A tropomyosin receptor kinase family protein, NTRK2 is a potential predictive biomarker for lung adenocarcinoma

Data acquisition and reanalysis using different bioinformatics tools

The relevant bioinformatics data analysis of NTRK2 was obtained from several bioinformatics web resources, which were summarized in Table S1. The flow diagram of NTRK2 screen is shown in Fig. S1.

Oncomine is a cancer microarray or high-throughput sequencing data-mining platform, from which we can get gene expression signatures in human cancer tissues and cells (Rhodes et al., 2004). The data in Oncomine could be also linked into other public databases, such as GEO and The Cancer Genome Atlas (TCGA) (Hutter & Zenklusen, 2018). We conducted the comparison of NTRK2 expression across eight analyses between the LUAD and normal tissues. Additionally, Gene Expression Profiling Interactive Analysis (GEPIA) (Tang et al., 2017a), GE-mini (Tang et al., 2017b), Cancer RNA-Seq Nexus (CRN) (Li et al., 2016) and UALCAN (Chandrashekar et al., 2017), four additional cancer microarray or high-throughput sequencing data-mining databases, were employed to verify the results.

Wanderer is an interactive viewer, providing gene expression and DNA methylation data in human cancer (Diez-Villanueva, Mallona & Peinado, 2015), which enables us to screen for the possible methylation sites in the NTRK2 DNA sequence and to analyze the correlation between clinical characteristic of LUAD patients and NTRK2 expression and methylation sites. For the prognostic analysis, the Kaplan–Meier Plotter, a tool that can be used to assess the effect of genes on survival (Wang et al., 2018a), was utilized to describe the relationship between NTRK2 expression level, overall survival time (OS) and post-progression survival time (PPS). Further, the association between NTRK2 expression and disease-free survival (DFS) was completed through the GEPIA database.

Two datasets of the treatment-related transcriptome microarray, GSE6400 (Wang et al., 2007) and GSE54293 (Denisova et al., 2014), were acquired from the GEO database (Barrett & Edgar, 2008). Subsequently, the effects of NTRK2 expression on the chemotherapy for LUAD were analyzed.

The expression and methylation of NTRK2 correlation analysis were implemented by MethHC, which provided the information of DNA methylation and gene expression in human cancer (Huang et al., 2015). For the relevance between the disease prognosis and the methylation sites of NTRK2, the MethSurv tool was employed (Modhukur et al., 2018).

Using the cBioportal web tool (Gao et al., 2013), genes coexpressed with NTRK2 in LUAD were downloaded. Then, the STRING database (Szklarczyk et al., 2017) and Cytoscape software (Reimand et al., 2019) were used to complete the protein–protein interaction (PPI) network of these coexpression genes. Then, we utilized the DAVID bioinformatics resource (Huang, Sherman & Lempicki, 2009) to conduct the GO and KEGG pathway analysis of NTRK2 coexpression genes in LUAD samples. The web tools of WebGestalt (Wang et al., 2017) and PATHVIEW (Luo et al., 2017) were used for building a graphic.

Statistical analyses

The statistical tests were performed using SPSS 12.0 software (IBM Analytics). The results were expressed as the mean ± SD. Student t test, one-way ANOVA and K independent samples test were performed when appropriate. P < 0.05 was considered statistically significant.

NTRK2 is downregulated in LUAD tissues

The NTRK family consists of three members, NTRK1, NTRK2 and NTRK3. Through the bioinformatics analysis of databases, we evaluated the transcriptional levels of NTRK family members in LUAD. First, we used the Oncomine database to observe the expression of NTRK1, NTRK2 and NTRK3 in eight LUAD datasets (Beer et al., 2002; Bhattacharjee et al., 2001; Hou et al., 2010; Landi et al., 2008; Okayama et al., 2012; Selamat et al., 2012; Stearman et al., 2005; Su et al., 2007). The results showed that NTRK2 had significantly lower expression in LUAD through the comparison among nine datasets, whereas NTRK1 and NTRK3 showed no statistical significance (Fig. 1A). Therefore, NTRK2 was chosen as the research target. To verify the trend, we examined the NTRK2 expression in LUAD by GEPIA and GE-mini, and we discovered the NTRK2 expression was clearly reduced in LUAD compared with the normal tissues (Figs. 1B and 1C). In addition, the heatmap from CRN database further indicated the low expression of NTRK2 in LUAD tissues (Fig. 1D). Next, given some activated oncogenes, such as Erb-B2 receptor tyrosine kinase 2 (ERBB2) and MET, have been demonstrated the driver roles in LUAD (The Cancer Genome Atlas Research Network, 2014), we want to evaluate the association between NTRK2 and these oncogenes. The data from UALCAN revealed the significantly downregulated NTRK2 (P < 0.01), upregulated ERBB2 (P < 0.01) and upregulated MET (P < 0.01) in LUAD tissues (Fig. S2A). Spearman correlation analysis showed the negative association between the expression of NTRK2 and ERBB2 or MET (Fig. S2B). Taken together, all of the above data suggested that the decreased expression of NTRK2 contributed to LUAD tumorigenesis, supporting its tumor-inhibiting function in LUAD.

NTRK2 expression is associated with the clinical characteristics of LUAD patients

After determining the expression of NTRK2 in LUAD, we further analyzed the correlation between the NTRK2 expression level and the clinical characteristics of patients. Using the Wanderer database, we obtained a series of clinical data, and a summary of clinical characteristic parameters is provided in Table 1. As shown in this table, NTRK2 expression was significantly associated with gender (P = 0.007), pathologic T (P = 0.021), pathologic M (P = 0.006) and age (P = 0.036). Then, the Kaplan–Meier Plotter tool was used to evaluate the effects of NTRK2 expression on OS and PPS, confirming that the downregulated of NTRK2 expression was significantly related to shorter OS (P = 0.00029) (Fig. 2A) and PPS (P = 0.021) (Fig. 2B). Furthermore, we found that low NTRK2 expression was associated with RFS (P = 0.012) through using the GEPIA database (Fig. 2C). In conclusion, NTRK2 could be as a potential biomarker both for diagnosis and prognosis.

The roles of NTRK2 in LUAD therapies

For the purpose of identifying the exact function of NTRK2 in LUAD chemotherapy, two treatment-related transcriptome microarray datasets, GSE6400 and GSE54293, were obtained from the GEO database. Previous studies have demonstrated that actinomycin D (Bai et al., 2019) and MK2206 (Dai et al., 2017) were two promising antitumor drugs. In the GSE6400 dataset, we discovered that the expression of NTRK2 was apparently higher in the actinomycin D treatment group than in the mannitol-control group (P = 0.008) (Fig. 3A). In addition, for the GSE54293 dataset, the AKT inhibitor MK2206 could enhance the NTRK2 expression levels significantly (P = 0.009) (Fig. 3B). Collectively, the findings observed above suggested that NTRK2 might enhance the response of cancer cells to the chemotherapeutics.

The relationship between NTRK2 methylation and the clinical characteristics of LUAD patients

It is well-known that there is a negative correlation between DNA methylation and gene expression (Shi et al., 2017; Zhou et al., 2019). From the MethHC database, we observed that global NTRK2 methylation was significantly higher in LUAD samples compared with normal samples (P < 0.005) (Fig. 4A) and was negatively related to its expression (P = 0.000) (Fig. 4B), which gives further support for the low expression of NTRK2 in LUAD. Subsequently, the methylation site cg03628748 was screened out of the data (P = 4.35E-12) (Table S2) acquired from the Wanderer database. Then, the relationship between cg03628748 and the clinical characteristics of LUAD patients was examined, and results showed that cg03628748 was significantly related to Kras mutation (P = 0.038) and pathologic T (P = 0.000) (Table 2). Moreover, there was a significant negative correlation between higher methylation value of cg03628748 and shorter OS in LUAD patients (P = 0.034), which was analyzed by using the web tool of MethSurv (Fig. 4C).

Functional enrichment analysis of NTRK2-associated coexpression genes

Using the cBioPortal database, 15,146 genes that were notably coexpressed with NTRK2 in the LUAD samples were acquired. The volcano plot was established for exhibiting between the altered and unaltered NTRK2 expression group (Fig. 5A). Next, we singled out 219 NTRK2-associated codifferentially expressed genes (co-DEGs) with the criteria of P value < 0.05 and |log Ratio| ≥ 2 (Table S3). Then, a PPT network of the co-DEGs was performed by using the STRING database and Cytoscape software (Fig. 5B). For the purpose of comprehending the biological function for these co-DEGs, GO and KEGG analyses were conducted by the WebGestalt and PATHVIEW web tools, respectively. The biological processes showed that these co-DEGs were mainly connected with biological regulation and metabolic processes (Fig. 5C). For the analysis of cellular components, the coexpression genes were mainly localized on cell membranes (Fig. 5D). For molecular function, protein binding was primarily enriched for these coexpression genes (Fig. 5E). Furthermore, the KEGG pathway demonstrated that these genes were involved in the process of xenobiotics and drug metabolism by cytochrome P450 (Table S4).