The presence of both viral particles and antiviral mucosal proteins may represent critical determinants of perinatal human immunodeficiency virus type 1 (HIV-1) transmission. In 60 HIV-1—infected women, concentrations of the innate mucosal protein, secretory leukocyte protease inhibitor (SLPI), were lower in vaginal fluid samples from 17 women whose babies became infected than in samples from nontransmitting women (mean ± SE, 57 ± 11 vs. 557 ± 177 ng/mL, respectively; P = .01). Rates of transmission among women with higher SLPI concentrations (>100 ng/mL) were lower than those among women with lower concentrations (<100 ng/mL; 8.7% vs. 40.5%, respectively; P = .01). Concentrations of other putative HIV-1—inhibitory innate immune factors were similar in both groups. Concentrations of vaginal HIV-1 tended to be higher in transmitting than in nontransmitting women (407 vs. 174 virions/mL; P = .09). Increased concentrations of selected innate mucosal immune factors, such as SLPI, seem to be associated with reduced rates of perinatal HIV-1 transmission and may contribute to natural antiretroviral defense.
A significant proportion of mother-to-child transmission of HIV-1 is thought to occur close to delivery , when exposure to HIV-1 in vaginal fluids (e.g., through skin and ingestion) represents a major risk for neonatal infection. Higher concentrations of viral particles in cervicovaginal lavage samples have been associated with an increased risk for mother-to-child transmission [2, 3]. However, this relationship is inconsistent, which suggests that other factors may modulate the risk for transmission. Local innate immune factors with HIV-1 inhibitory activity in mucosal fluids may contribute to protection against transmission in this brief but critical period.
Prominent among these mucosal factors is secretory leukocyte protease inhibitor (SLPI), an 11.7-kDa protein secreted by mucosal epithelial and acinar cells, which inhibits HIV-1 infection of monocytes and T cells in vitro . Current evidence suggests that SLPI, at a concentration ⩾100 ng/mL, blocks HIV-1 internalization, rather than binding, by CD4-expressing cells in vitro, by a mechanism involving inhibition of virus entry or capsid uncoating [5, 6]. Similarly, lactoferrin, an iron-binding glycoprotein, inhibits HIV-1 infection in vitro , as does lysozyme, an enzyme found in human placenta, breast milk, and phagocytes . The clinical roles of these factors are less well characterized. Thus, we examined the relationship of rates of mother-to-child transmission of HIV-1 among women in South Africa with concentrations and presence of HIV-1 in blood and vaginal fluids and with innate immune factors in vaginal fluids.
Subjects and Methods
We examined blood and vaginal fluids from HIV-1—infected pregnant women enrolled in a vitamin A trial, to reduce mother-to-infant transmission . As described elsewhere , HIV-1—seropositive women, recruited at 28–32 weeks' gestation, were randomized to receive either vitamin A supplementation or placebo through the antenatal period. The subjects were followed up regularly for 15 months postdelivery or for 3 months after cessation of breast-feeding. None of the women received antiretroviral therapy. All women enrolled into the trial were eligible for this substudy. A random sample of 80 women was recruited. This sample size was chosen to determine the effect of viral quantity in secretions on the risk for transmission. This sample size was selected to have 80% power (α = .05; 2-tailed test), to detect a difference of 0.5 log HIV-1 RNA quantity between transmitting and nontransmitting women. Three collected samples were inadequate and were discarded.
Vaginal fluids were obtained at entry into the study (28–32 weeks) and 6 weeks later by lavage of the vagina, using 5 mL of PBS. Samples were centrifuged at 400 g for 10 min within 5 h of sampling. One-mL aliquots of the supernatant were stored at −70°C for determination of free virus load and concentrations of SLPI, lactoferrin, lysozyme, and total IgG and IgA. Vaginal fluids were also collected from 10 HIV-seronegative pregnant women at 28–32 weeks' gestation.
SLPI was quantitated by ELISA (R&D Systems); other assays were performed as described elsewhere , and HIV-1 was quantitated by a highly sensitive polymerase chain reaction (PCR) assay (Roche Amplicor Version 1.5). Children were considered to be infected if they had a single positive PCR assay result or if they were antibody positive at 15 months, and were considered to be uninfected if they became antibody negative after 15 months or had persistently negative PCR results after 6 weeks.
HIV RNA levels below the detection threshold of the assay (i.e., <200 copies/mL) were coded as 199 copies/mL for calculations of mean RNA quantity. RNA levels were transformed using log10. Mean concentrations of innate immune factors were compared between groups by using t tests, and proportions between groups were compared by using χ2 tests. Pearson correlation coefficients were used to investigate associations between continuous variables. Logistic regression was used for univariate and multivariate analysis.
Study population and rates of HIV-1 transmission
Vaginal fluids were collected at study entry from 77 HIV-1—infected women (mean gestational age, 28.9 weeks; range, 28–32 weeks) and 6 weeks later (n = 34; mean gestational age, 36.5 weeks) and from 10 HIV-1—uninfected control women (mean gestational age, 29.2 weeks).
Transmission status was determined in 60 (78%) of 77 children (the remainder were lost to follow-up), of whom 17 (28.3%) were infected and 43 (71.7%) were uninfected. Among the 17 HIV-1—infected children, 7 had positive RNA PCR assay results at birth, 5 had negative results at birth but positive results within 6 weeks, and 1 child showed negative results at 9 months but positive results by 17 months. Three children with a positive HIV RNA test result by 1 month of age and one with a positive HIV RNA result at 4 months had no earlier negative test results. Among the 60 women followed up to delivery, 51 (85%) elected to initiate at least some breast-feeding. There were no significant differences in HIV-1 RNA or innate factors between mothers whose infants' HIV-1 status could be determined and those who were lost to follow-up.
Concentration of immune factors and viral RNA
Concentrations (mean [SE]) of SLPI, lactoferrin, and lysozyme in vaginal fluids were similar at enrollment and 6 weeks later (448  vs. 674  ng/mL; 5.3 [1.1] vs. 5.1 [1.5] µg/mL; and 5.5 [1.4] vs. 3.6 [1.1] µg/mL, respectively). Thus, only values atstudy entry were analyzed.
Compared with results in seronegative women, HIV-1—infected women showed a trend toward lower concentrations (mean [SE]) of SLPI (448  vs. 739  ng/mL, respectively; P = .43), higher concentrations (mean [SE]) of lactoferrin (5.3 [1.1] vs. 1.3 [0.3] µg/mL, respectively; P = .20), lysozyme (5.5 [1.4] vs. 0.6 [0.2] µg/mL, respectively; P = .17), total IgG (69.5 [12.0] vs. 28.3 [5.8] µg/mL, respectively; P = .2) and total IgA (7.8 [1.07] vs. 2.4 [0.75] µg/mL, respectively; P < .001). The mean HIV-1 concentration in vaginal fluids was 2.35 log virions/mL (range, 1.18–4.09).
Correlation with risk of transmission
Rates of perinatal HIV-1 transmission correlated with immunologic and virologic parameters. Concentrations of the innate protein SLPI in the vaginal mucosa were significantly lower (P = .01), and concentrations of vaginal HIV-1 tended to be higher (P = .09), among women who transmitted HIV-1 to their infants than among those who did not transmit the virus (table 1). In contrast, levels of other local innate immune factors (lactoferrin and lysozyme) were similar in both groups. Transmitting and nontransmitting women showed similar vaginal concentrations (mean [SE]) of total IgG (55.6 [10.4] vs. 62.6 [8.1] µg/mL, respectively; P = .9) and total IgA (6.2 [2.4] vs. 7.8 [1.3] µg/mL, respectively; P = .7).
No transmission occurred in women who had vaginal SLPI concentrations >200 ng/mL (figure 1). Women with higher concentrations of SLPI (>100 ng/mL), a level previously associated with effective inhibition of HIV-1 infection in vitro , had a transmission rate of 8.7% (2 of 23), compared with 40.5% (15 of 37) for those with lower concentrations (<100 ng/mL; P = .01, Fisher's exact test). The association of transmission rates with higher vs. lower SLPI concentrations was greatest among vaginal deliveries (5.9% [1 of 17] vs. 47.8% [11 of 23], respectively; P = .01) but was inapparent among women undergoing cesarean deliveries (20% [1 of 6] vs. 28.6% [4 of 14], respectively; P not significant [NS]).
Concentrations of SLPI in vaginal fluids correlated neither with maternal CD4+ T cell numbers (r = 0.06) nor with local HIV-1 RNA levels. The mean vaginal log10 HIV-1 RNA concentrations among women with higher and lower concentrations of SLPI were 1.98 and 2.16, respectively (P, NS). In the subgroup for whom both values were available (25 of 60), SLPI concentrations in vaginal fluids were negatively correlated with plasma HIV-1 RNA values (r = 0.41, P = .03); mean log10 HIV-1 RNA values in plasma were 4.00 and 4.80 among women with higher and lower SLPI concentrations, respectively (P, NS). The numbers of CD4+ T cells and CD4:CD8 ratios were lower in blood from HIV-1—transmitting women (table 1).
In univariate analysis, higher SLPI concentrations in vaginal fluids were associated with a risk of HIV-1 transmission significantly lower (odds ratio [OR], 0.14; 95% confidence interval [CI], 0.03–0.69) than that associated with lower concentrations. This reduction remained significant in multivariate logistic regression analysis adjusting for CD4+ and CD8+ T cell numbers and HIV-1 RNA concentrations in vaginal fluids (OR, 0.07; 95% CI, 0.01–0.61). After additionally adjusting for plasma HIV-1 RNA, the magnitude of the reduction in transmission associated with high SLPI concentration was similar (OR, 0.16) to that observed in unadjusted analyses (OR, 0.14), which suggests that HIV-1 RNA quantity did not explain the association between SLPI and transmission. However, plasma HIV-1 RNA data were available for only 8 transmitting mothers and 17 nontransmitting mothers. Hence, the confidence interval was wide (95% CI, 0.01–2.37), and the association was not statistically significant in this small subgroup.
Effects of vitamin A
Finally, as reported elsewhere, maternal vitamin A supplementation had no effect on the rate of mother-to-child transmission of HIV-1  or on plasma HIV-1 RNA quantity . Within this subsample, treatment with vitamin A had no effect on the concentration of SLPI, lactoferrin, or lysozyme in vaginal fluids from enrollment at 28–32 weeks' gestation to the time of sampling 6 weeks later.
We have identified a strong association between concentrations of the mucosal protein SLPI in vaginal fluids and rates of mother-to-child transmission of HIV-1. Indeed, in this cohort, the concentrations of SLPI in vaginal fluids that correlated most closely with low rates of transmission (>100 ng/mL) paralleled those that substantially inhibited HIV-1 infection in vitro . Moreover, the relationship between transmission rates and SLPI concentrations was striking among women who delivered vaginally but absent among women who delivered by cesarean section, a procedure that severely limits the infant's exposure to vaginal fluids. These results suggest that SLPI in vaginal fluids may inhibit infection of infants born by vaginal delivery, most likely by blocking entry of the virus into cells. None of these subjects were treated with antiretroviral (ARV) therapy, and, because there are no data on the effects of ARV therapy on concentrations of SLPI, we are unable to speculate on the role of SLPI in subjects treated with ARV therapy.
Other mucosal factors may also contribute to the efficiency of HIV-1 transmission, or protection against such transmission, in the vaginal mucosa. However, in contrast to the results with SLPI, no such association was found with lysozyme or lactoferrin, other mucosal proteins with putative HIV-1—inhibitory activity [7, 8]. Compared with values in vaginal fluid samples obtained from seronegative women, the SLPI value was lower and total IgG and IgA values were higher among HIV-1—seropositive women. IgG is produced locally in the vagina, but it may also derive from transudation of serum to mucosa, particularly during HIV-1 infection . Most HIV-1—specific antibodies at a variety of mucosal sites (e.g., breast milk, saliva, and intestinal fluids) are of the IgG rather than the IgA class . Parenterally administered HIV-1—specific human IgG monoclonal antibodies have provided protection against mucosal infection with simian immunodeficiency virus/HIV [13, 14]. However, we found that levels of total vaginal IgG were similar in fluids from HIV-1 transmitters and nontransmitters.
Another critical determinant of perinatal infection is the amount of HIV-1 present at the time and site of exposure. The rate of detection of virus was higher (97%) and the median concentrations were lower than those reported elsewhere [2, 15]. These differences may be due to differences in the sensitivity of assays used for viral quantitation and differences in sampling methodology. These concentrations showed a nonsignificant trend toward higher values among transmitters, consistent with findings among a non—breast-feeding population in Thailand . The ability to detect an association between mucosal concentrations of HIV-1 and transmission may be confounded by limitations in our ability to confirm the precise timing of infection. Such difficulties in ascribing risk may also relate to the interaction of virus with inhibitory factors, such as SLPI, at the mucosal surface at delivery, when most transmission appears to occur .
The lower CD4+ T cell counts and higher plasma concentrations of HIV-1 in the limited number of patients tested in this study confirm that rates of transmission are higher among women with more advanced disease . These parameters are not direct or consistent predictors of transmission, which suggests that perinatal infection—and, likely, all mucosal transmission of HIV-1—is a complex process. Careful functional analyses of the interaction of virus with the local milieu are essential if we are to identify effective interventions to prevent mother-to-child transmission of HIV-1.