Inflammatory proteomics of uterine fluid is potentially a non-invasive predictor for endometrial receptivity: a pilot study

This nested cohort study aimed to evaluate the potential of inflammatory protein detection in uterine fluid using OLINK technology as a non-invasive method for assessing the endometrial receptivity phase. We enrolled patients scheduled for frozen embryo transfer at the Reproductive Medicine Center of Xiangya Hospital. Inclusion criteria were regular menstrual cycles; female age ≥ 20 years; and body mass index (BMI) between 18 and 25 kg/m 2 . Exclusion criteria included polycystic ovarian syndrome (PCOS), severe hydrosalpinx, uterine adhesion, congenital reproductive tract malformations, chromosomal abnormalities, endometrial polyps, endometrial hyperplasia, untreated chronic endometritis, thin endometrium, submucosal myoma, intramural uterine myoma, endometriosis (stages III–IV), genital tuberculosis, or other severe systemic comorbidities (e.g., hypertension, diabetes mellitus, or malignant tumors). All patients underwent endometrial receptivity testing (ERT) under a hormone replacement therapy (HRT) cycle. Samples were collected 5 days after initiating progesterone supplementation (P + 5). Estradiol valerate was started at 4 mg/day and increased to 6 mg or higher until endometrial thickness reached > 7 mm. Progesterone was introduced after at least 12 days of estrogen priming, provided the endometrium measured > 7 mm. Progesterone supplementation consisted of 200 mg oral progesterone capsules (Qining, China) daily plus 600 mg vaginal progesterone capsules (Utrogestan, France). The first day of progesterone administration was designated P + 0, and sampling was performed on P + 5. Paired samples of uterine fluid (UF) and endometrial tissue were collected from 12 patients (24 samples total). After saline rinsing of the cervix, an embryo transfer catheter attached to a syringe was introduced into the uterine cavity, and gentle aspiration was applied to collect UF. The fluid was immediately placed in 500-µL normal saline (NS) and centrifuged to remove cellular debris, and the supernatant was stored at − 80 °C until proteomic analysis. Immediately afterward, an endometrial biopsy was obtained using a standard sampler. Each endometrial sample was divided into two portions: one was preserved in RNA stabilization solution for RNA sequencing and ERT modeling. RNA extraction, library preparation, and sequencing were conducted as described previously [ 15 ]. ERT results were interpreted using our established machine-learning model [ 15 ], which classifies endometrial status into three phases: pre-receptive, receptive, and post-receptive. The other portion was fixed in paraffin, sectioned, and stained with hematoxylin and eosin (H&E) for histological dating by pathologists according to Noyes’s criteria [ 5 ]. Two pathologists independently examined the H&E-stained sections, and a consensus dating was reported. In case of disagreement, a third senior pathologist made the final determination. Based on ERT and histological dating results, patients were categorized into either the window of implantation (WOI) group or the displaced WOI group. Differences in inflammatory protein expression in UF were compared between these groups. Finally, we integrated proteomic and transcriptomic data for bioinformatic analysis and developed a predictive model based on UF proteomics to identify the receptive phase. All samples were collected between March 2024 and June 2024. The study was approved by the Ethics Committee of Xiangya Hospital (Approval No. 2024003), and written informed consent was obtained from each participant. The overall study design is summarized in Fig.  1 . Fig. 1 The overall scheme of the study. NS, normal saline The overall scheme of the study. NS, normal saline Prior to the main cohort study, we conducted a preliminary experiment to assess the clinical feasibility and safety of detecting inflammatory proteins in UF using the OLINK platform and to determine the optimal dilution factor. We collected UF from three patients before blastocyst transfer in a frozen embryo transfer cycle. Each sample was diluted at three gradients: the initial UF was placed in 500-µL NS (first gradient) and centrifuged, and the supernatant was collected. This supernatant was further diluted at ratios of 1:5 (second gradient) and 1:10 (third gradient). Thus, three dilution samples per patient (nine total) were subjected to OLINK inflammation panel analysis. Reproductive outcomes were subsequently followed up. Results showed that 13 proteins (IL2, IL-2RB, IL4, IL-5, IL-10RA, IL13, IL-20, IL-22RA1, IL-24, TSLP, SLAMF1, FGF-5, and Beta-NGF) had missing data rates ≥ 88.9% (undetectable in at least 8 of 9 samples), and three proteins (IL10, NT-3, and FGF-21) had missing data rates ≥ 66.7% (undetectable in at least 6 of 9 samples). The remaining 76 proteins were detectable in at least six samples, with a missing data rate < 33.3%. The first dilution gradient was selected as optimal due to its lowest missing data rate. Follow-up revealed clinical pregnancy in two of the three patients. These preliminary results supported the safety and feasibility of UF proteomic profiling via OLINK. Accordingly, in the main study, UF was diluted in 500-µL NS for OLINK analysis. The design of the preliminary experiment is illustrated in Supplementary Fig.  1 . A full list of the 92 proteins included in the Olink inflammation panel is provided in Supplementary Table  1 . Inflammatory proteins in uterine fluid samples were quantified using the Olink Target-96 Inflammation panel (Olink Proteomics, Sweden), which simultaneously measures 92 inflammation-related proteins. All proteomic profiling was conducted by Lianchuan Bio Corp. The assay is based on the Proximity Extension Assay (PEA) technology. In this method, pairs of antibodies, each tagged with a unique DNA oligonucleotide, bind to the target protein. When the antibodies are in close proximity, the oligonucleotides hybridize and are extended by a DNA polymerase, creating a unique, protein-specific DNA barcode. This barcode is then amplified and quantified using real-time quantitative PCR (qPCR). The qPCR cycle threshold (Ct) value, defined as the cycle number at which the fluorescence signal crosses a predefined threshold, was recorded for each protein. Olink’s proprietary data processing pipeline converts these Ct values into Normalized Protein eXpression (NPX) values, an arbitrary log2-scale unit that facilitates relative quantification while minimizing technical variations. The calculation of NPX is detailed in Supplementary Equation S1 . Quality control and initial statistical analysis of the NPX data were performed using the “OlinkAnalyze” R package [ 25 ]. The complete dataset from the Olink inflammation panel is provided in Supplementary Table  2 . To construct a non-invasive classifier for endometrial receptivity, we developed a prediction model using the top five differentially expressed proteins (DEPs) identified in uterine fluid. A generalized linear model (GLM) was employed as the classifier. The model’s performance was evaluated using receiver operating characteristic (ROC) analysis, and its discriminatory power was reported as the area under the curve (AUC). Endometrial tissue transcriptome profiling was performed as previously described [ 15 ]. Briefly, total RNA was extracted from endometrial biopsies, followed by reverse transcription, amplification, and library preparation using standard commercial kits. Sequencing was conducted on an Illumina HiSeq 2500 platform, generating approximately 5 million raw reads per sample. The raw RNA sequencing data have been deposited in the Gene Expression Omnibus (GEO) database under the accession number GSE283724 ( https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE283724 ). The “DESeq2” (version 1.34.0) R Package [ 26 ] was used to screen the differentially expressed genes (DEG) between the WOI group and the displaced WOI groups. Raw gene counts were used as input. Genes with an absolute fold change > 2 and an adjusted P -value < 0.05 were considered differentially expressed genes (DEGs). Furthermore, gene set enrichment analysis (GSEA) was conducted using the “GSEABase” R package [ 27 ] to identify significantly enriched biological pathways. Considering the heterogeneity of clinical data, unpaired Wilcoxon tests were performed throughout this study unless otherwise specified. We presented the continuous variables as mean ± standard deviation (SD) and median (interquartile ranges). Significance was indicated by * P  < 0.05, ** P  < 0.01, *** P  < 0.001, and **** P  < 0.0001. All analyses were performed using the R software (version 4.4.1).

A total of 12 patients were included in the study. Histological evaluation via hematoxylin and eosin (H&E) staining confirmed that all endometrial samples were in the early to mid-secretory phase (Supplementary Fig.  2 ). The ERT results showed that seven patients were in the receptive phase and five patients were in the pre-receptive phase (Supplementary Fig.  3 ). Accordingly, seven patients were divided into the WOI group, and five patients were in the displaced WOI group. The characteristics of each patient are shown in Table  1 . Table 1 Characteristics of patients included in this study ID Age ERT result ERT model predicted optimal receptive timing Endometrial staging Group P01 24 Receptive phase P + 5 days + 16 h Secretory endometrium WOI P02 32 Pre-receptive phase P + 6 days + 4 h Early secretory endometrium Displaced WOI P03 40 Pre-receptive phase P + 6 days + 2 h Secretory endometrium Displaced WOI P04 28 Receptive phase P + 5 days + 5 h Early secretory endometrium WOI P05 35 Receptive phase P + 5 days + 7 h Secretory endometrium WOI P06 38 Receptive phase P + 5 days + 9 h Early secretory endometrium WOI P07 29 Receptive phase P + 5 days + 8 h Secretory endometrium WOI P08 38 Receptive phase P + 5 days + 9 h Secretory endometrium WOI P09 47 Pre-receptive phase P + 5 days + 21 h Early secretory endometrium Displaced WOI P10 37 Receptive phase P + 5 days + 5 h Secretory endometrium WOI P11 36 Pre-receptive phase P + 6 days + 4 h Secretory endometrium Displaced WOI P12 31 Pre-receptive phase P + 5 days + 21 h Proliferative to early secretory endometrium Displaced WOI Note: ERT , endometrial receptivity testing; P , progesterone; WOI , window of implantation Characteristics of patients included in this study Note: ERT , endometrial receptivity testing; P , progesterone; WOI , window of implantation The results revealed a distinct inflammatory protein expression profile in the uterine fluid between the two groups (Fig.  2 ). Specifically, the displaced WOI group exhibited elevated levels of multiple inflammatory factors (Fig.  2 A). A heatmap visualizing the NPX values of the differentially expressed proteins further illustrated the segregation between the groups (Fig.  2 B). Among the most significantly altered proteins, the top ten upregulated in the displaced WOI group compared to the WOI group were CXCL10, IL-17C, CST5, IL-10RB, VEGFA, LIFR, MMP-10, CDCP1, IL-18R1, and TNFB (Fig.  2 C). Fig. 2 Inflammatory proteins from uterine fluid were differentially expressed between the WOI and displaced WOI groups. A Volcano plot showing the differentially expressed proteins between the WOI and displaced WOI groups. B Heatmap showing the differentially expressed proteins between the WOI and displaced WOI groups. C Boxplot showing the top ten differentially expressed proteins. WOI, window of implantation Inflammatory proteins from uterine fluid were differentially expressed between the WOI and displaced WOI groups. A Volcano plot showing the differentially expressed proteins between the WOI and displaced WOI groups. B Heatmap showing the differentially expressed proteins between the WOI and displaced WOI groups. C Boxplot showing the top ten differentially expressed proteins. WOI, window of implantation We constructed a predictive model for endometrial receptivity using the top five differentially expressed proteins (DEPs). A combined model incorporating all five DEPs (CXCL10, IL-17C, CST5, IL-10RB, and VEGFA) achieved a discriminative performance, with an area under the curve (AUC) of 1.0 (Fig.  3 ). Furthermore, the predictive models based on each individual protein also showed strong performance, with AUC values of 0.886 for CXCL10, 0.886 for IL-17C, 0.857 for CST5, 0.857 for IL-10RB, and 0.857 for VEGFA. Fig. 3 ROC showing the AUC of the predictive model using the top five differentially expressed proteins. ROC, receiver operator curve; AUC, area under the curve ROC showing the AUC of the predictive model using the top five differentially expressed proteins. ROC, receiver operator curve; AUC, area under the curve Analysis of endometrial tissue transcriptomes revealed 677 significantly upregulated and 521 down-regulated genes in the displaced WOI group compared to the WOI group (Fig.  4 A). Among these, two genes—CCL20 and LIFR—were consistently differentially expressed at both the transcript level in endometrial tissue and the protein level in uterine fluid (Fig.  4 B). A full list of differentially expressed genes (DEGs) is provided in Supplementary Table  3 . Gene set enrichment analysis (GSEA) further indicated that the displaced WOI group exhibited significant enrichment in immune-related biological processes. These included adaptive immune response, B cell-mediated immunity, immunoglobulin-mediated immune response, and regulation of immune effector processes (Fig.  4 C). Notably, these immune processes were down-regulated in the WOI group relative to the displaced WOI group (Fig.  4 D). Complete GSEA results for biological processes are available in Supplementary Table  4 . Fig. 4 Transcriptomic and proteomic characteristics of endometrium during WOI by combined analysis of transcriptomic and proteomic data. A Volcano plot showing the differentially expressed genes of the endometrium between WOI and displaced WOI groups. B The intersection of DEGs of endometrium and DEPs of uterine fluid. C Top 15 enriched biological processes by GSEA. D GSEA plot showing the enriched immunological biological process of WOI and displaced WOI groups. WOI, window of implantation; DEGs, differentially expressed genes; DEPs, differentially expressed proteins; GSEA, gene set enrichment analysis Transcriptomic and proteomic characteristics of endometrium during WOI by combined analysis of transcriptomic and proteomic data. A Volcano plot showing the differentially expressed genes of the endometrium between WOI and displaced WOI groups. B The intersection of DEGs of endometrium and DEPs of uterine fluid. C Top 15 enriched biological processes by GSEA. D GSEA plot showing the enriched immunological biological process of WOI and displaced WOI groups. WOI, window of implantation; DEGs, differentially expressed genes; DEPs, differentially expressed proteins; GSEA, gene set enrichment analysis

Embryo implantation involves a finely tuned maternal immune response to the semi-allogeneic embryo. While it is well established that immune and inflammatory processes are critical to successful implantation, the specific characteristics and regulatory mechanisms of the endometrial immune microenvironment during the window of implantation remain poorly understood. Elucidating the expression patterns of immune and inflammatory factors across different endometrial receptivity stages will provide valuable insight into the immune landscape that supports embryo attachment and implantation. Our findings reveal a distinct inflammatory protein signature in uterine fluid, with significant differences observed between the WOI and displaced WOI groups. Specifically, the displaced WOI group exhibited upregulation of multiple inflammatory factors (Fig.  2 A). Among the differentially expressed proteins, the top ten—CXCL10, IL-17C, CST5, IL-10RB, VEGFA, LIFR, MMP10, CDCP1, IL18R1, and TNFB—were all decreased in the WOI group relative to the displaced WOI group (Fig.  2 C). Transcriptomic analysis of endometrial tissue, however, showed that CST5 mRNA was not detectable and TNFB was expressed at low levels in the endometrium, whereas the remaining eight corresponding genes were readily detected (Supplementary Table  3 , baseMean column). This suggests that CST5 and TNFB proteins in uterine fluid may originate from non-endometrial sources. Several of these proteins have established roles in endometrial function and embryo implantation. CXCL10, for instance, is implicated in immune cell recruitment to the endometrium during the implantation period in pigs [ 28 ], and its regulation by uterine fluid-derived Let-7b influences uterine receptivity in goats [ 29 ]. In humans, CXCL10 contributes to trophoblast adhesion and apposition in early pregnancy [ 30 , 31 ]. MMP10 is differentially expressed between pre-receptive and receptive endometrium [ 32 ] and participates in human decidualization [ 33 ]. LIF and its receptor signaling are critical for the acquisition of uterine receptivity and embryo implantation [ 34 ]. VEGFA supports endometrial receptivity and mediates embryo–endometrial crosstalk [ 35 ], while IL-17 and IL-18 have also been linked to uterine receptivity and implantation processes [ 36 , 37 ]. A predictive model constructed from the top five differentially expressed proteins successfully classified endometrial receptivity phases (Fig.  3 ). These findings demonstrate that inflammatory protein detection in uterine fluid using Olink technology is feasible and represents a promising non-invasive approach for defining the endometrial receptivity window. Gene set enrichment analysis (GSEA) of endometrial transcriptomic data revealed that differentially expressed genes between receptivity phases were predominantly enriched in immune-related processes. These included adaptive immune response, B cell-mediated immunity, immunoglobulin-mediated immune response, and regulation of immune effector processes (Fig.  4 C). Notably, the expression of immune-related gene sets was significantly lower in the WOI group compared to the displaced WOI group. This suppression of inflammatory immune responses during the implantation window may represent an adaptive mechanism to facilitate implantation of the semi-allogeneic embryo. The current study offers several advantages over previous approaches. (1) Non-invasive and convenient sampling: Unlike traditional invasive endometrial biopsy, uterine fluid collection can be performed during the same cycle as embryo transfer. Previous studies, including our own, have confirmed the safety of this procedure prior to embryo transfer [ 24 , 38 , 39 ]. Nevertheless, larger and more rigorous prospective studies are warranted to further validate its safety and confirm the absence of adverse effects on endometrial integrity or subsequent reproductive outcomes. (2) Direct representation of the uterine microenvironment: Compared to other non-invasive methods such as peripheral blood sampling, uterine fluid more directly reflects the local uterine environment, potentially offering greater biological relevance while minimizing interference from systemic physiological or pathological conditions. (3) Advantages of the OLINK platform: Based on Proximity Extension Assay (PEA) technology, the OLINK platform enables simultaneous quantification of up to 96 proteins with high specificity and sensitivity. The assay requires only 1 µL of sample, can be completed within 24 h, and does not require technical replicates. These characteristics provide distinct advantages over other cytokine detection platforms such as Luminex and Meso Scale Discovery, making OLINK particularly suitable for analyzing low-volume samples like uterine fluid. However, several limitations of this study should be acknowledged: (1) This is a pilot investigation with a relatively small sample size; (2) the molecular mechanisms through which the identified inflammatory factors influence embryo implantation remain unexplored; (3) the absence of matched blood samples analyzed with the same OLINK panel limits our ability to determine whether the detected proteins are locally secreted or systemically derived; (4) the analysis was restricted to 92 inflammatory proteins, omitting other potentially relevant factors. While the panel included known receptivity-associated proteins such as LIF and LIFR [ 34 ], other important mediators may have been overlooked. The recently introduced Olink Reveal panel ( https://olink.com/products/olink-reveal ), which quantifies > 1000 proteins per sample, could provide more comprehensive proteome coverage in future studies. This expanded panel would enable detection of additional key proteins implicated in endometrial receptivity and implantation—such as MUC1 [ 40 – 42 ] and FOXO1 [ 43 – 45 ]—that are not included in the current inflammation panel, thereby offering deeper insights into inflammatory processes during the peri-implantation period. Given the ethical considerations and practical challenges associated with performing this test prior to embryo transfer, we propose that it be integrated directly into the embryo transfer cycle rather than in a preceding cycle. This strategy would minimize potential inaccuracies caused by intercycle variability in endometrial receptivity. Furthermore, in light of growing evidence suggesting that natural cycles may yield better outcomes than hormone replacement therapy (HRT) cycles, we intend to include both cycle types in subsequent research. Comparing inflammatory protein expression in uterine fluid and its association with endometrial receptivity across natural and HRT cycles will enhance our understanding of their respective impacts on embryo implantation. Such comparative analysis will also support the development of individualized testing and treatment strategies tailored to specific patient conditions (e.g., natural versus HRT cycles). A rigorously designed prospective study with a larger sample size is necessary to establish and validate a non-invasive, precise predictive model for the endometrial receptivity phase. Future proteomic profiling of uterine fluid during embryo transfer cycles should be correlated with pregnancy outcomes to identify key inflammatory factors influencing implantation success.

Achieving a successful pregnancy depends on three key factors: a high-quality embryo, a receptive endometrium, and synchronized communication between the embryo and the endometrium [ 1 ]. Endometrial receptivity refers to a specific phase in the menstrual cycle—typically occurring between days 19 and 23—known as the window of implantation (WOI), during which the endometrium is primed to facilitate embryo implantation [ 2 ]. During this period, the endometrium undergoes transient changes in gene transcription and protein expression to enable embryo implantation. Impaired endometrial receptivity has been implicated in implantation failure and early pregnancy loss [ 3 ]. Displaced WOI leads to asynchrony between endometrium and embryo development, ultimately resulting in implantation failure or early pregnancy loss [ 4 ]. Current evidence suggests that advanced age, ovarian stimulation during assisted reproductive technology (ART) cycles, and recurrent implantation failure (RIF) are related to displaced WOI. Therefore, accurately identifying the receptive state of the endometrium is critical for improving ART outcomes. Methods for assessing endometrial receptivity primarily fall into four categories: histopathology examination, ultrastructural observation, imaging screening, and molecular detection. In 1975, Noyes et al. established morphological dating criteria that classify the endometrium into early-, mid-, and late proliferative phases, followed by early-, mid-, and late secretory phases [ 5 ]. The WOI generally opens in the middle-secretory phase. Although Noyes’ dating has been widely used as a histomorphological tool to evaluate receptivity [ 5 – 7 ], it is limited by subjectivity and inter-observer variability. Moreover, subtle structural changes are not reliably captured by this method. Subsequently, pinopodes—microprotrusions on the endometrial epithelium—were proposed as an ultrastructural marker of the WOI [ 8 ]. However, the specificity and clinical utility of pinopodes remain controversial. Some studies report that pinopodes persist throughout the luteal phase rather than being strictly confined to the WOI. For instance, Xu et al. found no significant differences in pinopode morphology or coverage between fertile and infertile women [ 9 ]. Another study reported discrepancies between receptivity status determined by RNA sequencing-based endometrial receptivity testing (rsERT) and pinopode evaluation [ 10 ]. Moreover, pinopode assessment is highly susceptible to technical variations: invasive biopsy can mechanically damage these delicate structures, leading to false negatives. Delayed fixation (beyond 30 min post-biopsy) or the use of inappropriate fixatives may cause structural collapse. Since pinopodes are unevenly distributed and evaluated subjectively under scanning electron microscopy (SEM), sampling and interpretation biases further limit their reliability. Imaging techniques such as ultrasound and MRI have also been used to evaluate endometrial receptivity [ 11 , 12 ]. Ultrasound is non-invasive and cost-effective, allowing assessment of endometrial thickness, pattern, and blood flow. However, it cannot provide molecular or genomic insights into endometrial status [ 2 ]. With advances in high-throughput sequencing, transcriptomic profiling has revealed significant gene expression changes across the menstrual cycle, offering a molecular characterization of endometrial receptivity [ 13 ]. In recent years, tools such as the endometrial receptivity array (ERA) or endometrial receptivity testing (ERT) have been developed to identify the WOI and guide personalized embryo transfer [ 14 , 15 ]. Additionally, a miRNA-based panel known as ER Map, utilizing RT-qPCR, has been introduced for receptivity assessment [ 16 , 17 ]. Notably, however, the efficacy of ERA/ERT in improving ART outcomes has been increasingly questioned. Several clinical trials and meta-analyses have reported no significant improvement in pregnancy or live birth rates with ERA-guided embryo transfer [ 18 – 22 ]. Most current methods for evaluating endometrial receptivity are invasive, requiring an endometrial biopsy, and cannot be performed during the same cycle as embryo transfer. Sequencing-based approaches are also relatively complex, time-consuming, and expensive. Hence, there is a clear need for non-invasive, economical, accurate, and rapid alternatives. Some researchers have explored non-invasive approaches—for example, Li [ 23 ] and Ibañez-Perez [ 24 ] identified associations between uterine fluid exosomal profiles and endometrial receptivity or implantation outcomes. However, exosome isolation remains technically challenging. Given that proteins are direct functional agents in biological processes, proteomic analysis may offer more precise insights into endometrial receptivity compared to transcriptomics. It is still unknown whether intrauterine fluid proteomics can reliably assess endometrial receptivity. Therefore, this pilot study aims to evaluate the feasibility of using OLINK proteomics to detect inflammatory proteins in uterine fluid as a non-invasive means of evaluating endometrial receptivity.

Below is the link to the electronic supplementary material. ESM 1 (DOCX 10.4 KB) ESM 2 (PNG 156 KB) ESM 3 (PNG 8.09 MB) ESM 4 (PNG 967 KB) ESM 5 (XLSX 37.5 KB) ESM 6 (XLSX 26.6 KB) ESM 7 (XLSX 1.35 MB) ESM 8 (XLSX 896 KB) (DOCX 10.4 KB) (PNG 156 KB) (PNG 8.09 MB) (PNG 967 KB) (XLSX 37.5 KB) (XLSX 26.6 KB) (XLSX 1.35 MB) (XLSX 896 KB)