Menopause and estrogen associations with gut barrier, microbial translocation, and immune activation biomarkers in women with and without HIV

Abstract

Objectives:

Estrogens may protect the gut barrier and reduce microbial translocation and immune activation, which are prevalent in HIV infection. We investigated relationships of the menopausal transition and estrogens with gut barrier, microbial translocation, and immune activation biomarkers in women with and without HIV.

Design:

Longitudinal and cross-sectional studies nested in the Women’s Interagency HIV Study.

Methods:

Intestinal fatty acid binding protein (IFAB), lipopolysaccharide binding protein (LBP), and soluble CD14 (sCD14) levels were measured in serum from 77 women (43 with HIV) before, during, and after the menopausal transition (~6 measures per woman over ~13 years). A separate cross-sectional analysis was conducted among 72 post-menopausal women with HIV with these biomarkers and serum estrogens.

Results:

Women in the longitudinal analysis were a median age of 43 years at baseline. In piece-wise linear mixed-effects models with cut-points 2 years before and after the final menstrual period to delineate the menopausal transition, sCD14 levels increased over time during the menopausal transition (Beta [95% CI]=38 [12, 64] ng/mL/year, p=0.004), followed by a decrease post-transition (−46 [−75, −18], p=0.001), with the piece-wise model providing a better fit than a linear model (p=0.0006). In stratified analyses, these results were only apparent in women with HIV. In cross-sectional analyses among women with HIV, free estradiol was inversely correlated with sCD14 levels (r=-0.26, p=0.03). LBP and IFAB levels did not appear related to the menopausal transition and estrogen levels.

Conclusion:

Women with HIV may experience heightened innate immune activation during menopause, possibly related to depletion of estrogens.

INTRODUCTION

HIV infection leads to persistent immune activation despite adherence to antiretroviral therapy (ART), which may contribute to higher risk of non-AIDS-related conditions (e.g., cardiovascular disease, neurocognitive decline, cancer) in people with HIV compared to those without HIV¹. Microbial translocation, a process in which damage to the gut epithelial barrier allows microbial products to translocate from the gut to the circulation, is a suspected cause of immune activation in people with HIV².

Recent experimental evidence suggests that estrogens play a protective role in both maintenance of gut barrier integrity^3–6 and HIV pathogenesis (i.e., HIV replication and immune response)^7,8. Consequently, the menopausal transition, accompanied by sharp declines in endogenous estrogens, may be a period of risk for microbial translocation and immune activation in women with HIV. In a cohort of women without HIV, biomarkers of gut barrier dysfunction, microbial translocation, and immune activation increased within-woman from pre- to post-menopause⁹, suggesting that menopause may increase microbial translocation. In the Women’s Interagency HIV Study (WIHS), we previously showed that biomarkers of innate immune activation (soluble CD14 [sCD14] and sCD163), but not gut barrier dysfunction, were higher in post-menopausal compared to pre-menopausal women with HIV, with sCD14 possibly increasing during the menopausal transition¹⁰. However, that study, which used a pre-existing dataset, was limited in several ways: (1) short follow-up (up to 2 years/3 samples per woman), resulting in few women with both pre- and post-menopausal biomarker data; (2) lack of a direct biomarker of microbial translocation; and (3) lack of a comparison group of women without HIV, to aid in understanding whether observed associations, and possible underlying mechanisms, are unique to women with HIV.

Here, we conducted a new longitudinal study of gut barrier, microbial translocation, and immune activation biomarkers in serum among women with and without HIV in the WIHS. Namely, these biomarkers are: intestinal fatty acid binding protein (IFAB), a cytosolic protein in enteroctyes that is released upon gut epithelial damage¹¹; lipopolysaccharide binding protein (LBP), an acute phase protein which binds to lipopolysaccharide (LPS), a cell wall component of gram-negative bacteria¹²; and sCD14, a marker of monocyte activation¹³. Importantly, we measured these biomarkers across a long time span (~13 years on average) before, during, and after the menopausal transition, allowing us to closely examine their relationship with the menopausal transition. Further, we explored associations of these biomarkers with serum estrogens, in a separate cross-sectional analysis of post-menopausal women with HIV from WIHS. We hypothesized that the biomarkers would increase during the menopausal transition, and show inverse associations with serum estrogens in post-menopause. Together, these analyses provide a more comprehensive view of estrogen-related impacts on gut barrier dysfunction, microbial translocation, and immune activation in women with and without HIV.

METHODS

Study population.

The WIHS was the largest multicenter cohort of women with HIV and socio-demographically similar women without HIV in the U.S., collecting clinical, demographic, and behavioral data semi-annually through interviews, physical exams, and laboratory tests ^14–16. Institutional review boards at all sites approved the study, and participants provided written informed consent. The current study features two analyses: (1) a longitudinal analysis of biomarkers across the menopausal transition (which we call “menopause analysis”), and (2) a cross-sectional analysis of biomarkers in relation to circulating estrogens (“serum estrogens analysis”).

For the menopause analysis, women were selected for biomarker measurement based on detailed inclusion and exclusion criteria, to ensure participants had a confidently observed natural final menstrual period (FMP; i.e., menopause) with sufficient follow-up before and after the FMP, while not experiencing major adverse events that may impact biomarker values (Supplementary Figure 1). From 4,982 participants and 101,667 person-visits in the WIHS, we restricted to women with a confidently observed natural FMP and age at FMP ≥40 years (n=426 women; explained further below), and excluded women with HIV seroconversion during the study (n=25 women) or hepatitis C virus (HCV) seropositivity at any time during the study (n=1,448 women). We also excluded person-visits with pregnancy (n=4,037 person-visits), hormonal contraceptive use (n=6,991 person-visits), or use of hormone therapy (n=31,989 person-visits), and person-visits that occurred at any time after a major medical event (e.g., AIDS event, cancer, cardiovascular event; n=48,312 person-visits). Finally, we restricted to women with at least 6 person-visits spread across the following time categories: 2 person-visits 3+ years before FMP; 1 person-visit 0–3 years before FMP; 1 person-visit 0–2 years after FMP; and 2 person-visits 2+ years after FMP. Based on these criteria, a total of 79 participants with 474 person-visits were selected, of which 77 participants (451 person-visits) were available in the biorepository and served as the final sample size for biomarker measurement (Supplementary Figure 1).

For the serum estrogens analysis, we used a cross-sectional dataset of post-menopausal women with measured sex hormones, previously described¹⁷. Among 197 women with measured sex hormones, 91 had biomarker data available. Due to the small sample size of women without HIV in the cross-sectional data (n=19), this analysis focused only on women with HIV (n=72).

Defining the natural FMP.

We used each participant’s longitudinal survey history since their enrollment in the WIHS to identify if/when they reached their natural or surgical FMP, using a multi-step process described previously¹⁸. Briefly, we obtained the date of FMP from the ‘first day of most recent period’ at the last visit with reported menses. For the purposes of this study, women were considered to have a confidently observed natural FMP if they: (1) did not report having a menstrual period after first reporting menopause (i.e., consistent self-report over time), (2) did not report a hysterectomy or bilateral oophorectomy before or around the time of FMP, (3) did not have >1 visit missing around the time of FMP, and (4) had ≥2 visits consistently reporting menopause after the FMP.

Biomarker measurement.

Biomarkers were measured in serum at the Laboratory for Clinical Biochemistry Research at the University of Vermont. IFAB and sCD14 were measured using Quantikine^™ enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems, Minneapolis, MN, USA), and LBP was measured using an R-PLEX^® electrochemiluminescence immunoassay kit (Meso Scale Diagnostics, Rockville, MD, USA), following manufacturer’s instructions. Longitudinal samples from the same participant were measured in the same batch.

Sex hormone measurement.

Sex hormones and sex hormone binding globulin (SHBG) were measured in serum at the pharmacogenomics laboratory at Université Laval, using validated methods described previously^19,20. Briefly, gas-chromatography (GC) coupled to mass spectrometry (MS) was used to quantify estrone (E1) and estradiol (E2)¹⁹, while liquid-chromatography (LC) tandem MS was used for estrone sulfate (E1-S)²⁰. SHBG was measured by ELISA (Diagnostics Biochem Canada, Inc., Ontario, Canada). Values below lower limit of detection were imputed with half the lower limit of detection. Free estradiol was calculated using a formula with estradiol, SHBG and a constant for albumin (43 g/L)^21,22.

Covariate data.

Data for analysis were taken from the WIHS core visit at which each serum sample was collected. Covariates considered for analysis are shown in Table 1. Estimated glomerular filtration rate (eGFR) was calculated using the CKD-EPI creatinine equation without race²³. Missing covariate data were imputed based on the immediately prior study visit with available data.

Table 1. Characteristics of participants at the baseline measurement of gut barrier, microbial translocation, and immune activation biomarkers, in the Women’s Interagency HIV Study.

	All	Women with HIV	Women without HIV	P-valuea
n	77	43	34
Age, years (median [IQR])	43.00 [40.00, 44.00]	42.00 [40.00, 44.00]	43.50 [40.00, 44.75]	0.52
Age at FMP, years (median [IQR])	50.00 [49.00, 53.00]	50.00 [48.00, 53.00]	50.00 [49.00, 52.00]	0.73
Number of person-visits (median [IQR])	6.00 [6.00, 6.00]	6.00 [6.00, 6.00]	6.00 [6.00, 6.00]	0.85
Total timeb, years (median [IQR])	13.39 [9.98, 14.05]	13.69 [12.69, 14.09]	12.43 [8.94, 13.96]	0.07
Race/Ethnicity, n (%)				0.93
Black non-Hispanic	49 (63.6)	27 (62.8)	22 (64.7)
White non-Hispanic or other	15 (19.5)	9 (20.9)	6 (17.6)
Hispanic	13 (16.9)	7 (16.3)	6 (17.6)
Born outside the U.S. 50 states/DC, n (%)	21 (27.3)	13 (30.2)	8 (23.5)	0.69
Annual income, n (%)				0.24
$12000 or less	37 (48.1)	22 (51.2)	15 (44.1)
$12001-$24000	14 (18.2)	5 (11.6)	9 (26.5)
$24001 or more	26 (33.8)	16 (37.2)	10 (29.4)
Educational attainment, n (%)				0.09
Less than high school	24 (31.2)	9 (20.9)	15 (44.1)
Completed high school	24 (31.2)	15 (34.9)	9 (26.5)
Any college	29 (37.7)	19 (44.2)	10 (29.4)
Employed, n (%)	40 (51.9)	23 (53.5)	17 (50.0)	0.94
Smoking status, n (%)				0.05
Never smoker	23 (29.9)	16 (37.2)	7 (20.6)
Current smoker	40 (51.9)	17 (39.5)	23 (67.6)
Former smoker	14 (18.2)	10 (23.3)	4 (11.8)
Current alcohol use, n (%)				0.39
Abstainer	35 (45.5)	22 (51.2)	13 (38.2)
>0–7 drinks/wk	34 (44.2)	16 (37.2)	18 (52.9)
>7 drinks/wk	8 (10.4)	5 (11.6)	3 ( 8.8)
Current substance use, n (%)	15 (19.5)	7 (16.3)	8 (23.5)	0.61
eGFR, ml/min/1.73m2 (median [IQR])	94.81 [82.37, 109.30]	99.09 [88.45, 111.01]	91.97 [80.47, 96.36]	0.02
AST, IU/L (median [IQR])	20.00 [15.00, 24.00]	21.00 [17.00, 24.50]	18.00 [15.00, 20.75]	0.05
ALT, IU/L (median [IQR])	17.00 [13.00, 25.00]	21.00 [14.00, 25.50]	14.50 [11.25, 22.00]	0.07
Waist circumference, cm (median [IQR])	91.20 [82.00, 99.80]	89.00 [81.75, 97.60]	92.75 [84.50, 101.23]	0.58
BMI, kg/m2 (median [IQR])	29.20 [25.60, 33.10]	27.30 [24.85, 32.50]	30.25 [27.40, 33.40]	0.19
Systolic blood pressure, mmHg (median [IQR])	120.00 [110.00, 128.00]	122.00 [107.50, 129.50]	117.50 [111.25, 126.75]	0.78
Diastolic blood pressure, mmHg (median [IQR])	75.00 [68.00, 80.00]	77.00 [67.00, 81.50]	72.00 [70.00, 77.75]	0.39
Diabetes, n (%)	3 ( 3.9)	1 ( 2.3)	2 ( 5.9)	0.84
Diabetes medication, n (%)	1 ( 1.3)	1 ( 2.3)	0 ( 0.0)	0.99
Lipid-lowering medication, n (%)	2 ( 2.6)	2 ( 4.7)	0 ( 0.0)	0.58
Anti-hypertensive medication, n (%)	10 (13.0)	7 (16.3)	3 ( 8.8)	0.53
Viral load, copies/mL (median [IQR])		80.00 [80.00, 1550.00]
CD4+ T-cell count, cells/mm3 (median [IQR])		470.00 [379.50, 674.00]
Antiretroviral therapy regimenc, n (%)
None		8 (18.6)
PI-based		17 (39.5)
II-based		0 (0)
NNRTI-based		12 (27.9)
NRTI-based or other		6 (14.0)
IFAB, pg/mL (median [IQR])	1255.63 [834.10, 1960.23]	1594.17 [988.27, 2037.16]	1216.42 [788.37, 1659.46]	0.12
LBP, ng/mL (median [IQR])	5452.00 [4391.00, 7123.00]	5402.00 [4533.50, 6806.00]	5890.00 [4437.25, 7471.75]	0.45
sCD14, ng/mL (median [IQR])	1931.27 [1673.90, 2284.49]	2121.68 [1860.15, 2402.24]	1754.84 [1557.56, 2106.79]	0.001

^aP-value comparing women with vs. without HIV from Wilcoxon rank-sum test for continuous variables and Chi-square test or Fisher’s exact test for categorical variables. ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; eGFR, estrimated glomerular filtration rate; FMP, final menstrual period; IFAB, intestinal fatty acid binding protein; II, integrase inhibitor; LBP, lipopolysaccharide binding protein; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nuceloside reverse transcriptase inhibitor; PI, protease inhibitor; sCD14, soluble CD14.

^bTime from first to last biomarker measurement.

^cPI-based: At least 1 PI and 1 NRTI; II-based: at least 1 II and 1 NRTI; NNRTI-based: at least 1 NNRTI and 1 NRTI; NRTI-based: 3 or more NRTIs.

Statistical analysis.

General principles.

Descriptive characteristics were compared for women with vs. without HIV using Wilcoxon rank-sum test for continuous variables and the Chi-square or Fisher’s exact test for categorical variables. Analyses were conducted among all participants, and stratified by HIV serostatus when sample size allowed. P<0.05 was considered statistically significant. All analyses were conducted using R 4.2.2.

Biomarker descriptive analysis.

Consistency of biomarkers within participants over time was assessed using intra-class correlation coefficients. Correlations between biomarkers over time were assessed using repeated-measures correlations (‘rmcorr’ package, R)²⁴. Multivariable generalized estimating equation (GEE) models with the working independence correlation structure were used to evaluate associations of HIV serostatus and other participant characteristics (i.e., those shown in Table 1) with biomarker outcomes.

Menopause analysis.

The purpose of this analysis was to differentiate whether biomarkers increase at a constant rate over time, or accelerate during the menopausal transition, suggesting an effect of ovarian aging beyond that of chronological aging. We used previously described methods^25,26 involving piece-wise linear mixed-effects models to estimate associations of years before/after FMP with biomarkers in time segments of pre-transition, menopausal transition, and post-transition. We chose a priori cut-points of 2 years before and after FMP to delineate time segments, based on evidence that estradiol begins to decline on average 2 years before, and stabilizes approximately 2 years after, the FMP²⁷. All models included a random intercept, a term for years before/after FMP, and terms for interaction of the second time segment (>2 years before FMP) and third time segment (>2 years after FMP) with years before/after FMP, to allow rate of change in biomarkers to vary by time segments. We considered nested models to serially adjust for potential confounders which have been associated with the biomarkers in prior literature^28–30 or in this study (see ‘Biomarker descriptive analysis’ described above; covariates with p<0.1 for one or more biomarkers). Model 1 (demographic model) adjusted for age at FMP, HIV serostatus, race/ethnicity, income, educational attainment, and employment status; Model 2 (behavioral model) additionally adjusted for smoking status, alcohol use, and substance use; Model 3 (cardiometabolic model) additionally adjusted for body mass index (BMI), diastolic blood pressure, use of lipid-lowering medication, use of anti-hypertensive medication, eGFR, serum aspartate transaminase (AST), and serum alanine transaminase (ALT); and Model 4 (HIV model, women with HIV only) additionally adjusted for viral load, CD4+ T-cell count, and antiretroviral therapy regimen. Covariates were time-varying, excepting time-fixed variables (e.g., age at FMP, race/ethnicity). Likelihood ratio tests were used to compare piece-wise models to their respective linear models (i.e., without interaction terms of time segments and years before/after FMP).

Serum estrogens analysis.

We used Spearman’s correlation to examine cross-sectional relationships of biomarkers with serum estrogens in women with HIV. We examined both unadjusted and partial correlations, the latter adjusting for factors most strongly associated with serum estrogens or SHBG in this study population (waist circumference, eGFR, hemoglobin A1c, HCV serostatus, diabetes medication), as identified in our previous research¹⁷.

Data availability statement.

Data in this manuscript were collected by the WIHS, now the MACS/WIHS Combined Cohort Study (MWCCS). Access to individual-level data may be obtained upon review and approval of a MWCCS concept sheet. Instructions for concept sheet submission are on the study website (https://statepi.jhsph.edu/mwccs/).

RESULTS

Participant characteristics.

Among 77 women (56% with HIV) included in the menopause analysis, median age was 43 years at first biomarker measurement, and median age at FMP was 50 years (Table 1). Biomarkers were measured in a median of 6 serum samples per participant, spanning a median of 13 years from first to last biomarker measurement (Table 1). At first biomarker measurement, 81% of women with HIV were on ART, and women with HIV had worse kidney and liver function than women without HIV (Table 1).

Characterization of biomarkers.

Among the biomarkers, LBP had the highest intra-class correlation coefficient (0.64), indicating moderate consistency over time, followed by sCD14 (0.54) and IFAB (0.38) (Supplementary Table 1). Women with HIV had higher consistency for LBP and sCD14, and lower consistency for IFAB, compared to women without HIV (Supplementary Table 1). Both IFAB and LBP were positively correlated with sCD14 (r=0.25 and 0.22, p<0.0001), while IFAB was not correlated with LBP (r=0.02, p=0.72), and these results were similar in women with and without HIV (Supplementary Figure 2). Additionally, IFAB (adjusted difference=465 pg/mL, p=0.006) and sCD14 (264 ng/mL, p=0.0003) were higher in women with HIV compared to women without HIV in multivariable GEE models, while LBP did not differ significantly by HIV serostatus (Supplementary Table 2). A number of other participant characteristics, including race/ethnicity, smoking, alcohol use, substance use, BMI, and diastolic blood pressure, were associated with one or more biomarkers in women with or without HIV, in multivariable GEE models (Supplementary Table 2). In women with HIV, higher CD4+ T-cell counts were associated with lower sCD14, while ART regimens (excepting integrase inhibitor-based regimens) were associated with higher sCD14 compared to women not on ART (Supplementary Table 2).

Menopause and biomarkers of gut barrier dysfunction, microbial translocation, and immune activation.

Nonparametric locally weighted scatter-plot smoothing curves suggested an increase in sCD14, but not IFAB or LBP, during the menopausal transition (Figure 1; Supplementary Figure 3). In piece-wise linear mixed-effects models with cut-points of 2 years before and after FMP to delineate the menopausal transition, and adjusting for potential confounders, we observed that sCD14 increased over time during the menopausal transition (Beta [95% CI]=38 [12, 64] ng/mL/year, p=0.004), followed by a decrease post-transition (Beta [95% CI]=-46 [−75, −18], p=0.001) (Table 2; Figure 2). The piece-wise model provided a better fit than a linear model (p=0.0006), consistent with an effect of ovarian aging beyond chronological aging (Table 2; Figure 2). This relationship of sCD14 with the menopausal transition was apparent when women with HIV were examined separately; however, in women without HIV, the piece-wise model did not provide a better fit than the linear model (Figure 2; Supplementary Table 3).

Figure 1. Scatter plots of gut barrier, microbial translocation, and immune activation biomarkers over years before/after the final-menstrual-period (FMP).

In scatter plots of (a) IFAB, (b) LBP, and (c) sCD14, lines connect points from the same woman. Plots are overlaid with nonparametric locally weighted smoothing curves, shown with black curves and 95% confidence intervals. Cut-points of 2 years before and after the FMP are shown with dashed vertical lines. Each plot features 77 women (451 person-visits).

Table 2. Relationship of years before/after the final menstrual period (FMP) with gut barrier, microbial translocation, and immune activation biomarkers measured repeatedly (~6 times) over ~13 years per participant.

Estimates are from piece-wise linear mixed-effects models with 3 time segments: pre-transition (>2 years before FMP), menopausal transition (2 years before to 2 years after FMP), and post-transition (>2 years after FMP). All models included a random intercept, a term for years before/after the FMP, and terms for the interaction of the second time segment (>2 years before FMP) and the third time segment (>2 years after FMP) with years before/after FMP. All models were adjusted for age at FMP, HIV serostatus, race/ethnicity, income, educational attainment, employment status, smoking status, alcohol use, substance use, BMI, diastolic blood pressure, use of lipid-lowering medications, use of anti-hypertensive medication, eGFR, serum AST, and serum ALT. The likelihood ratio test was used to compare the piece-wise model to a linear model (i.e., without the interaction terms of time segments and years before/after FMP, i.e., constant slope over time). Model includes 77 women (451 person-visits).

	IFAB			LBP			sCD14
	Beta pg/mL/year	95% CI	P-value	Beta ng/mL/year	95% CI	P-value	Beta ng/mL/year	95% CI	P-value
Slope in pre-transitiona	26.85	−33.42, 87.12	0.38	32.37	−43.99, 108.72	0.41	4.82	−11.46, 21.09	0.56
Slope in menopausal transitiona	78.3	−18.31, 174.9	0.11	−2.76	−125.11, 119.58	0.97	37.89	11.8, 63.97	0.004
Slope in post-transitiona	−23.71	−128.77, 81.35	0.66	50.95	−82.16, 184.06	0.45	−46.26	−74.64, −17.88	0.001
Difference in slopeb, menopausal transition vs. pre-transition	51.45	−87.56, 190.46	0.47	−35.13	−210.88, 140.62	0.70	33.07	−4.43, 70.57	0.09
Difference in slopeb, post-transition vs. menopausal transition	−102.01	−276.65, 72.63	0.25	53.71	−167.38, 274.81	0.63	−84.15	−131.3, −36.99	0.0005
Likelihood ratio test			0.48			0.90			0.0006

^aBeta represents slope of years before/after FMP with biomarker in the specified time segment.

^bBeta represents difference in slope of years before/after FMP with biomarker, for the specified time segment comparison.

Figure 2. Relationship of years before/after the final menstrual period (FMP) with soluble CD14 (sCD14).

Estimates are from a linear model (dashed line) or a piece-wise linear mixed-effects model (solid line) with 3 time segments: pre-transition (>2 years before FMP), menopausal transition (2 years before to 2 years after FMP), and post-transition (>2 years after FMP). The linear model included a term for years before/after the FMP, while the piece-wise model included terms for years before/after the FMP, and the interaction of the second time segment (>2 years before FMP) and the third time segment (>2 years after FMP) with years before/after FMP. All models included a random intercept. Sample sizes (number of women/person-visits): all (77/451), women with HIV (WWH; 43/255), women without HIV (WWOH; 34/196).

For the outcomes of IFAB and LBP, change over time did not differ for the pre-transition, menopausal transition, or post-transition time segments, either in the combined study population nor in women with or without HIV separately (Table 2; Supplementary Tables 4–5).

Serum estrogens and biomarkers of gut barrier dysfunction, microbial translocation, and immune activation in women with HIV.

Among 72 post-menopausal women with HIV (median age 56 years; Supplementary Table 6) with concurrent serum estrogens and biomarker measures, greater free E2 was correlated with lower sCD14 (r=-0.26, p=0.03), which remained significant after adjustment for potential confounders (Figure 3). Due to continuing estradiol decline after menopause, we conducted a sensitivity analysis additionally adjusting for years from FMP in the subset of women with an observed natural FMP (n=45); in this subset, free E2 remained correlated with sCD14 after years from FMP adjustment (r=-0.35, p=0.03).

Figure 3. Correlations of serum estrogens with serum biomarkers of gut barrier, microbial translocation, and immune activation among women with HIV (n=72).

Heatmaps show (a) unadjusted and (b) partial Spearman’s correlations adjusted for waist circumference, eGFR, hemoglobin A1c, HCV serostatus, and diabetes medication. *p<0.05. E1, estrone; E1-S, estrone-sulfate; E2, estradiol.

DISCUSSION

In this first long-term longitudinal analysis of gut barrier, microbial translocation, and immune activation biomarkers across the menopausal transition, we observed that serum sCD14, but not IFAB or LBP, increased specifically during the menopausal transition in women with HIV. This result is consistent with our previous short-term study showing an increase in serum sCD14 but not IFAB during the menopausal transition in women with HIV¹⁰; as well as with a large cross-sectional analysis of baseline data from the Randomized Trial to Prevent Vascular Events in HIV (REPRIEVE) trial, in which post-menopausal women had significantly higher sCD14 than pre-menopausal women independent of age²⁸. Further, we found that lower free E2 in serum was correlated with higher serum sCD14, but not LBP and IFAB, in women with HIV. While the inverse correlation of free E2 and sCD14 was consistent with the menopause analysis, the correlation was weak and requires replication in a larger study. Taken together, our findings suggest that women with HIV may experience heightened innate immune activation during menopause, possibly related to depletion of estrogens during menopause. However, the estrogen-related effects on immune activation may be independent of gut barrier dysfunction and microbial translocation, at least as measured by the biomarkers IFAB and LBP.

Serum biomarkers of innate immune activation, particularly sCD14, have been associated with increased risk of cardiovascular disease and mortality in treated HIV infection^31–37. As such, an increase in sCD14 during the menopausal transition in women with HIV may be concerning; however, we also observed a decrease in sCD14 during the post-transition time period, making it unclear whether the short-term menopause-related increase in sCD14 would have long-term health effects. Additionally, the magnitude of increase over time in sCD14 during the menopausal transition was lower than in our prior report¹⁰, possibly related to greater precision in the current analysis which had better within-woman longitudinal sampling. Future research may be able to determine whether trajectories of immune activation across the menopausal transition are linked with health outcomes in women with HIV. In women without HIV, we observed a more linear increase in sCD14 over time, suggestive of chronological aging, and consistent with positive associations of age and sCD14 in other populations without HIV³⁸. We speculate that in women with HIV, where HIV infection is driving higher sCD14 levels, the effect of aging on sCD14 is diminished, while the effect of the menopausal transition on sCD14 is more apparent; and that in women without HIV, the gradual increase in sCD14 with aging may diminish the effect of the menopausal transition on sCD14. Consistently, a lack of association between age and sCD14 in people with HIV on ART has been previously observed^39,40.

Estrogens may protect against viral replication and HIV-related immune activation in women with HIV, providing an alternative mechanism by which estrogen depletion may result in immune activation independent of microbial translocation. For example, estrogens have been observed in vitro to inhibit HIV transcription in T-cells⁷, and to downregulate the inflammatory response of monocyte-derived macrophages in response to HIV infection ⁸. Additionally, a recent study of women with HIV found that levels of inducible HIV RNA increased during reproductive aging⁴¹. Thus, it is possible that depletion of estradiol during menopause in women with HIV may impact HIV control, resulting in immune activation.

Although menopause and estrogens were unrelated to gut barrier dysfunction and microbial translocation in this study population, prior research suggests mechanisms unrelated to HIV by which estrogens may protect the gut barrier and reduce microbial translocation and immune activation. In a study of 65 women without HIV, women had higher IFAB, LBP, and sCD14 in serum after menopause than before menopause, while lower estradiol and higher follicle stimulating hormone (FSH) in serum were significantly correlated with higher serum IFAB and sCD14⁹. The protective effects of estrogens on the gut barrier are supported by experimental evidence showing that estrogen receptor-β signaling maintains colonic epithelial barrier function^4,5, estradiol upregulates tight junction proteins in intestinal epithelial cells⁶, and ovariectomy increases intestinal permeability in mice⁴². To our knowledge, no other human studies have evaluated the relationship of menopause, serum estrogens, and biomarkers of gut barrier function and microbial translocation. While our findings suggest that the menopausal transition and estrogens are not related to gut barrier function, additional large longitudinal studies are needed in both HIV and non-HIV human populations.

Interestingly, in this study population of women with and without HIV, serum LBP was not higher in women with vs. without HIV, in contrast to observations in other study populations where LBP is typically higher in people with HIV compared to those without^29,30,43,44. Additionally, serum IFAB and LBP were not correlated with each other, despite the theoretical mechanistic connection between gut barrier dysfunction, represented by IFAB, and subsequent microbial translocation, represented by LBP. Lack of correlation between serum IFAB and LBP has been observed consistently in previous literature^45–50, and may suggest an intermediate step between gut barrier dysfunction and the acute response to microbial translocation that is not captured by either biomarker. However, both IFAB and LBP were positively correlated with serum sCD14, indicating their link with immune activation. Research involving a larger panel of microbial translocation biomarkers (e.g., LPS, 16S ribosomal DNA, endotoxin core antibody) may aid in identifying a better surrogate, if any, for microbial translocation in this study population.

This study was strengthened by the long-term longitudinal measurement of biomarkers before, during, and after the menopausal transition in the same women, which allowed more precise characterization of their relationship with menopause, and is an approach that has never been undertaken previously. Additional strengths included our analysis of serum estrogens with biomarkers to support menopause-related findings, as well as the basis of our analyses in the well-characterized WIHS cohort, providing detailed information on many covariates. Our study was limited by moderate sample size, particularly among women without HIV. A further limitation was characterization of reproductive age by self-report, rather than an objective biomarker such as anti-mullerian hormone, which may be ideal in women with HIV who are more likely to have prolonged amenorrhea for reasons other than menopause⁵¹. However, we minimized misclassification risk by using longitudinal self-report data to define menopause, and by carefully selecting participants with an observed natural FMP for the longitudinal analysis. The correlation analysis of serum estrogens and biomarkers was additionally limited by its cross-sectional design which prevents causal and temporal inferences, inability to include women without HIV due to small sample size, and by inclusion of only post-menopausal women who have inherently low estrogen levels, precluding generalizability to pre- and peri-menopausal women. Lastly, our results may not be generalizable to other populations of women with or without HIV.

In conclusion, in this first longitudinal study of gut barrier, microbial translocation, and immune activation before, during, and after the menopausal transition, we found that immune activation increases specifically during the menopausal transition in women with HIV. Additional research is needed to determine whether this short-term menopause-related increase in immune activation may be associated with risk of co-morbidities in the longer-term, such as cardiovascular disease¹⁸. Better understanding of the risks associated with menopause in women with HIV may also encourage further research on the potential benefits of menopausal hormone therapy, which is currently under-utilized and under-researched in women with HIV⁵².