A public data-based molecular classification of small cell lung cancer by neuroactive signaling networks unveils distinct microenvironment landscapes and immunotherapy-related prognostic biomarkers

Introduction

Small cell lung cancer (SCLC), an aggressive neuroendocrine tumor accounting for 15% of all lung cancer cases, demonstrates remarkable intertumoral heterogeneity with distinct molecular subtypes defined by master transcriptional regulators: ASCL1 (SCLC-A), NEUROD1 (SCLC-N), POU2F3 (SCLC-P) and YAP (SCLC-Y) (1,2). This classification framework has revolutionized our understanding of SCLC biology, yet critical knowledge gaps persist regarding the microenvironmental determinants driving its characteristic rapid metastasis and therapeutic resistance.

Emerging evidence highlights the critical role of neurobiology in tumorigenesis, particularly through bidirectional crosstalk between cancer cells and the peripheral nervous system (PNS). The autonomic and sensory nerve fibers comprising the PNS establish functional synapses with neoplastic cells across multiple malignancies, such as gastric (3), breast (4,5), prostate (6), and lung cancers (7,8). Signaling transduction between neuron and other cells is mostly based on ligand and receptors, neurotransmitter, and neuropeptide (9-11). Notably, the sources of neuroactive signaling are not only neurons, but tumor cells and immune cells. Tumor cells and cells in tumor microenvironment (TME) express receptors of neurotransmitter, which indicates a potential for functional neuroactive signaling pathways. Such interactions may influence immune evasion, tumor progression, or therapeutic resistance (11,12). These findings highlight the underappreciated role of neurobiology in modulating the progression of SCLC.

Despite these advances, the neurobiological landscape of SCLC remains paradoxically underexplored. While SCLC’s neuroendocrine origin suggests inherent neuroactive signaling capacity, systematic characterization of neuroactive signaling networks across molecular subtypes is lacking. Key unanswered questions persist: (I) How do neuroactive signaling patterns vary between SCLC subtypes? (II) What is the spatial organization of neuroactive signaling components within tumor ecosystems? (III) Can neuroactive signaling signatures predict clinical outcomes or therapeutic vulnerabilities?

To address these knowledge gaps, we performed integrated multi-omics analysis of 79 treatment-naïve SCLC specimens, complemented by single-cell RNA sequencing (scRNA-seq) of 11 independent SCLC tumors. We identified SCLC tumors in three major clusters based on the expression of neuroactive signaling genes. This work provides the preliminary comprehensive atlas of neuroactive signaling networks in SCLC, offering new avenues for therapeutic targeting and precision medicine approaches. We present this article in accordance with the STREGA reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-620/rc).

Methods

Data source

Bulk-sequencing transcriptomic data were obtained from corresponding publications or Gene Expression Omnibus (GEO) (Table S1), including 79 primary SCLC tumors from George’s study (George’s Cohort), 65 tumors and 7 normal tissues from GSE60052 (Jiang’s Cohort), 8 tumors and 14 normal tissues from GSE40275 (Kastner’s cohort) (13-15), 206 cell lines from GSE32036. And for immunotherapy validation cohort, we collected and integrated tumor sequencing data from 21 SCLC patients included in our previously published study (16). All these patients subsequently received 1^st-line chemoimmunotherapy combined with anti-angiogenic therapy. Single-cell transcriptome data of 11 SCLC tumors were collected from Tang et al, comprising 5025 cells (17).

Neuroactive-signaling-related genes (NRGs) used for analysis

Totally, 492 NRGs were retrieved from gene set enrichment analysis (GSEA) database. The details of the gene list are shown in Table S2.

Consensus molecular clustering by “ConsensusClusterPlus”

ConsensusClusterPlus R package was used to classify SCLC patients based on the expression of NRGs in primary tumor. The final number of clusters was set to five, and these were defined as neuroactive signaling-related molecular clusters (NRCs).

Gene set variation analysis (GSVA) and single-sample gene set enrichment (ssGSEA) analysis

The features of biological processes in different clusters were investigated by GSVA enrichment analysis. The geneset, Hallmark of Cancer, was downloaded from MSigDB database (https://www.gsea-msigdb.org/gsea/index.jsp, accessed on 2024). The signature of cancer stemness was collected from published research (18). The stemness score was calculated using ssGSEA analysis (19,20).

TME infiltration evaluation

The XCELL algorithms were applied to assess tumor-infiltrating immune cell subgroups and neurons among NRCs (21). To confirm the stable TME infiltration patterns of clusters, we also utilized immune cell infiltration scores with the cell type signature from another study using ssGSEA analysis (20).

Genomic analyses

The somatic mutation data of George’s cohort were downloaded from the corresponding publication. We mapped top 30 mutated genes in NRCs and acquired TMB through ‘maftool’ package.

Drug susceptibility analysis

In the interest of the differences in the therapeutic effects of drugs of SCLC patients, the drug imputed sensitivity score of drugs was calculated using the “oncoPredict” package based on Sanger’s Genomics of Drug Sensitivity in Cancer (GDSC) v2 (22). The TIDE algorithm was applied to predict the response of immunotherapy (23).

Weighted gene co-expression network analysis (WGCNA)

Genes upregulated in SCLC from Kastner’s cohort and Jiang’s cohort were obtained, and clustering was performed on these genes using the R package “WGCNA”. Outlier samples were removed, and feature gene modules (module membership, MM) were selected. Pearson correlation was applied to cluster the samples and construct a scale-free network. The pickSoftThreshold function was used to choose an appropriate soft threshold. Based on the selected soft threshold, genes were grouped into modules containing at least 20 genes, and modules with a correlation lower than 0.3 were merged using dynamic tree cutting, dividing the gene set into multiple modules. Subsequently, the correlation between each module and different NRCs was evaluated. Gene modules significantly correlated with NRC3 were selected.

Single-cell RNA sequencing data analysis

The R package “Seurat” (4.1.0) was utilized to performed cell clustering and annotation. DoubleFinder R package (v2.0.3) was used to identify doublet events in scRNA-seq data. Highly variable genes were selected and used to conduct principal components analysis (PCA) dimension reduction. We defined the identity of each cell cluster according to the expression of well-known cell markers: tumor cells (EPCAM, KRT18), normal epithelial (EPCAM, KRT18, ALDH1A1), T/NK cells (CD3D, CD3E), B/Plasma cells (CD38, MS4A1, SDC1), myeloid cells (C1QA, CD14, CD68), fibroblasts (DCN, C1R, C1S), mast cells (CPA3, KIT). To estimate NRS level in scRNA-seq data, pseudo-bulk RNA-seq data were generated by aggregating single-cell transcriptomic counts from sample. Monocle2 R package was used to estimate the cellular trajectory to discover the state transition of malignant cells (24). Copy number variations (CNVs) of tumor cells were inferred from scRNA-seq data using InferCNV (v1.4.0; https://github.com/ broad institute/inferCNV).

Cell-cell communication analysis

The cell-cell communication network was inferred from scRNA-seq data using the CellChat R package (v1.1.1) according to canonical ligand-receptor pairs (25).

Cell culture

Human SCLC cell lines (H446, H146, H1688, H526), human normal bronchial epithelial cell lines (16HBE, BEAS-2B), human lung adenocarcinoma cell line (A549), and human large cell lung cancer cell line (H460) were purchased from the American Type Culture Collection (ATCC, RRID: CVCL_1562 for H446, CVCL_1473 for H146, CVCL_1487 for H1688, CVCL_1569 for H526, CVCL_0021 for 16HBE, CVCL_0168 for BEAS-2B, CVCL_0023 for A549, and CVCL_0459 for H460). Mouse SCLC cell line (SCLC-1) was purchased from National Collection of Authenticated Cell Cultures. The cell lines were authenticated by STR profiling and karyotyping upon initial receipt, and were tested negative for mycoplasma using a polymerase chain reaction (PCR)-based detection method. All cell lines were maintained in either RPMI-1640 medium or DMEM (Thermo Fisher Scientific, Waltham, MA, USA) medium supplemented with 10% fetal bovine serum. Cells were cultured at 37 ℃ in a humidified atmosphere with 5% CO₂.

Drugs

The drug was used at the indicated concentrated: CNP-38 10 µM (MCE, Cat# HY-P5127).

Histochemistry staining

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of The First Affiliated Hospital of Guangzhou Medical University (Guangzhou, China, No. 2020-189) and informed consent was obtained from all individual participants.

The sections of paraffin embedded tumorous tissue of SCLC patients (n=3) were deparaffinised and stained with hematoxylin and eosin (H&E). And for immunohistochemistry staining, deparaffinized and rehydrated sections were boiled in Na-citrate buffer (10 mM, pH 6.0) for 30 min for antigen retrieval. The tissues were blocked in 10% negative goat serum for 60 minutes at room temperature, washed with ddH2O and PBS, and then probed with the primary antibody (29352-1-AP) at 1/200 dilution overnight at 4 ℃. The detection was performed using an HRP conjugated compact polymer system. DAB was used as the chromogen.

Quantitative real‑time PCR (qRT-PCR)

First-strand cDNA was synthesized using the Vazyme R323-01 kit according to the standard protocol. And qPCR was performed using the SYBR Green Supermix and CFX96 real-time PCR detection system. Each experiment was performed in triplicate, and the mRNA expression of genes was analyzed using the 2^−ΔΔCt method. The following primers were used: 5'-GCA AAT ACA AAG GAG CCA ACA AG-3' and 5'-CAT GGA GCC GAT TCG GTC C-3' for Natriuretic Peptide C (NPPC), 5'-TCG TAC CCT GGT TCG CAC T-3' and 5'-GCT GCA CAC TGA GGT TGC T-3' for NPR2, 5'-TCG TGC GTG ACA TTA AGG AG-3' and 5'-ATG CCA GGG TAC ATG GTG GT-3' for ACTB.

Gene expression data were normalized to ACTB, and results are presented as mean ± standard deviation from three independent biological replicates.

Cell migration assays

Corning cell culture inserts (24-well, 8-µm pore size) were used according to the manufacturer’s instructions. Briefly, cells (2×10⁴) suspended in 200 µL serum-free medium containing CNP-38 were seeded into the upper chamber of the inserts. The lower chamber was filled with 500 µL of medium supplemented with 10% FBS. The chambers were incubated at 37 ℃ for 12 hours. After incubation, cells remaining on the upper side of the membrane were gently scraped off with a cotton swab. The membranes were then fixed with 4% paraformaldehyde and stained with 0.5% crystal violet solution. Migrated cells were counted in four random fields per well using a light microscope. Each experiment was repeated at least three times, and the migration ratio was normalized to the control group.

Statistical analysis

All statistical analyses were conducted using the R software (v4.2.0). The Student’s t-test, Wilcoxon rank sum test, Kruskal-Wallis test, Pearson’s Chi-squared test and log rank test were used in present study. Statistical significance was set at P<0.05 and shown as *, P<0.05; **, P<0.01; and ***, P<0.001.

Results

Molecular classification in SCLC based on neuro-interaction signaling genes

The analytical process of this study is illustrated in Figure S1A. First, we investigated gene sets related to neuroactive signaling from MSigDB database and took the collection of three gene sets (KEGG_NEUROACTIVE_LIGAND_RECEPTOR_INTERACTION, REACTOME_NEUROTRANSMITTER_RECEPTORS_AND_POSTSYNAPTIC_SIGNAL_TRANSMISSION, REACTOME_TRANSMISSION_ACROSS_CHEMICAL_SYNAPSES), a total of 492 genes (Figure S1B), to define them as NRGs. Gene Ontology (GO) enrichment analysis indicated that these genes are involved in various activities related to neurotransmitters, including NMDA receptor activation and postsynaptic events, GABAergic synapses, and calcium signaling pathways (Figure S1C). Expression of genes between SCLC and normal samples was compared, and 51 NRGs were found to be differentially expressed in both independent datasets (Figure S1D; logFoldChange >1.0 & adj.p.val <0.05).

To understand the expression patterns of NRGs involved in tumorigenesis comprehensively, 79 SCLC patients with primary tumor transcriptome sequencing data and survival information were included for further analyses, with their clinical characteristics summarized in Table S3.

A total of 5 clusters were identified through a consensus clustering algorithm (Figure S1E,S1F). Among them, 23 patients were classified as NRC1, 24 patients as NRC2, 29 patients as NRC3, 1 patient as NRC4, and 2 patients as NRC5 (Figure 1A; Table S4). In survival analysis, patients in NRC3 had worse outcomes compared to other groups (P=0.042, log-rank test, Figure 1B). NRC1–3 were the major subtypes, and subsequent analyses mainly focused on comparisons among these three subtypes. Different NRCs exhibited distinct clinical characteristics (Figure 1C). NRC2 and NRC3 had more stage I disease (58.33% vs. 46.67%, 0.018, Pearson’s Chi-squared test), compared with NRC1(18.18%). Notably, patients in NRC3 had the worst survival with the lowest 1-year overall survival (OS) rate (53.33%, P=0.03, Pearson’s Chi-squared test). Additionally, the results of the univariate and multivariate Cox regression analyses suggested that the NRC3 subtype is an independent prognostic risk factor for SCLC patients (Figure 1D; Figure S1G-S1J). The heatmap showed differential expression of NRGs across different clusters (Figure 1E).

Figure 1 Clinical characteristics and survival of different NRCs. (A) t-SNE visualization of clusters in tumors of George’s cohort, n=79. Dots are colored according to clusters. (B) Kaplan-Meier curves for overall survival of NRCs. (C) Heatmap shows clinical characteristics of three major NRCs. P value was determined by Pearson’s Chi-squared test. (D) Forest plot shows multivariate Cox regression of clinical factors and NRCs. (E) Heatmap shows the expression of NRGs among three major NRCs. OS, overall survival; CI, confidence interval; NRC, neuroactive signaling-related molecular cluster; NRG, neuroactive-signaling-related gene.

Biological molecular changes underlying different clusters in SCLC

To understand the relationship between these three NRCs and classical subtypes, we created a Sankey diagram. The results showed that within the classical ANPY classification of SCLC, each subtype could be further divided into subgroups (Figure 2A). What’s more, NRC subtyping does not show a significant correlation with the classical ANPY subtyping (Figure S2A). This suggests tumor heterogeneity within SCLC subtypes and highlights the need for further refinement of SCLC subtypes, to which NRC subtyping may contribute. Next, to explore the biological differences between NRCs, we performed GSVA analysis. The results revealed that NRC3 was enriched in the Hedgehog signaling pathway, while NRC2 was enriched in pathways related to MYC signaling and metabolism, such as mitotic signaling, glycolysis, and lipid metabolism. In contrast, NRC1 was enriched in immune response pathways, such as interferon-γ and inflammatory responses (Figure 2B,2C). Distinct metabolic profiles were observed among the clusters, with NRC2 showing broad enrichment in amino acid metabolism, glycan biosynthesis, and energy-related pathways, while NRC1 and NRC3 exhibited relatively lower activity in most pathways (Figure 2D).

Figure 2 Biological molecular changes underlying three clusters in SCLC. (A) Alluvial diagram of clusters with different molecular subtypes. (B) Heatmap shows molecular characteristics of NRCs. (C) Dot plot shows correlation between NRCs and Hallmarks gene sets. (D) Heatmap shows metabolic characteristics of NRCs. NRC, neuroactive signaling-related molecular cluster; GSVA, gene set variation analysis; SCLC, small cell lung cancer.

The analysis of tumor stemness showed that NRC2 had higher tumor stemness (Figure S2B). According to the analysis of neurotransmitter pathway activity, we found that NRC3 was generally more active (Figure S2C). Currently, systemic chemotherapy remains the conventional strategy for SCLC patients. We subsequently estimated median inhibition concentration (IC50) values of several chemotherapeutic drugs via oncoPredict algorithm. Drug sensitivity analysis indicated that different NRCs had distinct drug sensitivity profiles (Figure S2D). NRC2 was more sensitive to cell cycle inhibitors such as CDK9 inhibitor, which exhibits antitumor activity in SCLC in recent research (26,27). NRC3 was sensitive to kinase inhibitors like tozasertib (Aurora kinase inhibitor), axitinib (selective VEGFR 1-3 inhibitor), and cediranib (VEGFR2 inhibitor), while NRC1 showed low sensitivity to antitumor drugs.

Distinct TME in three tumor clusters

To investigate the genomic features associated with NRG-defined subtypes, we analyzed somatic mutation profiles across NRC1–3. ZFHX4 mutations were significantly enriched in NRC3, with mutation rates of 32%, 35%, and 62% in NRC1, NRC2, and NRC3, respectively (Figure 3A). ZFHX4 has been previously implicated in skeletal and neural development (28). A pan-cancer research reported that ZFHX4 mutant was related to better clinical outcomes for immunotherapy (29). These suggested the potential relevance of ZFHX4 mutation to the neuroactive phenotype of NRC3 tumors and better survival benefit for immunotherapy. We next explored the components of TME of each subtype by applying XCELL algorithm and ssGSEA to estimate the relative infiltration of various cell populations. NRC1 exhibited relatively robust immune cell infiltration, including higher levels of Macrophage M1 (Figure 3B; Table S5), activated dendritic cells, activated CD8⁺ T cells, and natural killer (NK) cells (Figure S2E), compared to NRC2 and NRC3. NRC2 displayed a moderate level of immune cell infiltration, while simultaneously showing the highest abundance of fibroblasts among the three subtypes, suggesting a complex immune-stromal microenvironment. In contrast, NRC3 displayed characteristics of an “immune-naïve” subtype, low immune cell and fibroblast infiltration but robust neuron component. This suggests that NRC3 may have a better anti-tumor effect in response to immune-stimulating treatments, such as chemotherapy combined with immunotherapy. Interestingly, the correlation analysis between neurotransmitter pathways and immune infiltration revealed that dopaminergic and GABAergic synapses were negatively correlated with fibroblast infiltration (Figure S2F), indicating the association of neurons component and fibroblast.

Figure 3 Genomic and immunological features of NRCs in SCLC. (A) Oncoplot shows top 30 somatic alterations of NRCs in George’s cohort. (B) Barplot shows the distribution of immune cell subtypes across three NRCs. The statistical difference of three clusters was compared through the Kruskal-Wallis test. (C) Barplot shows the immune function among three NRCs. (D) Heatmap shows the expression of genes related to immune functions across three NRCs. ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001. NRC, neuroactive signaling-related molecular cluster; TMB, tumor mutation burden; SCLC, small cell lung cancer.

To further assess immune functional status, we evaluated the activity of key immune-related pathways. NRC1 showed upregulated MHC class I pathway activity and enhanced cytotoxicity, whereas NRC3 exhibited suppression of both pathways (Figure 3C). In addition, NRC1 demonstrated higher expression of immune response-related genes compared with other subtypes (Figure 3D), supporting its designation as an immune-active tumor subtype. This immune-activated state may contribute to the relatively favorable prognosis observed in NRC1, despite its association with more advanced clinical staging.

Collectively, these findings delineate three biologically distinct SCLC subtypes based on NRGs stratification: NRC1, defined by immune activation and potentially sensitive to immunotherapy; NRC2, characterized by enriched metabolic and cell cycle activity; and NRC3, distinguished by neurodevelopmental features.

Identification of key genes by WGCNA analysis

To further characterize the key genes in NRC3, we collected all the up-regulated genes in tumor from two available SCLC cohorts (GSE60052, GSE40275, Figure S2G). We next performed WGCNA analysis to identify the gene module most associated with NRC3 (Figure 4A). Firstly, the optimal soft-thresholding power (β) was set to 7 to achieve a scale-free topology fit index (R²) of 0.9. Subsequently, the minimum number of genes per module was defined as 20 (Figure S2H). Based on the gene expression similarity, hierarchical clustering grouped the genes into 18 distinct modules (Figure 4B). The results showed that cyan gene module was associated with NRC2 and NRC3 [module eigengene (ME) =−0.43, P=1e−04; ME =0.43, P=1e−04; Figure 4C], whose genes were enriched in ion channels referring to cellular electrical activity (Figure 4D). And we defined neuroactive-signaling-related signature score (NRS) based on genes in cyan gene module. Patients were stratified into high and low NRS groups according to NRS (Figure 4E), and the TME components of each group were subsequently analyzed. The results revealed that tumors in the high NRS group exhibited increased infiltration of neurons and Th1 cells, whereas the low NRS group showed greater infiltration of T cells, macrophages, and fibroblasts (Figure 4F). Correlation analysis demonstrated that NRS was positively correlated with neuron components (r=0.83, P<0.001, Figure 4G), while negative with fibroblast (r=−0.54, P<0.001, Figure 4H). The distribution of NRS in George’s Cohort was displayed in Figure S2I. These findings indicate that the NRS reflects not only the neuroactive interaction profile of tumor but also key differences in the composition of the tumor microenvironment.

Figure 4 WGCNA identifies gene co-expression modules in SCLC. (A) Dendrogram of upregulated genes in SCLC samples. (B) Heatmap shows correlations between gene modules and clinical traits. (C) Scatter plot shows the module-trait relationship between the cyan module and NRC2 (top) or NRC3 (bottom). (D) Enrichment analysis of the cyan gene module. The color indicates the number of enriched genes in each pathway. (E) Heatmap illustrates the expression profiles of NRS genes. (F) Boxplot displays differences in the proportions of TME cell types between high- and low-NRS groups in George’s Cohort (top) and Jiang’s Cohort (bottom). Statistical significance was assessed using the Wilcoxon rank-sum test. (G,H) Scatter plot shows plots show that NRS is positively correlated with neuronal components and negatively correlated with fibroblast components of tumor in George’s Cohort. *, P<0.05; **, P<0.01; ***, P<0.001. NRC2, neuroactive signaling-related molecular cluster 2; NRC3, neuroactive signaling-related molecular cluster 3; NRS, neuroactive-signaling-related signature score; SCLC, small cell lung cancer; TME, tumor microenvironment; WGCNA, weighted gene co-expression network analysis.

ScRNA-seq analysis revealed transcriptional heterogeneity of malignant cells between NRS-defined SCLC tumors

To uncover the cellular composition of SCLC, the collected scRNA-seq data were clustered and identified seven major cell types based on their canonical cell marker genes (Figure S3A). Spearman’s correlation was performed to identify the relationship between clusters (Figure 5A-5D; Figure S3B,S3C; Table S6). Using the pseudo-bulk algorithm, we stratified single-cell samples into high and low NRS groups based on the median NRS value (Figure 5E).

Figure 5 Landscape of cell types in SCLC. (A) t-SNE visualization of cells derived from primary tumor samples of 11 patients, colored by clusters. (B) t-SNE visualization of cells from 11 SCLC patients, colored by main cell types. (C) Heatmap of Spearman’s correlations among each cluster. (D) Dot plot of representative cell markers among each cluster. Dot size represents abundance, and color represents expression level. (E) t-SNE visualization of cells from tumors stratified by high and low NRS scores via pseudo-bulk sequencing algorithm. (F) t-SNE visualization of malignant cells from 11 SCLC patients. (G) Differentiation trajectory of malignant cells inferred by Monocle, and colored by pseudo time (left) and NRS subtype (right). (H) The landscape of large-scale CNVs of malignant cells inferred by inferCNV. CNV, copy number variation; NRS, neuroactive-signaling-related signature score; SCLC, small cell lung cancer; t-SNE, t-distributed stochastic neighbor embedding.

To further investigate the dynamic states of these tumor cells, we performed pseudotime trajectory analysis. The results showed that malignant cells from H-NRS samples were positioned at later stages of the inferred evolutionary trajectory, whereas cells from L-NRS samples were primarily located in earlier developmental states (Figure 5F,5G). CNV analysis of malignant cells revealed distinct genomic alteration patterns between H- and L-NRS groups (Figure 5H). These findings imply that H-NRS tumors may represent a more advanced and aggressive stage of SCLC.

Given the pivotal role of cell-cell communication in tumor immunity and progression, we investigated the interactions between malignant cells and the TME cells. Analysis of cellular composition revealed that tumors in the H-NRS group exhibited a higher proportion of malignant cells compared to those in the L-NRS group (Figure 6A). Furthermore, cell-cell communication analysis demonstrated that the overall number and strength of cellular interactions were comparable between the two tumor groups (Figure 6B). It is noteworthy that immune cell interactions were more active in H-NRS tumors, whereas fibroblast interactions with other cell types were more prominent in L-NRS tumors (Figure 6C,6D). This revealed distinct microenvironmental dynamics associated with different NRS levels. The results of signaling patterns showed that CXCL signaling was active in H-NRS tumor (Figure 6E-6H). In contrast, collagen signaling pathway was more active in tumors with L-NRS, as evidenced by the enhanced activity of several ligand-receptor pairs compared to those in the H-NRS group (Figure 6I). In conclusion, cells in H-NRS tumors engage in more active interactions with various components of the TME, indicating a potentially more complex and dynamic TME.

Figure 6 Distinct tumor microenvironment infiltration of NRS-defined SCLC tumor. (A) Bar plot shows the distribution of different cell types within each tumor sample in two groups. (B) Histogram showing both the quantity and intensity of intercellular communication within tumors, grouped by NRS status. (C) Heatmap displays the number of inferred cell-cell interactions within TME. Red indicates higher activity in H-NRS tumors. Blue indicates higher activity in L-NRS tumors. (D) Scatter plot shows the outgoing and incoming interaction strength of each cell subset in H-NRS tumors and L-NRS tumors. Dot plot of ligand-receptor interaction from tumor cells to immune cells between H-NRS tumors and L-NRS tumors. Dot size represents the P-value, and color represents the interaction strength. (E,F) Heatmaps show the top 15 incoming and outgoing signaling in two groups. (G,H) Hierarchy plots illustrate CXCL signaling across different cell types in two groups. (I) Dot plot presents the interactions (L-R pairs) in COLLAGEN pathway from fibroblast to other cell types. Dot size represents p value, and color represents communication probability. CXCL, C-X-C motif chemokine ligand; H-NRS, high neuroactive-signaling-related signature score; L-NRS, low neuroactive-signaling-related signature score; L-R, ligand-receptor; NRS, neuroactive-signaling-related signature score; SCLC, small cell lung cancer; TME, tumor microenvironment.

Low-NRS patients display a poor response to chemoimmunotherapy

To explore the predictive value of NRS in immunotherapy response, we evaluated the response to anti-PD-L1 therapy in George’s Cohort and a cohort of SCLC patients from Guangzhou Medical University (GZHMU’s cohort). The high NRS group exhibited a response rate over 10% higher than that of the low NRS group in both cohorts (Figure 7A,7B). The clinical characteristics of patients in GZHMU’s cohort were summarized in Table S7. Due to the limited samples of GZHMU’s cohort, the results showed no statistically significant difference in TMB and tumor lesion change (Figure 7C,7D). We noticed that high NRS tumors tend to exhibit higher TMB, greater tumor shrinkage after immunotherapy, and better survival benefits compared to those with low NRS tumors (Figure 7E-7H).

Figure 7 NRS predicts immunotherapy efficacy. (A,B) Histogram shows the response rate of immunotherapy inferred by TIDE algorithm in George’s Cohort and GZHMU’s Cohort. (C,D) Boxplot shows TMB and percentage of lesion change of tumor in GZHMU’s Cohort. (E-H) Scatter plots illustrate the correlation between NRS and TMB (E), lesion change (F), PD-1 treatment cycles (G), and OS (H). (I,J) Kaplan-Meier curves for PFS and OS in SCLC patients stratified by NRS levels. Cutting off value was determined using the survminer R package. GZHMU, Guangzhou Medical University; NRS, neuroactive-signaling-related signature score; OS, overall survival; PD-1, programmed cell death protein 1; PFS, progression-free survival; SCLC, small cell lung cancer; TIDE, tumor immune dysfunction and exclusion; TMB, tumor mutation burden.

Furthermore, survival analyses revealed that H-NRS SCLC patients experienced longer progression-free survival (PFS) and OS when treated with first-line chemo-immunotherapy, compared with their L-NRS counterparts (Figure 7I,7J). These findings highlight the potential of NRS as a predictive biomarker for immunotherapy benefit in SCLC. In addition, immune infiltration scoring revealed that, consistent with our previous findings, tumors in the H-NRS group exhibit a higher neuronal component but lower immune infiltration and fibroblast content, reflecting an immune-naïve TME. This may be the underlying reason why patients in this group derive greater benefit from immunochemotherapy. In contrast, tumors in the L-NRS group, while having higher immune cell infiltration, also show a higher fibroblast component, indicating an immune-tolerant TME (Figure S3D).

NPPC plays oncogenic role in SCLC

To identify key genes within NRS associated with patient prognosis, we performed survival analysis on NRS components. Total 11 genes were found to be significantly correlated with OS (Figure 8A). Among them, NPPC emerged as a significant prognostic risk factor (hazard ratio: 1.034, 95% confidence interval: 1.011–1.057, P=0.003). Patients with high expression level of NPPC exhibited significantly worse outcomes compared to those with low NPPC expression (P<0.001, log-rank test, Figure 8B). Transcriptomic sequencing data from tissues and cell lines suggest that NPPC is highly expressed in small cell lung cancer (Figure S3E,S3F). To determine the cellular origin of NPPC expression, we mapped the expression of NPPC and its receptor gene NPR2 in our single-cell RNA-seq dataset. The results revealed that NPPC was specifically expressed in malignant tumor cells, while NPR2 was expressed in both tumor cells and T cells (Figure 8C,8D), suggesting a potential tumor-immune interaction axis mediated by the NPPC-NPR2 signaling pathway. Pseudo-time analysis results indicated that NPPC levels gradually increase during tumor progression (Figure S3G).

Figure 8 Identification of prognostic genes in NRS and Validation. (A) Forest plot shows prognosis-associated genes within NRS. (B) In George’s cohort, the distribution of high and low expression of NPPC. Cutting off value was determined using the survminer R package. High expression level of NPPC is associated with poor patient prognosis. (C,D) t-SNE visualization of cells shows the expression of NPPC and its receptor NPR2 in single-cell RNA-seq data. (E) Representative areas of HE and IHC of NPPC expression in tumorous tissues from three SCLC patients. Scale bar, 50 µm. (F-G) qPCR validation of NPPC and NPR2 expression across various human normal epithelial and SCLC cell lines. (H) Transwell migration assay of SCLC-1 cells treated with CNP-38 (product of the NPPC gene) or DMSO (control) for 12 hours (0.5% crystal violet dye, ×100). (I) Data processing and analysis of transwell migration assay. ***, P<0.001; CI, confidence interval; CNP, C-type natriuretic peptide; DMSO, dimethyl sulfoxide; HE, hematoxylin and eosin staining; IHC, immunohistochemistry; NPPC, natriuretic peptide C; NPR2, natriuretic peptide receptor B; NRS, nomogram/prognostic risk score; qPCR, quantitative polymerase chain reaction; RNA-seq, ribonucleic acid sequencing; SCLC, small cell lung cancer; t-SNE, t-distributed stochastic neighbor embedding.

Then, we further validated NPPC expression in human SCLC tissue and cell lines. The image of IHC demonstrated that the expression level of NPPC varies among different SCLC tumors. Of note, Patient 03, who exhibited elevated NPPC expression, succumbed following the first round of chemotherapy, demonstrating an exceptionally short OS (Figure 8E, Table S8). The results of qPCR showed NPPC was highly expressed in SCLC cells, compared with lung epithelial cells, lung adenocarcinoma cells and lung large cell carcinoma cells (Figure 8F,8G).

To evaluate the impact of NPPC on SCLC cells, we conducted transwell migration assay on SCLC-1. The results revealed that treatment with C-type Natriuretic Peptide-38 (CNP-38), the peptide product of NPPC, significantly promoted the migration of SCLC-1 cells compared to the DMSO control (Figure 8H,8I). These findings indicated that NPPC may contribute to tumor metastasis in SCLC.

Discussion

The emerging paradigm of cancer neuroscience has unveiled intricate neuro-tumoral crosstalk as a critical regulator of oncogenesis, extending beyond classical neuronal innervation to encompass autocrine/paracrine neuroactive signaling within tumor ecosystems (7,8,30-32). Neurotransmitters are now recognized as critical components of TME, functioning not only through neuronal secretion but also via autocrine and paracrine signaling from cancer and immune cells (33-35). These molecules influence not just tumor growth and metastasis but also modulate immune cell behavior, suggesting a complex, bidirectional interaction between neural signals and immune responses in cancer.

Our study sheds new light on these neuro-immune interactions in SCLC, a malignancy notorious for its aggressive clinical course and poor prognosis. We identified three distinct neuro-interaction-related molecular clusters (NRC1–3) that display markedly different transcriptomic, genomic, and microenvironmental features. NRC1, characterized by active immune signaling pathways and high immune infiltration, represents an “immune-active” phenotype, potentially more responsive to immunotherapy. In contrast, NRC3 showed minimal immune engagement and pronounced neurogenic signaling, defining an “immune-naïve” or “neuro-interaction-active” subtype. NRC2, enriched for cell cycle, DNA repair, and MYC-driven transcriptional programs, represents a “metabolically active” phenotype.

This classification not only improves our understanding of SCLC heterogeneity but also provides a rational framework for therapeutic stratification. The construction of NRS, a gene signature reflecting activity of neuroactive signaling, allowed further dissection of these phenotypes at single-cell resolution. Our findings suggest that tumors with H-NRS exhibit decreased immune infiltration and more advanced malignant trajectories, whereas L-NRS tumors are enriched for fibroblasts and immune cells. Notably, communication analysis revealed that fibroblasts in L-NRS tumors were the most interactive cell population, highlighting their possible role as physical and biochemical barriers to effective anti-tumor immunity.

This aligns with prior reports that cancer-associated fibroblasts (CAFs), through excessive extracellular matrix (ECM) production, increase intratumoral solid stress and physically hinder T cell infiltration (36). Such “immune-excluded” microenvironments represent a significant obstacle to the efficacy of immune checkpoint blockade. Importantly, we observed that L-NRS tumors in our institutional cohort derived limited benefit from first-line chemo-immunotherapy. This provides a potential explanation for the observed clinical resistance and underscores the need for combination therapies that target both the immune axis and stromal components—such as the inclusion of anti-angiogenic or ECM-modulating agents.

Mechanistically, we identified NPPC, a gene coding CNP, as a poor prognostic factor within the NRS. NPPC was specifically expressed in malignant cells, while its receptor, NPR2, was found in malignant cells and T cells, suggesting a direct neuro-immune communication axis within the TME. Functional assays indicated that NPPC/CNP may promote tumor cell migration, hinting at its potential role in metastasis. Although these findings are preliminary, they suggest that NPPC may be a key modulator of tumor–immune and tumor–neural interactions, and thus a potential therapeutic target.

Despite these advances, several limitations must be acknowledged. First, the lack of in vivo validation for NPPC’s role in shaping the immune microenvironment limits mechanistic interpretation. Second, although we used scRNA-seq data to infer communication patterns, functional validation of cell-cell interactions using spatial transcriptomics or co-culture models would enhance the robustness of our conclusions. Third, while NRS stratified patients in our cohort, prospective clinical validation in larger, independent cohorts and in the context of clinical trials is essential before clinical application.

Nevertheless, our study provides novel insights into the neural dimension of SCLC biology. By linking neuroactive signals to immune evasion, genomic instability, and clinical outcomes, we enchanted our understanding of both biomarker and anti-tumor therapy. As the field progresses, integrating neuroactive signaling metrics with conventional biomarkers may finally unravel SCLC’s paradoxical chemosensitivity/rapid relapse cycle, ultimately transforming this recalcitrant malignancy into a tractable chronic disease.

Limitation

Although NPPC was identified as a potential prognostic biomarker and therapeutic target, its biological role in modulating the TME and response to immunotherapy has not yet been validated through in vivo experiments. The study relied on a relatively small number of patients from retrospective datasets, which may limit the generalizability of the findings. Larger, prospective studies are needed to confirm the predictive value of NRS in clinical settings.

Conclusions

In conclusion, our research provides new insights into the classification of SCLC based on neuroactive-related gene expression. NRS serves as a potential biomarker for SCLC patients treated with chemoimmunotherapy, promoting more precise clinical management and risk stratification. And NPPC represents a novel therapeutic target for SCLC treatment.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STREGA reporting checklist. Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-620/rc

Data Sharing Statement: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-620/dss

Peer Review File: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-620/prf

Funding: This study was supported by grants from National Natural Science Foundation of China (No. 82270065), National Key Research and Development Program of China (Nos. 2022YFF1203300 and 2024YFA1108900), Natural Science Foundation of Guangdong Province of China (No. 2023A1515010886), Clinical and Epidemiological Research Project of State Key Laboratory of Respiratory Disease (No. SKLRD-L-202405), and Major Project of Guangzhou National Laboratory (No. GZNL2024A02004).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-620/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of The First Affiliated Hospital of Guangzhou Medical University (Guangzhou, China, No. 2020-189) and informed consent was obtained from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Rudin CM, Brambilla E, Faivre-Finn C, et al. Small-cell lung cancer. Nat Rev Dis Primers 2021;7:3. [Crossref] [PubMed]
2. Baine MK, Hsieh MS, Lai WV, et al. SCLC Subtypes Defined by ASCL1, NEUROD1, POU2F3, and YAP1: A Comprehensive Immunohistochemical and Histopathologic Characterization. J Thorac Oncol 2020;15:1823-35. [Crossref] [PubMed]
3. Zhi X, Wu F, Qian J, et al. Nociceptive neurons promote gastric tumour progression via a CGRP-RAMP1 axis. Nature 2025;640:802-10. [Crossref] [PubMed]
4. Xiong SY, Wen HZ, Dai LM, et al. A brain-tumor neural circuit controls breast cancer progression in mice. J Clin Invest 2023;133:e167725. [Crossref] [PubMed]
5. Chang A, Botteri E, Gillis RD, et al. Beta-blockade enhances anthracycline control of metastasis in triple-negative breast cancer. Sci Transl Med 2023;15:eadf1147. [Crossref] [PubMed]
6. Zahalka AH, Arnal-Estapé A, Maryanovich M, et al. Adrenergic nerves activate an angio-metabolic switch in prostate cancer. Science 2017;358:321-6. [Crossref] [PubMed]
7. Fnu T, Shi P, Zhang W, et al. Sympathetic Neurons Promote Small Cell Lung Cancer through the β2-Adrenergic Receptor. Cancer Discov 2025;15:616-32. [Crossref] [PubMed]
8. Peinado P, Stazi M, Ballabio C, et al. Intrinsic electrical activity drives small-cell lung cancer progression. Nature 2025;639:765-75. [Crossref] [PubMed]
9. Vermeer PD, Restaino AC, Barr JL, et al. Nerves at Play: The Peripheral Nervous System in Extracranial Malignancies. Cancer Discov 2025;15:52-68. [Crossref] [PubMed]
10. Mancusi R, Monje M. The neuroscience of cancer. Nature 2023;618:467-79. [Crossref] [PubMed]
11. Amit M, Anastasaki C, Dantzer R, et al. Next Directions in the Neuroscience of Cancers Arising outside the CNS. Cancer Discov 2024;14:669-73. [Crossref] [PubMed]
12. Winkler F, Venkatesh HS, Amit M, et al. Cancer neuroscience: State of the field, emerging directions. Cell 2023;186:1689-707. [Crossref] [PubMed]
13. Kastner S, Voss T, Keuerleber S, et al. Expression of G protein-coupled receptor 19 in human lung cancer cells is triggered by entry into S-phase and supports G(2)-M cell-cycle progression. Mol Cancer Res 2012;10:1343-58. [Crossref] [PubMed]
14. George J, Lim JS, Jang SJ, et al. Comprehensive genomic profiles of small cell lung cancer. Nature 2015;524:47-53. [Crossref] [PubMed]
15. Jiang L, Huang J, Higgs BW, et al. Genomic Landscape Survey Identifies SRSF1 as a Key Oncodriver in Small Cell Lung Cancer. PLoS Genet 2016;12:e1005895. [Crossref] [PubMed]
16. Liu M, Qiu G, Guan W, et al. Induction chemotherapy followed by camrelizumab plus apatinib and chemotherapy as first-line treatment for extensive-stage small-cell lung cancer: a multicenter, single-arm trial. Signal Transduct Target Ther 2025;10:65. [Crossref] [PubMed]
17. Tian Y, Li Q, Yang Z, et al. Single-cell transcriptomic profiling reveals the tumor heterogeneity of small-cell lung cancer. Signal Transduct Target Ther 2022;7:346. [Crossref] [PubMed]
18. Miranda A, Hamilton PT, Zhang AW, et al. Cancer stemness, intratumoral heterogeneity, and immune response across cancers. Proc Natl Acad Sci U S A 2019;116:9020-9. [Crossref] [PubMed]
19. Barbie DA, Tamayo P, Boehm JS, et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature 2009;462:108-12. [Crossref] [PubMed]
20. Finotello F, Trajanoski Z. Quantifying tumor-infiltrating immune cells from transcriptomics data. Cancer Immunol Immunother 2018;67:1031-40. [Crossref] [PubMed]
21. Aran D, Hu Z, Butte AJ. xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome Biol 2017;18:220. [Crossref] [PubMed]
22. Maeser D, Gruener RF, Huang RS. oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data. Brief Bioinform 2021;22:bbab260. [Crossref] [PubMed]
23. Jiang P, Gu S, Pan D, et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med 2018;24:1550-8. [Crossref] [PubMed]
24. Trapnell C, Cacchiarelli D, Grimsby J, et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 2014;32:381-6. [Crossref] [PubMed]
25. Jin S, Guerrero-Juarez CF, Zhang L, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun 2021;12:1088. [Crossref] [PubMed]
26. Valdez Capuccino L, Kleitke T, Szokol B, et al. CDK9 inhibition as an effective therapy for small cell lung cancer. Cell Death Dis 2024;15:345. [Crossref] [PubMed]
27. Wang Y, Wang M, Ma L, et al. Identification of a Potent and Selective CDK9 Degrader as a Targeted Therapeutic Option for the Treatment of Small-Cell Lung Cancer. J Med Chem 2025;68:2528-50. [Crossref] [PubMed]
28. Pérez Baca MDR, Palomares-Bralo M, Vanhooydonck M, et al. Loss of function of the zinc finger homeobox 4 gene, ZFHX4, underlies a neurodevelopmental disorder. Am J Hum Genet 2025;112:1388-414. [Crossref] [PubMed]
29. Mao Y, Xie H, Lv M, et al. The landscape of objective response rate of anti-PD-1/L1 monotherapy across 31 types of cancer: a system review and novel biomarker investigating. Cancer Immunol Immunother 2023;72:2483-98. [Crossref] [PubMed]
30. Kobayashi H, Iida T, Ochiai Y, et al. Neuro-Mesenchymal Interaction Mediated by a β2-Adrenergic Nerve Growth Factor Feedforward Loop Promotes Colorectal Cancer Progression. Cancer Discov 2025;15:202-26. [Crossref] [PubMed]
31. Sakthivelu V, Schmitt A, Odenthal F, et al. Functional synapses between neurons and small cell lung cancer. Nature 2025;646:1243-53. [Crossref] [PubMed]
32. Wang K, Ni B, Xie Y, et al. Nociceptor neurons promote PDAC progression and cancer pain by interaction with cancer-associated fibroblasts and suppression of natural killer cells. Cell Res 2025;35:362-80. [Crossref] [PubMed]
33. Jiang SH, Hu LP, Wang X, et al. Neurotransmitters: emerging targets in cancer. Oncogene 2020;39:503-15. [Crossref] [PubMed]
34. Klein Wolterink RGJ, Wu GS, Chiu IM, et al. Neuroimmune Interactions in Peripheral Organs. Annu Rev Neurosci 2022;45:339-60. [Crossref] [PubMed]
35. Khanmammadova N, Islam S, Sharma P, et al. Neuro-immune interactions and immuno-oncology. Trends Cancer 2023;9:636-49. [Crossref] [PubMed]
36. Kalluri R. The biology and function of fibroblasts in cancer. Nat Rev Cancer 2016;16:582-98. [Crossref] [PubMed]