To improve our understanding of the pathophysiology underlying this enigmatic disease, animal models have been employed due to the ethical limitations of performing controlled studies of infertile women with endometriosis. The procedures on animals were carried out following institutional guidelines and the Institutional Animal Care and Use Committee of Wuhan University approved the experimental protocol (Approval No. WP2020-08005).

Five-week-old female C57BL/6 mice (Vital River Laboratory, China) were housed under well-controlled conditions (12 h light/12 h dark cycle maintained at a temperature of 22–25°C). After a week of acclimation, mice were injected with 17 β-estradiol (Sigma-Aldrich, E2758) (3 μg/mouse, s.c.) for 1 week, then the endometriosis model was conducted by autologous transplantation of uterine tissue (17). Briefly, after euthanized, the left uterine horns were isolated. Obtained uterine tissue was cut into three equal-sized parts as implants, auto-transplanted was performed around three arteries of the intestinal mesentery. Sham-operated control mice (sham) were subjected to the same steps, but no implant was sutured to the intestinal mesentery. To allow the recuperation and development of endometriotic implants, the subsequent experiment began after 3 weeks.

After 3 weeks, mice were superovulated with 5 IU pregnant mare serum gonadotropin (PMSG) (Solarbio, P9970) followed 48 h later by 5 IU human chorionic gonadotropin (hCG) (LIVZON, China) to induce follicle development and ovulation. When mimicking the poor response to LH surge of ovary in vivo, the mice were treated with full-dose PMSG (5IU) followed by half-dose hCG (2.5 IU) to trigger ovulation. Ovarian tissues for follicular morphology were collected before superovulation or 48 h after PMSG administration. Granulosa cells for gene expression analysis were isolated at different times after hCG administration (0, 2, 4, and 8 h). To exam the number of ovulated oocytes, ampullae were collected at 14–16 h after hCG injection and then secured to release the clutch of cumulus–oocyte complexes (COCs). For morphology analysis of the post-ovulatory ovary, the samples were collected at 24 h after hCG administration.

For follicle counting, ovaries from each group were collected at 14–16 h after hCG injection and hematoxylin and eosin (H&E) staining was performed as described previously (18). Briefly, the right ovaries were fixed in 4% paraformaldehyde, routinely paraffin-embedded, then cut thoroughly into sections of 5-μm thickness followed by staining. In every fifth section, follicles containing oocytes with a visible nucleus were counted and properly classified into different follicle stages (19). The number of luteinized unruptured follicle and corpus luteum (CL) were also recorded.

Mice ovaries were collected as mentioned above and Immunohistochemistry was performed on formalin-fixed paraffin-embedded sections using immunoperoxidase staining kit (PV-9001, ZSGB-BIO, Beijing, China) according to the procedure of the manufacturer. After deparaffinized and rehydrated, antigen retrieval was carried out with sodium citrate. Approximately 3% hydrogen peroxide was used to eliminate the activity of endogenous peroxidase. The sections were then treated with bovine serum albumin (BSA) blocking, followed by incubation with primary antibody for LHCGR (19968-1-AP; 1:200 dilution; Proteintech), C/EBPα (18311-1-AP; 1:200 dilution; Proteintech), COX-2 (ab15191; 1:200 dilution; Abcam) and corresponding secondary antibody. The omission of the primary antibody served as a negative control. The H-score was processed and calculated as described previously (20). We used the following equation: H-score = ∑ Pi (i), where i was the intensity of staining with a value of 1, 2, or 3 (weak, moderate, or strong, respectively) and Pi was the percentage of stained cells for each intensity, varying from 0 to 100%.

The study protocol was approved by the Institutional Review Board (No. 2018047). Human GCs were obtained from patients undergoing in vitro fertilization (IVF)/intracytoplasmic sperm injection (ICSI) treatment due to tubal factor or male subfertility at the Reproductive Medical Center of Zhongnan Hospital. After controlled ovarian hyperstimulation, 10,000 IU hCG was administered to trigger ovulation. The follicular fluid was immediately collected and centrifuged for 10 min at 2,000 rpm after oocyte pick-up. Then the pellet was resuspended in an enzymatic solution to digest clusters of cells. GCs were highly isolated through Percoll density gradient and red blood cells were removed using lysis buffer. The cell pellet was resuspended in DMEM/F12 medium (Gibco) supplemented with100 U/ml penicillin, 100 µg/ml streptomycin, and 10% (v/v) fetal bovine. The cells were then seeded at a density of 2 × 10⁵ cells/ml in a 6-well plate and incubated for 3 days at 37°C in humidified atmosphere with 5% CO₂. The media was replenished every 24 h. To mimic the effect of LH and cAMP in vivo, we used hCG (10 IU/ml; LIVZON, China) and dibutyryl-cAMP (db-cAMP, 1 mM; HY-B0764A, MedChemExpress, USA) respectively to stimulate cells and further cultured for stated hours in vitro according to previous studies (21). KGN cell line was cultured in DMEM/F12 medium as mentioned above.

For gene silencing experiments, human GCs were transfected with 50 uM small interfering (siRNA) oligonucleotides against LGCGR, C/EBPα or negative control (NC) siRNA (Huzhou Hippo Biotechnology Co., Ltd.) using lipofectamine 3000 transfection reagent (Invitrogen, USA) according to the instructions provided by the manufacturer. The specific sequences of target genes were as follows: si-LHCGR, 5′-UGC CUU CAA AGU ACC UCU UAU TT-3′ (sense) and 5′-AUA AGA GGU ACU UUG AAG GCA TT-3′ (antisense); si-C/EBPα, 5′-GGA GCU GAC CAG UGA CAA UTT-3′ (sense) and 5′-AUU GUC ACU GGU CAG CUC CAG-3′ (antisense); si-LHCGR scrambled NC, 5′-GUC AUU AUC CUU UCG CAC UAA dTdT-3′(sense) and 5′-UUA GUG CGA AAG GAU AAU GAC dTdT-3′ (antisense); si-C/EBPα scrambled NC, 5′-GGU AAC GGG ACC GAC UUA AdTdT-3′ (sense) and 5′-UUA AGU CGG UCC CGU UAC CdTdT-3′ (antisense). For further experiments which were focused on the mechanisms of signal pathways, the cells were incubated with or without db-cAMP or hCG for further 24 h after 24 h of transfection (22).

Total RNA was extracted with an RNA extraction kit (RN0302, Aidlab, China). RNA (1 ug) was reverse transcribed with a cDNA Synthesis Kit (R212-01, Vazyme, China). Quantitative real-time PCR (qRT-PCR) was performed using ChamQ SYBR qPCR Master Mix (Q311-02, Vazyme, China) and a CFX96 PCR system machine (Bio-Rad Laboratories, USA). Each reaction was performed with a total volume of 20 μl, consisting of 2× ChamQ SYBR qPCR Master Mix (10 µl), 5’- and 3’-primer (0.4 µl, respectively), cDNA (1 µl), and ddH2O (8.2 µl). With the following primers: LHCGR: 5′-TCC TTT CCA GGG AAT CAA TC-3′ (sense) and 5′-GGC CGG TCT CAC TCG AC-3′ (antisense); C/EBPα: 5′-CAC GAA GCA CGA TCA GTC CAT-3′ (sense) and 5′-CGG AGA GTC TCA TTT TGG CAA G-3′ (antisense); COX-2: 5′-TAA GTG CGA TTG TAC CCG GAC-3′ (sense) and 5′-TTT GTA GCC ATA GTC AGC ATT GT-3′ (antisense); GAPDH: 5′-CTG TTC GAC AGT CAG CCG CATC-3′ (sense) and 5′-GCG CCC AAT ACG ACC AAA TCC G-3′ (antisense). Data analysis was performed using Bio-Rad CFX manager system, using GAPDH as a reference transcript.

Whole-cell protein extract was lysed and isolated from cultured cells or mouse ovaries. After measuring protein concentrations using a BCA Protein Assay Kit (P0010, Beyotime, China), equal amounts of denatured protein were separated by electrophoresis in 10% SDS polyacrylamide gels and transferred to polyvinyl difluoride membranes (Millipore, Billerica, USA), which were then saturated with blocking buffer for 1 h. After that, the membranes were incubated with rabbit polyclonal anti-LHCGR (19968-1-AP; 1:1,000 dilution; Proteintech), rabbit polyclonal anti-CEBPα (8178; 1:1,000 dilution; Cell Signaling), or rabbit polyclonal anti-COX2 antibodies (ab 15191; 1:1,000 dilution; Abcam) O/N at 4°C. The blots were incubated with HRP–conjugated anti-rabbit IgG for 1 h. Peroxidase activity was detected using the ECL system (Touch Imager, e-Blot, China).

After treatment for stated hours in vitro, cells were fixed with paraformaldehyde. Then, the cells were treated with 0.5% Triton solution to encourage intracellular labeling. Cells were next blocked with 5% BSA for 1–2 h at RT and incubated with primary antibody against CEBPα (8178; 1:200 dilution; Cell Signaling) O/N at 4°C. Secondary antibody incubation was performed at RT for 60 min. Cells were washed and treated with antifade mounting medium, which contains DAPI (P0131, Beyotime, China).

The effect of C/EBPα on the transcriptional activity of COX-2 promoter was determined by analyzing the dual-luciferase activities using a commercial assay kit (E2920, Promega, USA). The expression construct used for luciferase-based assays was pcDNA3.1 C/EBPα (NM_004364) while reporter construct used was pGL3-PTGS2 promoter (−1.2 kb/+137). KGN cells were co-transfected with the indicated plasmids with the help of lipofectamine 3000 transfection reagent (Invitrogen, USA) as previously described (23). The samples were lysed at 48 h after transfection, detection of luciferase activity was conducted. Firefly luciferase measurements were normalized to Renilla luciferase.

After treatment with hCG or db-cAMP, KGN cells (2 × 10⁷) were collected and processed as described previously (24). Briefly, cells were washed and fixed in 1% formaldehyde for 15 min then cross-linking was terminated using 0.125 M glycine. Then the chromatin immunoprecipitation (ChIP) assay was performed to determine whether CEBPα interacts with the putative binding site in COX-2 promoter using a Simple ChIP Kit (56383, Cell Signaling) according to the protocols of the manufacturer. Approximately 1% of the chromatin fragments were stored at −20°C to be used later for input for normalization. For each immunoprecipitation (IP) reaction, every 5 ug chromatin sample was incubated with 4 ug CEBPα antibody (18311-1-AP, Proteintech) O/N at 4°C or with 1 ul IgG antibody (2729, CST) as a negative control for nonspecific IP. The primers for the COX2 promoter used in ChIP-PCR analyses were as follows: 5′-TCTAGGAAGCCTTTCTCCTCCT-3′ (sense) and 5′-TGATCCACGCTCTTAGTTGAAAT-3′ (antisense). The resulting signals were normalized to input values, with the IgG-negative control values subtracted as background.

Data were calculated as percentages or ratios relative to the corresponding negative controls, presented as means ± SEM, and were appropriately analyzed by ANOVA, or unpaired t-test with GraphPad Prism (Version 8.1.1, California). Values of P <0.05 were considered statistically significant.

After confirming the induction of surgical-induced endometriosis three weeks after the operation, we initially collected the ovaries and evaluated the impact of endometriosis on the general morphology and ovarian reserve ( Figures 1A, B ). The anatomical observation indicated both morphology and ovarian weight were similar between EMs and sham mice. Counts of primary, secondary and antral follicles in the endometriosis model (78.00 ± 9.72, 46.67 ± 9.14, and 16.00 ± 2.00, respectively) were comparable to those of sham-operated mice (84.00 ± 11.24, 44.33 ± 10.33, and 16.67 ± 2.94, respectively) ( Figure 1C ). These results suggested that endometriosis mice presented similar healthy follicles to sham mice.

Ovarian responsiveness to gonadotropins was further evaluated to examine if folliculogenesis and ovulation are affected by endometriosis. Firstly, ovaries were collected 48 h after PMSG treatment when follicles developed to preovulatory stage ( Figure 1D ). Quantification of ovarian follicles indicated that the number of preovulatory follicles showed no significant differences in animals of both groups (7.33 ± 1.50 in sham vs. 7.17 ± 1.94 in EMs). Subsequently, the number of ovulated oocytes was assessed after a superovulation protocol to further investigate whether ovulation was affected by endometriosis ( Figure 1E ). After 16 h of hCG administration, fewer cumulus–oocyte complexes (COCs) were released in EMs mice compared to that in sham-operated mice (9.12 ± 2.31 in EMs vs. 13.67 ± 1.37 in sham), indicating the ovulatory capacity was compromised in endometriosis mice.

During ovulation, the follicle ruptures and oocyte is released, the remaining GCs and theca cells under the influence of LH are luteinized to form a corpus luteum (CL). EMs ovaries showed increased luteinized unruptured follicles ( Figures 1F, G ), characterized by the oocytes destined for ovulation becoming entrapped in preovulatory follicles or corpora lutea within a full investment of luteinized granulosa cells. These observations established that, in endometriosis, LUFs leads to reduced ovulation and abnormal CL formation, and it may be a cause of endometriosis-associated infertility.

To uncover the mechanism responsible for LUFs, we mimicked the poor response to LH surge of ovary in vivo by treating the mice with full-dose PMSG (5I U) followed by half-dose hCG (2.5 IU) to trigger ovulation. Although similar trends had been observed in full-dose hCG mice occurred in sham mice, the EMs mice displayed a significant reduction of ovulatory oocytes and more frequent incidence of LUFs after administration of half dose of hCG in vivo ( Figures 2A, B ). Moreover, the expression pattern of LHCGR protein in the GCs collected at different time points after hCG administration was determined by Western blotting ( Figure 2C ). The results showed that LHCGR protein levels were lower in the GCs of EMs mice in the early ovulatory phase compared with controls, the reduced trends even lasted to late ovulatory phases. IHC staining confirmed that the intensity of LHCGR staining appeared weaker and sporadic in GCs of hCG-primed (0 and 8 h) EM mice than that in sham mice ( Figures 2D, E ). These results suggested that decreased LHCGR induces a poor response to LH surge and therefore contribute to the pathogenesis of LUFs. More importantly, the dysfunction presents before endogenous LH surge.

LH signaling primes many key ovulatory genes in granulosa cells. To further look into the mechanism responsible for LUFs, we first investigate whether LHCGR is involved in the expression of ovulation-related genes (25–27). The qRT-PCR analysis indeed showed that the expression levels of known genes, such as VEGFA, COX-2, AREG, and EREG were significantly diminished in the human GCs of LHCGR knockdown than those of the negative control (NC) ( Figure 3A ). We further analyzed the gene expression between EMs and sham mice. Both VEGFA and COX-2 were significantly decreased in the GCs of EMs mice ( Figure 3B ). Since abnormal COX-2 function is also associated with ovulation failure (28), these observations led us to investigate the correlation between LH signaling and COX-2.

To determine the effect of LH signaling on COX-2, human GCs were stimulated with 10 IU/ml hCG to mimic the in vivo induction. The expression pattern of COX-2 in the human GCs collected at different time points after hCG administration was determined ( Figures 3C, D ). As expected, hCG treatment induced the expression of COX-2 in levels of mRNA and proteins at 24 h, and the levels remained appreciable even at 36 h after hCG treatment ( Figures 3C, D ). To reveal the functional role of LHCGR involved in the expression of COX-2 during the periovulatory period, RNA interference (RNAi) approach was employed to knock down LHCGR transcripts in the presence or absence of hCG. The knockdown of LHCGR per se, rather than negative control, recapitulated the hCG induced COX-2 upregulation ( Figures 3E, F ).These results revealed that LHCGR is involved in the hCG-induced upregulation of COX-2 expression in human GCs.

We further assess the expression of COX-2 during ovulation, granulosa cells were collected at different time points after hCG (0, 2, 4, and 8 h) treatment for analysis ( Figure 3G ). COX-2 protein of the granulosa cells from EMs mice were significantly decreased compared to sham mice. IHC staining for COX-2 showed that COX-2 was mainly localized to granulosa and theca cells of dominant follicles during both early and late ovulatory phases, and abundant COX-2 was found at 8 h after hCG priming, whereas little staining was observed in EMs mice ( Figure 3H ). Therefore, we conjecture that the endometriosis-related abnormal actions of LHCGR modulate downregulation of COX-2 in GCs, then results in reduced ovulation with impaired follicle rupture.

We further sought to identify underlying mechanisms of the LHCGR-induced COX2 upregulation. There is evidence indicating the expression and functional activation of C/EBP family members is essential for events associated with reproduction (29, 30). We reasoned that C/EBPα may participate in the LHCGR-induced COX2 expression. Accordingly, Western blot was conducted to investigate the expression pattern of C/EBPα in granulosa cells collected at different time points after hCG administration and the results showed that C/EBPα was diminished in EMs mice ( Figure 4A ). IHC staining showed that, after PMSG administration (48 h), follicles were either mature or in the process of ovulation, C/EBPα was expressed in the granulosa and theca cells of superovulated mouse ovaries. After treatment with hCG, the protein level was significantly increased. Interestingly, C/EBPα staining in the EMs mice appeared to be attenuated in the granulosa cells ( Figure 4B ).

The results demonstrated that hCG escalated the expression of C/EBPα in granulosa cells after 24 h in both mRNA and protein levels ( Figures 4C, D ). We also confirmed that hCG significantly induced C/EBPα expression in the nuclei of human GCs ( Figure 4E ).

Consistently, knockdown of LHCGR affected the basal levels of C/EBPα expression; it also further significantly diminished the hCG-induced C/EBPα expression ( Figures 4F, G ). These results provided evidence that C/EBPα may involve in regulation by LHCGR signaling in human GCs. To further substantiate our observation, siRNA-mediated down-regulation of endogenous C/EBPα was employed, and we found that the expression of COX-2 was down-regulated after knockdown endogenous C/EBPα in GCs ( Figures 4H–J ).

To interrogate the bona fide regulation of C/EBPα on COX-2, dual-luciferase reporter assay using KGN cells was conducted and the results showed that C/EBPα was sufficient to operate as a transactivator of COX-2 transcription since the luciferase activity of cells transfected with COX-2 wild-type reporter plasmid (COX-2-wt) was strongly improved after co-transfected with C/EBPα-overexpression (C/EBPα-oe) plasmid in a dose-dependent manner ( Figure 4L ). These results indicated that C/EBPα activated the transcription of COX-2. To address the potential binding region of C/EBPα in the COX-2 promoter region, a bioinformatics analysis was conducted by JASPAR database (http://jaspar.genereg.net/), a putative C/EBPα-binding site located at position −416/−403 of the COX-2 promoter was identified ( Figure 4K ). The results of dual-luciferase reporter assay verified that the luciferase activity in cells co-transfected with COX-2 mutant reporter plasmid (COX-2-mut) and C/EBPα-oe plasmid was not altered ( Figure 4L ). Furthermore, ChIP-PCR assays were conducted to validate the molecular interaction between C/EBPα and the identified biding site in COX-2 promoter in KGN cell line collected 24 h after treatment with hCG ( Figure 4M ). The result unveiled that the immunoprecipitation of the C/EBPα antibody-enriched DNA fragments containing identified binding sites, demonstrating CEBPα was strongly bound to the promoter region upstream from the transcriptional start site of COX-2 gene.

Further experiments showed that hCG-increased expression of COX-2 was attenuated by knock down of C/EBPα in human granulosa cells ( Figures 4N–P ). Characterization of C/EBPα showed the same pattern of expression. These results suggest that C/EBPα plays an important role in hCG-induced COX-2 expression in human GCs.

Cognate receptor of LH is G-protein coupled receptor that is predominantly mediated by activation of adenylate cyclase and cAMP-dependent mechanisms in ovarian follicle growth and maturation (31). As C/EBPα is identified with both constitutive and cAMP inducible activities (32), we next examined whether COX-2 expression was mediated through a cAMP- regulated pathway.

First, ChIP-PCR assay was conducted to clarify the enrichment of C/EBPα bound to the COX-2 promoter ( Figure 5A ). The interaction was also confirmed by results obtained in KGN cells treated with db-cAMP (24 h), indicating that C/EBPα directly binds to the promoter of COX-2 gene to regulate its expression in a cAMP-dependent manner.

To further determine the role of cAMP, human GCs were treated with db-cAMP in vitro. The observations showed that db-cAMP improved C/EBPα expression in the nuclei of human GCs as well ( Figure 5B ). Treatment of human GCs with db-cAMP for 24 h significantly induced the expression of C/EBPα and COX-2 ( Figures 5C–E ). Indeed, knockdown of LHCGR not only decreased the expression of C/EBPα and COX-2 in basal treatment, but it also exerted a significant inhibitory effect of cAMP-induced expression of C/EBPα and COX-2 ( Figures 5F–H ). Additionally, knockdown of C/EBPα counteracted the cAMP-induced COX2 upregulation ( Figures 5I–K ). Here, we documented that C/EBPα and COX-2 may be induced by LHCGR signaling in a cAMP-dependent manner in granulosa cells. Furthermore, cAMP can promote the transcriptional activity of C/EBPα.