The impact of pregnancy on anti-HIV activity of cervicovaginal secretions

Abstract

Background

Mucosal immunity of the female genital tract plays a critical role in defense against sexually transmitted infections like human immunodeficiency virus (HIV). Pregnancy is associated with both structural and immunological alterations in the genital mucosa, but the impact of these changes on its ability to suppress HIV infection is unknown. Current epidemiologic data are conflicting as to whether pregnancy increases the risk of HIV acquisition.

Objective

To define the association between antimicrobial peptides and chemokines in cervicovaginal secretions and in vitro HIV infectivity among pregnant and non-pregnant women.

Study Design

40 pregnant and 37 non-pregnant women were enrolled in a prospective longitudinal cohort study at a single tertiary care women’s hospital in Providence, Rhode Island. Cervicovaginal lavage (CVL) was performed at each study visit. For pregnant women, study visits occurred once per trimester and there was an optional postpartum visit. For non-pregnant women, study visits occurred across a single cycle timed to occur in the proliferative, ovulatory, and secretory phases based on presumption of a regular menstrual cycle. The impact of CVL on HIV infectivity was evaluated using a TZM-bl assay and compared between pregnant and non-pregnant women for each visit. The previously validated TZM-bl assay, utilizing a luciferase reporting gene to indicate HIV infection of TZM-bl cells, was measured with a luminometer, with higher relative light units indicating greater levels of in vitro HIV infection. Immune mediators were measured using a multiplex bead assay. HIV infectivity and median concentration of each mediator were compared between pregnant and non-pregnant groups using the Wilcoxon rank sum test.

Results

Cervicovaginal fluid from pregnant and non-pregnant women significantly decreased HIV infectivity in both groups compared to positive control (virus only),p<0.01, but infectivity was not different between groups (p>0.44). During the second and third trimesters, pregnant women experienced suppression of several cervicovaginal immune mediators, including Human beta defensin (HBD)-2, lactoferrin, Macrophage inflammatory protein (MIP)-3α, Regulated upon activation, normally T-cell expressed and secreted (RANTES), and Stromal cell-derived factor (SDF)-1 (all p<0.05). The antimicrobial peptide elafin was significantly correlated with HIV infectivity in both groups across all visits, except at postpartum in the pregnant group (n=16). Secretory leukocyte protease inhibitor (SLPI) was also significantly correlated with infectivity across all visits, but in non-pregnant women only (p<0.03).

Conclusions

Cervicovaginal secretions from both pregnant and non-pregnant women contain immune mediators associated with HIV infectivity in an in vitro assay but infectivity was not different between pregnant and non-pregnant groups. If pregnant women are at increased risk for HIV infection, it is unlikely to be mediated by alterations in the effectiveness of these protective secretions.

Condensation

While antimicrobial peptides in cervicovaginal secretions are altered during pregnancy, the protective effect of cervicovaginal fluid against HIV in an in-vitro model is not compromised.

INTRODUCTION

HIV/AIDS is the leading cause of death worldwide for women of reproductive age,¹ affecting 1.5 million pregnancies in 2013, a statistic that has not improved since 2009. Additionally, there were 240,000 new infections reported in 2013, largely as a result of perinatal transmission.² The disproportionate impact of HIV on young women may stem not only from social inequality, but also from biological patterns of heterosexual transmission. Studies of HIV discordant couples consistently demonstrate higher odds ratios of male-to-female transmission than female-to-male transmission, due to the relative anatomic and histologic susceptibilities of the female genital tract to HIV acquisition.^3–5

It is important to consider how the susceptibility of the lower genital tract to HIV acquisition may be altered in pregnancy, as risk of perinatal transmission magnifies the impact of HIV on reproductive age women. While several studies have demonstrated high incidence of HIV infection in pregnant women,^6–8 data from prospective studies comparing HIV acquisition in pregnant and non-pregnant women have been conflicting.^9–11 In a cohort study of over 10,000 women in Rakai, Uganda, Gray et al showed that the risk of HIV acquisition doubles during pregnancy, based on an incidence rate of 2.3 per 100 person-years in pregnancy compared to 1.1 per 100 person-years in non-pregnant women.⁹ This study was unique in its ability to establish this association after controlling for sexual behaviors in both men and women, indicating that there may be a physiological basis to the increased risk of HIV acquisition in pregnancy. However, a more recent meta-analysis reported that there are no significant differences in risk of HIV acquisition between pregnant and non-pregnant women based on pooled data from five prospective cohort studies in African populations.¹²

The first line of defense against HIV acquisition in the female genital tract is the mucosal lining of the epithelial barrier.¹³ There is evidence in literature of pregnancy-mediated alterations in the cytokines, chemokines, and antimicrobial peptides (AMPs) that contribute to genital immunity,^14–17 but the impact of these changes on HIV infectivity is not known. Studies in non-pregnant women have associated Macrophage Inflammatory Protein (MIP)-3α, RANTES (regulated upon activation, normally T-cell expressed and secreted), human beta defensin (HBD)-2, elafin, and several other immune mediators with protection against HIV.^18–20 In a previous study, we demonstrated in vitro anti-HIV-1 activity in cervicovaginal secretions from both pregnant and non-pregnant women using a TZM-bl indicator assay.²¹ This study employed the TZM-bl assay, preferred by the World Health Organization (WHO) as the most standardized way to measure HIV infectivity,²² to test cervicovaginal lavage (CVL) from pregnant women. The TZM-bl assay utilizes a luciferase reporting gene to indicate HIV infection of TZM-bl cells. The luminescence is measured with a luminometer, with higher relative light units indicating greater levels of in vitro HIV infection. The objective of the current study was to utilize this novel in vitro method to compare, in an adequately powered study, the risk of HIV infectivity between pregnant (1^st, 2^nd, and 3^rd trimester) and non-pregnant women, and to investigate the associations between HIV infectivity, pregnancy, and various immune mediators in CVL.

MATERIALS AND METHODS

Patients and Study Design

This prospective cohort study was conducted at a single tertiary care women’s hospital in Providence, Rhode Island. From 2010–2013, we enrolled HIV-negative pregnant and non-pregnant women between the ages of 18 and 35 in this longitudinal study. Pregnant women with gestational age of <14 weeks (by best obstetric estimate) and non-pregnant women with at least three months of regular menses were offered enrollment. All participants provided written, informed consent during enrollment. Institutional Review Board (IRB) approval was provided by the Women & Infants Hospital IRB.

Conditions that might alter the risk of HIV infectivity or pregnancy outcomes were criteria for exclusion. These included pre-gestational diabetes mellitus, chronic hypertension, recent antibiotic use, use of hormonal contraception, current or planned cerclage, planned termination of pregnancy, known fetal anomalies, recent sexually transmitted infection, or recent symptomatic vaginal discharge. Patients at our facility have HIV testing performed in an opt-out manner at the first prenatal visit. Study participants were asked to refrain from using intravaginal products or participating in sexual intercourse without a condom within 48 hours of study visits; those who did not comply were asked to return at a later date.

Study Visit Protocol

For the pregnant group, study visits occurred once per trimester (<14 weeks, 14–28 weeks, and after 28 weeks) followed by an optional postpartum visit at 4–8 weeks postpartum. For the non-pregnant group, visits occurred across the menstrual cycle such that there was one visit in each of the proliferative (days 7–10), ovulatory (days 12–16), and secretory (days 20–26) phases. Demographic and clinical data were collected at each visit and pregnancy outcome data were abstracted from medical records after delivery.

At each visit, participants underwent cervicovaginal lavage (CVL) as previously described.²¹ Prior to CVL, secretions from the posterior fornix were collected and rolled onto a dry slide for gram stain, in order to test for bacterial vaginosis by Nugent score. After collection, CVL samples were centrifuged at 1500 g for 10 minutes, and supernatant was stored at −80°C until TZM-bl assay performance.

Assessment of HIV Inhibition by CVL

The anti-HIV activity of CVL from pregnant and non-pregnant women was assessed using a TZM-bl inhibition assay.²¹ Infection is measured by expression of reporter genes under the control of an HIV-1 long terminal repeat (LTR). The HIV-1 strain used in this assay was X4-tropic, NL4.3. TZM-bl cells were prepared and maintained as previously described.¹⁸

TZM-bl cells were seeded at 2×10⁴ cells per well in a 96-well microtiter plate and allowed to achieve confluence overnight at 37°C. CVL from each patient was diluted 1:4 in TZM-b media and incubated with HIV for 1 hour at 37°C in a final volume of 100µl. Media was aspirated from the TZM-bl cells and replaced with the mixture of CVL and virus (100µl) plus 100µl of TZM-bl media. Viability of TZM-bl cells on treatment with CVL was quantified using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI) according to manufacturer’s instructions.

The light intensity of each well was measured using a luminometer and was expressed as relative light units (RLU). Uninfected TZM-bl cells in media only and cells in CVL only were used as negative controls and for determination of background luminescence. Cell viability was confirmed for each assay. TZM-bl cells incubated with HIV-1 alone was used as a positive control, and the median RLU of these wells was set to 100% after subtraction of background luminescence. Percent inhibition by CVL was calculated by taking the RLU values of CVL treated infected cells after subtraction of background luminescence and expressing them as percentages of the RLU values of the positive control.

Measurement of Immune Mediators

Concentrations of several AMPs and chemokines that have been previously correlated with HIV infectivity were measured in CVL of pregnant and non-pregnant women by multiplex bead assay. This assay utilized magnetic fluorescent microspheres (Luminex Corp, Austin, TX) coupled to capture antibodies as previously described.^23,24 Specific capture and detection antibodies, as well as standards for each AMP were sourced as described by Boesch et al.²⁴ A master mix of antibody-coupled microspheres was prepared by combining individual bead sets at a concentration of 500 microspheres for each antibody in a 10µl final volume in LL-37 ELISA kit dilution buffer (Hycult Biotech, Plymouth Meeting, PA) for each well. 40µl of CVL sample or standard, followed by 10µl of the microsphere master mix, was placed in each well (Greiner Bio-One, Monroe, NC). CVL samples were diluted in assay buffer when necessary. After addition of samples, the plate was incubated overnight (Biotek, Winooski, VT). Fluorescent signal was detected as previously described, using biotinylated detection antibodies with Streptavidin-R-phycoerythrin (Prozyme, Hayward, CA) as the secondary detection agent.²⁴ When the fluorescent signal was too low to measure, the lowest value on the standard curve for each mediator was substituted.

Sample size determination

The sample size was calculated using RLU infectivity estimates from preliminary work. Based on these estimates, we anticipated that the mean RLU in non-pregnant women was expected to be 1.5 million (+/−SD=1 million). The sample size was calculated using between factors repeated measures MANOVA. The correlation between repeated measures was set at 0.25. In order to detect a doubling of the anticipated RLU, a sample size of 35 women per group was required, using an alpha=0.01 to account for subgroup analyses, and power of 80%. 10% over-enrollment was planned a priori to account for anticipated incomplete data collection.

Statistical Analysis

Categorical demographic variables were compared between pregnant and non-pregnant groups by Chi-square or Fisher’s exact test. Normally distributed continuous variables were compared by t-test. RLU values, percent inhibition of HIV by CVL, and immune mediators concentrations were not normally distributed, even after logarithmic or other standard data transformations. Therefore, these variables were examined by nonparametric statistical tests instead of repeated measures ANOVA and other parametric tests. Percent inhibition of HIV by CVL and concentrations of immune mediators were compared between groups at each visit using the Wilcoxon rank sum test. Spearman rank correlations between percent HIV inhibition and concentration of immune mediators were calculated within each group. After stratifying by group, percent inhibition and immune mediator concentrations were compared across visits by the Friedman test. Two-tailed P-values <0.05 were considered statistically significant. The impact of comparing several immune markers on the statistical test results was assessed by using the Benjamini-Hochberg method to adjust the critical value for declaring statistical significance while controlling the false discovery rate (FDR) at 0.05 for each study time point.²⁵ Statistical analysis was performed using SAS v9.2 (SAS Institute, Cary, NC) and GraphPad Prism 6 (GraphPad Software Inc., LaJolla, CA).

RESULTS

Patient and Sample Characteristics

A total of 77 patients, 40 pregnant and 37 non-pregnant, were enrolled in this prospective cohort. The groups were similar in most demographic characteristics, although pregnant women were more likely to be married and less likely to consume alcohol (Table 1).

Table 1.

Demographic Characteristics of Patients at Enrollment

Demographic Characteristic	Pregnant (n=40)	Non-Pregnant (n=37)	P value
Age
Mean (SD)	26.5 (5.1)	26.3 (4.6)
Range	18.0–35.0	18.0–35.0	0.84a
BMI at first study visit
Mean (SD)	27.0 (5.3)	27.8 (7.3)	0.59a
Nugent score at enrollment	1.0 (0 – 10)	3.5 (0 – 10)	0.1
Median (range)
Race
White	25 (62.5)	15 (40.5)	0.11b
Black	6 (15.0)	10 (27.0)
Asian	0 (0.0)	2 (5.4)
American Indian	1 (2.5)	0 (0.0)
Hispanic or Latina	7 (17.5)	10 (27.0)
Other	1 (2.5)	0 (0.0)
Marital Status
Single	14 (35.0)	27 (73.0)	0.0005b
Living with Partner	5 (12.5)	2 (5.4)
Married	21 (52.5)	6 (16.2)
Divorced	0 (0.0)	2 (5.4)
Education Level
Some high school	0 (0.0)	2 (5.4)	0.12c
High school graduate/GED	14 (35.0)	9 (24.3)
1–3 Years College	12 (30.0)	9 (24.3)
College Graduate	6 (15.0)	13 (35.1)
Graduate work	8 (20.0)	4 (10.8)
Employment Status
Unemployed	14 (35.0)	11 (29.7)	0.67b
Full-time	20 (50.0)	16 (43.2)
Part-time	6 (15.0)	8 (21.6)
Other	0 (0.0)	1 (2.7)
Unknown	0 (0.0)	1 (2.7)
Annual family income
< $10,000	6 (15.0)	6 (16.2)	0.14c
$10,000-$24,999	3 (7.5)	11 (29.7)
$25,000-$49,999	8 (20.0)	8 (21.6)
$50,000-$74,999	10 (25.0)	6 (16.2)
$75,000 or more	7 (17.5)	4 (10.8)
Unknown	6 (15.0)	2 (5.4)
Insurance
Private	18 (45.0)	21 (56.8)	0.10b
Medicaid	21 (52.5)	11 (29.7)
Uninsured	1 (2.5)	4 (10.8)
Other	0 (0.0)	1 (2.7)
Average drinks/week (past 3 months)
0 drinks	36 (90.0)	23 (62.2)	0.01b
1–3 drinks	3 (7.5)	11 (29.7)
> 3 drinks	1 (2.5)	3 (8.1)
Tobacco use (past 3 months)
0 ppd	36 (90.0)	33 (89.2)	0.87b
< ½ ppd	3 (7.5)	2 (5.4)
> ½ ppd	1 (2.5)	2 (5.4)
Marijuana use (past 3 months)
0 times/week	38 (95.0)	34 (91.9)	0.67b
1–3 times/week	2 (5.0)	3 (8.1)
Number of times pregnant
Median (Range)	2 (0–6)	0 (0–5)	0.0004d

^aT-test,

^bFisher’s exact test,

^cChi-square test,

^dWilcoxon rank sum test

Vaginal swabs were collected at each visit for pH measurement and microscopic evaluation for presence of yeast, bacterial vaginosis, and trichomonads. There were no significant differences between groups in mean Nugent scores, or in number of participants with bacterial vaginosis at each visit. CVL samples were assessed in a reference laboratory for presence of various immune cells. Numbers of neutrophils, lymphocytes, eosinophils, and basophils did not differ between groups, but non-pregnant women had significantly higher numbers of macrophages in their CVL during enrollment (p=0.04) and first return visit (p=0.05).

HIV Inhibition by Cervicovaginal Fluid

HIV infectivity of TZM-bl cells was significantly reduced in the presence of CVL from both pregnant and non-pregnant women at all visits (p<0.01), and did not differ between groups at any visit (Figure 1). Median percent inhibition of HIV by CVL from pregnant women ranged from 87.0% in the third trimester to 92.9% postpartum, and did not change significantly between trimesters (Supplementary Table 1a–b). Percent inhibition of HIV in the non-pregnant group ranged from 83.9% in the ovulatory phase (return visit 1) to 87.5% in the secretory phase (return visit 2). There was no appreciable difference in the ability of CVL to inhibit HIV across phases of the menstrual cycle.

Figure 1. HIV Infectivity by Visit.

Inhibition of HIV Infection in TZM-bl cells by CVL

Comparison of HIV infectivity of TZM-bl cells without CVL (positive control), and TZM-bl cells with added CVL of pregnant and non-pregnant women. Infectivity measured by Relative Light Units (RLU). HIV infectivity was reduced significantly when CVL was added. Each data point represents one individual patient sample. The box plots represent median and 25^th and 75^th percentiles.

***p<0.0001, **p<0.001, *p<0.01.

*

• Pregnancy visits: Enrollment- <14 weeks gestation
  Return visit 1- 14–28 weeks gestation
  Return visit 2- >28 weeks gestation
• Non-pregnant visits: Enrollment- Proliferative phase
  Return visit 1- Ovulatory phase
  Return visit 2- Secretory phase

Immune Mediators Across Pregnancy and the Menstrual Cycle

Concentrations of AMPs and chemokines previously shown to impact HIV infectivity were measured in CVL samples (Table 2). In this cohort, we observed a generalized suppression of immune mediators during pregnancy, which was most apparent during the second and third trimesters. Concentrations of most mediators were similar between groups at enrollment, which occurred during the first trimester of pregnancy or proliferative stage of the menstrual cycle for non-pregnant women. MIP-3α was the only mediator that was suppressed during the first trimester of pregnancy (p=0.001).

Table 2.

Concentrations of Immune Mediators at Each Visit

Immune Mediator	Enrollment*	Return Visit 1*	Return Visit 2*	Return Visit 3*
Median (Range), pg/mL	n=76	n=67	n=58	n=16
Elafin
Pregnant	879 (173 – 4241)	581 (140 – 4457)	631 (144 – 3872)	769 (176 – 7535)
Non Pregnant	1215 (182 – 14661)	958 (210 – 8308)	1138 (197 – 5072)
P value	0.2	0.4	0.06
HBD-2
Pregnant	1292 (428 – 5828)	1361 (291 – 7411)	1077 (189 – 9999)	1455 (234 – 5896)
Non Pregnant	1758 (8 – 14298)	1914.50 (512 – 16246)	1840 (462 – 7596)
P value	0.3	0.03	0.03a
HBD-3
Pregnant	89 (6 – 2000)	51 (6 – 264)	38 (5 – 563)	6 (6 – 171)
Non Pregnant	39 (6 – 484)	85 (5 – 1884)	95 (6 – 414)
P value	0.1	0.1	0.04
Lactoferrin
Pregnant	51 (7 – 3239)	40 (12 – 4209)	26 (12 – 7318)	56 (12 – 853)
Non Pregnant	73 (8 – 697)	79 (8 – 1000)	101 (12 – 1766)
P value	0.2	0.05	0.001a
LL-37
Pregnant	88 (12 – 3396)	74 (12 – 1757)	53 (12 – 3563)	89 (12 – 1382)
Non Pregnant	77 (12 – 1502)	99 (12 – 810)	168 (12 – 1611)
P value	0.7	0.3	0.003a
MIP-3α
Pregnant	8 (5 – 68)	8 (6 – 73)	8 (6 – 37)	8 (6 – 83)
Non Pregnant	8 (7 – 741)	8 (5 – 1424)	8 (6 – 401)
P value	0.001a	0.05	0.006a
RANTES
Pregnant	6 (6 – 47)	6 (4 – 60)	6 (5 – 103)	6 (4 – 57)
Non Pregnant	6 (6 – 1063)	6 (6 – 242)	6 (4 – 54)
P value	0.06	0.04	0.03
SDF-1
Pregnant	13 (3 – 152)	12 (4 – 174)	9 (4 – 449)	15 (3 – 183)
Non Pregnant	14.50 (4 – 378)	24 (3 – 2000)	36 (4 – 873)
P value	0.6	0.04	0.003a
SLPI
Pregnant	768 (181 – 14222)	960 (100 – 15786)	1328 (142 – 15444)	1152 (115 – 8574)
Non Pregnant	1501 (69 – 14144)	1360 (278 – 16000)	1585 (178 – 11212)
P value	0.2	0.1	0.4

All p-values by Wilcoxon rank sum test

^aStatistically significant controlling for a false discovery rate of 5% for each time point.

^*

• Pregnancy visits: Enrollment- <14 weeks gestation
  Return visit 1- 14–28 weeks gestation
  Return visit 2- >28 weeks gestation
• Non-pregnant visits: Enrollment- Proliferative phase
  Return visit 1- Ovulatory phase
  Return visit 2- Secretory phase

Several immune mediators were progressively attenuated throughout the second and third trimesters of pregnancy. During the second trimester (and ovulatory phase of the menstrual cycle for non-pregnant women), HBD-2 (p=0.03), lactoferrin (p=0.05), RANTES (p=0.04), SDF-1 (p=0.04), and MIP-3α (p=0.05), varied significantly in comparison with the non-pregnant group. Median concentrations of most of these AMPs dropped from the first to the second trimester in pregnant women, while increasing from enrollment to the first return visit in their non-pregnant counterparts. Levels of HBD-2, for instance, dropped significantly from the second trimester to the third within the pregnant group (p=0.03). Data from the third trimester showed an even greater reduction of some of these mediators in pregnancy, as well as significant suppression of HBD-3 (p=0.04) and LL-37 (p=0.001). Levels of HBD-3 decreased significantly throughout pregnancy (p=0.006). CVL concentrations of SLPI and elafin did not differ significantly between groups at any point during the study. After adjusting for multiple comparisons, differences remained statistically significant for MIP-3α in the first trimester (proliferative phase for non-pregnant women) and MIP-3α, lactoferrin, LL-37, SDF-1, and HBD-2 in the third trimester (secretory phase for non-pregnant women).

Association of Immune Mediator Concentrations in CVL with HIV Infectivity

Associations between HIV infectivity and CVL concentrations of immune mediators at each visit were assessed using Spearman rank correlation (Table 3). Concentrations of the AMP elafin were significantly correlated with HIV infectivity in both pregnant and non-pregnant groups throughout all visits except postpartum. SLPI was associated with HIV infectivity throughout all phases of the menstrual cycle in non-pregnant women, but only in the second trimester in pregnant women. HBD-2 was also associated with HIV infectivity in this group, but only during the second (ovulatory phase) and third (secretory phase) return visits. In pregnant women, HIV infectivity was negatively correlated with LL-37 in the second trimester, HBD-3 in the third trimester, and lactoferrin and SDF-1 in both second and third trimesters. Lactoferrin was also negatively associated with HIV infectivity in the non-pregnant group, but at enrollment (proliferative phase) only. At the pregnant postpartum visit (n=16), previously observed correlations between infectivity and immune mediators were lost. However, associations between infectivity and CVL concentrations of HBD-2, as well as RANTES, were observed for the first time in the pregnant group at this visit. Adjusting for multiple comparisons, correlations between inhibition of HIV infectivity and elafin, lactoferrin, and SDF-1 remained significant for pregnant women in the second and third trimesters. For the non-pregnant group, all proliferative phase correlations remained statistically significant, whereas only elafin and HBD-2 remained significantly correlated with infectivity for the ovulatory and secretory phases, respectively.

Table 3.

Correlation of Percent HIV Inhibition with Immune Mediator Concentration

Immune Mediator	Enrollment*	Return Visit 1*	Return Visit 2*	Return Visit 3*
Rho p-value	n=76	n=67	n=58	n=16
Elafin
Pregnant	−0.361	−0.535	−0.485	0.065
p-value	0.022	0.002a	0.014a	0.812
Non Pregnant	−0.437	−0.480	−0.390
p-value	0.008a	0.003a	0.025
HBD-2
Pregnant	0.271	−0.189	0.194	0.506
p-value	0.090	0.309	0.353	0.046
Non Pregnant	−0.275	−0.353	−0.485
p-value	0.104	0.035	0.004a
HBD-3
Pregnant	−0.005	0.305	0.404	0.242
p-value	0.977	0.096	0.045	0.367
Non Pregnant	−0.159	−0.103	0.005
p-value	0.355	0.552	0.978
Lactoferrin
Pregnant	0.228	0.466	0.548	0.159
p-value	0.157	0.008a	0.005a	0.556
Non Pregnant	0.431	0.256	−0.039
p-value	0.009a	0.132	0.830
LL-37
Pregnant	0.248	0.361	0.360	0.033
p-value	0.122	0.046	0.077	0.903
Non Pregnant	0.186	0.055	−0.070
p-value	0.279	0.748	0.699
MIP-3a
Pregnant	0.053	0.012	0.280	−0.072
p-value	0.748	0.950	0.175	0.790
Non Pregnant	0.165	0.231	0.203
p-value	0.337	0.175	0.257
RANTES
Pregnant	0.185	0.235	0.040	−0.553
p-value	0.253	0.203	0.851	0.026
Non Pregnant	0.059	0.088	0.064
p-value	0.732	0.608	0.725
SDF-1
Pregnant	0.190	0.531	0.526	0.261
p-value	0.240	0.002a	0.007a	0.328
Non Pregnant	0.177	0.061	−0.084
p-value	0.301	0.722	0.642
SLPI
Pregnant	−0.159	−0.374	−0.093	0.256
p-value	0.328	0.038	0.658	0.339
Non Pregnant	−0.447	−0.363	−0.391
p-value	0.006a	0.030	0.025

All rho and p-values by Spearman rank correlation

^aStatistically significant controlling for a false discovery rate of 5% for each group and time points.

^*

• Pregnancy visits: Enrollment- <14 weeks gestation
  Return visit 1- 14–28 weeks gestation
  Return visit 2- >28 weeks gestation
• Non-pregnant visits: Enrollment- Proliferative phase
  Return visit 1- Ovulatory phase
  Return visit 2- Secretory phase

Delivery Outcomes

Of the 39 pregnancies for which data were available, there were 37 live births, one first trimester miscarriage at 6 weeks, and one case of intrauterine fetal demise (IUFD) at 19 weeks. Mean gestational age at delivery was 38.3 weeks, with a range of 24.4–41 weeks (Supplementary Table 2). Average birthweight was 3150 grams. Incidence of preterm birth was low, 8.1%.

COMMENT

While several studies have assessed the presence and behavior of AMPs and chemokines during pregnancy, few have focused on those present in the lower genital tract specifically as they relate to HIV infectivity. Our results demonstrate that CVL from pregnant women exerts significant antiviral activity across all trimesters of pregnancy and postpartum, and that this protection is not diminished in pregnancy compared to the non-pregnant state. We also demonstrated the ability of CVL from non-pregnant women to inhibit HIV across all phases of the menstrual cycle to a similar extent. This finding is important given conflicting epidemiological data regarding the risk of a woman’s susceptibility to HIV infection during pregnancy.^9,12 This study provides evidence that if pregnant women are indeed at increased risk for HIV, it is unlikely to be mediated by alterations in the effectiveness of protective cervicovaginal secretions. Other theories of infectivity risk in pregnancy include thinning of the vaginal epithelium by progesterone,²⁶ which was not assessed in this study. In addition, given that there were no significant differences in HIV infectivity across all study visits in both groups, our data calls into question whether changes in the hormonal milieu during pregnancy specifically modulate anti-HIV activity in CVL. Strengths of this study include the longitudinal nature, protocoled sample collection, and careful cohort selection. Limitations include sample size and incomplete follow-up for several participants. Additionally, LH surges were not measured in the non-pregnant group in order to precisely time ovulation and cyclic hormonal changes. While many immune mediators were measured, these are only a portion of those found in the female genital tract. Finally, we used a single strain of X4 virus in our assays. There could be differences in findings seen with the use of wild type or R5 viruses.

The antiviral activity of cervicovaginal fluid is presumed to be conferred in part by the AMPs and chemokines investigated in this study. There are several mechanisms by which these immune mediators are thought to directly inhibit HIV infection, including prevention of HIV binding to target cell co-receptors, or destabilization of the viral membrane.²⁷ Interestingly, most of these mediators, including HBD-2, HBD-3, lactoferrin, LL-37, MIP-3α, RANTES, and SDF-1 were suppressed during pregnancy, especially during the second and third trimesters. This result may be attributed to the unique characteristics of the immune milieu during mid-gestation. This period has been described as one of “immune quiescence,” characterized by suppression of pro-inflammatory cytokines by high levels of progesterone.²⁸ In light of the ability of pro-inflammatory cytokines to induce expression of AMPs,^29,30 this result is not surprising. However, the reason why CVL from pregnant women is equivalent to that of non-pregnant women in its ability to inhibit HIV, despite containing lower concentrations of AMPs and chemokines, is unknown.

In this study, we found that concentrations of the AMP elafin were correlated with increased HIV infectivity by CVL from both pregnant and non-pregnant women at all study visits. Higher concentrations of SLPI were also consistently correlated with increased HIV infectivity, but in non-pregnant women only. This is consistent with findings from a study by Ghosh et al, which established no correlation of either elafin or SLPI with HIV suppression in healthy, non-pregnant women.¹⁸ Considering the established anti-HIV properties of elafin and SLPI, our results were somewhat unexpected. However, high levels of these mediators in CVL have been previously associated with adverse pregnancy outcomes such as cervical shortening and preterm birth suggesting that the concentrations we found were indicative of a response to some stimulus in vivo.^31,32

In the second and third trimesters of pregnancy, we found lactoferrin, LL-37, SDF-1, and HBD-3 were positively correlated with HIV inhibition. These correlations were lost in the postpartum period, possibly due to sample size limitations of this optional follow-up visit. While these associations have not been previously tested in CVL from pregnant women, Shust et al found that lactoferrin, but not HBD-3, in CVL from non-pregnant women was positively correlated with inhibition of herpes simplex virus in vitro.³³ It is unclear why, in our study, these particular mediators demonstrated greater antiviral activity in context of the pregnant immune milieu.

In conclusion, this study demonstrates significant anti-HIV activity in the CVL of pregnant and non-pregnant women in a carefully selected longitudinal cohort. We did not find that pregnancy increased the risk of HIV infectivity in this in vitro model of cervicovaginal fluid. By detecting positive associations of several AMPs with in vitro HIV inhibition, this study provides a foundation for the development of preventative approaches at the level of the genital tract, such as vaginal microbicides based on a woman’s own immunity. Future pre-clinical studies could explore the possibility of utilizing specific AMPs in vaginal microbicide product development in order to prevent cellular infection with HIV.

Abbreviations

CVL

Cervicovaginal lavage

AMP

Antimicrobial peptide

MIP-3α

Macrophage inflammatory protein-3α

RANTES

Regulated upon activation, normally T-cell expressed and secreted

HBD

Human beta defensin

WHO

World Health Organization

CXCL

Chemokine ligand

IRB

Institutional Review Board

RLU

Relative light units

LTR

Long terminal repeat

SDF-1

Stromal cell-derived factor 1

SLPI

Secretory leukocyte protease inhibitor

PBS

Phosphate buffered saline

IUFD

Intrauterine fetal demise