Background

Pregnant women are at higher risk of acquiring Human Immunodeficiency Virus (HIV) infection compared to non-pregnant women [1–4]. If HIV positive, pregnant women may also transmit HIV more frequently to their uninfected partner than non-pregnant women (International Microbicides Conference 2010, abstract #8). The increased susceptibility for HIV infection during pregnancy is independent of sexual behavior and likely due to biological causes [1]. However, the underlying mechanisms for both the increased susceptibility and infectivity are unknown.

Previous studies have shown that sex hormones influence female genital tract immunity [5] and a large body of literature analyzed effects of proinflammatory cytokine concentration in cervicovaginal secretions with bacterial vaginosis and preterm birth [6]. However, comprehensive comparisons of cervicovaginal cytokine concentrations between pregnant and non-pregnant women have to our knowledge not been conducted.

Given the central role of cytokines and other mucosal immunomodulatory factors in HIV pathogenesis [7], and the profound systemic changes during immunity [8, 9], we hypothesized that pregnancy may result in shifts of the cervicovaginal cytokine profile that may increase the risk of HIV infection. To explore this hypothesis, we conducted a comprehensive analysis of immunomuodulatory factors in samples collected from pregnant and non-pregnant women.

Methods

Study population

The study enrolled 23 pregnant and 25 non-pregnant women attending the Obstetrics and Gynecology clinic at the University of Southern California Medical Center in Los Angeles between February and April 2008. Healthy women between 17 and 45 years of age were invited to enroll in the study if they were: not on hormonal contraception in the last 6 months, had no intrauterine device, did not report sexual intercourse within the last 24 h and were not actively menstruating. All women underwent a clinical examination. Women with bacterial vaginosis or candidiasis were subsequently excluded from the analysis, resulting in a final study population of 21 pregnant and 24 non-pregnant women.

All women provided written consent and the study was approved by the institutional review board at the University of Southern California, Los Angeles, CA and Children's Hospital Los Angeles, CA.

Data and sample collection

Socio-demographic, obstetric and gynecological data were collected by a structured questionnaire. A digital picture of the cervix was taken with an inserted endocervical wick (Tear-Flo™) serving as length standard. After removal of the endocervical wicks, a CVL sample was collected by bathing the cervical os in phosphate buffered saline (PBS). Fluid in the vaginal vault was then collected with a transfer pipette and stored on ice until transported to the laboratory for processing within 4 h. Blood was collected with ethylenediaminetetraacetic acid (EDTA) as anticoagulant and cells were separated from CVL sample and blood by low speed centrifugation. Supernatants were frozen at -80°C until measurement.

Lab assays

Measurements of immunomodulatory factors were conducted with the Milliplex™ Map Human Cytokine/Chemokine KIT for the measurement of 39 premixed cytokines (Millipore, Billerica, MA) using the Luminex technology (Luminex Corporation, Austin, TX) as specified in manufacturer's instructions. Protein concentration in CVL was measured using the Quick-Start Bradford Dye Reagent (Bio-Rad, Hercules, CA) according to the manufacturer's instructions. All samples were assayed in duplicate and the mean between these measurements used for all analysis.

Cervical ectopy

The size of the cervix as well as any visible endocervical epithelium covering the ectocervix was determined by planimetry and expressed in mm2 using an image processing program (Adobe Photoshop)[10, 11]. The degree of cervical ectopy was expressed as the percent of the visible endocervix covering the ectocervix.

Statistical analysis

Concentrations outside of the range of the multiplex assay were imputed with a value just below the lower detection limit of the kit plus 2 standard deviations or above the upper cut-off (10,000 pg/ml) of the assay. To adjust for variation in the volume of phosphate buffered saline (PBS) used during the CVL (10 or 12 ml) concentrations determined from samples with lower dilution (10 ml) were divided by a factor of 1.2. Where appropriate, variables were log10 transformed to approximate normal distributions.

We used the chi-square test to compare categorical variables, unless the expected cell counts below 5, in which case we used Fisher's exact test. The chi-square trend test was used to compare ordered categorical variables, the Wilcoxon rank sum test for non-normally-distributed continuous variables, and the T-test for normally distributed variables. Bonferroni correction was used to correct for multiple testing. Pearson correlation coefficient were calculated for normally distributed and Spearman correlation coefficient for non-normally distributed variables.

Linear regression modeling was conducted to adjust for confounding. Variables were retained in the final model if they remained significantly associated with log10-transformed CCL22 concentration or if they changed the effect estimate by more than 15%. Two missing datapoints for the time since last coitus were imputed with the population median of 7 days to retain all women in the regression analysis. All statistical analyses were performed using SAS software (Version 9.2, Cary, NC).

Results

Characteristics of the study population

As expected, pregnant women more frequently reported vaginal discharge and had more pronounced cervical ectopy than non-pregnant women. Pregnant women were also younger than non-pregnant women. They, however, did not differ from non-pregnant women in other socio-economic or gynecological variables (Table 1).

Table 1 Characteristics of 21 pregnant and 24 non-pregnant women included in the analysis.
Full size table

Twenty six (66%) of the 39 measured immunomodulatory factors were detectable in at least half of the CVL samples included in the analysis. With the exception of IL-1ra, none of immunomodulatory factors were frequently above the detection limit of the assay (Table 2).

Table 2 Concentrations of immunomodulatory factors in CVL of 21 pregnant and 24 non-pregnant women from Los Angeles.
Full size table

Pregnancy and immunomodulatory factors in CVL

CVL collected from pregnant women contained threefold lower concentrations of C-C motif chemokine 22 (CCL22; also known as macrophage-derived chemokine) than CVL from non-pregnant women (mean ± standard deviation [SD] of log10 pg CCL22 per ml 1.5 ± 0.4 versus 2.0 ± 0.5, p = 0.0011, geometric mean: 29.7 pg/ml vs 89.7 pg/ml). Pregnant women also had lower CVL concentrations of monocyte chemotactic protein-1 (MCP-1) than non-pregnant women (mean ± SD of log10 pg per ml among pregnant women: 1.6 ± 0.6 versus non-pregnant women 2.0 ± 0.7, p = 0.03; geometric mean: 41.1 pg/ml versus 111.4 pg/ml); however, only the difference in CCL22 remained significant when adjusted for multiple testing (p < 0.0013).

The difference in CVL CCL22 concentration between pregnant and non-pregnant women strengthened slightly when pregnant women were restricted to those in their third trimester (n = 13) (mean ± standard deviation [SD] of log10 pg CCL22 per ml 1.4 ± 0.4 versus 2.0 ± 0.5, p = 0.0011, geometric mean: 23.6 pg/ml vs 89.7 pg/ml). There additionally was a strong negative correlation between gestational age and CCL22 concentration in CVL (Spearman correlation coefficient [RS]: -0.49, p = 0.0006) when analyzed in the whole population assigning a gestational age of "0" to non-pregnant women (Figure 1). There was, however, no significant association of CCL22 concentration with gestational age when restricted to pregnant women (RS = -0.21, p = 0.34)

Figure 1
figure 1

Scatter plot of CCL22 concentration in cervicovaginal lavages with gestational age among 21 pregnant and 24 non-pregnant women. Non-pregnant women were assigned a gestational age of 0 weeks.

Full size image

Correlation of CVL CCL22 concentration with other immunomodulatory factors

As listed in Table 3, the concentration of CCLL22 concentration correlated or tended to correlated with that of Eotaxin, Fractalkine, GM-CSF, GRO, IL-17, IL-9, IP-10, MCP-1, MCP-3, TGFα, TNFβ and VEGF in CVL. There was, however, no correlation with the total protein concentration in CVL (RS = 0.10; p = 0.51) or with plasma CCL22 concentration (RS = 0.06, p = 0.69).

Table 3 Immunomodulatory factors associated with CCL22 in CVL
Full size table

Can the difference in CVL CCL22 concentration between pregnant and non-pregnant women be explained by other factors?

To test whether underlying differences between pregnant and non-pregnant women may have caused the difference in CCL22 concentration, we conducted linear regression modeling adjusting for possible confounders. In univariate analysis, age, time since last coitus and the size of cervical ectopy were additionally associated with CCL22 concentration in CVL (Table 4). When adjusted for age and time since last coitus, the association between pregnancy and CCL22 concentrations remained strong. None of the other variables shown in Table 1, including cervical ectopy or vaginal discharge, remained significantly associated with the CCL22 concentration or appreciably changed the effect estimate when additionally included in the model. The results are similar, when gestational age instead of pregnancy is included in the model or when the CCL22 concentration is expressed as ratio of CCL22 to total protein in CVL to adjust for possible variation during the sample collection.

Table 4 Linear regression modeling of log10 transformed CCL22 concentrations in CVL among 21 pregnant and 24 non-pregnant women.
Full size table

Discussion

Among this group of healthy American women, we found threefold lower concentration of CCL22 in CVL samples from pregnant women than in those from non-pregnant women. This difference remained significant when corrected for multiple testing or in adjusted analysis, suggesting that pregnancy may result in reduced concentration of cervicovaginal CCL22.

The strength of this study lies in the large number of analyzed cytokines and immunomodulatory factors among healthy women. To our knowledge, this is the most comprehensive analysis of these factors in CVL of pregnant and non-pregnant women. In contrast to at least some previous studies [12–15], we did not detect differences in proinflammatory cytokines between pregnant and non-pregnant women. This could be explained by the sample size, the exclusion of women with clinical bacterial vaginosis, the relatively early collection of samples or fluctuations of these cytokines throughout pregnancy. We are however the first study that tested CCL22 concentrations in CVL.

CCL22 has previously been detected in other mucosal sites including the intestine [16], the lung [17] and the endometrium [18] as well as in vaginal tissue in mice [19], thus its presence in vaginal tissues among humans seems plausible.

Previous studies have described fluctuations of CCL22 expression in endometrium during the menstrual cycle and increases in the same tissue during early pregnancy [20], suggesting a control by sex hormones. While it is unclear, what caused the decreased CCL22 concentrations among pregnant women in our study, progesterone has been shown to suppress the NF-κB transcription factor [21], which is an activator of CCL22 expression [22]. It is therefore possible that increased progesterone concentrations directly result in reduced CCL22 expression, which should be tested in vitro. In addition, CCL22 in our analysis was associated with a number of immunomodulatory factors, especially Ip10 and MCP-1. It was also increased shortly after coitus. Thus it is likely that there are a number of other physiological and immunological mechanisms that also influence CCL22 concentrations in CVL [23].

Intriguingly, CCL22 has been implied in the HIV pathogenesis in several ways. CCL22 is a T-helper cell type (TH) 2 cytokine that is highly expressed in macrophages and dendritic cells of the monocyte line [17] as well as in activated T-cells [24]. It is a strong chemoattractant for leukocytes expressing the CCR4 receptor [23] and has been suggested to be a key regulator of innate immunity in mice [25]. In at least some in vitro studies, CCL22 has been suggested to have HIV suppressive effects [26–29]. Such mechanisms could explain the increased risk of HIV infection with decreased CCL22 concentration. However, at least one other study suggested that CCL22 is secreted by CD16+ monocyte-derived macrophages to activate resting T-cells for HIV infection [30] and may therefore also increase the risk of HIV infection in certain situation. Thus further analysis of the effects of CCL22 on mucosal cytokine concentration is required.

As in all statistical analysis, we cannot exclude that differences in CVL concentrations of CCL22 are caused by chance. However, given the strong difference observed here, its possible regulation by sex hormones and its possible implication in HIV pathogens, a role of CCL22 in mediating a protection against HIV at the female genital mucosa seems plausible and should be investigated further.

Conclusion

In this cohort, pregnancy is associated with reduced CCL22 concentration in cervicovaginal secretion, which may influence the risk of HIV infection.