Abstract Background: Recent studies have identified several molecular subtypes of lung adenocarcinoma (LUAD) that have different prognoses to help predict the efficacy of immunotherapy. However, the prognostic prediction is less than satisfactory. Alterations in intracellular copper levels may affect the tumor immune microenvironment and are linked to cancer progression. Previous studies have identified some genes related to cuproptosis. The characteristics of the cuproptosis molecular subtypes have not been thoroughly studied in LUAD. Methods: The transcriptomic data and clinical information of 632 LUAD patients were used to investigate the LUAD molecular subtypes that are associated with the cuproptosis-related genes (CRGs), the tumor immune microenvironment, and stemness. The cuproptosis score was constructed using univariate Cox regression and the minor absolute shrinkage and selection operator (LASSO) to quantify the prognostic characteristics. Results: Three different molecular subtypes related to cuproptosis, with different prognoses, were identified in LUAD. Cluster A had the highest cuproptosis score and the worst prognosis. Patients in the high cuproptosis score group had a higher somatic mutation frequency and stemness scores. Patients in the low cuproptosis score group had more immune infiltration and better prognosis. Conclusion: Molecular subtypes of LUAD based on CRGs reflect the differences in LUAD patients. The cuproptosis score can be used as a promising biomarker, which is of great significance to distinguish the relationship between cuproptosis and the immune microenvironment. The cuproptosis signature based on the cuproptosis score and clinical characteristics of individual patients will be useful for guiding immunotherapy in LUAD. Keywords: cuproptosis, LUAD, molecular subtypes, prognosis, tumor microenvironment, immune infiltration 1 Introduction With an estimated 2 million new cases and 1.76 million deaths each year, lung cancer is one of the most commonly diagnosed cancers and the leading cause of cancer-related deaths worldwide. Lung adenocarcinoma (LUAD) is the most common histological type of lung cancer ([38]Thai et al., 2021). In recent years, immune checkpoint inhibitors alone or in combination with chemotherapy have significantly improved survival in patients with advanced LUAD. In the KEYNOTE-024 study, patients without EGFR/ALK aberrations treated with pembrolizumab had significantly improved overall survival (30.0 months; 95% CI, 18.3 months to not reached) compared with chemotherapy (14.2 months; 95% CI: 9.8–19.0 months) (HR, 0.63; 95% CI, 0.47–0.86) ([39]Reck et al., 2019). In the KEYNOTE-189 study, the overall survival (22.0 months; 95% CI, 19.5–25.2 months) for patients treated with pembrolizumab in combination with chemotherapy in nonsquamous non-small cell lung cancer was significantly longer than that in the placebo group (HR, 0.56; 95% CI, 0.45–0.70) ([40]Gadgeel et al., 2020). Although immunotherapy is an effective treatment, not all patients with LUAD can benefit from it. In addition, LUAD is a heterogeneous disease, which also makes identifying new subtypes essential to predicting prognosis and ensuring that patients receive personalized treatment ([41]Liu et al., 2021). Recently, an increasing number of molecular subtypes have been studied to predict the efficacy of immunotherapy. For example, Zhang et al. isolated three types of ferroptosis-related molecules in LUAD, which helped predict the prognosis, tumor microenvironment (TME) cell infiltration characteristics, and immunotherapy effects in patients with LUAD ([42]Zhang et al., 2021). Wang et al. identified two distinct subtypes of LUAD. The high-risk subtype was characterized by a lower TIDE score, increased programmed death-ligand 1 (PD-L1) expression, higher tumor mutation burden (TMB), elevated levels of the cell cycle modulators CDK4/CDK6, and TP53 mutations, and it was implicated for immune checkpoint blockade therapy ([43]Wang et al., 2020). [44]Wu et al. (2021) established a promising immunoprognostic model associated with TP53 to identify early-stage LUAD patients with a high risk of unfavorable survival. These studies have shed light on the molecular subtypes of LUAD. However, the above classification strategies for predicting immune efficacy were not sufficiently effective. Thus, finding new molecular subtypes of LUAD is vital for identifying potential benefits of immunotherapy. Copper-induced proptosis (cuproptosis) is a novel form of cell death induced by excessive intracellular copper ([45]Tang et al., 2022). Copper overload induces lipoylated dihydrolipoamide S-acetyltransferase (DLAT) aggregation, which is associated with mitochondrial tricarboxylic acid (TCA) cycle activation. This results in proteotoxic stress and leads to cuproptosis ([46]Wang et al., 2022). In recent studies, alterations in intracellular copper levels have been linked to cancer development and progression, including lung cancer ([47]Ge et al., 2022). Therefore, cuproptosis may serve as a novel target for treating LUAD. Recent studies have shown that intracellular copper regulates key signaling pathways mediating PD-L1-driven cancer immune evasion ([48]Voli et al., 2020). These findings suggest that cuproptosis may affect the tumor immune microenvironment, and identification of the characteristics of cuproptosis may effectively predict the efficacy of immunotherapy. Previous studies have identified certain genes related to cuproptosis ([49]Polishchuk et al., 2019; [50]Aubert et al., 2020; [51]Dong et al., 2021; [52]Ren et al., 2021; [53]Bian et al., 2022; [54]Kahlson and Dixon, 2022; [55]Tang et al., 2022; [56]Tsvetkov et al., 2022). Several studies have reported the predictive value of these cuproptosis-related genes (CRGs) in LUAD. Li et al. constructed a prognostic model for patients with radiotherapy resistance based on CRGs screened from RNA-sequencing data of radiation-treated cell lines ([57]Li et al., 2022). [58]Hu et al. 2022 and [59]Zhang et al. (2022) also used different CRGs to construct risk models to predict the prognosis of LUAD patients. However, the role of CRGs in affecting the immune microenvironment in LUAD has yet to be explored. In the present study, 632 LUAD samples were divided into three cuproptosis-related subtypes based on differentially expressed genes (DEGs) of LUAD subtypes according to CRGs and immune profiles. Additionally, a model with cuproptosis scores was established. The characteristics of the immune microenvironment between low and high cuproptosis score groups were explored. These findings show that the cuproptosis score might be an independent prognostic factor for LUAD patients and predict the clinical efficacy of immunotherapy. 2 Materials and methods 2.1 Data sources and preprocessing The transcriptomic data and clinical information of LUAD patients were downloaded from the TCGA database ([60]https://portal.gdc.cancer.gov/) and Gene Expression Omnibus (GEO) database ([61]https://www.ncbi.nlm.nih.gov/geo/). Patients of different sexes and races were included, and patients without survival information were excluded. The transcriptomic data and clinical information of 632 tumor samples were collected. Of these tumor samples, 516 were from the TCGA-LUAD dataset and 116 were from GEO ([62]GSE26939) ([63]Wilkerson et al., 2012). Additionally, information on 59 normal samples was collected from TCGA. The mutation information of 557 LUAD patients was obtained from the TCGA database. The GDC GISTIC copy number (gene-level) dataset of 531 LUAD patients was obtained from UCSC Xena ([64]https://xena.ucsc.edu/). All transcriptomic data were processed to log2 form. All gene expression levels that repeatedly appeared in multiple rows were averaged and kept in one row. All data were downloaded in June 2022. The R (version 4.2.0) and R Bioconductor packages were used for all data analyses. 2.2 Identification of CRGs in LUAD Initially, 19 cuproptosis-related genes (NFE2L2, NLRP3, ATP7B, ATP7A, SLC31A1, FDX1, LIAS, LIPT1, LIPT2, DLD, DLAT, PDHA1, PDHB, MTF1, GLS, CDKN2A, DBT, GCSH, and DLST) were identified from previous studies ([65]Polishchuk et al., 2019; [66]Aubert et al., 2020; [67]Dong et al., 2021; [68]Ren et al., 2021; [69]Bian et al., 2022; [70]Kahlson and Dixon, 2022; [71]Tang et al., 2022; [72]Tsvetkov et al., 2022). Cuproptosis-related gene expression was determined in normal cells and LUAD tumor cells from the TCGA-LUAD dataset. The difference between the normal and tumor groups was analyzed using the Wilcoxon test and displayed in a box plot. p values less than 0.05 were considered to indicate significant differences in gene ontology. 2.3 Detection of CRG-related mutations in LUAD The numbers of mutated genes were calculated using the mutation information of each sample obtained from the TCGA database. The mutation information of 19 CRGs and the clinicopathological characteristics were displayed in a waterfall plot using the R package “maftools”. The copy number variation (CNV) frequencies of 19 CRGs are displayed in a bar plot. The “RCircos” package in R was applied to show the location of the 19 CRGs on chromosomes. 2.4 Survival analysis of CRGs in LUAD patients The expression of the CRGs was extracted from 632 merged data from TCGA-LUAD and [73]GSE26939, and 17 of these CRGs (NFE2L2, ATP7B, ATP7A, SLC31A1, FDX1, LIAS, LIPT1, DLD, DLAT, PDHA1, PDHB, MTF1, GLS, CDKN2A, DBT, GCSH, and DLST) were extracted. The 95% confidence interval (CI), hazard ratios (HR), and p values of these 17 CRGs were calculated by univariate Cox regression using the “Survival” package in R. The interaction and impact of each CRG on the prognosis are shown in a network. 2.5 Correlation analysis of CRGs and LUAD immune estimation The ESTIMATE tool from the R package “estimate” and gene expression signatures were used to estimate the fraction of stromal and immune cells within the tumor samples and to estimate the elements of the TME, including StromalScore, ImmuneScore, ESTIMATEScore, and TumorPurity ([74]Yoshihara et al., 2013). CIBERSORT, a method that reduces noise and unidentified mixtures, can recognize the composition of tumor cells by using gene expression profiles ([75]Newman et al., 2015). Correlations between the CIBERSORT data for immune cell infiltration and the 4 CRGs (LIPT1, DLAT, PDHA1, and DLST) that were significantly associated with survival situations are displayed. Additionally, correlation tests between these 4 CRGs and the ImmuneScore were performed using Spearman analyses. 2.6 Clustering of LUAD patients based on CRGs The R package “ConsensusClusterPlus” was used for consensus clustering and result visualization ([76]Wilkerson and Hayes, 2010). The clustering was based on the 2 CRGs (PDHA1 and DLAT) with the two highest correlation scores, as determined by ImmuneScore. The efficacy of the consensus clustering was determined by performing principal component analysis (PCA). Two cuproptosis subtypes (CRG cluster A and CRG cluster B) were found. Clinical characteristics based on the cuproptosis subtypes are displayed in a heatmap. Functional and pathway enrichment analyses of the cuproptosis subtypes were performed. The “GSEABase” and “GSVA” R packages were applied for pathway enrichment analysis and to analyze the differences in biological functions among the different cuproptosis clusters ([77]Hänzelmann et al., 2013). The gene sets of “c2. cp.kegg.symbol” and “c5. go.symbols” were downloaded from Gene Set Enrichment Analysis (GSEA) ([78]https://www.gsea-msigdb.org/gsea/datasets.jsp) and were used to run GSVA enrichment analysis. The single sample gene set enrichment analysis (ssGSEA) method was used to evaluate the scores of the TME cells in each LUAD sample ([79]Barbie et al., 2009). The immune infiltration of the cuproptosis subtypes is displayed in a box plot. 2.7 Clustering of LUAD patients based on the DEGs between the cuproptosis subtypes Using the linear model and empirical Bayes statistics for differential expression in the R package “limma,” 122 DEGs between the two cuproptosis subtypes were identified (absolute logFC >0.585 and adjusted p values <0.05). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were used as references, and enrichment