Single-cell analysis of follicular fluid reveals dysregulation of ovulatory immune function in Polycystic Ovary Syndrome patients undergoing ovarian stimulation†.

We selected eight age- and BMImatched control and PCOS patients from our consented and sampled patient pool UW-Madison IRB #2018-1247. The patients satisfied the inclusion and exclusion criteria for the study ( Table 1 ): PCOS patients were diagnosed based on clinical satisfaction of Rotterdam criteria ( Table S1 ), and control patients had male factor, tubal, or genetic testing cause for IVF. Eight pairs of samples—TVOR appointment PBMC and follicular cells from each patient—were included in the study design. All patients had both peripheral blood and follicular fluid collected at the time of their transvaginal oocyte retrieval, following either an antagonist or Lupron overlap IVF protocol and hCG trigger. All follicles from the same patient were pooled. Patient demographics and matching are in Table 1 and Supplementary   Table   S1 , and group statistics are in Supplementary Table S2 . Peripheral blood mononuclear cells (PBMCs) and follicular fluid mononuclear cells (FFMCs) were separated from plasma and follicular fluid respectively by density gradient centrifugation and washed with RPMI +10% fetal bovine serum. ACK lysis buffer was added to FFMCs to remove red blood cell contamination. Cells were counted, cryopreserved, and stored at −80°C until experiments were conducted. To ensure cell suspension viability after thawing, we enriched the samples for live cells by using a Dead Cell Removal Kit (Miltenyi Biotec, Bergisch Gladbach, Germany, cat# 130-090-101) with magnetic microbeads and Octomacs Magnetic separation columns (Miltenyi, cat # 130-042-109), following manufacturer’s protocol. Prior to our experiment, we tested this protocol on previously banked test samples and ensured that we produced high quality samples with >90% viable cells. Experimental samples were further assessed using the Countess II (Invitrogen, Carlsbad, CA) upon submission to the UW Biotechnology Center Gene Expression Core and demonstrated 90%–97% viability prior to library preparation and sequencing. Single cell suspension was prepared after checking for viability and count, starting with 1x10 6 cells, suspended in Mg2+, Ca2+ and nuclease free buffer and incubated with Fc blocker (BioLegend, cat # 422301) for 10 mins at 40°C. The cell suspension was then incubated with appropriate multiplexing antibodies-TotalSeqB0252 antihuman Hashtag 1 (Biolegend cat #394631) for PBMC and TotalSeqB0252 antihuman Hashtag 2 (Biolegend cat #394633) for follicular cells for 30 mins at 40°C. The cell suspensions were washed and the tagged PBMC and follicular cell samples from each patient pooled into a single sample, to produce a total of eight samples. The pooled samples were washed, resuspended in a labeling solution including TotalSeq B Feature Barcode (FB) antibodies (according to manufacturer’s instructions) and incubated for 30 mins at 4°C. FB antibodies are listed in Supplementary Table S3 . After this final staining step, the cell suspension was washed, resuspended in a Mg2+, Ca2+ and nuclease free suspension buffer and submitted to the UW Biotechnology Center for library preparation and sequencing. Cell suspensions were submitted to UW Biotechnology Center where cell viability was further validated using the CountessTM II (Invitrogen). All samples met the initial quality control criteria and proceeded to single cell RNA library preparation and sequencing on the 10X Genomics platform. Libraries were prepared and downstream processing utilizing Illumina Novaseq X was performed by the University of Wisconsin Gene Expression Center in collaboration with the UWBC DNA Sequencing Facility, Madison, Wisconsin. All sample and library preparation was prepared concurrently to avoid batch effects. Sequence data were processed with CellRanger v6.1.2. scRNA sequences were aligned to the Homo sapiens hg38 assembly. Filtered gene matrices were analyzed in R using Seurat v5.1.0 except where otherwise specified. Doublets were annotated using ScDblFinder (Bioconductor) and removed, and then all eight samples were merged into a single Seurat object. For initial quality control, cells were retained that had between 200 and 10 000 features, lower than 15% mitochondrial transcripts, total RNA count below 150 000, and were annotated as singlets by scDblFinder ( Supplementary Table S4 ). Counts were depth-normalized and log-transformed, the top 2000 variable features determined, and counts scaled. Principle component analysis was used for dimensionality reduction. A k-nearest-neighbor graph was constructed (FindNeighbors, FindClusters; Louvain clustering with resolution 0.8) and visualized using uniform manifold approximation and projection (RunUMAP). Batch effects were tested for using Harmony integration and not observed, due to sample and library preparation at the same time. Harmony clusters aligned well with standard clusters ( Supplementary Figure S1H–J ). PBMCs were bound with HTO- 1 and FFMCs with HTO- 2. Hashtag expression levels were scaled and cells with HTO-1 expression >6 and HTO-2 expression <7 were annotated as PBMCs ( Supplementary Figure S1D–G ). Cells with HTO-1 expression <6 were annotated as FFMCs, and cells with expression of both hashtags <7 could not be reliably demultiplexed and were excluded. Demultiplexing was validated by the presence of the granulosa and theca clusters almost exclusively in the FFMC group. Clusters were annotated based on gene expression and canonical immune cell surface markers (read out by DNA-tagged antibodies) and using a combination of cell type marker genes as well as highly expressed genes identified by rank-sum tests (FindMarkers and FindAllMarkers). Genes used to determine cluster identities are found in the relevant figures. Cells were first clustered based on broad cell type and NK cells, follicular cells, and T cells were extracted as separate Seurat objects, subclustered, and annotated in finer detail before being merged back with the original object. Differential gene expression analyses were carried out using model-based analysis of single-cell transcriptomics (MAST) implemented through Seurat FindMarkers and plotted with EnhancedVolcano. Heatmaps were created with ComplexHeatmap, with expression data pseudobulked via AggregateExpression. For KEGG pathway and Gene Ontology analyses, genes with log2FC > 0.25 and FDR < 0.05 (Benjamini–Hochberg) were analyzed using clusterProfiler’s enrichKEGG and enrichGO functions. Seurat objects containing control and PCOS follicular cells were converted to CellChat objects and processed using the CellChat v2 pipeline and CellChatDB.human database. Each object was processed separately and then combined for comparison. Individual signaling pathways were visualized using the chord layout of netVisual_Aggregate. For compositionally balanced analysis, selected celltypes from each patient ( Supplementary Figure S7 ) were downsampled to match lowest sample numbers and CellChat analysis was repeated, using population = TRUE.

Eight patients were included in our study: two each in the non-obese PCOS, obese PCOS, non-obese control, and obese control groups ( Table 1 ). Patients were age- and body mass index (BMI)-matched, with patient identifiers increasing with BMI (i.e., PCOS A and Control A are lowest BMI and PCOS/Control D are the highest). Each group had the same average age and BMI and a similar age and BMI range ( Supplementary Tables S1 and S2 ), controlling effects of obesity across groups. Following isolation and sequencing [ 16 ], 68 601 cells passed quality control metrics ( Figure 1a , Supplementary Figure S1A–C , Supplementary Table S4 ). Cells were first clustered based on canonical immune marker genes into relatively broad categories ( Figure 1b and c ), with T cells being the most populous cell type ( Figure 1d ). T cells, NK cells, B cells, monocyte lineage cells, and follicular cells were then further subclustered for major known immune cell subsets based on feature barcoded antibody levels and canonical gene expression for all cell types ( Figure 1e–g , Supplementary Figure S2 ) [ 17–20 ]. Study groups and inclusion/exclusion criteria. Identification of peripheral blood mononuclear cells and follicular fluid mononuclear cells by transcriptomic and proteomic markers. (a) Blood and follicular fluid were collected from four PCOS and four control patients, and mononuclear cells isolated and sequenced via CITEseq. Cells passing quality control were passed through our previously used analysis pipeline and unsupervised clustering was performed via UMAP at resolution 0.8. (b) UMAP clusters were annotated as broad cell types, based on (c) expression of canonical immune-subset defining genes. (d) Overall cell type proportions by patient (blood and follicular fluid combined). (e) Following subclustering, cell identities were annotated with greater precision based on (f) gene expression and (g) proteomic analysis. We combined all cells (Peripheral blood mononuclear cell [PBMCs] and follicular fluid mononuclear cells [FFMCs], in both PCOS and control patients) for clustering and annotation. Following annotation, we compared these conditions ( Figure 2a ) to determine if follicular fluid is a distinct immune compartment and whether PCOS impacts its cellular composition. We found that major cellular categories were similar when analyzed with single-cell composition analysis (scCODA) ( Figure 2b ), [ 21 ] indicating that the ovarian follicle is not a dramatically distinct, immune privileged environment. That said, ovarian follicle is a uniquely regulated immune environment, as demonstrated by a set of specific cellular differences compared with matched systemic circulation. In particular, we identified a tissue-resident population of cDC2s, that more highly expressed inflammatory markers MRC1 and CD163 , but differed in profile from macrophages and monocytes ( Figure 1f , Supplementary   Figure   S2A ). Additionally, monocytes levels were lower in the follicle while macrophage proportion was increased, consistent with monocyte differentiation with tissue residence ( Figure 2c , Supplementary Table S5 ). We next tested the hypothesis that immune cellular composition is dysregulated by PCOS. Our data suggest that PCOS and controls are closely matched, consistent with flow cytometry analyses of follicular composition of PCOS and control patients undergoing IVF [ 4 ]. While our experimental design was not geared towards specific isolation of granulosa and theca cells (see Methods), a small proportion overall was identified in samples, and were by definition only present in the follicular fluid. Notably, there were no statistically significant differences between PCOS and controls of proportions of immune subsets in peripheral blood between PCOS and control patients ( Supplementary Figure S3 ). Consequently, the immune environment of the follicle is distinct compared with circulation, with specific immune cell subsets recruited to or resident in this environment, and with PCOS only directly impacting the proportion of follicular cDC2s. These data suggest that at least in PCOS patients undergoing ovulation induction and IVF, cellular composition resembles that of control patients. Follicular fluid mononuclear cell composition varies slightly from peripheral blood. (a) Cell identities in follicular fluid (top) and peripheral blood (bottom), compared in control (left), and PCOS (right). (b) Cell identity proportions in all four conditions. (c) Classical monocytes and intermediate/Non-classical monocytes are of a greater proportion in PBMCs, while macrophagesand tissue-resident cDC2 cells are of a greater proportion in FFMCs. Analysis was performed with scCODA, and significance was determined at P  < 0.05. To understand how immune dysregulation in PCOS presents in the ovary specifically, it is important to determine the immune processes of normal ovulation. Ovulation is a choreographed acute inflammatory process, with the luteinizing hormone (LH) surge resulting in infiltration of leukocytes, particularly macrophages and monocytes, as well as production of chemokines and cytokines including CCL2, CCL20, TNF, PGF, and IL1B [ 22–29 ]. To determine immunocellular drivers of these events in healthy controls, we first determined the overall follicular leukocyte transcriptome compared with matched peripheral blood and observed the expected acute inflammatory signature. Notable transcriptionally active genes compared to circulation were CCL2, TNF, CCL7 , and CXCL2 ( Figure 3a ). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways related to inflammation were highly upregulated, including cytokine-cytokine receptor interaction, NF-kappa-B signaling, and the phagosome ( Figure 3b ). The peri-ovulatory follicle is an active immune environment due largely to activity of monocytes, macrophages, and conventional dendritic cells. (a) Differential expression analysis (pseudobulked by patient) of genes upregulated (positive log2FC) and downregulated (negative log2FC) in control FFMCs compared to PBMCs. (b) Top KEGG Pathways of genes upregulated (log2FC > 0.25, FDR < 0.05) in control FFMCs compared to PBMCs. (c) Mean expression of signaling genes important for physiological ovulation. (d). Differential expression analysis of genes upregulated (positive log2FC) and downregulated (negative log2FC) in control follicular monocytes, macrophages, and cDC2s compared to the same cell types in circulation. Focusing on the cellular sources of cytokines and chemokines known to be important for ovulation, we found that the vast majority were expressed by monocytes and monocyte-derived cells, and a few by non-immune follicular (granulosa, theca, etc.) cells ( Figure 3c ). We specifically examined the differential transcriptional programming of follicular monocytes, macrophages, and cDC2s (compared to circulating, Figure 3d ) and found that it captured most of the immune activity related to ovulation. Taken together, our high-resolution analysis confirms and extends the current understanding of monocyte/macrophage lineage as primary orchestrators of normal immune ovulatory activity. Next, we asked if correcting the ovulatory defect in PCOS by external FSH, LH, and trigger also restores the normal acute ovulatory immune reaction. To systematically dissect this question, we first analyzed differential gene expression in peripheral blood, between PCOS and control patients. We found that PCOS patients displayed minimal differences in peripheral gene expression compared to controls ( Supplementary Figure S4A–C ), suggesting that ovulation induction may be sufficient to alleviate systemic inflammation in PCOS. This is consistent with flow cytometry data of a larger patient population, indicating that baseline differences in inflammatory cytokine levels are not seen at time of TVOR [ 4 ]. Having established that the peripheral transcriptome of PCOS patients undergoing IVF is similar to that of controls, we then asked how gene expression differs between PCOS peripheral and follicular cells, and if it mirrors that observed in normal/control ovulation (as in Figure 3 ). We found that PCOS follicular cells did upregulate over 200 genes compared to peripheral cells ( Figure 4a ), indicating some ovarian activity. We then examined expression of a set of cytokines implicated in both normal ovulatory function and PCOS-related systemic inflammation ( Figure 4b ), and found that these inflammatory genes were most strongly expressed in the follicular cells of control patients, with similar profiles in the periphery of both groups but significantly decreased expression in PCOS follicular cells. This suggested that PCOS follicles remain dysregulated through IVF, even as systemic inflammation was potentially decreased. Immune environments of PCOS and control follicles diverge. (a) Differential expression analysis (pseudobulked by patient) of genes upregulated (positive log2FC) and downregulated (negative log2FC) in PCOS FFMCs compared to PBMCs. (b) Mean expression of cytokines typically elevated in PCOS periphery. (c) Top 20 Gene Ontology terms of genes upregulated (log2FC > 0.25, FDR < 0.05) in PCOS FFMCs compared to PBMCs, enriched genes overlapping between PCOS and control follicles, and upregulated in control FFMCs compared to PBMCs. (d) Genes in KEGG pathways found in the same subsets as part C. To investigate differences in the PCOS follicle further, we compared the sets of genes uniquely upregulated in PCOS follicles and control follicles (log2FC > 0.25, false discovery rate (FDR) < 0.05, pseudo-bulked by patient) compared to their respective periphery, and the set of genes upregulated in both groups of follicles ( Figure 4c ). GO term matching indicated that genes related to antigen presentation and adaptive immunity are overrepresented in PCOS follicles, compared to a broader cytokine response in control follicles. This was confirmed by analysis of KEGG pathways ( Figure 4d ), as different genes were implicated in cytokine-cytokine receptor interactions, and antigen presentation was overrepresented in PCOS follicles once more. Upon examining the genes that were highly expressed in both PCOS and control follicular cells compared to their peripheral counterparts, we found that control macrophages displayed higher expression of these genes overall ( Supplementary Figure S5 ). Additional notable findings were selective genes expressed by theca and fibroblasts, CD56bright NK cells, and tissue resident cDC2s. Finally, when directly comparing PCOS and control FFMCs, there were modest differences in gene expression, especially decreased gene expression in PCOS tissue resident cDC2s and macrophages ( Supplementary Figure S6 ). Taken together, this suggests that while there is overall immune activity required for ovulation, this immune activity remains tightly regulated in the control follicles, whereas in PCOS follicles there is a more generalized and dysregulated inflammation consistent with the chronic systemic inflammation observed in PCOS. As our results suggested that several key immune signaling pathways were absent or downregulated in the PCOS follicles, we next asked if cell–cell information flow was disrupted in PCOS. We employed CellChat to evaluate how cell–cell receptor to ligand interactions compared between the PCOS and control datasets [ 30 ]. We found that the TNF pathway had the highest information flow in control compared to PCOS, consistent with its identification as an upregulated pathway in controls as well as its role as a key pathway in ovulation ( Figure 5a ) [ 26–28 ]. VEGF, IL1, and CCL signaling were similarly upregulated in controls. Most importantly, IL10 signaling was also upregulated in controls, indicating that anti-inflammatory balance is also necessary for normal ovulation. When these cell–cell communication pathways were examined in detail ( Figure 5b ), TNF and VEGF signaling were entirely absent in the PCOS dataset, and in CCL, IL1, and IL10 pathways, macrophage and monocyte signaling in particular were disrupted in PCOS compared to controls. TNF, IL1, IL10, and CCL signaling pathways are dysregulated in PCOS follicles. (a) Relative information flow of all signaling pathways detected in PCOS and control datasets by CellChat. Stars indicate pathways discussed further. (b) Signaling pathways with sending and receiving cell types specified. TNF and VEGF pathways were not detected in PCOS samples. (c) Differential expression analysis of genes upregulated (positive log2FC) and downregulated (negative log2FC) in PCOS monocytes and macrophages compared to controls. (d) Mean expression of ovulation-related immune signaling genes by select myeloid subsets. To further confirm the role of macrophage and monocyte signaling in PCOS-related immune dysregulation, we carried out differential gene expression of PCOS monocytes and macrophages compared to control monocytes and macrophages. We found that NFKB, IL1B , and CCL4 were significantly downregulated, consistent with the CellChat and previous KEGG pathway results ( Figure 5c ). Finally, signaling genes known to be important in ovulation were specifically downregulated in PCOS classical monocytes, macrophages, and tissue-resident cDC2s ( Figure 5d ). Analysis of other cell types confirmed that differences in gene expression between PCOS and control follicles were primarily confined to monocytes and macrophages ( Supplementary Figure S6 ), and repetition of the CellChat analysis with a compositionally balanced, down-sampled group of myeloid cells confirmed that absences and differences in key pathways were not due to insufficient power, but instead are due to differences in gene expression mostly in control patients ( Supplementary Figure S7A–D ).

While chronic inflammation in PCOS in the systemic compartment is well-understood, the tissue-specific ovarian inflammation and its relationship to normal ovulation is less clear [ 3 ]. Here, we hypothesized that PCOS follicles have unique immunological dysregulation, differing from both control follicles and from the systemic inflammation observed in PCOS. Due to the nature of our study population, collecting follicular fluid from IVF patients, we were also inherently testing if this systemic inflammation was present during stimulated ovulation . We annotated 30 cell identities within follicular fluid, contributing to our understanding of the peri-ovulatory immune environment. While our study was not powered to identify small differences in abundance of different immune cell populations, we did identify that macrophages were more abundant in follicular fluid, and classical monocytes were less abundant (as expected, since monocytes primarily reside in circulation). Notably, we identified a tissue-resident cDC2 population that was and transcriptionally distinct from their circulating counterparts, and implicated in many of the differences observed in PCOS and control follicles. In control follicles, multiple inflammatory signaling pathways were robustly upregulated, consistent with the requirements of ovulation. Signaling pathways such as TNF, IL1, and NF-kB were strongly activated in the control peri-ovulatory leukocytes ( Figures 3 and 5 ), consistent with the surge of cytokines and leukocyte recruitment triggered by the LH surge [ 14 ]. This pro-inflammatory cascade essential for follicular rupture and luteinization is acute and must be tightly regulated, and we also observed upregulation of the counter-regulatory IL10 signaling pathway in control follicles. This suggests that the control stimulated ovulation is tightly balanced at the immune level. Monocytes, macrophages, and cDC2s emerged as the central orchestrators of this regulation, consistent with prior evidence of their involvement in ovulation and subsequent tissue remodeling [ 14 , 31–35 ]. This is further supported by increased expression of CCL2 (MCP-1 ) in ovarian follicular cells, recruiting monocytes to the area. In contrast, PCOS follicles demonstrated a different profile more consistent with chronic low-grade inflammation, despite stimulated ovulation. We found enrichment of pathways related to extracellular matrix organization and fibrosis (i.e., increased collagen and laminin signaling; Figure 5a ). This fibrotic microenvironment may impede normal leukocyte migration, as typically collagen is broken down by matrix metalloproteinases during the ovulatory process [ 14 , 36 ]. and indeed ovarian collagen deposition is a known feature of PCOS and has been demonstrated histologically [ 37–40 ]. Our results are consistent with prior studies that have found upregulation of extracellular matrix-receptor interaction pathways in PCOS, consistent with a fibrotic follicular environment [ 37 , 41 ]. They are also consistent with our own prior reports that plasma both prior to ovulation induction and at time of oocyte retrieval contained more inflammatory cytokines than follicular fluid [ 4 ]. While our cell isolation workflow was not specifically designed to investigate theca and granulosa cells, other investigators have found dysregulated theca-immune cell signaling and distinct theca/granulosa transcriptomes in untreated PCOS, supporting the idea that normal immune activity is dysregulated at the ovarian level specifically [ 42 ]. Overall, our results and those of others suggest a model wherein normal ovulation is a tightly controlled, acute inflammatory process driven by monocyte-derived cells and balanced by anti-inflammatory signaling. Stimulated follicles in PCOS patients, however, do not successfully activate the necessary inflammatory pathways, especially regarding monocyte and macrophage recruitment, and instead experience increased fibrosis and chronic inflammation. While our study was limited by the use of exogenously FSH and LH-stimulated IVF patients which may not capture all aspects of typical ovulation, as well as a relatively small sample size, our data suggest that the fundamental inflammatory events of ovulation remain intact in control patients under IVF conditions [ 31–35 , 43 ]. Additionally, IVF protocols were similar for nearly all patients in our study to reduce variation due to inherent personalization of IVF, and gene expression results were pseudobulked to consider patient-level effects. It is important to note that in our study, mature oocytes were retrieved from PCOS patients, so anovulation or failed oocyte-cumulus complex development is not responsible for the differences between PCOS and control follicles. While our study was not powered to investigate differences in pregnancy/IVF outcomes, an equal number of PCOS and control patients in our study successfully conceived and had live births. Therefore, our findings emphasize that the PCOS ovary exhibits a fundamentally different immune state even when ovulation is stimulated and systemic inflammation is consequently decreased. Future studies may investigate the impact of specific androgen levels and/or resolution of the different PCOS phenotypes with regard to ovarian inflammation, as we matched patients with age and BMI but not androgen levels. Additionally, our study focused on the mononuclear/immune fraction of the follicle by design, but further investigation into immune relationships with ovarian stromal cells is warranted. Finally, it may be worth investigating if tissue-targeted immune modulation could become a potential approach for improving ovulatory function or oocyte quality in PCOS. In conclusion, our study presents a detailed immune cell atlas of the control and PCOS follicular environments, and suggest that chronic systemic inflammation in PCOS is accompanied by disruption of normal inflammatory pathways in the ovary. This immune dysregulation likely contributes to the ovulatory dysfunction in PCOS, and further investigation could inform targeted strategies to treat PCOS-associated infertility.

All statistical analyses were performed in R, using Seurat for single-cell analysis or base R packages for other statistics. Differential gene expression was carried out by pseudo-bulking gene expression by patient, either for an entire tissue or for each celltype, and DESeq2 was used for analysis with Benjamini–Hochberg multiple testing correction. For KEGG and GO analyses, the top 20 pathways/gene ontology terms by Bonferroni-corrected P -value are shown. For cell proportion comparisons, paired (for FFMCs and PBMCs between same patients) and unpaired (all other comparisons) scCODA was used with results considered significant at P  < 0.05 [ 21 ]. Further details are provided in figure legends and Supplementary Table S5 .

Polycystic ovary syndrome (PCOS) is the most common endocrine disorder, affecting millions of reproductive-aged individuals with ovaries and making it one of the most common causes of female factor infertility [ 1 ]. PCOS is most commonly defined by the Rotterdam Diagnostic Criteria as the presence of two out of three of the following: (1) Clinical or biochemical hyperandrogenism, (2) Oligomenorrhea, and (3) polycystic ovaries on ultrasound [ 1 ]. Other diagnostic criteria are also coming into use, recognizing the importance of anti-Müllerian hormone (AMH) and hyperandrogenism [ 2 ]. Visceral obesity, hyperinsulinemia, and insulin resistance are also common in PCOS, further influencing reproductive and metabolic health. PCOS is associated with low-grade systemic inflammation, including elevation of inflammation-related cytokines (TNF, IL6, MCP-1 (CCL2), MIF, IL18, and IL1B) [ 3 , 4 ] and higher M1/M2 macrophage and Th17/Th2 ratios [ 3 , 5–10 ]. While the inflammation associated with PCOS is well-characterized in circulation, the specific inflammation experienced by the ovarian follicle and cellular sources of inflammatory molecules are not defined. Follicular fluid, extracted at the time of transvaginal oocyte retrieval (TVOR) during in vitro fertilization (IVF) has shown higher levels of pro-inflammatory markers (TNF, CRP, IL6, IL2 and INFG) in individuals with PCOS compared to controls in some studies [ 11–13 ], but with some indications that follicular inflammation is lower than that of circulation at time of ovulation [ 4 ]. Thus, the relationship between systemic inflammation, ovarian inflammation, and ovulation is not clear. This is significant as normal ovulation is a controlled inflammatory process, with release of pro-inflammatory cytokines and chemokines and recruitment of cells including macrophages, monocytes, and T cells to the follicle immediately before ovulation [ 14 , 15 ]. Given potential overlap between this acute inflammatory environment in the peri-ovulatory ovary and the chronic inflammation observed in PCOS, we sought to characterize the immune environment in the ovarian follicle at a single-cell level in both control and PCOS patients. We hypothesized that PCOS follicles would display unique immune dysregulation, differing from both control follicles and from the systemic inflammation observed in PCOS. To test this hypothesis, we collected blood and follicular fluid at the time of transvaginal oocyte retrieval in patients undergoing in vitro fertilization at a university-associated fertility clinic. We used CITE-seq (Cellular Indexing of Transcriptomes and Epitopes) to sequence and characterize mononuclear cells in both the ovarian follicle and peripheral blood [ 16 ]. This led to identification of important inflammatory pathways in normal ovulation and disruption of these pathways in the PCOS follicle even as systemic inflammation may have been mitigated. Our data suggest that ovarian inflammation and immune dysregulation are not due to anovulation alone and that the PCOS follicle has a distinct immune profile.