Prognostic values, ceRNA network, and immune regulation function of SDPR in KRAS-mutant lung cancer

Serum Deprivation Protein Response (SDPR) plays an important role in formation of pulmonary alveoli. However, the functions and values of SDPR in lung cancer remain unknown. We explored prognostic value, expression pattern, and biological function of SDPR in non-small cell lung cancer (NSCLC) and KRAS-mutant lung cancers. SDPR expression was evaluated by quantitative real-time PCR (RT-qPCR), immunohistochemistry (IHC), and Western blot on human NSCLC cells, lung adenocarcinoma tissue array, KRAS-mutant transgenic mice, TCGA and GEO datasets. Prognostic values of SDPR were evaluated by Kaplan–Meier and Cox regression analysis. Bioinformatics implications of SDPR including SDPR-combined transcription factors (TFs) and microRNAs were predicted. In addition, correlations between SDPR, immune checkpoint molecules, and tumor infiltration models were illustrated. SDPR expression was downregulated in tumor cells and tissues. Low SDPR expression was an independent factor that correlated with shorter overall survival of patients both in lung cancer and KRAS-mutant subgroups. Meanwhile, ceRNA network was constructed to clarify the regulatory and biological functions of SDPR. Negative correlations were found between SDPR and immune checkpoint molecules (PD-L1, TNFRSF18, TNFRSF9, and TDO2). Moreover, diversity immune infiltration models were observed in NSCLC with different SDPR expression and copy number variation (CNV) patterns. This study elucidated regulation network of SDPR in KRAS-mutant NSCLC, and it illustrated correlations between low SDPR expression and suppressed immune system, unfolding a prognostic factor and potential target for the treatment of lung cancer, especially for KRAS-mutant NSCLC.

Recently, a series of compounds targeting KRAS-G12C variant have been developed and achieved promising effects in preclinical experiments and phase I clinical trials [18][19][20]. However, it is not clear whether KRAS mutation, especially G12V and G12D variant, can have any clinical benefits. Meanwhile, patients with co-occurring TP53/KRAS mutations showed remarkable clinical response to immune checkpoint inhibitors (CPI) [21]. Moreover, patients with KRAS mutation had favorable clinical benefits of anti-PD-1/PD-L1 immunotherapy [22], and high PD-L1 expression in tumor cells was associated with improved overall survival in KRAS mutant patients [23]. However, the loss of STK11/LKB1 promoted programmed PD-1/PD-L1 inhibitor resistance [24]. These studies indicated that the presence of co-occurring genetic events and the mutant KRAS allelic content increase biological heterogeneities of KRASmutant NSCLC, which complicates the treatment of KRAS-mutant lung cancers.
To investigate the gene expression signature in KRASoncogene-driven lung cancer, we compared the differences between KRAS-mutant tumors and normal lung tissue derived from a genetically engineered mouse model (GEMM), based on expression profiling and comprehensive bioinformatics analysis. Several differentially expressed genes (DEG) were screened according to the gene expression profile, but SDPR was the only DEG that decreased in both GEMM tumors. SDPR (also known as CAVIN2, NC_000002.12, gene ID 8436), a key substrate for protein kinase C, was found to play a critical role in inducing membrane curvature and participating in the formation of caveolae [25]. It has been reported that SDPR is a potential diagnostic indicator in cancers such as hepatocellular carcinoma and gastric cancer [26][27][28]. However, it remains unknown whether SDPR could be a predictor or target for lung cancer, especially in KRAS-mutant group. Moreover, SDPR is considered a suppressor gene in papillary thyroid cancer [29], but the regulatory mechanism of SDPR remains to be illustrated. Meanwhile, the connection between SDPR and tumor microenvironment (TME) has rarely been explored. Our study explored the gene signature, regulation, and effect of SDPR on tumor and immune infiltration, based on comprehensive bioinformatics analysis, evaluation of lung cancer specimens, and preclinical experiments.

Cell lines and reagents
Human non-small cell lung cancer cells (HCC4006, H23, H358, SK-LU-1 and H1299) were purchased from American Type Culture Collection (ATCC), Virginia., America. Human embryonal lung cell (MRC-5) was purchased from the Type Culture Collection of the Chinese Academy of Sciences, Shanghai, China [30]. HCC4006, H23, H358 and H1299 cells were maintained in RPMI 1640 supplemented medium, MRC-5 cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM), and SK-LU-1 cell lines was maintained in Minimum Essential Medium (MEM), respectively. All cells were cultured in standard environment as descried previously [31].

Transgenic mouse and establishment of KRAS-mutant lung cancer models
The LSL-KRAS mice (B6.129S4-KRAStm4Tyj/JNju) were purchased from Nanjing biomedical institution of Nanjing University, Nanjing, China, and housed in specific pathogen-free (SPF) institution of Experimental animal center in Southern Medical University, Guangzhou, China. Cre recombinase induced Adeno-associated viruses (AAV-CMV-bGloin-Cre) were purchase from Shanghai genechem Co., Ltd., China. AAV-CMV-bGloin-Cre virus was in tracheally instilled into LSL-KRAS mice to induce KRAS-oncogene expression [32]. After further 4-6 months, visible tumor nodules were observed in lung tissue. Finally, tumor-bearing mice were sacrificed, and tumor tissue and normal lung tissue were collected.

Clinical Specimens and Immunohistochemistry (IHC) staining
Tissue microarray with clinical pathological data of lung cancer (HLugA180Su06) was purchased from Shanghai Outdo Biotech Biotechnology Co., Ltd., China. This lung cancer microarray (HLugA180Su06) contains 94 tumors and 86 paired adjacent normal tissues. All the tissues were collected from lung adenocarcinoma patients who underwent surgical resection from July 2004 to June 2009. The follow-up was from August 2014 and ranged from 5 to 10 years. IHC staining were performed as described in Additional file 1: S1 [31]. SDPR expression were detected with a rabbit polyclonal antibody against SDPR (Proteintech, #12339-1-AP; RRID: 2183305).

Screening of differentially expressed genes (DEGs) and identification of the abundance of tumor immune infiltration
In this study, GSE18784, GSE49200, GSE72094 and GSE48414 were downloaded from GEO dataset, and datasets (PanCancer Atlas) contained lung adenocarcinoma expression profiles and paired normal tissues were downloaded from TCGA database through cBioPortal and sangerbox download tools. "EdgeR" R package in R version 3.6.2 (The R Foundation for Statistical Computing, Vienna, Austria; https ://www.r-proje ct.org/) was used to screen out the murine DEGs between normal and tumor tissues. "CIBERSORT.R" R package was used to explore the abundance of tumor immune infiltrations in KRAS-mutant lung adenocarcinomas, and TIMER 1.0 and 2.0 (Tumor Immune Estimation Resource, https :// cistr ome.shiny apps.io/timer /) were used to identify the abundance of immune cells, such as B cells, CD4 + T cells, CD8 + T cells, Neutrophils, Macrophages and Dendritic cells at different SDPR copy number variation (CNV) patterns. The description of the above datasets and analysis processing method were described in Additional file 1: S2.

Phylogenetic analysis of SDPR
SDPR (NC_000002.12, gene ID 8436), also known as CAVIN2 is located in Chromosome 2. Homo sapiens amino acid sequences of CAV and CAVIN family members were downloaded from Uniprot database. Subsequently MEGA-X (https ://www.megas oftwa re.net/) was used to conduct sequence alignment and infer phylogenetic trees (Additional file 1: S3). The phylogeny was inferred using the Neighbor-Joining method, and the tree is created and conducted using Interactive Tree Of Life (iTOL, https ://itol.embl.de/).

Bioinformatics mining of SDPR
The information of chromosome location site and gene structure of SDPR gene were analyzed through Gen-eCards (https ://www.genec ards.org/). Protein sequences among CAVIN and CAV family members were downloaded from Uniprot database (https ://www.unipr ot.org/). The sequence alignment was performed to analyze the identity between Homo sapiens, Mus musculus, Rattus norvegicus, Pan troglodytes, Macaca mulatta, Sus scrofa and Felis catus.

Statistical analysis
All the data were analyzed by SPSS, version 20, IBM Corp., Armonk, USA. SDPR or DACH1 expression was presented as the mean ± standard deviation (SD). Bars indicate SD. * P < 0.05; ** P < 0.01. Differences between the means were examined by student's t test or one-way analysis of variance (ANOVA). Multiple comparisons among the groups were performed using LSD method. Nonparametric test was used to analyze the SDPR scores in lung and tumor tissues, and correlation analysis was assessed by Pearson or Spearman correlation method. Kaplan-Meier method and Cox proportional hazard regression model were used to evaluate the prognostic value of SDPR in lung adenocarcinoma. Differences with a value of P < 0.05 were considered statistically significant. All of the experiments were performed at least thrice.

The discovery of gene expression signature in KRAS-oncogene-driven lung cancer
To uncover specific gene expression signature of KRASoncogene-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.
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 KRASoncogene-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 (GSE72094), 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 (R 2 = 0.509, P < 0.01; Fig. 4b) and negatively with WT-1 (R 2 = − 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 (Additional file 1: 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.

Discussion
Over 8 different variants of KRAS mutation have been identified at codons 12, 13 and 61 in NSCLC [37]. Several studies explored the therapeutic vulnerability and prognostic differences between the KRAS mutation subtypes [38][39][40]. However, KRAS mutant status may not be recommended to select NSCLC patients for specific treatment such as adjuvant chemotherapy. Meanwhile, there were no significant differences in the phosphorylation level of MEK/ERK kinase among the above variants, despite phosphorylation of AKT and activation of RAL seem to differ between KRAS-G12C and KRAS-G12V cells [4,38]. In summary, no specific KRAS variants were validated as ideal prognostic factors of survival or vulnerability indicators for treatment of KRAS-mutant tumors. In our study, we found that SDPR expression was not only decreased in KRAS-mutant NSCLC cells, and KRAS-driven murine tumor from GEMMs, but also downregulated in human NSCLC specimens based on GEO datasets, TCGA datasets, and lung adenocarcinoma tissue array (Fig. 2a-g). Moreover, SDPR expression was suggested to be an independent prognostic factor in lung cancer (Fig. 3a-d, Table 1). Our research provides a potential target for the prognosis and treatment of NSCLC independent of KRAS variants. More biological experiments and clinical trials are needed to validate and complement our conclusions.
Co-occurring genetic events were frequently observed in KRAS-mutant lung tumors, unlike other oncogenedriven lung cancers [40,41]. STK11 co-mutations (KL), TP53 co-mutations (KP), and CDKN2A/B inactivation plus low thyroid transcription factor-1 (TTF-1) expression (KC) were considered as classical models among KRAS-mutant tumors, and may induce different biological behaviors and characteristics of tumors [40,42]. Cooccurring STK11 or KEAP1 mutations were associated with worse OS in KRAS-mutant NSCLCs [9,40]. Moreover, the lowest levels of PD-L1 and deficient inflammatory immune cells infiltration were found in the KL group. In contrast, the KP group with the highest PD-L1 expression was infiltrated with active inflammation (mainly  The correlations between SDPR and two predicted TFs, DACH1 and WT-1 based on TCGA dataset (Atlas). c, d Prediction of upstream miRNAs of SDPR in KRAS-mutant lung adenocarcinoma. KRAS-mutant specific miRNAs were screened out based on TCGA dataset (GSE48414), then intersected with SDPR combined miRNAs predicted by miRDB, miranda and targetScan. Five miRNAs finally were screened out and the expression levels in normal tissue and KRAS-mutant lung carcinoma were shown with heat map. f Complementary sequences between hsa-miR-1, DACH1 and SDPR. g Construction of ceRNA network to visualize the regulation models of SDPR, TFs and miRNA in KRAS-mutant lung carcinoma T cell inflammation) [24]. Our research found negative correlations between SDPR, PD-L1, and immune cells in KRAS-mutant lung cancers (Fig. 6a-c). More studies should explore the influence of co-occurring genetic events on SDPR expression and malignant biological behaviors of tumors.
Previous research explored the prognostic and diagnostic significance of SDPR in gastric cancer [27], hepatocellular carcinoma [28], and papillary thyroid cancer (PTC) [29]. Our research originally found downregulation of SDPR in lung cancers as well as in KRAS-mutant subgroup ( Fig. 2a-g), and innovatively explored the immune checkpoint molecules and abundance of immune infiltrations at different SDPR expression and CNVs models (Fig. 6a-c, Table 2). Those results provide a novel theory for the immune regulatory functions of SDPR in tumorigenesis, progression and metastasis.
In terms of the regulation and function of SDPR in lung cancer, the reason leads to the depression of SDPR is unclear. It was reported that MiR-577 regulates TGF-β in gastric cancer through a SDPR-modulated positive-feedback loop [27]. Moreover, overexpression of SDPR inhibited the activity of ERK and NF-κB pathways in breast cancer [36]. In our study, a series of pathways, including the TGF-β pathway, were enriched between SDPR-low and SDPR-high specimens in KRAS mutant lung cancers (Fig. 5d). In addition, results of GO analysis indicated different distribution of extracellular components depending on SDPR expression (Fig. 5c). Our study screened out a series of TFs and miRNAs as promising candidates for the upstream targets of SDPR in KRAS-mutant cancers, and we constructed a ceRNA Fig. 6 Correlation between SDPR, immune checkpoint molecules and abundance of tumor immune infiltrations. a Correlations between SDPR expression and immune checkpoint molecules based on GEO dataset (GSE72094). b Abundance of immune infiltrates in KRAS-mutant lung cancer with different somatic CNV patterns of SDPR based on GEO dataset (GSE72094). Low SDPR expression group marked in blue, and high SDPR expression group marked in red. c Correlations between SDPR mRNA expression and abundance of immune infiltrates in lung adenocarcinoma using TIMER (comprehensive resource for the clinical relevance of tumor-immune infiltrations) system. d Abundance of immune infiltrates in lung adenocarcinoma with different somatic CNV patterns of SDPR using TIMER system regulation network of SDPR in KRAS mutant lung cancers, which provided useful information for the molecule regulatory network of SDPR in KRAS-mutant lung cancers.

Conclusions
In our study, a decrease of SDPR was found in lung cancers as well as in KRAS-mutant subgroup, and which may be a promising prognostic marker for the survival of patients with lung cancer. Moreover, systematic exploration of SDPR in gene location, species conservation, function, and potential regulatory network was illustrated in lung cancer, especially in KRAS-mutant tumors. In addition, our research originally unfolded the correlation between SDPR, immune checkpoint molecules, and abundance of immune infiltrations. In summary, SDPR could be a promising prognostic factor and potential target for the treatment of lung cancer, especially for KRAS-mutant adenocarcinomas.

Supplementary Information
The online version contains supplementary material available at https ://doi. Significant correlation between immune cell subgroups and SDPR expression were shown in Table 2 based on TIMER, CIBERSORT, quanTIseq, xCell, MCPcounter and EPIC algorithms. Positive correlation was marked in underline, while negative correlation was marked in italics. P< 0.01 were marked as 0.00