Colorectal cancer remains one of the most common and lethal cancers worldwide, and its notorious cellular heterogeneity has long frustrated efforts to understand why some tumors progress relentlessly while others respond to treatment. Tumors are not uniform masses of identical cells; they contain diverse populations of cancer cells, immune cells and stromal cells that communicate with one another and collectively shape disease course. Understanding which specific cell types harbor the molecular programs responsible for malignancy is therefore a central question in cancer biology, and answering it requires looking beyond conventional bulk analyses that average signals across thousands of mixed cells.

The research team assembled an extraordinary dataset to tackle this problem. They integrated data from 2,993 colorectal cancer samples spanning four complementary technologies: bulk RNA sequencing from 2,568 samples drawn from two overall survival and recurrence-free survival cohorts; single-cell RNA sequencing capturing 281,961 individual cells from 152 specimens; spatial transcriptomics from six samples, which preserves information about where genes are expressed within intact tissue; and proteomics from 267 samples, which measures the actual proteins produced by tumor cells. This integrated approach allowed the investigators to move from population-level associations down to individual cells and back up to clinically validated signatures, a strategy increasingly seen as the gold standard for dissecting tumor complexity.

Using computational methods to integrate and annotate the single-cell data, the researchers constructed a stage-stratified atlas of colorectal cancer and resolved eleven distinct malignant epithelial subsets within tumors. Among these, one cluster stood out. Designated Epi_4, this subset was enriched in late-stage tumors and carried strong signatures of epithelial-mesenchymal transition, the process by which epithelial cells acquire migratory and invasive properties; hypoxia, reflecting the low-oxygen conditions typical of growing tumors; and inflammatory programs. Critically, patients whose tumors showed high activity in this subset had significantly worse overall survival and recurrence-free survival across the bulk RNA sequencing cohorts.

The defining molecular marker of this aggressive subset proved to be GPRC5A, a gene encoding a G protein-coupled receptor, a class of cell-surface proteins renowned for their roles in cellular signaling and their historical success as drug targets. The researchers designated this population GPRC5A-positive epithelial cells. GPRC5A expression rose steadily from stage I through stage IV disease, tracking with tumor progression, and elevated levels were associated with poor outcomes across multiple independent cohorts. Spatial transcriptomics confirmed that GPRC5A-positive cells occupied specific locations within tumor tissue consistent with the single-cell findings, and proteomic measurements at the protein level corroborated the RNA-based observations, providing a rare degree of concordance across molecular layers.

Correlation alone, however, does not establish function. To test whether GPRC5A actively drives malignant behavior or merely marks it, the team performed CRISPR-based perturbation experiments in colorectal cancer cell lines, altering GPRC5A expression and observing the consequences. Disrupting the gene affected cell proliferation, migration and invasion, the hallmarks of metastatic potential. Immunoblotting revealed corresponding changes in epithelial-mesenchymal transition markers, indicating that GPRC5A influences the molecular machinery that governs cellular plasticity. In mouse xenograft models, manipulating GPRC5A altered tumorigenicity, the capacity of cancer cells to seed and sustain tumors in living tissue. Together, these experiments support the conclusion that GPRC5A is not simply a passive biomarker but a functionally important contributor to malignant phenotypes, at least in the models tested.

The investigators then turned their attention upward along the regulatory hierarchy, asking which molecular master switches control GPRC5A expression. Using SCENIC, a computational framework that infers transcription factor activity from single-cell expression data, combined with analysis of binding motifs in the JASPAR database, they identified FOSL1 as a candidate upstream regulator. FOSL1 belongs to the AP-1 family of transcription factors, well-established players in cancer cell proliferation, invasion and inflammation. Chromatin immunoprecipitation followed by quantitative PCR, a technique that detects whether a specific protein binds to a specific DNA sequence, provided experimental support that FOSL1 physically occupies the GPRC5A promoter region. This finding suggests a concrete regulatory pathway through which malignant epithelial states might be induced and maintained, and it raises the possibility that blocking this axis could suppress the aggressive cell population.

Perhaps the most intriguing dimension of the study concerns the tumor microenvironment, the ecosystem of non-cancerous cells that surrounds and interacts with tumors. Spatial analysis and ligand-receptor mapping, which predicts communication between cell types based on the expression of signaling molecules and their corresponding receptors, revealed a close physical and functional association between GPRC5A-positive epithelial cells and a population of cancer-associated fibroblasts marked by the expression of periostin, designated POSTN-positive fibroblasts. The computational analysis predicted reciprocal signaling between these two cell populations through several ligand-receptor pairs, including COL1A1 interacting with SDC4, COL1A1 and COL1A2 engaging ITGA2 and ITGB1, and PPIA binding BSG. Fibroblasts are known to remodel the extracellular matrix and secrete growth factors that support tumor growth, and this study suggests a potentially reciprocal dialogue in which epithelial cells and fibroblasts reinforce each other’s malignant behaviors. Importantly, patients whose tumors displayed concurrent high signatures of both GPRC5A-positive epithelial cells and POSTN-positive fibroblasts had the worst overall and recurrence-free survival, suggesting that this cellular partnership may be a powerful indicator of aggressive disease.

The translational implications of the work extend to drug response. Using OncoPredict, a computational tool that estimates drug sensitivity from gene expression profiles, the researchers found an association between GPRC5A status and sensitivity to trametinib, an FDA-approved MEK inhibitor used in other cancers. Molecular docking and molecular dynamics simulations produced a computational model of a possible direct interaction between trametinib and the GPRC5A protein, raising the speculative but tantalizing prospect that the drug might act partly through this receptor. The authors are appropriately cautious on this point, emphasizing that the docking model remains experimentally unvalidated and that direct binding studies will be required before any therapeutic conclusion can be drawn. Cell sensitivity assays provided additional exploratory support for the link between GPRC5A and trametinib response, but the researchers stress that this line of investigation is hypothesis-generating rather than definitive.

The study’s conclusions are carefully hedged in ways that reflect both its ambition and its limitations. The authors state that GPRC5A-positive epithelial cells represent a malignancy-associated state in colorectal cancer and that GPRC5A is functionally important for malignant phenotypes in the tested models, conclusions that are well supported by their convergent evidence. However, they explicitly note that the inferred relationships with POSTN-positive fibroblasts and the trametinib findings should be regarded as hypothesis-generating pending functional crosstalk experiments, direct binding validation and therapeutic testing. This level of rigor is notable in a field where single-cell findings are sometimes overinterpreted, and it sets a clear roadmap for follow-up studies: co-culture systems to test epithelial-fibroblast signaling, biophysical assays to confirm or refute trametinib binding to GPRC5A, and ultimately clinical evaluation of GPRC5A as a biomarker for patient stratification.

The work also received approval from institutional ethics committees and was conducted in accordance with the Declaration of Helsinki, with written informed consent obtained from all participants and animal procedures reviewed by the appropriate ethics board. Supported by funding from Anhui Provincial and Bengbu Municipal research programs, the study exemplifies how relatively modest clinical research institutions can now leverage large public datasets and advanced molecular platforms to make contributions of genuine translational significance. If subsequent studies validate the central role of the GPRC5A-positive epithelial state and its interaction with the stromal compartment, the findings could eventually inform diagnostic tests that identify high-risk patients and therapeutic strategies aimed at disrupting the epithelial-fibroblast axis or exploiting the drug sensitivity patterns uncovered here. For now, the study stands as a compelling demonstration of how multi-omics integration can transform a heterogeneous tumor mass into a legible map of malignant cell states, communication networks and therapeutic vulnerabilities.

The findings, published in Nature Communications, address one of the central problems in glioblastoma research: no two tumour cells are necessarily alike. Even within a single patient, cancer cells can differ in their growth patterns, biological behaviour and response to treatment. This cellular variation is one reason glioblastoma frequently returns after surgery, radiotherapy and chemotherapy. By examining the genetic activity of individual cells rather than treating a tumour as a uniform mass, the researchers were able to study its complexity at a far higher resolution.

The key focus of the investigation was isoforms, alternative versions of the molecular messages produced from the same gene. When a gene is activated, its DNA sequence is copied into RNA, which can then be processed in different ways. These alternative RNA molecules may produce proteins with distinct structures and functions. In cancer, changes in isoform production can influence how cells divide, invade surrounding tissue, resist treatment or interact with the immune system. Some isoforms may also create abnormal protein fragments that mark cancer cells for immune attack.

Until recently, scientists have struggled to investigate these molecules in individual tumour cells because standard single-cell sequencing technologies generally read only short fragments of RNA. Those fragments can reveal which genes are active, but they often cannot show the complete structure of a transcript. As a result, different isoforms may be difficult to distinguish, and rare tumour-specific versions can remain invisible. Long-read sequencing overcomes this limitation by reading much longer RNA molecules, allowing researchers to identify full-length transcripts and connect their molecular components accurately.

In this study, the team analysed data from more than 210,000 individual cells collected from 27 patients with glioblastoma. The analysis included malignant cells as well as immune cells and stromal cells in the tumour microenvironment—the network of non-cancerous cells and tissues that surrounds and supports a tumour. This broad approach enabled the researchers to examine not only the genetic diversity of the cancer itself, but also the molecular relationships between tumour cells and their local biological environment.

The resulting map contained thousands of previously unannotated isoforms, meaning genetic transcripts that had not been recorded in existing reference databases. Many of these newly identified isoforms appeared exclusively in tumour cells and were absent from healthy tissues examined in the analysis. Their restricted distribution is particularly important for drug development: a target found only in cancer cells could, in principle, be attacked without causing the widespread damage associated with targets shared by healthy organs.

The researchers then assessed whether some of the tumour-specific isoforms might generate neoantigens—abnormal protein fragments that can be displayed on the surface of cells by molecules known as major histocompatibility complex class I, or MHC class I. These molecules act like biological presentation platforms. They bind fragments derived from proteins inside a cell and display them to immune cells, including cytotoxic T cells. If the displayed fragment appears abnormal, T cells may recognise the cell as dangerous and destroy it.

A subset of the newly discovered isoforms was predicted to produce peptides capable of binding strongly to MHC class I molecules. This does not yet prove that the peptides trigger an effective immune response in patients, but it identifies candidates for further laboratory and clinical testing. If validated, such molecules could expand the supply of neoantigens available for personalised cancer vaccines, in which a vaccine is designed around the unique molecular features of an individual patient’s tumour. They could also support the development of engineered T-cell therapies or other precision immunotherapy strategies.

The implications are significant, but the discovery is not an immediate change to glioblastoma treatment. The predicted neoantigens must first be confirmed experimentally, including tests showing that they are naturally produced by tumour cells, presented on MHC molecules and recognised by human immune cells. Researchers will also need to determine how consistently these isoforms occur among patients and whether cancer cells can lose them under immune pressure. Nevertheless, the study provides a new route into a previously inaccessible layer of tumour biology, potentially revealing targets missed by gene-level and short-read analyses.

The project also demonstrates the value of combining clinical expertise, brain tumour surgery, oncology and population-scale genomics. By linking long-read sequencing with single-cell analysis, the team established and validated an analytical framework that could be applied to other cancers and clinical datasets. The researchers say the work strengthens Hong Kong’s role in translational genomics and precision medicine. For glioblastoma, a disease that has resisted many conventional therapeutic approaches, the newly exposed world of tumour-specific isoforms may offer one of the most promising sources of future targets for personalised vaccines and immune-based treatments.

Subject of Research: Glioblastoma and tumour-specific isoform diversity

Article Title: Mapping glioblastoma’s isoform diversity using long-read single-cell analysis

News Publication Date: 23 April 2026

Web References: https://www.nature.com/articles/s41467-026-72258-2

References: Nature Communications, DOI: 10.1038/s41467-026-72258-2

Image Credits: The University of Hong Kong

Keywords: glioblastoma, brain cancer, long-read sequencing, single-cell analysis, isoforms, neoantigens, personalised cancer vaccines, immunotherapy, precision medicine, tumour microenvironment

The therapeutic landscape surrounding STING agonists has long been characterized by a paradox: potent preclinical efficacy often fails to translate into consistent clinical success. The heterogeneity in tumor responsiveness has perplexed oncologists and researchers alike, prompting a rigorous examination of intra-tumoral molecular circuits that might elucidate these disparities. Predominantly, prior focus gravitated toward STING’s modulation of the tumor immune microenvironment, sometimes neglecting the autonomous tumor-cell-intrinsic pathways that potentially dictate sensitivity or resistance to these agonists.

A collaborative team of scientists spanning several prestigious Chinese institutions—including the First Medical Center of the Chinese PLA General Hospital and Peking University School of Life Sciences—sought to decode this enigma at a molecular resolution. Their seminal study, published in the renowned journal Cancer Biology & Medicine in late 2025, meticulously delineates how STING activation precipitates a STAT1-dependent transcriptional cascade culminating in the upregulation of HMGN2, which in turn orchestrates tumor cell apoptosis. These findings spotlight the critical role of tumor-intrinsic signaling cascades in mediating therapeutic outcomes and offer a mechanistic framework for leveraging STING agonists with newfound precision.

Experimental interrogation began with a comprehensive survey of diverse tumor cell lines to gauge their susceptibility to STING agonist treatment. Strikingly, neural-origin tumors—specifically neuroblastoma and glioblastoma—exhibited robust proliferative inhibition upon exposure to the synthetic STING agonist SR-717. Contrastingly, several non-neural carcinomas demonstrated scant response, underscoring the tissue-specific nature of this effect. Furthermore, murine xenograft models recapitulated these results, even under immunodeficient conditions, suggesting that the antitumor activity transcends conventional immune cell engagement and hinges upon intrinsic tumor cell pathways.

To unravel the molecular architecture underlying this phenomenon, the researchers deployed high-throughput RNA sequencing, which unveiled a constellation of differentially expressed genes post-STING activation. Central among these was HMGN2, a nucleosome-binding protein previously implicated in chromatin remodeling but now unveiled as a linchpin in apoptotic execution. Functional assays confirmed that overexpressing HMGN2 in tumor cells independently triggered apoptosis, whereas CRISPR-mediated HMGN2 knockout abrogated the antitumor efficacy of STING agonists, decisively establishing causality.

The axis linking STING activation and HMGN2 transcription was further explored through mechanistic studies revealing the indispensable role of STAT1, a well-characterized transcription factor commonly downstream of interferon signaling. Chromatin immunoprecipitation assays elegantly demonstrated STAT1’s direct binding to promoter regions of HMGN2, forming a previously uncharacterized STING–STAT1–HMGN2 signaling axis. This pathway operates as an autonomous molecular conduit within tumor cells, effecting apoptotic programs that culminate in tumor suppression independent of canonical immune responses.

Beyond mechanistic insights, bioinformatic analyses leveraging publicly accessible oncologic datasets reinforced the clinical relevance of HMGN2. Elevated HMGN2 transcript levels correlated positively with improved patient survival across various malignancies, positioning this protein not only as a therapeutic mediator but also as a valuable prognostic biomarker. This dual role opens the tantalizing prospect of employing HMGN2 expression as a stratification tool to identify patients most apt to benefit from STING-targeted interventions.

The implications of this research resonate deeply within the field of precision oncology. Historically, the application of STING agonists has been hamstrung by unpredictable patient responses and an absence of reliable biomarkers. By shifting the paradigm to acknowledge tumor cell-intrinsic regulation as a determinant of efficacy, clinicians may soon tailor treatments guided by molecular profiling of the STING–STAT1–HMGN2 axis. Such stratification promises to optimize therapeutic indices, maximize patient benefit, and reduce unwarranted exposure to ineffective regimens.

The neuro-oncology domain stands to gain substantially from these findings. Given the aggressive nature and immunosuppressive microenvironment characteristic of nervous system tumors such as glioblastoma, therapeutic strategies that circumvent immune evasion mechanisms are critically needed. STING agonists capable of directly engaging tumor-intrinsic apoptotic pathways may represent a paradigm shift, offering a novel modality to target these recalcitrant malignancies with enhanced specificity and potency.

Looking forward, this discovery invites exploration into combination therapies designed to potentiate HMGN2 expression or restore the sensitivity of resistant tumors to STING modulation. Agents that synergize with STAT1 activation or chromatin remodeling may amplify apoptotic cascades, thereby broadening the therapeutic window. Additionally, the universality of the STING–STAT1–HMGN2 axis suggests potential applicability beyond neural tumors, warranting investigation across diverse cancer types where STING pathway engagement remains underexploited.

This transformative insight into the molecular determinants governing tumor responses to STING agonists underscores the evolution of cancer immunotherapy from a primarily immune-centric strategy to an integrated approach encompassing direct tumor cell targeting. It exemplifies how dissecting intracellular signaling networks can reveal critical vulnerabilities previously masked by the complexity of tumor-immune interactions. Ultimately, these advances herald a future wherein STING agonists transcend their experimental roots to become tailored, mechanism-driven weapons in the oncologist’s arsenal.

As the clinical community wrestles with the formidable challenge of durable, effective cancer treatments, the STING–STAT1–HMGN2 axis emerges as a beacon illuminating new therapeutic and diagnostic horizons. By harnessing the intrinsic apoptotic machinery within tumor cells, this pathway offers a compelling template for the next generation of immunotherapeutic innovations. The journey from bench to bedside is poised to leverage these findings, translating molecular discovery into tangible clinical impact with the promise of improved patient outcomes.

Subject of Research: Not applicable

Article Title: Antitumor effects of STING agonists on nervous system tumors via tumor-intrinsic STING-STAT1-mediated HMGN2 expression

News Publication Date: 30-Oct-2025

References:
DOI: 10.20892/j.issn.2095-3941.2025.0326

Image Credits: Cancer Biology & Medicine

Keywords: Agonists, STING pathway, STAT1, HMGN2, tumor-intrinsic signaling, nervous system tumors, neuroblastoma, glioblastoma, apoptosis, immunotherapy, cancer biomarkers, precision oncology