Open Access

Characterization of long non-coding RNA expression profiles in lymph node metastasis of early-stage cervical cancer

  • Authors:
    • Chunliang Shang
    • Wenhui Zhu
    • Tianyu Liu
    • Wei Wang
    • Guangxin Huang
    • Jiaming Huang
    • Peizhen Zhao
    • Yunhe Zhao
    • Shuzhong Yao
  • View Affiliations

  • Published online on: March 29, 2016     https://doi.org/10.3892/or.2016.4715
  • Pages: 3185-3197
  • Copyright: © Shang et al. This is an open access article distributed under the terms of Creative Commons Attribution License.

Metrics: Total Views: 0 (Spandidos Publications: | PMC Statistics: )
Total PDF Downloads: 0 (Spandidos Publications: | PMC Statistics: )


Abstract

Pelvic lymph node metastasis (PLNM) is an independent prognostic parameter and determines the treatment strategies of cervical cancer. Increasing evidence indicates that long non-coding RNAs (lncRNAs) play a crucial role in the process of tumor biological functions. This study aimed to mine lymph node metastasis-associated lncRNAs and investigate their potential pathophysiological mechanism in cervical cancer lymph node metastasis. We applied the lncRNA-mining approach to identify lncRNA transcripts represented on Affymetrix human genome U133 plus 2.0 microarrays from Gene Expression Omnibus (GEO) and then by validation in clinical specimens. The biological role and molecular mechanism of these lncRNAs were predicted by bioinformatic analysis. Subsequently, a receiver operating characteristic (ROC) curve and survival curve were conducted to evaluate the diagnostic and prognostic value of candidate lncRNAs. In total, 234 differentially expressed lncRNAs were identified to significantly associate with pelvic lymph node metastasis in early-stage cervical cancer. Our qRT-PCR results were consistent with the mining analysis (P<0.05). The functional enrichment analysis suggested that these lncRNAs may be involved in the biological process of lymph node metastasis. The ROC curves demonstrated satisfactory discrimination power of MIR100HG and AC024560.2 with areas under the curve of 0.801 and 0.837, respectively. Survival curve also indicated that patients with high MIR100HG expression had a tendency of poor prognosis. This is the first study to successfully mine the lncRNA expression patterns in PLNM of early-stage cervical cancer. MIR100HG and AC024560.2 may be a potential biomarkers of PLNM and these lncRNAs may provide broader perspective for combating cervical cancer metastasis.

Introduction

Uterine cervical cancer is the third most common cancer and the second leading cause of cancer death in women between 20 and 39 years of age worldwide (1). The National Comprehensive Cancer Network (NCCN) (the version 1.2014) guidelines recommend the primary treatment for early-stage cervical cancer (FIGO IB and IIA) as either surgery or chemo radiotherapy. The primary chemoradiotherapy is recommended as the preferred treatment for patients with high-risk positive lymph node (2). However, accurate and efficient clinical detection methods for lymph node metastasis is difficult, so many patients receive initial unnecessary surgery with adjuvant chemoradiotherapy (3). The combination of surgery and chemoradiotherapy carries the worse morbidity, particularly long-term complications (4). Furthermore, pelvic lymph node metastasis (PLNM) has been identified as the strongest key prognostic parameter in cervical cancer, particularly early-stage cervical cancer (5); therefore, effective PLNM detection is essential to select an optimal therapy.

In clinical practice, imaging diagnostics, including magnetic resonance imaging (MRI), positron emission tomography (PET), and computed tomography (CT) scans, are conventional methods of PLNM detection before treatment. Nevertheless, additional statistical analyses demonstrate that these methods have poor sensitivity for detecting lymph node metastasis (6,7). Recently, Sentinel LN (SLN) biopsy in early-stage cervical cancer yields high diagnostic value for the status of lymph node (8), but this still need general anesthesia. Thus, a non-invasive and more accurate evaluation method is urgently needed.

Rapid advances in molecular biotechnology have increased attention to biomacromolecules or small RNAs, including microRNAs and short interfering RNAs. High-throughput technologies have been devoted to identifying lymph node-associated biomarkers at genomic levels (9) and the protein (10). Several recent studies have evaluated the gene expression profiles of PLMN in uterine cervical cancers (1113) but neglected long non-coding RNAs (lncRNAs), which are longer than 200 nt and do not encode proteins. LncRNAs have emerged as potentially powerful regulators involved in various biological processes. Accumulating evidence indicates that aberrantly expressed lncRNAs participate in the carcinogenesis and development of malignant tumors through binding proteins or modulating other short regulatory RNAs (14). In addition, lncRNAs have been identified as oncogenic RNAs that promote tumor cell invasion in uterine cervical cancer (15) and as independent predictors of overall survival (16,17). Hence, lncRNAs are expected to be excellent biomarkers for PLNM in cervical cancer.

In this present study, we used a data mining method and bioinformatics analysis to determine the PLNM-associated lncRNA profiles, which was fortuitously represented on the commonly used microarray platform. Using bioinformatics software and methods, we performed an initial exploration of the potential functional enrichment and pathway mechanisms of these lncRNAs. The candidate lncRNAs show promise to be biomarkers of diagnosis and prognosis in cervical cancer. This study provides a new frontier for uncovering the potential metastasis mechanisms to lymph node in uterine cervical cancer.

Materials and methods

GEO gene expression data

The raw PLNM gene expression data and corresponding related clinical parameters were downloaded from the publicly available GEO (http://www.ncbi.nlm.nih.gov/geo/) including GSE26511 and partial GSE2109 data. GSE26511 comprises 20 cervical cancer specimens without PLNM and 19 with PLNM. For these 39 samples, primary treatment consisted of type 3 radical hysterectomy and pelvic lymph node dissection (18). GSE2109 summarized 2,158 samples included in the Expression Project for Oncology (exp0). Based on TNM stage and primary tumor site, we selected 20 cervical cancer specimens from GSE2109, including 15 without PLNM and 5 with PLNM. All samples in these two panels were hybridized to Affymetrix human genome U133 plus 2.0 microarrays.

GeneChip Probe Re-annotation

Based on the lncRNA classification pipeline constructed in a previous study (19), we identified a number of lncRNAs represented on the Affymetrix microarrays. First, the latest version of NetAffx Annotation File (HG-U133_Plus_2 Annotations, CSV format, Release 34, 30 MB, 10/24/13) was obtained from the Affymetrix official website. This annotation file was mapped to the HG-U133_Plus_2 probe sets ID. Second, for the probe sets from the Refseq database, those IDs beginning with 'NR' were retained, and transcript IDs labeled with 'NP' were deleted. For the probe sets from the Ensembl database, the online software BioMart was applied to convert Affymetrix microarray IDs to Ensembl IDs together with the corresponding gene type. We only retained genes annotated as 'lincRNA', 'sense_intronic', 'processed_transcript', 'anti-sense', 'sense_overlapping', '3prime_overlapping_ncrna', or 'misc_RNA'. Next, based on the above two steps, we removed probe set IDs annotated as 'microRNA', 'snoRNAs', ' pseudo-genes' and other small RNAs.

Differential expression analysis

The expression level of lncRNAs is lower than that of protein-coding RNAs, and the robust multichip average (RMA) has higher detection efficiency for lncRNAs (20). Therefore, these raw CEL files were background-adjusted, normalized, and log-transformed using RMA rather than Microarray Analysis Suite 5.0 (MAS 5.0) using RMAexpress software (Windows version 1.1.0, written by Ben Bolstad). The differentially expressed probe sets were identified using a parametric two-sample t-test (with a random variance model) with a significance threshold of P<0.05 and validated using permutation testing across samples in BRB-Array Tools v4.4.0 Beta 1 (http://linus.nci.nih.gov/BRB-ArrayTools.html). We also entered the probe set data into Cluster3.0&TreeView (originally developed by Michael Eisen, Stanford University) to process the Hierarchical Clustering Analysis.

To further verify the outcome of the microarray analysis, we adopted the training-validation strategy explored by Michiels et al (21) to classify the microarray-based datasets. This method indicated that the percentage of misclassification would decrease as the number of samples in the training set increased. Thus, we defined the data set GSE26511 as the training group to identify the differentially expressed lncRNAs and GSE2109 as the validation group to evaluate misclassification.

Construction of the lncRNA-mRNA coexpression network

The Pearson correlation coefficient (PCC) and P-value were considered in the construction of an lncRNA-mRNA co expression network (22). In view of the above hierachical cluster analysis resulted from GSE26511 and significant changes, we selected top 20 differential expression lncRNAs and 189 mRNAs (fold change >1.5) to form the coexpression network based on the Pearson correlation coefficient (PCC, PCC≥0.60, P<0.05). The PCC was calculated using the coding and non-coding RNAs. For the same mRNA with different probe sets or transcripts, we used the mean value as the final gene expression value. The lncRNA-mRNA coexpression network was created using Cytoscape software (v2.6.3).

GO and KEGG pathway analysis of lncRNA-coexpressed mRNAs

Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed using Molecule Annotation System (MAS) V3.01 (CapitalBio Corp., Beijing, China). MAS provides a series of comprehensive functional annotation tools to elucidate the biological meaning of differentially expressed genes (23). GO is the product of a collaborative effort to address the need for consistent descriptions of gene products among several databases and covers the following three domains: Biological Process, Cellular Component and Molecular Function. The Fisher's exact test P-value and EASE-score were used to denote the significance of the GO enrichment terms correlated with the conditions, and the Fisher P-value was used for pathways. The lower the P-value, the more significant the GO term or pathway (24).

Gene set enrichment analysis (GSEA)

Gene set enrichment analysis (GSEA) is used to interpret gene expression data by determining the statistical significance of differences in predefined gene sets between biological states (25). In addition, GSEA can be used to find pathways that correlate to the expression of the gene. To probe the biological mechanisms of the differentially expressed lncRNAs MIR100HG and AC024560.2, all coding-gene mRNAs and these two lncRNAs were separately used to generate the expression data set. With the help of GSEA software V2.1.0 (Broad Institute, MIT, Cambridge, MA, USA), we constructed a 32,619 (genes) x39 (samples) expression matrix to perform GSEA. The predefined gene set 'c2.all.v4.0.symbols.gmt' is one of 7 major collections from the Molecular Signatures Database (MSigDB). A normalized enrichment score (NES) was calculated as the primary statistic of GSEA. In addition, the chromosome location of the lncRNAs MIR100HG and AC024560.2 was visualized using the UCSC Genome Browser (http://genome.ucsc.edu/).

Patients and tumor specimens

A total of 35 samples of fresh cervical cancer tissues were randomly collected from patients who underwent surgery at the First Affiliated Hospital of Sun Yat-sen University from May 2013 to December 2014 and provided written informed consent. All surgical specimens were immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction. The clinical study was approved by the Medical Ethics Committees at the First Affiliated Hospital of Sun Yat-sen University. All samples were confirmed pathologically, and the clinical characteristics are presented in Table I.

Table I

Clinicopathological characteristics of early-stage cervical cancer patients included in the qRT-PCR analysis.

Table I

Clinicopathological characteristics of early-stage cervical cancer patients included in the qRT-PCR analysis.

No.Age (years)FIGO stageTumor size (cm)Number of Positive PLNNumber of removed PLNLVSI DifferentiationPathological typeStromal invasion
Early-stage cervical cancer with PLNM
1  58Ib11.5138NegativeG2SCC>1/2
2  50Ib2 2   521NegativeG2SCC<1/2
3  52Ib1 1   156PositiveG1SCC<1/2
4  42IIa2 5   219NegativeG1SCC>1/2
5  35IIa2 5   215NegativeG1SCC>1/2
6  59IIa1 2   221NegativeG2SCC>1/2
7  62IIa1 3   322NegativeG1SCC>1/2
8  46Ib1 4   338PositiveG1SCC>1/2
9  48IIa1 4   136NegativeG1SCC>1/2
1058IIa1 3   153NegativeG1SCC>1/2
1135IIa1 2   172PositiveG1SCC>1/2
1243IIa1 2   221NegativeG2SCC<1/2
1343Ib1 3   217PositiveG1SCC<1/2
1449Ib2 5   416PositiveG2SCC>1/2
1534IIa1 1   217PositiveG1SCC>1/2
Early-stage cervical cancer without PLNM
1  50Ib1 3   017NegativeG1SCC<1/2
2  56Ib1 1   033NegativeG3SCC>1/2
3  44Ib1 3   0  9NegativeG2SCC>1/2
4  48Ib21.3041NegativeG2SCC<1/2
5  46IIa12.5044NegativeG1SCC>1/2
6  54IIa12.5036NegativeG2SCC>1/2
7  37Ib1 2   020NegativeG2SCC<1/2
8  39Ib1 3   022NegativeG1SCC<1/2
9  47Ib1 2   038NegativeG1SCC>1/2
1050IIa1 2   021NegativeG2SCC>1/2
1152Ib1 3   022NegativeG2SCC>1/2
1236Ib2 5   021NegativeG1SCC<1/2
1343Ib1 2   016NegativeG1SCC<1/2
1440IIa12.5012NegativeG1SCC>1/2
1541Ib1 3   019NegativeG1SCC<1/2
1640Ib1 2   025NegativeG2SCC>1/2
1740IIa1 2   015NegativeG1SCC>1/2
1838Ib1 3   0  5NegativeG1SCC<1/2
1952Ib1 4   020NegativeG1SCC>1/2
2036Ib1 2   015NegativeG2SCC<1/2
RNA extraction and quantitative real-time polymerase chain reaction (qRT-PCR) validation

Total RNA from fresh frozen tissues was extracted using RNAiso Plus reagent (Takara, Dalian, China), and complementary DNA was reverse-transcribed using PrimeScript RT Master Mix (Takara) according to the manufacturer's instructions. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed with SYBR Premix Ex Taq (Takara). All qRT-PCRs were performed in a 7500 Fast Real-time PCR System (Applied Biosystems, Carlsbad, CA). The RNA primers used in qPCR are presented in Table II.

Table II

RNA primer sequences used in our study.

Table II

RNA primer sequences used in our study.

PrimerSequence (5′ to 3′)
LINC01139F: TTCTCTCACCCTTCAAACAGC
R: ACCAAAGATGTCGCAGGACT
MIR100HGF: GGCGACATCAGACAGACAGA
R: AGGACCAGCTGAAAGGAACA
AC024560.2F: TGGGTCGCTCTGTATCTCTG
R: CGGTGGCTGTGAGTATGAAG
LINC01503F: TGGATTTTCATGCCTGCTG
R: GGCTGCATTACCAGAAAGGT

[i] F, Forward; R, Reverse.

TCGA cohorts

In order to indicate the prognosis significance of MIR100HG in early-stage cervical cancer, we used the MIR100HG expression data and clinical information from the Cancer Genome Atlas (TCGA, https://tcga-data.nci.Nih.gov/tcga/). Patients without intact follow-up and pathological data were excluded. The clinical data of all 131 patients from TCGA cohort are shown in Table III.

Table III

Clinical parameters of TCGA cohort enrolled in our study.

Table III

Clinical parameters of TCGA cohort enrolled in our study.

ParametersData
Total number of patients131
Age (years)
 Median45
 Range20–80
FIGO stage
 IA3
 IB111
 IIA17
Overall survival
 Deaths23 (17.5%)
Follow-up (years)
 Median0.65
 Range0–11.6
Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics 19.0 for Windows (Released 2010; IBM Corp., Armonk, NY, USA). Receiver operating characteristic (ROC) analysis was performed to assess the sensitivity and specificity of the measured markers. The cut-point of MIR100HG expression was defied as the median. Kaplan-Meier and the two-sided log-rank test were used to calculate the survival curves. Significance was defined at a P-value of <0.05.

Results

Gene expression data characteristics

The GSE26511 and GSE2109 series were used in our study. To ensure consistent clinical data of patients included in these two series, we listed the age and FIGO stage of patients in Table IV. Primary analysis was focused on GSE26511, which was a published dataset consisting of the largest number of specimens. The entire workflow of this process is presented in Fig. 1A.

Table IV

Characteristics of the gene expression data covariates (all squamous cell carcinoma).

Table IV

Characteristics of the gene expression data covariates (all squamous cell carcinoma).

GSE26511Pelvic lymph nodesAge at diagnosisFIGO stageGSE2109Pelvic lymph nodesAge at diagnosisFIGO stage
GSM651831Negative56.41b1GSM46919Negative60–701b1
GSM651832Negative45.81b1GSM46942Negative30–401b1
GSM651833Negative49.51b1GSM117626Negative60–701b1
GSM651834Negative34.72a  GSM152580Negative20–301b2
GSM651835Negative55.51b1GSM152635Negative20–301b1
GSM651836Negative38.51b1GSM152667Negative30–401b1
GSM651837Negative34.91b1GSM152719Negative30–401b1
GSM651838Negative47.41b1GSM179853Negative30–401b1
GSM651839Negative42.31b1GSM179907Negative50–601b1
GSM651840Negative35.81b2GSM179956Negative30–401b1
GSM651841Negative51.62a  GSM203742Negative50–601b1
GSM651842Negative72  1b2GSM325809Negative40–491b1
GSM651843Negative71  1b2GSM325835Negative30–391b1
GSM651844Negative35.91b2GSM89042Negative40–501b1
GSM651845Negative68.92a  GSM102476Negative40–501b1
GSM651846Negative47.41b2GSM76614Positive 40–501b1
GSM651847Negative31.51b1GSM89050Positive 30–401b1
GSM651848Negative72.72a  GSM152723Positive 50–602a  
GSM651849Negative39.91b1GSM152751Positive 40–501b2
GSM651850Negative50.71b1GSM203622Positive 50–601b2
GSM651851Positive 56.21b1
GSM651852Positive 29.11b1
GSM651853Positive 32.22a  
GSM651854Positive 60.61b1
GSM651855Positive 49.92a  
GSM651856Positive 34.91b2
GSM651857Positive 32.71b2
GSM651858Positive 40.41b1
GSM651859Positive 48.51b2
GSM651860Positive 37.41b1
GSM651861Positive 37  1b2
GSM651862Positive 32  1b1
GSM651863Positive 37.41b1
GSM651864Positive 45.51b2
GSM651865Positive 72.51b1
GSM651866Positive 42.31b1
GSM651867Positive 46.31b1
GSM651868Positive 34.21b2
GSM651869Positive 50.52a  
Differentially expressed lncRNA profile

We identified 3,432 probe sets (matching with 2,803 lncRNAs) via re-annotation of the Affymetrix human genome U133 plus 2.0 microarrays based on the Refseq and Ensembl databases. For the GSE26511 training group, 249 probe sets (matching with 234 lncRNAs) that were differentially expressed in cervical cancer specimens of varying lymph node metastasis status were identified using the method described above (data not shown). Of these 249 probe sets, 40 probe sets (31 lncRNAs) were upregulated in cervical cancer tissues without PLNM compared to the lymph node metastasis group, and 209 probe sets (203 lncRNAs) were downregulated. Hierarchical clustering maps of all samples from the GSE26511 training group and the GSE2109 validation group were constructed from these differentially expressed lncRNAs. GSE2109 was used to reduce the error due to the collection of specimens from cervical cancer tissues with or without lymph node metastasis (Fig. 1B and C).

LncRNA classification and distribution

The genetic location of biomolecules plays an important role in diverse biological and molecular functions (14). Using the UCSC genome browser and Ensembl database, the differentially expressed lncRNAs were characterized as lincRNA, 3prime_overlapping_ncRNA, antisense, processed_transcript, and sense_intronic.sense_overlapping based on the correlation between lncRNAs and their associated coding genes. A total of 100 and 17 lncRNAs were included in the up- and down-regulated lncRNA classifications, respectively. Each chromosome had different numbers of up- and down-regulated lncRNAs (Fig. 2E and F).

The overview of mRNA profile

Similar to the lncRNA analysis method, 2,980 probe sets (matching 228 lncRNAs) were identified as differentially expressed in cervical cancer specimens with PLNM. Among these mRNAs, 1,641 probe sets (matching 1,170 mRNAs) were upregulated in cervical cancer tissues without PLNM compared to the lymph node metastasis group, and 1,339 probe sets (matching with 1,111 mRNAs) were downregulated. In the subsequent analysis, we selected 335 mRNAs with a differential expression fold change >1.5 to construct a clear and significant network.

LncRNA-mRNA coexpression network

Although lncRNA functions have been implicated in many diseases, the biological functions of a large proportion of lncRNAs remain unknown. The tumor metastasis is a multifactor, multistep and multigene interactive process. Affymetrix human genome U133 plus 2.0 microarrays provided the expression levels of lncRNAs and mRNAs, and we constructed an lncRNA-mRNA coexpression network to interpret the potential biological roles of PLNM-associated lncRNAs in early-stage cervical cancer (26). The coexpression network revealed that one mRNA could target several lncRNAs and vice versa, suggesting a potential role of lncRNA and mRNA interactions in the process of lymph node metastasis in cervical cancer (Fig. 3).

GO and pathway analysis

To further evaluate the differentially expressed lncRNAs, we performed GO and pathway analyses of mRNAs co-expressed with lncRNAs. For biological process (Fig. 2C), i) signal transduction; ii) development; iii) regulation of transcription, DNA dependence; iv) cell adhesion; v) transcription; vi) cell differentiation; vii) cell cycle arrest; viii) positive regulation of cell proliferation; ix) nervous system development; x) and the Wnt receptor signaling pathway through β-catenin were involved. For cellular component and molecular function, nucleus and protein binding were included separately (Fig. 2A and B). KEGG pathway analysis indicated that the mRNAs co-expressed with the lncRNAs were mainly involved in the following: i) Gap junction; ii) Axon guidance; iii) melanoma; iv) Regulation of actin cytoskeleton; v) Phosphatidylinositol signaling system; vi) Adherens junction; vii) Prostate cancer; viii) MAPK signaling pathway; ix) Wnt signaling pathway; x) and Purine metabolism (Fig. 2D). The gap junction pathway plays a physiologically relevant regulator role in cervical tumor cells (27). We also identified the Wnt receptor signaling pathway through β-catenin, in agreement with the results of Noordhuis et al (18). These data imply that PLNM-lncRNAs may contribute to the lymph node metastasis process through the above mechanisms and pathways.

Bioinformatics analysis of lncRNA-MIR100HG and AC024560.2

To identify lncRNAs that play an important role in lymph node metastasis in early-stage cervical cancer, we analyzed the detailed information for two of the top lncRNAs from the upregulated group and downregulated groups (Table V). The lncRNA linc01503 had 4 transcripts, increasing the uncertainty of the primer and transcription products. MIR100HG was co-expressed with more coding genes than linc01139 in the upregulated group. In light of these findings, we selected MIR100HG and AC024560.2 as promising lncRNAs for further investigation.

Table V

The top four differentially expressed lncRNAs in cervical cancer specimens with lymph node metastasis.

Table V

The top four differentially expressed lncRNAs in cervical cancer specimens with lymph node metastasis.

ProbeSetGene symbolRegulation in N+Fold-changeParametric P-valueCo-expression coding genesTranscriptsBiotypebpChr
235599_atLINC01139Down2.050.0001141141lincRNA1540Chr1
225381_atMIR100HGDown1.96   0.002493471 sense_overlapping3082Chr11
238804_atAC024560.2Up1.610.000290931lincRNA1471Chr3
229296_atLINC01503Up1.500.009936354lincRNA901Chr9

The mir-100-let-7a-2 cluster host gene, MIR100HG, is located on chromosome 11q24.1 and is a regulator of hematopoiesis and oncogenes in the development of myeloid leukemia (28). We performed GSEA using the expression level of MIR100HG and the entire mRNA expression dataset. Based on the GSEA results, we identified 1,127 and 2,362 gene sets that were positively or negatively associated, respectively, with MIR100HG among the 3,489 'curated gene sets'. The curated gene sets were collected from various sources such as online pathway databases, publications in PubMed, and knowledge of domain experts. Gene sets with the highest NES in the positive and negative correlation groups included RICKMAN_METASTASIS_DN and, DELYS_THYROID_ CANCER_DN, respectively (Fig. 4A and B). We also estimate that MIR100HG participates in the gap junction pathway and TGF-β pathway, with NES values of 1.6079853 and −2.2482386, respectively (Fig. 4C and D). Disruption of the TGF-β/Smad signaling pathway may be conducive to the malignant progression of cervical dysplasia in human cervical cancer (29). These analysis results are consistent with our earlier data and further indicate that MIR100HG may participate in the regulation of lymph node metastasis in early-stage cervical cancer through several pathways.

AC024560.2 is an lncRNA located on chromosome 3 and is described as Homo sapiens long non-coding RNA OTTHUMT00000340266.1. We used the expression level of AC024560.2 in GSEA using the entire mRNA expression dataset. GSEA resulted in the identification of 2,823 and 666 gene sets that were positively or negatively associated, respectively, with AC024560.2 among the 3,489 'curated gene sets'. Gene sets with the highest NES in the positive and negative correlation groups included WILLERT_WNT_SIGNALING (NES=1.9029536) and KEGG_LINOLEIC_ACID_METABOLISM (NES= −1.7962813), respectively (Fig. 5A and B). A review of the literature revealed that the activation of the Wnt/β-catenin pathway promotes proliferation and tumor formation in cervical cancer cells (30), consistent with the GSEA report.

Clinical relevance

To validate the results of data mining and determine the clinical relevance of lncRNA dysregulation, we detected the expression levels of the above four lncRNAs in 35 clinical tissues with and without PLNM by qRT-PCR (Fig. 6). The qRT-PCR result was consistent with our analysis, in that expression of all 4 lncRNAs had statistical difference with the same trend (P<0.05). Next, we used existing data to access the discriminatory power for lncRNA MIR100HG and AC024560.2 by constructing ROC curves. Areas under ROC curves of two lncRNA signatures were 0.801 and 0.837, respectively (Fig. 7A and B). This suggested that MIR100HG and AC024560.2 achieve a fine diagnostic accuracy in diagnosing lymph node metastasis of cervical cancer. According to the TCGA cohort, the Kaplan-Meier analysis demonstrated that the patients with lncRNA MIR100HG high expression had poor prognosis (Fig. 7C). The prognosis value of AC024560.2 is not shown due to the lack of exact genetic match information in TCGA cohort.

Discussion

PLNM has critically significant implications for individual therapy in early-stage cervical cancer. The evaluation of PLNM status using an accurate, noninvasive method is a major focus of research. As part of the development of a molecular diagnosis for PLNM, various types of proteins and microRNAs that participate in the regulation of PLNM through various pathways have been identified. However, the clinical application of these molecules has been limited by sensitivity, sample size, regions and other factors. LncRNAs are emerging as crucial regulators of genomic activity and the expression of protein-coding genes and other non-coding transcripts, including microRNAs (14). A role of lncRNAs in oncogenesis, tumor progression, tumor cell apoptosis and cell period arrest has been confirmed in several human cancers (31). Therefore, lncRNA function may be relevant to predicting PLNM. The present study is the first to comprehensively investigate the lncRNA expression profile associated with PLNM in early-stage cervical cancer patients.

Although gene expression chip and protein microarrays have been widely used by researchers, the design of lncRNA expression microarrays is not mature. Therefore, mining lncRNA information from the most commonly used commercial microarrays in human cancer profiling, including the Affymetrix HG-U133 Plus 2.0 array, is an important supplement for this field. This approach is accurate, feasible, and low cost and has been adopted in several studies (19,32). Herein, we used this method for reference and successfully mined the ideal lncRNA expression profile associated with PLNM.

After screening, 2,803 lncRNA transcripts were filtered from the Affymetrix microarray, and 234 lncRNAs that were significantly associated with pelvic lymph node metastasis were identified. We validated these results by qRT-PCR. Using the lncRNA-mRNA CNC network and molecular analysis system, we assessed the major biological functions and molecular mechanisms in which PLNM-associated lncRNAs might be involved. lncRNA MIR100HG and AC024560.2 were further analyzed based on their location, co-expression with coding genes, gene set enrichment and clinical discriminatory power.

In our MAS report, the enrichment score of gap junctions was highest. Gap junctions are a specialized intercellular connection between two cells composed of proteins from the connexin family in vertebrates. Gap junctions allow various molecules, ions and electrical impulses to directly pass through the channel (33). The relationship between the aberrant expression of the gap junction protein connexin and lymph node metastasis has been verified for several cancers, including ovarian adenocarcinoma, human ductal breast cancer, colorectal cancer, and oral squamous cell carcinoma. Some membrane connexins are independent prognostic factors in various cancers. In uterine cervical cancers, some connexin proteins that mediate gap junction intercellular communication have been associated with carcinogenesis and tumor progression (34). The expression levels of gap junction beta-3 protein (GJA1) and gap junction alpha-10 protein (GJA10) were also significantly different in our mRNA analysis (P<0.05). These findings provide a theoretical basis to further explore this biological pathway.

The lncRNA MIR100HG was originally identified as highly expressed in acute megakaryoblastic leukemia (AMKL). MIR100HG acts as a mediator of hematopoiesis and oncogenes in the progression of AMKL, an aggressive form of hematological cancer (28). Moreover, there is an intronic coding region (BLID) in MIR100HG gene, which functions as a proapoptotic molecule through a caspase-dependent mitochondrial pathway of cell death (35). This activity provides a starting point to explore the expression pattern of MIR100HG in cervical cancer. GSEA revealed that the gene set RICKMAN_METASTASIS_UP had a higher normal enrichment score and positive correlation with the profile. This result confirms the key value of MIR100HG in cancer lymph node metastasis. Furthermore, MIR100HG is closely correlated with the gap junction and TGF-β pathways. Most of the mRNAs that were co-expressed with MIR100HG participate in the gap junction pathway. The TGF-β pathway is an important PLNM-associated pathway in cervical cancer and was analyzed by Noordhuis et al (18). Thus, our findings are consistent with previous research results. We explained the biological pathway of PLNM in cervical cancer from a novel molecular mechanism perspective and established a more comprehensive understanding based on molecular profile information.

lncRNA-AC024560.2 is a novel lncRNA that has not been previously associated with cancer. GSEA indicated an association of AC024560.2 with the linoleic acid metabolism pathway. Downregulation of peroxisome proliferator-activated receptors (PPARs) acting as nuclear receptors for linoleic acid metabolites participating in the apoptosis signaling pathway in colorectal cancer was reported as early as 2003 (36). The role of the Wnt signaling pathway, which is related to AC024560.2, was reported previously in cancer (18). Therefore, as regulators of biological function for certain proteins and microRNAs, these lncRNAs may play an important and valuable role in PLNM of early-stage cervical cancer. To increase the clinical significance of our findings, we analyzed the diagnostic capacity of MIR100HG and AC024560.2 using ROC curves. Our results provide strong evidence for the prediction of lymph node status in cervical cancer using these two lncRNAs.

However, there are limitations to our study. First, the sample size of the GSE2109 validation group was small, and thus hierarchical clustering was less obvious. Second, the lncRNAs selected here might not represent the entire lncRNA profile involved in PLNM, because the Affymetrix human genome U133 plus 2.0 microarrays did not include all the lncRNAs present. Third, only 131 patients of early-stage cervical squamous cell carcinoma from TCGA cohort were accorded with the inclusive criteria, so that the difference between survival curves was not significant. In addition, only one pathological type of cervical cancer (squamous cell cancer) was studied because of the limitation of GEO data. Finally and unfortunately, we had no more experimental evidence to further prove our analysis report. The focus of our study was the value of bioinformatics analysis in addressing important clinical topics and discovering potential role of lncRNAs in lymph node metastasis.

In summary, we successfully identified 234 differentially expressed PLNM-associated lncRNAs in early-stage cervical cancer using the data mining method. The qRT-PCR was carried out to further detect the lncRNA expression patterns in clinical cervical cancer tissues. Using the LncRNA-mRNA Coexpression network, we detailed the possible function of these lncRNAs from different perspectives, including molecular mechanism and biological pathways. Two promising lncRNAs MIR100HG and AC024560.2 were evaluated using GSEA reports and other databases to uncover their location, biological function and discrimination power. Our results fully revealed the significance of bioinformatics in analyzing clinical issues and will serve as a guide for future research. Our study also increases the understanding of lncRNAs in pathogenesis of lymph node metastasis in early-stage cervical cancer and may be a reference basis for the treatment of cervical cancer metastasis.

Acknowledgments

This study was supported by grants from the Natural Science Foundation of Guangdong Province, China (no. 2015A030313073) and Sun Yat-Sen University Clinical Research 5010 Program (no. 2007010). We thank the Gene Expression Omnibus (GEO) and TCGA database for providing their platforms and contributing their valuable data sets. The BRB-Array tools used in the study were developed by Richard Simon and the BRB-Array Tools Development Team.

References

1 

Siegel R, Ma J, Zou Z and Jemal A: Cancer statistics, 2014. CA Cancer J Clin. 64:9–29. 2014. View Article : Google Scholar : PubMed/NCBI

2 

Carlson RW, Larsen JK, McClure J, Fitzgerald CL, Venook AP, Benson AB III and Anderson BO: International adaptations of NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw. 12:643–648. 2014.PubMed/NCBI

3 

Selman TJ, Mann C, Zamora J, Appleyard TL and Khan K: Diagnostic accuracy of tests for lymph node status in primary cervical cancer: A systematic review and meta-analysis. CMAJ. 178:855–862. 2008. View Article : Google Scholar : PubMed/NCBI

4 

Landoni F, Maneo A, Colombo A, Placa F, Milani R, Perego P, Favini G, Ferri L and Mangioni C: Randomised study of radical surgery versus radiotherapy for stage Ib-IIa cervical cancer. Lancet. 350:535–540. 1997. View Article : Google Scholar : PubMed/NCBI

5 

Hosaka M, Watari H, Mitamura T, Konno Y, Odagiri T, Kato T, Takeda M and Sakuragi N: Survival and prognosticators of node-positive cervical cancer patients treated with radical hysterectomy and systematic lymphadenectomy. Int J Clin Oncol. 16:33–38. 2011. View Article : Google Scholar

6 

Dong Y, Wang X, Wang Y, Liu Y, Zhang J, Qian W and Wu S: Validity of 18F-fluorodeoxyglucose positron emission tomography/computed tomography for pretreatment evaluation of patients with cervical carcinoma: A retrospective pathology-matched study. Int J Gynecol Cancer. 24:1642–1647. 2014. View Article : Google Scholar : PubMed/NCBI

7 

Nogami Y, Banno K, Irie H, Iida M, Kisu I, Masugi Y, Tanaka K, Tominaga E, Okuda S, Murakami K, et al: The efficacy of preoperative positron emission tomography-computed tomography (PET-CT) for detection of lymph node metastasis in cervical and endometrial cancer: Clinical and pathological factors influencing it. Jpn J Clin Oncol. 45:26–34. 2015. View Article : Google Scholar

8 

Diaz JP, Gemignani ML, Pandit-Taskar N, Park KJ, Murray MP, Chi DS, Sonoda Y, Barakat RR and Abu-Rustum NR: Sentinel lymph node biopsy in the management of early-stage cervical carcinoma. Gynecol Oncol. 120:347–352. 2011. View Article : Google Scholar : PubMed/NCBI

9 

Grigsby PW, Watson M, Powell MA, Zhang Z and Rader JS: Gene expression patterns in advanced human cervical cancer. Int J Gynecol Cancer. 16:562–567. 2006. View Article : Google Scholar : PubMed/NCBI

10 

Wang W, Jia HL, Huang JM, Liang YC, Tan H, Geng HZ, Guo LY and Yao SZ: Identification of biomarkers for lymph node metastasis in early-stage cervical cancer by tissue-based proteomics. Br J Cancer. 110:1748–1758. 2014. View Article : Google Scholar : PubMed/NCBI

11 

Kim T, Choi J, Kim WY, Choi CH, Lee J, Bae D, Son D, Kim J, Park BK, Ahn G, et al: Gene expression profiling for the prediction of lymph node metastasis in patients with cervical cancer. Cancer Sci. 99:31–38. 2008.

12 

Huang L, Zheng M, Zhou QM, Zhang MY, Jia WH, Yun JP and Wang HY: Identification of a gene-expression signature for predicting lymph node metastasis in patients with early stage cervical carcinoma. Cancer. 117:3363–3373. 2011. View Article : Google Scholar : PubMed/NCBI

13 

Biewenga P, Buist MR, Moerland PD, Ver Loren van Themaat E, van Kampen AHC, ten Kate FJW and Baas F: Gene expression in early stage cervical cancer. Gynecol Oncol. 108:520–526. 2008. View Article : Google Scholar : PubMed/NCBI

14 

Bhan A and Mandal SS: Long noncoding RNAs: Emerging stars in gene regulation, epigenetics and human disease. Chem Med Chem. 9:1932–1956. 2014. View Article : Google Scholar : PubMed/NCBI

15 

Sun NX, Ye C, Zhao Q, Zhang Q, Xu C, Wang SB, Jin ZJ, Sun SH, Wang F and Li W: Long noncoding RNA-EBIC promotes tumor cell invasion by binding to EZH2 and repressing E-cadherin in cervical cancer. PLoS One. 9:e1003402014. View Article : Google Scholar : PubMed/NCBI

16 

Cao S, Liu W, Li F, Zhao W and Qin C: Decreased expression of lncRNA GAS5 predicts a poor prognosis in cervical cancer. Int J Clin Exp Pathol. 7:6776–6783. 2014.PubMed/NCBI

17 

Huang L, Liao LM, Liu AW, Wu JB, Cheng XL, Lin JX and Zheng M: Overexpression of long noncoding RNA HOTAIR predicts a poor prognosis in patients with cervical cancer. Arch Gynecol Obstet. 290:717–723. 2014. View Article : Google Scholar : PubMed/NCBI

18 

Noordhuis MG, Fehrmann RSN, Wisman GBA, Nijhuis ER, van Zanden JJ, Moerland PD, Ver Loren, van Themaat E, Volders HH, Kok M, ten Hoor KA, et al: Involvement of the TGF-beta and beta-catenin pathways in pelvic lymph node metastasis in early-stage cervical cancer. Clin Cancer Res. 17:1317–1330. 2011. View Article : Google Scholar : PubMed/NCBI

19 

Zhang X, Sun S, Pu JKS, Tsang ACO, Lee D, Man VOY, Lui WM, Wong STS and Leung GKK: Long non-coding RNA expression profiles predict clinical phenotypes in glioma. Neurobiol Dis. 48:1–8. 2012. View Article : Google Scholar : PubMed/NCBI

20 

Yang F, Zhang L, Huo XS, Yuan JH, Xu D, Yuan SX, Zhu N, Zhou WP, Yang GS, Wang YZ, et al: Long noncoding RNA high expression in hepatocellular carcinoma facilitates tumor growth through enhancer of zeste homolog 2 in humans. Hepatology. 54:1679–1689. 2011. View Article : Google Scholar : PubMed/NCBI

21 

Michiels S, Koscielny S and Hill C: Prediction of cancer outcome with microarrays: A multiple random validation strategy. Lancet. 365:488–492. 2005. View Article : Google Scholar : PubMed/NCBI

22 

Adler J and Parmryd I: Quantifying colocalization by correlation: The Pearson correlation coefficient is superior to the Mander's overlap coefficient. Cytometry A. 77:733–742. 2010. View Article : Google Scholar : PubMed/NCBI

23 

Liu JB, Dai CM, Su XY, Cao L, Qin R and Kong QB: Gene microarray assessment of multiple genes and signal pathways involved in androgen-dependent prostate cancer becoming androgen independent. Asian Pac J Cancer Prev. 15:9791–9795. 2014. View Article : Google Scholar : PubMed/NCBI

24 

Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, et al The Gene Ontology Consortium: Gene ontology: Tool for the unification of biology. Nat Genet. 25:25–29. 2000. View Article : Google Scholar : PubMed/NCBI

25 

Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, Paulovich A, Pomeroy SL, Golub TR, Lander ES, et al: Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA. 102:15545–15550. 2005. View Article : Google Scholar : PubMed/NCBI

26 

Liao Q, Liu C, Yuan X, Kang S, Miao R, Xiao H, Zhao G, Luo H, Bu D, Zhao H, et al: Large-scale prediction of long non-coding RNA functions in a coding-non-coding gene co-expression network. Nucleic Acids Res. 39:3864–3878. 2011. View Article : Google Scholar : PubMed/NCBI

27 

Macdonald AI, Sun P, Hernandez-Lopez H, Aasen T, Hodgins MB, Edward M, Roberts S, Massimi P, Thomas M, Banks L, et al: A functional interaction between the MAGUK protein hDlg and the gap junction protein connexin 43 in cervical tumour cells. Biochem J. 446:9–21. 2012. View Article : Google Scholar : PubMed/NCBI

28 

Emmrich S, Streltsov A, Schmidt F, Thangapandi VR, Reinhardt D and Klusmann JH: LincRNAs MONC and MIR100HG act as oncogenes in acute megakaryoblastic leukemia. Mol Cancer. 13:1712014. View Article : Google Scholar : PubMed/NCBI

29 

Iancu IV, Botezatu A, Goia-Ruşanu CD, Stănescu A, Huică I, Nistor E, Anton G and Pleşa A: TGF-beta signalling pathway factors in HPV-induced cervical lesions. Roum Arch Microbiol Immunol. 69:113–118. 2010.

30 

Chen Q, Cao HZ and Zheng PS: LGR5 promotes the proliferation and tumor formation of cervical cancer cells through the Wnt/β-catenin signaling pathway. Oncotarget. 5:9092–9105. 2014. View Article : Google Scholar : PubMed/NCBI

31 

Yang G, Lu X and Yuan L: LncRNA: A link between RNA and cancer. Biochim Biophys Acta. 1839:1097–1109. 2014. View Article : Google Scholar : PubMed/NCBI

32 

Hu Y, Chen HY, Yu CY, Xu J, Wang JL, Qian J, Zhang X and Fang JY: A long non-coding RNA signature to improve prognosis prediction of colorectal cancer. Oncotarget. 5:2230–2242. 2014. View Article : Google Scholar : PubMed/NCBI

33 

Lampe PD and Lau AF: The effects of connexin phosphorylation on gap junctional communication. Int J Biochem Cell Biol. 36:1171–1186. 2004. View Article : Google Scholar : PubMed/NCBI

34 

Steinhoff I, Leykauf K, Bleyl U, Dürst M and Alonso A: Phosphorylation of the gap junction protein Connexin43 in CIN III lesions and cervical carcinomas. Cancer Lett. 235:291–297. 2006. View Article : Google Scholar

35 

Broustas CG, Gokhale PC, Rahman A, Dritschilo A, Ahmad I and Kasid U: BRCC2, a novel BH3-like domain-containing protein, induces apoptosis in a caspase-dependent manner. J Biol Chem. 279:26780–26788. 2004. View Article : Google Scholar : PubMed/NCBI

36 

Shureiqi I, Jiang W, Zuo X, Wu Y, Stimmel JB, Leesnitzer LM, Morris JS, Fan HZ, Fischer SM and Lippman SM: The 15-lipoxygenase-1 product 13-S-hydroxyoctadecadienoic acid down-regulates PPAR-delta to induce apoptosis in colorectal cancer cells. Proc Natl Acad Sci USA. 100:9968–9973. 2003. View Article : Google Scholar : PubMed/NCBI

Related Articles

Journal Cover

June-2016
Volume 35 Issue 6

Print ISSN: 1021-335X
Online ISSN:1791-2431

Sign up for eToc alerts

Recommend to Library

Copy and paste a formatted citation
x
Spandidos Publications style
Shang C, Zhu W, Liu T, Wang W, Huang G, Huang J, Zhao P, Zhao Y and Yao S: Characterization of long non-coding RNA expression profiles in lymph node metastasis of early-stage cervical cancer. Oncol Rep 35: 3185-3197, 2016.
APA
Shang, C., Zhu, W., Liu, T., Wang, W., Huang, G., Huang, J. ... Yao, S. (2016). Characterization of long non-coding RNA expression profiles in lymph node metastasis of early-stage cervical cancer. Oncology Reports, 35, 3185-3197. https://doi.org/10.3892/or.2016.4715
MLA
Shang, C., Zhu, W., Liu, T., Wang, W., Huang, G., Huang, J., Zhao, P., Zhao, Y., Yao, S."Characterization of long non-coding RNA expression profiles in lymph node metastasis of early-stage cervical cancer". Oncology Reports 35.6 (2016): 3185-3197.
Chicago
Shang, C., Zhu, W., Liu, T., Wang, W., Huang, G., Huang, J., Zhao, P., Zhao, Y., Yao, S."Characterization of long non-coding RNA expression profiles in lymph node metastasis of early-stage cervical cancer". Oncology Reports 35, no. 6 (2016): 3185-3197. https://doi.org/10.3892/or.2016.4715