Advanced maternal age as a risk factor for endometrial receptivity in repeated implantation failure: intersectional transcriptomic analysis and gene validation.

With the increasing trend of delayed childbearing, more women of advanced maternal age (AMA, >35 years) seek pregnancy through in vitro fertilization and embryo transfer (IVF–ET) ( Ubaldi et al., 2019 ). To date, the success rate of IVF–ET has reached approximately 60% ( Margalioth et al., 2006 ); however, 5%–10% of patients still experience repeated implantation failure (RIF) despite the transfer of morphologically high-quality embryos. There is no universally accepted definition of RIF. In this study, RIF was defined as the failure of at least two embryo transfer (ET) cycles, each involving the transfer of one or more morphologically high-quality embryos ( Busnelli et al., 2021 ). In women with RIF, implantation may repeatedly fail even when high-quality embryos are transferred, suggesting that endometrial factors and impaired embryo-endometrium crosstalk contribute significantly to implantation failure ( Sehring et al., 2022 ; Neves et al., 2019 ; Hiraoka et al., 2023 ; Norwitz et al., 2001 ). Shapiro et al. (2016) found that patients older than 35 years exhibited lower implantation and live birth rates in IVF cycles. Given the considerable physical and psychological burden associated with RIF, earlier diagnosis and intervention have increasingly been advocated in clinical practice ( Ma et al., 2022 ). The window of implantation (WOI), typically occurring between days 20 and 24 of a 28-day menstrual cycle, represents the period when the endometrium is optimally receptive to embryo implantation ( Achache and Revel, 2006 ; Kliman and Frankfurter, 2019 ). In IVF–ET cycles using hormone replacement therapy (HRT), the fifth day after the initiation of progesterone administration (P+5) generally corresponds to the WOI for blastocyst transfer ( Gomez et al., 2015 ; Mumusoglu et al., 2021 ). In patients with RIF, this critical period may be temporally displaced or molecularly dysregulated, thereby contributing to implantation failure. Abnormal endometrial receptivity (ER) has been implicated in up to two-thirds of RIF cases ( Guo et al., 2023 ). Emerging evidence suggests that AMA not only impairs ovarian reserve and embryonic competence but may also alter ER through endocrine, immune, and transcriptomic mechanisms ( Devesa-Peiro et al., 2022 ). Historically, Noyes et al. (2019) first introduced histological criteria for evaluating ER in the 1950s. Since then, transcriptomics profiling has enabled more precise molecular characterization of endometrial status. RNA sequencing combined with machine learning approaches has been increasingly applied to infer the WOI based on endometrial gene expression patterns, providing a temporal framework for characterizing receptivity-related molecular features rather than directly measuring ER itself ( Diaz-Gimeno et al., 2011 ; Ruiz-A and lonso, 2021 ; He, 2021 ). However, evidence regarding the clinical utility of receptivity-guided strategies remains inconsistent. While several studies have suggested no significant improvement in pregnancy outcomes among unselected populations with RIF ( Fodina et al., 2021 ; Saxtorph et al., 2020 ), others have reported improved implantation and live birth rates in patients with RIF ( Xu, 2025 ; Ohara et al., 2022 ). Alterations in ER are closely linked to dynamic changes in gene expression ( Bashiri et al., 2018 ). Beyond advancing our understanding of the implantation process, transcriptomic features associated with receptivity may also provide potential therapeutic targets for implantation failure ( Diaz-Gimeno et al., 2014 ). Although maternal aging is well known to impair oocyte quality and embryonic competence, its impact on ER remains incompletely understood. Therefore, the present study aimed to investigate whether advanced maternal age is associated with displacement of the WOI and explore potential transcriptomic characteristics of the endometrium in women with RIF.

This retrospective observational study analyzed clinical data from 254 women diagnosed with RIF who were treated at the Department of Reproductive Medicine in Nanjing Women and Children’s Healthcare Hospital from January 2020 to March 2024. Inclusion criteria were age between 22 and 44 years and a planned subsequent ET at our center. Exclusion criteria included (i) undergoing IVF–ET at an outside institution (n = 11) and (ii) transcriptomic analysis indicating a post-receptive WOI (n = 7), due to insufficient sample size for meaningful analysis. After applying these criteria, a total of 236 eligible participants were included in the study. Considering ER represents a dynamic temporal continuum and that a 1-day pre-receptive endometrium may reflect a transitional physiological state, we defined receptive and 1-day pre-receptive samples as normal ER, whereas 2-day pre-receptive samples were classified as abnormal ER ( Figure 1 ). Flow chart. The study was approved by the Ethics Committee of Nanjing Women and Children’s Healthcare Hospital (ethics number [2020] YL-009), and all participants signed the informed consent form. Endometrial biopsy and transcriptomic analysis were performed once for each participant. Clinical data were retrieved from the hospital information management system. This study was conducted in accordance with the Declaration of Helsinki. All patients received a standardized HRT regimen for endometrial preparation. In brief, beginning on the second day of the menstrual cycle, patients received 4 or 6 mg of estradiol valerate daily (Progynova, Bayer, France). The dosage was subsequently adjusted based on endometrial thickness (EMT) and serum estradiol (E2) and progesterone (P) levels, typically reaching 6–10 mg per day after 1 week. Luteal phase support (LPS) was initiated when the EMT reached 7 mm, the serum E2 level was ≥200 pg/ml, and the serum P level was <1.5 ng/ml (P+0). Endometrial biopsies were performed on P+5 using a disposable endometrial sampler (Jiangxi Nuode Medical Device Co., Ltd., Jiangxi, China). Biopsied tissues (≥ 8 mm 3 ) were transferred to cryotubes containing sample preservation solution (Yikon, China) and immediately stored at −20 °C for RNA sequencing. Total RNA was extracted from endometrial tissues and quantified using the Qubit RNA HS assay. cDNA synthesis and amplification were performed using a low-input RNA amplification kit, followed by library preparation and sequencing on the Illumina NextSeq 550 platform. ER status and WOI timing were assessed using a validated transcriptome-based machine learning model developed by Yikon Genomics in collaboration with Xiangya Hospital of Central South University. The model integrates gene expression signatures of receptivity-associated genes to classify samples as pre-receptive, receptive, or post-receptive. Transcriptomic profiles were mapped to a reference framework to estimate temporal displacement from the optimal WOI. Bioinformatic processing and normalization were performed using the ChromGo analysis platform. Read counts were normalized using the geometric median ratio method implemented in the DESeq2 R package (DESeq2_1.50.2). To evaluate transcriptomic heterogeneity and ensure data quality, unsupervised dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP) was performed on normalized gene expression data as an exploratory quality-control step. Sample-level clustering was used to identify potential outliers and transcriptionally distinct subgroups. This analysis revealed three clusters, including one small and clearly separated cluster. Based on predefined analytic criteria, downstream analyses were restricted to the largest cluster (n = 194), representing the dominant transcriptional structure of the dataset and serving as the primary analytic cohort. This step was performed prior to differential expression analysis and was not guided by clinical phenotypes or outcome variables. Differentially expressed genes (DEGs) were identified using the DESeq2 package with an adjusted p -value <0.05 and |log 2 fold change| ≥1. Functional enrichment analysis of DEGs was performed using the clusterProfiler package (clusterProfiler v4.18.1) to identify biological pathways associated with age and ER status. Differential expression analyses were conducted separately for maternal age groups and ER status, and overlapping genes between the two DEG sets were identified. Total RNA from endometrial samples was extracted using TRIzol reagent. Reverse transcription was performed to generate cDNA, and quantitative real-time PCR was conducted using SYBR Green chemistry on a ViiA™ 7 Real-Time PCR system. GAPDH served as the internal control. Primer sequences are provided in Supplementary Table S1 . Continuous variables were expressed as the mean ± standard deviation for normally distributed data and as the median (interquartile range) for non-normally distributed data. Normally distributed variables were compared using the independent samples t-test, whereas non-normally distributed variables were analyzed using the Mann–Whitney U test. Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate. Multivariate logistic regression analysis was performed to identify independent risk factors for impaired ER. Variables with statistical significance in univariate analysis ( p < 0.1) were initially screened. In addition, clinically relevant variables were also included in the multivariate logistic regression model. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to evaluate associations between maternal age and ER status. Differential expression analysis was performed using DESeq2, and p -values were adjusted using the Benjamini–Hochberg method to control the false discovery rate. A two-sided p < 0.05 was considered statistically significant.

A total of 236 women with RIF were included in the final analysis based on transcriptome-derived ER assessment. Among them, 50 (21.2%) exhibited a receptive endometrium at P+5. The remaining 186 patients showed displaced receptivity, including 136 (57.6%) classified as 1-day pre-receptive and 50 (21.2%) classified as 2-day pre-receptive ( Supplementary Figure S1 ). For clinical comparison, patients with receptive or 1-day pre-receptive endometrium were grouped as normal ER, while those with 2-day pre-receptive endometrium were classified as abnormal ER ( Table 1 ). Baseline of RIF patients with different receptivity status. a: Data are expressed as the mean ± SD for continuous variables following a normal distribution. b: Data are expressed as the median (first quartile, third quartile) for continuous variables not normally distributed. c: Data are expressed as the numbers (percentages) for categorical variables. * Fisher’s exact test. bFSH, basal follicle-stimulating hormone; bLH, basal luteinizing hormone; AMH, anti-Müllerian hormone; PCOS, polycystic ovarian syndrome. Women in the abnormal ER group were significantly older than those in the normal ER group (34.3 ± 3.1 vs. 33.0 ± 4.1 years, p = 0.016). No significant differences were observed in BMI, infertility type, baseline reproductive hormones (FSH, LH, and E2), AMH levels, parity, EMT on progesterone initiation day (P+0), or comorbid gynecological and immune conditions (all p > 0.05). Multivariate logistic regression analysis using ER displacement (2-day shift vs. non-shift) as the dependent variable identified maternal age as an independent predictor of ER displacement ( p = 0.027) ( Table 2 ). Multivariate logistic regression analysis. Whether the ERT is 2-day displaced; AMH, anti-Müllerian hormone; PCOS, polycystic ovarian syndrome; EMT, endometrial thickness. We further analyzed the first personalized embryo transfer (pET) cycles following ER assessment. A total of 51 women underwent pET with the transfer of at least one high-quality embryo, including 21 patients in the young group (≤30 years) and 30 in the aged group (>35 years). After adjustment of embryo quality and WOI timing, implantation outcomes were comparable between the two groups. The clinical pregnancy rate was 47.6% in the young group and 40.0% in the aged group ( p = 0.649), while the implantation rate was 34.3% and 27.9%, respectively ( p = 0.568) ( Supplementary Table S2 ). As described in the Methods, unsupervised dimensionality reduction using UMAP was performed as an exploratory quality-control assessment of the RNA-seq data. Three clusters were identified, including one small cluster that was clearly separated from the main sample population. According to the predefined analytic strategy, downstream analyses were performed using the largest and most transcriptionally homogeneous cluster (n = 194) ( Supplementary Figure S2 ). The age threshold of 35 years was selected because it is the conventional clinical definition of AMA and has been widely adopted. Recent population-based evidence further supports its relevance to differences in prenatal care utilization and perinatal outcomes ( Geiger et al., 2021 ). To achieve a clearer comparison between distinct age groups, women aged 31–35 years were excluded from the age-stratified differential expression analysis. Differential expression analysis was performed between the young group (≤30 years, n = 43) and the aged group (>35 years, n = 51) ( Figure 2A ). A total of 43 differentially expressed genes (DEGs) were identified in the aged group, including 34 upregulated and 9 downregulated genes (adjusted p 35 years, n = 51) women with RIF. (A) Schematic diagram of the analysis. (B) Volcano plot showing DEGs between the endometrial tissues of women in the young and aged groups (adjusted p -value <0.05 and |log 2 FC| ≥1). (C) Heatmap of the DEGs based on hierarchical clustering. (D) Gene Ontology enrichment analysis of DEGs based on the identified genes. (E) KEGG pathway enrichment analysis of DEGs based on the identified genes. (F) Cnetplot visualization depicting the relationship between DEGs and their associated enriched GO or KEGG terms. (G) Protein–protein interaction network constructed using the STRING database for DEGs. Hierarchical clustering based on these DEGs showed clear separation between the two age groups ( Figure 2C ). Pathway enrichment analysis may indicate significant involvement of immune-related pathways. KEGG analysis revealed enrichment in cytokine–cytokine receptor interaction, chemokine signaling pathway, viral protein interaction with cytokine and cytokine receptor, and alcoholic liver disease pathways ( Figure 2D ). Gene Ontology analysis showed that upregulated genes in the aged group were enriched in blood microparticles and immunoglobulin complexes, whereas downregulated genes were enriched in chemokine-mediated signaling, neutrophil chemotaxis, and cellular response to chemokine stimulus ( Figure 2E ). Network analysis further highlighted key regulatory relationships among these genes. Cnetplot analysis revealed central hub genes connecting multiple enriched pathways ( Figure 2F ), and STRING-based protein-protein interaction analysis demonstrated a tightly connected gene interaction module ( Figure 2G ). To investigate molecular features associated with abnormal receptivity, transcriptomic profiles of women with 2-day displaced ER (n = 43) were compared with those with normal ER (n = 43) ( Figure 3A ). Differential transcriptomic landscape of the endometrium in women with normal receptivity (n = 43) and 2-day displaced receptivity (n = 43). (A) Schematic diagram of the analysis. (B) Volcano plot showing DEGs between the endometrial tissues of women in 2-day displaced and normal receptivity groups (adjusted p -value <0.05 and |log 2 FC| ≥1). (C) Heatmap of the DEGs based on hierarchical clustering. (D) Gene Ontology enrichment analysis of DEGs based on the identified genes. (E) KEGG pathway enrichment analysis of DEGs based on the identified genes. (F) Cnetplot visualization depicting the relationship between DEGs and their associated enriched GO or KEGG terms. (G) Protein–protein interaction network constructed using the STRING database for DEGs. Differential expression analysis identified a distinct set of DEGs between the two groups ( Figure 3B ; Supplementary Data Sheet 2 ; Figure 3C ). KEGG enrichment analysis revealed significant involvement of pathways including cytokine–cytokine receptor interaction, ECM–receptor interaction, PI3K–AKT signaling, and viral protein interaction with cytokine and cytokine receptor ( Figure 3D ). GO analysis may indicate that downregulated genes in the displaced ER group were associated with cytokine–cytokine receptor interaction, cornified envelope formation, and ECM–receptor interaction, whereas upregulated genes were enriched in neuroactive ligand–receptor interaction, hormone signaling, and cAMP signaling pathways ( Figure 3E ). Network visualization using Cnetplot identified hub genes with strong pathway connectivity ( Figure 3F ), and STRING-based analysis demonstrated a distinct interaction network among the identified DEGs ( Figure 3G ). To identify genes potentially linking maternal age and altered ER, DEGs identified from the age and ER comparisons were intersected. This analysis identified 10 overlapping genes, namely, ATF3 , CXCL1 , CXCL2 , IGHG1 , NCCRP1 , PTHLH , CYP24A1 , MTCO1P40 , CHP2 , and SDK2 ( Figures 4A,B ). Overlapping transcriptomic landscape of the endometrium between age and receptivity of women with RIF. (A) Schematic diagram of the analysis. (B) Heatmap of the overlapping DEGs based on hierarchical clustering. (C) Gene Ontology enrichment analysis of overlapping DEGs based on the intersected gene set. (D) KEGG pathway enrichment analysis of overlapped DEGs based on the intersected gene set. (E) Protein–protein interaction network constructed using the STRING database for overlapped DEGs. (F) Validation of overlapped DEGs by qRT-PCR in an independent cohort (young, n = 5; aged, n = 5). Given the limited number of overlapping DEGs (n = 10), enrichment analyses should be considered exploratory and interpreted with caution. The qRT-PCR validation results should be interpreted cautiously due to the limited sample size. Functional annotation suggested that these genes were associated with immune-related biological processes, including cell killing and humoral immune response ( Figure 4C ). KEGG pathway analysis indicated enrichment in IL-17 signaling, TNF signaling, NF-κB signaling, and cytokine–cytokine receptor interaction pathways ( Figure 4D ). Protein–protein interaction analysis suggested potential interactions among CXCL1 , CXCL2 , and ATF3 ( Figure 4E ). To validate these findings, qRT-PCR was performed using an independent cohort of endometrial samples (n = 5 per group). MTCO1P40 was excluded from qRT-PCR validation because it is a pseudogene, and IGHG1 was not selected because its expression mainly represents endometrial immune-cell infiltration rather than intrinsic endometrial cellular responses. Consistent with RNA-seq results, ATF3 , CXCL1 , PTHLH , and NCCRP1 were significantly downregulated in the aged group ( Figure 4F ).

With the increasing trend of delayed childbearing, the proportion of women of AMA undergoing IVF–ET has grown significantly. Notably, a growing subset of these women experience RIF, highlighting the need to explore potential age-related impairments in ER ( Bastu et al., 2019 ; Garcia et al., 2018 ; Li et al., 2025 ; Attali and Yogev, 2021 ). Although previous studies have primarily focused on aging-associated defects in oocytes and embryos, the role of endometrial aging remains incompletely understood and controversial ( Pathare et al., 2023 ). Endometrial factors are estimated to account for about two-thirds of the causes of embryo implantation failure ( Craciunas et al., 2019 ). A randomized controlled trial has reported that ER-guided pET may be associated with improved implantation and live birth rates in women of advanced maternal age with RIF ( Barbakadze, 2024 ). In this study, integrating transcriptomic profiling with WOI-based receptivity assessment, we observed that women with RIF and AMA exhibited distinct endometrial transcriptomic features during the window of implantation. When comparing patients with or without marked ER displacement, women with abnormal receptivity were significantly older ( P = 0.016). Multivariate regression analysis further confirmed that age was independently associated with ER status (OR: 1.118, 95% CI: 1.013-1.234, P = 0.027). Consistent with our findings, previous studies have found that advanced-age women are more likely to exhibit WOI displacement ( Fujii and Oguchi, 2023 ; Zhao, 2023 ; Yaron et al., 1993 ). Transcriptomic analysis revealed distinct gene expression patterns between aged and younger groups. Functional enrichment analyses may indicate dysregulation of pathways related to cytokine signaling and immune regulation. In particular, genes involved in chemokine-mediated signaling, neutrophil chemotaxis, and cellular responses to chemokines were downregulated in the aged endometrium. These findings suggest that maternal aging may alter the endometrial immune microenvironment, potentially impairing appropriate responses to embryonic signals during implantation. Similarly, patients with displaced ER exhibited enrichment of immune-related pathways among downregulated genes. Immune regulation plays a pivotal role in regulating uterine receptivity ( Robertson et al., 2022 ). Such alterations may be linked with the establishment of a receptive micro-environment for embryo-endometrial communication. In the proteomic analysis of endometrial samples from women with RIF and controls, GO analysis likewise showed enrichment in terms related to the immune response ( Wang et al., 2021 ). Shi et al. (2018) reported that in transcriptomic profiling of endometrial samples from women with RIF during the WOI, the downregulated mRNAs were likewise enriched in the pathways “ cytokine–cytokine receptor interaction ” and “ ECM–receptor interaction. ” To explore shared molecular signatures between AMA and displaced ER, we identified 10 overlapping DEGs. Functional annotation of this limited gene set suggested involvement in immune-related processes, such as neutrophil chemotaxis and humoral responses. However, given the small number of genes, these enrichment results should be interpreted cautiously and considered exploratory. Notably, several of these genes have previously been implicated in implantation biology. Boespflug et al. found that ATF3 is a regulator of mouse neutrophil migration ( Boespflug et al., 2014 ). Cheng et al. (2017) also reported a crucial role for ATF3 in enhancing ER and embryo attachment via upregulation of leukemia inhibitory factor (LIF). CXCL1 , a small cytokine of the CXC chemokine family, is known to promote endothelial cell proliferation, migration, and tube formation, critical steps in angiogenesis ( Baston-Buest et al., 2017 ). Its role in decidual angiogenesis has been demonstrated both in vitro and in vivo ( Ma et al., 2021 ; Izumi et al., 2015 ; Baston-Bust et al., 2013 ), and its downregulation, along with CXCL2 , may impair immune cell recruitment and maternal vascular remodeling essential for implantation. Transcriptomic analyses have shown significant downregulation of CXCL1 and C4BPA in RIF patients with a thin endometrium ( Kurmanova et al., 2024 ). PTHLH , which is important in early pregnancy, when antagonized, induces fetoplacental growth restriction through mitochondrial apoptosis and decreased platelet endothelial cell adhesion molecule expression ( Thota et al., 2005 ). NCCRP1 , associated with cell proliferation, has been detected in the ER profile in recurrent miscarriage ( Craciunas, 2021 ). Such displacement may reflect inadequate or delayed decidualization. Validation in an independent cohort confirmed differential expression of four genes, providing preliminary support for their involvement in receptivity alterations. Several limitations should be acknowledged. First, this was a retrospective single-center study, and potential selection bias cannot be excluded. Second, the validation cohort was relatively small, and larger studies are needed to confirm these findings. Third, the observational design limits causal inference between AMA and ER displacement. In addition, the grouping strategy for ER may not fully capture the continuous nature of the WOI, and alternative classification strategies should be explored in future studies. Further mechanistic studies are required to clarify how maternal aging affects endometrial function at the molecular level. Our study identified several candidate genes associated with age-related transcriptomic alterations in the endometrium. These findings provide preliminary molecular evidence of immune-related transcriptomic changes in the endometrium of women with AMA and RIF. Further studies with larger cohorts and functional investigations are warranted to elucidate the biological roles of these candidate genes in endometrial aging and reproductive outcomes.