Based on necroptosis identifying the immunological features and prognostic signatures of lung adenocarcinoma

Peng Xia ,De-Gui Wang* ,Si-Wei Ouyang ,Rong Shen ,Zhao Guo ,Xu-Guang Yang ,Xiang-Wen Liu,Kun Xie

1Department of Anatomy and Histology,School of Basic Medical Sciences,Lanzhou University,Lanzhou 730030,China.

Abstract Background:Lung adenocarcinoma is one of the most common pathological types of lung malignant tumor with high morbidity and mortality.Long non-coding RNAs are gradually recognized to play crucial roles in tumor occurrence and development.Necroptosis is a newly established way for cell programmed death,undertaking essential roles in anti-tumor.Therefore,identifying necroptosis-related long non-coding RNAs and based on them to evaluate the signatures of lung adenocarcinoma is essential for patients’ survival prediction and therapy.Methods: We collected data from the public database and performed the least absolute shrinkage to construct a 13-lncRNAs prognostic model.Based on the Consensus Clustering,ESTIMATE,CIRERSORT,and weighted gene co-expression network analysis to identify the immune signatures.Results: This study identified a 13-lncRNAs prognostic model.The model’s prediction accuracy was evaluated by receiver operating characteristic and independent-prognosis analysis;besides,a Gene Expression Omnibus dataset was applied for external validation.Furthermore,we analyzed the immune features of subgroups in multiple dimensions.A consensus clustering analysis based on the 41 genes was implemented to separate lung adenocarcinoma patients into two subgroups.We compared the features of subgroups in multiple dimensions,including survival,immune microenvironment,immune cells infiltration and gene co-expression network analysis.Conclusion: We established a prognosis necroptosis-related risk model to predict lung adenocarcinoma patients’ prognosis and systematically understood the correlation between immune and necroptosis.This study can applicate in clinical to predict the prognosis of lung adenocarcinoma patients and provide new insight into lung adenocarcinoma immune therapy.

Keywords:lung adenocarcinoma;long non-coding RNAs;necroptosis;the least absolute shrinkage and selection operator;prognostic model

Background

Lung cancer is one of the most common types of cancer with high morbidity and mortality [1,2].The subtypes of lung cancer can be roughly divided into two categories,non-small-cell lung cancer(NSCLC) and small cell lung cancer.NSCLC accounted for 85% of these two subtypes and can be further divided into squamous cell carcinoma,adenocarcinoma and large-cell carcinoma,while the lung adenocarcinoma (LUAD) is the most common one.However,the 5-year overall survival (OS) rate for patients with LUAD is only 15%[3].Therefore,it is crucial to identify novel efficient tumor biomarkers for LUAD precision diagnosis,targeted therapy and prognosis prediction.

Long non-coding RNAs (lncRNAs) belong to non-protein-coding RNAs with a length of more than 200 nucleotides.Although lncRNAs have limited ability for protein-coding,they still undertake multiple cellular and physiologic functions,such as transcription regulation and mRNA post-transcriptional regulation [4].In diverse tumors,including lung cancer,lncRNAs play essential roles in tumor initiation,invasion,and metastasis [4–6].Furthermore,many studies have demonstrated that lncRNAs can influence the tumor microenvironment (TME),affecting tumor infiltration and progression[7,8].Therefore,elucidating the specific functions of lncRNAs and recognizing their features are significant for LUAD individualized treatment and improving patient prognosis.

Necroptosis is a newly identified programmed death way for cells,independent of caspase,mediated by receptor interacting protein kinase1/3 (RIPK1/RIPK3) and executed by mixed lineage kinase domain like pseudokinase (MLKL) [9–11].Necroptosis plays a complicated role in regulating the TME and inflammatory,which profoundly affect tumorigenesis and metastasis [12,13].Recently,its impact on tumors was reported to correlate with anti-tumor immune,but its specific mechanism is not very clear until now.Therefore,exploring the interplay between necroptosis-related lncRNAs and anti-tumor immune is significant for LUAD therapy.

Up to now,the treatment and prognosis evaluation for LUAD mainly rely on tumor,node and metastasis staging criteria.However,the tumor,node and metastasis staging system ignoring the internal molecules’ interplay may sometimes cause poor performance in predicting the OS of patients.Therefore,we constructed a 13-necroptosis-related lncRNA signature to predict LUAD patients’survival in this study.To further explore the immune signature of LUAD,we also performed a consensus clustering analysis based on screening for necroptosis and prognosis and calculated the immune score and other features.Based on these works,we can have a robust basis for LUAD prognosis prediction and immune therapy.

Materials and methods

Data collection

We downloaded the clinical information and RNA-sequencing transcriptional profile of LUAD (T-494,N-59) from the cancer genome atlas (TCGA) database (https://portal.gdc.cancer.gov).And we transferred the fragments per kilobase million data into transcripts per million for subsequent analysis.For accuracy,we applied Perl language (version Strawberry-Perl-5.32.1.1,https://strawberryperl.com/) for data pre-processing and only retained samples with integral clinical information.A total of 494 tumor samples were retrieved and randomly divided into the train set and test set by caret R package,at a ratio of 7:3.Furthermore,67 necroptosis-related genes were retrieved from a published study [14].For external validation,we obtained a dataset from the Gene Expression Omnibus(GEO) database as a validation set,GSE30219,a large size independent dataset with 289 integral clinical information.

Identifying prognostic associated necroptosis-related genes

We extracted necroptosis-related genes expression information from the mRNAs expression matrix of LUAD.And then,we combined it with corresponding clinical data.A univariate Cox regression analysis was applied for these genes.Eventually,we obtained 19 necroptosis-related genes associated with the prognosis of LUAD patients.

Screening of necroptosis-related lncRNAs and establishing a co-expression network of lncRNAs-mRNAs

To explore the mutual relations between lncRNAs and necroptosis,we performed a correlation analysis for the 19 previously screened necroptosis mRNAs with lncRNAs.And we use the ∣Pearson correlation coefficients∣>0.4 andP<0.001 as the filter threshold.Furthermore,to visualize the correlation network,we utilized the“igraph”package of R to manifest their relationship.

Identification of prognosis related lncRNAs

Based on the previously screened necroptosis-related lncRNAs,we further conducted a univariate Cox regression analysis for these lncRNAs.And we utilizedP<0.05 as the filter threshold to screen prognosis-related lncRNAs.Subsequently,a forest plot was used to manifest the univariate Cox regression analysis results.And the expression in tumor and normal tissue of these identified lncRNAs was shown in a heat map.

Consensus clustering for the prognosis necroptosis-related lncRNAs

Based on the previously screened lncRNAs,which are related to both necroptosis and prognosis,we used the“ConsensusClusterPlus”package and“limma”package inRto perform a consensus clustering analysis.We divided all of these LUAD samples into two subgroups.Moreover,we compared the survival difference in these subgroups.And we display the relationship between subgroups and clinical information in a heat map.

Implementing immune signature analysis,weighted gene co-expression network analysis (WGCNA) and gene ontology(GO)analysis for consensus clustering established subgroups

We extracted the expression matrix of immune checkpoints,including PDCD1(PD-1),CD274(PD-L1),PDCD1LG2(PD-L2),CTLA4,and LAG3 from the mRNA matrix of LUAD.And we performed a correlation analysis between the previously identified prognosis necroptosis-related lncRNAs and these immune checkpoints.And we compared the immune checkpoints genes expression in different subgroups.In addition,the TME was evaluated by the ESTIMATE algorithm.And then,the CIRERSORT algorithm was applied to calculate the fraction of 22 types of immune cells and Wilcoxon rank-sum test was used to test the differences in different classes.

To identify hub genes in different clusters,WGCNA for the LUAD transcripts per million mRNA expression matrix to identify the encoding-gene expression signature for subgroups.Firstly,we constructed an adjacency matrix to describe the correlation of modules.Then we determined the soft-threshold β=9 (scale-freeR2=0.9) and transformed the adjacency matrix into a Topological Overlap Matrix,which can quantify the similarity of modules based on weighted correlation.Next,hierarchical clustering was performed to form different modules,and each module contains more than 30 genes.Finally,the eigengene of modules was calculated and similar modules were merged (cutoff=0.25).Based on the WGCNA algorithm,the different genes expression patterns were clustered into multiple modules,and then a correlation analysis was applied to explore the relationship between modules and clusters.Eventually,we screened the top-30 intramodule connectivity in the most relevant model for subgroups.

To recognize functional features for these top-30 genes,we performed GO analysis,which contains cellular component,molecular function and biological process 3 parts.We operated GO enrichment analysis based on theseRpackages“clusterProfiler”and“org.Hs.eg.db”,“org.Hs.eg.db”for genes annotation,“clusterProfiler”for GO analysis.According to the GO analysis,a network diagram based on a“cnetplot”was performed to demonstrate the relationship between genes and pathways.And then,we selected genes that have high connectivity to pathways.To identify gene expression signatures and the association with the prognosis of LUAD patients,we did a Kaplan-Meier survival analysis in Gene Expression Profiling Interactive Analysis for the selected hub genes [15].Besides,we validated its expression status in LUAD tumor tissue by the human protein atlas database.

Development and validation of a prognostic model for LUAD based on prognosis necroptosis-related lncRNAs

We randomly divided all samples of TCGA lncRNAs matrix into train set and test set at a ratio of 7:3.Based on the prognostic necroptosis-related lncRNAs,we operated the least absolute shrinkage and selection operator (LASSO) regression algorithm to filter variables and conducted by 10-fold cross-validation with“glmnet”and“survival”Rpackages.Eventually,a total of 13 lncRNAs were screened by lamda.min,which were used to construct our model.We calculated the risk score of LUAD patients based on the expression levels and the lambda coefficient from LASSO regression.The following equation was adopted to calculate the risk score:

Risk score=(β1 × Expression lncRNA1)+(β2 × Expression lncRNA2) +·+(βn × Expression lncRNAn).β is the lambda coefficient from LASSO regression;n is the number of the lncRNAs we construct model.

And the median risk score was adopted as a cut-off to group all LUAD patients into the high-risk and low-risk groups.Subsequently,we performed survival analysis based on“survival”and“survminer”Rpackages.Moreover,we also plotted the receiver operating characteristic (ROC) curves and calculated the areas under the time-dependent ROC curves for both the train and test set.

External validation for prognostic model and independent prognostic analysis

First,we compared the clinical and cluster features for the model-defined high and low-risk groups of LUAD patients.Then,we retrieved a dataset from the GEO database,GSE30219,with 289 integral clinical information for external validation.We evaluated the established prognosis model by this dataset.And the survival difference of the external validation set was compared.Furthermore,independent prognostic analysis was conducted to examine whether the clinical features and our established model are independent factors in deciding LUAD patients’ prognosis.

Screening of differential expression genes in the model and identifying signatures based on the model risk score

We implemented a differentially expressed genes (lncRNAs) analysis between the TCGA LUAD tumor samples and normal samples based on the“limma”package inR.Then,we selected the 13 lncRNAs in our established model from the difference analysis results.To narrow the scope of differentially expressed genes,we further filtered rely on this criterion:∣logFC∣>1 andP-value <0.05.Eventually,only one gene was left.And we plotted a box diagram to compare its expression in tumor and normal tissue.

To explore different phenotypic characteristics for LUAD from the perspective of the model,we also compared the risk score differences in different clusters (cluster1 and cluster2 based on our previous Consensus Clustering algorithm) and in the high/low immune score group (based on our previous ESTIMATE algorithm and use the median immune score to divide high/low immune score group).

Results

Identification of prognosis necroptosis-related lncRNAs

The workflow of this research is shown in Figure 1.Univariate Cox regression analysis was applied for 56 necroptosis-related genes.Eventually,19 prognosis-related genes were obtained,including FADD,MLKL,TRIM11,IPMK,MYC,TNFRSF1A,TRAF2,PANX1,MAP3K7,DIABLO,ID1,HSPA4,FLT3,HAT1,PLK1,ALK,TERT,HSP90AA1,DDX58.And then,a correlation analysis was performed(∣correlation coefficients∣ >0.4 andP<0.001) to filter necroptosis-related lncRNAs.267 lncRNAs were identified and a co-expression network of necroptosis-related lnRNAs was shown in Figure 2A.A univariate Cox analysis was performed to confirm necroptosis and prognosis associated lncRNAs and 41 lncRNAs were identified in the end(P<0.05) (Figure 2B).The expression profiles of the 41 lncRNAs in tumor and normal samples were also illustrated in Figure 2C (＊P<0.05,＊＊P<0.01,＊＊＊P<0.001).

Figure 1 The workflow of this study.LUAD,lung adenocarcinoma;TCGA,the cancer genome atlas;lncRNAs,long non-coding RNAs;ROC,receiver operating characteristic;AUC,area under the curve;GEO,Gene Expression Omnibus;WGCNA,weighted gene co-expression network analysis;GO,gene ontology;HPA,the human protein atlas.

Figure 2 Identification of prognosis necroptosis-related lncRNAs. A,A co-expression network between the necroptosis-related mRNAs and lncRNAs;B,forest plot for the 41 prognosis necroptosis-related lncRNAs;C,a heatmap to manifest the expression characteristics of the 41 filtered lncRNAs.lncRNAs,long non-coding RNAs.lncRNAs,long non-coding RNAs.

Consensus clustering for the prognosis necroptosis-related lncRNAs

Based on the 41 prognosis necroptosis-related lncRNAs,we conducted a consensus clustering analysis for all LUAD samples to explore the relationship between the expression signatures of the 41 lncRNAs and LUAD subtypes.According to the empirical cumulative density function plot,when k=2,the consensus clustering matrix shows that the LUAD samples can be classified into two subtypes (Figure 3A).According to the consensus clustering analysis result,we next compared the clinical information based on clusters (cluster1 and cluster2).The gender,N staging and tumor stage were shown to have significant differences (Figure 3B).Furthermore,to further explore the differences between the two subgroups,Kaplan-Meier survival curves for cluster1 and cluster1 showed that the OS of the two subgroups are different.The OS of cluster 2 is better than cluster 1 (Figure 3C).

The immune checkpoint and TME signatures of the cluster1 and cluster2 Subgroups

To explore the TME features of the two subgroups of LUAD,we performed a correlation analysis for the immune checkpoints PDCD1(PD-1),CD274(PD-L1),PDCD1LG2(PD-L2),CTLA4 and LAG3 with the prognosis necroptosis-Related lncRNA (Figure 4A).we also compared the immune checkpoints expression level in different clusters (Figure 4B–Figure 4F).Eventually,we found that except for CTLA4,all the other four immune checkpoints have expression differences,and the expression levels for cluster1 are all higher than cluster2.In addition,to evaluate the degree of immune activation of different clusters,we further calculated the TME score by“ESTIMATE”arithmetic,which showed that the immune score of cluster1 is lower than cluster2 while the stromal score and estimate score have no significant difference in cluster1 and cluster2 (Figure 4G–Figure 4I).

Immune cells infiltration analysis,WGCNA and GO analysis for clusters

To assess the immune cell infiltration conditions,we applied“CIBERSORT”arithmetic to evaluate the proportions of immune cells in LUAD.The 22 immune cells infiltration in cluster1 and cluster2 were displayed in Figure 5A.Based on the analysis results and simultaneously combining the immune cells with a relatively high fraction level,we ultimately found that the Plasma cells,T cells CD8,Macrophages M0 and Macrophages M1 are higher in cluster1.On the contrary,the T cells CD4 memory resting,Macrophages M2,Dendritic cells resting,Dendritic cells activated and Mast cells resting are higher in cluster2.

To further explore the signature of gene expression patterns in different clusters,we utilized the WGCNA algorithm to recognize features for cluster1 and cluster2.According to the WGCNA,we established that the MEgrey module has the highest correlation(∣correlation coefficient∣=0.56,P<0.001) with the cluster(Figure 5B).Therefore,we screened the top-30 intramodule connectivity genes from the MEgrey module,including ANXA10,ASCL1,ALCA,CDHR2,CDHR5,CSAG1,CSAG3,DDX3Y,EIF1AY,EPS8L3,FER1L6,HNF4A,KDM5D,MAGEA12,MAGEA3,MAGEA6,MT-ATP6,MT-CO1,MT-CO2,MT-CO3,MT-CYB,MT-ND2,MT-ND4,MT-ND4L,MYO1A,REG4,RPS4Y1,USP9Y,UTY,ZFY.

Figure 3 Consensus clustering analysis based on the 41-prognosis necroptosis-related lncRNAs.A,Consensus clustering matrix when k=2;B,clinical information comparison heatmap of clusters;C,kaplan-meier survival analysis for two clusters.lncRNAs,long non-coding RNAs.

Figure 4 Analysis of immune checkpoints and TME. A,Correlation heatmap for immune checkpoints and prognosis necroptosis-related lncRNA;B–F,immune checkpoints expression in clusters,LAG3(B),PDCD1(C),PDCD1LG2(D),CD274(E),CTLA4(F);G–I,evaluation of TME by“ESTIMATE”,stromal score(G),immune score(H),estimate score(I).TME,tumor microenvironment;lncRNAs,long non-coding RNAs.

And then,GO enrichment analysis was operated to identify functional features of the top-30 genes.A network diagram was plotted to manifest the results of GO analysis,which showed that the CDHR2 has high connectivity with enriched pathways (Figure 5C).Next,we performed a Kaplan-Meier survival analysis in Gene Expression Profiling Interactive Analysis,which showed that the OS time for the CDHR2 high expression group was worse than the low expression group in LUAD (P=0.033) (Figure 5D).Further,we validated the CDHR2 expression status in the the human protein atlas database,showing that CDHR2 has high antibody staining in some immunohistochemical samples and a high CDHR2 expression level has a lousy prognosis,which indicates that CDHR2 is a hub gene for the prognosis of LUAD patients (Figure 5E).

Figure 5 The immune cells infiltration landscape and WGCNA of clusters. A,Analysis of immune cells infiltration by“CIBERSORT”for clusters;B,diagram for Module-cluster correlation;C,a network plot for GO enrichment analysis result of the selected top-30 intramodule connectivity genes;D,Kaplan-Meier survival analysis for CDHR2;E,immunohistochemical staining for CDHR2 in LUAD from the HPA database.WGCNA,weighted gene co-expression network analysis;GO,gene ontology;LUAD,lung adenocarcinoma;HPA,the human protein atlas.

Development and validation of a prognostic model of LUAD based on prognosis necroptosis-related lncRNAs

A total of 494 tumor samples of the TCGA lncRNAs matrix were randomly divided into train set (348) and test set (146) in a ratio of 7:3.According to the previously screened 41 prognosis necroptosis-related lncRNAs,we processed LASSO regression and 10-fold cross-validation to filter variables for the prognostic model(Figure 6A,Figure 6B).In the end,13-lncRNAs were established to construct the model,including RHOQ-AS1,LINC02693,LINC02273,LINC01800,LINC00996,FGD5-AS1,CADM3-AS1,TMPO-AS1,LINC01116,SFTA3,LINC02802,CRNDE and PARP11-AS1.The risk score can be calculated by the following formula:(–0.1694 ×RHOQ-AS1 Expression)+(0.1762 × LINC02693 Expression) +(–0.1310 × LINC02273 Expression)+(-0.6575 × LINC01800 Expression)+(–0.0423 × LINC00996 Expression)+(0.1671 ×FGD5-AS1 Expression)+(–0.0160 × CADM3-AS1 Expression) +(0.0808 × TMPO-AS1 Expression)+(0.0845 × LINC01116 Expression)+(–0.0156 × SFTA3 Expression+(0.1348 ×LINC02802 Expression) +(–0.050× CRNDE Expression)+(–0.0505× PARP11-AS1 Expression).In this model,five genes (LINC02693,FGD5-AS1,TMPO-AS1,LINC01116,LINC02802) have positive coefficient,indicating that these are harmful factors for LUAD patients.On the contrary,eight genes (RHOQ-AS1,LINC02273,LINC01800,LINC00996,CADM3-AS1,SFTA3,CRNDE,PARP11-AS1)have negative coefficient that undertake a protective effect on the prognosis of LUAD patients and their higher expression level often means a better outcome.Based on this risk prognosis model,we divided train set into high and low-risk group by the median risk score.Kaplan-Meier survival analysis was conducted for two groups,and log-rank tests showed that the low-risk group had a better prognosis (P<0.001,Figure 6C).Next,the ROC curve of 1,3 and 5 years were plotted to estimate the model’s prediction efficiency and the area under the curve (AUC) was calculated,which respectively are 0.75,0.7,0.67 (Figure 6D).

To examine the model’s prognosis prediction efficacy,we also calculated the risk score of the test set.We separated samples into high and low-risk groups in the median risk score cutoff.The same as before,we compared the survival differences of the two groups,plotted the ROC curves and calculated the AUC for 1,3 and 5 years.Similar to before,the two groups have significant differences in survival.The low-risk group has a better prognosis than the high(P=0.002,Figure 6E).And the AUC for the 1,3,5 years ROC curves are 0.66,0.6 and 0.78 (Figure 6F).These results indicating this model has a good prediction effect.

Figure 6 LASSO regression analysis and construction of the prognostic model. A,LASSO coefficient profiles for screened 41 lncRNAs;B,distribution diagram of the partial likelihood deviation with the LASSO coefficient;C,the high/low-risk groups’ survival comparison in the train set;D,the ROC curves and AUC of 1,3 and 5 years in the train set;E,the high/low-risk groups’ survival comparison in the test set;F,the ROC curves and AUC of 1,3 and 5 years in the test set.LASSO,least absolute shrinkage and selection operator;lncRNAs,long non-coding RNAs;ROC,receiver operating characteristic;AUC,area under the curve.

External validation and independent prognostic analysis for model

For further confirmation of the efficacy of this model,we verified this model in an external validation dataset from the GEO database,GSE30219,with 289 samples from the GPL570 platform.Similarly,a Kaplan-Meier survival analysis was performed to compare the survival difference between the high-risk and low-risk groups based on the median risk score cutoff,which showed that the high-risk group is bad in prognosis (Figure 7A).Moreover,the AUC of the ROC curves in 1,3 and 5 years respectively,are 0.68,0.64 and 0.61 (Figure 7B).These results robustly indicate that this model has an excellent performance in LUAD patients’prognosis survival prediction.

To verify whether our model can be used as an independent prognostic factor independent of other clinical traits.We conducted independent prognostic analysis for this model in the entire LUAD TCGA dataset combined with clinical information.First,a univariate independent prognostic analysis showed that age and gender are not correlated with survival,but the risk score and tumor stage are survival-related (Figure 7C).Then,a multivariate independent prognostic analysis was implemented,which indicated the risk score and tumor stage could be independent of other factors as independent prognostic factors (Figure 7D).In conclusion,the results showed that the model-based risk score and tumor stage are independent factors for prognosis,while age and gender are not independent prognosis factors.Therefore,our established model can be an independent prognostic factor for LUAD prognosis.

Figure 7 External validation and independent prognostic analysis. A,The high/low-risk groups’ survival comparison in GEO external validation set;B,the ROC curves and AUC of 1,3 and 5 years in the GEO external validation set;C,univariate Cox regression analysis for clinical features and risk score to analyze the relationship between these signatures and OS;D,the ROC curves and AUC of 1,3 and 5 years in the train set;E,multivariate Cox regression analysis showed that the model-based risk score was an independent prognostic factor for the prognosis of LUAD patients.GEO,Gene Expression Omnibus;ROC,receiver operating characteristic;AUC,area under the curve;OS,overall survival;LUAD,lung adenocarcinoma.

Screening differential expression gene in model and identifying clusters’signature based on risk score

We utilized the“limma”package to analyze the differentially expressed genes (lncRNAs) and filtered genes rely on the following criteria:∣Log2 foldchange∣>1 andP-value <0.05.At last,we confirmed that SFTA3 was downregulated in tumor group (Log2 foldchange=–1.4548,P-value <0.001).And its expression level in the tumor and the normal group was compared in a box map,showing that SFTA3 is relatively high expression in normal but low expression in the tumor group (Figure 8A).According to the previous Cox analysis regression analysis (Figure 2B),the Hazard Ratio of SFTA3 is 0.842(P<0.001),which indicates that a high SFTA3 expression level is a protective factor for patients’survival and a low SFTA3 expression level may indicate bad prognosis.Meanwhile,this comparison has established that SFTA3 is low expressed in tumors.It also reminds us that detecting SFTA3 low expression may be an essential marker for LUAD diagnosis.

Moreover,for more profoundly investigating the multiple traits of LUAD.We compared the differences in clusters based on our model’s risk score.Results showed that cluster1 has a higher risk score than cluster2,consistent with our previous studies that cluster1 has a worse prognosis than cluster2 (Figure 3C,Figure 8B).Combined with our previous studies,cluster1 has a lower immune score than cluster2(Figure 4H),which can explain why cluster1 has a poor prognosis–due to cluster1 in low immune infiltration status that causes patients into high risk and has an unfavorable prognosis.Therefore,these findings suggested that the immune process plays an essential role in LUAD progression.

Figure 8 The differential expression of SFTA3 and the risk score difference in clusters. A,SFTA3 differential expressed between the tumor(low) and normal(high) samples;B,the risk score difference of cluster1 (high)and cluster2 (low).

Discussion

LUAD is one of the most common pathology types of NSCLC,which has an increasing incidence recent years and a poor prognosis for treatment [16,17].It’s really urgent for us to find an effective way to evaluate the prognosis of LUAD patients and deeply understand its immunological traits for its treatment.Necroptosis is a newly confirmed cell programmed death way based on RIPK1/RIPK3 activation and MLKL execution,which is also closely correlated with anti-tumor immune and TME.Increasing evidence indicates that upregulating the necroptosis level can reduce tumor growth and metastasis,which remarkably increases the survival time in vivo and in vitro [18,19].In addition,immune checkpoint inhibitors (ICIs) are a new rising targeted therapy for cancer in recent years.For instance,Atezolizumab was applied to treat lung cancer’s first-line chemotherapy,which has achieved good results [20].Therefore,the application of ICIs is a promising way for tumor therapy.Interestingly,a crosstalk between the necroptosis mediated anti-tumor immune and ICIs mediated anti-tumor effect was observed recently,which showed that both of them have a synergistic effect to enhance the effect of anti-tumor immunity [21].

In our study,we identified the immunological characteristics of LAUD based on necroptosis and consensus clustering.By comparing the immune score and combined with their survival status,we found that cluster1 has a lower immune score that accompanies a worse survival probability.It indicates that low immune activation status may mean a poor prognosis.Later,the infiltration of the immune cells in the LUAD TME was analyzed.We found that the tumor-associated macrophages take a large proportion of the whole immune cells system and macrophages M1 is higher in cluster1,while macrophages M2 is higher in cluster2.According to previous studies,both macrophages M1 and M2 are involved in anti-tumor immune related inflammation [22].But they are different in function that M1 promotes inflammation and anti-tumor,activation of macrophages M1 could promote multiple cytokines secretion,such as interleukin-1β and tumor necrosis factor-α [23,24],while M2 undertakes anti-inflammation and promotes tumor progress which involved the process of angiogenesis,neovascularization and the stromal remolding/activation [25,26].Unexpectedly,our results are conversely to this rule.Although cluster1 has a higher M1 and lower M2 than cluster2,it has a worse prognosis.To answer this question,we further focus on the crosstalk between necroptosis and the tumor-associated macrophages.We found that macrophages M1 mediated high expression of inflammation cytokines negatively correlated with necroptosis,upregulating inflammation cytokines may reduce RIPK-1 depend cell death [27,28].Therefore,we speculate that the necroptosis was downregulated in cluster1 because of M1-mediated high inflammation infiltration.We also identified that the immune checkpoints of cluster1 are universally higher than cluster2 (PD-1,PD-L1,PD-L2,LAG3).In conclusion,we considered that utilizing ICIs and combining them with elevated necroptosis levels has considerable prospects for LUAD treatment.

To predict the prognosis of LUAD patients,we established a novel necroptosis-related lncRNA-based prognostic model.This model combines the necroptosis trait with prognosis,and we emphasize that not only genes differential expression decides the prognosis for patients,but also the correlation between gene expression and characteristic phenotype (necroptosis).After the model was constructed and evaluated,we applied a differential gene expression analysis.Eventually,SFTA3 was filtered as the most differential expression lncRNA.LncRNA SFTA3 is a protective gene correlated with the necroptosis and patients’ prognosis;however,it was downregulated in LUAD tumor samples.Another study also showed that SFTA3 has highly functional enrichment [29].Therefore,specifying its function in the future is meaningful for LUAD therapy.In addition,another hub gene–CDHR2 was a screened coding gene via the WGCNA algorithm based on our consensus clustering results.Its high expression demonstrated a harmful effect for the OS time of LUAD patients and its high immunohistochemical staining was also found in some LUAD tumor samples.In short,we speculate that these multiple dimensional gene expression traits may help us diagnosis and find more new treatment ways for LUAD.

However,this study still has some limitations.Although we utilize bioinformatics analysis to identify the immunology signatures and construct a prognostic model for LUAD,experimental validation and clinical follow-up are needed。

Conclusion

We have constructed a novel 13 necroptosis-related lncRNAs model and identified the immune signatures of LUAD patients,which could apply in clinical to evaluate patients’ prognosis and provide a new insight for LUAD treatment.