Upregulation of desmoglein 2 and its clinical value in lung adenocarcinoma: a comprehensive analysis by multiple bioinformatics methods

Patients and tissues

Forty paired lung adenocarcinoma tissues and non-cancer tissues (>five cm adjacent to the tumor) were collected from patients who underwent pulmonary surgery in the department of thoracic surgery in our hospital from March 2016 to June 2019. All patients had no history of other cancers and underwent no other prior treatment including chemotherapy, radiotherapy or targeted therapy before surgery. This study was approved by the Ethics Committee of the Second Affiliated Hospital of Xi’an Jiaotong University (2016036). Written informed consent was obtained from all participants.

Tissue microarray (TMA) and immunohistochemistry

Tissue microarray (HLugA150CS03) containing 75 paired lung adenocarcinoma tissues and adjacent non-tumor tissues was purchased from Shanghai Outdo Biotech (Shanghai, China). Then immunohistochemistry (IHC) assay was performed. Tissue sections were treated with EDTA buffer under high pressure at high temperature to retrieve antigen. Sections were incubated with primary antibody DSG2 (1:250, ab150372, Abcam, USA) at 4 ^∘C overnight. Biotinylated anti-IgG was added and incubated at room temperature for 1 hour. The slides were stained with DAB and counterstained with haematoxylin after incubating with streptomycin-HRP for 30 minutes. Samples were visualized using diaminobenzidine system and hematoxylin re-dying, and analyzed under microscope (OLYMPUS CX41, Japan).

Tissue microarrays were scored independently by two pathologists who were blinded to the clinicopathologic features and outcomes of the patients. The staining was analyzed under low magnification (40×) and high magnification (200×). The intensity of staining was divided into four grades (intensity scores): negative (0), weak (1), moderate (2) and strong (3). Immunoreactivity was divided into five grades (percentage scores) according to the percentage of stained cells: no staining (0), 1%–25% (1), 26%–50% (2), 51%–75% (3) and >75% (4). Then, a final histological overall score was calculated by the multiplication of the two values. An overall score of 0–12 was calculated and graded as low (score < 7) or high (score ≥ 7).

Qrt-pcr

Fast1000 (Xfyangbio, China) was used to extract the total RNA from all the specimens according to the manufacturer’s protocol. After the reverse transcription, the TB Green^® Premix Ex Taq™ II (Takara Biotechnology Co., Ltd.) was utilized to detect the gene amplification and qRT-PCR was performed on the CFX96 Touch™ Real-Time PCR Detection System (Bio-rad) in standard mode for 40 cycles. The primers were as follows: ACTIN forward, 5′-TGGCACCCAGCACAATGAA-3′  and reverse, 5′-CTAAGTCATAGTCCGCCTAGAAGCA-3′; DSG2 forward, 5′- ATGACGGCTAG GAACACCAC-3′  and reverse, 5′-GGGTCAGTTTGTGGCTGACT-3′(Vite et al., 2013). All experiments were conducted in triplicate. The difference of DSG2 expression between cancer tissues and non-cancer tissues was calculated using the formula 2^−ΔCq.

Data from online databases

RNA-Seq data of DSG2 in LUAD were downloaded from TCGA (https://portal.gdc.cancer.gov/). After repeated values were deleted, LUAD consisted of 513 lung adenocarcinoma tissues and 59 lung normal tissues. Among them, there were 57 paired tissues. The original expression data were displayed in fragments per kilobase million (FPKM). All the data were transformed into TPM (transcripts per million reads) and normalized by log₂(TPM+1) for the following analyses. Clinical information including age, gender, tumor-node-metastasis (TNM) stage, smoking status were also obtained. The original data from other studies were downloaded from Oncomine database (http://www.oncomine.org/). Also, the expression level of DSG2 protein was detected from the Human Protein Atlas (http://www.proteinatlas.org/) (Ponten, Jirstrom & Uhlen, 2008).

Integrated meta-analysis

A comprehensive meta-analysis was performed to determine the combined expression value of DSG2 in LUAD groups and non-cancer groups by combining the three sources (qRT-PCR data, TCGA data and Oncomine datasets). I-squared test was used to evaluate heterogeneity. When I²<50%, a fixed-effects model was used, while I²>50%, the random-effects model was selected (Higgins et al., 2003). Standard mean difference (SMD) with a 95% confidential interval (CI) was used to assess the effect sizes of these pooled data. A summary receiver operating characteristic (SROC) curve was constructed to describe the diagnostic ability of DSG2 in LUAD. In addition, Begg’s test was carried out to confirm whether publication bias existed in the studies.

Co-expressed genes and pathways identification and analysis

In order to find out potential mechanism, Pearson test was used to identify co-expressed genes. The cutoff value was set as —R—>0.4 and P < 0.05. The interaction between DSG2 and co-expressed genes were processed by the Cytoscape 3.7.1 software to generate a visual image. GO enrichment and KEGG pathway analysis of these co-expressed genes were performed by Database for Annotation, Visualization and Integrated Discovery (DAVID) online database (https://david.ncifcrf.gov/) (Huang da, Sherman & Lempicki, 2009). P < 0.05 and false discovery rate (FDR) <0.05 were considered to have significance significantly. GEPIA (http://gepia.cancer-pku.cn/index.html) contains the RNA sequencing expression data of 9,736 tumors and 8,587 normal cases from TCGA and the Genotype-Tissue Expression (GTEx) projects (Tang et al., 2017). In the present study, GEPIA was used to compare the expression levels of the co-expressed genes between lung adenocarcinoma tissues and normal tissues and investigate the prognostic significance.

Statistical analysis

All data were analyzed using R software (version 3.5.1) and GraphPad Prism 8. Meta-analysis was analyzed using STATA 15. The student’s t-test was used to compare the differential expression level of DSG2 between LUAD tissues and adjacent normal lung tissues. The associations between the level of DSG2 and clinicopathological parameters were evaluated using Mann-Whitney U test or Kruskal–Wallis H test. The chi-squared test was utilized to determine whether there was a difference in the distribution of DSG2 samples among the different categories in the TMA. The prognostic value was evaluated by Cox regression analyses and Kaplan–Meier. Multivariate Cox analysis was used to compare the impact of DSG2 expression on survival along with other clinical traits, the median DSG2 expression was regarded as the cut-off value. The associations between DSG2 expression and co-expressed genes were determined by Pearson correlation coefficient. All p value <0.05 was considered statistically significant.

Clinical value of DSG2 expression in lung adenocarcinoma

In the current study, a total of 40 paired lung adenocarcinoma samples were utilized to determine the expression of DSG2 by qRT-PCR. As shown in Fig. 1A, DSG2 was significantly over-expressed in lung adenocarcinoma than corresponding non-tumorous normal tissues (p < 0.001). To further identify the importance of DSG2 in the lung adenocarcinoma progression, the correlation between DSG2 and clinicopathological features was analyzed. The expression level of DSG2 in patients with early TNM tumor stage (I–II) was lower than patients with advanced TNM stage (III–IV) (0.063 ± 0.040 vs 0.128 ± 0.098, p = 0.023). For tumors greater than three cm in size, a significant increase in DSG2 expression was observed (0.109 ± 0.087 vs 0.052 ± 0.027, p = 0.027). Patients with lymph node metastasis exhibited higher DSG2 expression than patients without lymph node metastasis (0.113 ± 0.087 vs 0.053 ± 0.034, p = 0.014). No significant association was found between DSG2 expression and other clinicopathological characteristics, including age, gender and smoking history (Table 1).

To further verify our results, data from other studies including TCGA data and Oncomine datasets were used. Original data consisted of 513 LUAD samples and 59 normal lung samples with clinicopathological information were downloaded from TCGA. Among them, there were 57 paired LUAD samples. As shown in Fig. 1B, increased expression of DSG2 was observed in cancer tissues compared to normal lung tissues (p < 0.001). The analysis of the paired LUAD samples exhibited a consistently higher expression level in lung cancer (Fig. 1C). In addition, the associations between DSG2 expression and clinicopathological parameters were shown in Table 2. A higher expression of DSG2 is significantly correlated with a higher T stage (p = 0.001), N stage (p < 0.001) and TNM stage (p < 0.001). Also, the expression of DSG2 was closely related with gender (p = 0.017). With regard to age, smoking history and M stage, no statistical significance was observed. Nine datasets about LUAD were downloaded from Oncomine online database. In all the studies, DSG2 has a significant higher expression in cancer tissues than lung normal tissues (Figs. 1D–1L).

In addition, tissue microarray was used to verify the expression level of DSG2. The system scoring was shown in Figs. 2A–2I. Immunohistochemistry staining analysis showed higher expression of DSG2 in LUAD tissues compared with the adjacent non-tumor tissues (Figs. 2J–2K, p < 0.001). Higher DSG2 expression was positively associated lymph node metastasis (p = 0.040). However, there was no correlation between DSG2 expression and age, tumor size or TNM stage (Table 3, p >  0.05). Higher DSG2 expression was also observed in the male group than the female group (p < 0.001). DSG2 was located in the cytoplasm or membranous. Moreover, immunostaining evidence from the Human Protein Atlas database supported the upregulation of DSG2 in lung adenocarcinoma tissues. Two different antibodies (HPA004896 and CAB025122) were utilized to detect the expression of DSG2 in 9 lung adenocarcinoma tissues and 6 normal lung tissues. By antibody CAB025122, 3 were high staining and 2 were medium. By using antibody HPA004896, of total 4 lung adenocarcinoma tissues, 3 were high staining and 1 was medium. However, staining in normal lung tissues were both low no matter which antibody was used (Figs. 2L–2O).

Integrated meta-analysis

The results of the integrated meta-analysis contained TCGA data, Oncomine datasets and qRT-PCR data. As shown in Fig. 3A, the pooled effect sizes from forest plots indicated that DSG2 was significantly highly expressed in lung adenocarcinoma than in non-cancer tissues (SMD=1.30, 95% CI [1.08–1.52], p = 0.014) using the random effect model (I²=54.8%). Area under curve (AUC) for the diagnostic value of DSG2 was 0.79 (95% CI [0.75–0.82]) (Fig. 3B). The pooled sensitivity and specificity were 0.73 (95% CI [0.69–0.76]) (Fig. 3C) and 0.96 (95% CI [0.89–0.98]) (Fig. 3D), respectively. Apart from that, Begg’s test revealed that no publication bias existed among all the studies analyzed (p = 0.615).

Survival analysis

To evaluate the relationship between DSG2 and prognosis, Kaplan–Meier survival analysis was performed. The exclusion criteria were overall survival of <30 days. The expression of DSG2 was cut into high expression level and low expression level according to the median value of DSG2. As the result shown in Fig. 4, high mRNA expression of DSG2 was associated with shorter overall survival (hazard ratio [HR]: 1.638; 95% CI [1.214–2.209], p = 0.001) and poor progression-free survival (HR: 1.475; 95% CI [1.102–1.974], p < 0.001). Furthermore, multivariate Cox regression analysis was performed to explore factors associated with survival. A trend was observed in which low DSG2 expression was associated with an increased survival time. However, no statistical significance was determined (Table 4).

Co-expressed genes and potential pathways associated with DSG2

Based on TCGA, 215 co-expressed genes with DSG2 were identified (Fig. 5). Then GO enrichment analysis and KEGG pathway analysis were processed. As the result shown in Fig. 6 and Table 5, those co-expressed genes were mainly enriched in ‘cell division’, ‘mitotic nuclear division’, ‘chromosome segregation’ in Biological Process; ‘cytosol’, ‘midbody’ and ‘membrane’ in Cellular Component; ‘ATP binding’, ‘protein binding’, ‘microtubule motor activity’ in Molecular Function. KEGG pathway analysis revealed a significant enrichment of the co-expressed genes in the pathway ‘cell cycle’. The ten genes most strongly correlated with DSG2 are DSC2, SLC2A1, BZW1, ARNTL2, ERO1L, CLDN12, ECT2, ANLN, LAMC2 and GPR115. The expression and the correlation with OS of these ten genes were validated from GEPIA database. The results showed that seven of ten co-expressed genes were both significantly higher in LUAD compared to normal lung tissues and with worse OS (DSC2, SLC2A1, ARNTL2, ERO1L, ECT2, ANLN and LAMC2) (Fig. 7).

In order to determine whether DSG2 plus the co-expressed genes could generate a synergistic effect for OS, all gene data were downloaded from TCGA. The expression of each gene was cut into high expression level and low expression level. Cox analyses were performed to calculate the HR. Unfortunately, HR of DSG2 +DSC2 was lower than DSG2 or DSC2 alone (1.349 vs 1.638 vs 1.645). There was no synergistic effect when DSG2 combined with the co-expressed genes. This phenomenon was also found in other six genes (SLC2A, ARTNL2, ERO1L, ECT2, ANLN and LAMC2) (Table 6).