Comparison of tumor related signaling pathways with known compounds to determine potential agents for lung adenocarcinoma.

BACKGROUND: This study compared tumor-related signaling pathways with known compounds to determine potential agents for lung adenocarcinoma (LUAD) treatment.
METHODS: Kyoto Encyclopedia of Genes and Genomes signaling pathway analyses were performed based on LUAD differentially expressed genes from The Cancer Genome Atlas (TCGA) project and genotype-tissue expression controls. These results were compared to various known compounds using the Connectivity Mapping dataset. The clinical significance of the hub genes identified by overlapping pathway enrichment analysis was further investigated using data mining from multiple sources. A drug-pathway network for LUAD was constructed, and molecular docking was carried out.
RESULTS: After the integration of 57 LUAD-related pathways and 35 pathways affected by small molecules, five overlapping pathways were revealed. Among these five pathways, the p53 signaling pathway was the most significant, with CCNB1, CCNB2, CDK1, CDKN2A, and CHEK1 being identified as hub genes. The p53 signaling pathway is implicated as a risk factor for LUAD tumorigenesis and survival. A total of 88 molecules significantly inhibiting the five LUAD-related oncogenic pathways were involved in the LUAD drug-pathway network. Daunorubicin, mycophenolic acid, and pyrvinium could potentially target the hub gene CHEK1 directly.
CONCLUSION: Our study highlights the critical pathways that should be targeted in the search for potential LUAD treatments, most importantly, the p53 signaling pathway. Some compounds, such as ciclopirox and AG-028671, may have potential roles for LUAD treatment but require further experimental verification.

Introduction

Lung adenocarcinoma (LUAD) is a subtype of non‐small cell lung cancer (NSCLC). It is the most common type of lung cancer in the world, with particularly poor prognosis in patients who have reached advanced stages.1, 2, 3, 4, 5 Because of the low diagnosis rates in the early stages of the disease, most patients are diagnosed with LUAD when it is too late to effectively treat with surgical removal (approximately 70% in China). With the development of tumor‐targeted drugs, some LUAD patients may receive a curative effect. EGFR‐tyrosine kinase inhibitors (TKIs) have become an indispensable therapeutic strategy in advanced NSCLC patients with EGFR sensitive mutations, such as those in exons 19 and 21. Nevertheless, drug resistance frequently occurs during therapy, minimizing the ability to control disease progression.6, 7, 8, 9, 10, 11, 12, 13, 14 Urgent investigation of the pathways related to the mechanisms of drug resistance is being conducted to determine the critical biological agents of this process. Differentially expressed genes (DEGs) in lung cancer and non‐cancerous lung tissues can be used to identify novel signaling pathways and reliable biomarkers that may lead to therapeutic targets.

The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/) is a collaborative database that features comprehensive, multi‐dimensional maps of the key genomic changes in cancer. The database contains both tumor tissues and matched normal tissues from a large number of patients and has been widely used in the research community.15, 16, 17, 18, 19, 20 The Connectivity Map (CMap; http://www.broadinstitute.org/cmap/) is a resource that uses transcriptional expression data to probe relationships between diseases, cell physiology, and therapeutics. CMap seeks to uncover novel treatments for a variety of diseases, including cancers, and neurological and infectious diseases.21, 22, 23, 24, 25

The combination of the enrichment inquiry of crucial cancer‐related genes from TCGA and the CMap analysis indicates that signaling genes may be prospective drug targets for therapy in certain tumors26 such as ovarian cancer21, 27 and hepatocellular carcinoma.25 In the current study, we performed an in silico investigation using CMap analysis of the DEGs gathered from a large population of LUAD patients (n = 483). This process led to the identification of prospective candidate drugs for future LUAD treatments.

Methods

Lung adenocarcinoma (LUAD) signaling pathways

TCGA provided 19 123 messenger RNA (mRNA) expression levels from 483 LUAD cases by RNA‐sequencing. For comparison, 347 cases of matched normal and genotype‐tissue expression controls were provided by the Gene Expression Profiling Interactive Analysis platform from the University of California Santa Cruz Xena Project (http://xena.ucsc.edu).18, 28, 29, 30, 31 All expression data were transformed into Log 2 (transcripts per kilobase million [TPM] + 1) for further calculation. The fold change of the mRNA expression levels was calculated using the difference between the LUAD samples and the non‐cancerous controls. The cutoffs to classify an mRNA as being abnormally expressed were Log 2 fold change > 1 and a q‐value < 0.05. Analysis of variance was utilized for the differential expression analysis. The DEGs identified from the LUAD tissues were subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathway analysis with the Database for Annotation, Visualization and Integrated Discovery (DAVID). A false discovery rate of < 0.1 was regarded as statistically significant for the KEGG pathway analysis.32, 33, 34, 35

Signaling pathways of known compounds from the Connectivity Map (CMap)

CMap provides a data resource to interpret a drug's prospective molecular mechanisms using genome‐wide gene signatures. Over 7000 expression profiles corresponding to 1309 compounds are publicly available from the online CMap database (build02). The microarray data for each of the compounds was downloaded and the DEGs were calculated by comparing treatment and control groups.21, 22, 23, 24, 25 A fold change of > 2 or < 0.5 was used to identify the DEGs, which were subsequently subjected to KEGG pathway analysis using the procedure followed in a previous report.25

Potential novel drugs for LUAD

The significant pathways involved in the onset of LUAD and those affected by the CMap drugs were integrated. The compounds involved in these overlapping pathways were considered to be potential drugs for the treatment of LUAD. Rank‐based pattern‐matching processes were employed to perform a similarity evaluation of various pharmaceutical substances. The statistical significance was computed based on random permutations. The top 10 compounds are listed in Figure S1 along with their molecular structures, as provided by PubChem Compound (https://www.ncbi.nlm.nih.gov/pccompound).

Gene ontology analysis and protein–protein interaction of the genes in the overlapping pathways

To reveal the overall function of genes enriched in the overlapping pathways, gene ontology analysis was performed by DAVID and the results were displayed in a gene ontology plot using R software ggplot2 packages (R Foundation for Statistical Computing, Vienna, Austria). Protein–protein interaction was conducted to draw the gene network.36, 37, 38, 39

Construction of the drug‐pathway network

Because the five overlapping pathways may be important for targeting novel molecules, a drug‐pathway network was constructed. Molecules corresponding to these five pathways were collected and entered into Cytoscape software (National Institute of General Medical Sciences, Bethesda, MD, USA). Lines between the pathways and molecules represent a potential relationship between the drugs and the corresponding pathways.

Clinical significance of hub genes enriched in the p53 signaling pathway

The p53 signaling pathway was the most significant pathway related to the effect of known compounds. The expression patterns and prognostic values of the hub genes from the p53 signaling pathway were comprehensively analyzed. The mRNA expression levels of these genes were obtained from the Gene Expression Profiling Interactive Analysis online database. Kaplan–Meier survival plots were used to assess the prognostic roles of these genes.18, 28, 29, 30, 31 The corresponding protein expression levels were collected from the Human Protein Atlas to validate the different expression trends between LUAD and non‐tumor lung tissues.40, 41, 42, 43, 44, 45, 46 As multiple targets provide a more powerful prognostic value and estimate of treatment potential, several hub genes were pooled as a group and their prognostic implications were assessed based on LUAD TCGA data (n = 475), as calculated by the SurvExpress database (47, http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp). All probe sets and records were averaged per sample. A prognostic index (PI) of these hub genes was calculated by Cox survival analysis and the corresponding hazard ratio was analyzed.47, 48, 49, 50, 51, 52, 53

Predicted binding affinities between hub genes and compounds using molecular docking

To obtain detailed information on how these novel drugs exert anti‐tumor functions, molecular docking between the compounds and the proteins of the hub genes enriched in the most significant pathway (the p53 signaling pathway) was carried out using the Surflex‐Dock program in Sybyl X‐2.0 (Tripos Inc., St Louis, MO, USA). Docking scores were calculated to represent binding affinities.54, 55, 56

Results

Identification of LUAD signaling pathways

Altogether, 1993 DEGs were gathered from 483 LUAD patients, including 514 upregulated and 1479 downregulated genes (Fig S2). KEGG analysis showed that 57 pathways were significantly related to LUAD. Unsurprisingly, several among these were well‐studied pathways involved in the initiation and progression of LUAD: focal adhesion; extracellular matrix‐receptor interaction; Rap1, PI3K‐Akt, Ras, chemokine, MAPK, p53, and VEGF signaling pathways; pathways in cancer; proteoglycans in cancer; cell adhesion molecules, and cell cycle (Fig 1; Table S1). Drugs that target these pathways may have the potential to treat LUAD, thus potential candidate molecules were identified.

Figure 1

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway of differentially expressed genes. The length of the bar represents the weight of the number of genes in the pathway. The color represents the significance of the enriched pathways.

Pathways affected by known molecular agents in CMap

The global associations between 1309 bioactive small molecular agents and the identified pathways were obtained from the CMap database. Altogether, 35 enriched KEGG pathways were obtained with a false discovery rate of < 0.1. The top 10 significant pathways are displayed in Figure 2a. After the integration of the 57 LUAD‐related pathways and the 35 pathways affected by the small molecules, five overlapping pathways were identified that could be targeted by known small molecules (Fig 2b), including pathways in cancer and chemokine, MAPK, p53, and Fc epsilon RI signaling pathways. The p53 signaling pathway was the pathway affected by the most molecules. A total of 44 (44/82) molecules are characterized as having an anti‐tumor function that targets the p53 signaling pathway.

Figure 2

Discovery of potential targeted pathways by Connectivity Map (CMap). (a) The enriched pathways that were obtained from CMap. (b) Integration of the 57 lung adenocarcinoma (LUAD)‐related pathways and the small molecules that affected these pathways.

Gene ontology and protein–protein interaction analyses

Because these five pathways play essential roles in LUAD tumorigenesis and are likely vital for the discovery of novel drugs, genes enriched in the five pathways were collected and gene ontology analysis was performed to reveal their characteristics and possible molecular mechanisms. The genes characterized as being involved in biological processes were most associated with cell chemotaxis, extracellular matrix organization, and the chemokine signaling pathway (Fig 3a). The genes characterized as being related to cellular components were most associated with the extracellular region, the extracellular space, and the plasma membrane (Fig 3b). As for molecular function, these genes were markedly involved in chemokine, PIK3CA, and Ras guanyl‐nucleotide exchange factor activity (Fig 3c). A protein–protein interaction network consisting of the genes in the five pathways was constructed, which revealed a short list of genes that were closely related to each other, including IL6, GNG11, FGF2, PIK3R1, and MMP9 (Fig 4).

Figure 3

Gene ontology of genes enriched in the five overlapping pathways. (a) Biological process; (b) cellular component; and (c) molecular function.

Figure 4

Protein–protein interactions of genes enriched in the five overlapping pathways. The size of the circle represents the weight of the gene in the network. The thickness of the line reflects the connection score between two genes.

Based on the five common pathways involved in the onset of LUAD and the effect of the unknown molecular drugs, a drug‐pathway interaction network was constructed. A total of 88 molecules significantly inhibiting the five important LUAD‐related oncogenic pathways were involved in the drug‐pathway network. These compounds may have potential as future anti‐LUAD drugs (Fig S3). To reveal the interacting effects of the genes involved in each pathway, a protein–protein interaction link was again constructed for each pathway (Fig S4).

Candidates for novel effective drugs to treat LUAD

The relationships between the five pathways and the novel molecules were analyzed. The top 10 significant drugs all interacted with the p53 signaling pathway (Fig 5): etoposide, irinotecan, camptothecin, pyrvinium, metyrapone, trifluridine, mycophenolic acid, resveratrol, AG‐028671, and daunorubicin (Fig S4).

Figure 5

p53 signaling pathway. The red star indicates differentially expressed genes of lung adenocarcinoma enriched in the p53 signaling pathway.

Clinical significance of the p53 signaling pathway in LUAD

The five hub genes of the p53 signaling pathway are CCNB1, CCNB2, CDK1, CDKN2A, and CHEK1. As expected, the mRNA expression levels of these genes were markedly upregulated in LUAD tissues compared to normal lung tissues (Fig S5). Protein expression levels also showed a consistent trend (Fig 6). Significantly, overexpression of these genes is associated with lower survival rates in LUAD patients (Fig 7). Consequently, a PI model was generated to predict a patient's overall chance of survival in combination with the upregulation of these five genes (Fig 8), using the formula: PI = (0.378* expression level of CCNB1) + (−0.781* expression level of CCNB1) + (0.704* expression level of CDK1) + (0.012* expression level of CDKN2A) + (0.315* expression level of CHEK1). Patients were divided into high and low‐risk groups based on the PI model, and the five genes were all upregulated in the high‐risk group (Fig 9a). Multivariate analysis demonstrated a hazard risk of 1.55 (95% confidence interval 1.14–2.11; P = 0.005), indicating that the p53 signaling pathway is a risk factor in the survival outcome of LUAD patients (Fig 9b).

Figure 6

The protein expression level of four hub genes in the p53 signaling pathway stained by immunohistochemistry. CCNB1 expression is significantly higher in (c,d) lung adenocarcinoma (LUAD) than (a,b) non‐tumor tissues. CCNB2 expression is significantly higher in (g,h) LUAD than (e,f) non‐tumor tissues. CDK1 expression is significantly higher in (k,l) LUAD than (i,j) non‐tumor tissues. CCNB2 expression is significantly higher in (o,p) LUAD than (m,n) non‐tumor tissues. Magnification: ×100 (a,c,e,g,i,k,m,o) or ×400 (b,d,f,h,j,l,n,p).

Figure 7

Kaplan–Meier analysis revealed the prognostic value of the five hub genes. Prognostic value of CCNB1 for predicting (a) overall survival (OS) and (b) disease‐free survival (DFS) of lung adenocarcinoma (LUAD) patients. Prognostic value of CCNB2 for predicting (c) OS and (d) DFS of LUAD patients. Prognostic value of CDK1 for predicting (e) OS and (f) DFS of LUAD patients. Prognostic value of CDKN2A for predicting (g) OS and (h) DFS of LUAD patients. Prognostic value of CHEK1 for predicting (i) OS and (j) DFS of LUAD patients.

Figure 8

Construction of prognosis index (PI). (a) Available clinical information related to risk group, PI, and clinical outcome. (b) A heatmap representation of the gene expression values. (c) The low and high score groups for the PI in lung adenocarcinoma patients. (d) A box plot across risk groups ordered by PI.

Figure 9

Evaluation of prognosis index in the survival of lung adenocarcinoma patients in The Cancer Genoma Atlas. (a) Different expression of the five genes between the high and low‐risk groups. (b) Kaplan–Meier analysis showed that patients in the high‐risk group had a shorter overall time than those in the low‐risk.

Correlations between hub genes and compounds identified using molecular docking

To examine if the potential effects of the molecules involved in the p53 signaling pathway were related to direct targeting between compounds and genes, molecular docking analysis was performed. The hub gene from the p53 signaling pathway that revealed the greatest clinical significance was CHEK1. Little research regarding the role of CHEK1 in LUAD has been reported, thus CHEK1 was selected for molecular docking analysis. Intriguingly, the top 10 molecules all exerted moderate binding capacity to CHEK1 (Table S2). Daunorubicin, mycophenolic acid, and pyrvinium exhibited the highest connection scores (Fig 10).

Figure 10

The top three most effective molecules related to CHEK1 from molecular docking. (a,b) Mycophenolic acid with a docking score of 6.9634; (c,d) daunorubicin with a docking score of 6.3656; and (e,f) pyrvinium with a docking score of 6.2614.

Discussion

In the current study, we speculated that prospective compounds for LUAD treatment and therapy are involved in the signaling pathways related to LUAD tumorigenesis. LUAD DEGs were identified in TCGA and subjected to KEGG signaling pathway analysis. The identified pathways were overlapped with those available from CMap, resulting in five significant common pathways. Among these five pathways, the p53 signaling pathway consisting of 13 genes was related to the most molecules from the CMap dataset. Based on multiple data mining, five hub genes from the p53 signaling pathway (CCNB1, CCNB2, CDK1, CDKN2A, and CHEK1) were found to play an essential role in the occurrence and development of LUAD. Molecule docking techniques revealed that daunorubicin, mycophenolic acid, and pyrvinium could bind directly to CHEK1. In addition, several known compounds that have never been associated with LUAD, including AG‐028671 and ciclopirox, were identified as being worthy of further investigation for the treatment of LUAD.

It is interesting to discover that five classical signaling pathways play pivotal roles in the tumorigenesis of LUAD and can be targeted by known compounds. These five pathways, including pathways in cancer, and chemokine, MAPK, p53, and Fc epsilon RI signaling pathways, have been studied in multiple cancers, including LUAD.57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68 After analyzing these five pathways in CMap, several known compounds were identified. Unsurprisingly, many among the top 10 have been studied for the treatment of LUAD, including etoposide,69, 70, 71 irinotecan,72, 73, 74, 75, 76 resveratrol,77, 78, 79 and daunorubicin.80, 81, 82 This finding confirms that several well‐known anti‐cancer drugs used in LUAD treatment fulfill their function by targeting the five aforementioned signaling pathways. This also demonstrates the feasibility of the current methods of combining pathways identified by DEGs and CMap.

Little is known of the relationship of four of the compounds identified in this analysis to LUAD: pyrvinium,83 metyrapone,84 trifluridine,85, 86 and mycophenolic acid.87, 88 These compounds have the potential to be used in future LUAD treatments, as they can inhibit vital LUAD pathways. Most importantly, two of these known compounds, ciclopirox and AG‐028671, have never been evaluated in LUAD. Ciclopirox is a protein synthesis inhibitor that decreases DNA and RNA replication and protein synthesis, and has been studied as a potential treatment for other cancers.89, 90, 91 However, no information could be obtained for AG‐028671, which may be a new drug for the treatment of LUAD. These in silico findings should be confirmed with further experiments.

Because the p53 signaling pathway was the most significant pathway in the CMap analysis and the top 10 compounds were all enriched in this pathway, we examined the clinical role of this pathway using different levels of data mining, including mRNA and protein. The five hub genes (CCNB1, CCNB2, CDK1, CDKN2A, and CHEK1) were all profoundly upregulated in the LUAD tissues. Moreover, all of the hub genes are risk factors for LUAD survival. When these five genes were placed into a risk pool, the PI verified them to be an independent factor to predict the overall survival of LUAD patients, confirming that the p53 signaling pathway has a critical function in the progression of the disease. These results verify the significance of the novel drugs that have been discovered. Using molecular docking, we found that the top 10 molecules can significantly affect the hub genes, some by targeting CHEK1 directly, especially daunorubicin, mycophenolic acid, and pyrvinium. As the genes involved in the overlapping pathways play substantial roles in the development of LUAD, the interference of these genes by powerful drugs can suppress their function. However, future in vivo or in vitro experiments are required to confirm whether these drugs could potentially provide meaningful assistance in LUAD treatment.

In conclusion, our study highlights critical pathways for the potential treatment of LUAD, most significantly the p53 signaling pathway. Some compounds known for treating LUAD target this pathway, including daunorubicin and mycophenolic acid. Pyrvinium targets the hub gene CHEK1 directly. However, the effects of several potential LUAD drugs are not well characterized, especially ciclopirox and AG‐028671. Future work should concentrate on in vitro and in vivo experiments for these novel compounds to determine if they can be advanced to clinical trials.

Disclosure

No authors report any conflict of interest.

Figure S1 Chemical structures of the top 10 significant molecules. (a) etoposide; (b) irinotecan; (c) camptothecin; (d) pyrvinium; (e) metyrapone; (f) trifluridine; (g) mycophenolic acid; (h) resveratrol; (i) AG‐028671; and (j) daunorubicin.

Click here for additional data file.

Figure S2 Differentially expressed genes (DEGs) between lung adenocarcinoma and normal tissues. (a) The DEG positions based on GRCh38.p2 (NCBI); and (b) a volcano plot displaying the DEGs.

Click here for additional data file.

Figure S3 Small molecular drugs and their predicted pathways in lung adenocarcinoma. The yellow nodes represent drugs, and the blue nodes represent pathway.

Click here for additional data file.

Figure S4 Protein–protein interactions of five overlapping pathways. (a) Pathway in cancer; and (b) MAPK, (c) p53, (d) chemokine, and (e) Fc epsilon RI signaling pathways.

Click here for additional data file.

Figure S5 The messenger RNA expression level of five hub genes in p53 signaling pathway: (a) CCNB1, (b) CCMB2, (c) CDK1, (d) CDKN2A, and (e) CHEK1.

Click here for additional data file.

Table S1 Significant Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of differentially expressed genes in lung adenocarcinoma.

Click here for additional data file.

Table S2 Molecular docking results for CHEK1.

Click here for additional data file.