Proteomic stratification reveals immune metabolic heterogeneity in lung adenocarcinoma

Abstract

Objective

Lung Adenocarcinoma (LUAD) is a leading cause of cancer-related mortality worldwide. The relationship between metabolic reprogramming and immune infiltration has been identified as having a crucial impact on LUAD progression. The aim of this study is to achieve a deeper understanding of the interplay between the immune system and metabolism in the LUAD microenvironment.

Methods

A total of 213 treatment-naïve LUAD patients from two independent cohorts were enrolled. Using the proteomic data from one of the cohorts (n = 103), three molecular subtypes of LUAD were identified based on their immune signatures, clinical characteristics, metabolic reprogramming, and genomic features were analyzed. The data from the other cohort (n = 110) were used to validate the findings.

Results

Three subtypes with distinct immune environment (IM1-IM3) were identified in LUAD based on the proteomic dataset, each of which has distinctive clinical, immune, and metabolic characteristics. Among these subtypes, IM2, which has the highest immune infiltration and is more likely to benefit from immunotherapy. In contrast, IM3 was found to have the poorest prognosis, exhibits active hypoxia and glycosaminoglycan biosynthesis, and is thought to be associated with extracellular matrix remodeling and epithelial-mesenchymal transition activation. Additionally, the active glycosaminoglycan biosynthesis pathway in the IM3 subtype was associated with an immune-desert microenvironment.

Conclusions

This study presents the proteomic immune stratification of LUAD, revealing the possible link between immune cells and reprogramming of LUAD metabolisms, which may be a viable therapeutic strategy to improve LUAD immunotherapy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05164-0.

Introduction

Lung cancer accounts for the highest cancer incidence and mortality worldwide. The 5-year survival rate is less than 20% [1]. Lung adenocarcinoma (LUAD) is the most common histological subtype of non-small cell lung cancer (NSCLC), which accounts for about 40% of lung malignancies [2]. Smoking, long-term exposure to radon, occupational exposure to carcinogens, and outdoor air pollution are among the major risk factors associated with the development of LUAD [3].

Although surgical resection, chemotherapy, and immunotherapy can improve the treatment outcomes for patients with lung adenocarcinoma to some extent, the high recurrence rate and the difficulty in treating metastatic lung adenocarcinoma remain major challenges in its treatment [4]. Immunotherapy plays a crucial role in the treatment of LUAD [5]. Compared to traditional radiation and chemotherapy, immunotherapy has relatively fewer side effects [6]. Moreover, immunotherapy can be personalized based on the patient’s immune status and tumor characteristics, enhancing the specificity and effectiveness of the treatment [7]. Immune subtyping of LUAD patients helps guide personalized treatment strategies and improve therapeutic outcomes. Additionally, immune subtyping provides a deeper understanding of the biological characteristics of different patients’ tumors, revealing the relationship between the immune microenvironment and tumor progression, and offering new insights for future research and treatment.

Advances in omics technologies permit the large-scale analysis of the molecular characteristics for LUAD and have defined several clinical patient subtypes [8]. Transcriptomic subtyping has revealed LUAD subclasses that differ in the expression of proteins related to proliferation, metabolism, stemness and lung function. Among these transcriptome-based subtyping, The The Cancer Genome Atlas (TCGA) study integrated genomic, transcriptomic, and clinical data to identify distinct subtypes of LUAD. By classifying LUAD based on gene expression patterns and molecular features, the research revealed significant heterogeneity in tumor biology and patient prognosis. Key findings include the identification of subtypes with varying degrees of aggressiveness, molecular characteristics, and treatment responses [9]. A recent study identified disulfidptosis-related molecular subtypes in lung adenocarcinoma and developed a prognostic signature based on disulfidptosis-related genes, revealing their associations with the tumor immune microenvironment, drug sensitivity, and patient survival [10]. A recent study using single-cell transcriptomic analysis revealed distinct tumor immune microenvironment landscapes between EGFR-mutant and EGFR–wild-type LUAD [11]. Another study identified a metabolism-related signature predicting immunotherapy response in LUAD and highlighted SLC25A1 as a key immune-exclusion gene [12].These insights have paved the way for personalized treatment approaches in lung adenocarcinoma, highlighting the importance of tailoring therapies based on the specific molecular features of each tumor.

Proteomics has proven to be a valuable approach for investigating cancer biological changes and disease classification, as it offers a comprehensive representation of cellular states [13]. Based on proteomic data, Xu et al. stratified the 103 LUAD cohort into molecular subtypes S-I, S-II and S-III. Among these subtypes, the S-III (characterized by proliferation and proteasome) was associated with poor outcomes [14], The LUAD proteomics study from Clinical Proteomic Tumor Analysis Consortium (CPTAC) classified LUAD patients into four subtypes, among which the C3 and C4 subtypes are considered to be associated with patients’ immune profiles [15]. Our study also utilized some features from this data to validate our findings. In contrast, the subtyping from PDC000220 mainly focuses on genetic mutation characteristics and does not elucidate the patients’ immune profiles [16].

Although previous studies have yielded important insights into LUAD, a proteome-based immune taxonomy, potentially integrated with metabolomic features and externally validated, has not yet been established to improve immunotherapy stratification. In fact, although immune checkpoint blockade (ICB) has been shown to enhance clinical survival in patients with LUAD, about 50% of cases exhibit no response to ICB and the underlying mechanism for therapeutic resistance remains unclear [17]. It is necessary to stratify clinical LUAD patients into different immune subtypes for targeted treatment design boosting the efficacy of cancer immunotherapy.

In this study, we identified three immune subtypes of LUAD based on the proteome and investigated the interactions between immune infiltration and metabolic reprogramming in the LUAD immune microenvironment. This research has the potential to shed light on the LUAD immune microenvironment and LUAD personalized treatment.

Materials and methods

Experimental design and statistical rationale

For the discovery cohort, we used “Xu et al. proteomics dataset”. To validate the findings, the other datasets “Gillette et al.” were established. We also designed a protocol to identify and characterize immune subtypes, examine the metabolic heterogeneity and its correlation with immune infiltration, and investigated the role of GAGs biosynthesis in the IM3 immune microenvironment.

Kruskal-Wallis test and Wilcoxon test was used to test whether there are significant differences in biological characteristics among three subtypes, such as immune cells, hypoxia pathways. Spearman analysis was used to examine the correlation score between two variables. Kaplan-Meier plots (Log-rank test) were used to analyze the prognosis differences. All statistical tests were two-sided, and statistical significance was defined as p-value < 0.05. To account for multiple testing, Benjamini-Hochberg FDR correction was applied to adjust the p-values. All these statistical analyses were performed in R.

Specimen collection

The collection procedure of lung adenocarcinoma (LUAD) tissue samples used for immunohistochemistry (IHC) analysis in this study was consistent with our previously published work [18]. Briefly, human LUAD tumor tissues and paired adjacent normal tissues were obtained from patients who underwent surgical resection at the National Cancer Center and Cancer Hospital of the Chinese Academy of Medical Sciences and Peking Union Medical College. None of the patients received chemotherapy, radiotherapy, targeted therapy, or immunotherapy prior to surgery.

Formalin-fixed and paraffin-embedded (FFPE) tissue specimens were prepared according to standard pathological protocols. Hematoxylin and eosin (H&E)-stained sections were independently reviewed by two experienced pathologists to confirm the histopathological diagnosis and assess tissue quality. Tumor samples with adequate tumor content and adjacent normal tissues without tumor cell infiltration were selected for subsequent IHC staining.

The use of these human tissue samples was approved by the Research Ethics Committee of the National Cancer Center and Cancer Hospital of the Chinese Academy of Medical Sciences and Peking Union Medical College.

Human ethics

This study consisted of two parts: analysis of publicly available datasets and validation using human tissue samples.

For the bioinformatics analysis, all proteomics datasets (including the “Xu et al.” and “Gillette et al.” cohorts) were obtained from previously published studies and publicly accessible resources. No new human participants were recruited for this part, and all data were de-identified; therefore, ethical approval and informed consent were not required for the use of these datasets.

For the experimental validation, human LUAD tissue samples were collected as described above. The study protocol was reviewed and approved by the Research Ethics Committee of the National Cancer Center and Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College.

Representative immune signatures

The 181 tumor-related immune signatures were collected from previous literature. The ssGSEA was used to calculate separate enrichment scores for each pairing of protein expression profile and immune signatures as implemented in the R package GSVA. WGCNA package in R was used to cluster the immune signatures of each samples. Finally, within each module, the signature with the top kME was considered as the most representative signature of that module, where the kME of a signature was measured by the signed KME function in WGCNA.

Identification and analysis of immune subtypes

The 16 representative signatures were used to cluster LUAD tumor samples using CancerSubtypes in R. A log-rank test was applied to compare the overall survival (OS) or disease-free survival (DFS) among three immune subtypes.

Tumor Immune Dysfunction and Exclusion (TIDE, http://tide.dfci.harvard.edu/) tool was used to estimate the TIDE score for the clinical responsiveness of patient to ICB therapy (anti-PD1/CTLA4).

Analysis of metabolic pathway activity

The representative proteins of metabolic pathways were downloaded from the KEGG database, and the ssGSEA method was used to analyze the scores of each metabolic pathway. The differences in metabolic pathway scores among the three subtypes were assessed using the Wilcoxon test or the Kruskal-Wallis test.

Functional analysis and GSEA

The R package “clusterProfiler” was applied for the Gene Ontology (GO) analysis of three immune subtypes. GSEA was applied to enrich hallmark gene sets downloaded form the Molecular Signatures Database (MSigDB). Input proteins were ranked in descending order according to the log2FC values. Enrichment significance was estimated using default settings and 1000 permutations. Benjamini-Hochberg-adjusted p-values less than 0.05 were considered significantly enriched.

Gene mutations analysis of immune subtypes

Significantly mutated genes were generated by Maftools for the CNHPP-LUAD cohort accessed from the mutation annotation file. MAF files were created to analyze the exome characteristics of the three immune subtypes, and the associated tumor driver genes were calculated.

Immunohistochemical staining and scoring

IHC staining and scoring of CD8 and PD-L1 were conducted in our previous research [18]. The patient cohort, scoring method, and result interpretation for lung adenocarcinoma were identical to our previous study. However, due to differences in the subtype classification method, we recollected the IHC staining images from the patients.

Potential drugs for the treatment of three immune subtypes

Cmap, a public tool that enables researchers to identify compounds that may inhibit diseases based on gene expression profiles, was employed to predict which drugs might be effective for the treatment of each LUAD subtype. We compared the expression profiles of each LUAD subtypes for Cmap analysis. Specific analysis based on Cmap tools was conducted to further interrogate the action mechanisms pattern of drugs that most likely inhibit LUAD.

Identification of cancer subtypes based on microenvironment features

To clarify the impact of the surrounding environment of the tumor on CPTAC-LUAD, nearest template prediction (NTP) was applied to assign patients into three immune subtypes. NTP performed class prediction using predefined protein markers and returned the significance level of each sample prediction with a nominal P-value.

Results

Immune subtypes of LUAD and their associations with patient clinical information

The biological mechanisms underlying the response to immunotherapy in the LUAD immune microenvironment are not yet fully understood. Analyzing the immune microenvironment status of lung adenocarcinoma patients based on proteomics data is crucial for designing immunotherapy strategies, predicting patient responses, and achieving precision medicine [19].

To perform immune-based molecular subtyping of lung adenocarcinoma patients, we undertook an extensive literature search and analyzed 181 signatures that are known to be associated with immune activity in tumor tissue [20]. Based on the protein expression profile, we scored 181 immune signatures across each of the 103 LUAD tumor samples using Single Sample Gene Set Enrichment Analysis (ssGSEA). Then, the Weighted Gene Co-expression Network Analysis (WGCNA) was used to cluster 181 immune signatures into two modules based on the enrichment scores. WGCNA identified representative immunological features in two modules. Based on 16 representative immune features, 103 LUAD patients were classified into three subtypes: IM1, IM2, and IM3. These three immune subtypes showed distinct representative immune signatures, with the IM2 exhibiting immune-active state (Fig. 1A, Table S1). From the clinical information, IM3 tumors have the lowest degree of differentiation and the highest tumor stage, with no statistically significant differences in age and gender (Fig. 1B). The IM1 subtype has the best OS, whereas both IM2 and IM3 has worse survival, DFS and OS show consistent trends (Fig. 1C, Table S2). For early-stage LUAD patients (Stage I), differences in prognosis can also be observed among the three subtypes (Figure S1). We also found that based on immune scoring, the distribution of the three subtypes of patients on the PCA map shows distinct separation (Figure S2A). We performed an overlap analysis of our immune subtype with the molecular subtype identified by Xu et al. (XMS: S-I, S-II, and S-III). We observed the highest rate of oncordant assignments occurs between IM3 and S-III subtypes. IM3 is mainly distributed in S-II and S-III subtypes, matching XMS finding that S-III subtype has the worst prognosis (Fig. 1D).

Fig. 1.

The immune subtypes in LUAD cohort. A Heatmap of immune signature scores calculated by ssGSEA. B heatmap of clinical information, XMS and IMS. C Kaplan-Meier curves of OS (left) and DFS (right) for each immune subtype in LUAD cohort. D Relationship between XMS and IMS

Characteristics of the immune microenvironment among three immune subtypes

Considering the crucial role immunomodulators in the formation of tumor microenvironment (TME), we examined the relative expression levels of major histocompatibility, immunostimulatory, and immunoinhibitory molecules amont three immune subtypes (Fig. 2A). We found that these immunomodulators exhibited higher expression in IM2, while many of them were relatively downregulated in IM1 and IM3. Among them, MHC-I expression was elevated in the IM2 subtype (Figure S3A). In contrast, MHC-II expression showed a relatively higher level in IM1, although the difference between IM1 and IM3 was not pronounced (Figure S3B). The IM3 subtype exhibits more pronounced immunosuppressive characteristics, accompanied by elevated levels of CD274 and CD276. In contrast, immune-stimulatory molecules are expressed at higher levels in the IM2 subtype. Meanwhile, using LUAD proteomics data to predict patients’ responses to immunotherapy, we found that IM2 subtype are expected to achieve better outcomes with immunotherapy (Fig. 2B). Currently, solely relying on PD-L1 molecules to identify suitable candidates for immunotherapy has faced some challenges [21]. However, our proteomics-based immune subtyping can assist in selecting appropriate patients for immunotherapy to a certain extent, thus achieving precision medicine for patients.

Fig. 2.

Characteristics of the Immune Microenvironment Among Three Immune Subtypes. A Expression Level of immunomodulatory proteins in three immune subtypes. B Immune response to treatment in the three immune subtypes of patients. C Immunohistochemistry staining images of the three immune subtypes of patients

To experimentally validate the immune characteristics of each subtype, two experienced pathologists independently evaluated the proportion of infiltrated immune cells on H&E-stained slides of 60 LUAD samples from our previous study in our own cohort [22]. As expected, the proportion of infiltrated immune cells was the highest in IM2, suggesting that IM2 was “hot” tumors. We also analyzed immunohistochemical (IHC) staining of CD8 and PD-L1 and evaluated their expression of positive cells (0-100%) and the average intensity of the positive staining (0, 1+, 2+, or 3+) into consideration. We found that CD8 and PD-L1, the key immune markers and predictors of immunotherapy efficacy, were the highest in IM2 (Fig. 2C).

Characteristics of tumor biological behavior among the three immune subtypes

To further elucidate the molecular mechanisms underlying the poor prognosis of IM3 subtypes, we conducted an in-depth proteomic analysis of IM3 subtype characteristics. GSEA revealed that IM3 subtypes exhibited significantly elevated scores in epithelial-mesenchymal transition (EMT), hypoxia, and glycolysis pathways. Conversely, their scores for fatty acid metabolism and oxidative phosphorylation pathways were notably lower. This indicates that IM3 tumors possess a higher propensity for metastasis and exhibit the Warburg effect (Fig. 3A).

Fig. 3.

Biological behaviors of highly expressed proteins among three subtypes. A GSEA plots of hallmark pathways with high and low enrichment scores in the IM3 subtype. B GO enrichment pathways of highly expressed proteins in the three immune subtypes. C Expression levels of cell cycle-related proteins in the three immune subtypes of patients. D Expression levels of focal adhesion-related proteins in the three immune subtypes of patients

Further, in-depth Gene Ontology (GO) analysis revealed that the extracellular matrix (ECM) pathway is highly expressed in IM3 subtype tumors, such as extracellular matrix organization, extracellular structures organization and extracellular encapsulation structures organization. In contrast, the IM2 subtypes are primarily associated with immune-related pathways such as antigen receptor-mediated signaling pathway, B cell activation, and immune response. The IM1 subtypes presenting intermediate features of both IM2 and IM3 subtypes which show high expression in cytokine-related pathways and mismatch repair-related pathways(Fig. 3B, Table S3).

Based on the analysis of representative protein expression levels in signaling pathways, the significantly altered proteins in the representative pathways of the three subtypes play roles in cell cycle and focal adhesion. This finding provides insights into the molecular mechanisms underlying the poor prognosis of the IM3 subtype (Fig. 3C and D). Additionally, lung-associated proteins are significantly overexpressed in the IM1 subtype (Figure S2B).

Metabolic heterogeneity and its correlation with immune infiltration among three subtypes

Tumor progression is a dynamic process involving continuous interactions between tumor cells and other cells within the TME, including immune cells and stromal cells [23]. As a result, tumor cells can profoundly influence various cell types within the tumor ecosystem to promote their survival and the spread of malignancies [24]. Among these factors, alterations in tumor metabolism are believed to significantly impact the internal microenvironment of the tumor, playing a crucial role in tumor progression [25].

Thus, we conducted the ssGSEA analysis of the metabolic characteristics of the three immune subtypes in patients with LUAD, exploring the interaction between immunity and metabolism in tumor progression (Table S4). Our research revealed that the IM1 subtype exhibited more active metabolic characteristics (Fig. 4A). In contrast, differential analysis between the IM3 subtype and other subtypes indicated a significant increase in the score of the Glycosaminoglycans (GAGs) biosynthesis pathway in the IM3 subtype (Fig. 4B). We also found that the expression of proteins related to the GAGs biosynthesis pathway was significantly elevated (Fig. 4C).

Fig. 4.

Metabolic heterogeneity among three immune subtypes. A Heatmap of representative metabolic pathways scores among three immune subtypes. B Volcano plot of differential metabolic pathways in IM3 immune subtypes. C Expression levels of key proteins involved in the glycosaminoglycan biosynthesis pathway among three immune subtypes. D Expression levels of hallmark hypoxia scores among three immune subtypes. E Correlation between hypoxia scores and glycosaminoglycan biosynthesis scores. F Correlation between glycosaminoglycan biosynthesis score and immune infiltration

GAGs are complex polysaccharides composed of repeating disaccharide units [26]. They play a crucial role in the extracellular matrix and on cell surfaces, participating in the construction of the tumor microenvironment [27]. GAGs can interact with receptors on the surface of immune cells, influencing their activity and migration capabilities [28]. Studies have reported that hyaluronic acid can bind to the CD44 receptor, affecting the function of T cells and NK cells, thereby helping tumor cells evade immune surveillance [29].

Through further analysis, we discovered that the hypoxia pathway was significantly overexpressed in the IM3 subtype of lung adenocarcinoma and was significantly positively correlated with GAG biosynthesis (Fig. 4D and E, S5). Studies have reported that under hypoxic conditions, the expression and activity of hyaluronic acid synthase significantly increase, leading to the accumulation of HA in the tumor microenvironment [30]. This accumulation affects tumor cell proliferation, migration, invasion, and immune response capabilities. In our correlation analysis, we found that immune signatures positively correlated with GAGs biosynthesis include the infiltration of exhausted T cells, immune checkpoint pathways, and MDSCs. Conversely, when GAGs biosynthesis is active, Naïve CD4 + T cells and activated B cells, which play roles in antitumor immune responses, are suppressed in the tumor immune microenvironment (Fig. 4F, S4C). This findings further confirmed that in the tumor microenvironment, hypoxia affects GAGs biosynthesis and secretion, leading to the remodeling of the tumor ECM. This remodeling not only influences tumor cell behavior, creating conditions favorable for tumor cell survival and spread, but also inhibits immune cell activity, promoting tumor immune evasion [31].

We found that the fatty acid biosynthesis pathway is significantly overexpressed in the IM1 subtype (Fig. 4A), with the expression levels of fatty acid biosynthesis-related proteins markedly increased in IM1 (Figure S4A). This expression is negatively correlated with patient immunity (Figure S4B). As fatty acid metabolites can influence macrophage polarization and arachidonic acid along with other fatty acid metabolites can modulate immune cell functions and affect tumor immune responses, this may represent a potential therapeutic target for IM1 subtype patients [32].

Genomic features and potential drugs of the three immune subtypes

Recent Analyses have linked the tumor genomic landscape with tumor cytolytic activity, indicating that gene mutations can generate neoantigens that stimulate immune responses but may also lead to immune evasion through mechanisms such as overexpression of immune checkpoints, affecting tumor immune surveillance and treatment outcomes [33]. The associated genomic data available in the LUAD-CNHPP datasets allowed us to investigate the underlying genomic features.

Analysis of gene mutation characteristics based on three immune subtypes reveals that TP53 is the most frequently mutated gene in the IM3 subtype, while EGFR is the most common in both IM1 and IM2 subtypes (Fig. 5A). The mutations are predominantly Missense mutations, though EGFR also exhibits a notable frequency of in-frame deletions, particularly in the IM1 subtype (Figure S6A). Tumor driver gene analysis suggests that KRAS and EGFR are the primary driver gene mutations in lung adenocarcinoma (Figure S6B). Compared to the other subtypes, IM3 shows a significant frequency of MMP16 gene mutations, indicating distinct tumor invasion features (Figure S6C).

Fig. 5.

Genomic features and potential drugs of the three immune subtypes. A Genomic features among three immune subtypes. B Potential drugs among three immune subtypes depend on the proteomics data

One of the most important applications of molecular subtype is to stratify patients for precision medicine. To measure whether the three immune subtypes were sensitive to different drug profiles, we employed Connectivity Map (Cmap) to look for potential drugs that might be useful against each LUAD subtypes [34]. We found that different immune subtypes were sensitive to distinct compounds, further confirming the significance of molecular subtyping based on immune signatures (Fig. 5B).

Validation of three immune subtype characteristics using an independent proteomics cohort

To clarify the potential impact and generalizability of proteomics-based immune subtyping in enhancing precision medicine for patients with LUAD, we validated the tumor-associated characteristics of the three subtypes using another high-quality proteomics cohort (CPTAC-LUAD) [15]. Using the NTP method, we performed subtype classification of the CPTAC cohort based on the CNHPP dataset. After calculating immune scores with xCell [35], we observed a significant upregulation of key immune effector cells, such as CD8 + T cells, CD4 + T cells, and NK cells in the IM1 subtype (Fig. 6A). This further confirms the reliability and generalizability of our immune subtyping approach.

Fig. 6.

Immune landscape in an independent proteomic cohort. A Heatmap of immune cell scores across the three subtypes in the CPTAC cohort. B Correspondence between immune subtypes and CPTAC subtypes in the CPTAC cohort. C Immune scores of the immune subtypes in the CPTAC cohort. D Expression levels of the glycosaminoglycan biosynthesis pathway across the three immune subtypes in the CPTAC cohort

In the CPTAC cohort, tumors are categorized as cold or hot tumors by Gillette et al. Matching analysis revealed that all hot tumors identified in Gillette et al. analysis corresponded to the IM2 subtype(Fig. 6B). Additionally, immune scores calculated by Gillette showed that IM2 subtype lung adenocarcinoma patients had significantly higher immune scores compared to IM1 and IM3 subtypes (Fig. 6C). Furthermore, the glycosaminoglycan biosynthesis pathway showed relatively high activity in the IM3 subtype in the CPTAC cohort, although the level was comparable to that observed in IM1 (Fig. 6D).

This supports the broader applicability of our proteomics-based immune subtyping and provides valuable insights for understanding the immunological mechanisms and therapeutic strategies for patients with LUAD.

Discussion

It is crucial to explore the clinical characteristics and molecular subtypes of patients with LUAD to find appropriate treatment methods and improve disease prognosis [36]. Here, we identified three immune subtypes (IM1, IM2, IM3) based on the proteome of LUAD patients and analyzed the correlation between the immune microenvironment and tumor progression. We found that the IM2 subtype has the highest immune score and immune cell infiltration ratio, suggesting a potentially better response to immunotherapy. The metabolic characteristics of the IM1 and IM3 subtypes are distinct. The IM3 subtype has the worst prognosis and exhibits high GAG biosynthesis capacity. These findings hold promise in providing insights into the immune landscape of LUAD and guiding personalized treatment approaches for this disease.

We conducted a matching analysis between our immune subtypes and the subtypes identified by Xu et al. (XMS) [14]. Our subtypes are based on immune pathway scores from 103 LUAD tumor samples, while XMS subtypes are based on the top 25% most variable proteins. In XMS classification, the S-I subtype is considered a high-metabolic phenotype, S-II is a mixed phenotype, and S-III is a proliferative phenotype. Our IM3 subtype mainly matches XMS S-III subtype, which partially explains the poor prognosis observed in IM3 patients due to their biological functions. In terms of the immune landscape, XMS subtypes do not show significant differences among the three subtypes in immunological indicators such as antigen presentation and CD8A protein expression. Therefore, the IMS subtypes can better elucidate the immune landscape of LUAD patients. In XMS classification, the S-I subtype with high metabolic characteristics primarily matches our IM1 subtype. Our metabolic profiling also shows that the IM1 subtype has higher metabolic scores, including pathways related to fatty acid metabolism and glycolysis. Thus, improving the metabolic landscape of the IM1 subtype may be a potential therapeutic approach for treating IM1 patients.

To provide some possible future directions to improve immunotherapy efficacy in LUAD, we have identified differences in factors related to immunotherapy response across the immune subtypes, including immune cell infiltration status and T cell dysfunction, through immune scoring. Additionally, we used the TIDE score to predict the immunotherapy response of the three subtypes [37]. A lower TIDE score indicates a reduced risk of tumor immune escape and a higher likelihood of benefiting from immunotherapy. We found that patients with the IM2 subtype are more likely to benefit from immunotherapy. Therefore, immune scoring based on proteomics can help us more accurately identify patients suitable for immunotherapy, with potential clinical applications [38]. However, it is worth noting that these conclusions are primarily based on TIDE scores and need to be further validated with larger cohorts and prospective studies.

By analyzing the metabolic profile changes in LUAD patients with the three immune subtypes, we found that the GAG biosynthesis pathway is highly expressed in the IM3 subtype. In fact, GAG biosynthesis can influence the components of the tumor microenvironment by altering the extracellular matrix [39], further affecting the biological functions of the tumor. Through functional analysis, we found that the hypoxia pathway score is significantly elevated in IM3 patients, which may be related to the GAG biosynthesis pathway [40]. The alteration of this pathway induced by hypoxia can also impact the expression recognition of immune interaction molecules, thereby affecting the tumor immune response. Our analysis revealed that high glycosaminoglycan expression is negatively correlated with tumor immune response, which may explain the poor prognosis of IM3 LUAD patients.

In this study, we utilized proteomics data from Chinese LUAD patients to perform immune pathway scoring and immune subtyping. We not only elucidated the immune characteristics and prognostic correlations of the three subtypes but also comprehensively analyzed the metabolic profiles, genomic characteristics, and corresponding therapeutic recommendations for the three immune subtypes. This study provides insights for the precise treatment of LUAD patients.

Conclusions

This study revealed for the first time the the immune subtypes of LUAD patients based on proteomics and conducted an in-depth analysis of the prognosis, metabolic characteristics, and immunotherapy responses of these three immune subtypes. We found that a subset of LUAD patients with an active immune microenvironment are more likely to benefit from immunotherapy. In contrast, another subset of patients exhibited an immune-desert microenvironment characterized by hyperactive GAGs biosynthesis, high infiltration of immunosuppressive cells (exhausted T cells), and the poorest prognosis. Modulating the metabolic features of these patients may be a strategy to reverse their suppressive immune microenvironment. Our study provides insights into the interaction between immunity and metabolism at the proteomic level, offering valuable information for the precision treatment of lung adenocarcinoma patients.

Abbreviations

LUAD

Lung adenocarcinoma

NSCLC

Non-small cell lung cancer

ICB

Immune checkpoint blockade

ssGSEA

Single Sample Gene Set Enrichment Analysis

WGCNA

Weighted Gene Co-expression Network Analysis

TME

Tumor microenvironment

EMT

Epithelial-mesenchymal transition

IHC

Immunohistochemical

GO

Gene Ontology

GAGs

Glycosaminoglycans

ECM

Tumor extracellular matrix

Cmap

Connectivity Map

OS

Overall survival

DFS

Disease-free survival

MSigDB

Molecular Signatures Database

TIDE

Tumor Immune Dysfunction and Exclusion

NTP

Nearest template prediction

CPTAC

Clinical Proteomic Tumor Analysis Consortium

Author contributions

Yaru Wang conceived and designed the study, performed data analysis and interpretation, and drafted the manuscript. Miao Miao, Qingqing Wang, and Yuying Yin participated in preliminary analyses. Haijuan Zhao, Shuang Zhao, and Han Yang contributed to clinical information organization and result validation. Ting Xiao provided critical resources and professional guidance on proteomics and molecular oncology, and revised the manuscript for important intellectual content. Xin Wang supervised the overall project, contributed to study conception and design, and critically revised the manuscript.

Funding

This work was funded by the National High Level Hospital Clinical Research Funding (BJ-2023-088), the National Key R&D Program of China (No. 2022YFF0705004); the National Natural Science Foundation of China (NSFC 81172035, 82504178).

Data availability

The datasets analysed during the current study are publicly available. The quantitative proteomics data of the discovery cohort were obtained from the PRIDE database under the accession number PXD0001804000(41). The CPTAC lung adenocarcinoma proteomics dataset was downloaded from the CPTAC Data Portal (https://cptac-data-portal.georgetown.edu/cptac/s/S056)(42).

Declarations

Ethical approval and consent to participate

This study was reviewed and approved by the Ethics Committee of the Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College. The research was conducted in accordance with the Declaration of Helsinki and the guidelines of the approving ethics committee. Archived, anonymized residual tumor tissue samples collected during routine clinical care were used. This study utilized previously published data with prior ethical approval and informed consent [14], as well as publicly available datasets that are de-identified and open-access [15]. In addition, archived residual tumor tissue samples collected during routine clinical care were included. Where applicable, informed consent was obtained from all participants or their legal guardians.

Accordance statement

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.