This study was approved by the Institutional Review Board of the Center for Reproductive Medicine, Shandong University, China. Written consent was obtained from all participants.

All participants were recruited from Shandong University Affiliated Hospital for Reproductive Medicine. The criteria for enrolment in the RIF group were: 1) implantation failure of more than three serially transferred high-quality embryos or more than 10 embryos from multiple transfers; 2) a normal karyotype for both individuals in the couple; 3) a maternal age < 40 years; 4) no uterine abnormalities, spontaneous abortions, or autoimmune diseases; and 5) a regular menstrual cycle (21–35 days) with no steroid hormone use for more than 1 month. Participants were included in the control group if they had conceived within the first three high-quality embryo transfers or had a history of successful pregnancy before in vitro fertilization (IVF) or intracytoplasmic sperm injection and met inclusion criteria 2–5 for the RIF group. Nineteen RIF patients and 13 control IVF patients were recruited. The characteristics of the control group and the RIF patients are presented in Table 1 .

Endometrial biopsies were performed 5–7 days after ovulation during a natural menstrual cycle for patients and control subjects. Samples were immediately snap-frozen in liquid nitrogen and stored at -80°C for subsequent processing (RNA isolation and western blotting). Samples for immunohistochemistry (IHC) were fixed in 4% paraformaldehyde for 24 hours.

Total RNA was extracted from endometrial samples or telomerase-transformed human endometrial stromal cells (T-HESCs) using TRIzol reagent (TaKaRa, Dalian, China) following the manufacturer’s protocol. RNA was reverse transcribed into cDNA using a reverse transcription kit (TaKaRa). Gene expression levels were determined by quantitative (q) polymerase chain reaction (PCR) using SYBR Green chemistry (TaKaRa) and a LightCycler^® 480 instrument (Roche, Basel, Switzerland). Oligonucleotide-specific primer sequences were designed using the National Center for Biotechnology Information primer design tool (https://www.ncbi.nlm.nih.gov/tools/primer-blast/). The PCR primers were as follows: RARA forward, 5′-GGGCAAATACACTACGAACAACA-3′ and reverse, 5′-CTCCACAGTCTTAATGATGCACT-3′; CEBPB forward, 5′-CGACGAGTACAAGATCCGGC-3′ and reverse, 5′-TGCTTGAACAAGTTCCGCAG-3′; prolactin (PRL) forward, 5′-CATATTGCGATCCTGGAATGAG-3′ and reverse, 5′-GATGAACCTGGCTGACTATCA-3′; insulin-like growth factor-binding protein 1 (IGFBP1) forward, 5′-GGCACAGGAGACATCAGGAGAA-3′ and reverse, 5′-GATGAACCTGGCTGACTATCA-3′ and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) forward, 5′-GGAGCGAGATCCCTCCAAAAT-3′ and reverse, 5′- GGCTGTTGTCATACTTCTCATGG-3′. mRNA expression levels were normalized to GAPDH expression levels.

Total proteins were extracted from T-HESCs using a lysis buffer (Beyotime, Shanghai, China) supplemented with a protease inhibitor cocktail (1:100 dilution; CWBio, Beijing, China). All extracted proteins were heated at 100°C for 10 min and then stored at -80°C. The proteins were separated on a 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis system and transferred to polyvinylidene difluoride membranes (Millipore, Burlington, MA, USA). The membranes were blocked with 5% nonfat milk in Tris-buffered saline with Tween-20 and incubated with the following primary antibodies: anti-RARα (1:1,000 dilution; 62294, lot #1, Cell Signaling Technology, Danvers, MA, USA), anti-C/EBPβ (1:150 dilution; SC-7962, lot#12017, Santa Cruz Biotechnology, Dallas, TX, USA) and anti-GAPDH (1:5,000 dilution; SA00001-1/SA00001-2, lot#20000275/20000311, Proteintech, Wuhan, China). The membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (1:5,000 dilution, Proteintech). Western blotting was then performed, and the labeled protein bands were developed using HRP (Millipore). The intensity of the bands was determined using Image Lab software (Bio-Rad, Hercules, CA, USA).

Endometrial tissues were obtained from control IVF patients and RIF patients. Tissue samples were fixed in 4% paraformaldehyde for 24 hours, dehydrated, embedded in paraffin, and stored at -20°C. Tissue sections were heated at 37°C overnight and then de-waxed, hydrated through a graded alcohol series (100%, 95%, 75%), and washed with distilled water. After immersing the sections in an ethylenediaminetetraacetic acid solution and boiling for 15 min for antigen retrieval, they were incubated in 3% H₂O₂ for 15 min to block endogenous peroxidase activity and then blocked with bovine serum albumin for 1 hour. Tissue sections were incubated in a moist chamber overnight at 4°C with an anti-RARα primary antibody (1:50 dilution, 62294, lot #1, Cell Signaling Technology) or an anti-C/EBPβ primary antibody (1:50 dilution; SC-7962, lot#12017, Santa Cruz Biotechnology, Dallas, TX, USA). After incubation at room temperature for 20 min with a secondary antibody (ZSGB-bio, Beijing, China), the signal was developed using a diaminobenzidine kit (ZSGB-bio). Negative control is incubated with antibody diluent, without the primary antibody included, and then followed the same procedure above. Images were captured at the same magnification and under the same exposure conditions for each sample. Representative images of all staining intensities were captured at random. IHC results were quantified using the H-score method, in which the percentage of positive cells (0–100%) and staining intensity (0–3+) were considered. A final score was calculated on a continuous scale from 0 to 300 using the following formula: H-sore =1 × (% of cells 1+) + 2 × (% of cells 2+) + 3 × (% of cells 3+).

Immortal T-HESCs were cultured in phenol red-free Dulbecco’s modified Eagle’s medium (DMEM)/F12 containing glutamine (Gibco, Grand Island, NY, USA), 1% penicillin-streptomycin (HyClone, Logan, UT, USA), 1% NaHCO₃ and 10% dextran-coated charcoal-stripped fetal bovine serum (C-FBS; Biological Industries, Beit HaEmek, Israel) at 37°C and 5% CO₂. To induce in vitro decidualization, the cells were incubated for 4 days with differential medium consisting of phenol red-free DMEM/F12 with 2% C-FBS, 0.5 mM 8-bromo-adenosine-3′,5′-cyclic monophosphate (cAMP; Sigma-Aldrich, St. Louis, MO, USA) and 10^-6 M medroxyprogesterone-17-acetate (MPA, Sigma-Aldrich). The differential medium was changed every 48 hours. The PRL and IGFBP-1, which were classic decidual markers, were used to assess the differentiation status of HESCs in culture.

T-HESCs were seeded in antibiotic-free medium in six-well plates; after reaching 60% confluence, they were transfected with a small interfering RNA (siRNA) specific for RARA (si-RNA RARA, GenePharma, Shanghai, China) or with a negative control (NC) siRNA (GenePharma) for 24 hours using Lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA). All procedures were performed according to the manufacturer’s instructions. For the transfection of siRNA RARA, 40 pmol of siRNA and 3.5 µL of Lipofectamine 3000 were diluted with 100 µL of reduced-serum medium (Opti-MEM, Gibco) per mL. Both transfection mixtures were incubated for 15 min at room temperature and added dropwise to each well. After 6 hours, the transfected cells were washed with phosphate-buffered saline (PBS), and in vitro decidualization was induced as described above for 4 days. T-HESCs in the undifferentiated group were incubated with non-selective medium for 4 days. In a rescue experiment, an adenoviral vector engineered to overexpress CEBPB (Vigene, Jinan, China) was transduced into T-HESCs (volume ratio: 1:5,000) 24 hours after transfection with siRNA RARA to restore CEBPB expression. The cells were then incubated in differential or non-selective medium. Twenty-four hours after adenovirus transduction, the differential or non-selective medium was removed, and the cells were washed with PBS and incubated in differential medium for another 48 hours. The siRNA sequences were as follows: RARA, forward 5′-GGGUGAUCACGCUGAAGAUTT-3′ and reverse, 5′-AUCUUCAGCGUGAUCACCCTT-3′.

RARα expression plasmid (pET-RARA, Vigene, Jinan, China) were used for overexpression. For the transfection of pET-RARA, 2.5ug plasmid and 5 µL of Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) and 5ul P3000 were diluted with 100 µL of reduced-serum medium (Opti-MEM, Gibco) per ml. All procedures were performed according to the manufacturer’s instructions. The other steps of experiment were following the “siRNA transfection”.

The RARα binding site was predicted using the transcription factor binding profile database JASPAR (http://jaspar.genereg.net/). The 2200 nucleotides upstream of the coding sequence were selected as the CEBPB promoter region for input into JASPAR. The primers used for chromatin immunoprecipitation (ChIP)-PCR were designed using the tool at http://pcrsuite.cse.ucsc.edu/ (20). The sequences of the CEBPB (-2,009/-1781) primers were as follows: forward, 5′-AAGGTCAGGGAGGGTTTCC-3′ and reverse, 5′-CCAGCCACCATTATCCTAGC-3′. T-HESCs were exposed to differential medium when cell confluence reached 80%. After 4 days of incubation, ChIP was performed using an EZ-Magna ChIP™ A/G Chromatin Immunoprecipitation Kit (17-10086, Millipore) according to the manufacturer’s instructions. Undifferentiated T-HESCs were incubated in standard non-selective medium for 4 days; when they reached 80% confluence, they were used in ChIP experiments. In brief, T-HESCs were fixed with 1% formaldehyde at room temperature for 10 min. The cells were lysed with cell and nuclear lysis buffer, after which sonication was performed to shear the chromatin into 300–1,000 bp fragments. The chromatin fragments were immunoprecipitated with an anti-RARα antibody (Cell Signaling Technology) and IgG (Abcam, Cambridge, UK). Finally, the precipitates were collected and analyzed by ChIP-PCR and ChIP-qPCR. After 32 cycles of PCR, the amplification products were analyzed by 1.5% agarose gel electrophoresis, and the level of chromatin enrichment was quantified.

Human embryonic kidney (HEK293) cells were cultured in DMEM (Gibco) with 10% fetal bovine serum (Biological Industries) and 1% penicillin-streptomycin (HyClone) in a 37°C incubator with 5% CO₂.

Firefly luciferase reporter constructs were designed with the upstream -2,200 CEBPB promoter region (pGL3-CEBPB) and with a mutated predicted binding site (-2,009/-1,781) in the CEBPB promoter (pGL3-CEBPB-mutant). An RARα expression plasmid (pET-RARA, Vigene, Jinan, China) and the reporter plasmids (pGL3-CEBPB or pGL3-CEBPB-mutant) were used to transfect HEK293 cells at a 1:3 ratio with Lipofectamine 3000 (Invitrogen). All procedures were performed according to the manufacturer’s instructions. Forty-eight hours after transfection, luciferase activities were measured using a dual-luciferase assay (Promega). Firefly/Renilla fluorescence ratios were calculated to determine the role of RARα in the regulation of CEBPB promoter activity.

The data presented here are representative of three or more biological replicates. The data are presented as the mean ± standard error of the mean. Significant differences between two groups were analyzed by Student’s t-test, using Prism Version 7 (GraphPad, San Diego, CA, USA). P-values < 0.05 were considered to be statistically significant.

The basal characteristics of the control and RIF groups are listed in Table 1 . There were no significant differences in maternal age, body mass index, basal antral follicle count, basal follicle-stimulating hormone concentration, or luteinizing hormone concentration between the RIF and control groups. As shown in Figures 1A, B , RARA mRNA and RARα protein expression levels were significantly lower in the mid-luteal phase endometria of RIF patients than control participants (P = 0.010 and P =0.04). IHC analysis ( Figure 1C ) showed high levels of RARα protein localized in the nuclei of endometrial stromal cells. These results implied that RARα mainly plays its role in the nuclei of stromal cells in the mid-luteal phase. IHC analysis also showed that RARα protein levels were clearly decreased in the nuclei of stromal cells in the endometria of RIF patients compared with the endometria of control participants. The H-score of RARα in the nuclei of endometrial stromal cells was lower in the RIF group than the control group (P < 0.001).

To determine whether decreased RARα expression levels affected the decidualization process, we transfected decidualized T-HESCs with siRNA RARA or an NC siRNA. T-HESCs were treated with cAMP and MPA for 4 days to induce decidualization. The induction of decidualization for 4 days was successful. The PRL and IGFBP-1, two classic decidual marker, were detected, and the morphology change of T-HESCs were recorded ( Figures 2A , 3A ). The PRL and IGFBP-1mRNA level were significantly increased (P<0.01 and P<0.001). After 4-days induction, the expression of RARα were significantly decreased ( Figure 2B , P=0.009).

After induction, PRL and IGFBP1 mRNA levels were significantly decreased (P = 0.008 and P = 0.004, respectively) in the decidualized T-HESCs with RARα knocked down ( Figure 4A ). The consequent cellular morphology is shown in Figure 3B . Without RARα knockdown, the T-HESCs transformed into large, round decidual cells after stimulation with decidualization-inducing conditions; however, when RARα was knocked down, decidualization was inhibited and the T-HESCs remained fibroblast-like.

The RARα expression decreased after 4-days induction, however, RARα knockdown impaired decidualization. To further explore the role of RARα in decidualization, plasmid of RARA was used to overexpressing RARα. After induction, PRL mRNA levels were significantly increased (P=0.022) and IGFBP-1 mRNA levels were clearly higher in the decidualized T-HESCs with RARα overexpression ( Figure 4B ). Results of RARα knockdown and overexpression demonstrated that, although its expression decreased compared with the proliferative status, certain amount of RARα expression is crucial in decidualization.

RARα participates in numerous physiological processes by forming heterodimers with RXR and regulating the expression of a series of genes. RARα may therefore affect decidualization by regulating the expression of a determinant gene. We used the protein-protein network database STRING (https://string-db.org/) to predict protein interactions, using crucial genes for decidualization as the STRING inputs (12). As shown in Figure 5A , a potential interaction was found between RAR/RXR and CEBPB.

CEBPB is a crucial gene for embryo implantation and decidualization (21). CEBPB mRNA (P = 0.006) and C/EBPβ protein expression levels were significantly decreased in T-HESCs subjected to RARα knockdown and 4 days of in vitro decidualization ( Figures 5B, C ). This result implied that RARα might influenced decidualization via regulating C/EBPβ transcription.

To further determine the relationship between C/EBPβ and RARα during decidualization, we restored C/EBPβ expression in T-HESCs with RARα knocked down. C/EBPβ overexpression rescued the decidualization of these T-HESCs ( Figure 5D ). After C/EBPβ overexpression, the transcription levels of PRL and IGFBP-1 were increased in T-HESCs with RARα knocked down compared with control cells. The consequent morphology of C/EBPβ-overexpressing T-HESCs with RARα knocked down is shown in Figure 3C . Without C/EBPβ overexpression, the T-HESCs with RARα knocked down transformed into larger, rounder cells compared with those shown in Figure 3B , which were more decidual-like after stimulation with decidualization-inducing conditions.

To determine whether RARα regulates CEBPB directly, the CEBPB -2,200 nucleotide region was input into JASPAR for putative binding site prediction ( Figure 6A ). The results indicated a potential RARα/RXRα binding site in the region of CEBPB (-2,009/-1,993). To confirm whether RARα was capable of binding to this region, we performed ChIP assays on T-HESCs that had been cultured with or without cAMP and MPA for 4 days, using primers specific for CEBPB (-2,009/-1,781). PCR analysis of the ChIP precipitates showed that RARα bound to and enriched the predicted region of CEBPB from a lysate of decidualized T-HESCs ( Figure 6B ). To further calculate the fold enrichment of the CEBPB promoter region bound to the pulled-down chromatin, ChIP-qPCR was performed, and the results were normalized to the amount of input DNA. As Figure 6B shows, compared with IgG alone, an anti-RARα antibody significantly enriched the CEBPB promoter region (P = 0.01) pulled down from the lysate of decidualized T-HESCs.

To further explore the relationship between RARα and CEBPB and to identify the stimulatory agent that influences CEBPB transcription, we detected changes in CEBPB mRNA levels in non-selective medium. In the absence of culture medium containing cAMP and MPA, the CEBPB mRNA expression levels did not change significantly (P = 0.171) after RARα knockdown ( Figure 6C ). In ChIP experiments ( Figure 6D ) without cAMP and MPA stimulation, the promoter region of CEBPB (-2,009/-1,781) was not pulled down by an anti-RARα antibody in undifferentiated T-HESCs (P = 0.109).

To further determine the binding site of RARα/RXRα, we used a plasmid directing the expression of RARα and the firefly luciferase reporter constructs pGL3-CEBPB and pGL3-CEBPB-mutant. A luciferase reporter assay was used to monitor the transcriptional activity of the 2,200 upstream nucleotides of the CEBPB promoter region and the CEBPB (-2,009/-1,781) mutant in HEK 293 cells overexpressing RARα protein. RARα protein was expressed in transfected cells, and transcriptional activity was then measured from a responsive reporter. As shown in Figure 6E , cells transfected with pGL3-CEBPB showed increased transcription of the luciferase gene compared with those transfected with pGL3-CEBPB-mutant(P<0.001).

To determine whether the expression of C/EBPβ was affected by decreased RARα levels in the mid-luteal phase endometria of patients with RIF, we measured C/EBPβ protein levels in endometrial samples of patients with RIF (the same samples used for RARα measurement). As presented in Figures 7A, B , C/EBPβ mRNA were significantly decreased (P=0.0024) and protein levels were clearly lower in patients with RIF than in control IVF patients. IHC analysis ( Figure 7C ) showed that C/EBPβ was mainly localized in the nuclei of endometrial stromal cells. C/EBPβ levels in stromal cell nuclei clearly decreased in the mid-luteal phase endometria of patients with RIF. The H-score of C/EBPβ in the nuclei of endometrial stromal cells was lower in the RIF group than the control group (P < 0.001).