The discovery of gene expression signature in KRAS-oncogene-driven lung cancer
To uncover specific gene expression signature of KRAS-oncogene-driven lung cancer, we analyzed transcriptional expression profiles of normal lung tissues and KRAS-mutant lung tumor tissues based on GEO datasets (GSE18784, GSE49200), respectively, and identified differentially expressed genes (DEGs) with statistical difference (P < 0.05) between normal and tumor tissues. As shown in Fig. 1a-b, 25 upregulated DEGs and 45 downregulated DEGs were screened out based on GSE18784 dataset, using “EdgeR” R package. Using the same method, 155 upregulated DEGs and 120 downregulated DEGs were screened out based on GSE49200 dataset. The signatures between the two DEGs sets were different, indicating the heterogeneity of KRAS-driven tumors. Interestingly, SDPR was the only DEG that decreased in KRAS-mutant tumor tissues based on GEO datasets (GSE18784, GSE49200), which suggested that the downregulation of SDPR might be a specific signature during the development of KRAS-mutant lung cancer.
Structure and phylogenetic conservative analysis of SDPR
SDPR, also named CAVIN2, is a member of CAVIN family, which is located at chromosome 2, q32.3 (Fig. 1c). The structures of SDPR gene include 5’UTR exon, two exons, 3’UTR exon, and one intron. Protein sequences were compared to explore conservation of SDPR during molecule and species evolution, and the alignment results showed that Homo sapiens SDPR shared 82.82%, 81.88%, 99.53%, 96%, 87.29% and 89.18% identity with Mus musculus, Rattus norvegicus, Pan troglodytes, Macaca mulatta, Sus scrofa, and Felis catus, respectively, which indicates that SDPR is highly conserved in mammals (Fig. 1d).
CAV and CAVIN family members play important roles in the formation and stability of pulmonary alveoli (35). Moreover, CAVIN members could regulate the expression of CAV members. Thus, we analyzed the phylogenetic conservation of CAV and CAVIN family members. As shown in Fig. 1e, CAV and CAVIN family members are divided into two major clusters, and CAVIN2 shares a closer evolutionary relationship with CAVIN3, compared with CAVIN1 and CAVIN4.
SDPR is downregulated in human lung adenocarcinoma, including KRAS-mutant group
To identify the SDPR expression level in mouse and human lung tissues and tumors, we established KRAS-oncogene-driven lung cancer models (32) and detected SDPR expression using RT-qPCR. As shown in Fig. 2a, higher SDPR expression was detected in pulmonary than in bronchial tissue. Moreover, lower SDPR expression was observed in KRAS-mutant tumor tissues (P < 0.05). We further confirmed SDPR expression in human tissues and found a similar result in KRAS-mutant tumors. As shown in Fig. 2b-e, SDPR expression significantly decreased in KRAS-mutant specimens as well as all lung tumors compared with normal tissue (P < 0.05). In addition, low SDPR expression was detected in KRAS-mutant and KRAS-wild type NSCLC cell lines compared with immortalized normal lung cells, MRC-5 (Fig. 2f-g).
Low expression of SDPR is associated with a poor prognosis in NSCLC patients
As shown in Fig. 3a-c, low expression of SDPR was associated with shorter OS in NSCLC patients as well as in KRAS-mutant group, based on GEO dataset and lung cancer microarray (GSE72094, HLugA180Su06, P < 0.05). Similar results were found in NSCLC patients using GEPIA (Fig. 3d, P < 0.05). Meanwhile, univariate survival analysis indicated that low SDPR expression was associated with the shorter OS in NSCLC patients as well as in KRAS-mutant group (KRAS-mutant lung adenocarcinoma, P < 0.05, hazard ratio [HR] = 0.7; lung adenocarcinoma, P < 0.05, hazard ratio [HR] = 0.7; Table 1). Moreover, multivariate survival analysis showed that SDPR expression and stage were independent predictors of prognosis in lung adenocarcinoma patients as well as in KRAS-mutant group (Table 1). These data highlight the prognostic value of SDPR in human lung adenocarcinoma, especially in KRAS-mutant subgroup.
Construction of competing endogenous RNA (ceRNA) network of SDPR in KRAS-mutant lung adenocarcinoma pathway
To identify the upstream regulatory structure of SDPR in KRAS-mutant lung cancer, DEGs based on GSE72094 and three public predicted websites (TargetScan, miRDB and miranda) were used (Fig. 4a). Briefly, 139 expression profiles of KRAS-mutant patients with complete clinical information were collected (GSE72074), and DEGs sets between low and high SDPR group were screened out using “EdgeR” R package. Three public websites, TargetScan, miRDB and miranda, were used to predict potential combinations between SDPR and transcription factors. As shown in Fig. 4a, two transcription factors (DACH1, WT-1) were identified based on DEGs and TargetScan websites. Moreover, SDPR correlated positively with DACH1 (R2=0.509, P < 0.01; Fig. 4b) and negatively with WT-1 (R2=-0.218, P < 0.05; Fig. 4c). We detected the expression of DACH1 in NSCLC cell lines, MRC5 cells and the KRAS-oncogene-driven lung cancer mice. The DACH1 expression in bronchial tissue was lower than that in normal lung tissue based on KRAS oncogenic mice models. Meanwhile, DACH1 expression was lower in tumor tissue than in normal lung tissue. Moreover, The DACH1 expression in NSCLC cells was lower than that in MRC5 cells (Supplementary Figure S1).
Similar to the above screening method of transcription factors, a set of miRNAs was predicted, and five miRNAs (hsa-miR-1, hsa-miR-204, hsa-miR-144, hsa-miR-105 and hsa-miR-363) were ultimately screened out, which were observed in the above 4 miRNA sets (Fig. 4d). All of them were downregulated in KRAS-mutant lung adenocarcinoma compared with normal lung tissues (Fig. 4e). Interestingly, we found some potential complementary sequences between hsa-miR-1 and DACH-1 (Fig. 4f), indicating that the above miRNAs and TFs may form a complex network to regulate SDPR expression. Thus, we screened a series of miRNAs with potential combination sequence with SDPR-related TFs, and constructed a competing endogenous RNA (ceRNA) network of SDPR in KRAS-mutant lung adenocarcinoma (Fig. 4g).
Biological enrichment analysis of SDPR downstream pathway
To explore the downstream pathway of SDPR, DEGs based on GSE72094 were explored to identify biological differences between tissues with low and high SDPR expression in KRAS-mutant lung cancer. Gene ontology analysis was performed using DAVID online software to unfold the biological function of biological process, cellular component and molecule function among the above DEGs. As shown in Fig. 5a-c, biological processes were mainly associated with cell mitosis and cell cycle, and the differences of cellular components were mainly located in the extracellular space, exosomes, and matrix. In addition, there were a series of members related to redox balance and energy transfer, indicating the close interaction between SDPR expression and metabolism. Moreover, GSEA analysis results showed that G2 pathway and TGF-beta pathway were most likely associated with the above DEGs (Fig. 5d-e).
Correlation between SDPR, immune negative regulatory molecules and immune infiltration models
Recently, SDPR was reported to play an important role in cancer progression and metastasis via epithelial mesenchymal transition (EMT) in gastric and breast cancers (27, 36). However, the function of SDPR in lung cancer, especially in KRAS-mutant group, remains unclear. Since different SDPR expression levels are accompanied with changes in extracellular components (Fig. 5c), we hypothesized that SDPR expression may be closely related with tumor environment. Thus, we explored the correlation between SDPR, immune checkpoint molecules and immune infiltration models.
As shown in Fig. 6a, SDPR expression level correlated negatively with PD-L1(CD274), GITR(TNFRSF18), 4-1BBR(TNFRSF9) and TDO2 (R2 =-0.247, -0.327, -0.183, -0.233, respectively; P < 0.05). Since the role of SDPR in immune infiltration is unclear, we analyzed the abundance of immune cells in lung cancers at different SDPR expression levels and copy number variation (CNVs) patterns. In KRAS-mutant subgroups, cancer tissue with lower expression of SDPR was accompanied with less infiltration of γ T cells and resting mast cells but higher abundance of plasma cells, CD4+ memory activated T cells and M1 macrophages (Fig. 6b). Meanwhile, SDPR expression in lung adenocarcinoma correlated positively with infiltration of memory B cells, endothelial cells, M1and M2 macrophages, myeloid dendritic cells, neutrophils, memory resting CD4+ T cells, CD8+ T cells, but correlated negatively with M0 macrophages, plasma B cells, and CD4+ memory activated T cells based on TIMER 2.0 website (Table 2). In addition, lung adenocarcinoma with SDPR arm-level deletion showed less infiltration of CD4+ T cells, macrophages and neutrophils in TME (Fig. 6c).
These results illustrated close relationship between SDPR, PD-L1(CD274), GITR(TNFRSF18), 4-1BBR(TNFRSF9), TDO2, and abundance of immune cells in human lung adenocarcinoma, especially in KRAS-mutant subgroups.