Vaginal tissues were obtained from female adult patients who underwent pelvic or cosmetic surgeries. Abdominal skin tissues from female donors were also collected under protocols approved by the Institutional Review Board (IRB). Tissues were not procured from any individual who was pregnant or infected with HIV, hepatitis C virus, or tuberculosis. Tissues with any severe acute inflammation (with or without microbial infections) were also excluded. Severe acute inflammation at the time of surgery was determined by the surgeon’s evaluation of tissue redness, swelling, and pain.

Blood from healthy female donors was collected under the IRB protocol approved with written informed consent. Blood samples were not from the same patients who provided tissues, as vaginal tissues and abdominal skin tissues were not from the same patients. Blood samples were not procured from any individuals who were pregnant or infected with HIV, hepatitis C virus or tuberculosis. None of the tissue or blood donors was under any therapy with hormones for at least 6 months before the time of tissue collection. The demographic information of study participants is summarized in Supplementary Table 1 . The numbers of tissues and blood samples used in each experiment and donor information are indicated in methods and figure legends. All experiments were performed in accordance with relevant guidelines and regulations.

Both VM and skin tissue samples were processed as previously described (10, 27). In brief, epithelial sheets and LP pieces were separated after overnight treatment of tissues with Dispase type 2 (Roche Applied Science), and further incubated for 2 days before isolating single cells by Ficoll-sodium diatrizoate gradient (Lymphocyte Separation Medium, MP Biomedicals). Cells were then stained with antibodies and sorted with FACS Aria (BD Biosciences), as previously described (6, 10). Peripheral blood mononuclear cells (PBMCs) from age-matched healthy female volunteers were isolated by density gradient centrifugation using Ficoll-Paque PLUS (GE Healthcare). Blood myeloid DC (mDC, Lin1^-HLA-DR⁺CD11c⁺CD123^-) were enriched (PanDC kit, Stemcell) and sorted (6, 10).

To isolate VM APC subsets, single cell suspension of VM tissues was stained with 7-AAD (Biolegend), anti-CD14 (IgG1, 61D3, eBioscience), anti-CD1c (IgG1, L161, BioLegend), anti-Langerin (IgG2b, 15B10, in-house), anti-HLA-DR (IgG2a, L243, BioLegend). HLA-DR⁺ cells were gated and Langerin⁺ LCs, CD1c⁺CD14^- DCs, CD1c⁺CD14⁺ DCs, and CD1c^-CD14⁺ MØ were sorted by FACS Aria II (BD Biosciences) ( Supplementary Figure 1 ). HLA-DR^- cells were also sorted. To isolate skin DC subsets, a single cell suspension of skin biopsies was stained with the same antibodies used for the isolation of VM APC subsets. sLCs from the epidermis as well as CD1c⁺CD14^- DCs and CD1c⁺CD14⁺ DCs from the dermis were sorted by FACS Aria II (BD Biosciences). To isolate blood mDCs, PBMCs were stained with 7-AAD, Lin1 (BD Biosciences), anti-CD123 (IgG1, 9F5, BD), anti-CD11c (IgG1, B-ly6, BD) and anti-HLA-DR (IgG2a, L243, BioLegend). mDCs (Lin1^-HLA-DR⁺CD11c⁺CD123^-) were then sorted by FACS Aria (BD Biosciences).

For transcriptomics analysis, 8 different APC subsets were isolated from VM tissues, skin, and blood. Four APC subsets were isolated from the VM tissues of 41 patients. The numbers of individual APC subsets isolated were variable among tissues from different donors. Due to the small number of cells isolated from VM tissues, most samples tested in this study were prepared by combining the same cell types isolated from at least two tissue donors. A total of 48 samples were prepared with vaginal APC subsets. These include 9 samples of vLC isolated from 26 donors, 12 samples of vCD14^- DC isolated from 35 donors, 13 samples of vCD14⁺ DC isolated from 41 donors, and 14 vMØ isolated from 37 donors. It was of note that the majority of cDCs in lamina propria of human VM tissues were CD1c⁺ which could represent cDC1. Three skin DC subsets were isolated from tissues of 10 patients. We had 9 samples of sCD14^- DC, 8 samples of sCD14⁺ DC, and 10 sLC samples (total of 27 samples). Blood mDC from 6 female donors were isolated (6 samples). HLA-DR^- (after depletion of CD3⁺ T, CD19⁺/CD20⁺ B, and CD56⁺ NK cells) and HLA-DR⁺ non-APCs ( Supplementary Figure 1 ) in VM tissues were also isolated from indicated numbers of tissue donors, as previously described (27).

Methods for mRNA preparation, hybridization, data processing, and batch correction for the transcriptomics analysis were previously described (10). In brief, total RNA was isolated from cell lysates using the ArrayPure-Nano-scale RNA Purification Kit (Epicentre, Madison, WI). Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA) was used to measure RNA Integrity Numbers (RIN) for each sample. All samples with RIN values greater than seven were retained for further processing. RNA concentration was measured using a Nanodrop 1000 (Nanodrop Technologies, Wilmington, DE). 250 ng of RNA from all samples passing quality control were amplified and labeled using the TargetAmp™ 2-Round aRNA Amplification Kit 2.0 (Epicentre). 750 ng of amplified labeled RNA were hybridized overnight to Illumina HT12 V4 beadchips (Illumina, San Diego, CA). Chips were scanned on an Illumina BeadStation 500 following the manufacturer’s protocols. To identify technical sources of variability, we conducted principal variance component analysis (PVCA) using the PALO list of 27,935 expressed genes, which identified a significant source of batch effect between samples run in batches 1, 2, 3, and 4. Of note, batch 3 samples only contained vaginal DC or macrophages, and batch 4 only contained mDC and vaginal HLA-DR^- cells. To correct for this batch effect, we conducted ComBat correction using R/Bioconductor. One-way Welch analysis of variance (ANOVA) was conducted using an adjusted P-value cutoff of 0.05 with Benjamini-Hochberg multiple testing correction.

The weighted average proportion variance was calculated with the R/Bioconductor package ‘pvca’ (version 1.0.0) [http://www.bioconductor.org/packages/release/bioc/html/pvca.html]. The threshold used for the minimum amount of variance explained by the selected principal components was 0.5. The entire transcriptomics dataset described in this manuscript is deposited in the NCBI Gene Expression Omnibus [http://www.ncbi.nlm.nih.gov/geo] (GEO series accession number GSE54480). Statistical significance was determined using the analysis of variance test, as indicated in individual figure legends, using the Prism 8 software (GraphPad Software). Significance was set at p<0.05.

Total RNAs from indicated cell types were isolated by using the RNeasy Mini Kit (Qiagen). Real-time polymerase chain reaction (PCR) was performed. SYBR Green qPCR Master Mix (Thermo Fisher) was used for amplifying ESR1 and PGR. Primer sequences were 5′-GAATCTGCCAAGGAGACTCGC-3′ and 5′-ACTGGTTGGTGGCTGGACAC-3′ for ESR1 (protein ERα) and 5′-TGCTCAAGGAGGGCCTGCCGCAGGT-3′ and 5′-CTACTGAAAGAAGTTGCCTCTCGCC-3′ for PGR (protein PR). Target gene transcripts were normalized to the level of β2-microglobulin (B2M) mRNA. Amplification was carried out for 35 cycles of denaturation at 94°C, annealing at 55°C, and extension at 72°C.

Cryosections were fixed in cold acetone, dried, and blocked for nonspecific fluorescence with an Fc receptor blocker and background buster (Innovex Biosciences). Sections were stained with the indicated Abs. Digital images were taken using an Olympus BX51 microscope utilizing the Planapo20/0.7 or Planapo40/0.95 objective, Roper Coolsnap HQ camera (Olympus), and Metamorph software (Molecular Devices). Confocal images were taken with the Leica SP1 (Leica) utilizing the Planapo63/1.32 objective. Images were acquired using the same exposures for antibody and isotype staining and identical scaling was applied.

Mouse anti-human ERα (IgG1, 6F11, Abcam) and rabbit anti-human PR (IgG, SP2, Abcam) along with goat anti-mouse IgG, goat anti-rabbit IgG as secondary antibodies were used for staining ERα and PR, respectively. Anti-human PR (SP2, Abcam) recognizes both isoforms of PR (PR-A and PR-B). Mouse anti-CD10 (IgG1, HI10a, BD Biosciences), and mouse anti-vimentin (IgG1, RV202, Abcam), were used. Appropriate isotype antibodies were used for all tissue/cell stainings performed in this study.

Statistical significance was determined using the analysis of variance test, Pearson correlation analysis, and t-test as indicated in individual figure legends, using the Prism 8 software (GraphPad Software). Significance was set at p < 0.05.

The 81 samples obtained from VM tissues (48 samples), skin (27 samples), and blood (6 samples) were normalized to the median of all samples. With one-way ANOVA, we identified 16 differentially expressed transcripts (DETs) of NHRs among 81 samples for the 8 different APC subsets isolated from VM (4 subsets), skin (3 subsets), and peripheral blood (1 subset, mDC). Those 16 NHR transcripts were hierarchically clustered and presented as a heatmap in Figure 1A . We found that samples were primarily clustered by tissues of origin and then by different cell populations. Tissue-specific clustering was further substantiated by PVCA analysis ( Figure 1B ). vLC, vCD14^- DC, and vCD14⁺ DC clustered together in the VM, while sLC, sCD14⁺ DC and sCD14^- DC clustered in the skin. vMØ were also closer to vCD14⁺ DC than skin APCs and blood mDC. Blood mDC were closer to skin APC subsets than vaginal APC subsets.

To compare the expression levels of individual NHRs in different subsets of APCs tested, raw expression data are presented in Figure 1C and Supplementary Figure 2 . Of note, ESR1 (NR3A1) expression was detected mainly in the VM APC subsets. However, ESR1 expression level was highly variable among individual donors, as it could be influenced by the hormonal status of patient donors (28, 33–35) which is also related to age, menstrual cycle, and menopausal status. ESR2 or PGR expression was not detected. CD14⁺ vDC and vMØ, expressed increased levels of NR1H3 (LXRA, oxysterol receptor) and RXRA (NR2B1) ( Figure 1C ). vMØ also expressed increased levels of ESRRA, VDR, and PPARG when compared to most of the other APC subsets tested ( Supplementary Figure 2 ).

We also found that skin APCs and blood mDC expressed increased levels of both NR4A3 (NOR1) and NR4A2 (NURR1). RARA expression was also higher in blood mDCs than in other APC subsets tested ( Figure 1C ). RARA/RXRA/PPAR and VDR are known to contribute to vitamin A- and D-mediated immune modulation, respectively (20).

Taken these data together, we concluded that NHR expression profiles in the 8 different human APC subsets are mainly dependent on the tissues of origin. This highlights the potential roles of VM tissue-specific microenvironments in the expression of specific NHRs that might further influence the immune response in the VM and lower female genital tract.

We next assessed NHRs that are differentially expressed by the 4 VM APC subsets and HLA-DR^- cells isolated from VM tissues of 7 donors. A total of 49 samples obtained from vaginal tissues was normalized to the median of all samples. One-way ANOVA identified 28 DETs of NHRs among the five populations (HLA-DR^- cells and 4 VM APC subsets) ( Figure 2A ).

Although ESR1 (NR3A1) expression was detected mainly in the VM APC subsets, when compared to APCs isolated from skin and blood ( Figure 1 ), we found that HLA-DR^- cells, not the vaginal APCs, are the major cell types expressing ESR1 ( Figure 2A, B ). In addition to ESR1, HLA-DR^- cells were also the major cell types expressing RORA (NR1F1), PPARD (NR1C2), NOR1 (NR4A3), RARB (NR1B2), NR2C2 (TR4), NR1D1 (NR1D1), PGRMC2, NR1D2, NR3C1 (GCR), and PGRMC1. Of the VM APC subsets, vMØ expressed higher levels of NR1H3 (LXRA), PPARG (NR1C3), RXRA (NR2B1), and VDR (NR1I1) than HLA-DR^- cells. Thyroid hormone receptors, NR1A1 (THRA) or NR1A2 (THRB), or androgen receptor (AR, NR3C4) expression was not observed. We could not detect two other major receptors for female sex hormones, PGR (NR3C3) and ESR2 (NR3B2).

Taken these data together, we concluded that HLA-DR^- cells are the major cell types expressing a variety of NHRs in the human VM tissues. Except for the increased expression of VDR by vMØ, most of the steroid hormone receptors detected were also mainly expressed by HLA-DR^- cells, suggesting the potential roles of HLA-DR^- cells in the types and magnitude of immune response elicited in the VM and lower female genital tracts in response to such hormones. In addition, future studies will also need to investigate the potential roles of NR1H3, PPARG, RXRA, and VDR in the immune response elicited by vMØ.

Female sex hormones are known to play important roles in the immunity in the VM and lower female genital tract. We therefore further examined the expression levels of ESR1 and PGR in the VM APC subsets and HLA-DR^- cells by qPCR ( Figure 3A ). In line with the data in Figure 2 , ESR1 expression was higher in HLA-DR^- cells than in any of the VM APC subsets. PGR expression was detectable by qPCR with expression by HLA-DR^- cells, though its expression level was lower than that of ESR1. Similar to ESR1, PGR expression was also higher in HLA-DR^- cells than in the VM APC subsets tested. Of note, approximately 90% of the total human VM tissues are HLA-DR^- cells (10) that contain stromal cells and smooth muscle cells in the LP of the VM tissues.

It is also important to note that the majority (more than 85-95%) of HLA-DR⁺ cells in the human VM tissues are not belonging to the VM APCs (expressing Langerin, CD1c, or CD14) (10). In addition, approximately 70% of HLA-DR⁺ non-APC populations expressed CD34 ( Supplementary Figure 1 ). We found that ESR1 expression was higher in HLA-DR⁺CD34⁺ cells than HLA-DR^- cells, whereas PGR expression was higher in HLA-DR^- cells than HLA-DR⁺CD34⁺ cells. No significant expression of ESR1 or PGR was detected in HLA-DR⁺CD34^- cells. These data indicate that the differential outcomes of the actions of estrogen and progesterone (14–16) could be, at least in part, due to their ability to target different cell types in the same tissue microenvironment. Both endothelial and epithelial cells in intestinal mucosal tissues were also previously reported to constitutively express HLA-DR (36–38). In support of the qPCR data in Figure 3B , ERα expression was detected in both epithelium and LP of the VM tissues, whereas PR expression was detected only in LP of the VM tissues ( Figure 3C ; Supplementary Figure 3 ).

Similar to our previous observation (27), a significant fraction of cells in the VM tissues expressing ERα or PR co-expressed vimentin (upper panels in Figures 4A, B ) that is known to be expressed in a variety of cell types, including smooth muscle cells and stromal cells (39, 40). However, neither ERα nor PR was detected in human skin (lower panels in Figure 4A ). This is in line with the data published in a previous study (41), showing that protein ER or PR expression was not detected in normal human skin. This also supports an idea for a tissue-specific action of female sex hormones via ERα and PR in a steady state. We further found that some of the ERα⁺ cells in the LP of VM tissues express CD10 ( Figure 4C ). None of the ERα⁺ cells in the epithelium expressed CD10. No significant colocalization of CD10 and PR were observed (not shown). Summarized data of the percentage of CD10⁺ERα⁺ in total ERα⁺ cells are presented in Figure 4D . Taken all these data together, we concluded that ERα and PR expression is specific for not only cell types, but also tissues of origin.