Effects of the levonorgestrel‐releasing intrauterine device on the immune microenvironment of the human cervix and endometrium

Problem There is little information regarding the impact of the intrauterine device on immune parameters of the upper female reproductive tract related to risk of HIV acquisition. Method of Study We collected cervical and endometrial samples from women using the hormonal intrauterine device to study its effects on endocervical cytokines/chemokine concentrations, phenotypic markers of T cells, responses of endometrial T cells to activation, and alterations of endometrial cellular infiltrates. Results Hormonal intrauterine device use was associated with: increased concentrations of inflammatory cytokines/chemokines (endocervix); increased coexpression of CXCR4 and CCR5 (endocervix and endometrium); increased coexpression of CD38 and HLADR (endocervix and endometrium); increased intracellular IL‐10 production after T‐cell stimulation (endometrium); and increased density of T cells, most notably regulatory T cells (endometrium). Conclusion Hormonal intrauterine device use resulted in both inflammatory and immunosuppressive alterations. Further research is needed to determine the significance of these changes for HIV risk.

acquiring HIV than women using no contraceptive or non-hormonal contraceptives. 6,7 These findings raise questions about the effects of progestin contraceptives with regard to susceptibility to HIV infection: is risk specific to DMPA, which has unique steroid actions; is risk related to systemic delivery of progestin; or is risk related to long-term continuous progestin exposure? Thus, it is important to more fully understand the effects of locally released LNG from the LNG-IUD because IUD contraceptives are highly effective and increasingly used. The biological basis for a possible association between progestins and HIV risk is not well understood. Some in vitro studies have suggested that endogenous progesterone may lead to increased markers of HIV susceptibility. 5,8 However, the effect of exogenous progestins on immune function differs by progestin type. 9 There is little information regarding the possible impact of LNG on innate and adaptive immune responses in the female reproductive tract. Additionally, IUDs are foreign bodies that result in local inflammation in the uterus, which may contribute to both pre-and post-fertilization contraceptive effectiveness. [10][11][12] Inflammation resulting from the IUD could also result in recruitment of HIV target cells, increasing HIV susceptibility.
Any effect of LNG-IUDs on susceptibility to HIV infection has not been evaluated rigorously in epidemiologic studies. The limited research that is available has not indicated an association of use of the coppercontaining IUD with an increased risk of HIV infection, but more complete assessment is needed. 13 The World Health Organization (WHO) medical eligibility criteria for contraceptive use states, "women at high risk of acquiring HIV can generally use LNG-IUDs," indicating that "the advantages of using this method generally outweigh the theoretical or proven risks". 14 In 2012, the WHO convened a technical panel to address ongoing concerns that hormonal contraceptives might increase the risk of HIV infection. This panel concluded that there is an urgent need for further research in both epidemiology and basic science to evaluate effects of hormonal contraceptives on HIV susceptibility. 15 The purpose of this study was to determine the effect of a commonly used progestin-releasing foreign body, the LNG-IUD, on the properties of mucosal immunity of the upper female reproductive tract. To characterize the unperturbed immune microenvironment, the control group consisted of samples from women not using hormonal or intrauterine contraception that were collected at the time of peak progesterone levels (6-10 days after ovulation). We studied levels of cytokines/chemokines in the endocervical canal, phenotypic markers of endocervical and endometrial T cells, responses of endometrial T cells to activation, and the characteristics of endometrial cellular infiltrates, in order to provide a comprehensive characterization of the effects of LNG-IUD use on the immune microenvironment of mucosal sites that may support HIV transmission.

| Study design
This is a cross-sectional non-randomized comparison of women using LNG-IUD with women using no hormonal contraception. The UCSF Committee on Human Research approved the study protocol, recruiting, and consent materials (approval # 10-01063).

| Recruitment of human volunteers
Healthy women volunteers from San Francisco and the greater Bay Area were recruited via flyers placed in a variety of venues and advertisements in local publications. Participants agreed to refrain from using vaginal products (creams, douches) for at least 10 days prior to study biopsies. Additionally, participants agreed to either abstain from vaginal intercourse or to use non-lubricated condoms for 72 hr prior to biopsy procedures. Women in the LNG-IUD groups were required to have had the IUD in place for at least 6 months (Table 1). Women in the control group reported no use of exogenous sex steroids for at least 3 months and a history of at least three normal and consecutive menstrual periods since discontinuation. In addition, if they gave birth in the past year, women in the control group were required to have a history of at least six normal and consecutive menstrual periods. Exclusion criteria included: age <18 or ≥45 years, positive HIV serology, positive urine nucleic acid amplification test for Neisseria gonorrheae or Chlamydia trachomatis, current or recent pregnancy or breastfeeding, recent gynecological symptoms, recent history of irregular menstrual cycles and/or use of vaginal products, current or frequent genital herpes recurrences, an abnormal cervical cytology in past year, use of other hormonal treatments in past year, use of systemic corticosteroids or immune-modulating therapies, or daily use of non-steroidal anti-inflammatory agents. Also excluded were women unwilling/unable to refrain from vaginal intercourse during the 72 hr before specimen collection. Women received payment to compensate for time and effort required for study participation.

| Clinical study procedures
Women in the control group were instructed to measure their urine for luteinizing hormone (LH) detection using a home detection kit (Clearblue ® Ovulation Test DIGITAL, Proctor & Gamble, Cincinnati, OH, USA). Within 7-11 days of urine LH detection (i.e., after ovulation), a study visit occurred for collection of study specimens. Prior to specimen collection, participants were asked whether they had engaged in vaginal intercourse within the prior 72 hr, and if they had, the visit was canceled and the participants were asked to reschedule during the subsequent menstrual cycle. Peripheral blood was collected in EDTA tubes, and serum progesterone was measured to validate the occurrence of ovulation. A speculum was inserted into the vagina, the cervix was visualized, and assessment made for the presence of vaginitis or cervicitis. If vaginal discharge was present, a wet mount was performed. If bacterial vaginosis, candidiasis, or trichomoniasis was diagnosed, the participant was offered treatment and biopsies were not performed. If the exam was normal, the following specimens inserted into the canal for 90 s, followed by a second identical collection); endocervical sample using a endocervical cytobrush sample (Cytobrush ® Plus Cell collector, CooperSurgical, Trumbull, CT, USA) turned three times; a cervical biopsy (performed at the transformation zone if visible or at the os using a Mini-Tischler punch biopsy forceps); and an endometrial biopsy using with a 3-mm cannula (Miltex brand | 139 Softflex) inserted through the internal os into the endometrial cavity.
For the latter procedure, if insertion of the cannula was difficult, local anesthesia was provided via cervical injection with lidocaine, and then a tenaculum was applied to the ectocervix. A second pass with the curette was made if the amount of tissue was observed to be insufficient after the first pass.

| Endocervical wick cytokine/chemokine measurements
Endocervical wick samples were snap-frozen at the time of collection and stored at −80°C until analysis in bulk (Fig. 1). Wick samples were weighed and then extracted following published techniques into 300 μL ice-cold extraction buffer (PBS, 0.25 mol/L NaCl + 0.1 mg/mL aprotinin), centrifuged, and extracted a second time in 300 μL of extraction buffer. 16 Wicks were allowed to air-dry for 24 hr, then weighed again; the dry wick weight was subtracted from the initial wet weight to determine the net weight (i.e., volume of fluid extracted from each wick). Two wick samples were collected consecutively from each participant, and the two samples were pooled after extraction for each participant for analysis. Samples were assayed on the Milliplex panel

| Phenotypic and intracellular cytokine staining and flow cytometry
For cell surface phenotyping, endocervical and endometrial cells were stained immediately following sample processing and leukocyte isolation as previously described (Table 2; Fig. 4). 18 For measurement of intracellular cytokine production, endometrial cells were rested overnight in complete medium prior to performing stimulation assays. The next morning, 2-3 × 10 6 cells in 200 microliters complete medium F I G U R E 2 Flow cytometry gating used in the analysis of endocervical and endometrial T-cell phenotypes. After initial gating of lymphocytes (based on forward vs side scatter) and doublet discrimination, dead cells were excluded by staining with live/dead viable amine; viable cells that were CD3 + CD66b − (not shown) were then subdivided into CD4 + or CD8 + populations. The resulting CD4 + or CD8 + T cells were then assessed for expression of three pairs of phenotypic markers, as described in the text. Shown from left to right, these were as follows: differentiation markers CCR7 and CD45RA; activation markers CD38 and HLA-DR; chemokine receptors CCR5 and CXCR4. Quadrant gates were drawn based on fluorescence-minus-one (FMO) controls. Numbers in each quadrant indicate percentages of CD4 + or CD8 + T cells expressing various combinations of markers. Data shown are from a representative participant using LNG-IUD Complete medium containing anti-CD107, monensin, and brefeldin A served as a negative control. Following a five-hour incubation, cells were incubated for 5 min in phosphate-buffered saline (PBS)/2% FCS/0.5 mmol/L EDTA, stained for surface markers and cell viability using aqua amino reactive dye in PBS/2% FCS for 20 min at 4°C, fixed in 4% formaldehyde, then permeabilized using FACS Perm 2 (BD Biosciences). Cells were then washed in PBS/2% fetal calf serum, stained for intracellular cytokines and CD3, washed again, then stored at 4°C in PBS/1% formaldehyde until analysis within 24 hr. The expression of CD107, IL-2, IL-10, IL-17, TNFalpha, IFNgamma, and MIP-1beta was measured as described previously. Flow cytometry data were acquired and analyzed as described above. were used for detection of antibody binding. Each analysis included an appropriate negative control using mouse IgG1 or IgG2a (DAKO).

| Immunohistochemistry of endometrium
Photomicrographs of randomly selected fields using the 40× objective of an Olympus BX51 microscope were captured using a CCD camera (Spot Camera; Diagnostic Instruments, Sterling Heights, MI, USA). Positively stained cells were counted on photomicrographs of up 5-10 fields from each slide. As the tissue in the photomicrograph did not always fill the field, the area that contained tissue was outlined and measured on a grid; counts represent the number of stained cells divided by the number of grids included in the image. Tissue areas covering more than 96 grids (of a total possible 382.5 grids) were included in the analysis. To avoid counting projections from a single CD68 + cell as multiple cells, CD68 staining had to be associated with a nucleus to be counted as a CD68 + cell. To be counted as a FoxP3 + cell, the staining had to be nuclear.

| Participant characteristics
We summarized and compared characteristics of study participants by exposure group (Table 1). For continuous characteristics (age and duration of current LNG-IUD exposure), we report median (min, max) and Wilcoxon rank-sum test P-values. For categorical characteristics (race, smoking status, and ovulation status), we report percentages per group and Fisher's exact test P-values.

| Wick analyses
Wick sampling and processing yielded one observation per biomarker per participant (Fig. 1). Preliminary histograms revealed that distributions of protein concentrations (pg/mL) of all biomarkers measured were right-skewed; consequently, concentrations were analyzed as linear functions of exposure group (LNG-IUD vs unexposed) and age (centered at 30 years old) using generalized estimating equation (GEE) models, assuming negative binomial distributions and using a log link F I G U R E 4 Panel (a) intracellular cytokine production by CD4 + and CD8 + T cells following polyclonal stimulation. For analysis of endometrial T-cell responses to stimulation, production of six cytokines (IFN-γ, IL-2, IL-10, IL-17, MIP-1β, and TNF-α) and the granule-associated membrane protein CD107 were measured by flow cytometry as described in the text. Initial gating was performed to identify lymphocytes and remove doublets (not shown), followed by gating of viable CD3 + T cells on subsets expressing either CD4 or CD8, and finally for individual responses as indicated on bivariate plots. Numbers indicate the percentages of CD4 + or CD8 + T cells in each quadrant. Data shown are from a representative participant using LNG-IUD with SEB-stimulated cells. Panel (b) production of IL-10 by CD4 + and CD8 + T cells from endometrial tissue of LNG-IUD users ("IUD") and control women ("REF") following stimulation with PMA-ionomycin (blue boxes, left) or SEB (red boxes, right). Box plots illustrate the percent responding cells by subject group, cell type, and stimulus. The length of each box represents the interquartile range (IQR; the distance between the 25th and 75th percentiles), and the interior line represents the median (50th percentile). A symbol (diamond) denotes the mean. "Whiskers" are drawn to the most extreme observations that lie within the fences. The upper fence is defined as the third quartile plus 1.5 times the interquartile range (IQR), and the lower fence is defined as the first quartile minus 1.5 times the interquartile range. Observations outside the fences are identified with small circles | 143 and robust standard errors. Age-adjusted group-specific mean (95% CI) concentrations per biomarker were plotted side-by-side, sorted by concentrations in the control group. The corresponding age-adjusted exposure effects, expressed as a mean (95% CI) relative concentration (fold-change) among exposed versus unexposed participants, were plotted in the same order. We also report Wald statistic P-values from the GEE analyses.

| Flow cytometry
Flow cytometry assays, based one endocervical and one endometrial sample per participant, generated person-level bivariate distributions of three pairs of phenotypes-CCR7/CD45RA, CD38/HLADR, and CXCR4/CCR5-separately for CD4 + and CD8 + cell types (Fig. 3). The  CONTROL LNG-IUD generated by the latter model which are based on empiric standard errors. We present these results via plots of the mean bivariate distributions by cell type, tissue, and arm (Fig. 3).

| Immune mediators
Studies of the effects of stimulation of CD4 + and CD8 + cells from endometrial tissue by PMA-ionomycin or Staphylococcal enterotoxin B (SEB) yielded four observations per immune mediator per participant, expressed as percentages ( Table 2; Fig. 4). We used GEE models, assuming negative binomial distributions and using log links and empirical standard errors. Each model expressed log percentage of a phenotype as a linear function of exposure group, stimulant, and their interaction, stratified by cell type. For each biomarker, we report the mean (95% CI) percentage in the control group and the mean (95% CI) exposure effect, expressed as the ratio (fold-change) among exposed versus unexposed participants (Table 2). Finally, we present box plots of the raw data per exposure group for biomarker IL-10 to allow comparison of model-based findings with raw data (Fig. 4b).

| Immune cell densities
For each specimen per participant, biomarker cell counts were determined in multiple standardized areas (see Methods for Immunohistochemistry above) ( Table 3). Only participants with 5-10 qualifying areas were included in the analysis, generating different sample sizes per marker. Counts of each biomarker were analyzed as above (negative binomial distribution, logarithm link, robust variance estimator) as a function of study group; additionally, an offset used to adjust for the sizes of the areas counted. Results were transformed back to the measurement scale for reporting.
Statistical analyses were conducted using SAS version 9.4. All P-values cited are two-sided and values less than α = .05 were considered statistically significant.

| Demographics
Nineteen LNG-IUD users and 27 control women were included in the study (Table 1). LNG-IUD users were younger than control women

| Effects of LNG-IUD use on chemokines, cytokine, and innate immune factors in the endocervical canal
To determine whether LNG-IUD altered the immune milieu of the endocervix, we studied concentrations of 13 proteins in endocervical fluid in samples from 19 LNG-IUD users and 24 control women (Fig. 1).
In general, protein concentrations were elevated among IUD users relative to controls, with statistically significant age-adjusted effects for 6 of 13 endocervical cytokine and chemokine levels. Comparing LNG-IUD users with control women, significant effects were identified for MCP1

| Effects of LNG-IUD use on the phenotype distributions of endocervical T cells
Representative flow cytometry gating used in the analysis of endocervical and endometrial T-cell phenotypes is shown in Fig. 2. To determine whether LNG-IUD use was associated with changes in immune characteristics of the T cells present in endocervix, the cell surface markers on CD4 + and CD8 + T-cell subsets collected via endocervical brushings from 25 controls and 17 IUD users were analyzed by flow cytometry (Fig. 3). In previous studies, we found that when compared with T cells in peripheral blood, a greater proportion of endocervical T cells were activated, effector memory cells, 18 a pattern we again observed among samples from control women from this study (data not shown). The proportion of CD4 + T cells co-expressing both HIV co-receptors (CXCR4 + CCR5 + ) was increased among LNG-IUD users compared to controls. (Fig. 3c). For both CD4 + and CD8 + T cells, the proportion expressing markers of the central memory compartment (CCR7 + CD45RA − ) was reduced among LNG-IUD users compared to controls (Fig. 3a). In addition, endocervical CD8 + T cells from LNG-IUD users demonstrated an increase in the proportion of cells expressing the activation marker HLADR (CD38 -HLADR + and CD38 + HLDR + ) (Fig. 3b).

| Effects of LNG-IUD use on the phenotype distributions of endometrial T cells
To determine whether LNG-IUD use was associated with changes in immune characteristics of the T cells present in endometrium, the cell surface markers on CD4 + and CD8 + (Figs 2 and 3) T cell subsets collected via endometrial biopsies from 16 women per group were analyzed by flow cytometry. In endometrium obtained from control women, >50% of CD4 + T cells expressed markers of an effector memory phenotype as well as CD38 and CCR5, as previously reported. 18 In samples from LNG-IUD users, the proportions of endometrial CD4 + T cells expressing the activation marker HLADR were increased compared with controls (CD38 − HLADR + and CD38 + HLADR + ), whereas the proportions of CD4 + T cells expressing CD38 were decreased (CD38 + HLADR − ) (Fig. 3b). The proportions of endometrial CD4 + T cells expressing the HIV co-receptor CXCR4 were significantly increased (CXCR4 + CCR5 − , CXCR4 + CCR5 + ), whereas those expressing CCR5 without CXCR4 expression were decreased (CXCR4 − CCR5 + ) ( Fig. 3c).
For CD8 + T cells, the proportions of endometrial cells expressing the activation marker HLADR were increased (CD38 − HLADR + and CD38 + HLADR + ), whereas the proportion of CD8 + T cells expressing CD38 without HLADR was decreased (CD38 + HLADR − ). The proportions of CD8 + T cells expressing the HIV co-receptor CXCR4 were significantly increased (CXCR4 + CCR5 − , CXCR4 + CCR5 + ) among LNG-IUD users, whereas the proportion expressing CCR5 without CXCR4 was decreased (CXCR4 − CCR5 + ). Figure 3 also allows the T-cell distribution to be compared between the endocervix and the endometrium. As reported previously, T-cell phenotypes in the two tissues are markedly different despite their anatomic proximity. 18 By comparing the graphs from control participants for cervix and endometrium, it is apparent that endocervix had significantly greater proportions of naïve, central memory, and terminally differentiated effector CD4+ T cells, as well as higher proportions of naïve and terminally differentiated effector CD8 + T cells than T cells recovered from the endometrium (Fig. 3a). However, endometrium had a significantly higher proportion of activated CD4 + and CD8 + T cells (CD38 + HLADR + ) than endocervix (Fig. 3b). With regard to the distribution of T cells that expressed HIV co-receptors, the proportions T A B L E 3 Quantification of cell types by immunohistochemistry in endometrial biopsy specimens, based on samples from 28 participants of CD4 + and CD8 + T cells expressing CCR5 (CXCR4 − CCR5 + and CXCR4CCR5 + ) were higher in the endometrium (Fig. 3c).

| Effects of LNG-IUD use on the response of endometrial T cells to stimulation
To determine whether the function of endometrial T cells differed between the LNG-IUD users and control women, we analyzed intracellular expression of soluble immune mediators for CD4 + (Table 2A) and CD8 + (Table 2B) Fig. 4b).

| Effects of LNG-IUD use on densities of immune cells in the endometrium
To determine whether the quantity of key immune cell types in endometrium differed between LNG-IUD users and control women, we performed IHC on endometrial biopsy tissue sections using markers for NK cells (CD56), macrophages (CD68), T cells (CD4 + and CD8 + ), and regulatory T cells (FoxP3) ( Table 3). In endometrium from control women, NK cells were present at the highest densities, followed by macrophages. CD8 + T cells were present at approximately twice the density of CD4 + T cells. FoxP3 + regulatory T cells were the least common cell type, at a density that could account for approximately half of the population of CD4 + T cells in the endometrium.
In endometrial samples obtained from LNG-IUD users, the densities of NK cells and macrophages were comparable to controls (Table 3)

| DISCUSSION
Our results from control participants not using hormonal con- The strengths of our study are the wide range of immune parameters that were measured, and the collection of samples from comparison participants timed to a narrow window of the ovulatory cycle, thus minimizing the effects of ovarian sex steroid variations on our findings. In addition, by enrolling women who had chronic exposure to the LNG-IUD, we were able to characterize the impact of this IUD in a timeframe that better mirrors actual use than post-insertion observations. This study is limited by relatively small numbers of individuals and samples analyzed, the fact that absence of recent sexual activity was documented by self-report only, and that the LNG-IUD group had a lower percentage of women who had ovulated compared to controls. Ultrasound studies have demonstrated that ovulation occurs in a subset of women using the LNG-IUD. 27 We assume that the local release of LNG would dominate any effects of progesterone released cyclically in the women who ovulated, but were unable to determine whether the release of progesterone from ovulation might have contributed to our findings, or perhaps masked a larger effect from the LNG-IUD had we controlled for this variable. Additionally, we cannot infer the effects of LNG-IUD directly as we did not study women prior to and after device insertion. We were unable to determine whether our results reflect the presence of a foreign body versus the local release of LNG in the endometrium. Comparison of effects of the copper (non-hormonal) IUD to those of LNG-IUD will help distinguish between these possibilities.
Our results indicate LNG-IUD use resulted in both inflammatory and immunosuppressive changes (the latter being perhaps compensatory in nature) in the local immune microenvironment of the endocervix and endometrium. Further research is needed to determine the effects of these changes on HIV infection of target cells from LNG-IUD users, and on overall risk of HIV acquisition.