The study was approved by the ethics committee of our hospital (2013-408081-01). Written informed consent was signed by each participant for the use of their samples.

RIF was defined as the failure to attain a clinical pregnancy following at least three consecutive embryo transfers with a cumulative transfer of more than four high-quality cleavage-stage embryos or more than two high-quality blastocysts (Coughlan et al., 2014). The control group comprised patients who underwent assisted reproductive treatment due to male infertility and successfully conceived after the first embryo transfer.

Individuals were excluded if they had any of the following conditions: repeated pregnancy loss (two or more biochemical pregnancies or two or more abortions); a history of adverse pregnancy; or any clear cause of embryo implantation failure, including, but not limited to, moderate-to-severe intrauterine adhesions, a thin endometrium (less than 7 mm before transformation), adenomyosis, endometriosis, uterine fibroids (submucosal fibroids, non-submucosal fibroids larger than 4.0 cm and/or endometrial compression), untreated hydrosalpinx, reproductive tract malformations, severe immune diseases, severe coagulation abnormalities, endocrine system diseases; karyotype anomalies in one or both partners; contraindications to pregnancy or assisted reproductive technology; infectious diseases, sexually transmitted diseases, or mycoplasma and/or chlamydia.

In this study, we enrolled 26 patients with RIF and 35 patients as controls. As summarized in Table 1, all endometrial samples from the RIF group were in the secretory phase. Among the 35 control samples, 5 were in the proliferative phase and 30 in the secretory phase, of which 4 were utilized for the isolation of primary hESCs.

Metabolomics analysis was carried out on endometrial tissues from ten patients with RIF and ten fertile control patients. Endometrial tissue samples (100 mg) were homogenized in a high-throughput tissue grinder and suspended in pre-chilled 80% methanol. After placing in an ultrasonic cleaner for 10 min, the samples were centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatants were filtered through a 0.22 µm filter and then injected into a LC-MS/MS system for analysis (Novogene Co., Ltd., Beijing, China). Differentially expressed metabolites (DEMs) were identified through partial least squares discriminant analysis (PLS-DA) using metaX, with screening criteria set as variable importance in projection (VIP) > 1, fold change (FC) > 1.2 or FC < 0.833, and p < 0.05. The resulting DEMs were visualized using the ggplot2 package in R.

Total RNA was extracted from 10 mg endometrial tissues or cultured hESCs using TRIzol reagent (Vazyme, #R401-01) according to the manufacturer’s protocol. RNA integrity and concentration were assessed by agarose gel electrophoresis and spectrophotometry at 260/280 nm (One Drop, OD1000+). The extracted RNA was then reverse-transcribed into cDNA using the 5X All-In-One RT MasterMix (ABM, #ABS-G492). RT-qPCR was performed in a 20 µL reaction volume with ChamQ Universal SYBR qPCR Master Mix (Vazyme), using 18S rRNA as the endogenous control. The primer sequences are listed in Supplementary Table S1.

Fresh endometrial tissues were collected from fertile control women during the secretory phase. The endometrial tissues were minced and enzymatically digested with 0.1% (w/v) type I collagenase (Worthington, Freehold, NJ, United States) for 30 min at 37 °C. The stromal cells and glands were then separated by filtering the digested tissues through a 30 µm sieve. The isolated cells were resuspended in DMEM/F12 (Gibco, #10-092-CVRC) supplemented with 10% FBS (Gibco, #1645615) and 1% penicillin–streptomycin (Gibco, #SV30010), and then incubated at 37 °C with 5% CO₂ until confluent. Stromal cell purity was verified by immunofluorescence assay for vimentin and E-cadherin as described below. Cells were cryopreserved in liquid nitrogen upon reaching 80%–90% confluence. No more than three passages were used for any cell strain. The hESCs were seeded into 60 mm or 35 mm culture dishes. At 90% confluence, the medium was replaced with serum-free DMEM/F12. After an overnight starvation period, cells were switched to phenol red-free DMEM/F12 containing 2.5% charcoal-stripped FBS. The hESCs were induced for 72 h and subsequently harvested for later experiments.

The hESCs were incubated with 1 μM medroxyprogesterone acetate (MPA, Millipore Sigma, #M1629) and 0.5 mM 8Br-cAMP (Sigma #B7880) to generate an in vitro decidualization model, with medium replacement at 48 h. To determine the effect of estrogen (E2) and progesterone (P4) on LPAR1 and LPAR6 expression, hESCs were treated with E2 (10⁻⁸ M, Millipore Sigma, #E2758), P4 (10⁻⁶ M, Millipore Sigma, #P0130), E2+P4, and 8Br-cAMP + MPA in a time-dependent manner for 4 days. The cells and/or supernatants of hESCs were collected at different times for subsequent experiments.

To elucidate the influence of oleoyl-LPA on decidualization, hESCs were treated with oleoyl-L-alpha-lysophosphatidic acid (Selleck #E2992) at concentrations of 0.1 μM and 1 μM. The hESCs and/or supernatants from different groups were harvested for subsequent experiments following 72 h of induction.

The PRL protein level in supernatants of hESCs after different treatments was determined by ELISA kit (Elabscience, #E-EL-H0141).

The hESCs were prepared as described previously (Cai et al., 2022). Briefly, hESCs were fixed with 4% paraformaldehyde (Biosharp, #BL539A) for 20 min and permeabilized with 0.1% Triton X-100 (Sangon Biotech, 9002-93-1) at room temperature for 5 min. The hESCs were then blocked with 1% Triton X-100 containing 3% BSA for 30 min at 37 °C, incubated with primary antibodies (Supplementary Table S2) at 37 °C for 2 h, then with secondary antibody at 37 °C for 1 h. Nuclei were stained with DAPI. Cells were kept from light before being examined and digitally imaged with a fluorescence microscope and CCD camera.

Primary hESCs at passages no higher than three were seeded into 6-well plates at a density of 2 × 10⁵ cells per well. At 60% confluence, the medium was replaced with serum-free and pen/strep-free basal medium. Cells were transfected with small interfering RNAs (siRNAs) targeting LPAR6 (siLPAR6-2/3) or negative control siRNA using Lipofectamine 3000 reagent (Invitrogen). After 6–8 h, the transfection medium was replaced with fresh complete culture medium. Decidualization was induced 48 h post-transfection, and total RNA and protein were harvested 72 h after induction from the respective experimental groups. All siRNAs for LPAR6 were synthesized and purified by Gene Pharma Co., Ltd. (Shanghai, China). The siRNA sequences were as follows:

siLPAR6-2:

sense: 5′-GGUGUUUGUGCUUGGGUUATT-3′,

antisense: 5′-UAACCCAAGCACAAACACCTT-3’;

siLPAR6-3:

sense: 5′-GCAUAACCUACAGACCUUATT-3′,

antisense: 5′-UAAGGUCUGUAGGUUAUGCTT-3’.

Following resuscitation of cryopreserved 2^nd-passage cells in DMEM/F12 (Gibco, #10-092-CVRC) plus 10% FBS (Gibco, #1645615) and 1% penicillin-streptomycin (Gibco, #SV30010), they were seeded into 96-well plates at a density of 5 × 10³. The cell were placed in an incubator with 5% CO₂ at 37 °C for 24 h to ensure adequate cell adherence. After adhesion, the original medium was removed from the well, cells were transfected with siRNAs using Lipofectamine 3000 as mentioned before or treated with the following reagents: oleoyl-LPA, AM095 (4'-[3-methyl-4-((R)-1-phenyl-ethoxycarbonylamino)-isoxazol-5-yl]-biphenyl-4-yl-Na, MCE, HY-16040), and Ki16425 (3-(4-[4-([1-(2-chlorophenyl)ethoxy]carbonyl amino)-3-methyl-5-isoxazolyl] benzylsulfanyl) propanoic acid, Selleck #S1315) and incubated at 37 °C for 24 h. Lastly, CCK-8 reagent was added, the plates were incubated at RT for 20 min, and the absorbance at 450 nm was measured using a plate-reader.

Total protein was extracted as previously described (Cai et al., 2022). The proteins were separated by 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes. After blocking, the membranes were incubated overnight at 4 °C with primary antibodies (Supplementary Table S2), followed by thorough washing and incubation with the corresponding secondary antibody for 1 h at RT. The band density was determined using ChemiCapture software (Beijing Sage Creation Science Co.).

Endometrial tissues were obtained from endometrial biopsy and fixed in 4% PFA at room temperature. The tissues were dehydrated, embedded in paraffin and cut into 5 μm-thick sections. Antigen retrieval was conducted by autoclaving the samples at 121 °C for 15 min in the presence of a citrate antigen retrieval solution. Sections were blocked in 5% BSA, incubated overnight at 4 °C with primary antibody (Supplementary Table S2), washed and incubated with HRP-conjugated secondary antibody. The specific antibody signal was visualized by incubation with diaminobenzidine and counterstaining with hematoxylin. Images were captured using a Leica DM 2000 microscope.

Metabolomic data was processed with R (version 3.4.3). Graph Pad Prism 9.0 was used for statistical treatment of experimental data, and the quantitative data were presented as the mean ± SEM. Student’s t-test was used to assess significant differences between two groups, and ANOVA and Tukey’s test were used among multiple groups. The threshold for statistical significance was set at p < 0.05.

In both positive and negative ion modes, the PCA score plots of quality control (QC) samples (Figures 1A, 2A) showed tight clustering, indicating high instrumental stability during data acquisition. Based on the diagnostic criteria described in the Methods section, 39 DEMs were identified in positive ion mode and 15 in negative ion mode. PCA score plots revealed significantly distinct metabolic profiles between the RIF group and the control group (Figures 1B, 2B). The PLS-DA score plots further demonstrated clear separation and clustering patterns between the two groups (Figures 1C, 2C). Permutation test results (Figures 1D, 2D) confirmed the robustness of the PLS-DA model, with a permutation-derived variability of less than 0.25, indicating that the model was not overfitted and had strong predictive capability. Expression patterns of the DEMs in the RIF and control groups were visualized (Figures 1E, 2E) and summarized in Supplementary Table S3, showing that in positive ion mode, 27 metabolites were upregulated and 12 were downregulated, while in negative ion mode, 5 were upregulated and 10 were downregulated. Notably, lysophosphatidic acid (LPA) was identified as a differentially abundant metabolite in both ion modes (Figures 1F, 2F). In positive ion mode, LPA had a FC value of 0.56 and a p-value of 0.009; in negative ion mode, its FC value was 0.29 with a p-value of 0.032.

To further validate the reduced level of LPA observed in the RIF group, we examined the mRNA expression of key enzymes involved in LPA synthesis (ATX, PLA1, and PLA2) and degradation (PPAP2A, PPAP2B, and PPAP2C) in endometrial tissues from both control and RIF groups. The results showed that ATX and PLA1, genes associated with LPA synthesis, were slightly downregulated in the RIF group, although not significantly. In contrast, PPAP2A involved in LPA degradation was slightly upregulated without statistical significance, while PPAP2B and PPAP2C were significantly upregulated in RIF (Supplementary Figure S1, p < 0.05). These findings suggest that the significant decrease in LPA levels in RIF may result from both a moderate reduction in its synthesis and a pronounced enhancement of its degradation.

To characterize the expression profiles of LPA receptor (LPAR) family members (LPAR1-6), we analyzed human endometrial transcriptome data from the GEO database, selecting datasets with sample sizes ≥20 that included both women with RIF and healthy controls. The datasets GSE111974 (24 RIF and 24 controls) and GSE58144 (43 RIF and 72 controls) were retrieved to evaluate LPAR expression. Analysis revealed that LPAR3 was significantly upregulated and LPAR6 significantly downregulated in RIF compared with controls in GSE111974, while in GSE58144, LPAR2 was significantly upregulated and LPAR6 significantly downregulated (Figures 3A,B). We further validated these findings using RT-qPCR on endometrial tissues from 12 RIF patients and 12 fertile controls, which demonstrated significant downregulation of LPAR1, LPAR5, and LPAR6 in the RIF group (Figure 3C).

Given the potential indirect influence of male-factor infertility on endometrial receptivity, we additionally examined the expression of key endometrial functional markers—FOXO1, HAND2, HOXA10, and KI67—in both groups. Immunohistochemistry results indicated that HOXA10 was significantly downregulated (p < 0.05) and KI67 significantly upregulated (p < 0.01) in the RIF group compared to controls. FOXO1 and HAND2 also exhibited downward trends, although these changes were not statistically significant (p > 0.05) (Supplementary Figure S2A). Consistent with these observations, Western blot analysis showed marked downregulation of HAND2 (p = 0.002) and non-significant decreasing trends for HOXA10 (p = 0.057) and FOXO1 (p > 0.05) in the RIF group (Supplementary Figure S2B).

Endometrial stromal cells, as the primary cellular constituents of the endometrium, play essential roles in decidualization. To ensure high purity and minimal epithelial contamination, stromal cells were isolated from fresh endometrial tissues obtained from fertile control women. The isolated cells were characterized using flow cytometry and immunohistochemistry for the stromal marker vimentin and the epithelial marker cytokeratin. Results demonstrated a stromal cell purity of 98.54%, as indicated by the proportion of vimentin-positive/cytokeratin-negative cells (Supplementary Figure S3), confirming the high quality of the isolated cell population for subsequent experiments. Given the function of ESCs as receptors of LPA, the relative expression of LPAR1-6 during decidualization was measured in hESCs by RT-qPCR. LPAR1 and LPAR6 were highly expressed throughout the decidualization process (Figure 4A), and consequently, the effects of E2 and P4 on LPAR1 and LPAR6 regulation were determined. Primary cultured hESCs were treated with E2, P4, E2+P4, 8Br-cAMP + MPA, and cells were harvested at various time points (Figures 4B,C). In addition, the impact of LPA on decidualization was evaluated by incubating hESCs with a range of LPA concentrations. LPA treatment (1 μM) significantly increased the mRNA levels of IGFBP1 and PRL, and the PRL protein level (Figures 4D–F). F-actin staining demonstrated the anticipated cytoskeletal reorganization and morphological changes, consistent with the transition from a fibroblast-like to a decidual phenotype. Specifically, the cells changed from long, spindle-shaped forms to larger, more rounded shapes (Figure 4G). These results collectively demonstrated that LPA promoted the decidualization of hESCs at the mRNA, protein, and morphological levels. LPA had no discernible influence on cell proliferation (Figure 4H).

To determine whether LPAR1 or LPAR6 was involved in regulating decidualization, we inhibited LPAR1 in hESCs using pharmacological antagonists (AM095 or Ki16425) and knocked down LPAR6 expression with specific siRNAs (siLPAR6-2/3). Inhibition of LPAR1 did not significantly alter the mRNA levels of IGFBP1 and PRL, nor affect PRL protein secretion (Figures 5A–C). Furthermore, LPAR1 blockade had no observable effect on cytoskeletal organization, as assessed by F-actin staining (Figure 5D). In contrast, LPAR6 knockdown (validated by RT-qPCR; Figure 5E) markedly reduced IGFBP1 and PRL mRNA expression, decreased PRL protein secretion, and disrupted cytoskeletal morphology (Figures 5F–I). These results indicate that LPAR6, but not LPAR1, is critical for decidualization, influencing transcriptional, translational, and structural aspects of the process. Additionally, neither LPAR1 inhibition nor LPAR6 knockdown affected cell proliferation (Figure 5J).

Based on the above results, we determined the expression of LPAR6 protein in clinical samples from RIF and control groups and also made comparisons between the proliferative and secretory phases. Western blotting and immunohistochemical staining revealed that the protein expression of LPAR6 was lower in the endometrium of RIF patients compared with controls (Figures 6A,B). Moreover, the protein expression of LPAR6 was higher in the secretory phase than in the proliferative phase (Figure 6C). Further experiments in hESCs demonstrated that after the inhibition of LPAR6, LPA no longer had an effect on the mRNA levels of IGFBP1 and PRL, or on PRL protein level (Figures 6D–F). It is worth noting that the addition of LPA after inhibition of LPAR6 did not rescue the expression of IGFBP1 and PRL compared with the sole inhibition of LPAR6 (Figures 6D,E). These results together support our hypothesis that LPAR6 is involved in the LPA-mediated regulation of decidualization.

Next, we attempted to identify the downstream signaling pathways that could mediate decidualization in clinical samples and hESCs. There were no significant differences in protein levels of mTOR, P-mTOR, and PTEN between RIF and control groups, which may be due to the limited sample size (Figure 6G). The Western blotting results showed that LPA had no evident influence on the expression of P-mTOR and PTEN but substantially reduced the expression of mTOR (Figure 6H). The inhibition of LPAR6 also had no effect on PTEN expression, but it did unexpectedly upregulate the expression of mTOR and P-mTOR (Figure 6I).