Skip to main content

Can biomarkers identified from the uterine fluid transcriptome be used to establish a noninvasive endometrial receptivity prediction tool? A proof-of-concept study



Embryo implantation in a receptive endometrium is crucial for successful pregnancy. Endometrial receptivity (ER) prediction tools based on endometrial transcriptome biomarkers by endometrial biopsy have been used to guide successful embryo implantation in in vitro fertilization (IVF) patients. However, no reliable noninvasive ER prediction method has been established, and one is greatly needed. We aimed to identify biomarkers from uterine fluid transcriptomic sequencing data for establishing noninvasive ER prediction tool and to evaluate its clinical application potential in patients undergoing IVF.


The non-invasive RNA-seq based endometrial receptivity test (nirsERT) was established by analyzing transcriptomic profile of 144 uterine fluid specimens (LH + 5, LH + 7, and LH + 9) at three different receptive status from 48 IVF patients with normal ER in combination with random forest algorithm. Subsequently, 22 IVF patients who underwent frozen-thaw blastocyst transfer were recruited and analyzed the correlation between the predicted results of nirsERT and pregnancy outcomes.


A total of 864 ER-associated differentially expressed genes (DEGs) involved in biological processes associated with endometrium-embryo crosstalk, including protein binding, signal reception and transduction, biomacromolecule transport and cell-cell adherens junctions, were selected. Subsequently, a nirsERT model consisting of 87 markers and 3 hub genes was established using a random forest algorithm. 10-fold cross-validation resulted in a mean accuracy of 93.0%. A small cohort (n = 22) retrospective observation shows that 77.8% (14/18) of IVF patients predicted with a normal WOI had successful intrauterine pregnancies, while none of the 3 patients with a displaced WOI had successful pregnancies. One patient failed due to poor sequencing data quality.


NirsERT based on uterine fluid transcriptome biomarkers can predict the WOI period relatively accurately and may serve as a noninvasive, reliable and same cycle test for ER in reproductive clinics.

Trial registration

Chinese Clinical Trial Registry: ChiCTR-DDD-17013375. Registered 14 November 2017,


An ideal synchrony between the embryo and the receptive endometrium is necessary for successful implantation. The receptive period of the endometrium, referred to as the window of implantation (WOI), normally occurs on the 19th to 24th day of a 28-day cycle. Previous studies have demonstrated that the pregnancy rate would significantly reduce when implantation is not performed during the WOI [1, 2]. However, the optimal WOI lasts for less than 48 hours and varies widely between individuals [3]. Abnormal endometrium receptivity (ER), including WOI shifts and pathologic injury, has been observed in numerous patients with repeated implantation failure (RIF) [4,5,6]. Therefore, an approach for evaluating ER status is urgently needed, especially in the field of assisted reproductive technology (ART).

To fulfill this requirement, several methods have been proposed in recent decades, such as ultrasound examination [7,8,9], histologic analysis [10], and morphological markers [11,12,13]. However, none have proven to be an ideal predictor of endometrial receptivity. With advances in molecular biological technologies, our understanding of the molecular mechanism of embryo implantation has significantly improved. In 2011, a 238-gene endometrial receptivity array (ERA) using an RNA expression microarray was developed by Diaz-Gimeno et al. [14]. The ERA method is capable of identifying different stages of the endometrial cycle, which are known as the pre-receptive (PR), receptive (RE), and post-receptive (PO) stages. Although not independent confirmations, the method’s accuracy and reproducibility have been reported to be reliable in a series of studies [15, 16]. Recent studies have demonstrated that pregnancy outcomes of patients with displaced WOI and infertile couples with conventional IVF can be improved by personalized embryo transfer (pET) guided by the ERA test, even without a history of RIF [17, 18]. In addition, relevant results indicate that transcriptomic and proteomic markers serve as promising tools for ER assessment [19, 20]. Although numerous differentially expressed genes (DEGs) involved in endometrial receptivity have been identified by previous studies, the overlap between these results is rather limited. One explanation might be that sample sizes, individual differences and microarray platforms differ between studies. Next-generation, high-throughput RNA sequencing (RNA-seq) is another powerful tool for comprehensively analyzing the whole transcriptome. RNA-seq is better than a microarray in terms of dynamic range, background noise, and identifying different transcripts [21, 22]. Another limitation for current diagnostic tools of endometrial receptivity is rooted in the need for invasive tissue sampling by endometrial biopsy. The endometrial RNA expression profile could be altered due to the small injuries caused by invasive sampling [23]. In addition, local injury to the endometrium was reported to have a negative impact on implantation [24]; therefore, it is inappropriate to perform endometrial tissue sampling tests and guide implantation in the same active cycle. It is necessary to develop a noninvasive diagnostic tool to accurately predict the WOI.

Uterine fluids are an important medium of communication between the embryo and endometrium and include an admixture of endometrial secretions, plasma transudates, and oviductal fluid [25]. Uterine fluid contains extracellular vesicles, RNAs, DNAs, regulatory proteins, ions, lipids and other bioactive factors and plays an important role in embryo implantation [26]. Thus, the high-throughput sequencing of uterine fluid provides an opportunity to find noninvasive biomarkers of endometrial receptivity for clinical use. The aspiration of uterine fluid prior to embryo transfer does not affect the embryo implantation rate [27], suggesting the feasibility of developing a noninvasive diagnostic tool based on uterine fluid. However, few transcriptional studies have focused on endometrial receptive markers from uterine fluid. A previous study [28] identified 53 candidate genes predictive of endometrial receptivity by using microarray technology to analyze uterine fluid, clinical diagnostic tests have not been conducted.

The aim of our study was to investigate the feasibility of predicting ER with biomarkers from uterine fluid and to establish a noninvasive RNA-seq-based endometrium receptivity test (nirsERT) with the potential to be used in reproductive clinics.

Materials and methods

Study design

The main objective of this study was to establish a prediction tool for endometrial receptivity using transcriptome sequencing data and to evaluate the feasibility of a noninvasive endometrial receptivity test using uterine fluid specimens. First, from November 2017 to December 2018, participants were recruited for us to identify differentially expressed genes (DEGs) in pre-receptive, receptive and post-receptive uterine fluid by transcriptome sequencing and expression profile analysis and to build the nirsERT model applying a random forest (RF) machine learning algorithm. To limit interference from confounding variables affecting ER, the inclusion criteria for IVF patients were set as follows: 20–39 years of age; body mass index (BMI) = 18–25 kg/m2; patients with a history of a intrauterine pregnancy/pregnancies who underwent the first IVF cycle due to tubal factors alone or patients who undergoing the first IVF cycle due to male factors alone; a regular menstrual cycle length (25–35 days) with spontaneous ovulation; normal ovarian reserves (baseline FSH < 10 mIU/mL, antimullerian hormone > 1.5 ng/ml, and antral follicle count > 5); able to be followed up to assess the pregnancy outcome; and successful intrauterine pregnancy after the first embryo transfer (ET). Intrauterine pregnancy was defined as the presence of a gestational sac with or without fetal heart activity in the uterine cavity as evaluated by ultrasound 4–5 weeks after ET. To establish the prediction tool, normal ER status was defined as a successful intrauterine pregnancy.

Second, from January to April 2019, participants were recruited to demonstrate the accuracy of the nirsERT in predicting the WOI. The inclusion criteria for patients from which we collected uterine fluid on the day of cryothaw blastocyst transfer were as follows: 20–39 years of age; BMI = 18–25 kg/m2; ultrasound showing an endometrial thickness of ≥8 cm and an endogenous serum progesterone level of ≤1.2 ng/ml on the day of progesterone administration/LH peak; and transferred embryos with high-quality blastocysts (blastocysts ≥3 BB on Day 5 and Day 6, graded based on the Gardner system) [29].

The following exclusion criteria were applied: endometrial diseases (including intrauterine adhesions, endometrial polyps, endometritis, endometrial tuberculosis, endometrial hyperplasia, and a thin endometrium); hydrosalpinx without proximal tubal ligation; submucous myomas, intramural hysteromyomas, or adenomyomas protruding toward the uterine cavity; endometriosis (stages III–IV); uterine malformations; and other medical or surgical comorbidities identified by consulting medical records, physical examination, blood tests, B-ultrasound and X-ray examination.

In the validation group, all patients received the nirsERT and were followed up to 4–5 weeks after ET to determine intrauterine pregnancy by ultrasound. Subsequently, all patients diagnosed with an intrauterine pregnancy were followed up until delivery.

Ethics statement

The present study was conducted at the Center for Reproductive Medicine at Xiangya Hospital of Central South University with permission from the Ethics Committee of Reproductive Medicine. This study is registered with the Chinese Clinical Trial Registry (No. ChiCTR-DDD-17013375).

Uterine fluid collection, processing and transcriptome sequencing

All patients provided written informed consent before sample collection. For patients included in model construction, uterine fluid samples were collected at three time points in the natural cycle preceding the first IVF cycle. Ultrasound was initiated from day 10 of the menstrual cycle to monitor ovulation. Blood LH levels were dynamically measured daily when the follicle diameter was > 14 mm. Patients continued to undergo daily ultrasound monitoring of ovulation until follicular discharge. Uterine fluid was collected using an embryo transfer catheter (Cook Medical; America) on days 5, 7, and 9 (LH + 5, LH + 7, and LH + 9, respectively) after the LH surge (denoted as LH + 0). For patients in the model validation group, uterine fluid was collected on the day of blastocyst transfer before embryo transfer. (Transfers of frozen-thawed blastocysts were performed 7 days after the LH surge of the natural cycle/5 days after the progesterone supplementation of hormone replacement (HRT) cycles).

Sampling was performed as follows. The cervix was cleansed with saline before sampling. After the outer catheter of the embryo transfer catheter was inserted through the cervix to a depth of 4 cm from the external cervical os, the inner catheter was introduced into the uterine cavity to a point 1–2 cm from the uterine fundus to avoid contamination with cervical mucus. A 2.5 mL syringe was connected to the inner catheter, and suction was applied. The inner catheter was withdrawn within the external catheter before the external catheter was withdrawn from the uterus. Approximately 5–10 μL of uterine fluid obtained was immediately placed into 20 μL of RNA-later buffer (AM7020; Thermo Fisher Scientific, Waltham, MA, USA) for RNA stabilization, sealed, and cryopreserved at − 20 °C. Sequencing analysis was carried out within 7 days after sampling.

Total RNA was extracted by using an RNeasy Micro Kit (74,004; Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Quality control of RNA was performed with a Qubit HS RNA Kit (Q32855, Thermo Fisher Scientific, Waltham, MA, USA) and Agilent Bioanalyzer 2100 (G2939BA, Agilent Technologies, Santa Clara, CA, USA). Reverse transcription and library preparation were conducted using the MALBAC® Platinum single-cell RNA amplification kit and Transposon Library Prep Kit (KT110700796 and XY045, Yikon Genomics, Suzhou, China). Qualified libraries were sequenced using the Illumina HiSeq 2500 platform with a single-end read length of 140 bp. An average number of 5 million reads was generated for each library. Low-quality bases and adapters were filtered or trimmed by the Trimmomatic tool (version 0.33). Filtered reads were then mapped to the human reference genome (ensembl primary assembly, version GRCh37) using STAR [30]. The RNA expression level was normalized based on FPKM (fragments per kilobase million) of each gene by RNA-SeQC (version 1.1.8) [31]. Base-2 logarithmic transformation of FPKM was conducted for further analyses.

Detection of differentially expressed genes

Differentially expressed genes (DEGs) among different endometrial receptivity conditions were identified by analysis of variance (ANOVA). The equation is written as follows:

$${Y}_{gi j}={\mu}_g+{T}_{gi}+{S}_{gj}+{\varepsilon}_{gi j}$$

where μg represents the mean expression level of gene g; Tgi is the gene-specific treatment effect referring to the status of a natural cycle or hormone replacement therapy when uterine fluid was obtained, \({T}_{gi}\sim \left(0,{\sigma}_{T_g}^2\right)\); Sgj is the gene-specific endometrial receptivity stage effect with three levels (pre-receptivity, receptivity, and post-receptivity), \({S}_{gj}\sim \left(0,{\sigma}_{S_g}^2\right)\); and εgij is the gene-dependent residual error, \({\varepsilon}_{gij}\sim \left(0,{\sigma}_{\varepsilon_g}^2\right)\). An F-test was applied to statistically assess the equality of variances between Sj and εijk for each gene, showing whether each gene was differentially expressed among different endometrial receptivity stages. Because RNA-Seq analysis involves multiple statistical tests, the false discovery rate (FDR) was used to adjust the p-value (q-value) for statistical inference. A Gene Ontology (GO) annotation and functional analysis of these DEGs was conducted with the DAVID tool [32].

Co-expression network construction and visualization

Co-expression modules in the endometrial receptivity process were detected by weighted gene co-expression network analysis (WGCAN) [33]. Applying WGCNA, we then identified key modules significantly correlated with endometrial receptivity stages. Cytoscape software (version 3.8.1) was then used to visualize the interaction networks with different co-expression key modules [34].

Biomarker identification and performance validation

To identify biomarkers for predictive model construction, a post hoc Tukey HSD (honestly significant difference) test from an ANOVA was applied for pairwise comparisons of three receptive levels. Genes showing significant differences in all pairwise tests were detected to maximally distinguish each receptive stage. Expression values of these biomarkers were then inputted as features for the machine learning method-random forest to train the pattern on three ER conditions (pre-receptivity, receptivity, and post-receptivity). The most important features (gene expression) were further selected by R package random forest based on two measures (mean decrease accuracy and mean decrease gini). Mean accuracy, sensitivity, specificity, the positive predictive value and the negative predictive valuewere determined through 10-fold cross-validation.

Statistical analysis

Continuous data subject to a normal distribution were expressed as the mean ± SD and were compared using independent-samples t-tests. Continuous data subject to a skewed distribution were expressed as the median and interquartile range (IQR) and were compared using an independent-samples Mann-Whitney U test. Categorical data were expressed as counts and percentages and were determined to be statistically significant using a chi-square test or Fisher’s exact test. A two-sided P-value equal to or less than 0.05 was considered to be statistically significant. Statistical analysis was performed using IBM SPSS software (Version 23.0, IBM Corp.)



To establish the nirsERT model, we collected uterine fluid of three different receptive stages (pre-receptive, receptive and post-receptive) in the same cycle from IVF patients with normal WOI timing for RNA-seq. Sixty-nine participants were recruited, 21 patients who were not pregnant in the first embryo transfer cycle after the sampling cycle were excluded, and 48 patients with successful intrauterine pregnancies were studied to build the nirsERT model (Fig. 1). Baseline clinical characteristics are shown in supplementary Table S1.

Fig. 1
figure 1

Flow diagram of the establishment and application of the noninvasive RNA-seq-based endometrial receptivity test

Uterine fluid RNA extraction and sequencing

To perform transcriptome sequencing, we collected 144 uterine fluid specimens from 48 participants and extracted total RNA using a commercial kit. As expected, the yield of RNA was relatively low, ranging from 0 to 1160 ng with an average of 148 ng. Almost one-third of RNA samples were below the detection limit of the Qubit RNA HS assay kit (0.25 ng/μL). Normally, it is difficult to construct sequencing libraries with less than 1 ng of total RNA. To address this issue, we utilized a commercial kit for reverse transcription and amplification with a low amount of RNA.

We first validated the repeatability of transcriptome sequencing combined with the above mentioned kit (see supplementary methods). The Spearman correlation between different initial amounts of RNA was above 0.95, demonstrating the stability and repeatability of this method with at least 0.2 ng RNA (Supplementary Fig. S1). Then, we processed the 144 RNA samples according to the same protocol. As a result, 140 NGS libraries were successfully constructed and sequenced, generating an average of 5.5 million raw reads per library. All sequencing data of this study have been deposited into CNGB Sequence Archive (CNSA) of China National GeneBank DataBase (CNGBdb, with accession number CNP0002025. A total of 632 million high-quality reads, representing approximately 82.1% of raw data, were mapped to the human reference genome (ensembl primary assembly, version GRCh37). The number of mapped genes ranged from 9591 to 17,913 in each library.

DEGs detection and functional analysis

To identify differentially expressed genes (DEGs) among the pre-receptivity, receptivity, and post-receptivity stages, an ANOVA (analysis of variance) was applied to process the log2 transformed transcriptomic data. As a result, 864 DEGs were detected across the three different ER statuses. Notably, there were relatively more downregulated DEGs between the post-receptivity and receptivity stages (Fig. 2A). An unsupervised hierarchical clustering of the DEGs revealed three distinct groups. An GO analysis of these DEGs was conducted with the DAVID tool [32]. The DEGs were significantly enriched in 71 biological process (BP) terms, 38 cellular component (CC) terms and 25 molecular function (MF) terms. The top 1 enriched terms for each category were identified as signal transduction (GO:0007165), cytoplasm (GO:0005737), and protein binding (GO:0005515) (Table 1 and Fig. 2B).

Fig. 2
figure 2

Differential expression analysis and functional enrichment across endometrial receptivity conditions. A Volcano plot for RNA profile of PR, PO and RE samples. The cut-off is set as 0.05 for p-value and 2 for fold-change. B Bubble chart for functional enrichment analysis of DEGs. X-axis: fold of enrichment. Color of bubble: log10 (p-value). Size of bubble: number of genes enriched in the corresponding GO term

Table 1 GO enrichment analysis of DEGs from uterine fluid samples

To further investigate the functional module of DEGs in the uterine fluid samples, we used the weighted gene coexpression network analysis (WGCNA) algorithm to analyze transcription regulatory networks. As a result, 4 coexpression network modules, 3 of which were MEturquoise, MEyellow and MEblue modules, were found to be highly significantly correlated with the ER stage. Four hub genes, ECI2 (MEturquoise), ATP6V1B2 (MEyellow), CXCL16 (MEblue) and SELP (MEgray), were then identified based on the highest level of intramodular connectivity found in the four coexpression modules (Table 2). The MEturquoise module included most DEGs, representing 59.1% (511/864) of the DEGs. The analysis also shows the strongest correlation with the ER stage with a correlation value of − 0.7. A functional enrichment analysis shows that genes in the MEturquoise module are involved in transcription regulation, such as epigenic modification-related pathways. MEblue genes are enriched in GTPase-mediated signal transduction, while MEyellow genes play roles in biomacromolecule transport and cell-cell adherens junctions. This result reflects the overall involvement of DEGs detected in the uterine fluid in endometrium-embryo crosstalk related biological processes, which include cell-cell communication, signal reception and transduction, and a series of cellular responses including the transcription and translation of proteins responsible for embryo implantation.

Table 2 WGCNA analysis of DEGs from uterine fluid

Establishing and validating the ER predictive tool

With Tukey’s test from an ANOVA, we selected genes with varying expression identified from each pairwise comparison of receptive stages (pre-receptivity versus receptivity, receptivity versus post-receptivity, and pre-receptivity versus post-receptivity). We therefore applied the expression pattern of these DEGs as training features for ER status classification using the random forest method. A random forest-based feature importance analysis with model prediction based on mean decrease accuracy and the Gini index was performed [35], resulting in 87 predictive markers (Table 3). To improve the power of the predictive tool, we included three hub genes as additional markers (Fig. 3), resulting in an nirsERT. Linear discriminant analysis (LDA) shows that the three ER conditions (pre-receptivity, receptivity, and post-receptivity) were distinctly classified by the expression patterns of these transcriptomic markers (Fig. 4A). To assess the performance of the present predictor, 10-fold cross-validation were applied. We obtained a mean accuracy of 93.0%, a mean specificity of 95.9%, and a mean sensitivity of 90.0%. Uterine fluid samples of different ER conditions could be well separated by setting a probability threshold of 0.6 (Fig. 4B).

Table 3 List of predictive markers selected by random forest algorithm
Fig. 3
figure 3

Partial predictive markers of the nirsERT. A Inferred source of marker and hub genes for the nirsERT; B Co-expression modules of uterine fluid DEGs generated via WGCNA

Fig. 4
figure 4

Establishment and validation of the nirsERT. A Clustering the training set via LDA using selected predictive markers; B Prediction results of training set samples with a probability threshold of 0.6. Pink dot: samples predicted as post-receptivity. Green triangle: samples predicted as pre-receptivity. Blue square: samples predicted as receptivity

Retrospective observation of a small cohort of patients undergoing IVF

To further evaluate the accuracy of the nirsERT, we analyzed the correlation between the predicted results of the nirsERT and the pregnancy outcomes. Twenty-two participants were recruited and their baseline clinical characteristics are shown in Table 4. Twenty-two uterine fluid samples from IVF patients were collected on the day of blastocyst transfer before embryo transfer and tested. Intrauterine pregnancy was determined by ultrasound 28 days after embryo transfer. The success rate of sequencing was 95.4% (21/22) with 1 library failing to pass the quality control procedure. As a result, 18 patients (85.7%, 18/21) were predicted to have a normal WOI, whereas 3 (14.3%, 3/21) and 0 were predicted to have delayed and advanced WOIs, respectively. The intrauterine pregnancy rate (IPR) was 77.8% (14/18) and live birth rate (LBR) was 72.2% (13/18) for patients with a normal WOI. There was no successful pregnancy in patients with a displaced WOI, significantly differing from those with a normal WOI (P < 0.05). The overall IPR and LBR for all patients were recorded as 63.6% (14/22) and 59.1% (13/22), respectively (Table 4).

Table 4 Baseline clinical characteristics, nirsERT results and clinical outcomes of 22 patients undergoing IVF


In recent decades, researchers have investigated a variety of means to evaluate the condition of endometrial receptivity. However, limited progress was made until transcriptomic markers were established [28, 36]. Diagnostic tools resulting from the endometrial tissue transcriptome are accurate and reproducible, but their application is hindered by the need for invasive sampling. Thus, developing a noninvasive, precise and reliable method of ERT is a major challenge in reproductive medicine. In this study, a noninvasive ERT method based on RNA-seq was examined for the first time, and we found the following benefits relative to previous studies. (1) RNA-seq can be used to identify more genes and in a more accurate manner than the conventional gene microarray. (2) Rather than sampling over two time points, we collected samples of uterine fluid at three different time points (the pre-receptive, receptive, and post-receptive stages) from the same patient at 48-h intervals during the same cycle. Thus, the study period was shortened, and a highly correlated sample cohort was established, allowing for a more precise analysis of DEGs to identify marker genes for ER. (3) Over 800 DEGs in uterine fluid were analyzed, providing insight into the functions and roles of multiple genes in embryo implantation. It was difficult to perform transcriptome sequencing with uterine fluid samples, as nearly 1/3 of the samples yielded total RNA of less than 0.25 ng/μL. To address this, we utilized a commercial kit designed for single-cell RNA reverse transcription and amplification. The results show high levels of stability and repeatability, and the Spearman correlation between different amounts of total RNA ranging from 0.2 ng to 20 ng was above 0.98. In using this kit, we successfully prepared 140 RNA-seq libraries and constructed the training dataset. However, 4 libraries failed to pass quality control testing, and we assume this might be caused by extremely low amounts of RNA in these uterine fluid samples. To ensure the availability of the nirsERT, it is important to investigate the distribution of the amount of total RNA in the population. In addition, the improvement of uterine fluid aspiration could be helpful in further studies.

According to our previous study [37], 3571 DEGs were identified from endometrial tissue across the ER statuses, and a predictive tool (rsERT) consisting of 175 marker genes was established based on these DEGs. By analyzing integrated data of the two studies, a total of 864 DEGs were identified, including 468 common DEGs shared with the rsERT study, and 396 uterine fluid-specific DEGs. We found that these common DEGs were significantly enriched in extracellular exosome (GO:0070062), cytoplasm (GO:0005737), cytosol (GO:0005829), nucleoplasm (GO:0005654) and protein binding (GO:0005515), supporting the conclusion that RNAs in uterine fluid originate from endometrial tissue cells with exosomes secreted outside of the cell. Unexpectedly, 396 DEGs were specifically observed in uterine fluid samples. These genes are significantly involved in the integrin-mediated signaling pathway (GO:0007229) and in immune responses, such as leukocyte migration (GO:0050900), inflammatory responses (GO:0006954) and responses to lipopolysaccharide (GO:0032496). Uterine fluid may play an independent role in regulating embryo implantation in terms of adhesion and immunity. In addition, approximately 38.2% (330 of 864) of all DEGs were previously reported [13, 14, 38,39,40,41], while 61.8% (534 of 864) were first identified to be differentially expressed in all three states of receptivity. Our findings highlight the importance of genes involved in protein binding, signal transduction, and leukocyte migration in uterine fluid. For instance, DEGs enriched in extracellular exosomes (GO:0070062), including SLC25A1 (ENSG00000100075), PLSCR1 (ENSG00000188313), and NME3 (ENSG00000103024), were observed to be significantly related to dynamic changes in sequential receptivity stages, which are assumed to mediate communication between the endometrium and embryo. Other cellular responses and signal transduction-related factors, e.g., RAC2 (ENSG00000128340) and ESR1 (ENSG00000091831), were also observed in our study (see Supplementary Tables S2 and S3).

Four hub genes, ECI2, ATP6V1B2, CXCL16 and SELP were identified via WGCNA. ECI2 encodes a key mitochondrial enzyme involved in the beta-oxidation of unsaturated fatty acids, which may provide energy necessary for the embryo implantation course. The presence of SELP implies the possible mechanism of P-selectin-mediated cell adhesion involved in endometrium-embryo interactions. CXCL16 and its receptor CXCR6 have been reported to play a role in decidualization during pregnancy [42]. ATP6V1B2 (ATPase H+ Transporting V1 Subunit B2) is a transmembrane transporter that may be responsible for transporting biomacromolecule-like secretory proteins to their target locations, such as the extracellular matrix. It is evident that these hub genes may play an important role in endometrium-embryo talk and embryo implantation. Hub genes ECI2, ATP6V1B2, and CXCL16 in three of the coexpression modules, MEturquoise, MEyellow, and MEblue, which are highly correlated with ER, were used as marker genes of ER to build the nirsERT model, increasing the predictive efficacy of the model.

nirsERT consisting of 87 markers and 3 hub genes was selected using a random forest algorithm among 864 DEGs. We compared two predictive tools, the nirsERT and rsERT, established using endometrial tissue samples in our previous study, and only 22 markers were shared for both uterine fluid and tissue samples (Supplementary Table S4). According to the Human Protein Atlas, proteins generated by these genes are located in a variety of subcellular locations [26], such as vesicles (BAG5, RAMP2), the nucleus or nucleoplasm (ZNF652, TRAK1), the cytosol (MAP2K6, RNF125) and cell junctions (PKP2). In addition, a strong correlation for the expression patterns of these genes was observed between uterine fluid and endometrial tissue samples (Supplementary Fig. S2). These results indicate that the source of the common markers could be exfoliated endometrial cells or extracellular vesicles. The performance of the nirsERT with rsERT was also examined using the same standard. 10-fold cross-validation revealed comparable mean accuracy (93.0% vs. 98.4%), mean specificity (95.9% vs. 98.9%) and mean sensitivity levels (90% vs. 97.8%).

We also investigated the markers selected in previous studies [14, 28, 37], and few commonalities were observed (Supplementary Fig. S3). No common marker was selected in all three studies. This may be related to the great differences in RNA expression profiles due to different samples from different populations, different RNA profiling technologies and marker gene screening methods used in each study. However, there is no universal standard for selecting marker genes for endometrial receptivity, and the mechanism of uterine transcriptomic changes involved during the process of embryo implantation is still unknown. Further investigations are required to improve the power and reproducibility of endometrial receptivity prediction.

To verify the accuracy of the nirsERT in predicting endometrial receptivity, uterine fluid collected on the day of blastocyst transfer was subjected to the nirsERT. The accuracy of nirsERT prediction was evaluated by analyzing the correlation between the predicted results and subsequent pregnancy outcomes. The results show that 77.8% (14/18) of patients predicted with a normal WOI had successful intrauterine pregnancies, of which 72.2% (13/18) had live birth, while none of the 3 patients with a displaced WOI had successful pregnancies. It is suggested that the failure of embryo implantation in patients with a displaced WOI may be the result of embryo-endometrial asynchrony. Although four unsuccessful intrauterine pregnancies in patients with a normal WOI were predicted by the nirsERT, 77.8% of IRP is consistent with the view that endometrial factors are responsible for approximately two-thirds of embryo implantation [43, 44]. Therefore, the results also further clinically validate the reliability of the nirsERT in predicting the WOI. Personalized embryo transfer (pET) guided by an nirsERT may contribute to restoring the synchronicity of embryonic and endometrial development, which promotes successful embryo implantation. In addition, the clinical pregnancy rate of routine blastocyst transplantation in our reproductive center reached 55–60%, while the overall intrauterine pregnancy rate of patients with aspiration of uterine fluid on the day of embryo transfer reached 63.6%, suggesting that the aspiration of uterine fluid did not affect embryo implantation. This is similar to another study conducted by our team at the same time, which showed the clinical pregnancy rate and implantation rate were higher in the endometrial fluid aspiration group (62.3% versus 50.8%; 57.1% versus 40.8%, respectively) [45]. The results of this study also indicate the feasibility of uterine fluid aspiration as a non-invasive method. Thus, the nirsERT based on RNA-seq of uterine fluid has the potential to detect and guide pET in the same active cycle, contributing to successful embryo implantation.

It follows that our method currently serves as the most promising approach for ideal pET. However, there are some issues has to confront. Firstly, the nirsERT is not applicable for clinical application in current. There are several aspects we need to improve in the further studies, such as detection period, accuracy, convenience, etc. We aim to provide a rapid turnaround for embryo transfer in a same cycle, by developing techniques like RT-qPCR. In fact, if we could identify several crucial RNA markers for endometrial receptivity, RT-qPCR might be promising. We are planning such studies in the near future. We believe this work is hard but well worth the effort. Secondly, the sample size of this validation study was small, and whether an nirsERT can improve the pregnancy outcomes of IVF patients by guiding pET is not yet known. We think it would be better to design a randomized clinical trial in the future to verify the clinical application value of the nirsERT. Lastly, the mechanism of endometrial receptivity marker genes also needs further investigation to provide a theoretical basis for clinical treatment strategies.


In summary, we established a noninvasive RNA-seq-based endometrial receptivity test (nirsERT) by a transcriptome sequencing analysis of uterine fluid combined with a random forest algorithm. Endometrial receptive DEGs in uterine fluid may be derived from endometrial tissue cells and have an independent role in embryo implantation. The nirsERT can predict the WOI period relatively accurately and may serve as a noninvasive, reliable and same cycle test for ER in reproductive clinics.

Availability of data and materials

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.



endometrial receptivity


window of implantation


RNA sequencing


a noninvasive RNA-Seq-based endometrial receptivity test


personalized embryo transfer


repeated implantation failure


assisted reproductive technology


in vitro fertilization


in vitro fertilization and embryo transfer


differentially expressed genes


body mass index


endometrial receptivity array


follicle-stimulating hormone


luteinizing hormone


  1. Macklon NS, Stouffer RL, Giudice LC, Fauser BC. The science behind 25 years of ovarian stimulation for in vitro fertilization. Endocr Rev. 2006;27(2):170–207.

    Article  PubMed  Google Scholar 

  2. Simon A, Laufer N. Repeated implantation failure: clinical approach. Fertil Steril. 2012;97(5):1039–43.

    Article  PubMed  Google Scholar 

  3. Prapas Y, Prapas N, Jones EE, Duleba AJ, Olive DL, Chatziparasidou A, et al. The window for embryo transfer in oocyte donation cycles depends on the duration of progesterone therapy. Hum Reprod. 1998;13(3):720–3.

    Article  CAS  PubMed  Google Scholar 

  4. Norwitz ER, Schust DJ, Fisher SJ. Implantation and the survival of early pregnancy. N Engl J Med. 2001;345(19):1400–8.

    Article  CAS  PubMed  Google Scholar 

  5. Galliano D, Bellver J, Diaz-Garcia C, Simon C, Pellicer A. ART and uterine pathology: how relevant is the maternal side for implantation? Hum Reprod Update. 2015;21(1):13–38.

    Article  CAS  PubMed  Google Scholar 

  6. Sebastian-Leon P, Garrido N, Remohi J, Pellicer A, Diaz-Gimeno P. Asynchronous and pathological windows of implantation: two causes of recurrent implantation failure. Hum Reprod. 2018;33(4):626–35.

    Article  CAS  PubMed  Google Scholar 

  7. Zhao J, Zhang Q, Wang Y, Li Y. Endometrial pattern, thickness and growth in predicting pregnancy outcome following 3319 IVF cycle. Reprod BioMed Online. 2014;29(3):291–8.

    Article  PubMed  Google Scholar 

  8. Zhao J, Zhang Q, Li Y. The effect of endometrial thickness and pattern measured by ultrasonography on pregnancy outcomes during IVF-ET cycles. Reprod Biol Endocrinol. 2012;10:100.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Hou Z, Zhang Q, Zhao J, Xu A, He A, Huang X, et al. Value of endometrial echo pattern transformation after hCG trigger in predicting IVF pregnancy outcome: a prospective cohort study. Reprod Biol Endocrinol. 2019;17(1):74.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Coutifaris C, Myers ER, Guzick DS, Diamond MP, Carson SA, Legro RS, et al. Histological dating of timed endometrial biopsy tissue is not related to fertility status. Fertil Steril. 2004;82(5):1264–72.

    Article  PubMed  Google Scholar 

  11. Zhu L, Che HS, Xiao L, Li YP. Uterine peristalsis before embryo transfer affects the chance of clinical pregnancy in fresh and frozen-thawed embryo transfer cycles. Hum Reprod. 2014;29(6):1238–43.

    Article  CAS  PubMed  Google Scholar 

  12. Qiong Z, Jie H, Yonggang W, Bin X, Jing Z, Yanping L. Clinical validation of pinopode as a marker of endometrial receptivity: a randomized controlled trial. Fertil Steril. 2017;108(3):513–517 e512.

    Article  PubMed  Google Scholar 

  13. Hu S, Yao G, Wang Y, Xu H, Ji X, He Y, et al. Transcriptomic changes during the pre-receptive to receptive transition in human endometrium detected by RNA-Seq. J Clin Endocrinol Metab. 2014;99(12):E2744–53.

    Article  CAS  PubMed  Google Scholar 

  14. Diaz-Gimeno P, Horcajadas JA, Martinez-Conejero JA, Esteban FJ, Alama P, Pellicer A. Simon C: A genomic diagnostic tool for human endometrial receptivity based on the transcriptomic signature. Fertil Steril. 2011;95(1):50–60 60 e51–15.

    Article  CAS  PubMed  Google Scholar 

  15. Diaz-Gimeno P, Ruiz-Alonso M, Blesa D, Bosch N, Martinez-Conejero JA, Alama P, et al. The accuracy and reproducibility of the endometrial receptivity array is superior to histology as a diagnostic method for endometrial receptivity. Fertil Steril. 2013;99(2):508–17.

    Article  PubMed  Google Scholar 

  16. Garrido-Gomez T, Ruiz-Alonso M, Blesa D, Diaz-Gimeno P, Vilella F, Simon C. Profiling the gene signature of endometrial receptivity: clinical results. Fertil Steril. 2013;99(4):1078–85.

    Article  CAS  PubMed  Google Scholar 

  17. Simón CGC, Cabanillas S, Vladimirov I, Castillón G, Giles J, Boynukalin K, et al. A 5-year multicentre randomized controlled trial comparing personalized, frozen and fresh blastocyst transfer in IVF. Reprod BioMed Online. 2020;41(3):14.

    Article  Google Scholar 

  18. Ruiz-Alonso M, Blesa D, Diaz-Gimeno P, Gomez E, Fernandez-Sanchez M, Carranza F, et al. The endometrial receptivity array for diagnosis and personalized embryo transfer as a treatment for patients with repeated implantation failure. Fertil Steril. 2013;100(3):818–24.

    Article  PubMed  Google Scholar 

  19. Altmäe S, Esteban FJ, Stavreus-Evers A, Simón C, Giudice L, Lessey BA, et al. Guidelines for the design, analysis and interpretation of 'omics' data: focus on human endometrium. Hum Reprod Update. 2014;20(1):12–28.

    Article  PubMed  Google Scholar 

  20. Demiral İ, Doğan M, Baştu E, Buyru F. Genomic, proteomic and lipidomic evaluation of endometrial receptivity. Turk J Obstetr Gynecol. 2015;12(4):237–43.

    Article  Google Scholar 

  21. Nagalakshmi U, Waern K, Snyder M. RNA-Seq: a method for comprehensive transcriptome analysis. Curr Protoc Mol Biol. 2010;Chapter 4:Unit 4.11.11–13.

    Google Scholar 

  22. Sîrbu A, Kerr G, Crane M, Ruskin HJ. RNA-Seq vs dual- and single-channel microarray data: sensitivity analysis for differential expression and clustering. PLoS One. 2012;7(12):e50986.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Kalma Y, Granot I, Gnainsky Y, Or Y, Czernobilsky B, Dekel N, et al. Endometrial biopsy-induced gene modulation: first evidence for the expression of bladder-transmembranal uroplakin Ib in human endometrium. Fertil Steril. 2009;91(4):1042–9 1049.e1041–1049.

    Article  CAS  PubMed  Google Scholar 

  24. Karimzade MA, Oskouian H, Ahmadi S, Oskouian L. Local injury to the endometrium on the day of oocyte retrieval has a negative impact on implantation in assisted reproductive cycles: a randomized controlled trial. Arch Gynecol Obstet. 2010;281(3):499–503.

    Article  PubMed  Google Scholar 

  25. Bhusane K, Bhutada S, Chaudhari U, Savardekar L, Katkam R, Sachdeva G. Secrets of endometrial receptivity: some are hidden in uterine secretome. Am J Reprod Immunol. 2016;75(3):226–36.

    Article  PubMed  Google Scholar 

  26. Thul PJ, Åkesson L, Wiking M, Mahdessian D, Geladaki A, Ait Blal H, et al. A subcellular map of the human proteome. Science. 2017;356(6340):eaal3321.

  27. van der Gaast MH, Beier-Hellwig K, Fauser BC, Beier HM, Macklon NS. Endometrial secretion aspiration prior to embryo transfer does not reduce implantation rates. Reprod BioMed Online. 2003;7(1):105–9.

    Article  PubMed  Google Scholar 

  28. Chan C, Virtanen C, Winegarden NA, Colgan TJ, Brown TJ, Greenblatt EM. Discovery of biomarkers of endometrial receptivity through a minimally invasive approach: a validation study with implications for assisted reproduction. Fertil Steril. 2013;100(3):810–7.

    Article  CAS  PubMed  Google Scholar 

  29. Gardner DK, Lane M, Stevens J, Schlenker T, Schoolcraft WB. Reprint of: Blastocyst score affects implantation and pregnancy outcome: towards a single blastocyst transfer. Fertil Steril. 2019;112(4 Suppl1):e81–4.

    Article  PubMed  Google Scholar 

  30. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics (Oxford, England). 2013;29(1):15–21.

    CAS  PubMed  Google Scholar 

  31. DeLuca DS, Levin JZ, Sivachenko A, Fennell T, Nazaire MD, Williams C, et al. RNA-SeQC: RNA-seq metrics for quality control and process optimization. Bioinformatics (Oxford, England). 2012;28(11):1530–2.

    CAS  PubMed  Google Scholar 

  32. Jiao X, Sherman BT, Huang da W, Stephens R, Baseler MW, Lane HC, et al. DAVID-WS: a stateful web service to facilitate gene/protein list analysis. Bioinformatics (Oxford, England). 2012;28(13):1805–6.

    CAS  PubMed  Google Scholar 

  33. Zhang B, Horvath S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol. 2005;4:Article17.

    Article  PubMed  Google Scholar 

  34. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003;13(11):2498–504.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Menze BH, Kelm BM, Masuch R, Himmelreich U, Bachert P, Petrich W, et al. A comparison of random forest and its Gini importance with standard chemometric methods for the feature selection and classification of spectral data. BMC Bioinformatics. 2009;10(1):213.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Suhorutshenko M, Kukushkina V, Velthut-Meikas A, Altmäe S, Peters M, Mägi R, et al. Endometrial receptivity revisited: endometrial transcriptome adjusted for tissue cellular heterogeneity. Hum Reprod. 2018;33(11):2074–86.

    Article  CAS  PubMed  Google Scholar 

  37. He A, Zou Y, Wan C, Zhao J, Zhang Q, Yao Z, et al. The role of transcriptomic biomarkers of endometrial receptivity in personalized embryo transfer for patients with repeated implantation failure. J Transl Med. 2021;19(1):176.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Kao LC, Tulac S, Lobo S, Imani B, Yang JP, Germeyer A, et al. Global gene profiling in human endometrium during the window of implantation. Endocrinology. 2002;143(6):2119–38.

    Article  CAS  PubMed  Google Scholar 

  39. Borthwick JM, Charnock-Jones DS, Tom BD, Hull ML, Teirney R, Phillips SC, et al. Determination of the transcript profile of human endometrium. Mol Hum Reprod. 2003;9(1):19–33.

    Article  CAS  PubMed  Google Scholar 

  40. Altmäe S, Reimand J, Hovatta O, Zhang P, Kere J, Laisk T, et al. Research resource: interactome of human embryo implantation: identification of gene expression pathways, regulation, and integrated regulatory networks. Mol Endocrinol. 2012;26(1):203–17.

    Article  PubMed  Google Scholar 

  41. Altmae S, Koel M, Vosa U, Adler P, Suhorutsenko M, Laisk-Podar T, et al. Meta-signature of human endometrial receptivity: a meta-analysis and validation study of transcriptomic biomarkers. Sci Rep. 2017;7(1):10077.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Mei J, Yan Y, Li SY, Zhou WJ, Zhang Q, Li MQ, et al. CXCL16/CXCR6 interaction promotes endometrial decidualization via the PI3K/AKT pathway. Reproduction (Cambridge, England). 2019;157(3):273–82.

    Article  CAS  PubMed  Google Scholar 

  43. Messaoudi S, El Kasmi I, Bourdiec A, Crespo K, Bissonnette L, Le Saint C, et al. 15 years of transcriptomic analysis on endometrial receptivity: what have we learnt? Fertil Res Pract. 2019;5:9.

    Article  PubMed  PubMed Central  Google Scholar 

  44. Craciunas L, Gallos I, Chu J, Bourne T, Quenby S, Brosens JJ, et al. Conventional and modern markers of endometrial receptivity: a systematic review and meta-analysis. Hum Reprod Update. 2019;25(2):202–23.

    Article  CAS  PubMed  Google Scholar 

  45. Hou Z, He A, Zhang Q, Liu N, Liu D, Li Y, et al. Endometrial fluid aspiration immediately prior to embryo transfer does not affect IVF/vitrified-warmed embryo transfer outcomes - a prospective matched cohort study. Reprod BioMed Online. 2022;44(3):486–93.

    Article  CAS  PubMed  Google Scholar 

Download references


We thank all patients and their family for the participation. We also thank Elsevier Author Services for preparation the Fig. 3.


This project is funded by the National Natural Science Foundation of China (grant no. 8187061497) and the National Key Research and Developmental Program of China (grant no. 2018YFC1004800).

Author information

Authors and Affiliations



A.H, H.W, Y.L, S.L conceived and designed the study. A.H, H.W, J.Z, Q.Z, N.L, D.L, Y.L, J.F, H.L, X.H, Y.S, T.Y, J.W collected clinical data. A.H, N.L, D.L, Y.L, J.F, X.H, T.Y, Z.H collected samples. Y.Z, C.W, C.H perform sequencing experiments and recordings. A.H, H.W, Y.Z, C.W, C.H, X.D performed the data analysis, model building and prepared the figures. A.H, H.W, Y.Z, C.W, C.H drafted the manuscript. Y.L, S.L, A.H, H.W, Y.Z, C.W supervised and revised the manuscript for important intellectual content. All authors reviewed the manuscript. The authors read and approved the final manuscript.

Corresponding authors

Correspondence to Sijia Lu or Yanping Li.

Ethics declarations

Ethics approval and consent to participate

The present study was conducted at the Center for Reproductive Medicine at Xiangya Hospital of Central South University with permission from the Ethics Committee of Reproductive Medicine. This study is registered with the Chinese Clinical Trial Registry (No. ChiCTR-DDD-17013375).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: 

Supplementary Materials and Methods. Fig. S1. Analysis of stability and repeatability for the transcriptome sequencing with low amount of RNA. Fig. S2. Expression pattern of 22 common markers between nirsERT and rsERT. Fig. S3. Venn diagram of predictive markers selected from three independent studies of endometrial receptivity.

Additional file 2: Table S1.

Clinical characteristics of the participants.

Additional file 3: Table S2.

Functional enrichment of uterine fluid DEGs.

Additional file 4: Table S3.

GO enrichment analysis of common and specific DEGs from uterine fluid samples.

Additional file 5: Table S4.

Functional annotation and Subcellular Location of common markers for nirsERT and rsERT.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

He, A., Wu, H., Zou, Y. et al. Can biomarkers identified from the uterine fluid transcriptome be used to establish a noninvasive endometrial receptivity prediction tool? A proof-of-concept study. Reprod Biol Endocrinol 21, 20 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:


  • Endometrial receptivity
  • Window of implantation
  • Transcriptomic profiling
  • Machine learning
  • Random forest algorithm
  • Noninvasive biomarker