Introduction

Appropriate communication between the fetus and endometrium is critical for healthy pregnancy outcomes [1, 2]. Infertility is an important social concern encountered by approximately 15% of couples during their reproductive age worldwide [3]. In vitro fertilization and embryo transfer (IVF–ET) have been developed in the last 40 years, providing healthy births to infertile couples. However, by improving the techniques to select good-quality sperms, oocytes, and fertilized embryos, only 30% of ETs succeed in having pregnancies [2]. Recurrent implantation failure (RIF) is one of the major issues in IVF–ET programs. Therefore, an inappropriate endometrial condition is believed to be a critical factor to cause RIF [1]; however, how the endometrial dysfunction causes RIF remains unclear.

Embryo implantation is a complex process of communication between blastocysts and endometria, molecularly and physically [1, 4]. Due to similarities in hormonal cycles and the manner of pregnancy maintenance, rodents are often used as a model to analyze embryo implantation processes as an alternative to human studies [1, 2, 4]. After coitus, ovaries provide increasing progesterone (P4) levels to prepare the endometria receptive to blastocysts [5]. With the influence of P4, the endometrial epithelium stops growing and begins to differentiate [6, 1]. Embryo implantation anomalies have been reported to severely affect pregnancy maintenance spanning species including humans [2]. Most of the studies using KO mice have shown that several single molecules influence the implantation and following pregnancy conditions, whereas their impacts in humans remain largely uncertain [1, 4]. The recent development of next-generation sequencing found that gene expressions are dynamically altered during menstrual cycles, possibly contributing to endometrial regenerations and implantation window opening [37, 38]. Although this beneficial information regarding menstrual tissues has already been reported, it remains unclear how gene expression signatures differ between endometria with successful and failed implantation. In our study, we dissected DEGs in human peri-implantation endometrium obtained from RIF and fertile patients. Enrichments of PRC2–H3K27me3-targeting genes were evident in DEGs, implying the possible roles of this axis to regulate genome-wide expressions during embryo implantation. Furthermore, the essential roles of Ezh2 in early pregnancy were identified by applying next-generation sequencing in Ezh2 uKO endometria. Previously, Nanjappa et al. and Fang et al. reported that Ezh2 uKO uteri exhibit abnormal epithelial integrities in their old ages; however, detailed mechanisms were not shown [27, 28]. They also compared gene expressions between the control and Ezh2 uKO under the treatment of exogenous P4 and E2, which might explain increased epithelial cell expansions in this milieu [39]. Recently, Osokine et al. investigated how Ezh2 contributes to decidual functions upon wound healing [40], focusing on post-implantation events after day 8 of implantation when deciduae were already terminally differentiated [48]. PRC1-induced H2AK119u1 can recruit PRC2 to promote H3K27 methylations [47]. As pharmaceutical inhibitions of either DNA methylations or PRC1 resulted in terminations of pregnancy in mice [49, 50], these epigenetic regulations may prepare uterine genomes for the following PRC2 binding. We also have not yet assessed how PRC2-induced H3K27me3 is canceled during the pregnancy progression. A report demonstrated that uterine tissues exhibit acetylation on lysines 4 and 27 of histone H3, the two major active histone markers, upon parturition [51]. This implies that some system works to determine pregnancy stages, switching epigenetic modifications. Future endeavors should examine how uterine cells respond to pregnancy conditions with appropriate gene expressions at each pregnancy condition.

Methods

Collection of human endometrial tissues in the peri-implantation period

Human endometrial tissues in the peri-implantation period were obtained from patients undergoing IVF–ET treatment aged under 40 years. Endometrial biopsy was performed as previously described [20]. The specimens obtained from those with uterine fibroids, adenomyosis, endometrial polyps, endometrial hyperplasia, and endometriosis were excluded from the study because these diseases have the possibility to affect endometrial receptivity. To minimize individual differences in hormonal status, the same protocol of hormonal replacement cycle for frozen ET was used for all patients in the cycle of endometrial biopsy. Endometrial biopsies were performed on day 7 of P4 administration during a hormonal replacement cycle, which is considered the peri-implantation period in humans. Patients underwent ET in subsequent cycles after the endometrial biopsy, and the outcome of clinical pregnancy was monitored. RIF patients were defined as those who had more than two failed embryo transfer cycles using good-quality embryos. The fertile controls were defined as patients who had clinical pregnancy in the subsequent cycle after the endometrial biopsy. Twelve and 26 independent endometrial samples were collected from patients with RIF and fertile controls, respectively. Table 1 demonstrated the information of the patients with RIF and the fertile controls. The study protocols using human specimens were approved by the institutional review board of the University of Tokyo (IRB numbers: 10991 and 2019241 G), and each woman signed informed consent for the use of tissues.

RNA-seq of the human endometrium

RNA extraction from human endometrial tissues was performed using the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). RNA-seq was performed at Macrogen Japan (Tokyo, Japan) on 38 specimens: 26 patients clinically pregnant as a result of embryo transfers after endometrial biopsy (fertile controls; successful implantation group) and 12 who did not become clinically pregnant (the RIF group; failed implantation group). The resulting raw read files were aligned on the human genome sequence (GRCh38/hg38) using the Hisat2 version 2.1.0 [52], and the number of reads at each locus was counted using the featureCount function [53] in the Subread tool. To compare expression levels between successful and failed implantation samples, read count files were submitted to DESeq2 version 1.16.1 [54]. DEGs were defined as genes showing >twofold difference in expression levels with a significant difference of P < 0.5 (P-values were adjusted for multiple testing using the Benjamini–Hochberg method). Moreover, these genes proceeded to enrichment analyses in Enrichr (https://maayanlab.cloud/Enrichr/) [62]. Table S1 shows qPCR primer sequences used to detect each gene.

Three-dimensional visualization of ISs

As previously reported, three-dimensional (3D) visualization on day 6 ISs was performed [9, 60]. Briefly, an anti-E-cadherin antibody (3195, Cell Signaling Technology, 1:300) was used to stain luminal and glandular epithelia. Alexa Fluor 555-conjugated anti-rabbit antibody (A21428, Thermo Fisher Scientific, 1:300) was used as a secondary antibody. The 3D images were acquired using LSM 800 (Zeiss, Oberkochen, Germany) and AXR (Nikon, Tokyo, Japan). The surface tool in Imaris (v 9.8, Oxford Instruments, Abingdon-on-Thames, UK) was used to construct 3D images.

Native ChIP

Day 6 uteri from the control and Ezh2 uKO were used. Uterine tissues surrounding the embryos were longitudinally opened at the mesometrial side and kept at −80 °C until use. Endometrial tissues were disrupted with 35 strokes in a Dounce homogenizer on ice, with a loose-fitting pestle in PBS-containing protease inhibitor cocktail and phosphatase inhibitors 2 and 3 (Sigma, St. Louis, MO, USA). After centrifugation with 1000g for 5 min, pellets were received in the nuclei EZ lysis buffer (Sigma) to isolate the nuclei. Native ChIP was then performed as previously described [63], with some modifications. For the fragmentation, chromatins were treated by 20 U/μl MNase at 37 °C for 5 min. Input DNA was analyzed with a 2100 Bioanalyzer system using High-Sensitivity DNA Reagent kit (Agilent, Santa Clara, CA, USA) to confirm DNA fragmentations at approximately 200–300 bps. Chromatin immunoprecipitation was performed using Magna ChIP G-Chromatin Immunoprecipitation Kit (Millipore, Burlington, MA, USA) following the manufacturer’s protocol. Anti-H3K27me3 antibody (39155, Active motif, Carlsbad, CA, USA) or anti-rabbit IgG (2729, Cell Signaling Technology) was used to precipitate the target chromatins. Chromatin-immunoprecipitated DNAs were purified by extracting phenol–chloroforms.

ChIP-seq

ChIP samples were subjected to ChIP-seq using Novogene Inc. service. The resulting raw read files were aligned on the mouse genome sequence (GRCm38/mm10) using bowtie2 version 2.3.3.1, and peak cells were made using MACS2 [64] with their default settings. Tag density plots and heatmaps in the vicinity were generated using ngs.plot version 2.47.1 [65]. BEDtools intersect [66] (version 2.26.0) was used to compute the numbers of H3K27me3 peaks within ±2 kb around the upregulated genes in Ezh2 uKO on day 6. Enrichment of H3K27me3 peaks around the target loci was then compared between the control and Ezh2 uKO with Mann–Whitney U test using R (4.0.2). For the analysis focusing on G2M-related cell cycle genes, human cell cycle genes previously reported [67] were converted into mouse gene names using biomaRt in R (4.0.2). To identify the enrichment of known motifs within H3K27me3 enriched loci, we used the HOMER [68] (version 4.9) function findMotifsGenome.pl with default parameters and a fragment size denoted by the argument -gain. To visualize H3K27me3 peaks, the IGVTools count function [69] (Broad Institute, Cambridge, MA) was used to create TDF files from the sorted BAM files. The TDF files were processed in the IGV browser [69] (Broad Institute) to show continuous tag counts over the target loci.

In vivo bromo-deoxyuridine (BrdU) incorporation assay

In vivo BrdU incorporation assay was performed as previously described [70]. Briefly, female mice were injected with BrdU (100 mg per kg body weight) at 1000 h on day 6 of pregnancy. Two hours later, they were sacrificed and the uteri were frozen immediately. Frozen sections (12 μm) were fixed in methanol for 10 min at room temperature and immersed in 2 N HCl for 20 min at 37 °C to denature DNA for immunohistochemical detection. BrdU-positive area was quantified using Image J (NIH).

Immunofluorescence

Frozen sections (12 μm) were used for immunofluorescence. BrdU (ab6326, Abcam, 1:250) and phospho-histone H3 (pHH3) (9701, Cell Signaling Technology, 1:300) antibodies were used as primary antibodies. Alexa Fluor 488-conjugated anti-rat immunoglobulin G (A11006, Thermo Fisher Scientific, 1:300) and Alexa Fluor 555-conjugated anti-rabbit immunoglobulin G (A21428, Thermo Fisher Scientific, 1:300) were used for signal detection with nuclear staining using 4,6-diamidino-2-phenylindole (Do**do, Kumamoto, Japan 1:500). Images were obtained using LSM 800 (Zeiss) and AXR (Nikon). The decidual area without pHH3 staining was quantified by Image J (NIH).

Senescence-associated β-galactosidase (SAβgal) staining

To evaluate the terminal differentiation of decidua, SAβgal activity staining was performed as previously described [35, 36]. To compare the intensity of SAβgal staining between the control and Ezh2 uKO, frozen sections from both genotypes were processed on the same slide. Sections were counterstained with eosin.

Statistical analysis

Statistical analyses were performed using two-tailed Student’s t-test in GraphPad Prism 9 (GraphPad Software, San Diego, CA); otherwise, they are described in detail in each experimental section.