Abstract

   Differential expressions of certain genes during tumorigenesis may
   serve to identify novel manageable targets in the clinic. In this work
   with an integrated bioinformatics approach, we analyzed public
   microarray datasets from Gene Expression Omnibus (GEO) to explore the
   key differentially expressed genes (DEGs) in non-small cell lung cancer
   (NSCLC). We identified a total of 984 common DEGs in 252 healthy and
   254 NSCLC gene expression samples. The top 10 DEGs as a result of
   pathway enrichment and protein–protein interaction analysis were
   further investigated for their prognostic performances. Among these, we
   identified high expressions of CDC20, AURKA, CDK1, EZH2, and CDKN2A
   genes that were associated with significantly poorer overall survival
   in NSCLC patients. On the contrary, high mRNA expressions of CBL, FYN,
   LRKK2, and SOCS2 were associated with a significantly better prognosis.
   Furthermore, our drug target analysis for these hub genes suggests a
   potential use of Trichostatin A, Pracinostat, TGX-221, PHA-793887,
   AG-879, and IMD0354 antineoplastic agents to reverse the expression of
   these DEGs in NSCLC patients.

   Keywords: Non-small cell lung cancer, differential gene expressions,
   integrative bioinformatics analysis, drug target potential analysis

Introduction

   Lung cancer is one of the deadliest diseases all around the world. The
   GLOBOCAN estimations by the International Agency for Research on Cancer
   predict approximately 13 million new cancer cases by the year 2040.
   Currently, lung cancer is the most commonly diagnosed form of cancer
   (11.6% of total cases) and the leading cause of cancer deaths (18.4% of
   total deaths).^ [29]1 Histologically, lung cancer can be divided into
   non-small cell lung cancer (NSCLC), which accounts for approximately
   85% of lung cancer cases and small cell lung cancer.^ [30]2 The lack of
   early-stage symptoms and effective diagnostic markers restrain the
   treatment success in NSCLC. Thus, most of the patients are diagnosed at
   an advanced stage, and half of them have distant metastatic disease at
   initial diagnosis.^ [31]3 Over the last decade, there have been
   considerable improvements in chemotherapy, radiation therapy, surgery,
   and targeted therapy for lung cancer. Especially with the recent
   advances in molecular biology, significant progress has been made
   through molecule-targeted therapy in NSCLC. For instance, it has been
   shown that about 20% of Caucasian and 50% of Asian NSCLC patients had
   mutations on their EGFR (epidermal growth factor receptor) genes.^
   [32]4 However, with the application of EGFR targeting small tyrosine
   kinase inhibitors (EGFR-TKIs) erlotinib and gefitinib, both response
   rate and median survival of these patients were found to be improved.^
   [33]5 Also, approximately 7% of NSCLC patients bear an activated ALK
   gene, but treatment with ALK inhibitor crizotinib has been shown to
   improve both response and 6-month progression-free survival rates in
   these patients.^ [34]6 Nonetheless, the 5-year survival rates of NSCLC
   patients remain low with a poor prognosis due to the development of
   intrinsic or acquired chemoresistance against therapeutic drugs.^ [35]7

   Non-small cell lung cancer is a result of the accumulation of several
   genetic and epigenetic modifications, which could have originated from
   multiple reasons.^ [36]8 The discovery and characterization of new
   prognostic or diagnostic markers together with enhanced therapeutic
   approaches for NSCLC are of top priority for the successful treatment
   of this disease. Unfortunately, information on the heterogeneous nature
   of the tumor and the involvement of affecting factors in the process of
   NSCLC tumor development are far from completely resolved. Therefore, it
   is highly important to shed light on the molecular mechanisms governing
   the pathogenesis of NSCLC and to identify effective diagnostic and/or
   prognostic biomarkers for novel treatment options. High-throughput
   technologies such as microarrays and integrated bioinformatics methods
   are used to obtain gene alterations during tumorigenesis and to
   identify novel prognostic markers in patients with cancer.^[37]9,[38]10
   For instance, in a recent study, Huang and Gao^ [39]11 have
   demonstrated that CDC20, CENPF, KIP2C, and ZWINT genes were
   differentially expressed in NSCLC tissues. Similarly, Xiao et al^
   [40]12 have identified CCNB1, CCNA2, CEP55, PBK, and HMMR as hub genes
   and key differentially expressed genes (DEGs) associated with NSCLC by
   bioinformatics analyses. Interestingly, Wang et al^ [41]13 have shown
   CCND1 as the most enriched gene and a potential prognostic biomarker in
   NSCLC through a gene set enrichment analysis. More recently, Zhang et
   al^ [42]14 have identified TOP2A, CCNB1, BIRC5, and TTK as well as
   miR-21-5p and miR-31-5p to be significantly associated with NSCLC
   prognosis through an integrative analysis of mRNA and miRNA expression
   profiles. Nevertheless, the genes that are discovered by 1 cohort might
   be difficult to be identified in other cohorts.^ [43]15 For this
   reason, it is essential to validate genes in several independent
   studies.

   In this study, we sought to identify potential therapeutic targets or
   prognostic biomarkers among the DEGs associated with NSCLC through an
   integrated bioinformatics approach. For this purpose, we retrieved 4
   different microarray datasets from Gene Expression Omnibus (GEO)
   database and screened for DEGs between NSCLC tumor and neighboring
   normal tissues. After gene set enrichment analysis to identify
   associated biological processes, a protein–protein interaction (PPI)
   network analysis was performed to elucidate potential key DEGs. We also
   explored the significance of candidate key DEGs and their correlation
   to patient prognosis through survival analysis. Finally, potential
   therapeutic drugs that may target and reverse the expression of these
   key DEGs were predicted by using the L1000CDS2 signature search engine.

Materials and Methods

Microarray datasets

   A comprehensive database search was conducted for identifying
   appropriate datasets including NSCLC tumor tissue and matched adjacent
   normal samples from the public GEO database.^ [44]16 To avoid
   microarray platform differences, the datasets originating from
   Affymetrix microarrays utilizing Human Genome U133 Plus 2.0 chips
   (Thermo Fisher Scientific, Inc., Waltham, MA, USA) were selected. Four
   datasets with accession numbers [45]GSE18842, [46]GSE19804,
   [47]GSE27262, and [48]GSE102287 were identified and downloaded for
   further analysis ([49]Table 1). [50]GSE19804^ [51]18 included 60 pairs
   of NSCLC tumor and matched adjacent normal lung tissue, while
   [52]GSE18842^ [53]17 comprised 46 NSCLC tumors and 45 paired controls,
   [54]GSE27262^ [55]19 contained 25 tumors and normal tissue pairs from
   stage I lung adenocarcinoma, and [56]GSE102287^ [57]20 contained 66
   matched NSCLC tumor and normal tissues. A total of 506 gene expression
   samples including 252 healthy and 254 NSCLC tissues were evaluated in
   this study.

Table 1.

   Transcriptome datasets employed in the present study.
   Source—ID Purpose No. of tumor samples No. of control samples
   References