In Silico, In Vitro, and In Vivo Analysis Identifies Endometrial Circadian Clock Genes in Recurrent Implantation Failure

Abstract

Context

Previous work has demonstrated the role of the circadian clock in ovarian steroid hormone synthesis and attributed embryo implantation failure associated with arrhythmic circadian clock genes to insufficient ovarian-derived progesterone synthesis. Research on expression of core circadian clock genes in the endometrium itself and possible roles in compromised endometrial receptivity and recurrent implantation failure (RIF) are limited.

Objective

We aimed to assess the core circadian clock gene profiling in human endometrium across the menstrual cycle and the possible gene interaction networks in the endometrial receptivity of window of implantation (WOI) as well as RIF.

Methods

The study was initially an in silico study, with confirmatory lab-based data from primary human endometrial stromal cells (hESCs) as well as endometrial biopsies obtained from 60 women undergoing gynecological surgery in a clinical research center. The study included 30 RIF women and 30 age-matched and body mass index–matched controls.

Results

Initial data mining and bioinformatics analysis of human endometrial microarray datasets across the menstrual cycle and between RIF women versus controls demonstrated the varied expression of core circadian clock genes across menstrual cycle, including the key role of PER2 in WOI and RIF. A PER2-centered network was investigated in the regulation of endometrial receptivity. We also confirmed the evidently increased mRNA expression of SHTN1, RXFP1, KLF5, and STEAP4 in the endometrium of RIF women, displaying the same trend as PER2 did, without any changes in MT1E and FKBP5. Treatment of PER2 siRNA in hESCs verified the positive regulation of PER2 to SHTN1, KLF5, and STEAP4.

Conclusion

Aberrant expression of endometrial PER2 might contribute to impaired endometrial receptivity and development of RIF via regulating SHTN1, KLF5, and STEAP4.

An embryo can implant only when the endometrium is receptive to it during the restricted “window of implantation” (WOI), 7 to 10 days after ovulation in the middle secretory phase (MSE) of the menstrual cycle (1, 2). Endometrial “receptivity” generally refers to the competence to accept an embryo during the WOI, which is closely related to successful implantation (3). Recurrent implantation failure (RIF) is defined as failure of 3 embryo transfer cycles in which 1 or 2 high-grade quality embryos are transferred into the uterus in each cycle in the WOI (4, 5). Pathological alterations in the endometrium such as hyperplasia, submucosal myomas, or endometrial polyps can result in RIF (6); aneuploidy in the embryos, and other factors either acting directly as embryo-toxics or that adversely affecting endometrial receptivity can also lead to RIF. Accumulating evidence has also suggested impaired endometrial receptivity at the molecular level in patients with unexplained RIF (7, 8). Thus, endometrial receptivity is linked to WOI and plays vital roles in embryo implantation as well as the pathogenesis of RIF (9).

Circadian rhythms are physical, mental, and behavioral changes that follow a 24-hour cycle. The regulation of the circadian clock is important for human physiological function and disease. In mammals, the master circadian clock is located in the suprachiasmatic nucleus (SCN) of hypothalamus. Daily changes are controlled by several interlocked transcriptional and post-translational autoregulatory feedback loops composed of core circadian clock genes. The first feedback loop begins when circadian locomotor output cycles kaput (CLOCK) and aryl hydrocarbon receptor nuclear translocator-like protein 1 (ARNTL) heterodimers enter the nucleus and drive the transcription of period 1/2/3 (PER1/2/3) and cryptochrome 1/2 (CRY1/2) genes by binding to an E-box (CACGTG) of promoters (10, 11). Once the concentration of PERs/CRYs proteins reaches a critical level, they re-enter the nucleus and repress the transcription of ARNTL/CLOCK, which results in the oscillations of ARNTL/CLOCK and PERs/CRYs expression following a circadian pattern (12). In addition, the ARNTL/CLOCK heterodimer also drives the transcription of orphan nuclear receptors, especially nuclear receptor subfamily 1 group D member 1 (NR1D1) and retinoic acid receptor-related orphan receptors α (RORα). In turn, NR1D1 represses ARNTL transcription while RORα induces its transcription, forming another feedback loop that interconnects the positive and negative limbs of the core circadian clock genes (13, 14). Neuronal PAS domain protein 2 (NPAS2), a paralog of CLOCK, can also form a heterodimer with ARNTL and effectively functions in the regulation of mammalian circadian rhythms (15).

A large literature has demonstrated functional roles of circadian rhythms in reproductive processes (16). Epidemiological evidence reveals that shift work is associated with menstrual irregularities and reproductive difficulties in women (17). In vivo experiments also demonstrate the role of core circadian clock genes in female reproduction. Arntl^−/− female mice display irregular estrous cycles, absence of the proestrus luteinizing hormone surge, and implantation failure; similar phenotypes are also present in Clock^Δ19 mutated mice (18-20). Loss of Arntl in ovarian steroidogenic cells also leads to embryonic implantation failure (21). Thus, both global and peripheral disruption of core circadian clock genes can affect embryonic implantation. However, most studies attribute aberrancies in implantation success to insufficient progesterone due to dysregulated ovarian function. It should be noted that exogenous progesterone supplement can only partially (38%) rescue the implantation failure in Arntl^−/− female mice (18), implicating dysfunction of endometrium. Per2 oscillation is reportedly enhanced in endometrial stromal cells during implantation, whereas it is attenuated during decidualization in mice (22). Thus, endometrial circadian clock genes may also play key roles in endometrial receptivity for embryo implantation and pregnancy. However, few studies have focused on the role of core circadian clock gene in the endometrial receptivity, the WOI, or the pathogenesis of RIF.

Using microarray data from publicly available Gene Expression Omnibus (GEO) database, we designed this in silico and lab-based study to clarify the expression of core circadian clock genes in endometrium across the menstrual cycle, including proliferative (PE), early secretory (ESE), mid-secretory (MSE), and late secretory (LSE) phases. Moreover, we investigated endometrial circadian clock genes and associated target gene networks in endometrial receptivity, the WOI, and RIF. The data aimed to suggest novel roles and potential personalized applications to diagnose and treatment for unexplained RIF, a clinically challenging disorder.

Methods

Datasets

Using the keywords Endometrium, Menstrual cycle phase, and Homo sapiens, we searched the associated gene expression profiles in public available GEO database (https://www.ncbi.nlm.nih.gov/geo/). We chose 4 samples from PE, 3 samples from ESE, 8 samples from MSE, and 6 samples from LSE in GSE4888 (23) for further analysis and excluded the other 6 ambiguous samples. All selected samples were from normally ovulatory women. We retained gene expression datasets from the Affymetrix Human Genome U133 Plus 2.0 Array (HG-U133_Plus_2) and detected expressions of core circadian clock genes throughout the menstrual cycle.

In order to investigate the role of core circadian clock genes in the pathogenesis of RIF, we searched the gene expression profiles of the endometrium of women with and without RIF in GEO database using the keywords Endometrium, Recurrent implantation failure, and Homo sapiens. We chose GSE111974 (24 controls and 24 RIF patients) for further investigation. Meanwhile, we detected the expression of circadian clock genes and potential target gene network in RIF patients and controls.

Selection of Core Circadian Clock Genes

We first assembled a list of 12 core circadian clock genes from published literature and review (12, 24, 25), including CLOCK, ARNTL, CRY1/2, PER1/2/3, casein kinase 1ε (CK1ε), NR1D1, RORA, NPAS2, and timeless (TIM). Then we restricted the list to 9 genes with available RNA expression data from GSE4888 and GSE111974 datasets. This yielded a total of 9 core circadian clock genes, including CLOCK, ARNTL, CRY1/2, PER2/3, NR1D1, RORA, and NPAS2. We systematically compared the expression of these 9 genes in PE versus SE, PE versus MSE, MSE versus LSE, and RIF patients versus controls using Wilcox test in R software (*P < 0.05, **P < 0.01, ***P < 0.001).

Correlations and Interactions Among Core Circadian Clock Genes

Interactions among core circadian clock genes were analyzed using the STRING (http://www.string-db.org/). Moreover, the correlation of PER2 and other potential target genes was identified using Spearman correlation in the “Corr” and “Corrplot” package of R software. P < 0.001 was considered as significantly correlated to each other.

Differentially Expressed Genes

After downloading GSE4888 and GSE111974 datasets from GEO, we re-annotated the probes. The “impute” package of R software was used to impute the missing expression data while “limma” package was used to normalize the gene expression and identify the differentially expressed genes (DEGs) separately between PE versus MSE, MSE versus LSE in GSE4888 and RIF versus controls in GSE111974. The significant difference was defined as $\left|\mathrm{l}\mathrm{o}{\mathrm{g}}_2\mathrm{F}\mathrm{C}\right|>1$ and adjust P < 0.05.

Weighted Gene Co-expression Network Analysis

To verify the co-expression and potential target genes of differential expressed core circadian clock genes, we used another method to analyze the DEGs in PE versus MSE, MSE versus LSE in GSE4888 and RIF versus controls in GSE111974, respectively. Weighted gene co-expression network analysis (WGCNA) assigns a connection weight to each gene pair in the network, this being more meaningful relative to traditional methods that use binary information. Thus, WGCNA can be used to find modules of highly correlated genes and identify candidate target genes (26). We used WGCNA to analyze the DEGs and identified the potential target genes of differentially expressed core circadian clock genes.

Enrichment Analysis

Gene Ontology (GO) analysis was used to identify the possible molecular function and visualize the potential biological meaning behind DEGs, whereas Kyoto Encyclopedia of Genes and Genomes (KEGG) was used to analyze the potential functions of these genes. After WGCNA, the co-expressed genes of differential expressed core circadian clock genes were enriched to investigate the biological process and pathways of circadian clock genes that were involved in endometrium. The gene ID was set using the “org.Hs.eg.db” of R software, and then GO and KEGG pathway enrichment analyses were performed with “clusterProfiler.”

Clinical Sample Collection

Endometrial biopsies from 30 controls and 30 RIF patients were obtained for the validation cohorts. All participants in the study were retrospectively recruited from the Center for Reproductive Medicine, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University who underwent in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI) treatment. Women who gave live birth after IVF/ICSI treatment were recruited as controls. RIF patients were women who failed to get implantation after 3 embryo transfer cycles in which 1 or 2 high-grade quality embryos were transferred to the patient in each cycle. Endometrial tissues were collected on the seventh day after ovulation (WOI) during natural menstrual cycles either before controlled ovarian hyperstimulation or in natural frozen embryo transfer cycle of IVF/ICSI. The Institutional Review Board of the Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, reviewed and approved all procedures (2017041409).

Cell Culture

Endometrial tissues (PE phase) for stromal cell isolation and culture were obtained prospectively by biopsy from participants undergoing hysteroscope. Fresh endometrial tissues were digested with collagenase type I and hyaluronidase in Hanks buffered salt solution as previously described (27). Contaminant red cells were lysed, and the dissociated cellular elements were treated with DNase. Then cells were filtered using a 40-μm cell strainer (BD Biosciences, CA, USA) to separate stromal cells from epithelial glands. Finally, primary human endometrial stromal cells (hESCs) selectively attached to cell culture flasks and then were cultured in Dulbecco’s modified Eagle medium/Hams F12 (Gibco, Grand Island, NY) containing 10% fetal bovine serum (Gibco) and 1% antibiotic antimycotic solution (Gibco). All cells were incubated at 37 °C in a humidified atmosphere, with 5% CO₂. Cell purity was tested by immunofluorescence staining for cytokeratin 18 and vimentin.

Immunofluorescence

For immunofluorescence of hESCs, cells were seeded on 4-well chamber slides (Millipore, Massachusetts, USA) and immunofluorescence was done as previously described (28). The primary antibodies were cytokeratin 18 (1:100 dilution; 10830-1-AP, Proteintech, Wuhan, China) and vimentin (1:100 dilution; 60330-1-Ig, Proteintech). Analysis of signals was carried out by fluorescence microscopy (Zeiss, Oberkochen, Germany) and digitally photographed.

Immunohistochemistry

Five μm-thick tissue frozen sections were prepared using optimal cutting temperature compound blocks of endometrium tissues (WOI) in both controls and RIF patients. Slides were rehydrated and then blocked using blocking buffer for 1 hour at room temperature. Heat-mediated antigen retrieval was carried out with 10 mM sodium citrate, 0.05% Tween 20, pH 6, and then slides were incubated in anti-PER2 antibody (1:100 dilution; ab179813, Abcam, Cambridge, UK) overnight at 4 °C. After being washed with phosphate-buffered saline, the slides were processed with the secondary antibody (1:400) for 1 hour at room temperature, and then the color reaction was visualized by exposure to diaminobenzidine. Slides were counterstained with hematoxylin and dehydrated through graded alcohols and xylene before visualizing using a microscope (Zeiss).

Real-Time Quantitative Polymerase Chain Reaction

Total RNA from cells and tissues was extracted using an Animal Total RNA Isolation Kit (FOREGENE, Chengdu, China) and then reverse-transcribed into cDNA (TAKARA, Dalian, China). The mRNA expression of target genes was detected using real-time quantitative polymerase chain reaction (RT-qPCR). Results were analyzed using the ΔΔCt method. The ratio of target gene to β-ACTIN expression was calculated and reported as the target mRNA level. The primer sequences of targeted genes are presented below:

PER2 (human): 5′-CTGGCCATCCACAAAAAGAT-3′ (forward) and 5′-CCTCCCAATGATGAAGGAGA -3′ (reverse);

FKBP5 (human): 5′-AATGGTGAGGAAACGCCGATG -3′ (forward) and 5′-TCGAGGGAATTTTAGGGAGACT -3′ (reverse);

KLF5 (human): 5′-TTCCACAACAGGCCACTTACT -3′ (forward) and 5′-GAGCATCTCTGCTTGTCTATCTG -3′ (reverse);

RXFP1 (human): 5′-CCTGAAGCCGGGTGTTTTTG -3′ (forward) and 5′-GGGAAATTCGACTGAGGTGATT -3′ (reverse);

MT1E (human): 5′-TCAGGTTGGGAGGGAACTCAA -3′ (forward) and 5′-GAAAGCCTGGAGAGGGAATGA -3′ (reverse);

SHTN1 (human): 5′-GAAAGAAAAGGCAACTC AACCAG -3′ (forward) and 5′-GCTGTCTGATTAACG GGTCTAAG -3′ (reverse);

STEAP4 (human): 5′-GGCTTTGGGAATACTTGGGTT -3′ (forward) and 5′-TGGACAAATCGGAACTCTCTCC -3′ (reverse);

β-ACTIN (human): 5′-GGGAAATCGTGCGTGACA TTAAG -3′ (forward) and 5′-TGTGTTGGCGTAC AGGTCTTTG -3′ (reverse).

Transfection of Small Interfering RNA With Liposome

Primary hESCs of passage 1 were seeded on 6-well plates and maintained in culture medium for 24 hours prior to the start of transfection experiment. PER2 siRNA was transfected into cells using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. The hESCs were plated onto 6-well plates and cultured until 80% confluence. After refreshing the medium, Lipofectamine RNAiMAX reagent and small interfering RNA (siRNA; GenePharma, Shanghai, China) against PER2 were mixed with Opti-MEM (Gibco) and then evenly added the same amount to each well. Cells were further incubated for 48 hours before measuring the efficiency of knock-down and the target gene abundance. The sequence of siRNA was presented below:

PER2 siRNA: 5′-GGACAUGAGACCAACGAAATT -3′;

Negative control (NC) siRNA: 5′-UUCUCCGAACGUGUCACGUTT-3′.

Statistical Analysis

Results are presented as mean ± SEM or SD. The differences between controls and RIF patients were analyzed in unpaired Student’s t test with SPSS software (IBM, NY, USA). Statistical significance is shown as *P < 0.05, **P < 0.01, or ***P < 0.001.

Results

Expression and Correlation of Core Circadian Clock Genes Throughout the Menstrual Cycle in Normo-Ovulatory Women

We reanalyzed the expression of 9 core circadian clock genes across the menstrual cycle from GSE4888. STRING analysis suggested that 9 circadian clock genes interacted closely with each other (interaction score ≥ 0.7) (Fig. 1A). Compared with PE, the expression of endometrial CLOCK and PER3 in the secretory phase (including ESE, MSE, and LSE) was greatly decreased (P = 0.05 and P = 0.001, respectively), especially in LSE (Fig. 1B-1D). Contrary to CLOCK and PER3, expression of PER2, ARNTL, NPAS2, CRY2, and RORA was significantly increased in SE versus PE (P = 0.002, 0.004, 0.009, and 0.02, respectively) (Fig. 1B, 1E-1I). While the expression of ARNTL, NPAS2, and CRY2 was elevated throughout the entire secretory phase (Fig. 1F-1H), PER2 expression significantly increased in ESE and MSE and dramatically decreased in LSE (Fig. 1E), suggesting a possible role for PER2 in endometrial receptivity of the WOI. No evident changes were observed in the expression of CRY1 and NR1D1 across the menstrual cycle (Fig. 1J and 1K).

Figure 1.

Expression and interactions of core circadian clock genes in endometrium of normo-ovulatory women across menstrual cycles.

A, Interactions among core circadian clock genes from STRING. B, Expression of 9 core circadian clock genes in the endometrium across menstrual cycles (PE, ESE, MSE, and LSE). C-K, Expression trend of CLOCK, PER3, PER2, ARNTL, NPAS2, CRY2, RORA, CRY1, and NR1D1 across menstrual cycles. Data are presented as means ± SEM. * P < 0.05 and ** P < 0.01 represent MSE against PE; # P < 0.05 and ### P < 0.001 represent LSE against MSE.

Target Gene Network and Functional Enrichment of Differentially Expressed Circadian Clock Genes in MSE Versus PE

To gain better understanding of the human endometrial signaling pathways and molecular events controlled by core circadian clock genes during PE to MSE transition, we detected the differences in the expression of core circadian clock genes between PE and MSE as previously described (29). Using “limma” package, we found 1123 transcripts differentially expressed (676 upregulated and 447 downregulated) in the endometrium of MSE compared with that of PE (Fig. 2A). The expression of ARNTL, RORA, PER2, NPAS2, and CRY2 significantly increased in MSE compared with that in PE (Fig. 2B), although almost all core circadian clock genes remained unchanged between ESE and MSE.

Figure 2.

Target gene network and functional enrichment of differential expressed circadian clock genes in endometrium of normo-ovulatory women in MSE versus PE.

A, Heatmap of DEGs in endometrium of MSE versus PE. B, Expression of core circadian clock genes in endometrium of MSE versus PE. * P < 0.05, ** P < 0.01. C, WGCNA analysis of DEGs in endometrium of MSE versus PE. D-E, GO and KEGG enrichments of DEGs that co-expressed with differential expressed circadian clock genes in MSE versus PE. F, Potential target genes of PER2 in endometrium of MSE versus PE.

WGCNA was further applied to identify the co-expression and potential target genes of differentially expressed core circadian clock genes. Firstly, we found differentially expressed circadian clock genes (ARNTL, RORA, PER2, NPAS2 and CRY2) belonged to the turquoise and gray modules (Fig. 2C). Genes co-expressed with these 5 genes (in the same module) were identified and the biological process and pathways were enriched using GO and KEGG analyses, including “circadian rhythm,” “response to steroid hormone stimulus,” “fibroblast proliferation,” “cell junction organization,” “extracellular matrix interaction,” “cell cycle,” and other biological processes (Fig. 2D and 2E), which might contribute to endometrial function changes during PE to MSE. The network of the genes that were significantly co-expressed with the 5 differentially expressed circadian clock genes in WGCNA is shown in Fig. 2F. A total of 59 co-expressed genes were found, which might be considered as the target genes of PER2 and participate in the changes of endometrial function changes from PE to MSE (threshold = 0.7). No significant co-expressed genes of ARNTL, RORA, NPAS2, and CRY2 were identified when we set the threshold at 0.7.

Target Gene Network of Differentially Expressed Circadian Clock Genes in LSE Versus MSE

We further investigated DEGs and expression profiles of core circadian clock genes in MSE and LSE, and identified 586 differentially expressed transcripts (354 upregulated and 232 downregulated) (Fig. 3A). The expression of CLOCK, PER2, and PER3 all were significantly decreased in LSE versus that in MSE (Fig. 3B). Combined with the expression in MSE versus PE and in LSE versus MSE, PER2 was the only core circadian clock gene highly expressed in MSE (WOI) and low-expressed in LSE and PE. Thus, we supposed that PER2 was involved in the preparation of appropriate endometrial receptivity during WOI. WGCNA was also used to identify co-expression and potential target genes of PER2 that might participate in the changes from MSE to LSE. PER2 was included in the gray module, and the co-expressed genes as well as enriched pathways were detected (Fig. 3C). Immune response contributed to the biological functions from MSE to LSE, including “regulation of leukocyte proliferation,” “negative regulation of immune system process” and “Th cells differentiation.” In addition, “cell adhesion” and “response to metal ion” were also involved in this process (Fig. 3D and 3E). We identified 86 potential target genes of PER2 (Fig. 3F) in LSE versus MSE. Combining the target genes of PER2 in MSE versus PE and in LSE versus MSE, we finally determined 13 potential target genes of PER2 in endometrium that might be involved in the regulation of endometrial receptivity for WOI (Fig. 3G) (Table 1).

Figure 3.

Target gene network and functional enrichment of differential expressed circadian clock genes in endometrium of normo-ovulatory women in LSE versus MSE.

A, Heatmap of DEGs in endometrium of LSE versus MSE. B, Expression of core circadian clock genes in endometrium of LSE versus MSE. * P < 0.05 and *** P < 0.001. C, WGCNA analysis of DEGs in endometrium of LSE versus MSE. D-E, GO and KEGG enrichments of DEGs that co-expressed with differential expressed circadian clock genes in LSE versus MSE. F, Potential target genes of PER2 in endometrium of LSE versus MSE. G, Venn diagram showing the number of DEGs that co-expressed with PER2 both in MSE versus PE and in LSE versus MSE.

Table 1.

Potential target genes of PER2 involved in the endometrial receptivity of the window of implantation

Genes	Biological processes
FKBP5	peptidyl-prolyl cis-trans isomerase activity
MT1G	response to metal ions
STEAP4	ion transport/cupric reductase activity
MT1F	metal ion binding
KLF5	RNA polymerase II proximal promoter sequence-specific DNA binding
MT1X	metal ion binding
RXFP1	G protein-coupled receptor signaling pathway
MT1E	response to metal ions
MT1M	response to metal ions
SHTN1	substrate-dependent cell migration, cell extension
MT1HL1	metal ion binding
MT1H	response to metal ions
PLCL1	lipid metabolic process

Role of Core Circadian Clock Genes and Target Gene Network in Endometrial Receptivity and Pathogenesis of RIF

Since our results indicated the potential roles of core circadian clock genes, especially PER2, in the regulation of endometrial receptivity, we further analyzed the endometrial gene expression profile of WOI in RIF patients and controls (GSE111974) to further demonstrate the functional role of circadian clock genes, especially PER2, in the pathogenesis of RIF. The expressions of CLOCK, CRY1/2, NR1D1, and PER1/2 were differentially expressed in the endometrium of RIF patients (Fig. 4A). WGCNA was used to identify the co-expression and target genes of differentially expressed circadian clock genes. PER2 belonged to the blue module (Fig. 4B) and 42 potential target genes of PER2 were identified from this module (Fig. 4C).

Figure 4.

Target gene network of differential expressed circadian clock genes in endometrium of RIF patients versus controls.

A, Expression of core circadian clock genes in endometrium of RIF patients versus controls. * P < 0.05, ** P < 0.01, and *** P < 0.001. B, WGCNA analysis of DEGs in endometrium of RIF patients versus controls. C, Potential target genes of PER2 in endometrium of RIF patients versus controls.

Moreover, we combined the potential target genes of PER2 in MSE versus PE, in LSE versus MSE, and in RIF patients versus controls. Finally, we identified 5 PER2-regulated genes that were involved not only in the endometrial receptivity for WOI but also in the pathogenesis of RIF (Fig. 5A). These genes were metallothionein (MT) 1F, MT1X, MT1E, FKBP prolyl isomerase 5 (FKBP5), and relaxin family peptide receptor 1 (RXFP1). Spearman correlation analysis revealed strong positive correlations among these 5 genes and PER2 (P < 0.001) (Fig. 5B). These 5 genes also showed the same expression trend as PER2 did across the menstrual cycle, increased in MSE and decreased in PE and LSE (Fig. 5C).

Figure 5.

Target genes of PER2 that involved in the regulation of endometrial receptivity and RIF.

A, Venn diagram showing the number of DEGs that co-expressed with PER2 in MSE versus PE, in LSE versus MSE, and in RIF patients versus controls. B, Spearman correlations between PER2 and 5 target genes in endometrium of normo-ovulatory women. P < 0.001 was considered as significant. C, Expression of PER2 and 5 target genes in endometrium of normo-ovulatory women across the menstrual cycles.

Validation of the Target Genes of PER2 Involved in the Pathogenesis of RIF

We detected the expression of PER2 and its potential target genes involved in the regulation of endometrial receptivity to validate our data mining results. Clinical characteristics of recruited controls and RIF patients are demonstrated in Table 2. PER2 mRNA level was obviously increased in the endometrium of RIF patients during WOI when compared with controls (Fig. 6A), which was same as our in silico results. In the detection of target genes, we found that the mRNA expressions of shootin 1 (SHTN1), RXFP1, kruppel like factor 5 (KLF5) and six transmembrane epithelial antigen of the prostate 4 (STEAP4) were evidently increased in the endometrium of RIF patients (Fig. 6D-6G), displaying the same trend as PER2, without any changes in MT1E and FKBP5 mRNA expressions (Fig. 6B and 6C). We also verified the strong positive correlations between PER2 and KLF5, RXFP1, FKBP5, SHTN1, and STEAP4 (P < 0.001) (Fig. 6H). Moreover, we also did the immunohistochemistry to detect the protein expression of PER2 in both controls and RIF patients. The protein level of PER2 was also significantly increased in RIF patients versus controls during WOI (Fig. 6I).

Table 2.

Clinical characteristics of recruited controls and RIF patients

Clinical parameter	Control (n = 30)	RIF (n = 30)
Age (year)	28.74 ± 3.89	29.12 ± 3.84
Infertility duration (year)	3.66 ± 2.32	4.75 ± 3.17
BMI (kg/m2)	20.95 ± 3.57	21.63 ± 5.11
AMH (ng/mL)	5.13 ± 2.34	5.01 ± 3.02
Basal FSH (IU/L)	7.19 ± 2.08	6.87 ± 2.45
Basal LH (IU/L)	5.02 ± 2.61	4.89 ± 3.23
Basal E2 (pg/mL)	39.44 ± 19.27	37.27 ± 17.96
Antral follicle count	13.17 ± 3.80	12.81 ± 4.54
Endometrial thickness (mm)	9.23 ± 3.28	9.03 ± 3.35
P4 of WOI (ng/mL)	19.71 ± 3.55	21.27 ± 4.62

All data are presented as the mean ± SD values.

Abbreviations: AMH, anti-Müllerian hormone; BMI, body mass index; E₂, estradiol; FSH, follicle-stimulating hormone; LH, luteinizing hormone; P₄, progesterone; RIF, recurrent implantation failure; WOI, window of implantation.

Figure 6.

Validation of PER2 and target genes in the endometrium of RIF patients and controls.

A-G, mRNA levels of PER2, MT1E, FKBP5, SHTN1, RXFP1, KLF5, and STEAP4 in endometrium of RIF patients (n = 30) and controls (n = 30) detected by RT-qPCR analysis. All endometrial samples were in MSE (WOI). β-ACTIN was used as loading controls for RT-qPCR analysis. Data are presented as means ± SEM. * P < 0.05. H, Spearman correlations of PER2, MT1E, FKBP5, SHTN1, RXFP1, KLF5, and STEAP4 mRNA levels. P < 0.001 was considered as significant. I, Protein expression of PER2 in the endometrium of controls and RIF patients using immunohistochemistry. All the endometrium samples were obtained from WOI.

In order to confirm the regulation of PER2 to downstream target genes, we knocked down PER2 in hESCs. The apparently reduced expression of KLF5, SHTN1, and STEAP4 after PER2 siRNA treatment showed the positive regulation of PER2 on them without any changes in MT1E, RXFP1, and FKBP5 (Fig. 7B-7H). Thus, increased PER2 was supposed to reduce endometrial receptivity via facilitating KLF5, SHTN1, and STEAP4, which further promoted the pathogenesis of RIF. The expression of core circadian clock genes and potential target genes of PER2 was analyzed using Spearman correlation in 3 datasets: MSE versus PE, LSE versus MSE, and RIF versus control. PER2 was closely correlated to 11 other genes, including KLF5 and STAEP4 (P < 0.05 and ρ > 0.9) (Fig. 7I).

Figure 7.

Regulation of PER2 to potential target genes in hESCs.

A, Immunofluorescence of cytokeratin 18 and vimentin in primary hESCs. B-H, mRNA levels of PER2, KLF5, SHTN1, STEAP4, MT1E, RXFP1, and FKBP5 after PER2 siRNA treatment in hESCs. β-ACTIN was used as loading controls for RT-qPCR analysis. Data are presented as means ± SEM from 4 to 5 experiments. ** P < 0.01 and *** P < 0.001. I, Spearman correlations of PER2, potential target genes, and core circadian clock genes. P < 0.05 and ρ > 0.9.

Discussion

The development of endometrial receptivity driven by coordinated changes in ovarian production of estradiol (E₂) and progesterone (P₄) across the menstrual cycle is essential to initiating embryo implantation (30). Impaired endometrial receptivity and aberrant signaling pathways have been demonstrated to contribute to the pathogenesis of RIF, which has been studied mainly in women undergoing IVF and transfer of euploid or morphologically good embryos (31). Circadian clock genes play crucial roles in various aspects of female reproduction from ovulation to parturition (32-34). However, expression of endometrial circadian clock genes across the menstrual cycle as well as their roles in the endometrial receptivity in the WOI and the pathogenesis of RIF remain unknown.

Herein, we found expressions of core circadian clock genes significantly changed across the menstrual cycle with increased expressions of ARNTL and NPAS2 as well as decreased expressions of CLOCK and PER3 in SE versus PE. Notably, PER2 expression was markedly increased in ESE and MSE and dramatically decreased in PE and LSE, indicating the possible role of endometrial PER2 in the normal preparation of endometrial receptivity. A previous research study using pregnant Per2 promoter-dLuc transgenic rats demonstrated that Per2 oscillation is enhanced in endometrial stromal cells during implantation but is attenuated during decidualization (22). Both Per1^-/- and Per2^-/- female mice are reported to have more implantation sites but fewer live offspring compared with wild-type mice (35). We identified 13 potential target genes of PER2 in WOI using WCGNA, including FKBP5, MT1G, STEAP4, MT1F, KLF5, MT1X, RXFP1, MT1E, MT1M, SHTN1, MT1HL1, MT1H, and phospholipase C like 1 (inactive) (PLCL1). These potential target genes were significantly co-expressed with PER2 and displayed consistent trend of expression changes in MSE versus PE and in LSE versus MSE, providing a possible target network of gene interactions mediating the contribution of PER2 to endometrial receptivity.

Furthermore, we compared DEGs in endometrium from women with and without RIF in the WOI and uncovered the involvement of endometrial core circadian clock genes, especially PER2, in the pathogenesis of RIF. Similar to our results, a significant inverse correlation between midluteal PER2 transcript level and the number of previous miscarriages has previously been reported, which also strongly infers that dysregulation of PER2 increases the likelihood of persistent miscarriages (36). Moreover, we identified 5 target genes of PER2 possibly involved in the regulation of endometrial receptivity in RIF through in silico analysis, including MT1E, MT1F, MT1X, FKBP5, and RXFP1. However, no change was detected in the endometrial mRNA expression of MT1E and FKBP5 between RIF patients and controls in our validation cohort. Meanwhile, MT1E and FKBP5 mRNA expression maintained a constant level after PER2 knock-down in hESCs, suggesting that MT1E and FKBP5 might not participate in the pathogenesis of RIF.

Relaxin and its receptor RXFP1 play a role in the decidualization of endometrial stromal cells that is essential for pregnancy success (37). They are both upregulated in the peri-implantation period in the conception cycles in the marmoset monkey (38, 39). While relaxin and its receptor RXFP1 are involved in implantation, our in silico and lab-based study demonstrated the involvement of decreased RXFP1 in RIF but it was not under the regulation of PER2.

Besides of RXFP1 and PER2, we also identified and validated increased mRNA expression of KLF5, SHTN1, and STEAP4 in the endometrium of RIF patients versus controls. Aberrant KLF5 in the uterus impairs implantation and decidualization causing female infertility in mice (40). KLF5 is expressed dynamically in the luminal epithelium and stromal cells throughout pre-implantation process. Klf5^-/- mice display embryonic lethality due to developmental defects in the pre-implantation embryo. Therefore, although there was no significant change of KLF5 in our in silico analysis of RIF and control dataset, endometrial KLF5 is indeed essential for implantation. SHTN1a is considered as a brain-specific cytoplasmic protein and expression of SHTN1b has recently been identified in peripheral tissues (41). The current literature supports the critical role of SHTN1 in neuronal polarization and migration and axon differentiation (42, 43), and whether SHTN1 functions in reproduction and female fertility remains to be determined. STEAP4 is a metalloreductase involved in metabolism and cancer progression, and its expression is significantly decreased at the receptive phase (44). Altered STEAP4 contributes to the implantation failure in thin endometrium (45) as well as impaired endometrial receptivity due to mifepristone treatment (46). Thus, STEAP4 is a likely candidate involved in regulation of endometrial receptivity and implantation. In the current study, KLF5, SHTN1, and STEAP4 were identified as target genes of PER2 that appear to contribute to regulation of endometrial receptivity (Table 1). More importantly, PER2 positively regulated the expression of KLF5, SHTN1, and STEAP4 in hESCs. Hence, we concluded that increased PER2 might contribute to development of RIF via facilitating KLF5, SHTN1, and STEAP4. More detailed research studies on the possible role of the regulation of PER2 to target genes in endometrial receptivity are warranted.

It is well acknowledged that P₄ is crucial for decidualization and establishment of implantation. Blastocyst implantation and subsequent development might not take place if serum P₄ is below physiological value (47). However, we did not observe the difference of serum P₄ during WOI between RIF women and controls (Table 2). Thus, it might not be the decreased serum P₄ level that causes impaired endometrial receptivity, but rather the endometrium itself or its response to P₄ function. P₄ synchronizes the circadian oscillator of endometrial stromal cells via feedback loops of clockwork system (48). Animal models also show that high local concentration of P₄ facilitates implantation of xenogenic cultured cells in the uterus (49). Therefore, altered endometrial PER2 might result in impaired endometrial receptivity or changes of WOI phase due to aberrant P₄-mediated signaling pathway. The specific mechanism warrants further research.

Melatonin is a circadian-regulating hormone released from the pineal gland into the circulation. Clinical studies suggest that melatonin supplementation in IVF cycles might lead to better pregnancy rates by affecting implantation (50, 51). Emerging evidence has reported the in vivo and in vitro functions of melatonin on uterine physiological processes, such as decidualization and implantation (52). Thus, melatonin treatment may partly improve the endometrial receptivity and implantation in RIF. Since PER2 expression can be regulated by melatonin (53) and our results suggest a role for PER2 in RIF, further studies on safety and efficacy of melatonin as a potential target for clinic treatment of RIF are needed.

In summary, this is the first report, to our knowledge, of a dual in silico and lab-based approach to study the expressions of core circadian clock genes in endometrium across the menstrual cycle and the possible gene network in the endometrial receptivity as well as in the pathogenesis of RIF. Aberrant endometrial PER2 may contribute to impaired endometrial receptivity in RIF via regulating SHTN1, STEAP4, and KLF5. Although we investigated regulation networks using WGCNA and correlation analysis, the actual mechanisms and possible signaling pathways underlying the role of PER2 in endometrial receptivity of WOI as well as RIF await further investigation. Moreover, the therapeutic function of melatonin treatment in RIF targeting PER2 still needs to be validated.

Abbreviations

ARNTL

aryl hydrocarbon receptor nuclear translocator-like protein 1

CLOCK

circadian locomotor output cycles kaput

CRY1/2

cryptochrome 1/2 (gene)

DEG

differentially expressed gene

ESE

early secretory phase

GEO

Gene Expression Omnibus (database)

GO

Gene Ontology

hESCs

human endometrial stromal cells

ICSI

intracytoplasmic sperm injection

IVF

in vitro fertilization

KEGG

Kyoto Encyclopedia of Genes and Genomes

LSE

late secretory phase

MSE

middle secretory phase

NPAS2

neuronal PAS domain protein 2

NR1D1

nuclear receptor subfamily 1 group D member 1

PE

proliferative phase

PER1/2/3

period 1/2/3 (gene)

RIF

recurrent implantation failure

RORα

retinoic acid receptor-related orphan receptors α

RT-qPCR

real-time quantitative polymerase chain reaction

siRNA

small interfering RNA

WGCNA

weighted gene co-expression network analysis

WOI

window of implantation.

Additional Information

Disclosures: All authors declare no conflicts of interest.

Data Availability

The two public gene expression datasets (GSE4888 and GSE111974) can be downloaded from GEO database (https://www.ncbi.nlm.nih.gov/geo/). All other data presented in this study are included in the article/supplementary material.