A machine learning pipeline that combines multiple omics datasets was used to derive a new transcriptomic footprint correlated with positive outcomes in patients with advanced kidney cancer who underwent immunotherapy. The approach identified the molecular characteristics of an immune signaling hub distinguished by a human leukocyte antigen (HLA) repertoire with a higher preference for tumoral neoantigens. This approach has the potential to advance the field by utilizing multimodal omics and spatial analyses and developing new immune-community-driven biomarkers for the practical management of patients with renal cell carcinoma.
The research article “A spatial architecture-embedding HLA signature to predict clinical response to immunotherapy in renal cell carcinoma” was published in Nature Medicine.
Cross-omics, cross-cohort, and cross-species efficacy
Immunotherapy is highly attractive for advanced renal cell carcinoma because it significantly infiltrates CD8+ T cells. Unlike other forms of cancer that can be effectively treated with immune checkpoint blockade, advanced renal cell carcinoma exhibits atypical immunological features. Advanced renal cell carcinoma, for example, has a relatively low tumor mutational burden, similar to nonimmunogenic tumors. Moreover, a significant association exists between increased levels of immune cells infiltrating advanced renal cell carcinoma and unfavorable results.
Although immune checkpoint blockade has been clinically approved for treating advanced renal cell carcinoma, there is currently a lack of approved biomarkers to help select patients or guide effective immunotherapy combinations. Many biomarkers used to assess the effectiveness of immune checkpoint blockade in advanced renal cell carcinoma have low accuracy or show inconsistencies across different studies. Therefore, it is still unclear whether the biomarkers confirmed in recent advanced renal cell carcinoma immunotherapy trials are effective in real-world medical settings.
Scientists from KU Leuven in Belgium analyzed a large dataset of 220 patients with advanced renal cell carcinoma who received immune checkpoint blockade treatment. A cross-omics machine learning pipeline helped find a new tumor transcriptomic signature that integrates germline HLA characteristics with a distinct spatial community of CD8+ T cells and tumor-associated macrophages (TAMs), showing immunogenic interactions that predict the response to immunotherapy in patients with advanced kidney cancer. This footprint was then validated at single-cell and spatial resolutions. This machine learning signature correlated with positive outcomes following immune checkpoint blockade treatment in the initial cohort, with 12 other cohorts of 1,377 patients.
After using this signature to bridge human-to-mouse tumor transcriptomes, the researchers found that a rationally designed combination immunotherapy that stimulated both CD8+ T cells (through anti-PD1) and macrophages (via CD40 agonism) was needed to achieve maximal tumor control. More specifically, using a mouse model of renal adenocarcinoma, researchers found that combining PD1 blockade and CD40 agonism, which induce dendritic cell activation to enhance the activation of anti-tumor T cells and reprogram macrophages to eliminate tumor stroma, increased the activity of proinflammatory TAMs and CD8+ T cells.
Prospective clinical studies will be necessary to clarify the prognostic versus predictive impact of the molecular signature.
