (OR, 3.18; 95% CI, 1.12-8.97), or intrapartum and breastfeeding combined (OR, 2.93; 95% CI, 1.18-7.26). After adjusting for viral RNA load and CD4 cell counts at delivery, the OR for in-utero versus intrapartum transmission by subtype C-env versus subtype A-env or subtype D-env remained significant (OR, 6.49; 95% CI, 1.76-23.90; and OR, 4.23; 95% CI, 1.07-16.76, respectively) for mothers with the same CD4 cell counts and log₁₀ viral RNA load. Because subtype A-env and subtype D-env displayed similar transmission patterns, their data were combined and compared to mothers infected with subtype C-env. This analysis showed significant differences between in-utero and intrapartum, as well as between in-utero and intrapartum and breastfeeding infections combined independent of viral RNA load and CD4 cell counts (Table 3). The analysis was repeated, adjusting for the time of rupture of membrane, disease stage, gestational stage, and vitamin regimen. The adjusted OR and P values remained similar to those obtained when only viral RNA load and CD4 cell counts were adjusted for (not shown).

Table 3

To assess if the quality or amount of input DNA was a factor in the detection of in-utero infections, a β-globin PCR was performed for all samples that were HIV negative at birth but positive at 6 weeks. A positive signal was obtained in all samples, suggesting that the sample was suitable for amplification. In addition, the distribution of subtypes did not change with the calendar time of transmission (Kruskal-Wallis test, P = 0.70) nor with the number of cells used to prepare the lysate (Kruskal-Wallis test, P = 0.46).

Discussion

Little information has been collected on the timing of transmission for genotypes of HIV-1 that account for the major epidemics in sub-Saharan Africa, where 87.5% of all infant infections occur [¹]. The major epidemics of Africa involve HIV-1 C, HIV-1 A, HIV-1 D and CRF02_AG, with HIV-1 C accounting for the largest fraction of infections [^9,10]. As CRF02_AG and HIV-1 A have the same envelope, they would be classified together in the current analysis. However, CRF02_AG has not been described in Tanzania.

Dar es Salaam, Tanzania is unusual in that a sufficiently high prevalence of infection occurs with HIV-1 A, HIV-1 C, and HIV-1 D to allow a comparison of all of the major envelope genotypes of Africa in the same population. As the viral envelope is the region of the genome responsible for cell tropism, we decided that the env genotype might be the most important basis for subtype categorization in this study. An earlier study suggested that HIV-1 C was more likely to be transmitted by pregnant women as compared to HIV-1 D, but the study did not contain data on viral load, and information on the timing of transmission was not available [^14,17].

This study compared the distribution of risk factors among the subtypes, and evaluated whether risk factors were associated with transmission timing of MTCT within the group of transmitting mothers. We determined the risk factors for transmitting mothers, estimated the transmission time for 94.4% of infected infants (239/253), and characterized the HIV-1 env subtype from 98% of the infants (249/253). There were no statistically significant differences among subtypes in maternal age, gestational age, time of rupture of membranes, disease stage, CD4 cell counts, or viral RNA load (Table 1). The most prominent findings included differences in the ratios of transmission times among env subtypes (Table 2, Fig. 1) and the higher risk of in-utero transmission by subtype C (Table 3).

Although HIV-1 subtypes were determined in a breastfeeding cohort in Kenya, the small number of subtype C and D infections made it difficult to compare the role of these subtypes in time of transmission [²⁷]. The transmission times of subtype A-env and D-env infections in Tanzania resembled the pattern described in other breastfeeding populations with predominantly subtype A and/or D infections, where the smallest contribution was given by in-utero infections (Table 2, Fig. 1) [^28-30]. In contrast, we found that subtype C-env was characterized by a larger proportion of in-utero infections. The propensity of subtype C to infect in utero was suggested by a report from Zimbabwe that showed that a large fraction of infants had a positive birth sample at a site where the epidemic is caused primarily by HIV-1 subtype C [¹⁸].

The predominance of in-utero infection in Tanzania by subtype C-env was confirmed when we pooled the intrapartum and breastfeeding infections and compared the data to in-utero transmissions. This analysis took into account 91.4% of all possible data by including 231 of the 253 known transmissions (Table 3). Subtype C-env-infected mothers were more likely to transmit in utero than at any other time-point than mothers infected with subtypes A-env, D-env, or both combined.

That we did not detect significant differences in the levels of maternal RNA loads between different HIV-1 subtypes may suggest that a biological factor, independent of plasma viral RNA load, was responsible for the earlier transmission of subtype C. Even though in-utero infections could result from either free virions or HIV-1-infected cells passing through or infecting placental cells, previous studies have shown that placental cells are infected more efficiently by cell-to-cell contact with lymphoid cells than by free virions [³¹]. It has also been shown recently that in subtype C-infected pregnant women, cell-associated HIV-1 DNA load is a stronger predictor of MTCT than is plasma RNA viral load [³²]. Therefore, it is plausible that mothers infected with subtype C had higher proviral DNA loads, or that subtype C virions could have an intrinsic property that renders them more infectious to placental cells. If these variables are responsible for the differences in transmission time remains to be determined.

Because the Amplicor test was optimized for the detection of subtype B-infected samples, it seems implausible that the observed preference for in-utero transmission of subtype C might be a result of the Amplicor test's greater sensitivity to detect subtype C sequences over subtype A or D sequences. The Amplicor test has been shown to be able to detect 10 or more copies of HIV-1 subtype B proviral DNA 100% of the time, and HIV-1 D is more closely related to HIV-1 B than HIV-1 C is to HIV-1 B [⁸]. Several independent studies, including one using Tanzanian samples, showed that the Amplicor sensitivity for subtype D-infected samples is also 100% [^24,33,34]. Because subtypes A, C, and D are the prevalent circulating subtypes in Tanzania, a modification of the standard protocol of the Amplicor test that was developed detected 63 of 64 known HIV-1-infected samples, regardless of the HIV subtype or the number of CD4 cells in a given sample [²⁵]. Based on these studies, we are confident that if any bias had been introduced by the test, it would have been in favor of earlier detection of subtype D rather than subtype C transmissions.

In this study and others, the definition of in-utero, intrapartum, and breastfeeding infections is based solely on the timing of the sample analysis. This approach may underestimate the number of in-utero infections, because infants are classified as infected in utero only on the basis of a positive PCR result when sampled at the time of birth. Because the development of PCR positivity should take 1-3 weeks to occur, it seems highly likely that some infections scored as intrapartum might have occurred as a result of exposure during the last 2 weeks in utero.

Several short-term protocols for prevention, centered on the time of delivery, have been proposed as cost-effective alternatives in countries with limited resources and minimal health care systems [^20,21,30]. The shortest single-drug protocol shown to significantly reduce MTCT was the HIVNET 012 conducted in Uganda, a population characterized by HIV-1 subtype A and D infections [²⁰]. In Tanzania, the proportion of intrapartum transmissions that could be prevented with chemoprophylactic interventions initiated at labor appears to be significantly reduced in subtype C-infected mothers (Fig. 1). Our results suggest that a large fraction of MTCT in geographical regions with predominant or expanding subtype C-env epidemics might only be prevented when prophylactic interventions are initiated earlier during pregnancy. Our results also reinforce the importance of interventions to minimize the risk of breastfeeding transmission for all of the major epidemics in Africa.

Acknowledgments

We thank C. E. VanWinkle for editorial assistance.

Sponsorship: This study was supported in part by NIH grants R35 CA39805, RO1 HD32257, RO1 HD37783 and by grant D43 TW00004 from the Fogarty International Center.