Douching Is Associated With Dysregulated Rectal Mucosal Immunity in Sexual Minority Men

Abstract

Background

Receptive condomless anal sex (CAS) associates with elevated rectal inflammation and mucosal injury, increasing HIV acquisition risk. Although douching may amplify rectal inflammation and alter microbial communities, this has not been well characterized in sexual minority men (SMM).

Methods

Ninety-two SMM (median age, 34.6 years) who were HIV negative and reported receptive CAS provided rectal swabs during sexually transmitted infection (STI) clinic visits. Associations among rectal douching, rectal cytokine/chemokine levels, and microbial communities, evaluated via immunoassay and 16S rRNA gene sequencing, respectively, were assessed.

Results

When compared with nondouching SMM (n = 27), SMM who douched (n = 64) reported more receptive CAS partners and displayed elevations in rectal cytokine/chemokines linked to immune activation and inflammation. Lower microbial richness, evenness, and Shannon diversity in SMM who reported douching were observed. Significant associations were identified between microbial alpha diversity metrics and rectal chemokine/cytokine levels. Finally, significant correlations were observed between rectal cytokine/chemokine levels and individual microbial genera.

Conclusions

Among SMM engaging in receptive CAS, douching may identify those with amplified biobehavioral HIV and STI risk. Elucidating the mechanisms whereby douching dysregulates rectal immune function and alters rectal microbial communities could yield targets for biomedical approaches to optimize HIV/STI prevention in SMM during receptive CAS.

Among sexual minority men engaging in condomless anal sex, douching was associated with elevated rectal chemokines and cytokines linked to immune activation, which correlated with alterations in rectal microbial community structure, suggesting that douching may identify those with amplified biobehavioral HIV and sexually transmitted infection risk.

In 2020, 71% of new HIV-1 diagnoses in the United States occurred among sexual minority men (SMM), with 68% attributable to male-to-male sexual contact [1]. Among SMM, receptive condomless anal sex (CAS) is characterized by increased inflammatory immune cell frequencies, elevated proinflammatory cytokine production, heightened epithelial cell proliferation, and upregulated injury and repair gene signatures as compared with men with no anal intercourse history [2–4]. Rectal microbial community perturbations occur in SMM reporting receptive CAS, including enrichment in Prevotella, as compared with men who have never engaged in anal intercourse and independent of HIV infection status and condom usage [2, 5–7]. Importantly, microbial community structure shifts may contribute to elevated immune activation and cellular proliferation, creating an environment conducive to HIV acquisition and viral replication [2–7]. Thus, rectal mucosal immune and microbial disturbances due to receptive CAS could amplify HIV acquisition risk among SMM. Examining behavioral factors that modify rectal mucosal immunity is essential to guide comprehensive HIV and sexually transmitted infection (STI) prevention efforts in SMM engaging in receptive CAS.

Rectal douching is the practice of rinsing the rectum in preparation for, or to clean after, receptive anal sex, which is common among SMM [8–11]. In a US national sample of SMM, two-thirds engaged in rectal douching, and those who douched reported more receptive and insertive CAS partners in the last 3 months [12]. Mechanical and chemical trauma from douching could damage the rectal epithelial lining and incite mucosal inflammation that synergistically combines with receptive CAS to prime the rectal microenvironment for HIV/STI acquisition [8, 10, 13]. Indeed, commercial enemas and water with soap are commonly used as douching agents and have been shown to cause rectal surface epithelium loss [14]. Additionally, douching with hyperosmolar vehicles caused epithelial denudation and sloughing, suggesting that water-based douches may trigger epithelial disruption [15, 16]. Moreover, 1 in 10 men report bleeding after douching, indicating rectal injury [12]. An important knowledge gap is whether douching associates with alterations in rectal mucosal immunity (eg, cytokine/chemokine levels) and microbial community structure that could amplify HIV/STI risk. In this cross-sectional study, we hypothesized that SMM reporting receptive CAS and douching would have elevated rectal inflammation and microbial community structure alterations as compared with SMM who did not report douching.

METHODS

Study participants were recruited from 4 STI clinics operated by the AIDS Healthcare Foundation in either Fort Lauderdale or Miami, Florida; via flyers at local restaurants, clubs, and preexposure prophylaxis (PrEP) clinics; or through social media advertisement. Eligibility criteria included being ≥18 years old, self-reporting as cisgender SMM who were HIV negative, and the ability to complete a questionnaire in English. Participants completed a brief screening questionnaire to assess eligibility. Community-recruited participants completed the screening questionnaire by telephone. After eligibility confirmation, participants provided written informed consent, completed a quantitative survey on psychosocial and behavioral factors, and submitted urine for toxicology testing. Rectal swabs were collected from participants who reported at least 1 of the following: taking PrEP, having any CAS partners in the past 3 months, or engaging in stimulant use (cocaine, methamphetamine) in the past 3 months. In total, 279 participants were enrolled. A subset of participants who had rectal swabs, receptive CAS, and no antibiotic use in the past 3 months was selected for the current analysis (n = 92). All procedures were approved by the University of Miami Institutional Review Board (20180044).

Behavioral Measures

Demographics

Sociodemographic factors were assessed via questionnaire and included age, race, ethnicity, education, income, sexual orientation, marital status, and country of birth.

Receptive CAS

Participants reported the number of men with whom they engaged in receptive CAS within the last 3 months.

Alcohol and Substance Use

Hazardous drinking was quantified by the Alcohol Use Disorders Identification Test; scores ≥4 were coded as positive for hazardous alcohol use [17]. Participants were classified as using cannabis or stimulants (ie, methamphetamine, powder cocaine, or crack-cocaine) based on any self-reported use in the past 3 months or reactive urine toxicology results.

Depressive Symptoms

Based on the Center for Epidemiologic Studies Depression 10-item scale (study Cronbach α = 0.85), scores ≥10 were classified as screening positive for depression [18].

Rectal Douching

Douching was assessed by asking, “During the last three months, have you ever douched to clean out your butt?”

Specimen Collection and Storage

Urine samples and rectal swabs were self-collected by study participants. Urine samples were immediately used for on-site toxicology screening by iCup Drug Screens (Redwood Toxicology Laboratory) to detect recent use of stimulants (methamphetamine, cocaine, amphetamines), cannabis, and opiates. Immediately after self-collection, rectal swabs were placed into collection tubes containing 2 mL of DNA/RNA Shield (Zymo Research) and kept at room temperature for no more than 48 hours, followed by storage at −80 °C.

Quantification of Rectal Cytokine Levels

The LEGENDPlex Human Inflammation Panel (BioLegend) was used to characterize rectal levels of 13 human inflammatory cytokines and chemokines, per the manufacturer's instructions. Briefly, rectal swabs frozen in DNA/RNA Shield were thawed and vortexed, and duplicate 25-μL aliquots of the DNA/RNA Shield reagent were used in the assay. Flow cytometric acquisition was performed on a BD LSRII Flow Cytometer with FACSDiva software (version 8). FCS files were analyzed in LEGENDPlex Data Analysis Software for Mac (version 7.1).

Characterization of the Rectal Microbiota via 16S rRNA Gene Sequencing

Rectal microbial communities were assessed by 16S rRNA gene sequencing. Rectal swabs frozen in DNA/RNA Shield were thawed, vortexed, and centrifuged (1 mL for 3 minutes at 13 000g). DNA was extracted from the cell pellet by sequential incubations: 200 μL of lysis buffer (30mM Tris-HCl, 10mM EDTA, 200mM sucrose, pH 8.2; all from Sigma-Aldrich) for 10 minutes at 65 °C; 10 mg/mL (final concentration) of egg white lysozyme solution (Sigma-Aldrich) in nuclease free water (Cytiva) for 1 hour at 37 °C; and 1% sodium dodecyl sulfate (final concentration w/v; Sigma-Aldrich) for 10 minutes at 56 °C. Next, the DNeasy Blood and Tissue Kit (Qiagen) was used starting at step 4 per the manufacturer's instructions. Extracted DNA was processed following Earth Microbiome Project protocols [19]. First, the V4 region of the 16S rRNA gene was polymerase chain reaction amplified in triplicate by using previously described primers: 515F (5′GTGYCAGCMGCCGCGGTAA-3′)–806R (5′-GGACTACNVGGGTWTCTAAT-3′) [20]. Amplicon products were normalized, pooled, and cleaned, followed by KAPA quantification (KAPA Biosystems) and sequencing via a 2 × 150–base pair run on an Illumina MiSeq with 20% PhiX (Illumina). Demultiplexed and trimmed FASTQ files were processed with the DADA2 R pipeline (version 1.14.0); taxonomic determination was conducted with the Silva classifier (version 138); and a phylogenic tree was generated in the phangorn package (version 2.5.5). To remove rare taxa prior to analysis, a 5% prevalence taxa threshold was applied (ie, taxa detected in at least 5% of samples). The microbiome package (version 1.18.0) was used to calculate alpha diversity metrics, including Shannon diversity, observed richness, and evenness (Pielou). Beta diversity was calculated via dissimilarity distance methods (Bray-Curtis and weighted and unweighted UniFrac) and visualized by principal coordinates analysis. Microbial analysis was visualized in ggplot2 (version 3.3.5).

Statistical Analysis

Descriptive and cytokine analyses were conducted in SAS Studio 3.81. Graphs were produced with PRISM 9.0.0. Statistical testing involving microbiome data was performed in R. For all statistical analyses, α = 0.05, although corrections for the false discovery rate (FDR) were used as indicated. There were limited missing data (see Table 1 note); thus, a listwise deletion approach was utilized. Bivariate association tests (Mann-Whitney U test, χ² test, and Fisher exact test) were conducted between participants who did and did not report douching to examine differences in demographics (age, race and ethnicity, country of origin [US born vs not], sexual orientation [dichotomized into gay vs everyone else due to cell sizes], education, income) and behavioral variables (number of receptive CAS partners, PrEP use, hazardous drinking, cannabis use, stimulant use, depression).

Table 1.

Participant Characteristics and Bivariate Associations Between Sociodemographic Factors and Douching Status

	Median [IQR] or No. (%)
	Overall (n = 92)	Any Douching (n = 64)	No Douching (n = 27)	Estimate	P Value
Age, y	34.6 [20.6]	34.6 [20.1]	35.0 [23.4]	U = 1213.5	.808
Race and ethnicity				…a	.785
Black, non-Hispanic/Latino	12 (13.0)	7 (10.9)	5 (18.5)
White, non-Hispanic/Latino	42 (45.7)	30 (46.9)	11 (40.7)
Hispanic/Latino	33 (35.9)	23 (35.9)	10 (37.0)
Asian, non-Hispanic/Latino	1 (1.1)	1 (1.6)	0 (0.0)
Not listed, non-Hispanic/Latino	2 (2.2)	2 (3.1)	0 (0.0)
Born in the United States	72 (78.3)	51 (79.7)	20 (74.1)	χ2(1) = 0.54	.464
Sexual orientation				χ2(1) = 4.29b	.038
Gay	73 (79.4)	54 (84.4)	18 (66.7)
Heterosexual	1 (1.1)	1 (1.1)	0 (0.0)
Bisexual	15 (16.3)	7 (10.9)	8 (29.6)
Another not listed	1 (1.1)	1 (1.1)	0 (0.0)
Education				…a	.918
Less than high school	1 (1.1)	1 (1.6)	0 (0.0)
High school graduate	12 (13.0)	9 (14.1)	3 (11.1)
Some college/trade school	28 (30.4)	20 (31.3)	7 (25.9)
College graduate	31 (33.7)	22 (34.4)	9 (33.3)
Postgraduate	19 (20.7)	12 (18.8)	7 (25.9)
Income, $				…a	.094
<4999	9 (9.8)	7 (10.9)	2 (7.4)
5000–11 999	6 (6.5)	4 (6.3)	2 (7.4)
12 000–15 999	3 (3.3)	3 (4.7)	0 (0.0)
16 000–24 999	8 (8.7)	3 (4.7)	5 (18.5)
25 000–34 999	7 (7.6)	4 (6.3)	3 (11.1)
35 000–49 999	17 (18.5)	9 (14.1)	8 (29.6)
50 000–74 999	17 (18.5)	15 (23.4)	2 (7.4)
75 000–99 999	5 (5.4)	4 (6.3)	1 (3.7)
100 000–124 999	10 (10.9)	9 (14.1)	1 (3.7)
Current PrEP use	26 (28.3)	24 (37.5)	2 (7.4)	χ2(1) = 8.00	.005
No. of receptive CAS partners	2.0 [2.0]	2.0 [2.5]	1.0 [1.0]	U = 906.00	.003
Screened positive for hazardous drinking	42 (45.7)	33 (51.6)	8 (29.6)	χ2(1) = 3.69	.055
Use in the past 3 moc
Cannabis	48 (52.2)	37 (57.8)	10 (37.0)	χ2(1) = 2.52	.112
Stimulant	24 (26.1)	19 (29.7)	5 (18.5)	χ2(1) = 1.10	.295
Screened positive for depression	22 (23.9)	17 (26.6)	5 (18.5)	χ2(1) = 0.59	.441

Percentages may not add to 100 due to missing data: douching, n = 1; race and ethnicity, n = 2; nativity, n = 1; sexual orientation, n = 1; education, n = 1; income, n = 10; PrEP, n = 1; cannabis, n = 3; stimulant, n = 5; depression, n = 3. Bold indicates P < .05.

Abbreviations: CAS, condomless anal sex; PrEP, preexposure prophylaxis; U, Mann-Whitney U test (Wilcoxon rank sum test); χ², Chi-squared test.

^aFisher exact test used due to not meeting cell size assumption for χ² use. Because of this, there is no statistic to report, only a P value.

^bSexual orientation was collapsed into a binary: gay = 1 vs everyone else = 0.

^cSelf-report or positive toxicology screen.

Due to nonnormality, cytokine concentrations were log transformed. Initially, bivariate association tests (Mann-Whitney U tests via the exact method) were used to examine concentrations of each cytokine between participants who did and did not report douching. Multivariable log-level regression models were then fit to test the association between rectal douching and individual cytokine levels when controlling for PrEP use, number of receptive CAS partners, race and ethnicity (dummy coded into being a person of color [Black, Hispanic/Latino, other minority] or White, non-Hispanic/Latino [reference]), and age [21]. Model assumptions and potential outlier issues were examined for each model. Four models (tumor necrosis factor α [TNF-α], interleukin 1β [IL-1β], IL-10, IL-18) each had 1 extreme outlier significantly driving nonnormality of residuals. Each outlier was made equal to the second-highest value, and models met assumptions.

For microbial analyses, alpha diversity differences between individuals who did and did not report douching were determined by a Wilcoxon test. PERMANOVA (vegan version 2.6-4) was used to assess differences in microbial community composition between douching groups. Differential abundance testing was conducted, which adjusts for sequencing depth and controls for the FDR (Benjamini and Hochberg method) [22].

Linear regression models were fit to test associations between rectal douching and alpha diversity indexes. Log-level regression cytokine data were fit to test the association between cytokine levels and alpha diversity indexes, as well as bacteria found to be correlated with cytokine makers through Spearman correlation. Variables that were controlled for in the descriptive and cytokine statistical analysis were also controlled for in the microbiota models. Spearman correlations (FDR corrected) identified associations between microbial genera and cytokine levels. Analyses were performed in pscl (version 1.5.5), car (version 3.0-12), and psych (version 2.19).

RESULTS

Study Population Characteristics

Among study participants (n = 92), the median age was 34.6 years (IQR, 20.6), and 46% (n = 42) were White non-Hispanic/Latino, 36% (n = 33) Hispanic/Latino, 13% (n = 12) Black non-Hispanic/Latino, 1% (n = 1) Asian non-Hispanic/Latino, and 2% (n = 2) another race not listed (Table 1). Most participants self-identified as gay (79%, n = 73), followed by bisexual (16%, n = 15), heterosexual (1%, n = 1), and another orientation not listed (1%, n = 1). Most participants had at least a high school education (98%, n = 90), and over half (54%, n = 50) had a college or postgraduate degree. A greater proportion of participants had an income of at least $35 000 (53%, n = 49), while 35% (n = 32) reported incomes >$50 000. Study participants had a median 2 receptive CAS partners (IQR, 2) and 28% (n = 26) reported PrEP use. In terms of substance use, 46% (n = 42) screened positive for hazardous drinking, 52% (n = 48) used cannabis, while 26% (n = 24) used stimulants in the past 3 months. Approximately 70% (n = 64) of participants reported any douching in the past 3 months.

We conducted bivariate association tests to assess differences between demographic and behavioral factors as a function of whether participants reported douching in the past 3 months. Sexual orientation (χ²[1] = 4.29, P = .038), current PrEP use (χ²[1] = 8.00, P = 0.005), and number of receptive CAS partners (U = 906.00, P = .002) were significantly associated with rectal douching (Table 1). Participants who reported rectal douching were more likely to identify as gay, to be currently taking PrEP, and to have more receptive CAS partners in the past 3 months.

Associations Between Douching and Markers of Rectal Inflammation

Bivariate association tests were used to examine the concentration of each chemokine and cytokine between douching groups. For these analyses, 91 of the 92 total participants were included in the models due to 1 missing the douching variable. There was a significant association between the douching group and 10 of the chemokines and cytokines assessed (P < .001 for all; Figure 1). Median levels of the following were significantly higher in douching vs nondouching participants: TNF-α, interferon γ (IFN-γ ), IL-6, IL-8 (CXCL8), IL-10, IL-12p70, IL-17a, IL-18, IL-23, and IL-33. TNF-α, monocyte chemoattractant protein 1 (MCP-1), and IL-1β were not significantly different between douching groups.

Figure 1.

Levels of rectal inflammatory markers by douching behavior in sexual minority men. Mann-Whitney U tests were run to test bivariate associations between douching behavior and level of each cytokine and chemokine (n = 91). All significant findings indicate that cytokines and chemokines were not evenly distributed across douching groups. Examining medians and IQRs further indicated that there were higher levels of cytokines and chemokines among those engaged in douching behavior when compared with those who did not engage in douching. Data are depicted as log-transformed values of each cytokine and chemokine assessed. Participants who reported douching are shown as open red circles. Participants who did not report douching are shown as filled blue circles. Black horizontal line and whiskers indicate median and IQR.

Log-level multivariable models were then used to assess the relationship between douching and rectal chemokine/cytokine outcomes while controlling for covariates. Douching was significantly associated with 8 of 13 chemokines and cytokines, and models had robust R² (Table 2). Specifically, levels of the following were higher among douching vs nondouching participants: IFN-γ, IFN-α, IL-6, IL-8, IL-12p70, IL-17A, IL-18, and IL-23. The overall models for TNF-α, MCP-1, IL-1β, IL-10, and IL-33 were not significant. Of note, the assessed covariates were not significantly related to chemokine/cytokine outcomes, aside from current PrEP use, which had a significant positive relationship with higher levels of IL-6 (b = 0.57, P = .01), IL-8 (b = 1.20, P = .016), IL-17A (b = 0.91, P = .026), and IL-23 (b = 0.6, P = .045).

Table 2.

Log-Level Regression Models Examining Association Between Douching and Rectal Cytokine Levels in Sexual Minority Men Who Were HIV Negative (n = 88)

Model	Outcome (Log)	R 2	b	SE	95% CI	P	Expected Change, %a
1	IFN-γb	0.30	0.96	0.22	.53, 1.39	<.0001	+161
2	IFN-αb	0.58	0.58	0.14	.31, .86	<.0001	+78
3	TFN-α	0.06	0.17	0.14	−.10, .44	.214	…
4	MCP-1	0.08	0.19	0.14	−.09, .48	.187	…
5	IL-1β	0.07	0.05	0.22	−.39, .48	.830	…
6	IL-6b,c	0.41	1.01	0.20	.61, 1.42	<.0001	+175
7	IL-8b,c	0.23	1.19	0.46	.28, 2.10	.011	+229
8	IL-10	0.15	0.36	0.15	.06, .65	.018	…
9	IL-12p70b	0.37	0.84	0.16	.52, 1.17	<.0001	+132
10	IL-17ab,c	0.32	1.56	0.38	.80, 2.32	<.0001	+376
11	IL-18b	0.30	0.63	0.14	.35, .92	<.0001	+88
12	IL-23b,c	0.36	1.37	0.28	.82, 1.92	<.0001	+294
13	IL-33	0.14	0.31	0.15	.004, .62	.047	…

Predictor: douching. Outcomes vertically on the left side represent model runs (ie, each row is a regression model). All rectal cytokine biomarkers were log transformed (n = 88 due to missing data). All models controlled for current preexposure prophylaxis use, number of condomless anal sex partners, race and ethnicity, and age.

^aExpected percentage change in outcome for those who douche vs those who do not. Percentage change calculated via (exp[b] – 1) × 100.

^bOverall model was significant; all other independent variables did not have a significant relationship with any of the outcomes unless denoted otherwise.

^cPreexposure prophylaxis use had a significant positive relationship with outcome.

Associations Between Douching and Rectal Microbial Community Composition

Of all 92 study participants, 80 were included in the taxonomic analysis, with 12 excluded due to missing samples or metadata, low DNA concentrations, or low read counts. Participants who douched had significantly lower richness and Shannon diversity and trended toward lower evenness than those who did not (Figure 2A). Linear regression modeling was used to examine associations between alpha diversity metrics and douching while controlling for covariates. Douching remained significantly negatively associated with richness, but the overall P values for all models were not statistically significant (Table 3).

Figure 2.

Rectal microbial community composition by douching behavior in sexual minority men. Rectal microbial communities were profiled in sexual minority men who reported douching as compared with participants who did not report douching (n = 80). A, Alpha diversity metrics included richness, evenness (Pielou), and Shannon diversity of bacterial communities in study participants. Box and whisker bars represent 25th–75th percentile and minimum and maximum number of the calculated metric. Black dots that overlay box and whisker plots represent individual participants for each alpha diversity metric. Horizontal bar within each box represents the median. B, Principal coordinates analysis of beta diversity in study participants. Shaded ovals for each group represent data ellipses. In panels A and B, participants who reported douching are shown in red, while participants who did not report douching are shown in blue. C, Relative abundance taxonomic plot for the top 15 most abundant genera.

Table 3.

Linear Regression Models Examining Association Between Alpha Diversity Metrics and Douching Status in Sexual Minority Men Who Were HIV Negative (n = 80)

Model	Outcome	R 2	b	SE	95% CI	P Value	Overall Model P Value
1	Richnessa	0.071	−59.241	23.275	−105.661, −12.819	.0131	.1117
2	Evenness	−0.019	−0.0270	−0.0231	−.073, .019	.2440	.5916
3	Shannon diversity	−0.001	−0.2440	0.1361	−.515, .027	.0770	.4577

Predictor: douching. Outcomes vertically on the left side representing model runs (ie, each row is a regression model; n = 80 due to missing data or unsuitable sample quality). All models controlled for current preexposure prophylaxis use, number of condomless anal sex partners, race and ethnicity, and age.

^aPreexposure prophylaxis use had a significant relationship with outcome.

To compare microbial composition between douching groups, we used beta diversity assessments (Bray-Curtis, unweighted UniFrac, and weighted UniFrac). Unweighted UniFrac, a measure that accounts for differences in the phylogenetic relatedness of present and absent taxa, indicated that participants who reported douching had significantly different rectal microbial communities than participants who did not douche (P = .005, r = 0.03; Figure 2B). At the phylum level, the relative abundance of microbial communities of all participants was dominated by Bacteroidota and Firmicutes, with minor representation of Fusobacteriota, Proteobacteria, Actinobacteriota, and Campilobacterota (Figure 2C). At the genus level, the rectal microbiota of all participants was dominated by Prevotella, followed by minor abundances of Fusobacterium, Finegoldia, Snethia, Corynebacterium, Peptoniphilus, Faecalibacterium, Bacteroides, and Anaerococcus (Figure 2C). There were no significant differences in bacterial abundance at the phylum or genus levels between douching groups.

Relationship Between Microbial Community Composition and Rectal Cytokine Levels

Linear regression modeling was used to examine associations between microbial alpha diversity metrics and log-transformed rectal chemokine/cytokine levels while controlling for covariates. Significant negative associations were observed between richness and IL-6, IL-12p70, and IL-23; between Shannon diversity and IL-6 and IL-8; and between evenness and IL-8 (Table 4). Spearman correlations with FDR corrections for multiple comparisons indicated that Prevotella and Dialister were negatively correlated with IL-6 (rho = −0.22 and −0.22, P = .046 and .049, respectively); Prevotella was negatively correlated with IL-23 (rho = −0.23, P = .044); Finegoldia and Peptoniphilus were negatively correlated with TFN-α (rho = −0.25 and −0.23, P = .024 and 0.044); and Bacteroides was positively correlated with MCP-1 (rho = 0.26, P = .02), IL-6 (rho = 0.27, P = .02), IL-17A (rho = 0.27, P = .02), IL-23 (rho = 0.29, P = .01), IL-33 (rho = 0.33, P < .01), and IL-12p70 (rho = 0.22, P = .04; Figure 3). After covariate adjustment, all associations remained significant except for those involving Prevotella and Bacteroides with IL-12p70. Race and ethnicity were significant with many of the linear regressions.

Table 4.

Log-Level Regression Models Examining Association Between Alpha Diversity Metrics and Rectal Cytokine Levels in Sexual Minority Men Who Were HIV Negative (n = 80)

Model	Outcomes (Log)	R 2	b	SE	95% CI	P Value	Overall Model P Value
Predictor: richness
1	IFN-γ	0.0372	−0.0026	0.0013	.9948–.9998	.0417	.2329
2	IFN-α	0.0619	−0.0018	0.0008	.9966–.9997	.0218	.1385
3	TFN-α	−0.0852	0.0005	0.0009	.9988–1.0022	.5263	.9689
4	MCP-1a	0.0006	−0.0002	0.0007	.9983–1.0013	.8085	.4440
5	IL-1β	−0.0652	0.0013	0.0012	.9989–1.0036	.2614	.8947
6	IL-6b,c	0.1158	−0.0027	0.0012	.9948–.9996	.0279	.0362
7	IL-8c	0.0592	−0.0022	0.0025	.9929–1.0026	.3694	.1470
8	IL-10	−0.0215	0.0011	0.0010	.9991–1.0030	.2740	.6001
9	IL-12p70b,c	0.0779	−0.0024	0.0009	.9914–1.0002	.0129	.0556
10	IL-17ac	−0.0562	−0.0042	0.0022	.9972–1.0015	.0629	.0957
11	IL-18	0.1160	−0.0007	0.0011	.9927–.9990	.5491	.8455
12	IL-23b,c	0.0269	−0.0041	0.0016	.9974–1.0006	.0128	.0360
13	IL-33c	0.0269	−0.0010	0.0008	.9948–.9998	.2436	.2839
Predictor: evenness
1	IFN-γ	0.0085	−1.9578	1.3354	.0098–2.0250	.1471	.3920
2	IFN-α	−0.0117	0.0897	0.8349	.2069–5.7818	.9148	.5315
3	TFN-α	−0.0508	−1.4286	0.8681	.0424–1.3537	.1043	.8127
4	MCP-1c	0.0169	−0.8474	0.7687	.0925–1.9852	.2741	.3402
5	IL-1β	−0.0748	−0.9823	1.2222	.0327–4.2853	.4243	.9366
6	IL-6c	0.0662	−1.3425	1.3057	.0193–3.5309	.3074	.1257
7	IL-8b	0.1299	−6.2924	2.4561	.0000–.2481	.0126	.0245
8	IL-10	−0.0386	−0.3011	1.0667	.0880–6.2186	.7786	.7224
9	IL-12p70c	0.0438	−0.6208	1.0121	.0009–11.0630	.5416	.3153
10	IL-17ac	−0.0512	−2.2851	2.3509	.0422–3.6577	.3344	.2043
11	IL-18	0.0512	−0.9341	1.1186	.0043–4.3786	.4065	.8149
12	IL-23c	0.0117	−1.9775	1.7320	.1167–3.4681	.2574	.1750
13	IL-33	0.0112	−0.4519	0.8501	.0098–2.0250	.5967	.3713
Predictor: Shannon diversity
1	IFN-γ	0.0529	−0.5154	0.2191	.3858–.9244	.0214	.1685
2	IFN-α	0.0022	−0.1381	0.1392	.6599–1.1496	.3245	.4336
3	TFN-α	−0.0549	−0.2274	0.1460	.5953–1.0658	.1238	.8380
4	MCP-1	0.0313	−0.1933	0.1281	.6384–1.0640	.1357	.2612
5	IL-1β	−0.0792	−0.1229	0.2055	.5869–1.3324	.5518	.9520
6	IL-6	0.1028	−0.4272	0.2148	.4250–1.0011	.0506	.0513
7	IL-8	0.1389	−1.1131	0.4101	.1450–.7443	.0084	.0189
8	IL-10	−0.0393	0.0331	0.1763	.7271–1.4693	.8516	.7271
9	IL-12p70	0.0683	−0.2612	0.1674	.2392–1.1311	.1232	.1833
10	IL-17a	−0.0481	−0.6535	0.3895	.5756–1.2158	.0978	.1198
11	IL-18	0.0923	−0.1784	0.1975	.3095–.9823	.3445	.7979
12	IL-23	0.0325	−0.6055	0.2843	.6244–1.0966	.0367	.0672
13	IL-33	0.0317	−0.1893	0.1412	.3858–.9244	.1843	.2552

Outcomes vertically on the left side represent model runs (ie, each row is a regression model; n = 80 due to missing data or unsuitable sample quality). All models controlled for current preexposure prophylaxis use, number of condomless anal sex partners, race and ethnicity, and age. Bold indicates P < .05.

^aPreexposure prophylaxis use had a significant relationship with outcome.

^bOverall model was significant; all other independent variables did not have a significant relationship with any of the outcomes unless denoted otherwise.

^cRace and ethnicity had a significant relationship with outcome.

Figure 3.

Correlation between microbial community composition and rectal cytokine levels. A Spearman correlation matrix was used to identify associations between the top-most abundant genera and cytokine/chemokine levels in the rectal mucosa of sexual minority men who did and did not report douching (n = 80). Blank squares indicate that the association between the variables was not significant. Circles indicate significant correlations after false discovery rate adjustment. The size of the circle indicates the strength of the correlation, while the color of the circle indicates the direction of the correlation.

DISCUSSION

This study demonstrated that rectal douching associates with amplified biobehavioral HIV/STI risk among SMM engaging in receptive CAS. Rectal douching was associated with current PrEP use and more receptive CAS partners. After adjusting for behavioral factors, SMM who douched displayed elevations in rectal chemokines and cytokines relevant to immune activation, inflammation, and T-helper 17 (Th17) cell expansion. These alterations were paralleled by lower alpha diversity metrics among SMM who douched, adjusting for the same covariates. Finally, individual bacterial genera were significantly associated with rectal chemokine/cytokine levels among douching participants, suggesting a link between rectal microbial community structure and inflammation. Thus, our findings indicate that douching may contribute to rectal immune and microbial dysregulation, which could prime the mucosal microenvironment for HIV/STI acquisition.

A notable finding was that levels of IL-17A (predominantly produced by Th17 cells [23]), IL-6 (Th17-cell differentiation factor [24]), IL-23 (Th17-cell expansion and maintenance [23]), IFN-γ (immune activation [25]), and IL-8 (neutrophil chemoattraction [26]) were elevated among douching participants. This is consistent with work demonstrating that SMM engaging in receptive CAS had higher rectal Th17-cell levels, greater CD8+ T-cell IFN-γ production, and neutrophil function gene upregulation [2]. Importantly, Th17 cells are critical in mucosal epithelial barrier maintenance and are preferential targets of HIV [27–29]. Additionally, excess mucosal neutrophil recruitment is linked with increased HIV susceptibility [30–34]. Thus, our findings critically extend previous work and identify rectal douching as a potential contributor to a proinflammatory mucosal state, which could facilitate HIV/STI acquisition.

Here, SMM engaging in CAS who reported douching exhibited decreased rectal microbial richness, evenness, and Shannon diversity. This is consistent with work indicating that SMM engaging in CAS have lower rectal mucosal Shannon diversity as compared with controls [2], and it extends these findings to identify douching as a contributing mechanism to rectal microbial alterations. Notably, the predominant genera among all participants regardless of douching was Prevotella, which is consistent with work demonstrating enrichment for Prevotella in the rectal and fecal microbiota of SMM [2, 5–7]. Additional work is needed to elucidate how species- or strain-level community shifts caused by rectal douching may influence rectal homeostasis.

We identified significant associations between varied levels of multiple mucosal cytokines/chemokines and microbial alpha diversity metrics and genus abundance. Given the links between mucosal inflammation and HIV acquisition risk [2–4], our data may indicate that douching-associated microbial community perturbations elicit proinflammatory mucosal responses that increase HIV acquisition risk. Indeed, gut microbial changes, including decreased Bacteroides and elevated inflammatory cytokine levels, were shown to precede HIV acquisition [35]. A better understanding of the biological mechanisms by which douching contributes to HIV/STI acquisition risk is critically needed.

Our observed associations among douching, chemokines/cytokines, and microbial composition remained after adjusting for PrEP use. This is notable as previous work showed that after 48 to 72 weeks of daily oral PrEP consisting of tenofovir disoproxil fumarate/emtricitabine, SMM had significantly decreased Streptococcus and increased Erysipelotrichaceae family members [36]. Conversely, a separate study demonstrated that oral tenofovir disoproxil fumarate/emtricitabine was associated with an increased abundance of Streptococcus, Mitsuokella, and Fusobacterium and decreased Escherichia/Shigella in SMM [37]. Finally, PrEP was associated with an increased abundance of Catenibacterium and Prevotella 2 and decreased Finegoldia and Corynebacterium 1 [38 ]. Although these studies agreed that PrEP was associated with altered rectal microbial abundance, there were discrepancies in the differentially abundant microbial taxa, which may be explained by differences in study design, sample size and collection, bioinformatics/statistical analysis, and duration of PrEP [39]. In sum, our data suggest that douching is associated with rectal microbial community structure alterations outside of the potentially confounding influence of PrEP.

A major strength here was the use of rectal swabs collected from SMM who reported receptive CAS, which allowed for examination of microbial and immune associations directly in the rectal microenvironment. A caveat is that rectal swabs were self-collected, which could cause sample-to-sample variations [40]. Additionally, although our approach allowed for measurement of cytokine/chemokine levels in the same sample used for 16S rRNA sequencing, we acknowledge that the DNA/RNA Shield medium has not yet been explicitly validated for protein preservation. Future work will be needed to verify protein stability in DNA/RNA Shield medium over time.

A study limitation is that we were able to collect information on only the number of CAS partners and not the number of sex acts, CAS or douching frequency, or douching-specific behaviors. Given that there may be substantial variability in the intensity, frequency, and methods of douching, including behaviors such as sharing douching equipment [41], and/or impacts from the number of sex acts or CAS frequency, future studies should collect more nuanced data on sexual behavior and douching practices to fully characterize the impacts of these factors on mucosal disruptions. Similarly, we were unable to objectively test for STIs; thus, the collection of objective STI status will be important in future studies. Additionally, larger cohort studies are warranted that include groups matched for PrEP use, number of CAS partners, and controls who douche but do not engage in CAS. Morever, in vitro mechanistic experiments will be beneficial that elucidate the precise effects of douching on inflammation and microbial community structure in the absence of CAS. 16S rRNA gene sequencing allowed us to characterize bacterial taxonomy; however, assessment of microbial functionality and other community members, including viruses and fungi, was not captured. Finally, the use of DESeq2 for differential testing has some limitations, such as difficulties with zero inflation, a common issue with microbiome data due to the assumption of a log-normal distribution. Other tools are available, although they come with alternative limitations and have lower statistical power. Taken together, future longitudinal studies which include data collection on the frequency of CAS and other behaviors, specific douching behaviors, circulating and mucosal immune cell frequencies and function, and microbial functionality and additional organisms are warranted to fully elucidate how douching could potentiate mucosal immune dysregulation and thereby amplify susceptibility to HIV and other STIs in SMM.

Our findings have important implications for clinicians by highlighting that douching behavior may identify SMM with amplified biobehavioral risk for HIV/STIs. Expanded efforts are needed to support uptake of PrEP and DoxyPEP (doxycycline post-exposure prophylaxis) among SMM who douche and to develop douching products that do not disrupt the rectal mucosa or microbiome. Our results also highlight that understanding the mechanistic relationship among rectal douching, immune and microbial disruptions, and HIV/STI risk among SMM engaging in receptive CAS could assist with identifying novel targets for microbicides that can be administered prior to engaging in receptive CAS.