Lysosome Gene Signature Offers New Prognostic Tool for Lung Cancer

Lung adenocarcinoma, the most commonly diagnosed form of lung cancer, remains one of medicine’s most stubborn adversaries. Despite decades of progress in targeted therapy and immunotherapy, the five-year survival rate for patients still sits below 15 percent, and once the disease has spread, that figure collapses to less than 3 percent. Now, a team of researchers in China has turned an unexpected cellular workhorse—the lysosome—into the foundation of a new predictive tool that may help clinicians decide, with greater precision, which patients need aggressive treatment and which therapies are most likely to work.

Lysosomes are membrane-enclosed organelles, formed by the Golgi apparatus and endoplasmic reticulum, that serve as the cell’s digestive and recycling centers. They contain soluble hydrolases capable of breaking down virtually every class of biological macromolecule, along with membrane proteins that regulate nutrient sensing, energy balance, and the trafficking of key signaling molecules. In healthy cells, this system maintains homeostasis. In cancer cells, however, lysosomes are hijacked. Tumors, with their voracious appetite for energy and building blocks, become heavily dependent on lysosomal degradation and recycling pathways to fuel their rapid growth.

The research team, led by Jingze Yan and colleagues at the First Affiliated Hospital of Nanjing Medical University and published in Clinical Cancer Bulletin, systematically examined 61 lysosome-related genes drawn from the Molecular Signatures Database. Their analysis of 522 lung adenocarcinoma samples from the TCGA database revealed that these genes are far from passive bystanders. Nearly 39 percent of the 577 tumor cases evaluated carried mutations in at least one of 35 lysosome-related genes, with LRP2 showing the highest mutation frequency, followed by TYR and USP6. Widespread copy number variations were also detected across the gene set, underscoring how frequently the lysosomal machinery is genetically perturbed in this cancer.

To understand how lysosomal biology shapes tumor behavior, the researchers applied consensus clustering to the TCGA cohort, sorting patients into two distinct lysosome-associated subgroups. The differences between them were striking. One cluster showed markedly higher infiltration of immune cells and richer immune-related activity, while the other displayed the opposite pattern. Functional enrichment analysis of 871 genes differentially expressed between the clusters pointed toward pathways central to cancer immunity: cytokine-cytokine receptor interactions, cell adhesion molecules, Th1 and Th2 cell differentiation, antigen processing and presentation, and the phagosome pathway. The message was clear—lysosomal gene expression is entangled with the immune architecture of the tumor microenvironment.

From this foundation, the team built what they call the lysosome-related prognostic signature, or LRPS. Using a rigorous statistical pipeline—univariate Cox regression to screen 214 survival-linked genes, multivariate Cox analysis to trim the list, and LASSO regression with 1000-fold cross-validation to select the final set—they distilled the signal down to 13 genes: DKK1, RHOV, DLGAP5, NTSR1, BCAN, GREB1L, OLAH, ACSM5, SPOCK1, LY6K, MS4A1, SEC14L3, and ELOVL2. Each gene contributes a weighted coefficient to a risk score, and patients are classified as high- or low-risk based on the median value.

The signature’s predictive power held up under scrutiny. In the training cohort, Kaplan-Meier curves showed that high-risk patients survived significantly worse than their low-risk counterparts, and time-dependent receiver operating characteristic analysis yielded area-under-the-curve values of 0.771, 0.744, and 0.671 for one-, three-, and five-year overall survival, respectively. Crucially, the pattern replicated in two independent validation cohorts, GSE72094 and GSE31210, with comparable discrimination. Multivariate Cox regression confirmed that the LRPS risk score remained an independent prognostic factor even after accounting for age, sex, and tumor stage, and stratified analyses showed the signature retained its reliability across every clinical subgroup tested.

Single-cell RNA sequencing of more than 208,000 cells from 58 lung adenocarcinoma tumors added a cellular dimension to the story. The 13 signature genes were not scattered randomly across the tumor. DKK1 and RHOV concentrated in type II alveolar epithelial cells, the likely cell of origin for many adenocarcinomas; DLGAP5 appeared mainly in proliferating cells; ACSM5 in myeloid cells; SPOCK1 in fibroblasts; MS4A1 in B cells; and SEC14L3 in ciliated cells. This spatial mapping suggests the signature captures biology from multiple compartments of the tumor ecosystem—malignant epithelium, immune infiltrate, and stroma alike—rather than reflecting a single cell type.

The immune connections ran deep. Using CIBERSORT, ssGSEA, and ESTIMATE algorithms, the researchers found that higher risk scores correlated positively with macrophages, neutrophils, and activated memory CD4-positive T cells, but negatively with resting memory CD4 T cells, monocytes, resting dendritic cells, and memory B cells. Low-risk patients carried higher immune and stromal scores and showed greater abundance of infiltrating immune cells and immune pathway activity, consistent with stronger anti-tumor immunoreactivity. The high-risk group also carried a higher tumor mutation burden and expressed immune checkpoint molecules such as CTLA4, CD27, and CD28 at significantly different levels, along with several m6A RNA modification regulators including METTL3 and ALKBH5.

Perhaps most clinically consequential were the therapeutic predictions. TIDE and immunophenoscore analyses indicated that high-risk patients were less likely to respond to immune checkpoint inhibitors, while low-risk patients showed signatures of greater sensitivity to anti-PD1 and anti-CTLA4 treatment. Drug susceptibility modeling told a complementary story: patients with higher risk scores appeared more likely to benefit from EGFR tyrosine kinase inhibitors such as gefitinib, erlotinib, and osimertinib, as well as crizotinib, cisplatin, and 5-fluorouracil, whereas axitinib showed predicted efficacy in the low-risk group. The team also folded the risk score into a nomogram alongside standard clinical variables, achieving three- and five-year AUCs of 0.777 and 0.693 with well-calibrated predictions.

The authors are candid about the limitations. Only 61 lysosome-related genes were considered, a list that will surely grow as the field expands; the signature was built and validated retrospectively on public data; and laboratory experiments will be needed to confirm the biological roles of the less-studied genes in the panel, such as DLGAP5, NTSR1, BCAN, OLAH, ACSM5, SEC14L3, and ELOVL2. Still, the work demonstrates how a cellular organelle long viewed as little more than the cell’s garbage disposal can encode a prognostic fingerprint of a deadly cancer. If prospective studies bear out the findings, a simple gene expression test could one day help oncologists tailor lung adenocarcinoma treatment—choosing between immunotherapy, targeted drugs, and chemotherapy—based on the lysosomal signature written into each patient’s tumor.

Subject of Research: A lysosome-related gene signature for predicting prognosis, tumor microenvironment, and therapeutic response in lung adenocarcinoma

Article Title: Identification of a lysosome-related prognostic signature to predict prognosis, tumor microenvironment and therapeutic responses in lung adenocarcinoma

Article References: Yan, J., Liu, Z., Sun, X., & Xia, X. (2025). Identification of a lysosome-related prognostic signature to predict prognosis, tumor microenvironment and therapeutic responses in lung adenocarcinoma. Clinical Cancer Bulletin, 4(1), Article 2. https://doi.org/10.1007/s44272-025-00029-z

Image Credits: AI Generated

DOI: 10.1007/s44272-025-00029-z

Keywords: lung adenocarcinoma, lysosome, prognostic signature, tumor microenvironment, immunotherapy, LASSO regression, biomarker, TCGA, single-cell RNA sequencing, drug sensitivity, tumor mutation burden, nomogram