Single-cell transcriptomics reveals ADRB1 as a key immune checkpoint regulating T cell exhaustion in lung cancer and its prognostic value in lung cancer

Abstract

Background

Lung cancer, LUAD, and LUSC are both among the leading causes of cancer-related deaths. T cell exhaustion within the tumor microenvironment compromises the immune response and limits the benefits of immune checkpoint therapy. Adrenergic signaling primarily through ADRB1 induces T cell exhaustion and immune suppression. ADRB1 + T cells have not yet been studied regarding their role in lung cancer progression or prognosis. scRNA-seq analysis for tumor and adjacent nontumor tissues from patients with LUAD and LUSC. Dimensionality reduction and clustering with Scanpy, followed by ssGSEA for adrenergic signaling activity. CNV was processed with inferCNV, and cell-to-cell communication was inferred via CellChat. Integration with TCGA bulk RNA-seq data enables prognostic modeling of ADRB1⁺ T cells. Kaplan-Meier survival analysis and univariate Cox regression were used to study the signature of ADRB1 + T cells for prognostic value.

Results

scRNA-seq data confirm the presence of exhausted markers in ADRB1 + T cells. The higher proportion in LUAD tumors as compared to LUSC was linked to more pronounced immune suppression and decreased T cell diversity. CellChat predicted extensive contacts between ADRB1 + T cells and macrophages, mainly through CCL and IFN-II signaling pathways. Inferred CNV evidenced evidence of malignant cells interacting with ADRB1 + T cells. In prognostic modeling based on ADRB1 + T cell signatures, high-risk populations were characterized by very poor overall survival, and ADRB1 was an independent prognostic factor in LUAD and LUSC.

Conclusion

This paper describes ADRB1 for the first time as a novel immune checkpoint involved in T cell exhaustion in lung cancer. Indeed, ADRB1 + T cells are promising biomarker candidates; they may even become therapeutic targets in lung cancer. Indeed, proposing a pathway by which ADRB1 may be targeted, anti-tumor immunity may be boosted and better patient prognosis secured. Thus, functional studies and clinical trials will have to be done to test and discuss this proposed concept under a new heading in therapeutic strategy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05112-y.

Background

Lung cancer, including lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), are one of the common types of cancers globally and one of the causes of death from cancer. Intensive morbidity and mortality result from lung cancer [1, 2]. Secondly, Despite advances in targeted therapy and immune checkpoint inhibitors, the prognosis for most patients with advanced lung cancer remains poor due to recurrence and treatment resistance [3]. T cell exhaustion has been one of the most recently longed-for phenomena associated with advanced cancers. It is the reduction of T cell function due to prolonged antigen stimulation [4–8]. T cell exhaustion is a new approach not yet used to enhancing anti-tumor immunity in cancer therapeutics [9–11]. Both forms of lung cancers, LUAD and LUSC, show common but variable degrees of T cell exhaustion, critical for their responses to the immunotherapy line [12]. Adrenaline signaling within the tumor microenvironment has quickly grown to be a very critical modulator of immune response [13]. It varies from norepinephrine, modulating the immune cells’ activity through adrenergic receptors. There are three main types of adrenergic receptors and subtypes: ADRB1, 2, and 3 [11, 14]. Underchronic stress conditions, elevated levels of norepinephrine in the tumor microenvironment would activate ADRB1, push T cells into exhaustion, and permit tumor immune evasion with the establishment of a permissive, immunosuppressive microenvironment [15]. Recent advancements about the complexity of adrenergic effects in modulating T cell functions across different cancer contexts sparked interest in the development of new therapeutics to overcome immune resistance in lung cancer [13].

Single-cell RNA sequencing revolutionized the study of tumor biology by providing unprecedented detail in dissecting intra-tumor heterogeneity and providing clues on how tumor-associated microenvironments shape tumor growth [16]. The technique enabled the once-in-dream possibility of studying dynamic interactions between tumor and immune cell populations to discover novel subtypes, which significantly advance tumor progression [17–19]. In this study, we aimed to investigate the involvement of ADRB1⁺ T cells in lung cancers LUAD and LUSC. For this, we performed in-depth scRNA-seq tumor and adjacent normal tissue analysis of the molecular mechanisms of ADRB1-mediated T cell exhaustion in the tumor microenvironment. It conducts the TCGA dataset bulk RNA-seq data integration in modeling the expression of ADRB1 + T cells for survival prognosis in patients, paving the way for subsequent personalized therapy.

Methods

Dimensionality reduction, clustering and differential expression analysis of scRNA-seq data

The cells are suspended in a viable solution and thus pass QC with cell numbers above 500 genes per cell or < 5% of the library size due to the expression of mitochondrial genes. The following pipeline used Scanpy (version 1.9.2) [20] for quality control, normalization, and clustering. Normalization within a single-cell setting using a global scaling method was performed with scanpy.pp.normalize_per_cell, and highly variable genes were calculated using scanpy.pp.highly_variable_genes. To eliminate the influence of cell cycle heterogeneity, cell cycle scores were calculated using predefined gene sets and regressed out during data scaling (scanpy.pp.regress_out). To correct for potential batch differences between samples, we applied BBKNN (Batch Balanced KNN) as an integration method following principal component analysis. After this, principal component analysis was run for dimensionality reduction, and UMAP was used for visualization. For clustering, the scanpy.tl.leiden function was used (resolution = 1). Each cell type was identified, and differential expression analysis between clusters, based on scanspy.tl.rank_genes_groups, was performed with adjusted p-values < 0.05.

Deconvolution-based cell-type feature score analysis

Cell-type–specific gene signatures derived from single-cell RNA-seq data were applied to TCGA bulk RNA-seq datasets using a deconvolution approach to estimate cell-type feature scores for each patient. Patients were stratified into high- and low-score groups based on the median value within each cohort. Kaplan–Meier survival curves were generated using overall survival (OS) and OS.time, and differences were assessed using the log-rank test.

Single-sample gene set enrichment analysis (ssGSEA)

The ‘GSVA’ R package, version 1.40.1 and the Molecular Signatures Database (MSigDB) gene sets, were used to assess the activity of the specific biological pathway Specifically, we used “Adrenergic receptor” as the keyword to search the MSigDB and selected three representative gene sets from the C5 (GO gene sets) category: GOBP_ADRENERGIC_RECEPTOR_SIGNALING_PATHWAY, GOMF_ADRENERGIC_RECEPTOR_ACTIVITY, and GOMF_ADRENERGIC_RECEPTOR_BINDING. These gene sets were chosen to comprehensively cover the signaling pathway, molecular function, and binding activity of adrenergic receptors. Pathway enrichment scores were heat mapped, and Spearman’s rank correlation with clinical variables was performed to identify possible positive correlations.

Copy number variation analysis using inferCNV

To assess copy number variations (CNVs) in single-cell RNA-seq data, we used the inferCNV package (version 1.18.0). Briefly, after filtering, the input single-cell expression matrix into inferCNV was info. By default, CNV analysis was performed while comparing the sample tumor to the reference normal of the same tissue type. The CNV score is then computed for each single cell, and cell type CNV heat maps are saved.

Cell-cell interaction analysis using CellChat

In this study, we use CellChat v1.5.0, an R package underpinned on a ligand-receptor pair database, to analyze cell-cell interaction on the basis of the CellChatDB. After QC and clustering, the CellChat pipeline was used to infer the cell interaction pattern across different cell types. Cell communication networks were thus aggregated based on the expression of the ligand-receptor pair. The strength of this communication across clusters was visualized as a directed graph and heatmap, which summarizes potential signaling pathways mediating interactions among tumor microenvironment cells.

Animal experiments

Animal experiments were conducted under an approved protocol (Affiliated Hospital of Qingdao University, Approval No. QYFYWZLL30218) and complied with the Basel Declaration and relevant regulations. Male BALB/c nude mice (6–8 weeks old, weighing 18–22 g) were obtained from GemPharmatech (Jiangsu, China) and housed under specific pathogen-free (SPF) conditions with a 12-hour light/dark cycle, constant temperature (22–25 °C), and humidity (40–60%), with free access to sterilized food and water. For tumor inoculation, 1 × 10⁶ A549 cells suspended in 100 µL PBS were subcutaneously injected into the right flank of each mouse. Mice were monitored at least once daily for general condition and tumor growth. Tumor size was measured with calipers and volume estimated as Institutional humane endpoints included ulceration, impaired mobility, or body weight loss > 15%. The institutional maximal tumor burden is 1.5 cm in diameter or 10% of body weight, and no animal exceeded these limits. Animals reaching endpoints were humanely euthanized according to approved procedures.

Immunoblotting

T cells were lysed in ice-cold RIPA lysis buffer supplemented with a cocktail of protease and phosphatase inhibitors (Roche) to ensure protein stability. Protein concentration was quantified using the bicinchoninic acid (BCA) method (Pierce, Thermo Scientific) according to the manufacturer’s instructions. Equal amounts of total protein (30 µg per lane) were loaded onto a 10% SDS-polyacrylamide gel and resolved by electrophoresis. Proteins were transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore) using a semi-dry transfer system. Membranes were incubated with 5% non-fat milk in TBST (TBS with 0.1% Tween-20) for 1 h at room temperature to block nonspecific binding. Blots were then incubated overnight at 4 °C with ADRB1 antibodies (ab85037) diluted 1:1000 in blocking buffer. After washing, membranes were incubated with appropriate HRP-conjugated secondary antibodies (1:5000 dilution) for 1 h at room temperature. Protein signals were detected using enhanced chemiluminescence (ECL) reagent (ECL Prime, GE Healthcare), and band intensities were quantified using ImageJ software (NIH). GAPDH was used as a loading control to normalize protein expression.

RNA extraction and qRT-PCR

Total RNA was isolated from T cells using the Universal RNA Extraction Kit (EZBioscience, EZB-RN4) following the manufacturer’s protocol. For reverse transcription, 400 ng of purified RNA was converted into complementary DNA (cDNA) using the PrimeScript RT Master Mix (TaKaRa, RR036A). Subsequently, 100 ng of cDNA was used as a template for quantitative PCR, which was carried out using PerfectStart Green qPCR SuperMix (TransGen, AQ601-04) on a CFX96 Real-Time PCR System (Bio-Rad). Gene expression levels were calculated using the ΔΔCt method, with Tbp serving as the internal control for normalization.

Murine T cell culture and treatment

Spleens from male C57BL/6 mice (6–8 weeks old) were collected and gently mashed in cold PBS to prepare a single-cell suspension. Red blood cells were removed using ACK lysis buffer (Yeasen). The resulting splenocytes were resuspended in T cell culture medium (TCM) and transferred to CD3/CD28 pre-coated plates for stimulation. TCM consisted of RPMI-1640 supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 0.1 mg/mL streptomycin, 10 mM HEPES, and 1 mM sodium pyruvate. Murine IL-2 (10 ng/mL) was added to promote T cell activation. After incubation at 37 °C for 2 days, propranolol (1 µM, MCE, HY-B0573B) was added to the culture. C57BL/6 mice were purchased from GemPharmatech (Jiangsu, P. R. China). Animal experiments were approved by the Medical Ethics Committee of the Affiliated Hospital of Qingdao University (Approval Number: QYFYWZLL30218).

Running deconvolution on LUAD and LUSC samples using CIBERSORT

Running Deconvolution on CIBERSORT to Estimate Immune Composition to Deconvolute Tumor Samples According to Cibersort, Later Inferring Bulk RNA-seq Data to Estimate Proportions of Immune Cell Types, the Cibersort Algorithm Run. First, the feature matrix corresponding to a single-cell expression was used, followed by deconvolution to TCGA LUAD and LUSC expression matrices. The scores were extracted for further analysis.

Statistical analysis

All statistical analyses were performed using R version 4.3.0 software. A p-value < 0.05 was considered statistically significant unless otherwise specified. Data Visualization was done using the ggplot2 package in R.

Results

Single-cell landscape of ADRB1 + T cells in lung cancer

To systematically map adrenergic signaling in the NSCLC ecosystems, we integrated scRNA-seq data for 52 fresh surgical resections of lung tumors and adjacent tissues (32 LUAD and 20 LUSC) [21]. Uniform manifold approximation (UMAP) resolved 15 major cellular compartments (Fig. 1a), including immune cells (B, B cells; Plasma; Mo: Monocyte; Gran, Granulocyte; DC, Dendritic cells; Plasma DC; Mast; CD4, CD4 + T cells; CD8, CD8 + T cells; Treg; ExCD8_T, CD8⁺ exhaust T cell; NK, Natural killer cell; NK_prolif, Natural killer cell proliferation), stromal cells (EC, Endothelial cells; Fib, Fibroblasts, Basal), and parenchymal cells (ATII, alveolar type II cells; Club, club cells; Cilia, ciliated cells). Figure 1b illustrates the single-cell landscape of adrenergic receptor expression in LUAD and LUSC tumor tissues. The results showed that genes related to adrenergic receptors were expressed in multiple cells. Marker gene analysis identified the characteristic genes of each cell population (Fig. 1c).

Fig. 1.

Single-cell landscape of adrenergic receptor in lung cancer. a Uniform manifold approximation and projection (UMAP) visualization of annotated cells in cell atlas of lung tumors. B, B cells; DC, Dendritic cells; Gran, Granulocyte; Mo, Monocyte; EC, Endothelial cells; Club, club cells; Cilia, ciliated cells; ATII, alveolar type II cells; CD4, CD4 + T cells; CD8, CD8 + T cells; NK, Natural killer cell; Fib, Fibroblasts. b UMAP plots showing expression of adrenal receptor pathway-related genes. c Dot plot showing marker genes for lung cancer cell populations. The size of the dot represents the proportion of cells expressing a gene. Color denotes the scaled ex-pression level. d The bar chart shows the proportion of different tissue types in each cell population. e The bar chart shows the proportion of patients with different stages in each cell population

The sample distribution map showed a striking result: LUAD had significantly more medium monocytes, dendritic cells, and CD8 + exhausted T cells than LUSC, while other immune cells, such as mast cells and Tregs, did the opposite. LUAD patients also had a much higher number of endothelial cells in stromal cells than LUSC patients, but Basal cells were found almost exclusively in LUSC patients. In addition, LUAD patients appear to have a higher proportion of parenchymal cells, especially in para-cancerous tissues (Fig. 1d). In the restaging distribution map, NK cells, monocytes, mast cells and dendritic cells accounted for nearly 50% of stage I patients, and CD8 + exhausted T cells accounted for a relatively high proportion of stage II patients. In early-stage patients, the proportion of parenchymal cells is usually higher, while stromal cells show heterogeneity (Fig. 1e).

Expression of exhaustion markers in ADRB1 + T cells

In order to investigate the differences in the tumor microenvironments of various types of lung cancer, we identify immune and stromal cell sets. The It was reported in the previous article that ADRB1 is a novel immune checkpoint in colon cancer. However, our analysis of immune cell subtypes identified six T cell subtypes, one of which is ADRB1 + T cells; therefore, this population of depletion-associated T cells also appears to have infiltrated lung cancer tissue (Fig. 2a). In addition, we also carried out further subtype analysis of the mechanism cells (Fig. 2b).

Fig. 2.

Microenvironment of lung tumor. a UMAP plot of immune cell subtypes annotation in lung tumors. Treg, regulatory T cell; Temra, Temra cytotoxic T cell; ExCD8 + T, Exausted CD8 + T cell; AlvMs, Alveolar macrophages; Gran, Granulocytes; DC, Dendritic Cells; plasma_DC, Plasmacytoid DCs. b UMAP plot of stroma cell subtypes annotation in lung tumors. acAFs, activated alveolar fibroblasts; CAF, cancer associated fibroblast; EC, endothelial cell; ArtEC, arterial vascular endothelial cell; VeinEC, venous vascular endothelial cell; EMT, Epithelial-Mesenchymal Transition c Bar chart showing the Jensen–Shannon divergence scores of each immune cell subtypes. d Bar chart showing the Jensen–Shannon divergence scores of each stroma cell subtypes. e Bar chart showing the percentage of immune and stromal cells in different cell cycle stages among different tissue. f Dot plot indicating the downregulation of ADRB1 ex-pression in LUAD tissue. g The violin plot shows the enrichment degree of exhausted-related genes in ADRB1 + T cells in cancer tissues and adjacent tissues of lung adenocarcinoma and lung squamous cell carcinoma

In order to detect the degree of cell dispersion in cancer tissue and para-cancer tissue of different lung cancer types, we calculated the JSD scores of different cell types in LUAD and LUSC cancer and para-cancer tissue, respectively. The results showed that JSD scores of almost all immune cells and stromal cells were decreased in LUAD patients’ cancerous tissues compared with adjacent tissues, indicating that most types of cells were less diffuse in cancerous tissues, reduced intracellular information entropy, and reduced cell heterogeneity. In contrast, ADRB1 + T cells and B cells showed increased intracellular entropy and diversity in cancer tissues (Fig. 2c). In LUSC patients, the JSD scores of ADRB1 + T cells in cancerous tissues were comparable to those in para-cancerous tissues. The results in most other immune cells and stromal cells were similar to those in LUAD (Fig. 2d). Subsequently, we examined cell proliferation in LUAD and LUSC patients. The analysis showed that the proportion of S stage and G2M stage in LUAD patients was higher than that in paracancer tissue, while the proliferation of cancer tissue and paracancer tissue in LUSC patients was basically the same (Fig. 2e). In addition, we detected the expression changes of ADRB1 gene in ADRB1 + T cells in cancer tissue and paracancer tissue in LUAD and LUSC patients respectively, and the results showed that the expression of adRB1 gene in cancer tissue was significantly lower than that in paracancer tissue in LUAD patients, while the difference was not significant in LUSC patients (Fig. 2f). These results suggest the existence of an exhausted T cell subpopulation characterized by the expression of the adrenergic receptor ADRB1 in LUAD. In cancer tissue, these cells express more adrenergic receptor-related genes, making cellular immune function decline. Subsequently, we compared the exhaustion scores of ADRB1⁺ T cells between tumor and adjacent normal tissues in both LUAD and LUSC. We calculated an exhaustion score for each ADRB1⁺ T cell using a set of established exhaustion-related genes, including PDCD1, LAG3, HAVCR2 (TIM-3), CTLA4, TOX, TIGIT, and ENTPD1 (CD39). The results, presented in the Fig. 2g, demonstrate that ADRB1⁺ T cells in LUAD tumor tissues exhibit significantly higher exhaustion scores compared to those in adjacent normal tissues (Wilcoxon test, p < 0.001). In contrast, no significant difference was observed in LUSC (p > 0.05), which is consistent with the ADRB1 expression pattern shown in Fig. 2f.

Immune cell composition in the TME post-treatment

To further characterize parenchymal cell populations, we applied the inferCNV algorithm to detect large-scale chromosomal CNVs across all epithelial cells, using immune and stromal cells as internal reference populations representing genetically stable, non-neoplastic cell types. The resulting heatmap displays the relative expression intensities across individual chromosomes, revealing extensive chromosomal amplifications and deletions within a subset of epithelial cells, indicative of significant genomic alterations compared to reference normal cells (Fig. 3a).

Fig. 3.

Identification and analysis of malignant cells. a Heatmap displaying large-scale segmental copy number variations across the genome for each cell population. b UMAP plot of parenchymal cell subtypes annotation in lung tumor. c UMAP plot showing cell cycle distribution of epithelial and malignant cells. d Dot plot showing copy number scores for each cell sub-group. The size of the dot represents the proportion of cells with copy number variation (CNV) score. Color denotes the level of the CNV score. e Bar chart showing the percentage of parenchymal cell in different cell cycle stages among different tissue. f The box plot shows the scores of CD44-labeled tumor stemness in the cancer tissues and adjacent tissues of lung adenocarcinoma and lung squamous cell carcinoma. g The violin plot shows the enrichment fraction of characteristic genes of dry tumor cells in CD44 + and CD44- malignant cells

Based on CNV profiles, epithelial cells were effectively distinguished from other parenchymal cell types. Furthermore, by evaluating the expression levels of the CD44 gene, a well-established marker of cancer stem-like cells, we successfully identified a population of malignant cells with stem-like properties. Cell cycle analysis demonstrated that a subset of these malignant cells was enriched in the G2/M phase, reflecting active proliferation (Fig. 3c). Quantitative analysis of CNV burden revealed that malignant cells exhibited the highest copy number scores among parenchymal cell populations (Fig. 3d). Notably, comparative analysis showed that cancerous tissues from patients with LUSC harbored a higher proportion of G2/M phase cells relative to adjacent non-tumor tissues (Fig. 3e). In parallel, assessment of stemness scores revealed significantly elevated stem-like features in cancerous tissues from both LUAD and LUSC compared to their corresponding adjacent normal tissues (Fig. 3f). To further validate the stem-like characteristics of CD44-positive malignant cells, we compared the expression of established stemness-associated markers, including SOX2, ALDH1A1, PROM1, and ITGA6, between CD44-positive and CD44-negative malignant cells. CD44-positive malignant cells exhibited significantly elevated expression of these stemness-related genes (Fig. 3g), supporting their stem-like transcriptional phenotype.

Cellchat analysis among all cell subtypes

To better understand such a complex network of interactions among different cell types and complicated mechanisms of intercellular signaling, we started with an analysis and visualization of the communication patterns using CellChat. That certainly was the first step toward constructing a network of interactions among different cell subtypes. To illustrate the extent of cellular communications, we initially mapped the intercellular connections and then visualized these using lines of different thicknesses connecting various clusters in the Fig. (Fig. 4a). For the subtle contributions of different cell subtypes to the input/output of intercellular communications, all the signals going to and coming from each cell class were illustrated in a dot plot indicating the strength of these signals (Fig. 4b).

Fig. 4.

Characterization of cellular interaction networks in lung tumors. a Circular plot depicting the overall number of predicted intercellular interactions among all cell types within the lung tumor microenvironment. b Scatter plot illustrating the cell–cell communication landscape between all cell types and the ADRB1⁺ T cell subset. Dot color represents different cell types, while dot size reflects the relative abundance of each cell population. c Circular plot highlighting the interaction network with ADRB1⁺ T cells as the source population and tumor cells as the target. The width of the connecting lines corresponds to the predicted interaction strength. d Circular plot illustrating the strength of interactions between ADRB1⁺ T cells and other tumor cell populations specifically through the CCL signaling pathway. e Western blot analysis confirming the effect of propranolol on ADRB1 expression in CD8 + T cells isolated from murine spleens (51 kDa for ADRB1 protein, 35 kDa for GAPDH as loading control). f Quantitative PCR (qPCR) results demonstrating significant downregulation of Ccl3, Ccl4, Ccl5, Ccr1, Ccr3, and Ccr5 transcript levels following Adrb1 knockdown. g Circular plot showing the predicted interaction strength between ADRB1⁺ T cells and other tumor cells mediated via the Type II interferon (IFN-II) signaling pathway. h Heatmap illustrating the intercellular interaction patterns between distinct cell populations within the IFN-II signaling pathway, with interaction strength represented by color intensity

Further interactions between ADRB1 + T cells and other cell subtypes, along with the T cells expressing immune checkpoint markers, were summarized, attending to almost every other cell type (Fig. 4c). These highlighted the enormous impacts that ADRB1⁺ T cells have on the composition and functionality of cell population changes. It was possible to relative characterize ADRB1⁺ T cells along with critical pathways mapping the presentation of antigens and the activity of cytotoxic T cells. ADRB1⁺ T cells mainly contributed to the communication with SPP1⁺ macrophages through CCL, which is responsible for the recruitment of immune cells to tissue specific sites (Fig. 4d).

To verify the regulatory role of ADRB1 in chemokine expression in T cells, we used propranolol (1 µM) to pharmacologically inhibit ADRB1. Western blot analysis confirmed a time-dependent decrease in ADRB1 protein expression from 0.5 to 48 h post-treatment, demonstrating effective suppression by propranolol (Fig. 4e). We then examined how ADRB1 inhibition affected chemokine transcription. Quantitative PCR analysis showed that propranolol treatment significantly reduced the mRNA levels of key chemokines, including Ccl3, Ccl4, and Ccl5, as well as their corresponding receptors Ccr1, Ccr3, and Ccr5 (Fig. 4f). These results suggest that ADRB1 expression in T cells is required for maintaining the transcriptional program of chemokines and their receptors, thereby potentially influencing T cell-mediated crosstalk within the tumor microenvironment. Besides that, the ADRB1 + T cells act here as ligand cells in the IFN-II signaling pathway, whereas the responding cells are Temra, Gran, and DC (Figs. 4g-h).

Prognostic analysis of gene expression features in single cell transcriptome

In the context of cancer progression, SPP1⁺ macrophages and stem-like malignant cells have been increasingly recognized as key contributors to adverse patient outcomes. To systematically evaluate the prognostic significance of these cellular populations and elucidate the role of the adrenergic receptor signaling pathway in LUAD and LUSC we performed transcriptomic deconvolution analyses. Specifically, we utilized the scRNA-seq expression profiles of ADRB1⁺ T cells, CD44⁺ malignant cells, and SPP1⁺ macrophages as reference feature matrices and integrated them with bulk RNA-seq datasets from LUAD and LUSC patients within TCGA cohort (Fig. 5a). Based on the median deconvolution scores for each cell type, patients were stratified into high- and low-score groups. The results demonstrated that elevated abundance of these cellular populations is significantly associated with patient prognosis in both LUAD and LUSC, underscoring their potential as independent prognostic biomarkers (Fig. 5b–i).

Fig. 5.

Deconvolution-based prognostic analysis of the TCGA lung cancer cohort integrating scRNA-seq data. a Schematic illustration of the deconvolution analysis workflow, integrating scRNA-seq profiles from lung cancer with bulk RNA-seq data from the TCGA cohort. b–e Kaplan–Meier survival analyses demonstrating that the abundance of ADRB1⁺ T cells, SPP1⁺ macrophages, and CD44⁺ parenchymal cells is significantly associated with overall survival in patients with LUAD. f–i Kaplan–Meier survival analyses confirming the significant prognostic impact of ADRB1⁺ T cells, SPP1⁺ macrophages, and CD44⁺ parenchymal cells in the LUSC cohort. j Boxplots showing the distribution of risk scores across different clinical subgroups within the TCGA cohort, based on deconvolution analysis. *P < 0.05; NS, not statistically significant

Further subgroup analyses revealed that ADRB1⁺ T cell deconvolution scores varied significantly across different clinical stages of LUAD. Notably, higher deconvolution scores were observed in patients with more advanced disease stages, suggesting that the presence of these T cells is not only linked to prognosis but may also reflect disease progression and tumor aggressiveness (Fig. 5j). Although LUAD and LUSC exhibited distinct receptor expression patterns and cellular characteristics, survival analysis based on cell-type feature scores did not reveal consistent prognostic differences between the two cancer types. These findings suggest that LUAD and LUSC may utilize distinct molecular programs that ultimately converge toward comparable clinical outcomes.

adrenergic receptor characterization construction prognosis and nomogram

To elucidate the prognostic relevance of ADRB1⁺ T cells in LUAD and LUSC patients, we intersected the gene signatures of ADRB1⁺ T cells with genes involved in the adrenergic receptor signaling pathway. Subsequently, a single-sample gene set enrichment analysis (ssGSEA) was performed to calculate individual risk scores for patients within the TCGA-LUAD and TCGA-LUSC cohorts. Patients were stratified into high-risk and low-risk groups based on the median risk score threshold. Kaplan–Meier survival analyses revealed a statistically significant difference in overall survival between high- and low-risk groups for both LUAD and LUSC, highlighting the clinical prognostic value of this risk model (Fig. 6a, b). Furthermore, univariate and multivariate Cox proportional hazards regression analyses confirmed that the ADRB1-based risk score serves as an independent prognostic factor for overall survival in both LUAD and LUSC (P < 0.05) (supFig. 1, 2).

Fig. 6.

Construction of prognostic model of related features of adrenergic receptor signaling pathway. a Kaplan-Meier analysis confirmed the prognostic significance of the adrenergic receptor signaling pathway-related features in the LUAD cohort. b Kaplan-Meier analysis confirmed the prognostic significance of the adrenergic receptor signaling pathway-related features in the LUSC cohort. c The nomogram prediction model for LUAD patients’ overall survival

To further enhance individualized prognostic assessment, we constructed nomogram models incorporating multiple clinical parameters for LUAD and LUSC patients, respectively (Fig. 6c, supFig. 4a). These models demonstrated that the ADRB1-based risk score exerted the most pronounced influence on patient prognosis. The 1-, 3-, and 5-year calibration curves showed excellent agreement between the predicted survival probabilities and actual outcomes, indicating robust predictive accuracy of the nomograms. Moreover, decision curve analyses (DCA) at 1, 3, and 5 years revealed substantial net clinical benefit, underscoring the strong translational potential and practical applicability of the ADRB1-based prognostic models (supFig. 3, supFig. 4b-c).

Discussion

Our study, therefore, demonstrates strong evidence that the role of ADRB1 in mediating immune suppression within the lung cancer via T cell exhaustion is significant. We identified an LUAD-enriched population of ADRB1 + exhausted T cells, characterized by a drastic depletion in T cell diversity and an increase in potentially immunosuppressive interactions in the tumor microenvironment. This enters more profoundly into the already well-structured persistence connection between adrenergic signaling activating the β-adrenergic receptor over time, depending on stress, and the fueling of immune escape mechanisms during tumor development [22].

Cell-to-cell interaction analysis performed by us showed how significant the cross-talk between T cells, including myriad subtypes, was primarily brought about through the CCL and IFN-II signaling pathways with macrophages, including SPP1 + macrophages. Indeed, all this recently published evidence shows that this β-adrenergic signaling modifies macrophage polarization, resulting in the formation of an immune-suppressing TME that facilitates tumor development [23]. They must be regarded as some of the immune escape mechanisms at play in lung cancer, thus pointing to ADRB1 therapeutic potential. While our CellChat analysis and preliminary experiments suggest that ADRB1 regulates CCL chemokine expression in T cells, the precise molecular mechanisms by which CCL and IFN-II signaling pathways mediate ADRB1-driven T cell exhaustion remain to be fully elucidated. Future studies employing co-culture systems with ADRB1-manipulated T cells and macrophages, along with chemokine blocking experiments, will be necessary to clarify the upstream and downstream regulatory relationships within these pathways.

An intriguing observation in our study is the differential expression of ADRB1 in ADRB1⁺ T cells between LUAD and LUSC tumor tissues, with significantly lower expression in LUAD tumor tissues compared to adjacent normal tissues, while no significant difference was observed in LUSC (Fig. 2f). This phenomenon likely stems from subtype-specific differences in the tumor microenvironment. First, LUAD tumors harbor a higher proportion of CD8⁺ exhausted T cells and distinct immune composition (Figs. 1d and 2c-d), creating a unique cytokine milieu that may downregulate ADRB1 expression as a feedback mechanism to prevent overactivation. Second, our CellChat analysis revealed stronger interactions between ADRB1⁺ T cells and SPP1⁺ macrophages through CCL and IFN-II signaling pathways in LUAD (Fig. 4d and g-h), and macrophage-derived factors may modulate adrenergic receptor expression on T cells. Third, potential differences in local adrenergic ligand availability between the two subtypes could lead to ligand-induced receptor internalization. These observations highlight the complexity of adrenergic signaling regulation within the tumor microenvironment and underscore the need for subtype-specific considerations in developing ADRB1-targeted therapeutic strategies.

While ADRB1-targeted therapy holds promise for enhancing anti-tumor immunity, several challenges must be addressed before clinical translation. The widespread expression of adrenergic receptors in normal tissues, particularly in the cardiovascular system, raises concerns about potential off-target effects and toxicity associated with systemic ADRB1 blockade. Additionally, chronic inhibition of ADRB1 may lead to compensatory upregulation of other adrenergic receptor subtypes or activation of alternative signaling pathways, potentially resulting in acquired resistance. The subtype-specific differences in ADRB1 expression observed in our study further suggest that patient selection based on tumor histology and ADRB1 expression levels may be necessary to maximize therapeutic benefit.

Combining ADRB1 inhibitors with existing immune checkpoint inhibitors represents a particularly promising strategy. Adrenergic signaling has been implicated in promoting T cell exhaustion and impairing responses to PD-1/PD-L1 blockade. Therefore, ADRB1 inhibition may synergize with checkpoint inhibitors by reversing T cell exhaustion and restoring the efficacy of immunotherapy. Our finding that ADRB1⁺ T cells communicate with macrophages through CCL and IFN-II signaling pathways further supports this rationale, as these pathways are known to modulate the tumor immune microenvironment and influence checkpoint blockade responses. However, the optimal combination regimen, including dosing schedules and patient selection criteria, requires systematic evaluation in future preclinical and clinical studies.

Of enormous relevance are those evident in clinical practice. This study contrives a prognostic model wherein ADRB1 + T cell signatures cannot be valued as preferential survival predictors; thereby, high-risk patients undergo highly lamentable consequence conditions. This may be noted since adrenergic receptor blockade is capable of improving anti-tumor immunity via the dismantling of norepinephrine-driven immune suppression [24]. From these observations, it follows that β-blockers or ADRB1-specific inhibitors might be adjunctive agents enhancing the benefits derived from immune checkpoint inhibitors in the treatment of lung cancer.

This study has several limitations. First, our scRNA-seq data were derived only from surgically resected primary tumors, lacking samples from metastatic lung cancer. Second, the prognostic model was developed solely within the TCGA cohort and has not been validated in external independent datasets. Third, functional validation of ADRB1-mediated T cell exhaustion through in vitro and in vivo experiments is still needed. Fourth, the systemic effects of adrenergic signaling modulation, including potential cardiovascular impacts, remain poorly understood. Future studies incorporating metastatic samples, external cohort validation, and functional experiments will be necessary to address these limitations and further validate our findings.

Conclusions

To conclude, in our investigation, ADRB1 emerges as a novel immune checkpoint in T cell exhaustion in lung cancer. In the integration of single-cell and bulk transcriptomic data, we indeed recapitulated the presence of an ADRB1 + T cell subtype involved in immune suppression and linked with reduced patients’ survival. Translating these findings into human lung cancer, blockade of ADRB1-related pathways could represent a novel therapeutic approach to resurfacing anti-tumor immunity in lung cancer. Further studies will confirm these findings and their relevance to adjunct immunotherapy through ADRB1 inhibition.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

XY, HX conceptualized, supervised for the study. ZC, FS and XD performed the experiments and analyzed the data. QL, HX, PS, ZX and SW modified the manuscript. XY and HX are co-corresponding authors, and ZC and FS are co-first authors. All authors read and approved the final manuscript.

Funding

Not applicable.

Data availability

The single-cell transcriptomic data analysed during the current study are available in the chenlulab repository. TCGA data were obtained from bulk RNA-seq and matched clinicopathological characteristics of lung cancer patients (Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Carcinoma (LUSC)) https://portal.gdc.cancer.gov/.

Declarations

Ethics approval

All animal experiments were performed in accordance with institutional, national, and international guidelines, including the Basel Declaration, and were approved by the Medical Ethics Committee of the Affiliated Hospital of Qingdao University (Approval No. QYFYWZLL30218). Mice used in this study were obtained from GemPharmatech (Jiangsu, P. R. China). Tumor monitoring and humane endpoints were applied throughout, and the institutional maximal tumor burden (1.5 cm in diameter or 10% of body weight) was not exceeded.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.