Aberrantly expressed long noncoding RNAs in recurrent implantation failure: A microarray related study

ABSTRACT

Long noncoding RNAs (lncRNAs) are a class of noncoding RNAs longer than 200 nucleotides. They were long regarded as transcription noise for their low expression and non-protein coding features. Recent published reports indicate that lncRNAs are involved in virtually every aspect of human biology. We aimed to profile the endometrial lncRNA expression pattern in women with recurrent implantation failure (RIF) and predict the function of the genes of the dysregulated lncRNA transcripts. Endometrial samples (24) were collected during window of implantation (14 RIF women and 10 women who conceived after embryo transfer). For the microarray study, 7 RIF endometrium and 5 control endometrium were selected, and quantitative real-time PCR (RT-qPCR) was performed on the rest of the endometrial samples to validate the microarray results. After that, lncRNA-mRNA co-expression analysis, GO analysis, KEGG analysis, and lncRNA-transcript factor (TF) analysis were carried out to analyze the gene functions of the dysregulated lncRNA transcripts. We detected a total of 197 lncRNA transcripts that were dysregulated in RIF endometrium compared with the control group. The relative expression levels of eight selected lncRNA transcripts were validated by RT-qPCR and were in accordance with the microarray outcomes. GO and KEGG analyses revealed that the coexpressed mRNA transcripts were involved in pathways that may affect endometrial receptivity such as cell adhesion. The lncRNA target predictions provided potential TF targets of the dysregulated lncRNA transcripts. Our results indicate that lncRNA expression profiles of RIF endometrium were different from that of normal receptive endometrial, suggesting that lncRNAs may regulate endometrial receptivity.

Abbreviations: GO: Gene Oncology; GFs: growth factors; KEGG: Kyoto Encyclopedia of Genes and Genomes; lncRNAs: long noncoding RNAs; PCA3: prostate cancer antigen 3; RT-qPCR: quantitative real-time PCR; RIF: recurrent implantation failure; STK: serine/threonine kinase; TF: transcription factor; WOI: window of implantation

KEYWORDS:

• Endometrial receptivity
• long noncoding RNA
• microarray
• recurrent implantation failure

Introduction

Successful embryo implantation is a process that requires both a synchronous development and coordinate cross-talk between the endometrium and the blastocyst. During the window of implantation (WOI), which spans 6−8 days after the LH peak, the endometrium acquires changes allowing embryo adhesion and subsequent invasion. The changes include histological changes under the influence of steroid hormones [Noyes et al. 1975], formation of pinopodes [Sudoma et al. 2011], and expression changes of integrins (α_vβ₃, α₄β₁), leukemia inhibitory factors, and other potential molecular markers of WOI [Koler et al. 2009]. Recurrent implantation failure (RIF) was defined as a failure of pregnancy after > 3 embryo transfers with high quality embryos or the transfer of ≥ 10 embryos in multiple transfers [Thornhill et al. 2005; Coughlan et al. 2014]. The cause of implantation failure is diverse and is due to maternal factors such as endometrial pathologies, hormonal or metabolic disorders, and immune factors, as well as other pathologies [Timeva et al. 2014]. It is now known that some critical genes are aberrantly expressed in RIF endometrium at the WOI [Tapia et al. 2011; Koler et al. 2009; Tepekoy et al. 2015], but due to the deficiencies in the specificity and sensitivity of these genes, this knowledge has yet to be put into wide clinical use. New gene markers are required for a better diagnosis or prediction of endometrial receptivity.

Long noncoding RNAs (lncRNAs) are a class of noncoding RNAs longer than 200 nucleotides. They were long regarded as transcription noise for their low expression and non-protein coding features. Accumulating published reports indicate that lncRNAs are involved in virtually every aspect of human biology, including chromosomal dosage compensation, control of imprinting, chromatin modification, chromatin structure, transcription, splicing, translation, cellular differentiation, integrity of cellular structures, cell cycle regulation, intracellular trafficking, reprogramming of stem cells, and heat shock response [Rinn and Chang 2012; Guttman and Rinn 2012]. Prostate cancer antigen 3 (PCA3) is a long noncoding RNA that is highly overexpressed in prostate cancer and can be detected in the urine of prostate cancer patients [Wei et al. 2014]. In 2012, the US Food and Drug Administration registered PCA3 as a risk assessment marker for prostate cancer with an indicated use to facilitate biopsy decision making among men with prior negative prostate biopsies [Huang et al. 2015]. LncRNAs such as PCA3 often play regulatory roles in pathologies, and dysfunction of some lncRNAs may result in diseases, such as cancer, heart failure, and endometriosis [Sun et al. 2014; Wang et al. 2015; Yang et al. 2015]. However, reports on the lncRNA expression profiles of RIF patients are still lacking. As such, in this study, we performed an RIF endometrial lncRNA expression profile analysis.

Results and discussion

Patients and endometrial samples

A total of 24 women were recruited for our study; 14 were RIF patients, and the other 10 women were used for the control group. Endometrial thicknesses and endometrial types were detected by trans-vaginal ultrasound on the day of LH surge. The endometrial thickness in each of the subjects was ≥ 0.7 cm, and the endometrial types were type A or type A−B according to the Gonen Y endometrial classification method [Gonen and Casper 1990]. The histological diagnoses of all endometrial samples were middle secretory phase endometrium. We applied a Student’s t-test (nonparametric tests when the distribution was abnormal) and found that there were no significant differences between the RIF group and the control group in age, BMI (body mass index), menstruation cycle length, menses duration, and endometrial thickness (Table 1). Supplemental Table 1 provides additional information of each subject.

Table 1. Characteristics of the women undergoing endometrial biopsy sampling.

	RIF(n=14)	Ctr (n=10)	P value
Age	32.43±0.98	32.10±2.89	0.82
BMI	22.53±2.51	21.71±2.22	0.42
Cycle length (days)	30.43±3.03	30.40±4.34	0.99
Menses duration (days)	5.07±1.14	5.05±0.98	0.96
LH+X (days)	LH+6.64	LH+7.30	/
Endometrial thickness (cm)	0.93±0.22	0.97±0.26	0.68

RIF: recurrent implantation failure; Ctr: control; BMI: body mass index; LH+X(days): days since the LH surge when the biopsy was taken; Endometrial thickness: the thickness of the endometrium on the day of LH surge.

The A260/A280 and A260/A230 ratios assessed by nanodrop and 28S/18S ratio assessed in 2% agarose gel electrophoresis indicated that the total RNA was suitable for further analysis. The first 12 endometrial samples collected (RIF 1 to RIF 7 and Control 1 to Control 5 in Supplemental Table 1) were selected to perform a microarray study. After that, the 12 endometrial samples collected later (sample RIF 8 to RIF 14 and Control 6 to Control 10 in Supplemental Table 1) were used for quantitative real-time PCR (RT-qPCR) validation.

LncRNA expression profile of RIF endometrial tissue

We found 197 lncRNA transcripts that were significantly dysregulated in RIF WOI endometrial samples compared with that in the control samples, with 132 lncRNA transcripts that were identified to be consistently up-regulated and 65 lncRNA transcripts that were down-regulated (FC ≥ 2 and P < 0.05). ENSG00000229385.1 (p9212), ENSG00000267549.1 (p8861), and LOC100505912 (p25300) were the most up-regulated lncRNA transcripts. In comparison, ENSG00000227068.1 (p16922), ENSG00000260228.1 (p30996), and ENSG00000263724.1 (p7859) were the most down-regulated lncRNA transcripts. Table 2 lists the top 20 up- and down-regulated lncRNA transcripts in our microarray study. Supplemental Table 2 lists all of the dysregulated lncRNA transcripts. All of our microarray gene expression information has been deposited in GEO as GSE71331.

Table 2. Top 20 dysregulated lncRNA transcripts in RIF endometrium.

Up-regulated lncRNAs			Down-regulated lncRNAs
Gene name	Probe NO.	FC	Gene name	Probe NO.	FC
ENSG00000229385.1	p9212	13.39	ENSG00000227068.1	p16922	9.64
ENSG00000267549.1	p8861	7.64	ENSG00000260228.1	p30996	9.24
LOC100505912	p25300	7.07	ENSG00000263724.1	p7859	7.46
ENSG00000267549.1	p8859	5.97	XLOC_000266	p32242	5.78
ENSG00000261645.1	p2900	5.71	XLOC_008826	p17962	5.71
ENSG00000267194.1	p7482	5.54	ENSG00000245750.3	p5635	5.61
ENSG00000261645.1	p2902	5.47	ENSG00000258733.1	p30863	5.16
XLOC_004399	p22979	5.32	XLOC_011294	p19780	5.05
ENSG00000237807.3	p15697	4.67	ENSG00000230024.1	p647	4.03
XLOC_003146	p29863	4.53	ENSG00000235142.2	p14287	3.90
ENSG00000261645.1	p2899	4.35	ENSG00000253666.1	p16303	3.86
ENSG00000248832.1	p33512	3.95	ENSG00000227121.1	p1614	3.74
ENSG00000239501.1	p4151	3.84	ENSG00000236366.1	p31892	3.68
ENSG00000253929.1	p16370	3.76	XLOC_011294	p19778	3.67
LOC100507516	p25504	3.64	ENSG00000235885.3	p9838	3.33
ENSG00000197291.4	p6832	3.52	ENSG00000263531.1	p6780	3.21
ENSG00000263427.1	p6668	3.47	ENSG00000235706.3	p30868	3.15
BC128552	p25729	3.43	XLOC_006335	p23606	3.04
XLOC_005748	p23230	3.42	ENSG00000237174.2	p14594	3.04
ENSG00000224259.1	p6832	3.37	AX747026	p25714	3.00

lncRNA: long noncoding RNAs; RIF: recurrent implantation failure. IncRNAs in bold font were selected to perform RT-qPCR validation.

Recently, there have been a few reports on the expression pattern of endometrial lncRNAs [Sun et al. 2014; Wang et al. 2015]. We compared our top 20 dysregulated lncRNA transcripts in RIF WOI endometrium with two reported endometriosis studies. However, we found that none of our top 20 dysregulated lncRNA transcripts coincided with lnRNAs in these two reports. This may be due to the following reasons: (1) the pathology of poor endometrial receptivity is not the same as the pathology of endometriosis; (2) the microarrays we applied were different than those used in the other studies; and (3) the biopsy times were different across studies. Wang et al. [2015] collected endometrium at the late secretory phase.

Clustering analysis

In an unsupervised hierarchical clustering analysis, the differentially expressed lncRNA transcripts were used for generating a heat map. In Figure 1, each column represents an endometrial sample, and each row presents a dysregulated lncRNA transcript. Red stripes imply that the lncRNA was up-regulated in the endometrial sample, and the green stripes imply the opposite. We can see that the RIF endometrial samples are gathered on the left side of the heat map, and the control samples are gathered on the right side. This result may indicate that the presence of specific lncRNAs may be able to distinguish normal receptive endometrium from RIF endometrium.

Figure 1. Clustering analysis results. Each line represents one long noncoding RNA (lncRNA) and each column represents one endometrial sample. The relative lncRNA expression is depicted according to the color scale. Red column indicates up regulation and green column indicates down regulation. Recurrent implantation failure (RIF) 1~7 represent RIF endometrial samples and Control 1~5 represent receptive control endometrial samples.

However, in Figure 1, the RIF 4 and RIF 6 samples were grouped with the 5 control endometrial samples in the second topmost clustering. In a follow-up visit, we found that RIF 4 and RIF 6 were both pregnant with the next embryo transplantation after this biopsy. We suggest that the etiologies of these two RIF patients were different from that of the other 5 RIF patients.

RT-qPCR analysis

We selected 8 lncRNAs to perform RT-qPCR validation. LncRNA selection was based on our prior RIF lncRNA expression profile analysis and the later lncRNA function prediction. The 8 selected lncRNAs reached at least 2 of the following criterion: (1) significantly dysregulated in RIF WOI endometrium (P < 0.05); (2) relatively larger fold change (FC > 2); and (3) predicted to be coexpressed with some known mRNAs related to endometrial receptivity. As shown in Figure 2, the fold changes of the 8-selected dysregulated lncRNA transcripts in the RT-qPCR validation were mostly consistent with that of the microarray. As the complete names of the 8 selected lncRNA transcripts were too long to show in the figure, we replaced them with the probe number. The complete names of our 8 selected lncRNA transcripts can be found in Supplemental Table 2. The consistency of the RT-qPCR validates the microarray results.

Figure 2. RT-qPCR validation of the eight selected lncRNAs transcripts. The grey columns represent results of microarray study, the dark (red) columns represent results of RT-qPCR validation. The heights of the columns represent the fold changes. Fold change is positive when the lncRNA is up regulated (RIF/Control) and negative when down regulated. The whole name of the selected lncRNA transcripts were replaced with probe number. lncRNA: long noncoding RNA; RIF: recurrent implantation failure.

LncRNA-mRNA co-expression analysis

LncRNAs can control gene expression via binding to and targeting chromatin regulators, activating gene enhancers, and recruiting chromatin modifying complexes to DNA [Rinn and Chang 2012]. However, the functions of the great majority of lncRNAs have not been defined. We could only predict the potential mRNA targets of our dysregulated lncRNA transcripts through co-expression analysis. Because a single lncRNA could be co-expressed with tens of mRNA transcripts, we set a rigid restriction on the Pearson correlation value and p value to determine the most reliable lncRNA-mRNA co-expression pairs. LncRNA transcripts and mRNA transcripts with a significant correlation and with top 100 Pearson correlation coefficients were selected to draw the network.

Figure 3. LncRNA-mRNA co-expression network. The yellow circles represent lncRNAs and the green rings represent mRNAs. Red connection lines indicate that the lncRNAs positively co-express with the connected mRNA transcripts, and the blue lines mean the contrast. The diameter of each yellow circle implies degree centrality. lncRNA: long noncoding RNA.

Figure 3 shows the top 100 most significant correlation lncRNA-mRNA pairs. Forty lncRNA transcripts (yellow circles) coexpressed with 64 mRNA transcripts (green circles). LncRNA transcripts with higher degree centrality are presented in bigger yellow circles. A positive co-expression relationship is displayed with a red line, and a negative co-expression relationship is displayed with a blue line. According to the statistics, ENSG00000267194.1 (p7482) was coexpressed with the greatest number of mRNA transcripts (coexpressed with 10 mRNA transcripts in total); it was positively expressed with 9 mRNA transcripts (MAP2K6, MEGF10, TRPM6, MAB21L3, SLAIN1, SOX17, NOV, CTNNA2, and NDRG2) and negatively expressed with one lncRNA transcript (GPX3). The gene functions of the above 10 coexpressed mRNAs could be classified into the following 6 functional categories.

1. Genes coding for cellular components: TRPM [Zhang et al. 2014b], MAP2K6 [Fong et al. 2007] and CTNNA2 [Denis et al. 2014] are mainly involved membrane structure. NDRG2 is a component of the of the Golgi apparatus. We know that the abnormal membrane composition and Golgi could affect cellular functions and cellular activities. So, it is reasonable that a defective endometrial cellular component may result in poor endometrial function [Zhu et al. 2015; Mukherjee et al. 2007].

2. Genes coding for proteins that target types of growth factors (GFs): NOV binds to IGF [Sarkissyan et al. 2014], and NDRG2 binds to VEGF [Ma et al. 2012]. Tazuke and Giudice [1996] reported that the expression of various growth factors and their receptors in the endometrium are cell specific and temporal during implantation, suggesting that some of these GFs are important for endometrial receptivity [Tazuke and Giudice 1996]. Guzeloglu-Kayisli et al. [2009] summarized that several well-known growth factors like EGF, TGF-β, and VEGF families are important for implantation.

3. Genes coding for proteins involved in adhesion, such as CTNNA2 [Denis et al. 2014] and MEGF10 [Suzuki and Nakayama 2007]. During the early stages of implantation, the blastocyst enters the uterine cavity, apposes and then adheres to a receptive endometrium to initiate implantation. Abnormalities in adhesion during the very early stage of implantation could result in implantation failure [Cuman et al. 2015].

4. Genes coding for proteins involved in the regulation of enzyme activity. TRPM6 [Zhang et al. 2014b] and MAP2K6 [Fong et al. 2007] control the activity of protein serine/threonine kinase (STK), and GPX3 controls glutathione peroxidase activity. Both STK and glutathione peroxidase are involved in intracellular signaling pathways and cellular activities like proliferation, differentiation, and apoptosis [McGuire et al. 2014]. Disruption of intracellular signaling and cellular activities in endometrial cells could finally result in poor endometrial receptivity in WOI.

5. Genes coding for proteins that act on ion channels. TRPM6 [Zhang et al. 2014b] regulates ion channel activity and metal ion transportation (including Ca²⁺ and Mg²⁺). Metal ions are essential for endometrial receptivity. For example, Ca²⁺ could bind to E-cadherin (a Ca²⁺-dependent transmembrane adhesion protein) and then mediate embryo adhesion [Zhang et. 2013].

6. Genes coding for proteins involved in cellular protection. TRPM6 [Zhang et al. 2014b], MAP2K6 [Fong et al. 2007], and GPX3 [Farimani Sanoee et al. 2014] rescue cells from the damage of toxicants, stress, and ROS. Several studies have linked the prevalence of female infertility with an increase in oxidative stress and toxicant levels in the various critical micro- or macro-environments in the body or in the uterus [Gupta et al. 2014].

In all, lncRNA-mRNA co-expression, ENSG00000267194.1 (p7482) may target 10 mRNA transcripts and alter endometrial receptivity through several pathways. However, the authentic gene functions of ENSG00000267194.1 (p7482) and other dysregulated lncRNA transcripts need further investigation.

In Figure 3, XLOC_005748 (p23230) and XLOC_004399 (p22979) are shown to be coexpressed with the second and third largest numbers of mRNA transcripts. The lncRNA-mRNA co-expression network contains substantial information about dysregulated genes. The corresponding gene names of the yellow circles in Figure 3 are listed in Supplemental Table 2.

Gene Oncology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of coexpressed mRNA

Because only a small portion of lncRNAs have been functionally annotated, the most common method to predict the functions of lncRNA is to investigate their coexpressed mRNAs. Thus, we applied two enrichment analyses, the GO (US) analysis and the KEGG (Japan) analysis. The GO analysis summarized the conceivable biological processes, cellular components, and molecular functions of the coexpressed mRNAs. The KEGG analysis summarized the potential pathways of the coexpressed mRNAs.

Figure 4 shows the biological functions of the targets of dysregulated lncRNA transcripts (Figure 4A). The coexpressed mRNAs mainly participate in biological processes such as anatomical structure regression, Mullerian duct regression, and epithelium development. The coexpressed mRNAs are mainly involved in synthesizing cellular components such as the extracellular matrix, basement membrane, and trailing edge (Figure 4B). According to the GO molecular function analysis (Figure 4C), the coexpressed mRNAs play important roles in modulating TGF-β receptor pathway specific activity, N-acetyllactosaminide β-1,6-N-acetyglucosaminyl transferase activity and 1-phosphatidylinositol-4-phosphate 5-kinase activity. Further, the KEGG analysis (Figure 4D) revealed that the targets significantly participate in the following pathways: the phosphatidylinositol signaling system, biosynthesis of unsaturated fatty acids, and the glycosphingolipid biosynthesis-lacto series.

Figure 4. GO enrichment analysis and KEGG pathway annotations of targets of co-expressed mRNAs. (A) Biological processes; (B) cellular components; (C) molecular functions; (D) KEGG pathway annotation. The horizontal ordinate shows the enrichment score. The ordinate shows the list of top ten biological processes, cellular components, molecular functions, and KEGG pathways. GO: Gene Oncology; KEGG: Kyoto Encyclopedia of Genes and Genomes.

The above GO and KEGG analyses revealed that the coexpressed mRNAs may regulate endometrial receptivity through several molecular mechanisms and pathways. Overall, the coexpressed mRNAs are mainly involved in pathways and gene functions related to basic and necessary cellular processes and components, such as anatomical structure regression and extracellular matrix. It is reasonable that abnormalities in basic and necessary cellular components and biological processes in the endometrium would finally result in damage to endometrial receptivity. We also noticed that the PPAR signaling pathway was sixth in the KEGG enrichment analysis. In recent decades, intensive study of PPARs has shed novel insight on female reproduction. PPARs regulate cell proliferation, secretion of hormones in the ovary, invasion of gestational trophoblast cells, and the synchronization of prostaglandin, steroids, and cytokines in the endometrium [Yang et al. 2008].

Transcription factor (TF) prediction

A TFSearch was applied to predict the TF binding sites of the dysregulated lncRNAs. Using the threshold of core matrix match =1, 170 lncRNA-TF pairs were found, covering 100 lncRNA transcripts and 23 TFs. Then, we generated a lncRNA-TF network (Figure 5). We can see that Nkx2-5 is in the core of the lncRNA-TF network and participated in 74 lncRNA-TF pairs. Nkx2-5 is among the earliest known markers of the cardiac mesoderm and could affect cardiac development through modulation of Wnt/β-catenin [Cambier et al. 2014]. Wnt/β-catenin has been reported to regulate endometrial gland formation, and ablation of specific Wnt signaling components results in infertility due to implantation failure [Tepekoy et al. 2015]. Due to the lack of studies on Nkx2-5 in the endometrium, the mechanisms of Nkx2-5 in the endometrium need to be studied further. The remaining 22 TFs were predicted to be targeted by no more than six lncRNA transcripts each. The gene symbols of the lncRNA transcripts are listed in Supplemental Table 3.

Figure 5. LncRNA-TF core network. LncRNA-TF core network consisting of 170 lncRNA-TF pairs with the highest matrix match score. The purple circles represent TFs and yellow circles represent lncRNAs. lncRNAs: long noncoding RNAs; TF: transcription factor.

Materials and methods

Sample collection

Participants were recruited from the Reproductive Medical Center of Peking University People’s Hospital. The inclusion criteria for the participants were as follows: < 40 y in age, have a regular menstrual cycle, have had no positive findings after trans-vaginal ultrasound and physical examination, have not taken oral contraceptives or used an intrauterine contraceptive device in 3 months, have a good hormonal reserve (FSH <10 mIU/mL), and have a good response to hormonal stimulation. Participants who suffered from hysteromyoma, adenomyosis, metrosynizesis, polycystic ovary syndrome, endometriosis, and /or hydrosalpinx were excluded in this study. The cause of sterility in the two groups were fallopian tube factor, male factor, and/or unknown reason. In the control group, all volunteers conceived after their first embryo transfer. While in the RIF group, patients failed after ≥ 3 times of high quality embryo transfers or ≥ 10 embryos were transferred. In addition, there was no other explanation for RIF after a thorough infertility work-up.

Each participant had undergone continuous trans-vaginal ultrasounds since menses +9 − 11 d until ovulation was detected. Meanwhile, urine LH was recorded at the same time to validate ovulation. Endometrial thickness and endometrial type were measured on the day of the LH surge. Endometrial samplings occurred at LH +6 − 8 d. Endometrial biopsies were performed during hysteroscopy or curettage. Biopsies were transferred to the laboratory in 1 h. Part of the tissue was embedded in paraffin for histological evaluation, and the other part was stored in liquid nitrogen for later RNA extraction. Sectioning and hematoxylin and eosin (HE) staining of endometrial samples were carried out by the Pathology Department of the Peking University People’s Hospital (Beijing, China). All participants signed informed consent forms following the approval of the institutional ethics committee (number 2011-87).

RNA isolation

Total RNA was extracted from endometrial samples by using Trizol reagent (Invitrogen, USA). For the lncRNA and mRNA microarray analyses, total RNA was purified with an RNeasy Mini Kit (Qiagen, Germany) and RNase-Free DNase Set (Qiagen). The RNA concentrations were quantified by Nanodrop ultraviolet spectrophotometer (Thermo Fisher, USA), and the quality of RNA was assessed by 2% agarose gel electrophoresis.

Microarray analysis

The Human LncRNA+ mRNA Array V3.0 (Agilent, USA) was designed with four identical arrays per slide (4 x 180K format), with each array containing probes investigating approximately 37,000 human lncRNAs and 34,000 human mRNAs. Those lncRNA target sequences were merged from multiple databases including GENCODE/ENSEMBL, Human LincRNA Catalog, RefSeq, and USCS. Microarray assays were performed by the CapitalBio Corporation (Beijing, China). RNA was converted into cDNA and then labeled with Cy3. After hybridization overnight, the microarray was washed with a Gene Expression Wash Buffer Kit (Agilent). Slides were scanned by Agilent Microarray Scanner with default settings, and data were extracted with Feature Extraction software 10.7 (Agilent).

Data analysis

The microarray data were analyzed by using the GeneSpring software V13.0 (Agilent) for data summarization, normalization, and quality control. To select the differentially expressed genes, we used threshold values of ≥ 2 and ≤ −2 -fold change (FC) and a Benjamini-Hochberg corrected p value < 0.05. The data were Log2 transformed and median centered by genes using the Adjust Data function of CLUSTER 3.0 software and then further analyzed with hierarchical clustering with average linkage. Finally, we performed tree visualization by using Java Treeview (Stanford University School of Medicine, Stanford, CA, USA).

Hierarchical clustering

Unsupervised two-way hierarchical clustering analysis was performed to classify the samples with similar patterns of gene expression and disclose the relationships between the endometrial samples.

RT-qPCR assay

Eight lncRNAs were selected to perform RT-qPCR validation due to their large fold change in the microarray study or their predicted gene functions. The primer sequences are listed in Supplemental Table 4. Total RNA extraction was described in the previous portion of this article. The reverse transcription reactions were carried out with reverse transcriptase (SuperScript III, Invitrogen, USA), and quantitative PCR reactions were then performed on ABI 7900 in a total volume of 20 μL. Every quantitative PCR reaction was performed in triplicate. The expression levels of the 7 chosen lncRNAs were normalized to β-actin and calculated using the 2^−∆∆Ct method.

Coexpression network construction

The coding-noncoding gene coexpression network (CNC network) was constructed based on the correlation analysis between the significantly differentially expressed lncRNA transcripts and mRNA transcripts in our microarray [Carlson et al. 2006; Prieto et al. 2008; Pujana et al. 2007]. For each pair of lncRNA and mRNA, the Pearson correlation was calculated, and the significant correlation pairs were chosen to construct the coexpression network. LncRNA transcripts and mRNA transcripts with Pearson correlation coefficients not less than 0.96 were selected to draw the network with Cytoscape (www.cytoscape.org). In the coexpression network analysis, degree centrality is defined as the number of links one node has to the other, which has been considered the most significant measure of the importance of a gene.

GO and KEGG enrichment analyses

Most lncRNAs have not been functionally annotated yet, and the probable functions of lncRNAs are mainly predicted by the annotation of their coexpressed mRNAs [Zhang et al. 2014a]. GO analysis [Ashburner et al. 2000] and KEGG pathway annotation [Kanehisa et al. 2012] were applied to the coexpressed targets of the dysregulated lncRNAs. GO analysis indexes the biological processes, molecular functions, and cellular components of coexpressed targets of the dysregulated lncRNAs. Meanwhile, KEGG analysis lists the pathways in which coexpressed targets of the dysregulated lncRNAs may be involved.

TF prediction

LncRNAs participating in certain biological pathways are regulated by vital TFs that regulate pathways [Lopez-Pajares et al. 2015]. TFSearch (http://www.cbrc.jp/research/db/TFSEARCH.html) was applied to predict these lncRNAs that possibly participate in pathways regulated by the corresponding TFs. The lncRNA-TF network was drawn with Cytoscape.

Declaration of interests

This work is supported by CapitalBio Technology Corporation, China (NO.RDU2011-04). The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Acknowledgments

The authors thank all the patients who participated in our study.

Supplemental material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Notes on contributors

Li-juan Fan

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Hong-jing Han

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Jing Guan

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Xiao-wei Zhang

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Qing-hua Cui

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Huan Shen

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.

Cheng Shi

Carried out the experiments, analyzed the experimental results, and wrote the manuscript: L-jF; Designed the experiments: CS, HS; Collected and filtered the endometrial samples: H-jH, JG. Checked the manuscript: X-wZ. Carried out the gene function prediction analysis: Q-HC.