Original Article Potential biomarkers and their regulatory ......Tissues were immediately frozen in liquid nitro-gen and shred using a homogenizer (IKA, Staufen, Germany). For the

Int J Clin Exp Pathol 2016;9(5):5103-5116www.ijcep.com /ISSN:1936-2625/IJCEP0023432

Original Article

Potential biomarkers and their regulatory relationships in laryngeal squamous cell carcinoma with lymph node metastasis revealed by integrating mRNA, microRNA and long non-coding RNA profiles

Wei Gao1,2*, Chunming Zhang1,2*, Teng Ma3, Shuxin Wen1,2, Rong Fu2, Dan Zhao2, Yongyan Wu2,4,5, Binquan Wang1,2

1Department of Otolaryngology, Head & Neck Surgery, The First Hospital Affiliated With Shanxi Medical University, China; 2Shanxi Key Laboratory of Otolaryngology Head & Neck Cancer, Taiyuan, Shanxi, China; 3Department of Radiation Toxicology and Oncology, Beijing Institute of Radiation Medicine, Beijing, China; 4College of Veterinary Medicine, Northwest A&F University and 5Key Laboratory of Animal Biotechnology, Ministry of Agriculture, Northwest A&F University, Yangling, Shaanxi, China. *Co-first authors.

Received January 6, 2016; Accepted March 20, 2016; Epub May 1, 2016; Published May 15, 2016

Abstract: Mechanisms about regulation of mRNAs, microRNAs (miRNAs), and long non-coding RNAs (lncRNA) in cervical lymph node metastatic laryngeal squamous cell carcinoma (LSCC) remained unclear. Two sets of three microarrays of mRNA, miRNA and lncRNA in fresh surgical LSCC samples were established, respectively from three paired tumor and adjacent normal mucosa epithelial tissues in cN + (Positive metastasis of cervical lymph node) patients, and three paired tumor and adjacent normal tissues in cN0 (non-cervical lymph node metastatic) pa-tients. Bioinformatic methods including differentiation analysis and enrichment analysis were recruited to identify crucial mRNAs, miRNAs and lncRNAs. Pearson Correlation Coefficients, TargetScan 6.2 and Miranda software were used to establish lncRNA-mRNA, miRNA-mRNA and lncRNA-miRNA regulatory networks. Venn analysis was per-formed to select specific integrated network for cervical lymph node metastatic LSCC. In total, 2253/600 differen-tially expressed (DE)-mRNAs, 45/34 DE-miRNAs and 71/48 DE-lncRNAs were identified as specific for lymph node metastatic/non-metastatic LSCC samples. Cell cycle and proteasome pathways were over-represented for specific mRNAs in lymph node metastatic samples. In the interacted networks, PVT1 was associated with the downregu-lated hsa-miR-1207-5p and the upregulated G6PD, which was predicted targeted by hsa-miR-1207-5p; whereas LOC157273 was closely related to the downregulated hsa-miR-145-5p and upregulated CDK4, SEH1L and SMC1A, which were predicted targeted by hsa-miR-145-5p. Several important mRNAs (G6PD, SMC1A and CDK4), miRNAs (miR-1207-5p and hsa-miR-145-5p) and lncRNAs (PVT1 and LOC157273) were identified as potential biomarkers for lymph node metastatic LSCC diagnosis. Additionally, PVT1 might regulate miR-1207-5p, whose potential target was G6PD; and SMC1A and CDK4 might be targets of hsa-miR-145-5p.

Keywords: Laryngeal squamous cell carcinoma, lymph node metastasis, mRNA, microRNA, long non-coding RNA, cell cycle

Introduction

Laryngeal squamous cell carcinoma (LSCC) is the most frequent type of Head and neck squa-mous cell carcinomas (HNSCC), which is the fifth frequent malignancy worldwide [1, 2]. It is the second predominant upper respiratory tract tumor with high mortality and is able to spread to regional cervical lymph nodes [3]. Though advanced technologies have been de-

veloped in the diagnosis and prevention of the disease, the survival rates were not significantly improved or even decreased in the United State [4, 5]. The overall 5-year survival rate for LSCC was 64.2% in a meta-analysis [6]. Cervical lymph nodes metastases is believed to be one major reason of the poor prognosis in LSCC patients [7]. Additionally, in patients with lymph node metastases, the 5-year survival rate was less than 50% [8].

http://www.ijcep.com

Biomarkers and their regulatory relationship in lymph node metastatic LSCC

5104 Int J Clin Exp Pathol 2016;9(5):5103-5116

Therefore it is necessary to explore new strate-gies such as identifying novel biomarkers asso-ciated with metastasis for prediction and prog-nosis of LSCC. In previous studies, a spectrum of gene makers related to LSCC have been pro-posed such as matrix metalloproteinase-2 (MMP-2) [9], cluster of differentiation 24 (CD24) [10] and U3 small nucleolar ribonucleoprotein protein (IMP3) [11]. Also, several achievements have been made in the identification of bio-markers linked to lymph node metastasis in LSCC, including RPN2 and eukaryotic transla-tion initiation factor 3, subunit A (eIF3a) [2].

Known as a class of small non-coding RNAs, microRNAs (miRNAs) normally serve as regula-tors of gene expression at post-transcriptional level, and play significant roles in various human cancers [1, 12]. MiR-34a was reported to involve in cell proliferation suppression and thus influence the occurrence of LSCC [13]. The elevated expression of miR-1297 was discov-ered in LSCC, and further experiments in vitro indicated that this miRNA suppressed the expression of its target PTEN, a tumor suppres-sor gene whose dysregulation might result in tumor metastasis and proliferation [14].

Recently, increasing attentions have been paid to the long non-coding RNAs (lncRNAs), which have the length of more than 200 nt and are widely distributed in human genome [15]. Different with miRNAs, lncRNAs could exert its regulatory functions at transcriptional and post-transcriptional levels [16]. Inspiringly, the roles of several potential lncRNA markers have been elucidated. The lncRNA Metastasis-associated lung adenocarcinoma transcript 1 (MALAT-1), overexpressed in LSCC tumor tissues and closely related to metastasis, was involved in the regulation of cell apoptosis [17]. To date, the elaborated lncRNA markers with experi-mental validations were AC026166.2-001 and RP11-169D4.1-001 [15].

However, the potential regulatory correlations among mRNAs, miRNAs and lncRNAs as well as the comparisons between cervical lymph node metastatic and non-metastatic LSCC, are rarely reported. In the present study, we established two microarray sets of mRNAs, miRNAs and lncRNAs, respectively obtained from three paired cervical lymph node meta-static LSCC tissues and adjacent normal muco-sa epithelial tissues in cN1 (sentinel lymph

node) patients, and from three paired LSCC tis-sues and adjacent normal mucosa epithelial tissues in cN0 (non-cervical lymph node meta-static) patients. Bioinformatic methods were recruited to screen differently expressed mRNAs, miRNAs and lncRNAs (DE-mRNAs, DE-miRNAs and DE-lncRNAs) and seek their potential regulatory relationships. Following that, venn analysis was performed to further explore the metastatic specific integrated net-work, attempting to uncover the underlying mechanisms in the progression of LSCC metas-tasis as well as provide novel related miRNA or lncRNA biomarkers.

Material and methods

Patient sample collection

All the LSCC patients recruited in this study did not undergo prior radiotherapy, chemotherapy, biotherapy and other molecular targeting thera-py, and were obtained from the Department of Otolaryngology Head and Neck Surgery of The first Hospital Affiliated with Shanxi Medical University. LSCC and the lymph node metasta-sis were all diagnosed using frozen section by two different senior pathologists. Paired tumor and control (adjacent normal mucosa epithelial tissues) were obtained during surgery. All the patients were ensured without the history of hepatitis A, B and C, human papillomavirus (HPV), human immunodeficiency virus (HIV), syphilis, other chronic diseases and familial genetic disorder. We also determined the tumor stages, with the criteria of tumor-node-metas-tasis (TNM), which was based on American Joint Committee on Cancer (AJCC). Detailed characteristics of patients are presented in Table 1. This study was conducted according to the Helsinki declaration and was approved by the Research Ethics Committee at Shanxi Medical University. All the patients signed the informed consent.

Total RNA extraction

Tissues were immediately frozen in liquid nitro-gen and shred using a homogenizer (IKA, Staufen, Germany). For the mRNA and lncRNA microarray preparation, the total RNA was extracted by TRIzol reagent (Invitrogen, Carlsbad, CA), while the total RNA for miRNA microarray was abstracted and purified applying mirVana™ miRNA Isolation Kit (Ambion, Austin, TX, US)



and miRNeasy Mini Kit (QIAGEN, GmBH, Germany). The samples were qualified and examined recruiting an Agilent Bioanalyzer 2100 (Agilent technologies, Santa Clara, CA, US). We evaluated the purity of total RNA through the detection of the absorbance ratio at 260 to 280 nm.

Microarray analysis

The RNA was amplified and transferred to cDNA after purification. Subsequently, based on the relevant instructions of the manufacture, the cDNA was labeled and homogenized. For mRNA and lncRNA microarray, the labeling was per-formed by Agilent Quick Amp Labeling Kit (Agilent Technologies, Santa Clara, CA) and the hybridization was conducted on the Agilent Human LncRNA Microarray (4*180 K) V19.0 via Agilent Gene Expression Hybridization Kit (Agilent technologies, Santa Clara, CA, US), in a SureHyb Hybridization Chamber (Agilent) at 65°C for 17 h. When the washing was complet-ed, the slides were scanned with an Agilent DNA Microarray Scanner (Agilent technologies, Santa Clara, CA, US), and afterwards the data were abstracted using Agilent Feature Extraction software. Further data analysis was conducted using Agilent GeneSpring GX v12.0 software (Agilent technologies, Santa Clara, CA, US).

For miRNA microarray, the labeling was per-formed by miRNA Complete Labeling and Hyb Kit (Agilent technologies, Santa Clara, CA, US) and hybridization was conducted with 100 ng Cy3-labeled RNA using miRNA Complete Labeling and Hyb Kit (Agilent technologies, Santa Clara, CA, US) on the Agilent Human MiRNA Microarray (8*60 K) V19.0. Based on the manufacture’s instruction, the hybridization was carried out in an Oven (Agilent technolo-gies, Santa Clara, CA, US) at 55°C for 20 h.

Afterwards, slides were washed in staining dishes (Thermo Shandon, Waltham, MA, US) using Gene Expression Wash Buffer Kit (Agilent technologies, Santa Clara, CA, US), followed by scanning through Agilent Microarray Scanner (Agilent technologies, Santa Clara, CA, US) and Feature Extraction software 10.7 (Agilent tech-nologies, Santa Clara, CA, US). Raw data were normalized by Quantile algorithm, Gene Spring Software 11.0 (Agilent technologies, Santa Clara, CA, US).

Identification of DE-mRNAs, DE-miRNAs and DE-lncRNAs

All the microarray data were preprocessed using Affy package in Bioconductor [18]. The raw data were subjected to background correction, quantile data normalization and probe summarization recruiting the Robust Multi-array Average (RMA) algorithm [19]. DE-mRNAs, DE-miRNAs and DE-lncRNAs between cervical lymph node metastatic LSCC and control samples, lymph node non-metastatic LSCC and control samples were screened based on t-test using linear models for microarray data (limma) package of Bioconductor R (http://www.bioconductor.org/packages/release/bioc/html/limma.html) [20]. The threshold for selection were P < 0.05 and |log2FC (fold change)| > 1. Moreover, by comparing these two kinds of DE-mRNAs, DE-miRNAs and DE-lncRNAs using Venn analy-sis [21], the common or specific ones in cervi-cal lymph node metastatic and non-metastatic LSCC tissues were selected for further analyses.

Functional enrichment analyses for DE-mRNAs

The DE-mRNAs were mapped to gene ontology (GO, http: //www. geneontology. org/) [22] and

Table 1. Clinical parameters of the six LSCC patients

Samples Hospitalization starting time Gender Age Types Differentiation TNM

StageClinical stage

C1/N1 2012/12/5 Female 61 Supraglottic LSCC Poorly differentiated T1N0M0 IC2/N2 2012/6/10 Male 75 Supraglottic LSCC Poorly differentiated T2N0M0 IIC3/N3 2013/1/12 Male 58 Glottic LSCC Poorly differentiated T3N0M0 IIIC4/N4 2012/8/13 Male 60 Glottic LSCC Highly differentiated T4N1M0 IVC5/N5 2013/4/28 Male 73 Supraglottic LSCC Moderately differentiated T2N1M0 IIIC6/N6 2012/5/4 Male 54 Supraglottic LSCC Moderately differentiated T4N2M0 IVLSCC, laryngeal squamous cell carcinoma; C, LSCC tissues (1-3 represent lymph node non metastatic tissues, 4-5 represent cervical lymph node metastatic tissues); N, the paired adjacent normal tissues; TNM, tumor-node-metastasis.



Figure 1. The heat map of clustering analysis of differentially expressed (DE) mRNAs, miRNAs and lncRNAs. X-axis represents samples, while Y-axis represents mRNAs, miRNAs and lncRNAs. A: DE-mRNAs in lymph node metastatic laryngeal squamous cell carcinoma (LSCC); C: DE-miRNAs in lymph node metastatic LSCC; E: DE-lncRNAs in lymph node metastatic LSCC; B: DE-mRNAs in lymph node non-metastatic LSCC; D: DE-miRNAs in lymph node non-meta-static LSCC; F: DE-lncRNAs in lymph node non-metastatic LSCC.



Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/pathway.html) [23] databases to recognize their poten-tial functions and pathways, with the method of Fisher test [24] using the tool of Database for Annotation, Visualization and Integrated Discovery (DAVID, http://david.abcc.ncifcrf.gov/) [25]. The P values were adjusted into false discovery rate (FDR), and the criteria for significantly enriched GO terms and path-ways were both FDR < 0.05. For the common or specific DE-mRNAs in cervical lymph node metastatic and non-metastatic LSCC tissues, functional enrichment analyses were also performed.

Correlation analysis of mRNA, miRNA and ln-cRNA

The potential interaction pairs between DE-lncRNA and DE-mRNAs with the Pearson Correlation Coefficients (absolute value) > 0.95 [26] were screened out to establish the co-expression network between them.

The target genes of the DE-miRNAs were predicted by TargetScan 6.2 (http://www.tar-getscan.org) database [27]. Then combining with the DE-mRNA information, the interaction pairs of miRNA-mRNA (which had reverse expression patterns) were recognized to con-struct a regulatory network between them.

By recruiting miRanda software (http://www.microrna.org/microrna/getDownloads.do) [28], the correlations between DE-miRNA and DE-lncRNA were predicted and the interaction pairs of miRNA-lncRNA with inverse expression patterns were selected to construct the res- ponding network.

Based on the interaction pairs in the above three networks, the interactions between mRNA and lncRNA sharing common miRNA were predicted. Afterwards, the integrated reg-ulatory network of mRNA-miRNA-lcnRNA was established.

Furthermore, we conducted parallel compari-sons between all the networks separately built for lymph node metastatic and non-metastatic LSCC tissues, to recognize the common inter-action pairs and specific ones and then estab-lished the corresponding mRNA-miRNA-lcnRNA networks.

All the networks were visualized by Cytoscape (http://cytoscape.org/) [29].

Results

Screening of DE-mRNAs, DE-miRNAs and DE-lncRNAs

Based on the t-test method using limma pack-age, there identified a total of 1183/2836 DE-mRNAs, 52 (30 up-regulated and 22 down-regulated)/63 (34 up-regulated and 29 down-regulated) DE-miRNAs and 81/104 DE-lncRNAs in lymph node metastatic/non-metastatic LSCC tissues, compared with control samples, as shown in the heat map of clustering analysis (Figure 1). By comparing the two sets of DE-mRNAs, DE-miRNAs and DE-lncRNAs, a total of 583 DE-mRNAs, 18 DE-miRNAs and 33 DE-lncRNAs were identified as common ones, while 2253/600 DE-mRNAs, 45/34 DE-miRNAs and 71/48 DE-lncRNAs were deemed as the specific ones in lymph node metastatic and non-metastatic LSCC samples, respectively.

Enriched functions and pathways for DE-mRNAs

GO and KEGG pathway enrichment analyses were conducted for the common and specific DE-mRNAs identified in lymph node metastatic and non-metastatic LSCC samples, respective-ly. As a result, the significant GO terms for specific DE-mRNAs in lymphatic metastasis LSCC included cell cycle related issues (GO: 0000278, GO: 0051301 and GO: 0000087), protein heterodimerization activity (GO: 0046- 982) and protein binding (GO: 0005515), while that in non-lymphatic metastasis LSCC tissues embraced cell adhesion (GO: 0007155), extra-cellular region (GO: 0005576) and AMP deami-nase activity (GO: 0003876) (Table 2). For the overlapped DE-mRNAs in both lymphatic metastasis and non-lymphatic metastasis LSCC tissues, the significant enriched GO terms were extracellular matrix structural constituent (GO: 0005201), proteinaceous extracellular matrix (GO: 0005578) and mitotic cell cycle (GO: 0000278) (Table 2). With regards to the path-ways, 12 pathways such as viral carcinogene-sis, proteasome and cell cycle were enriched for the specific DE-mRNAs in lymphatic metas-tasis LSCC, while 15 pathways such as cell adhesion molecules (CAMs), ECM-receptor interaction, small cell lung cancer and arrhyth-mogenic right ventricular cardiomyopathy (ARVC) were over-presented amongst the spe-cific DE-mRNAs in non-lymphatic metastasis



Table 2. Significant GO terms for DE-mRNAs in lymphatic metastasis and non-lymphatic metastasis LSCC (top 5)

Category Term Count P-Value FDRLymph node metastatic LSCC BP GO: 0000278~mitotic cell cycle 87 2.59199E-13 9.15748E-10

BP GO: 0051301~cell division 69 1.09578E-07 0.000153002

BP GO: 0060337~type I interferon-mediated signaling pathway

26 1.29919E-07 0.000153002

BP GO: 0000087~M phase of mitotic cell cycle 31 4.33335E-07 0.000339685

BP GO: 0000236~mitotic prometaphase 29 4.81476E-07 0.000339685

CC GO: 0005654~nucleoplasm 161 4.00857E-09 2.33299E-06

CC GO: 0005829~cytosol 330 1.26826E-07 3.69065E-05

CC GO: 0000786~nucleosome 24 1.56763E-06 0.000304119

CC GO: 0000777~condensed chromosome kinetochore 22 1.55833E-05 0.002267366

CC GO: 0005694~chromosome 24 8.59121E-05 0.010000169

MF GO: 0046982~protein heterodimerization activity 63 9.86116E-05 0.079094272

MF GO: 0005515~ protein binding 611 0.000132486 0.079094272

MF GO: 0003725~double-stranded RNA binding 13 0.001753337 0.639062494

MF GO: 0004298~threonine-type endopeptidase activity 8 0.003387402 0.639062494

MF GO: 0032393~MHC class I receptor activity 7 0.004296241 0.639062494

Non- lymph node metastatic LSCC BP GO: 0007155~cell adhesion 26 0.000193068 0.277052439

BP GO: 0043101~purine-containing compound salvage 4 0.00052869 0.30003844

BP GO: 0045578~negative regulation of B cell differentia-tion

3 0.000627258 0.30003844

BP GO: 0061180~mammary gland epithelium develop-ment

3 0.001298148 0.327228705

BP GO: 0071560~cellular response to transforming growth factor beta stimulus

5 0.001540479 0.327228705

CC GO: 0005576~extracellular region 64 0.000149173 0.037889837

CC GO: 0005788~endoplasmic reticulum lumen 11 0.002764353 0.351072807

CC GO: 0016942~insulin-like growth factor binding protein complex

2 0.005243474 0.378368834

CC GO: 0031143~pseudopodium 3 0.006865974 0.378368834

CC GO: 0031093~platelet alpha granule lumen 5 0.007448205 0.378368834

MF GO: 0043027~cysteine-type endopeptidase inhibitor activity involved in apoptotic process

4 0.002292289 0.30080397

MF GO: 0031732~CCR7 chemokine receptor binding 2 0.00298602 0.30080397

MF GO:0003876~AMP deaminase activity 2 0.004901537 0.30080397

MF GO:0005201~extracellular matrix structural constitu-ent

6 0.005991223 0.30080397

MF GO: 0070061~fructose binding 2 0.007241483 0.30080397

Common BP GO: 0000278~mitotic cell cycle 34 3.67755E-10 5.67079E-07

BP GO: 0030198~extracellular matrix organization 22 3.20136E-09 2.46825E-06

BP GO: 0006260~DNA replication 21 1.79475E-08 9.22502E-06

BP GO: 0006271~DNA strand elongation involved in DNA replication

10 9.81425E-08 3.78339E-05

BP GO: 0000084~S phase of mitotic cell cycle 16 4.89256E-07 0.000150887

CC GO: 0031012~extracellular matrix 22 1.55311E-08 4.93888E-06

CC GO: 0042555~MCM complex 6 1.20137E-06 0.000127918

CC GO: 0005615~extracellular space 49 1.20677E-06 0.000127918

CC GO: 0000776~kinetochore 10 1.86817E-05 0.001485193

CC GO: 0005578~proteinaceous extracellular matrix 18 3.49719E-05 0.002224211

MF GO: 0003777~microtubule motor activity 13 7.95852E-07 0.000421006

MF GO: 0008201~heparin binding 14 3.61926E-05 0.009572936

MF GO: 0003678~DNA helicase activity 6 0.000116293 0.020506412

MF GO: 0005201~extracellular matrix structural constitu-ent

8 0.000758276 0.100281966

MF GO: 0017111~nucleoside-triphosphatase activity 9 0.000956806 0.101230105GO, gene ontology; LSCC, laryngeal squamous cell carcinoma; BP, biological process; CC, cellular components; MF, molecular function; Count, number of differentially expressed mRNA in a specific term; FDR, false discovery rate.



LSCC. The overlapped DE-mRNAs between these two kinds of samples were predominant-ly enriched in 11 pathways such as focal adhesion, ECM-receptor interaction, mismatch repair, protein digestion and absorption and small cell lung cancer (Table 3).

Correlation analysis of mRNA, miRNA and ln-cRNA

The interaction networks of lncRNA-mRNA, miRNA-mRNA and miRNA-lncRNA were estab-lished for metastatic and non-metastatic sam-ples, respectively. In the networks, the ‘degree’ of a node (DE-lncRNA, DE-mRNA, or DE-miRNA) represents the number of interactions with other nodes. The most highly connected nodes (with a high degree) were considered to be the network ‘hubs’. After venn analysis, the specific and common networks were extracted.

For metastatic samples the specific lncRNA-mRNA network was comprised of 316 nodes and 874 edges with predominant hub lcnRNAs such as LOC157273 (degree = 57, interacted with miRNAs such as hsa-miR-145-5p and hsa-miR-4532), TPTE2P5 (degree = 28), PVT1 (degree = 46, interacted with hsa-miR-1207-5p and hsa-miR-143-3p) and LINC00640 (degree = 21), while the specific miRNA-lncRNA network was consisted of 109 nodes and 463 edges, embracing the hub lncRNAs such as RNA45S5 (degree = 20), LOC728377 (degree = 14) and LOC729732 (degree = 10) (Figure 2).

By contrast, for non-metastatic samples the specific lncRNA-mRNA network was consisted of 62 nodes and 70 edges, in which hub lncRNAs such as LOC643837 (degree = 6), LOC401164 (degree = 8) and CT49 (degree = 7) were highlighted, whereas in the specific miRNA-lncRNA network, 70 nodes and 189

Table 3. Significant over-presented pathways for DEGs in lymphatic metastasis and non-lymphatic metastasis LSCC (top ten)Category Pathway ID Pathway Term Count P-Value FDRLymph node metastatic LSCC PATH:05322 Systemic lupus erythematosus 44 7.66E-08 1.84585E-05

PATH:05034 Alcoholism 48 1.97E-06 0.000237039

PATH:05203 Viral carcinogenesis 46 0.000138 0.011074786

PATH:03050 Proteasome 15 0.000818 0.049256664

PATH:03030 DNA replication 13 0.001158 0.05580516

PATH:00240 Pyrimidine metabolism 24 0.002565 0.103016861

PATH:04110 Cell cycle 27 0.003681 0.126739329

PATH:03013 RNA transport 30 0.014665 0.441790181

PATH:03440 Homologous recombination 8 0.028306 0.726838393

PATH:03460 Fanconi anemia pathway 12 0.030159 0.726838393

Non- lymph node metastatic LSCC PATH:04512 ECM-receptor interaction 9 0.000654 0.106099001

PATH:04514 Cell adhesion molecules (CAMs) 11 0.001192 0.106099001

PATH:05144 Malaria 6 0.003067 0.181948833

PATH:05412 Arrhythmogenic right ventricular cardiomyopathy (ARVC) 7 0.0044 0.195791255

PATH:05200 Pathways in cancer 16 0.015871 0.483007823

PATH:05150 Staphylococcus aureus infection 5 0.017294 0.483007823

PATH:04916 Melanogenesis 7 0.019571 0.483007823

PATH:05146 Amoebiasis 7 0.024334 0.483007823

PATH:05322 Systemic lupus erythematosus 8 0.029801 0.483007823

PATH:05222 Small cell lung cancer 6 0.029896 0.483007823

Common PATH:03030 DNA replication 10 3.62E-07 6.27067E-05

PATH:04110 Cell cycle 15 7.12E-06 0.00061566

PATH:04512 ECM-receptor interaction 10 0.000261 0.015035502

PATH:03430 Mismatch repair 4 0.005477 0.236901812

PATH:04974 Protein digestion and absorption 7 0.009596 0.332026564

PATH:04510 Focal adhesion 12 0.013616 0.392580959

PATH:05146 Amoebiasis 7 0.032787 0.757183804

PATH:05222 Small cell lung cancer 6 0.038953 0.757183804

PATH:05414 Dilated cardiomyopathy 6 0.044312 0.757183804

PATH:05200 Pathways in cancer 15 0.047133 0.757183804LSCC, laryngeal squamous cell carcinoma; Count, number of differentially expressed mRNA in a specific term; FDR, false discovery rate.



edges were involved, in which hub lncRNAs were DTX2P1-UPK3BP1-PMS2P11 (degree = 9), COL6A4P2 (degree = 10), LOC643837 (degree = 9) and SNORD115-9 (degree = 7) (Figure 3).

For the overlapped DE-mRNAs, DE-miRNAs and DE-lncRNAs, the common lncRNA-mRNA net-work embraced 13 nodes and 10 edges, in which the lncRNA CHEK2P2 was predicted as a central node correlating to multiple genes

such as E2F1, PCNA, CHEK2, HIST1H4E, MCM5, RAD51, RBL1 and FEN1, while the com-mon miRNA-lncRNA network involved 24 nodes and 52 edges, in which lncRNAs including FENDRR (degree = 12), LOC339524 (degree = 9) and FLJ43663 (degree = 7) were remarkable (Figure 4).

Although miRNAs with multiple predicted tar-gets such as hsa-miR-145 (degree = 7) and hsa-miR-128 (degree = 5) were remarkably

Figure 2. The specific lncRNA-mRNA and lncRNA-miRNA networks for lymph node metastatic laryngeal squamous cell carcinoma samples. A: The specific lncRNA-mRNA network; B: The specific lncRNA-miRNA network with down-regulated lncRNA and upregulated miRNA; C: The specific lncRNA-miRNA network with upregulated lncRNA and downregulated miRNA. The wedge-shape represents lncRNA with blue color denoting upregulation, and yellow color denoting downregulation; the square represents miRNAs with pink color denotes upregulation, and green color denoting downregulation; the round represents mRNA with red color denoting upregulation, and green color denot-ing downregulation. The lncRNAs or miRNAs in the networks serve as the ‘nodes’, and each pairwise interaction is represented by an undirected link.



exhibited in miRNA-mRNA network separately for lymph node metastatic and non-metastatic LSCC (data not shown) samples, no specific or common pairwise interactions were identified after venn analysis. Therefore, the integrated networks of mRNA-miRNA-lncRNA for the spe-cific or common DE-mRNAs, miRNAs and lncRNAs failed to be constructed. However, the separate integrated networks for metastatic and non-metastatic samples were constructed: in the network for metastatic LSCC, there were 269 pairwise interactions, with central miRNA nodes of hsa-miR-1207-5p (with predicted tar-gets of DVL3 and G6PD and the interacted lncRNA LOC157273), hsa-miR-142-3p (with predicted targets of CA5B and CTNND1 and the interacted lncRNAs LOC96610) and hsa-miR-

145-5p (with predicted targets of SMC1A, SEH1L, CDK4 and E2F3 and the interacted lncRNA LOC157273); in the network for non- metastasis LSCC cells, 103 interaction pairs were recognized in the integrated network, in which the predominant miRNAs were hsa-miR-4716-3p, hsa-miR-4428, hsa-miR-4317 and hsa-miR-193a-3p (Figure 5).

Discussion

In previous studies relating to LSCC, the target-ed regulations between miRNA and mRNA have been extensively investigated, while the rela-tionships between lncRNA and mRNA were involved in a handful of studies. Our present study indicated that there were a total of

Figure 3. The specific lncRNA-mRNA and ln-cRNA-miRNA networks for non-metastatic laryn-geal squamous cell carcinoma samples. A: The specific lncRNA-mRNA network; B: The specific lncRNA-miRNA network with downregulated ln-cRNA and upregulated miRNA; C: The specific lncRNA-miRNA network with upregulated ln-cRNA and downregulated miRNA. The wedge-shape represents lncRNA which in blue denotes upregulation, while in yellow denotes downreg-ulation; the square represents miRNAs which in pink denotes upregulation, whereas in green denotes downregulation; the round represents mRNA which in red denotes upregulation, while in green denotes downregulation.



2836/1183 DE-mRNAs, 52/63 DE-miRNAs and 81/104 DE-lncRNAs between cervical lymph node metastatic/non-metastatic LSCC tissues and control samples, among which 2253/600 DE-mRNAs, 45/34 DE-miRNAs and 71/48 DE-lncRNAs were the specific for meta-static/non-metastatic LSCC samples. Pathways related to cell cycle and proteasome were over-represented for metastatic specific DE-mRNAs, while the predominant pathways for non-meta-static specific ones were ECM-receptor interac-tion and CAMs. In the lncRNA-miRNA and inte-grated mRNA-miRNA-lncRNA networks, up-reg-ulated PVT1 and LOC157273 were hub meta-

ported the previous results, suggesting the negative regulatory relationship might also exist in lymph node metastatic LSCC. Glucose-6-phosphate dehydrogenase (G6PD) is a cyto-solic enzyme that plays significant roles in the production of NADPH, the main electron donor in the defense against oxidizing agents. In pulmonary artery smooth muscle cells, G6PD underwent the ubiquitin proteasome pathway mediated degradation [33] under normoxia condition, implying its involvement in protea-some pathway. Additionally, G6PD was validat-ed to be the target of miR-1207-5p in mesan-gial cells [34]. Combining with our functional

Figure 4. The common lncRNA-mRNA and lncRNA-miRNA networks of non-metastatic and metastatic laryngeal squamous cell carcinoma samples. The wedge-shape represents lncRNA which in blue denotes upregulation, while in yellow denotes downregulation; the square in pink denotes upregulated miRNAs; the round in red upregulated mRNA. A: The specific lncRNA-mRNA network; B: The specific lncRNA-miRNA network.

static specific lncRNAs, and PVT1 was associated with the downregulated miRNAs of hsa-miR-1207-5p as well as the upregulated genes of G6PD, HIST2H4B and MCM6, among which G6PD was pre-dicted to be targeted by hsa-miR-1207-5p; whereas LOC- 157273 was closely related to the downregulated miRNAs of hsa-miR-145-5p as well as numerous upregulated genes such as CDK4, SEH1L and SMC1A, in which SEH1L and SMC1A were the predicted targets of hsa-miR-145.

The elevated expression of PVT1 was implicated in multi-ple human cancers including colorectal cancers (CRC) [30] and gastric cancer [31]. Moreover, PVT1 was verified to be involved in apoptosis inhibition in CRC pathophysi-ology. Meanwhile, miR-1207-5p was also implicated in this process and the knockdown of PVT1 resulted in elevated miR-1207-5p expression, sug-gesting the potential regula-tion between PVT1 and miR-1207-5p [30]. The downregu-lation of miR-1207-5p was also observed in HNSCC [32]. Our study indicated that the up-regulated PVT1 was linked to the downregulated miR-1207-5p, which indirectly sup-



enrichment analysis, which revealed specific mRNAs in lymph node metastatic samples were highly correlated with proteasome path-way, it might be speculated that in the progres-sion of lymph node metastasis, G6PD might have the potential to participate in the protea-

some pathway and be targeted by miR-1207-5p, which was predicted regulated by the lncRNA PVT1.

Large amount of studies have explored the functions of miRNAs in LSCC pathology using

Figure 5. The intergrated lncRNA-miRNA-mRNA networks. A: The network for lymph node metastatic laryngeal squa-mous cell carcinoma (LSCC) samples with downregulated mRNA, upregulated miRNA and downregulated lncRNA; B: The network for lymph node metastatic LSCC with upregulated mRNA, downregulated miRNA and upregulated lncRNA; C: The network for non-metastatic LSCC with downregulated mRNA, upregulated miRNA and downregulated lncRNA; D: The network for non-metastatic LSCC with upregulated mRNA, downregulated miRNA and upregulated lncRNA. The wedge-shape represents lncRNA which in blue denotes upregulation, while in yellow denotes down-regulation; the square represents miRNAs which in pink denotes upregulation, whereas in green denotes down-regulation; the round represents mRNA which in red denotes upregulation, while in green denotes downregulation.



miRNA microarray profiling. Downregulation of hsa-miR-145-5p was detected in various human malignancies and was experimentally demonstrated by RT-PCR in LSCC [35, 36]. The restoration of miR-145 was confirmed to inhibit the cell proliferation in human lung adenocarci-noma via the mediation of its targets EGFR [37], an important regulator in cell growth relat-ed signaling. Thus miR-145 was considered as a tumor suppressor in cancer development. Structural maintenance of chromosomes 1A (SMC1A) is one component of the cohesin mul-tiprotein complex, which is essential during cell division. SMC1A is predicted to be involved in cell cycle related events in irradiated IM9 human B lymphoblastic cell [38] and the gene expression was discovered in mitosis [39]. Cyclin-dependent kinases (CDKs) consist of many members that play significant roles in dif-ferent stages of cell cycle. CDK4 is considered as the factor for disordered progression of cell cycle in G1/S phase. The dysregulation of CDK4 has been detected in many cancers such as breast cancer [40], CRC [41] and hepatocellular carcinoma (HCC) [42] in previous studies. Moreover, a recent study verified the high expression of CDK4 in LSCC tumorigenesis [2]. Furthermore, the expression of CDK4 was found be suppressed by miR-145 in cell cycle regulation in non-small cell lung cancer cell pro-liferation [43]. Combining that miR-145 was downregulated, while SMC1A and CDK4 were both upregulated in lymph node metastatic LSCC with that specific metastatic DE-mRNAs were mainly enriched in cell cycle related path-way in our results, it might be inferred that these genes and miRNA might have crucial roles in metastatic LSCC progression via regu-lating cell cycle related pathway, and that a regulatory relationship might exist between this miRNA and the two genes.

The multifunctional transcription factor, TP53 regulates cell cycle arrest and apoptosis relat-ed gene expressions. The mutations of TP53 are implicated in various human cancers includ-ing HNLSCC and are correlated with poor prog-nosis, and thus TP53 normally serves as a tumor suppressor [44, 45]. Interestingly, LOC157273 was overexpressed in TP53 mutat-ed tumors in HCC [46], hinting the potential role of LOC157273 in HCC progression. However, more detailed functions about the lncRNA LOC157273 were rarely explored and remained

unknown, especially with regards to LSCC. Given that the upregulation of LOC157273 was revealed in lymph node metastatic LSCC sam-ples in our present study, this lncRNA might be a novel biomarker in the prognosis of metastat-ic LSCC. However, all these putative correla-tions are warranted to be validated by more experiments.

In conclusion, several important mRNAs (G6PD, SMC1A and CDK4), miRNAs (miR-1207-5p and hsa-miR-145-5p) and lncRNAs (PVT1 and LOC157273) were identified as potential bio-markers for lymph node metastatic LSCC diag-nosis. Additionally, PVT1 might be in the upstream of miR-1207-5p, which might target G6PD; and SMC1A and CDK4 might be the tar-gets of hsa-miR-145-5p and involved in cell cycle related pathways.

Acknowledgements

This work were supported by the National Natural Science Foundation of China (Grant No.81172584, 81402256), Scientific and Te- chnological Innovation Programs of Higher Education Institutions in Shanxi (STIP, 2014- 1102), The Research Project of Shanxi Province Health and Family Planning Commission (201301073, 2014028), and The Science and Technology Innovation Fund (01201315) & The Youth Research Fund (02201322) of Shanxi Medical University. We are grateful to Dr. Tao BAI (Pathology Department of NO.1 Hospital Affiliated to Shanxi Medical University, Taiyuan, Shanxi, 030001, China) for technical assis-tance. We also thank Dr. Jian SA (Department of Statistics, Shanxi Medical University, Taiyuan, Shanxi, China) for guidance with the statistical analysis.

Disclosure of conflict of interest

None.

Address correspondence to: Dr. Binquan Wang, De- partment of Otolaryngology, Head & Neck Surgery, The First Hospital Affiliated with Shanxi Medical University, Shanxi Key Laboratory of Otolaryngology Head & Neck Cancer. 85 Jiefang Road South, Taiyuan 030001, Shanxi, China. Tel: +86-351-4639988; Fax: +86-351-4639218; E-mail: [email protected]

mailto:[email protected]

mailto:[email protected]



References

[1] Hui AB, Lenarduzzi M, Krushel T, Waldron L, Pintilie M, Shi W, Perez-Ordonez B, Jurisica I, O’Sullivan B, Waldron J, Gullane P, Cummings B, Liu FF. Comprehensive MicroRNA profiling for head and neck squamous cell carcinomas. Clin Cancer Res 2010; 16: 1129-1139.

[2] Lian M, Fang J, Han D, Ma H, Feng L, Wang R and Yang F. Microarray gene expression analy-sis of tumorigenesis and regional lymph node metastasis in laryngeal squamous cell carci-noma. PLoS One 2013; 8: e84854.

[3] Wang W, Lin P, Han C, Cai W, Zhao X and Sun B. Research Vasculogenic mimicry contributes to lymph node metastasis of laryngeal squa-mous cell carcinoma. J Exp Clin Cancer Res 2010; 29: 1-9.

[4] Hoffman HT, Porter K, Karnell LH, Cooper JS, Weber RS, Langer CJ, Ang KK, Gay G, Stewart A, Robinson RA. Laryngeal cancer in the United States: changes in demographics, patterns of care, and survival. Laryngoscope 2006; 116: 1-13.

[5] Li D, Feng J, Wu T, Wang Y, Sun Y, Ren J and Liu M. Long intergenic noncoding RNA HOTAIR is overexpressed and regulates PTEN methyla-tion in laryngeal squamous cell carcinoma. Am J Pathol 2013; 182: 64-70.

[6] Rudolph E, Dyckhoff G, Becher H, Dietz A and Ramroth H. Effects of tumour stage, comorbid-ity and therapy on survival of laryngeal cancer patients: a systematic review and a meta-anal-ysis. Eur Arch Otorhinolaryngol 2011; 268: 165-179.

[7] Zhang B, Liu W, Li L, Lu J, Liu M, Sun Y and Jin D. KAI1/CD82 and CyclinD1 as biomarkers of invasion, metastasis and prognosis of larynge-al squamous cell carcinoma. Int J Clin Exp Pathol 2013; 6: 1060.

[8] Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T and Thun MJ. Cancer statistics, 2008. CA Cancer J Clin 2008; 58: 71-96.

[9] Liu WW, Zeng ZY, Wu QL, Hou JH and Chen YY. Overexpression of MMP-2 in laryngeal squa-mous cell carcinoma: a potential indicator for poor prognosis. Otolaryngology Head and Neck Surgery 2005; 132: 395-400.

[10] Shi Y, Gong HL, Zhou L, Tian J and Wang Y. CD24: a novel cancer biomarker in laryngeal squamous cell carcinoma. ORL J Otorhino- laryngol Relat Spec 2011; 74: 78-85.

[11] Chen K, Cornejo KM, Ye W, Wu Q, Liang J and Jiang Z. Oncofetal protein IMP3: a new diag-nostic biomarker for laryngeal carcinoma. Hum Pathol 2013; 44: 2126-2131.

[12] Farazi TA, Hoell JI, Morozov P and Tuschl T. MicroRNAs in human cancer. In: MicroRNA Cancer Regulation Springer; 2013. pp 1-20.

[13] Shen Z, Zhan G, Ye D, Ren Y, Cheng L, Wu Z and Guo J. MicroRNA-34a affects the occur-rence of laryngeal squamous cell carcinoma by targeting the antiapoptotic gene survivin. Med Oncol 2012; 29: 2473-2480.

[14] Li X, Wang Hl, Peng X, Zhou HF and Wang X. miR-1297 mediates PTEN expression and con-tributes to cell progression in LSCC. Biochem Biophys Res Commun 2012; 427: 254-260.

[15] Shen Z, Li Q, Deng H, Lu D, Song H and Guo J. Long Non-Coding RNA Profiling in Laryngeal Squamous Cell Carcinoma and Its Clinical Significance: Potential Biomarkers for LSCC. PLoS One 2014; 9: e108237.

[16] Geisler S and Coller J. RNA in unexpected plac-es: long non-coding RNA functions in diverse cellular contexts. Nat Rev Mol Cell Biol 2013; 14: 699-712.

[17] Feng J, Tian L, Sun Y, Li D, Wu T, Wang Y and Liu M. Expression of long non-coding ribonu-cleic acid metastasis-associated lung adeno-carcinoma transcript-1 is correlated with prog-ress and apoptosis of laryngeal squamous cell carcinoma. Head Neck Oncol 2012; 4: 46.

[18] Gautier L, Cope L, Bolstad BM and Irizarry RA. affy-analysis of Affymetrix GeneChip data at the probe level. Bioinformatics 2004; 20: 307-315.

[19] Allison DB, Cui X, Page GP and Sabripour M. Microarray data analysis: from disarray to con-solidation and consensus. Nat Rev Genet 2006; 7: 55-65.

[20] Smyth GK. Limma: linear models for microar-ray data. In: Bioinformatics and computational biology solutions using R and Bioconductor Springer; 2005. pp. 397-420.

[21] Shade A and Handelsman J. Beyond the Venn diagram: the hunt for a core microbiome. Environ Microbiol 2012; 14: 4-12.

[22] Consortium GO. The Gene Ontology (GO) data-base and informatics resource. Nucleic Acids Res 2004; 32: D258-D261.

[23] Kanehisa M, Goto S, Furumichi M, Tanabe M and Hirakawa M. KEGG for representation and analysis of molecular networks involving dis-eases and drugs. Nucleic Acids Res 2010; 38: D355-D360.

[24] Liew AWC, Law NF, Cao XQ and Yan H. Statistical power of Fisher test for the detection of short periodic gene expression profiles. Pattern Recognition 2009; 42: 549-556.

[25] Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC and Lempicki RA. DAVID: data-base for annotation, visualization, and inte-grated discovery. Genome Biol 2003; 4: P3.

[26] Wang J. Pearson Correlation Coefficient. In: Encyclopedia of Systems Biology Springer; 2013. pp. 1671-1671.



[27] Franchina T, Amodeo V, Bronte G, Savio G, Ricciardi GR, Picciotto M, Russo A, Giordano A and Adamo V. Circulating miR-22, miR-24 and miR-34a as Novel Predictive Biomarkers to Pemetrexed-Based Chemotherapy in Advanced Non-Small Cell Lung Cancer. J Cell Physiol 2014; 229: 97-99.

[28] Hua Z, Lv Q, Ye W, Wong CK, Cai G, Gu D, Ji Y, Zhao C, Wang J, Yang BB, Zhang Y. MiRNA-directed regulation of VEGF and other angio-genic factors under hypoxia. PLoS One 2006; 1: e116.

[29] Smoot ME, Ono K, Ruscheinski J, Wang PL and Ideker T. Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics 2011; 27: 431-432.

[30] Takahashi Y, Sawada G, Kurashige J, Uchi R, Matsumura T, Ueo H, Takano Y, Eguchi H, Sudo T, Sugimachi K, Yamamoto H, Doki Y, Mori M,Mimori K. Amplification of PVT-1 is involved in poor prognosis via apoptosis inhibition in colorectal cancers. Br J Cancer 2014; 110: 164-171.

[31] Ding J, Li D, Gong M, Wang J, Huang X, Wu T and Wang C. Expression and clinical signifi-cance of the long non-coding RNA PVT1 in hu-man gastric cancer. Onco Targets Ther 2014; 7: 1625.

[32] Chen KM, Stephen JK, Havard S, Shah V, Gardner G, Schweitzer VG and Worsham MJ. Novel dysregulated microRNAs in primary la-ryngeal squamous cell cancer. Cancer 2012 3: 76-81.

[33] Chettimada S, Kobelja R, Gupte R and Gupte S. Role Of Glucose-6-Phosphate Dehydro- genase In Pulmonary Artery Smooth Muscle Cell Proliferation & Apoptosis. In: AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE Am Thoracic Soc, 2011.

[34] Alvarez ML, Khosroheidari M, Eddy E and Kiefer J. Role of microRNA 1207-5P and its host gene, the long non-coding RNA Pvt1, as mediators of extracellular matrix accumulation in the kidney: implications for diabetic ne-phropathy. PLoS One 2013; 8: e77468.

[35] Ayaz L, Görür A, Yaroğlu HY, Özcan C and Tamer L. Differential expression of microRNAs in plas-ma of patients with laryngeal squamous cell carcinoma: potential early-detection markers for laryngeal squamous cell carcinoma. J Cancer Res Clin Oncol 2013; 139: 1499-1506.

[36] Cao P, Zhou L, Zhang J, Zheng F, Wang H, Ma D and Tian J. Comprehensive expression profiling of microRNAs in laryngeal squamous cell carci-noma. Head & Neck 2013; 35: 720-728.

[37] Cho WC, Chow AS and Au JS. MiR-145 inhibits cell proliferation of human lung adenocarcino-ma by targeting EGFR and NUDT1. RNA Biol 2011; 8: 125-131.

[38] Cha HJ, Shin S, Yoo H, Lee EM, Bae S, Yang KH, Lee SJ, Park IC, Jin YW, An S. Identification of ionizing radiation-responsive microRNAs in the IM9 human B lymphoblastic cell line. Int J Oncol 2009; 34: 1661-1668.

[39] Yang F and Wang P. The Mammalian synapto-nemal complex: a scaffold and beyond. Genome Dyn 2009; 5: 69-80.

[40] Yu Q, Sicinska E, Geng Y, Ahnström M, Zagozdzon A, Kong Y, Gardner H, Kiyokawa H, Harris LN, Stål O, Sicinski P. Requirement for CDK4 kinase function in breast cancer. Cancer Cell 2006; 9: 23-32.

[41] Pishvaian MJ, Yang S, Zouhairi ME, Wu CS, Mishra L and Avantaggiati ML. Synergistic anti-cancer activity of the CDK4/6 inhibitor PD-0332991 in combination with 5-fluorouracil-based chemotherapy in human colon cancer cells. Cancer Res 2010; 70: 5047.

[42] Rivadeneira DB, Mayhew CN, Thangavel C, Sotillo E, Reed CA, Graña X and Knudsen ES. Proliferative suppression by CDK4/6 inhibi-tion: complex function of the retinoblastoma pathway in liver tissue and hepatoma cells. Gastroenterology 2010; 138: 1920-1930. e1922.

[43] Chen Z, Zeng H, Guo Y, Liu P, Pan H, Deng A and Hu J. miRNA-145 inhibits non-small cell lung cancer cell proliferation by targeting c-Myc. J Exp Clin Cancer Res 2010; 29: 151.

[44] Petitjean A, Achatz M, Borresen-Dale A, Hainaut P and Olivier M. TP53 mutations in hu-man cancers: functional selection and impact on cancer prognosis and outcomes. Oncogene 2007; 26: 2157-2165.

[45] Ganci F, Sacconi A, Bossel Ben-Moshe N, Manciocco V, Sperduti I, Strigari L, Covello R, Benevolo M, Pescarmona E, Domany E, Muti P, Strano S, Spriano G,Fontemaggi G, Blandino G. Expression of TP53 mutation-associated mi-croRNAs predicts clinical outcome in head and neck squamous cell carcinoma patients. Ann Oncol 2013; 24: 3082-3088.

[46] Woo HG, Wang XW, Budhu A, Kim YH, Kwon SM, Tang ZY, Sun Z, Harris CC and Thorgeir- sson SS. Association of TP53 mutations with stem cell-like gene expression and survival of patients with hepatocellular carcinoma. Gas- troenterology 2011; 140: 1063-1070. e1068.

Documents

Original Article Potential biomarkers and their regulatory ......Tissues were immediately frozen in liquid nitro-gen and shred using a homogenizer (IKA, Staufen, Germany). For the