Single-cell analysis of ovarian myeloid cells identifies age-associated changes in macrophages and signaling dynamics

Abstract

The aging of mammalian ovary is accompanied by an increase in tissue fibrosis and heightened inflammation. Myeloid cells, including macrophages, monocytes, dendritic cells, and neutrophils, play pivotal roles in shaping the ovarian tissue microenvironment and regulating inflammatory responses. However, a comprehensive understanding of the roles of these cells in the ovarian aging process is lacking. To bridge this knowledge gap, we utilized single-cell RNA sequencing and flow cytometry analysis to functionally characterize CD45⁺ CD11b⁺ myeloid cell populations in young (3 months old) and aged (14–17 months old) murine ovaries. Our dataset unveiled the presence of five ovarian macrophage subsets, including a Cx3cr1^lowCd81^hi subset unique to the aged murine ovary. Most notably, our data revealed significant alterations in ANNEXIN and TGFβ signaling within aged ovarian myeloid cells, which suggest a novel mechanism contributing to the onset and progression of aging-associated inflammation and fibrosis in the ovarian tissue. In summary, our study revealed age-related changes in ovarian myeloid cells using single-cell RNA sequencing and flow cytometry, and identified distinct macrophage subsets and signaling alterations that may contribute to the inflammaging process of the ovary.

Graphical Abstract

Graphical Abstract.

Created by BioRender

Introduction

It has been well established that the reproductive potential of women diminishes decades earlier than the deterioration of many of their other physiological systems. The average life expectancy for women is 79.1 years [1], while menopause takes place at an average age of 51 [2]. Furthermore, the quality of oocytes produced by women significantly decreases in their late 30s due to an increase in aneuploidy [3, 4]. This age-associated decline in female fertility has been attributed to a diminishing ovarian reserve and the alterations in the ovarian tissue microenvironment.

In both humans and mice, ovarian aging is associated with excessive deposition, as well as altered organization of collagen and hyaluronan in the ovarian stroma, signifying tissue fibrosis [5–8]. Furthermore, levels of pro- and anti-inflammatory cytokines/chemokines, including interferon-γ (IFNγ), tumor necrosis factor α (TNFα), interleukin (IL)-6, IL-10, C-C Motif Chemokine Ligand 2 (Ccl2), Ccl5, IL-4, and IL-13, have been shown to be significantly increased in aged mouse ovaries compared to that in the young ones [7, 9], indicating an increase in chronic inflammation in the tissue. These changes result, in part, in a stiff tissue texture and a dysregulated signaling environment that have been linked to impairment of follicle growth [5, 7], ovulation [10], and oocyte quality [11], as well as potentially pathological events, including the development of ovarian cancers [12, 13]. Understanding the etiology and progression of these aging-associated changes in the ovarian tissue environment is crucial for the advancement of health and reproductive longevity in women.

Immune cells are pivotal in the surveillance and regulation of the ovarian tissue microenvironment [14–16]. Myeloid cells are a group of immune cells that derive from a common myeloid progenitor in the bone marrow [17] and consist of innate immune cells, including granulocytes, monocytes, macrophages, and dendritic cells [17]. They play essential roles in tissue maintenance and remodeling, as well as the orchestration of inflammatory response [17, 18]. In the mammalian ovary, macrophages are the most abundantly present myeloid cells and are functionally indispensable to many aspects of ovarian functions, including tissue organization, vasculature development, follicle growth, and corpus luteum formation [14, 16, 19, 20]. For instance, selective depletion of CD11c + macrophages using CD11c-diphtheria toxin receptor mice impairs folliculogenesis and vascular integrity [21]. Depletion of CD11b⁺ myeloid cells causes profound ovarian hemorrhage and follicular atresia [14]. Ovulation in the mouse ovary is significantly inhibited when total ovarian macrophages are depleted through intrabursal injection of clodronate liposome [22]. Furthermore, ovarian macrophages originate from multiple progenitors and exhibit heterogeneous phenotypes [23–25]. Both embryonic- and bone marrow–derived ovarian macrophages are present in the adult mouse ovary with different levels of major histocompatibility complex class II (MHCII) expression [23]. In a separate study, two ovarian macrophage subsets distinguished by F4/80 and CD11b expression were identified and exhibited different lifespans [24]. Additionally, the localization pattern of ovarian macrophages within the tissue is highly dynamic throughout the estrous cycle, correlating with their distinct phenotypes and various functions [26, 27]. Therefore, a comprehensive understanding of the heterogeneity of ovarian macrophages using an unbiased approach is essential for interrogating their roles in homeostasis and pathological conditions.

Ovarian myeloid cell populations undergo dynamic changes over the course of aging. Studies in mice and humans have reported aging-associated changes in the number and activation status of macrophages that accompany the onset of inflammation [9, 25, 28–30], though discrepancies exist concerning the specific directions of these changes. While increased macrophage presence was observed to accompany ovarian aging in some mouse studies [9, 28], others have reported a decreased macrophage presence in aged mouse or human ovaries compared to that in the young [25, 29, 30]. Additionally, an aged murine ovarian stroma also harbors a higher number of eosinophils and Ly6C⁺ monocytes compared to that in the young [8, 25]. Considering that myeloid cells are major producers of pro- and anti-inflammatory cytokines, these findings underscore the potential involvement of myeloid cells in the initiation and progression of the aging-associated inflammation.

In the past decade, advancements in single-cell RNA sequencing (scRNAseq) techniques have redefined transcriptomic research. This powerful technique has greatly benefited the study of ovarian aging and enabled the profiling of transcriptomic changes in aging ovarian cells, including the oocytes, granulosa cells, theca cells, and stromal cells [31–33]. However, due to the relative scarcity of myeloid cells within the ovarian tissue, a comprehensive high-resolution analysis of their population heterogeneity and transcriptomic changes in the aging process remained largely unexplored.

In this study, we used fluorescence-assisted cell sorting (FACS) in combination with scRNAseq to obtain a focused view of myeloid cells in young (3 months old) and aged (14–17 months old) murine ovarian tissue (Figure 1A). These cells mainly consist of macrophages, monocytes, dendritic cells (DCs), and neutrophils. Our dataset identified five distinct macrophage subsets in the murine ovarian tissue, including a Cx3cr1^low Cd81^hi subset that is almost exclusively present in the aged ovary. Our data further revealed elevated ANNEXIN and TGFβ signaling among the ovarian myeloid cells at advanced reproductive age and highlighted the potential role of neutrophils in the onset of aging-associated inflammation. Our findings provide a detailed atlas of macrophages and other myeloid cell populations in ovarian aging, laying the foundation for understanding their roles in diminished ovarian functions.

Figure 1.

Single-cell transcriptomic analysis captured the landscape of myeloid cells in young and aged murine ovaries. (A) Schematic drawing illustrating the scRNAseq experimental setup. (B) The clustering of CD45⁺CD11b⁺ ovarian myeloid cells with UMAP as the dimension reduction method. The plot includes cells from all ages and HTO groups. (C) The relative expression of the representative markers of each cluster shown in the combined clustering plot. The high and low expressions are represented by blue and gray colors, respectively. (D) Dotplot showing the expression of markers used for the identification of each cluster. Created by BioRender.

Methods

Animals

Virgin female C57BL/6 J mice that were 2–3 months (when mice reach sexual maturity) or 14–17 months of age (when mice became mostly acyclic), as well as Cx3cr1^CreER and Rosa26^{floxed-tdTomato} mice, were purchased from The Jackson Laboratory.

Prior to each experiment, the estrus stage of all mice was examined using vaginal cytology as described [34]. Mice at the diestrus stage were used for all experiments. Euthanasia of the animals was performed by CO₂ inhalation with a pressure regulator and a CO₂ rate at ~50% of the chamber per minute (two to three animals in the chamber at a time). All procedures on live animals were conducted in accordance with the protocol reviewed and approved by the American Association for Laboratory Animal Science.

Sample processing and dissociation

Ovaries were harvested from mice aged 2–3 months old and 14–17 months. For the 2–3 months old age group, ovaries from six animals were assigned to three replicates (ovaries from two animals per replicate). For the 14–17 months old age group, ovaries from five animals were assigned to three replicates (two replicates with ovaries from two animals and one replicate with ovaries from one animal). The number of ovaries included in each sample was selected based on the expected yields of cells as determined by a pilot experiment. All samples were dissociated into a single-cell suspension in digestion buffer (Gibco sterile 1× Phosphate buffered saline (PBS), 5% fetal bovine serum (FBS), 200 U Worthington Biochemicals collagenase IV, 200 U Qiagen RNase-free DNaseI) at 37°C using a gentleMacs dissociator with the program 37C_m_LDK1 (35-min program). The suspensions were washed with 5 mL of Gibco sterile 1× PBS with 5% FBS and 2 nM Ethylenediaminetetraacetic acid (EDTA) (UltraPure 0.5 M EDTA pH 8.0) and passed through a 100 μm cell strainer to remove debris.

Single-cell RNA sequencing and data analysis

The ovarian cell suspensions were stained with AmCyan Live/Dead stain at a 1:500 dilution in 1× PBS for 15 min and BioLegend TotalSeq HTO B series hashtags at a concentration of 10 μg/mL for 15 min. The cells were then pooled and stained with fluorochrome-conjugated primary antibodies BV650-CD45.2 (BioLegend 109835) and PerCP/Cy5.5-CD11b (BioLegend 101228) in 1× PBS with 5% FBS for 20 min. Stained cells were then washed with 1× PBS with 5% FBS and resuspended in 1× PBS with 5% FBS. Fluorescence-assisted cell sorting was performed by the Flow Core at UAMS using the BD FACSAria III cell sorter, and single live cells with positive BV650 and PerCP/Cy5.5 staining were collected into FBS/Dulbecco's Modified Eagle Medium (DMEM) (Supplemental Figure S1A).

The encapsulation of the sorted cells and single-cell library preparation were performed by the UAMS Genomics Core using the 10× Chromium Controller and Chromium Next GEM Single Cell 3′ Reagent Kits v3.1 (Dual Index) with Featured Barcode Technology for Cell Surface Proteins following the standard protocol provided by 10X Genomics. In brief, single cell suspensions with more than 70% live cells were loaded onto Chromium Controller (10X Genomics) to generate gel beads-in-emulsions. Then, co-partitioned cells were lysed, primers were released from Gel Beads, and barcoded full-length complementary DNA (cDNA) was produced and amplified. 3′ Gene expression libraries were generated from cDNA by fragmentation, end repair, A-tailing, adaptor ligation, and index polymerase chain reaction (PCR) amplification. The concentration and size distribution of final libraries were assessed by Qubit 1× double stranded DNA (dsDNA) HS Assay (Thermo Fisher Scientific, catalog # Q33231) and the Fragment Analyzer System (Agilent). The libraries were sequenced using the Illumina NextSeq 500 platform.

Raw scRNAseq data were filtered (remove low-quality reads with Q-score ≤30) and aligned to reference mm10–3.0.0, and the count matrices were generated for the young and aged libraries (performed by the Genomics Core using Cell Ranger v 7.0.0). The analysis of the data was performed using R 4.2.0 [35], with R package Seurat v.4 [36]. In brief, the young and aged datasets were filtered so that cells associated with one single HTO, %mt < 10; nFeatures between 500 and 6000 were retained. The filtered datasets were combined, normalized, scaled, and clustered using Uniform Manifold Approximation and Projection (UMAP) as a reduction method [37]. Gene set enrichment analysis (GSEA) results were generated using GSEA v.4.2.1 [38] with pre-ranked gene lists produced using Seurat and R 4.2.0. Pseudotime trajectory analysis was performed using the TSCAN package v4.3 [39]. Cell–cell communication analysis was performed using CellChat package v1.6.0 [40].

Flow cytometry

The ovarian cell suspensions were stained with eBioscience Fixable Viability Dye eFluor506 at a 1:500 dilution in 1× PBS for 15 min. Cells were then washed and stained with fluorochrome-conjugated primary antibodies at the recommended concentration for 20 min at room temperature. Monoclonal antibodies specific to mouse CD45 (BD Bioscience 566073), CD11b (BioLegend 101267), F4/80 (BioLegend 123149), CD68 (BioLegend 137023), Ly6C (BioLegend 128,017), CX3CR1 (Biolegend 149027), CD81 (BioLegend 104911), MHCII (BioLegend 107605), CD206 (Biolegend 141731), and CD11c (BioLegend 117365) were purchased from commercial manufacturers. The Foxp3/Transcription Factor Staining buffer set was used for Ki67 staining. Cells were resuspended in 200 μL of 1× PBS for subsequent analysis. To identify blood-derived cells in the ovary, 300 μL sterile PBS containing 3 μg of BV650 CD45 antibody was injected intravenously. Ovaries were collected after 3 min of injection.

Flow cytometry was performed by the Flow Core at UAMS using the Cytek Northern Lights Full Spectrum Flow Cytometer. Measurements were taken on biologically independent samples. The analysis of the data was performed using SpectroFlo, Flowjo v10.7, and Flowjo plugins UMAP v3.1. For UMAP projections of young and aged ovarian macrophages, CD45⁺F4/80⁺CD11b⁺ cells were concatenated to one Flow Cytometry Standard (FCS) file before creating UMAP projections. UMAP projections were performed using default parameters.

Fate mapping

To induce Cre recombinase activity for tracing CX3CR1+ macrophages, female Cx3cr1^CreERT/Rosa26^tdTomato mice were generated by crossing Cx3cr1^CreER and Rosa26^{floxed-tdTomato} mice. These mice were aged until they reached either 2–3 months or 14–17 months of age. To induce the recombination, tamoxifen was dissolved in corn oil and administered via oral gavage at a dose of 4 mg/day for three consecutive days. Following the last tamoxifen treatment, the mice were sacrificed at three different time points: 3, 10, and 24 days. Ovarian tissue was collected from the sacrificed mice for further analysis. All ovaries were harvested at the diestrus stage at each timepoint for analysis.

Statistical analysis

The statistical analyses in this study were conducted using GraphPad Prism 9 software. The Mann–Whitney test and one-way Analysis of Variance (ANOVA) were employed to determine the statistical significance of the flow cytometry data. For the identification of differentially expressed genes between the two age groups, the criteria for statistical significance were defined as a minimum log fold change of 0.3 and a false discovery rate–adjusted P-value lower than 0.05. False discovery rate (FDR) correction was performed using the Bonferroni method (default method used by Seurat).

Results

Single-cell transcriptomic analysis determines the complex landscape of myeloid cells in young and aged murine ovaries

To examine the aging-related transcriptomic changes in ovarian myeloid cells at single-cell resolution, scRNAseq with FACS-sorted CD45⁺ CD11b⁺ cells from the ovaries of young and aged C57BL/6 J virgin females was conducted. To ensure the reliability and reproducibility of our findings, three biological replicates for each age group were included and barcoded with hashtag oligos (HTOs).

A total of 5769 cells from the two age groups passed our quality control criteria. These cells were segregated into 15 clusters based on their transcriptomic profiles (Figure 1B). The UMAP cell cluster patterns within each age group were found to be similar among the HTO-labeled replicates, indicating high robustness and reproducibility of the results (Supplemental Figure S1B).

We annotated the clusters based on their transcriptomic signatures and identified a total of eight cell types. The most abundant cell types include macrophages (Adgre1^hi  C1qa^hi), monocytes (Adgre1^low  C1qa^lowCx3cr1^hi), DCs (Adgre1^low, Cd11c^hi), and neutrophils (S100a9^hi  S100a8^hi). The dataset also captured small numbers of natural killer (NK) cells (Gzma^hi, Nk7g^hi), B cells (Cd79a, Ly6d), fibroblasts (Col3a1, Col1a2), and endothelial cells (Cldn5, Cdh5) (Figure 1C and D). Among these cell types, macrophages were further segregated into five distinct subsets (referred to as Mac_1 to Mac_5 based on decreasing population size), monocytes consisted of two subsets distinguished by Ly6c2 expression (Ly6c2^low Monocyte_1 and Ly6c2^hi Monocyte_2), and DCs consisted of three subsets distinguished by Cd209a, Xcr1, and Ccr7 expression (Cd209a^hi DC_1, Xcr1^hi DC_2, and Ccr7^hi DC_3) (Figure 1C and D).

Ovarian macrophage population consists of multiple functionally distinct subsets

Our initial analysis focused on the ovarian macrophages since it is the most abundant among all cell types and displayed great heterogeneity in their transcriptional profiles. The dataset uncovered the presence of five macrophage subsets in the ovarian tissue (Figure 2A); each exhibits unique expression signatures indicative of different origins and functional specializations (Figure 2B).

Figure 2.

Ovarian macrophage population consists of multiple functionally distinct subsets. (A) The clustering of ovarian macrophages that are extracted from the CD45⁺CD11b⁺ ovarian myeloid dataset. (B) The expression of the top 15 represented markers for each macrophage subpopulation ranked by fold change. (C) The top relevant gene sets enriched in markers that are uniquely upregulated in each cluster. The color of the data point represents the Z-score of the net enrichment score (NES) of the gene set. Orange represents high value, while blue represents low value. (D) The violin plot showing the expression of Cd81 and Cx3cr1 in the ovarian macrophage subpopulations.

Notably, Mac_5 displayed a high expression of Gata6 (Figure 2B), which has been observed in a peritoneal macrophage population [41].

Among the four ovarian macrophage subsets (Mac_1, Mac_2, Mac_3, and Mac_4), GSEA indicates a significant enrichment of cell cycle regulation–related genes in the transcriptomic signatures of Mac_4 (Figure 2B and C), indicating its identity as a cluster of proliferating macrophages. Mac_1 was highly associated with the classic macrophage function of antigen processing/presentation, while Mac_2 was associated with ECM organization. Mac_3 was associated with activation of immune response and antigen–receptor-mediated signaling, suggesting its role in regulating immune response and inflammation. The results of the gene set enrichment were validated and visualized with the module scores for the relevant gene sets (Supplemental Figure S2). Consistently, the terms “antigen presentation,” “actin and ECM organization,” “activation of immune response,” and “cell cycle” scored high in Mac_1, Mac_2, Mac_3, and Mac_4, respectively (Supplemental Figure S2).

A closer examination of the transcriptomic signatures of the macrophage subsets revealed that the three most abundant macrophage subsets, Mac_1, Mac_2, and Mac_3, could be distinguished based on their expression of the chemokine receptor Cx3cr1 and the tetraspanin Cd81 (Figure 2D). Mac_1 exhibited high expressions of both Cx3cr1 and Cd81 (Cx3cr1^hi, Cd81^hi), Mac_2 displayed low expression of both markers (Cx3cr1^low, Cd81^low), and Mac_3 showed low Cx3cr1 and high Cd81 expression (Cx3cr1^low, Cd81^hi) (Figure 2D).

Since the functions of macrophages are heavily influenced by their activation status, we examined the expression pattern of markers that are commonly associated with classical (M1) and alternative (M2) macrophage activation status among the ovarian macrophage subsets [42]. Interestingly, a mixture of M1 and M2-associated markers was found to be expressed in all macrophage subsets (Supplemental Figure S3A). Thus, none of the ovarian macrophage subpopulations shows obvious bias toward M1 or M2 polarization.

Cx3cr1^hiCd81^hi Mac_1 represents a self-maintained low turnover ovarian macrophage subset

To validate the scRNAseq findings, we performed flow cytometry analysis with CD68⁺ F4/80⁺ ovarian macrophages (Supplemental Figure S4). Consistent with the sequencing data, ovarian macrophages segregate into distinct CX3CR1^hiCD81^hi, CX3CR1^lowCD81^low, and CX3CR1^lowCD81^hi subsets that correspond to Mac_1, Mac_2, and Mac_3 (Figure 3A). Additionally, the analysis captured a CX3CR1^hi CD81^low macrophage subset not observed in the scRNAseq data (Figure 3A), which may represent a transitional population.

Figure 3.

CX3CR1^hiCD81^hi macrophages represents a stable self-maintaining ovarian macrophage subset. (A) Representative flow cytometry plot of CX3CR1 and CD81 expression in ovarian macrophages in young and aged murine ovary. (B) Clustering of pooled CD68⁺ F4/80⁺ according to the expression of Ki67, CX3CR1, CD68, CD206, F4/80, MHCII, CD81, and Ly6C. UMAP was used as the dimension reduction method. The individual plots showed the four CX3CR1/CD81 defined macrophage subsets and the expression of Ki67, CX3CR1, CD68, CD206, F4/80, MHCII, CD81, and Ly6C, respectively. (C) Pseudotime trajectory model calculated based on the clustering of Monocyte_2, Mac_1, Mac_2, Mac_3, and Mac_4. (D) Schematic drawing illustrating the fate mapping experiment with CX3CR1^CreERT/Rosa26^tdTomato mice. (E) Percentage of total tdTomato-positive cells in the CX3CR1/CD81-defined macrophage subsets 3 days (n = 3), 10 days (n = 7), and 24 days (n = 6) post-TAM treatment. Mean ± SEM are shown. One-way ANOVA was used to test statistical significance in (E). ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

The characterization of these ovarian macrophage subsets using dimensional flow cytometry analysis led to several interesting observations. First, CX3CR1^hiCD81^low and CX3CR1^lowCD81^low (Mac_2) subsets exhibited high expression of Ly6C (Figure 3B) and lower levels of F4/80 and CD68, which is consistent with macrophages that are newly differentiated from monocytes [43]. Secondly, CX3CR1^lowCD81^hi (Mac_3) cells and CX3CR1^hiCD81^hi (Mac_1) cells showed high expression of F4/80 and CD206 (Figure 3B), suggesting their identities as tissue-resident macrophages [43]. Thirdly, although Ki67 expression was observed in all macrophage subsets, the highest Ki67 expression mostly overlaps with CX3CR1^hiCD81^hi (Mac_1) (Figure 3B), which suggests Mac_1 as a self-maintained macrophage population with a high number of proliferative cells [44]. The expression patterns of markers observed in the flow cytometry analysis largely aligned with those observed in the scRNAseq dataset (Supplemental Figure S6). Taken together, our data indicated that Mac_1 represents a mature, self-maintained ovarian macrophage subset, while Mac_2 and CX3CR1^hiCD81^low macrophages appeared to have more recently differentiated from monocytes.

To verify the relationships between the ovarian macrophage subsets, we generated a pseudotime trajectory incorporating the four ovarian macrophage subsets and the Ly6c2+ monocyte subset (Monocyte_2) using the TSCAN package [39]. Monocyte_2 was set as the root of the trajectory, as it represents the least differentiated state. The pseudotime model identified Mac_3 and Mac_4 as two endpoint clusters, inferred as derivatives of the Mac_1 subset (Figure 3C). Mac_2 was identified as an intermediate state between Monocyte_2 and Mac_1 (Figure 3C). Overall, the pseudotime analysis provided further evidence supporting the close association of Mac_1 with Mac_3 and the proliferative macrophage population (Mac_4).

To validate the self-maintaining status of Mac_1, we conducted fate-mapping analysis of the CX3CR1/CD81-defined ovarian macrophage subsets. All ovarian macrophages express CX3CR1 and thus can be labeled in the Cx3cr1^CreERT inducible fate mapping mouse model, where a tamoxifen-inducible Cre recombinase is expressed under the control of the endogenous Cx3cr1 promoter (Figure 3D). To track all ovarian macrophage subsets and measure their turnover rate in aged mice, the Cx3cr1^CreERT mouse was crossed to a Rosa26  ^tdTomato reporter mouse and a pulse-chase experiment was conducted. For the four macrophage subsets, we calculated the ratio of % tdTomato+ cells at the later time points to % tdTomato+ cells at day 3 post-tamoxifen treatment. Thus, a reduced % tdTomato retention indicates a decrease in tdTomato+ cells over time. Notably, while all four macrophage subsets were labeled by tdTomato, the % tdTomato retention decreased dramatically in both the CX3CR1^lowCD81^low (Mac_2) and the CX3CR1^hiCD81^low subsets. However, the CX3CR1^lowCD81^hi (Mac_3) and CX3CR1^hiCD81^hi (Mac_1) subsets displayed much slower decreases in % tdTomato retention (Figure 3E). These data indicate that the four ovarian macrophage subsets exhibited differences in turnover properties, with Mac_1 and Mac_3 subsets being relatively long-lived in the aged ovary, supporting the stable, self-maintaining property of Mac_1 and the relationship between Mac_1 and Mac_3.

The CX3CR1^low CD81^hi Mac_3 macrophage subset drastically expands in aged murine ovaries

The comparison of scRNAseq data between the two age groups unveils several age-associated changes. First, Mac_1 and Mac_2 exhibit drastic transcriptomic shifts between the age groups, with 1949 and 568 differentially expressed genes (DEGs), respectively. Gene set enrichment analysis showed enrichment of a distinct set of terms among DEGs from the two subpopulations in the aged group (Supplemental Figure S5). Genes associated with aerobic respiration and the electron transport chain were highly upregulated in aged Mac_2 compared to young Mac_2, while genes related to response to IFNγ and inflammation were upregulated in aged Mac_1 compared to young Mac_1.

Second, while the abundance of Mac_1, Mac_2, Mac_4, and Mac_5 did not show statistically significant differences across the two age groups, the Mac_3 subset was almost exclusively present in the reproductively old tissue (Figure 4A). This observation was further confirmed by the flow cytometry data, as the proportion of CX3CR1^lowCD81^hi subset in the total macrophage population increased significantly in aged ovarian tissue (Figure 3A and4B).

Figure 4.

A CX3CR1^low CD81^hi macrophage subset is uniquely present in aged murine ovaries. (A) The clustering of ovarian macrophages split by age, shown with UMAP as the reduction method. (B) The quantification of the CX3CR1^hiCD81^low, CX3CR1^hiCD81^hi, CX3CR1^lowCD81^hi, and CX3CR1^lowCD81^low macrophage subpopulations in young (n = 5) and aged (n = 7) murine ovaries based on the flow cytometry analysis. For CX3CR1^lowCD81^hi macrophages, P = 0.0013 (aged vs. young). (C) The top relevant terms in the GSEA results of the expression signatures of the Mac_3 macrophage subset compared to Mac_1, Mac_2, Mac_4, and Mac_5. (D) The top pathways in the outgoing and incoming signaling from and to the Mac_3 subset. The ranking was performed based on information flow. Higher value indicates higher overall signaling strength. (E) Bubble plot showing the relative strength of specific ligand–receptor interaction within the outgoing CCL and incoming GALECTIN pathways from and to the Mac_3 subset. The color of the bubble represents the relative communication probability. The Mann–Whitney test was used to test statistical significance in (B). ^**P < 0.01.

To gain insights into the properties of the Mac_3 subset, we analyzed its transcriptomic profile using GSEA [38]. This analysis revealed significant downregulation of genes related to chemotaxis, and a simultaneous upregulation of genes associated with antigen receptor signaling, immune response activation, and phagocytosis (Figure 4C). These results suggest Mac_3 as a subset of low-mobility macrophages with increased phagocytic activities. Remarkably, Mac_3 exhibits high expression of Pparg (Supplemental Figure S3A), a transcription regulator essential to promote lipid metabolism in macrophages [45]. This suggests that Mac_3 may engage in a metabolic program that enables efficient lipid utilization.

To explore the influence of Mac_3 within the tissue microenvironment, we inspected its interactions with other cell clusters within the dataset. Employing the CellChat R package, we inferred the signaling activities to and from Mac_3 in the aged myeloid cell populations based on ligand–receptor interactions [40]. Intriguingly, the analysis reveals that the outgoing signals from Mac_3 are predominantly those of the chemokine ligand (CCL) pathway, while TGFβ, vascular endothelial growth factor (VEGF), and CXCL are also among the top outgoing signaling pathways with relatively high information flow. Conversely, Mac_3 mainly receives signaling through the GALECTIN, TGFβ, and COMPLEMENT pathways (Figure 4D). A closer examination of the CCL signaling from Mac_3 indicates that the subset mainly produces Ccl6, Ccl3, and Ccl4, and the strongest communication takes place between Mac_3 and the neutrophils through Ccl6-Ccr1 and Ccl3-Ccr1 interactions (Figure 4E). GALECTIN signaling received by Mac_3 predominantly originates from other macrophage subsets and Monocyte_2, featured with Lgals9-CD45 interactions (Figure 4E).

Aged ovarian myeloid cells are featured with intensified ANNEXIN and TGFβ signaling.

To explore the temporal changes in the ovarian myeloid cells and elucidate their roles in the ovarian aging process, we analyzed and compared the cell–cell communication network in the two age groups based on the single-cell transcriptomic data. Overall, we observed an increase in both the number and the strength of cell–cell interactions among myeloid cells in the aged ovary compared to that in the young, with an exception being the outgoing signaling activities from Mac_2 (Figure 5A).

Figure 5.

Elevated ANNEXIN and TGFβ signaling are observed among aged ovarian myeloid cells. (A) The relative changes of the number and strength of signaling communication between different cell clusters in aged vs. young myeloid cells. The color of the cells represents the direction of changes. Red represents an increase, while blue represents a decrease. The bars at the end of columns and rows represent the column or row sums, respectively. (B) The ranking of pathways that are present in the communication network among both the young and aged myeloid cells based on is functional distinction between the two age groups. Higher value indicates greater changes in strength, sender/receiver identification, and dominant ligand–receptor pairs. (C) Relative information flow of the top-ranked pathways in (B) in young and aged communication networks. The length of the bar represents relative signaling strength. (D) The chord plot that visualizes the sender and receivers of ANNEXIN and TGFβ signaling in young and aged communication networks. The color of the outer and inner bars represents the identity of the sender and receiver of the signaling, respectively. The size of the bars represents the signal strength. (E) Fpr2 and Tgfbr1 in relevant cell clusters split by age. P-values of the age-associated expression changes were derived from t-tests. (F) The quantification of the count and proportions of Ly6C^hi and Ly6C^low monocytes in young (n = 8) and aged (n = 8) murine ovaries based on the flow cytometry analysis. P-values for the comparisons from left to right are: Ly6C-hi monocyte absolute count aged vs. young, P = 0.028127; Ly6C-low monocyte absolute count aged vs. young, P = 0.000466; Ly6C-hi monocyte as %Monocyte aged vs. young, P = 0.000155; Ly6C-low monocyte as %Monocyte aged vs. young, P = 0.000155. The Mann–Whitney test was used to test statistical significance in (F). ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

Pathways that are present in the signaling network of both age groups may exhibit significant functional differences due to alterations in the senders and receivers of the signaling and the predominant ligand–receptor pairs [40]. We ranked pathways shared between the two age groups based on their functional deviations in the young and aged cell–cell communication networks. Strikingly, the most prominent age-associated alterations were observed in the ANNEXIN and TGFβ pathways (Figure 5B). Additionally, BMP, VEGF, VISFATIN, and TNF pathways also ranked among the most functionally distinct between the two age groups (Figure 5B). Interestingly, all these markedly altered pathways exhibited significantly intensified signaling activities in the older age group (Figure 5C).

We delineated the most significantly altered ANNEXIN and TGFβ signaling pathways within young and aged cells and found drastic changes in both the sources and targets of these signaling interactions (Figure 5D). Among young ovarian myeloid cells, ANNEXIN signaling originates from multiple sources, including Mac_2, Mac_4, DC_1, DC_2, monocytes, and neutrophils, with neutrophils being the sole target. In aged ovarian myeloid cells, however, Mac_1 and Mac_5 emerged as additional sources of ANNEXIN signaling, while the target cell types expanded to include neutrophils, Mac_3, Mac_5, and Monocyte_1 (Figure 5D). Regarding the TGFβ pathway, there was a marked increase in overall signaling activities and a drastic expansion of incoming signaling to macrophage subsets including Mac_1, Mac_2, and Mac_3 (Figure 5D).

Upon examination of the relevant ligand and receptor expression, it became evident that the changes in ANNEXIN signaling activities were mainly a result of an upregulation of Anxa1 receptor Fpr2 in Monocyte_1 and neutrophils (Figure 5E)., as well as increased presence of Mac_3, which expresses the highest level of Fpr2 among all macrophage subpopulations (Supplemental Figure S3B). Changes in TGFβ signaling can be attributed to an elevation in Tgfbr2 expression in the Mac_2 subset (Figure 5E).

Notably, Anxa1 encodes Annexin A1 (ANXA1), an anti-inflammatory protein that orchestrates the resolution of inflammation. ANXA1 is a chemoattractant of monocytes [46–48]. Flow cytometry analysis showed that the aging-associated increase in ANNEXIN signaling to Monocyte_1 is accompanied by a significant increase in both the number and proportion of Monocyte_1 in the aged ovary (Supplemental Figure S4, Figure 5F), providing further evidence for the elevated signaling activities.

Discussion

In this study, we focused on elucidating the phenotypic heterogeneity and transcriptional changes of ovarian macrophages in young and aged mice. Five distinct ovarian macrophage subpopulations have been identified through single-cell transcriptomic analysis. Based on the expression of Ly6C, CD81, and CD11b, as well as their turnover rates in the aging ovarian tissue, the CX3CR1^hi CD81^hi Mac_1 and CX3CR1^low CD81^low Mac_2 subsets recapitulate the long-lived F4/80^hi CD11b^int and short-lived F4/80^int CD11b^hi ovarian macrophage subtypes, respectively, identified by Li et al. using flow cytometry in a recent publication [24]. In our study, Mac_1 was characterized as a mature, self-maintaining macrophage subset that resembles tissue-resident macrophages, while Mac_2 exhibited a much higher turnover rate. Moreover, GSEA analysis revealed the enrichment of “Response to interferon gamma” and “Oxidative phosphorylation” in aged Mac_1 and Mac_2 respectively, the signature of M1 and M2 polarization, implicating the distinct polarization states and functions of these two macrophage subsets in aged ovaries. Notably, these two macrophage subsets appear to share signature genes with macrophage subsets identified in the human ovary using scRNAseq [30]. Consistent with Mac_1, C1QC+, and HLA-DQA1+ macrophage subsets in human ovaries also highly express CD81 along with other activation markers and are enriched in the “Response to interferon gamma” pathway [30]. Mac_2 highly expresses Spp1, resembling SPP1+ macrophages found in human [30]. These data implicate ovarian macrophage subsets between murine and human are highly conserved and share many transcriptional similarities.

Interestingly, our data suggested the accumulation of a unique subset of CX3CR1^lowCD81^hi macrophages in aged ovarian tissue. The high expression of Pparg in this subset may implicate a metabolic adaptation to increased lipid metabolism, a signature of M2 macrophages [45]. Aging and obesity often result in a similar inflammatory milieu in the ovarian tissue, where aging is associated with increased visceral adipose accumulation and altered lipid profiles in and around the ovary [49, 50]. Thus, the higher expression of Pparg in this macrophage subset might suggest a metabolic adaptation to the altered ovarian tissue environment at advanced reproductive age. Moreover, it is known that PPARγ activation can enhance Fc-receptor-mediated phagocytosis in macrophages [51–54]. Aligning with this, genes associated with the regulation of phagocytosis are found to be upregulated in this specific macrophage subpopulation, suggesting increased phagocytic activity. Another intriguing discovery about the CX3CR1^lowCD81^hi macrophage subset is the predominance of the CCL pathway in its signaling output. This subset expresses a relatively high level of Ccl3 and Ccl4, known as macrophage inflammatory proteins that can stimulate the production of pro-inflammatory cytokines, including TNFα, IL-1β, and IL-6, within macrophages [55]. Additionally, it expresses Ccl6, a chemoattractant for monocytes and macrophages [56, 57] that has been linked to pulmonary inflammation and fibrosis [58, 59]. Importantly, inhibition of M2 macrophage polarization and/or elimination of fibrotic collagen using antifibrosis drugs pirfenidone and BGP-15 or metformin restore ovulations in aged mice [9, 60], which implicates that the CX3CR1^lowCD81^hi macrophage subset exhibiting lipid metabolic and profibrotic signatures identified in our study may be responsible for fibrosis in aged ovaries. Future studies utilizing mouse models that allow for specific depletion of this population are required to define their function in ovarian aging. It will also be interesting to examine the impact of small molecular inhibitors targeting lipid metabolism on this macrophage subset and fibrosis in the aged ovary. Moreover, histological studies have identified F4/80+ multi-nuclei giant cells that are uniquely present in aged ovarian tissue [7, 61]. The relationship between these cells and CX3CR1^lowCD81^hi macrophage subset identified in our study merits further investigation.

In addition to the emergence of the CX3CR1^lowCD81^hi macrophage subset, one of the most conspicuous changes associated with aging in ovarian tissue is the alterations observed in the cell–cell communication networks, particularly within the ANNEXIN and TGFβ pathways. Studies into the mechanisms and progression of inflammation suggested a potential interconnection between the upregulation of these two pathways. ANXA1, a pivotal player in the resolution of inflammation, performs crucial roles during infection and wound healing. It is released by infiltrating myeloid cells, particularly neutrophils and macrophages [62], to attenuate further neutrophil infiltration, recruit monocytes, and activate the expression of anti-inflammatory cytokines and TGFβ1 in macrophages [62, 63]. Our data aligned with this concept, as an increase in the presence of Monocyte_1 (Figure 5F) and expression of Tgfb1 in Mac_3 (Figure 5E) were observed in the older age group, both of which are targets of the heightened ANNEXIN signaling in aged ovarian tissue. Given that TGFβ signaling promotes the alternative activation of macrophages, the increased ANNEXIN signaling may contribute to the rise of M2-like macrophages in the reproductively aged ovary [9, 25]. Consequently, our findings cast ANNEXIN signaling as a compelling avenue warranting further investigation and offering novel therapeutic potentials.

Furthermore, neutrophils emerge as a major source of ANXA1 during wound healing [62]. Our single-cell data reflect an increase in the presence of neutrophils in aged ovarian tissue (Supplemental Figure S7A), as well as an elevation in the intensity of both incoming and outgoing ANNEXIN signaling within aged neutrophils (Supplemental Figure S7B). In each ovarian cycle, the release of the egg leads to wounding of the ovarian surface and subsequent wound healing that recruits neutrophils. The inability to effectively clear neutrophils during this process could potentially lead to the accumulation of these granulocytes and their detrimental contents in the ovarian tissue over time, which may contribute to the onset of chronic tissue inflammation and fibrosis. Our data underscore a potential role of neutrophils as a previously overlooked cellular cohort that might be intricately linked to ovarian aging.

Our scRNAseq dataset also revealed the presence of several rare cell populations with unique phenotypes, including fibroblasts, endothelial cells, and a GATA6⁺ macrophage subset (Mac_5). The fibroblast and endothelial cell clusters might simply represent cells that were carried along by true CD45⁺ CD11b⁺ cells during cell sorting. However, this fibroblast cluster expresses low levels of macrophage markers in our dataset, including Adgre1, C1qa, and Cx3cr1, reflecting the feature of myofibroblasts (Figure 1D). Myofibroblasts have been identified in the ovary and can be differentiated from macrophages through macrophage to myofibroblast transition [60]. GATA6 is expressed in granulosa cells in the ovary and is essential for follicle assembly, growth, and luteinization. Intriguingly, our scRNAseq dataset captured a small GATA6⁺ macrophage subset resembling the peritoneal macrophages, consistent with another recent study. Intriguingly, GATA6⁺ peritoneal macrophages have been shown to be recruited for tissue repair in other organs within the peritoneal cavity, such as the liver and intestine [64, 65]. Considering that the mammalian ovary undergoes tissue damage and repair in each ovarian cycle, it is possible that the peritoneal macrophages are recruited and play a functional role in the ovarian tissue repair process. Further investigation into the recruitment and functional importance of peritoneal macrophages in ovarian tissue repair would be an interesting area of research. Notably, the numbers of cells in these clusters are too low for meaningful statistical analysis between age groups in our dataset.

While our dataset provided insight into the age-associated changes in ovarian myeloid cells, we acknowledge the potential limitations in our approach. Myeloid cells are sensitive to environmental changes. The process of cell collection, staining, and sorting may have impacts on their phenotypes and transcriptomic profiles. This is a common limitation to nearly all flow cytometry and scRNA-seq analyses on primary cells isolated from tissues using an enzymatic dissociation approach, which may impact the expression of certain surface markers and lead to selective loss of sensitive cell types. With the advancement of the fixed RNA profiling and spatial transcriptomics workflow, there may be better solutions to minimize this limitation in the near future. We also noted that in order to minimize processing time, perfusion was not performed during cell collection; thus, the influence of myeloid cells from the circulation cannot be fully ruled out. However, intravenous administration of fluorescence-labeled CD45 antibodies showed that only ~4% of CD11b + cells from the murine ovaries were labeled with the CD45 fluorescence marker (Supplemental Figure S8), implicating that inclusion of myeloid cells from the circulation might be minimal.

Although we provided an in-depth analysis on macrophages in young and aged mouse ovaries using scRNAseq and flow cytometry, the current study does not directly examine the transcriptional profiles of different macrophage subsets in a spatiotemporal manner. Exploitation of spatial transcriptomic analysis with immunofluorescence microscopy on ovaries at different ages will allow a more precise and comprehensive understanding on functions of various ovarian macrophage subsets during ovarian aging in the context of tissue structure.

In addition to aging, immunological changes in ovarian tissue have been increasingly associated with various pathological conditions, such as diminished ovarian reserve [28], ovarian cancers [13], and polycystic ovary syndrome [66]. The scRNAseq dataset generated in this study has the potential to provide valuable insights into the understanding of these diseases as well. The findings of this study may have significant translational value if they can be validated in ovarian specimens from human patients.

Author contributions

ZZ is responsible for the design and execution of the experiments, data analysis, and the writing of the manuscript. LB advised the design of the experiments and writing of the manuscript. LH and MB supervised the study and advised the design/execution of the experiments and the writing of the manuscript.

Conflict of interest: The authors have declared that no conflict of interest exists.

Data availability

The data associated with this article are available in GEO database, with accession code GSE236712 and reviewer token yvktwqwotbkrrex. Any additional information required to reanalyze the data reported in this manuscript will also be available from the corresponding author upon request after publication.