Analysis of hub genes in small-cell lung carcinoma by weighted gene co-expression network※

XING Yonghua，SONG Peipei，SU Zhanhai，WANG Haiyan，LIU Yan，AN Juan，GUO Yinghua

(1.Department of Medicine，Qinghai University，Antitumor Research Team of Chinese and Tibetan Medicine，Xining 810016，China；2.Global Health Institute，Bureau of International Health Cooperation，National Center for Global Health and Medicine，Japan)

Abstract Objective The hub genes of Small-cell lung carcinoma(SCLC)were preliminarily screened to provide clues for multi-target therapy.Methods The characteristics of SCLC gene expression profile and related hub genes were analyzed by weighted gene co-expression network.Results A total of 7 844 differential genes were identified，and 8 gene modules were related to the occurrence and development of SCLC.NLN，TMEM40，TOP2A，COL5A2，SNX1，NDUFA13，C14orf159 and GIMAP4 were identified as the hub genes of SCLC.Conclusions The eight hub genes identified in the paper may be valuable diagnostic and therapeutic targets.

Keywords lung carcinoma；gene；expression；network；analysis；module；hub

1 Introduction

Small-cell lung carcinoma(SCLC)is an aggressive disease which is inclined to early dissemination，accounting for approximately 15% of all lung cancer.The final diagnosis of SCLC relies on morphology and immunohistochemistry.The disease is divided into limited and extensive stages based on the World Health Organization(WHO)guidelines.SCLC grows rapidly with a 5-year survival rate of less than 5%.The treatment standard of SCLC consists of platinum-based chemotherapy and radiotherapy without great change in the last decades[1].In the limited stage(LS)，radiotherapy is necessary for thorax and mediastinum and it has the potential to prevent metastasis.However，it is difficult to detect SCLC in time because of the lack of specific symptoms and effective screening approaches.In the extensive stage(ES)，chemotherapy is the main treatment for patients with metastatic disease.Nevertheless，there are few effective drugs with significant efficacy and minor side effects for ES-SCLC patients.Although topotecan is recommended as a standard second-line treatment of SCLC，it is not beneficial for all patients because of severe hematologic toxicity[2].Metronomic chemotherapy is more effective than single-agent regimens therapy which shows unexpected effects and significant toxicity.For example，the overall survival rate of patients treated with metronomic chemotherapy regimen of cisplatin，etoposide and irinotecan increased when comparing with patients who only taking topotecan[3].In addition，it is demonstrated by some clinic trails that three-drug chemotherapy regimen of atezolizumab，carboplatin and etoposide shows greater improvement to overall survival rate compared with the monotherapy than those of the three drugs[4].The outcomes of SCLC patients may be improved by combining early diagnosis，chemotherapy and irradiation.Hence，it is urgent to understand the biology of SCLC.

Recently，advances at molecular mechanism research provide numerable in-depth information of SCLC，leading the development of prognostic biomarkers and target agents.In contrast to histological evaluation，patients may be classified into subgroups with different genetic characteristics by whole-exome sequencing of tumor mutational burden analysis[5].The important biomarkers are categorized into three groups including cancer biomarkers，immune micro-environment biomarkers and circulating biomarkers.

Several cancer biomarkers were detected in pre-clinical mouse model and human SCLC specimens.The inactivation ofRbandTP53，mutations inPTEN，inSLIT2 andEPHA7，and focal amplifications of theMYCfamily andFGFR1 were detected by using SNP array，transcriptome and genome sequencing on SCLC specimens.The mutations in kinase，G-protein gene family members were detected in SCLC patients.The amplification and fusion of these genes may share in a related way[6].The Notch signaling is inhibited by delta-like ligand 3(DLL3)which specifically expresses on the surface of SCLC tumor cells with higher level.It seemed that the expression of DLL3 was related to the tumor progression in SCLC，and was involved in its migration and invasion.In terms of the treatment of target DLL3 in SCLC，several novel therapies were developed including antibody-drug conjugate(ADC)rovalpituzumab tesirine，the bispecific T cell engager(BiTE)immuno-oncology therapy AMG 757 and chimeric antigen receptor(CAR)T cell therapy AMG119.It was found in pre-clinical and clinical studies that the sensitivity and response of SCLC to chemotherapy cisplatin and etoposide may be improved by rovalpituzumab tesirine.It can be seen from the research of AMG 757 that redirecting T cells can kill DLL3-positive cancer cells.The figure of killing tumor cell was showed by CAR T cells.Immunotherapy may be another treatment strategy because of the abilities of recognizing and eradicating tumor cells.A few immune checkpoint pathways are determined by pre-clinical and clinical studies.The autoimmunity and antitumor immunity can by inhibited by Cytotoxic T-lymphocyte-associated protein 4(CTLA-4).The positive results of removing the inhibitory signal of CTLA-4 and stimulating antitumor immunity were obtained in some trials with ipilimimab which is the first therapeutic agent[7].Programmed cell death protein-1(PD-1)receptor can bind to its ligand PD-L1 which is expressed on the cells and immune micro-environment of tumor.In phase I/II trial of nivolumab(monoclonal antibody against PD-1)in patient with recurrent ES-SCLC displayed a 1-year survival[8].Inhibitory activity of macrophages of CD47 which over expressed on the surface of human SCLC cells was found，and in addition，CD47 was involved in the process of immune escape by tumors[9].Therefore，the imbalance between immune-inhibition and immune-activity is crucial in the process of tumor progression[10].The efforts of pursuing the sensitivity，specificity and reproducibility of circulating biomarkers have been made in recent years.It was demonstrated that Neuron specific enolase(NSE)plays a role in the diagnosis and treatment of SCLC.The levels of NSE with good specificity and sensitivity is an indication to the developmental stage of SCLC[11].Recently，it is inspiring to pursuit the biomarkers of SCLC even with limited and contradictory data.In general，multiple biomarkers are more specific and sensitive than a single biomarker.More importantly，multiple biomarker profiles can be used for the progression，diagnosis and treatment of SCLC.The purpose was to explore the genetic profiles of SCLC by using weighted gene co-expression network analysis(WGCNA)technology，so as to identify specific gene modules and hub genes，and provide biological information of SCLC[12].

WGCNA is a system biology method for detecting the correlation patterns among genes in high-dimensional datasets.It can be used for finding specific gene modules which consists of highly correlated genes and relating modules to external sample traits.In addition，the specific gene profiles of SCLC could be detected by the method which could provide some forward-looking ideas.The hub genes，found in the center of gene modules，were highly connected with other genes.They may be valuable for identifying the biomarkers of diagnosis and therapeutic targets for SCLC.

2 Materials and Methods

2.1 Microarray Data Pre-Processing

The transcription profiles of GSE43346 and GSE4498(Affymetrix Human Genome)were obtained from Gene Expression Omnibus(GEO).The datasets were divided into two groups including control group and SCLC group.The downloaded microarray raw data were filtered，corrected and normalized by rma function of Affy Bioconductor package in R[13，14].

The differentially expressed genes(DEs)of the t-test between control group and SCLC group were identified at the level ofP≤0.05.The weighted gene co-expression networks were constructed with the identified DEs.A total of 7 844 DEs were identified in this research.

2.2 Establishment of Weighted Gene Co-expression Network

Weighted gene co-expression network was constructed using the blockwise Modules function in WGCNA package of R software based on 7 844 DEs.Pearson′s correlation coefficient was computed between any two DEs and applied to build similarity matrix.Then，the similarity matrix was converted to a weighted adjacency matrix based on scale-free topology criterion β=9.Subsequently，the co-expression matrix and the topological overlap matrix(TOM)were constructed to analyze the connection among the genes[15].

2.3 Screening of Interesting Gene Modules and Hub Genes

The module preservation statistics Zsummaryscore(Zsummary=(Zdensity+Zconnectivity)/2)was applied to evaluate changes of properties of gene modules between control network and SCLC network.In common，module preservation is categorized into gradation by Zsummaryscore.Specifically speaking，Zsummary<2 means poor preservation，210 suggests strong preservation.The modules with poor preservation were considered as interesting ones in which hub genes were identified[18].It is noteworthy that there is a positive correlation between Zsummaryvalue and the size of module.

Median Rank which was independent of the size of module was used as another module preservation statistics in this study.There is a negative correlation between preservation and the value of median rank.The gene modules of interest were identified considering Zsummaryscore and Median Rank in the present study.

2.4 Analysis of Gene Significance(GS)and Module Membership(MM)

The GS indicates the correlation between gene expression profile and phenotypic trait.In this study，phenotypic trait is SCLC[12].GS is computed by GSi=-log(Pi)，where Piwas the P value from the t-test.MM was defined as：MMi=|cor(x(i)，ME)|.The MM is a measure of the correlation between gene expression profile and ME.Thei-th gene does not belong to the module when the value of MMiis close to 0，while it belongs to the module when that of MMiis close to 1.In general，it was indicated by a higher value of MM that a gene is of great importance to the given module[15].

2.5 Hub genes identification

The hub genes with high GS，high MM and high intramodular connectivity within module were identified in this study[19].

3 Results

3.1 Weighted Gene Co-expression networks of control group and SCLC group

Using the WGCNA package in R software，the weighted gene co-expression networks of control group and SCLC group were constructed respectively based on DEs(n=7 844).A total of 24 gene modules were identified using Dynamic Tree Cut(deep split=2，cut height=0.99).Especially，the grey gene modules were ignored in subsequent analysis as the genes in the modules did not belong to any ones.(Figure1)

A total of(A)12 and 11modules were identified in normal small airway epithelial cells group and SCLC group，respectively.Colors in the horizontal bar represent the different modules

3.2 Evaluation of module preservation that specifically related to SCLC

The network of normal airway epithelial cells was taken as the reference network，and the SCLC network was taken as the test network in the preservation statistics.Salmon gene module that was found only in SCLC group might provide some clues of the importance of SCLC.The other 7 gene modules，including turquoise，green，greenyellow，purple，yellow，magenta and red modules were significantly different between the reference network and the test network on the basis of Zsummary(Zsummary<10)and Median Rank[18].These 7 modules may play a role in the origin and development of SCLC.In specific，the preservation assessment of greenyellow and brown modules were disturbed by the size of module combining Zsummarywith Median Rank.Therefore，the preservation of brown module was mainly assessed by Median Rank.(Figure 2)

Each point represents a module，labeled with a color.In the preservation Zsummary graph on the left，the blue and green horizontal lines show the thresholds of Zsummary(y-axis)=2 and Zsummary(y-axis)=10，respectively.Zsummary>10 is indicative of strong preservation of the modules.In the preservation Median Rank graph on the right，the Median Rank of the modules close to zero indicated a high degree of module preservation.The comprehensive analysis is based on the Median Rank and Zsummary.The blue，black，pink，and tan modules were ultimately considered to be well preserved

3.3 Hub genes related to the origin and development of SCLC

A gene with high values of GS，MM and intramodular connectivity in the network was regarded as a hub gene in this research.In the modules of interest，there was a positive correlation between MM and intramodular connectivity，but no correlation was found between MM and GS.Therefore，it was found that a total of 8 hub genes were identified based on the values of MM and the centrality of genes in each module，includingNLN(MM.salmon=-3.299，P-value=0.002)，TMEM40(MM.turquoise=7.587，P-value=5.67E-09)，GIMAP4(MM.yellow=-4.197，P-value=0.000)，TOP2A(MM.red=-12.544，P-vlaue=1.05E-14)，COL5A2(MM.purple=-5.859，P-value=1.07E-06)，SNX1(MM.magenta=3.945，P-value=0.000)，NDUFA13(MM.green=3.499，P-value=0.001)andC14orf159(MM.brown=-3.365，P-value=0.001)(Figure 3).

Suddenly, out of breath, she snatched the tabor from Abdallah with her left hand, and, holding the dagger in her right hand, held out the tabor to her master

3.4 Gene ontology(GO)enrichment analysis of the 8 gene modules

The software DAVID(http：//david.abcc.ncifcrf.gov/)[20]was used in the present study to complete GO functional annotation and enrichment analysis.Eight top biological processes were enriched in the 8 specific gene modules，including regulation of RNA polymerase Ⅱ transcriptional preinitiation complex assembly(salmon)，skin development(turquoise)，immune response(yellow)，cell cycle process(red)，membrane organization(green)，extracellular matrix organization(purple)，regulation of cellular amino acid metabolic process(magenta)and cation transmembrane transport(brown).(Table 1)

The color indicates the module，and the dot indicates the gene within the module.The identification of hub genes based on the presence of high MM values.MM，module membership；GS，gene significance

Table 1 GO analysis of gene modules of SCLC group

续表

3.5 Pathway enrichment analysis of the eight gene modules

The pathway enrichment analysis for eight modules of interest was accomplished using software DAVID based on Kyoto Encyclopaedia of Genes and Genomes(KEGG)(http：//www.genome.jp/kegg/)database.Six top enriched pathways for six specific gene modules were as follows：neuroactive ligand-receptor interaction(P-value=6.26E-02)in the turquoise module，neuroactive ligand-receptor interaction glycolysis/gluconeogenesis(P-value=4.54E-03)in yellow module，MAPK signaling pathway(P-value=6.34E-02)in red gene module，chemokine signaling pathway(P-value=1.77E-02)in purple module，natural killer cell mediated cytotoxicity(P-value=6.14E-02)in green module，Huntington disease(P-value=8.44E-03)in brown gene module.However，no pathways were significantly enriched in the salmon and magenta modules.(Table 2).

Table 2 KEGG enrichment of SCLC group

Following shows the interaction of gene co-expression patterns in the eight modules of interest.The nodes with the degree of ≥25 were displayed for clarity.The network was visualized using software Cytoscape 3.0.The sizes and edge widths of these nodes were proportional to their degree and connection strength respectively.The red dots were indications to the hub gene of each gene module.(Figure 4)

Figure 4 Interaction of gene co-expression patterns in the moduls of interest

4 Discussion

Small-cell lung carcinoma(SCLC)is an aggressive neuroendocrine neoplasia with poor survival rate.The treatment options for recurrent SCLC were limited，and its prognosis has not been improved significantly for more than 3 decades.The studies of molecular profiling may provide information of biomarkers and therapeutic targets for SCLC.

Eight gene modules and eight hub genes were identified to be associated with SCLC.The salmon module was found only in SCLC networks and was enriched in the regulation of RNA polymerase Ⅱ transcriptional preinitiation complex assembly based on GO analysis.Neurolysin(NLN)was regarded as a hub gene in this module and was up-regulated in SCLC group(P-value=0.002).NLN，a zinc metallopepeptidase with His-Glu-Xaa-Xaa-His(HEXXH)motif，was widely distributed in cytosol，endoplasmic reticulum，mitochondria and nucleus of different mammalian tissues and tumor cells[21-24].The biologically inactive fragments may be generated by cleaving neurotensin(NTS)at Pro10-Tyr11 with NLN.NTS has long been recognized as a neurotransmitter or neuromodulator.Many studies have shown that NTS plays a role in endocrine，autocrine and paracrine growth stimulation of various types of cancer.Ocejo-Garcia M and his colleagues found that high concentrations were presented in non-small cell lung cancer and SCLC[25].NTS could promote tumor growth and product massive nodal metastasis[26].In addition，NLN could hydrolyze bradykinin(BK)，which plays a role in tumor-associated angiogenesis synergistically with interleukin-1[27].Angiogenesis is a primary process that brings tumor growth by providing nutrients and oxygen to tumor cells.Apoptosis may happen to tumor cells when the oxygen was limited from the diffusion of capillary.Furthermore，tumor growth could be promoted by increasing the permeability of tumor neo-vasculature with BK[28-30].Thaysa Paschoalin demonstrated that angiogenesis was inhibited in the culture supernatant(containing NLN and Thimet oligopeptidase)of BI6FI0-Nex2 melanoma cells in a dose-dependent manner[31].As a member of metalloproteases family，NLN may participate in the invasion and metastasis of tumor cell by degrading collagen and other components of connective tissues[32].It was assumed thatNLNis associated with tumor biology and it may be a potential valuable biomarker and therapeutic target for SCLC based on the above evidences.

The turquoise module was enriched in skin development on the basis of GO analysis.Multi-pass membrane proteinTMEM40 which was identified as hub gene was down-regulated in SCLC group(P-value=5.67E-09).Zhenfei Zhang and his colleagues found that the expression level of TMEM40 was related to pathologic grade，clinical stage，histological grade and pT status of patients with bladder cancer.They also investigated the up-regulation of TMEM40 in the tissues and cell lines of bladder cancer by using western blotting and quantitative RT-PCR assy.The inhibitions of proliferation，migration and invasion in EJ and 5 637 cells were observed by using wound healing assay，transwell assay and cell cycle analysis.The increase of p53 together with the activities of caspase-3，caspase-9 and PARP was detected in EJ and 5 637 cells with the knockdown ofTMEM40.In addition，they assumed that TMEM40 was one of oncogenic factors in the development of bladder cancer via p53 signaling pathway[33].Qing Yan Zhang et al found that TMEM40 was highly regulated in tongue squamous cell carcinoma(TSCC)comparing with adjacent normal tongue tissues.The expression of TMEM40 was positively correlated with pathological TNM(pTNM)status，histological grade and clinical stage.They also demonstrated that the apoptosis of TSCC cells might be inhibited by TMEM40 via the pathways associated with P53 and Bax[34].Another report found that higher level of TMEM40 mRNA was associated with the atrophy of parietal lobe and potential inflammatory function arthritis in mice[35，36].But in present study TMEM40 was down-regulated in SCLC group comparing with the control group.No literature was found on the role of TMEM40 in SCLC.An assumption that TMEM40 may possess specific activities which were different from TSCC was required to be confirmed in further research.TMEM40 may be a novel candidate target for cancer treatment since SCLC is generally considered as a major pulmonary neuroendocrine tumor.

The yellow module was enriched in immune response on the basis of GO analysis.GIMAP4 was considered as hub gene and up-regulated in SCLC group(P-value=1.69E-03)in the module.GIMAP4，expressed in the cytosol of mature T cells，was a member of GTPase of the immunity-associated protein(GIMAP)family.It is localized with trans-Golgi complex network(TGN).GIMAP4 has two unique characteristics including GTPase activity and calcium-binding IQ domain which is involved in calcium signaling.GIMAP4 is known as regulating the apoptosis of T cells by means of inducing the activation of caspase-3，and its activation and the apoptosis closely correlated with phosphorylation.Moreover，the regulatory apoptosis of GIMAP4 is associated with Bax protein[37，38].GIMAP4 may contribute to the regulation of immune homeostasis，T cell function and survival[39].TFN-γ secretion and subsequently JAK/STAT signalling pathway were impaired significantly in GIMAP4-depleted Th1 cells[40].In addition，it was strongly demonstrated by the accumulated association evidences that GIMAP4 was involved in several autoimmune diseases and cancers including diabetes，asthma，allergy，head and neck squamous cell carcinoma as well as ovarian cancer[41-43].Based on the evidences mentioned above，it was assumed that GIMAP4 may be a potential candidate immunological target for SCLC and it deserves further investigation.

The red module was enriched in cell cycle process by GO analysis.TOP2A(Topoisomerase II alpha)was identified as hub gene which was up-regulated in SCLC group(P-value=1.05E-14).Nuclear enzyme is encoded byTOP2Agene，which plays an important a role in DNA synthesis and transcription as well as chromosome segregation and cell cycle progression[44].The enzyme activities of TOP2A have been identified in several types of malignancies.For example，patients with prostate cancer with higher levels of TOP2A had shorter biochemical recurrence-free survival with angiolymphatic invasion by using survival analysis and immunohistochemistry[45].The higher expression levels of TOP2A accompanying PETN protein and PIK3CA mutation were found in patients with metastatic breast cancer.Some researchers demonstrated thatTOP2Agene amplification was a favorable prognostic factor in HER2(human epidermal growth factor receptor-2)-positive metastatic breast cancer treated with trastuzumab rather than anthracyclines[46].Elisa Neubauer et al found that higher levels of TOP2A，Ki-67 and RacGAP1 were identified in bronchopulmonary neuroendocrine neoplasms including SCLC patients with poor survival，and these three proliferation markers were distinctly higher in metastases than primary SCLC tumors.Furthermore，it was assumed that TOP2A was a better predictive marker acting as a direct predictive therapeutic value in other cancers and it increased with the loss of differentiation，T stage and grading.Moreover，it was the target of topoisomerase inhibitors.In the present，TOP2A inhibitor etoposide is approved for the treatment of SCLC[47].Using single gene’s expression analysis，it was showed by another report thatTOP2Aexpression was significantly correlated with the shorter Progression Free Survival(PFS)in limited and extend stage of patients with SCLC.In contrast，the lower level ofTOP2Awas correlated with better PFS both in LS-SCLC and ES-SCLC[48].Based on the accumulated evidences，it can be concluded thatTOP2Amay be a valuable prediction and therapeutic target for SCLC.

The purple module was enriched in extracellular matrix organization by GO analysis andCOL5A2 was regarded as hub gene which was up-regulated in SCLC group(P-value=1.07E-06).Collagen type V alpha 2 chain(COL5A2)is a member of collagen V which is a fibrillar collagen.The mutations ofCOL5A2 gene could lead to a classic syndrome Ehlers-Danlos which is a kind of human heritable connective tissue disorder.COL5A2 plays role in the formation of collagen which an important a major component of the extracellular matrix(ECM).ECM is known to play a role in cellular biological process including cell shape，proliferation，migration，differentiation，apoptosis and carcinogenesis.An interaction between cancer cells and ECM is a key mechanism of tumor cell invasion.The decent stroma microenvironment is considered as a prerequisite for tumor growth[49].In the research of colorectal carcinoma，COL5A2 was not expressed in normal epithelium but up-regulated in carcinomas.It was assumed that carcinogenesis were associated with an embryological expression pattern in the carcinoma tissue[50].It was showed by a relative report that high level of COL5A2 was identified in the bladder cancer compared with normal tissues.The patients with higher level of COL5A2 were associated with severe invasion and poor clinical outcome[51].A similar results showed thatCOL5A2 were differentially expressed between DCIS(ductal carcinoma in situ)and IDC(invasive ductal carcinoma)[52].Rongfang Qiu demonstrated that BRMS1/LSD1 complex could suppresses breast cancer cells metastasis by targeting the COL5A2 as well[53].An increase of COL5A2 expression had also been observed in mouse adenomas and human osteosarcoma[54，55].Therefore，it was assumed that COL5A2 may be valuable candidate gene for potential hallmark of SCLC metastasis and development.

According to GO analysis，the magenta module was enriched in the regulation of metabolic process of cellular amino acid and SNX1 was identified as hub gene in the module.SNX1 is a member of sorting nexin protein family，presented one phospholipid-binding motif termed for phox homology(PX)domain that is bond to phosphatidylinositol phosphates in endosomal membranes[56].Interacted with epidermal growth factor receptor(EGFR)，SNX1 has been participated in the process of endosome-to-lysosome trafficking so as to accelerate its degradation[57].Revealed by many researches，SNX1 is also involved in cell endocytosis，protein sorting，efflux，cell signal transduction，membrane transport and organelle movement[58-60].For example，the reduction of SNX1 could lead to the phosphorylation of hepatocyte growth factor(MET)in human lung cancer cell line.It could activate RTK/RAS signaling pathway and then promote cell proliferation and inhibition of cell apoptosis so as to enhance the growth of cancer cells and the malignancy may be increased.SNX1 is also involved in reducing EGFR membrane recycling and triggering EGFR/PI3K/AKT signaling pathway in lung cancer cell line[61].Besides，it was indicated by plenty of evidence that the loss of SNX1 played a role in colony formation，migration，invasion and proliferation.The apoptosis of cancer cell and susceptibility of cancer to drugs are promoted by the overexpression of SNX1 in several types of carcinoma[62-65].Furthermore，Zhang et al demonstrated that protein level of SNX1 was positively related to survival time and prognosis of gastric cancer[64].SNX1 may be a valuable therapeutic target or biomarker for the treatment and prognosis of SCLC.

The green module was enriched in a membrane organization on the basis of GO analysis.NDUFA13(NADH ubiquinone oxidoreductase subunit A 13)was identified as hub gene which was down-regulated significantly in SCLC group.NDUFA13 is the HGCN symbol ofGRIM19(Genes associated with Retinoid-IFN induced Mortality)，by which a mitochondrial complex Ⅰsubunit was encoded.The mutation ofGRIM19 may induce the instability of mitochondrial complex，hypotonia，dyskinesia and sensorial deficiencies.The decreased expression ofGRIM19 was found in many types of tumor.The expression levels ofGRIM19 was negatively correlated with that of surviving in cervical tumors[66].The Stat3-driven transcription of Bcl2 members was inhibited by GRIM19 interacting with STAT3，which resulted in the inhibition growth of tumor cell[67].Bcl-2 is up-regulated in carcinoma and involved in apoptosis resistance.GRIM19/p16INK4Acomplex was tightly associated with CDK4，preventing the binding of cyclin D1 and CDK4 which led to the progression of cell cycle[68].It was demonstrated that the expression level ofGRIM19 in primary tumor was lower than that in advanced tumor[69].It was found that GRIM19 could inhibit matrix metalloproteases(MMPs)，especially MMP7 could enhance mitogentic signaling and invasion[70].In addition，GRIM19 mutation tumour showed higher metastases capacity[71].Further works indicated that the loss ofGRIM19 could lead to the accumulation of H2O2in cytosol and increase the dimerization of STAT3 so as to activate antiapoptotic signaling[72].It was showed in some latest reports that the down-regulation ofGRIM19 was almost involved in the hallmark traits of cancerous cell including anti-apoptosis，invasive capacity，sustained proliferation，blood vessel formation，immune evasion and “Warburg effect”.It was assumed thatNDUFA13(GRIM19)may be a valuable candidate target tumor suppressor for SCLC.Such scenario remains to be investigated.

The brown module was enriched in cation transmembrane transport on the basis of GO enrichment analysis.C14orf159 was considered as hub gene which was up-regulated in the SCLC tissues(P-value=1.83E-03).C14orf159 is a D-glutamate cyclase(DGLUCY)which reversibly coverts free D-glutamate to 5-oxo-D-proline and H2O.It was found that the role of C14orf159 in gastric cancer was discussed only in one literature.According to the literature，the expression of C14orf159 in gastric cancer was significantly lower than that in normal gastric tissue.Moreover，its expression level was negatively associated with high TNM stage，while positively associated with the metastasis of lymph node.It was found that the overexpression of C14orf159 decreased the expression level of Snail，and the up-regulation of E-cadherin was also involved in the invasion of cancer cells.It was demonstrated in the literature that C14orf159 may block ERK-P90RSK signaling pathway associated with the invasion and proliferation in gastric cancer with the overexpression of C14orf159 in gastric cancer 803 cell line[73].However，C14orf159 was up-regulated in SCLC group in the present study.The role of C14orf159 in SCLC remained to be clarified，which may provide ideas for the diagnosis and treatment of SCLC.

In summary，8 gene modules were significantly altered in SCLC group based on the effective biological systemic technique WGCNA.These gene modules suggest that SCLC induced by chloroprene is a multistep process involving responses of multiple biological activities，including protein processing，nutrition metabolism，extracelluar matrix degradation，cell motility，signal transduction，vesicle transportation and immune response.Moreover，it is increasingly clear that these biological activities are regulated by the interaction networks of RNA，DNA and proteins.Hub genes are the most important nodes and potential therapeutic targets in these networks.A total of eight hub genes were identified in this study，includingNLN，TMEM40，TOP2A，COL5A2，SNX1，NDUFA13，C14orf159 andGIMAP4.We assumed that these hub genes might be valuable for the diagnosis and treatment for SCLC.

5 Conclusions

These findings may not only provide new research directions for future study，but also highlight potential biomarkers and novel therapeutic targets for SCLC.

Conflictofinterest

The authors declare no potential conflicts of interest.

Acknowledgements

The authors wish to show thanks to AMERICAN NATIONAL CANCER INSTITUE Surveillance，Epidemiology and End Results Program for providing epidemiological data of SCLC.