Immunotherapy with checkpoint inhibitors has transformed the treatment of advanced clear cell renal cell carcinoma (ccRCC), producing durable responses in some patients. However, a substantial proportion of patients do not benefit from these treatments, and predicting who will respond before starting therapy remains a major clinical challenge.
This study used single-cell RNA sequencing (scRNA-seq), a technology that measures gene activity in individual cells within a tumor, to investigate why some ccRCC tumors resist immunotherapy. By examining thousands of individual cells from multiple tumors, the researchers identified a previously unrecognized population of cancer cells marked by high expression of a gene called CCL2.
CCL2 encodes a protein called monocyte chemoattractant protein-1 (MCP-1), a signaling molecule that attracts monocytes and macrophages to sites of inflammation. When expressed by cancer cells themselves, CCL2 can reshape the immune environment within the tumor in ways that suppress anti-tumor immunity and reduce the effectiveness of checkpoint inhibitor treatments.
Using the biological insights from single-cell analysis, the researchers built a machine learning-based prognostic scoring model incorporating CCL2-associated genes. This model predicts both overall survival and likely response to immunotherapy, offering a potential tool for personalizing treatment decisions in ccRCC.
Conventional bulk RNA sequencing measures average gene activity across thousands of cells in a tissue sample, which obscures the cellular heterogeneity within tumors. Tumors contain many different cell types, including cancer cells, immune cells, blood vessel cells, and structural support cells, each with distinct gene expression profiles. Single-cell RNA sequencing overcomes this limitation by profiling each cell individually.
In ccRCC, prior bulk sequencing studies had identified general patterns of gene expression associated with outcomes, but could not determine which specific cell types drove those patterns. The scRNA-seq approach in this study allowed researchers to map gene expression to specific cell populations, revealing that CCL2-high cancer cells form a distinct subpopulation with different properties from CCL2-low cancer cells in the same tumor.
The study analyzed scRNA-seq data from multiple publicly available ccRCC datasets, collectively representing samples from dozens of patients. Using computational methods to integrate data across datasets, the researchers were able to identify consistent patterns of CCL2+ cell biology across different patients and tumor contexts, reducing the risk that findings reflect a single unique tumor rather than a general phenomenon.
The analysis also incorporated bulk RNA sequencing data from the TCGA-KIRC cohort and clinical outcome data to link the single-cell findings to patient survival. This integration of single-cell biological insight with population-level clinical data is a powerful approach for translating molecular discoveries into clinically actionable findings.
After clustering and annotating the single-cell data by cell type, the researchers identified a subset of neoplastic (cancer) cells characterized by high CCL2 expression. These CCL2+ cancer cells were distinct from other cancer cells in the same tumors in their gene expression patterns, suggesting they have adopted a different cellular state or identity with different functional properties.
Analysis of cell-to-cell communication networks within the tumor microenvironment showed that CCL2+ cancer cells are major sources of signaling molecules that attract and activate macrophages. Specifically, CCL2 produced by these cancer cells draws in monocytes that differentiate into tumor-associated macrophages (TAMs), immune cells that commonly adopt immunosuppressive roles within tumors rather than attacking cancer cells.
Tumors with higher proportions of CCL2+ cancer cells showed a more immunosuppressive microenvironment overall, with higher levels of regulatory T cells and immunosuppressive macrophages and lower levels of active cytotoxic T cells. This immune landscape is associated with reduced immune surveillance and decreased likelihood of responding to checkpoint inhibitor immunotherapy.
Trajectory analysis, which computationally reconstructs the developmental history of cells within the tumor, suggested that CCL2+ cancer cells represent a distinct cellular state that may arise from the acquisition of specific mutations or epigenetic changes. Understanding how these cells develop could inform strategies to prevent their emergence or target them specifically once present.
To translate the single-cell findings into a clinically applicable tool, the researchers identified genes associated with the CCL2+ cancer cell state and used these as the starting point for building a prognostic model. Two machine learning methods were applied: LASSO (Least Absolute Shrinkage and Selection Operator) regression for gene feature selection and a survival support vector machine (SVM) for prediction.
LASSO regression reduces a large number of potential predictor genes to a smaller subset of the most informative ones by penalizing variables that do not improve prediction. This step is critical for preventing overfitting and for identifying a parsimonious gene signature that could realistically be measured in clinical practice using standard gene expression assays.
The survival SVM was then trained to predict overall survival using the LASSO-selected gene features. Unlike standard classification, survival models must account for censored data, where some patients are still alive at the end of follow-up and have not yet experienced the outcome of interest. The survival SVM handles this appropriately while leveraging the discriminative power of the SVM framework.
The final model achieved a concordance index (C-index) of 0.650 in the validation cohort. While this represents modest but meaningful discriminative ability (0.5 would be random chance, 1.0 would be perfect prediction), the model provides information beyond conventional clinical staging and may be most valuable when combined with other prognostic factors rather than used in isolation.
Pathway analysis of the CCL2+ cancer cell gene expression profile highlighted the Notch signaling pathway as a central regulatory mechanism in these cells. Notch signaling is a fundamental developmental pathway that controls cell fate decisions, differentiation, and the adoption of stem-like properties. Its activation in cancer has been linked to treatment resistance and immunosuppression.
In ccRCC specifically, Notch signaling has been associated with tumor progression and with the adoption of mesenchymal-like characteristics that enable cancer cells to become more invasive and resistant to immune attack. The enrichment of Notch pathway activity in CCL2+ cells suggests that this pathway may be driving the immunosuppressive properties of this cell subpopulation.
Notch pathway inhibitors are being investigated in clinical trials across multiple cancer types. The identification of Notch as a key pathway in CCL2+ ccRCC cells raises the hypothesis that combining Notch inhibitors with immunotherapy might overcome the immunosuppressive barrier created by this cell population. This represents a testable treatment strategy for future preclinical and clinical investigation.
The connection between Notch signaling, CCL2 expression, and macrophage recruitment creates a coherent biological narrative: Notch-activated cancer cells upregulate CCL2, which recruits immunosuppressive macrophages, which in turn create an environment that protects both the CCL2+ cancer cells and the broader tumor from immune attack. Breaking any link in this chain could potentially restore immune sensitivity.
The central clinical implication of this study is that tumors with high CCL2+ cancer cell content are likely to be less responsive to standard checkpoint inhibitor immunotherapy. If this finding is validated in prospective studies, measuring CCL2 or the full CCL2+ gene signature in tumor biopsies could become a tool for stratifying patients before starting immunotherapy.
Patients with low CCL2+ scores might be preferentially offered immunotherapy, while those with high scores might benefit from approaches that first address the immunosuppressive microenvironment, such as therapies targeting the CCL2-macrophage axis or Notch inhibitors, before or alongside checkpoint inhibitors. This kind of biomarker-guided treatment selection is the foundation of precision oncology.
The CCL2-macrophage axis itself represents a potential therapeutic target. Drugs that block CCL2 signaling or deplete or reprogram immunosuppressive macrophages are under investigation in various cancers. Applying these approaches to ccRCC patients identified as having high CCL2+ burden could convert a resistant tumor microenvironment into one that is more amenable to immune attack.
For patients currently receiving or considering immunotherapy for advanced ccRCC, it is worth discussing with their oncologist whether tumor biomarker testing, including assessment of tumor immune microenvironment characteristics, has been performed. While CCL2-based testing is not yet standard practice, the underlying concepts of immune microenvironment assessment are increasingly incorporated into treatment decision frameworks at specialized centers.
This study demonstrates the power of single-cell RNA sequencing to uncover biologically meaningful cancer cell subpopulations that are invisible to conventional bulk analysis approaches. The discovery of the CCL2+ neoplastic cell population provides a new mechanistic framework for understanding immunotherapy resistance in ccRCC.
The integration of single-cell biology with machine learning prognostic modeling and clinical outcome data produces a multi-layered body of evidence connecting molecular cell biology to patient prognosis. Each layer of evidence reinforces the others: CCL2+ cells create immunosuppression in single-cell data, tumors enriched for CCL2-associated genes have worse outcomes in population data, and the model predicts survival at a level better than chance.
Future research priorities include prospective validation of the CCL2+ scoring model in patients receiving immunotherapy, experimental testing of whether CCL2 blockade or Notch pathway inhibition can sensitize ccRCC tumors to immunotherapy in preclinical models, and development of standardized assays for measuring CCL2+ cell burden in clinical biopsy samples.
The broader lesson of this work is that tumors are not homogeneous masses but complex ecosystems of interacting cell types. Understanding the role of specific cancer cell subpopulations in shaping the immune environment is increasingly recognized as essential for predicting treatment response and designing effective combination therapies. Studies like this one are building the foundation for that understanding in kidney cancer.