Intrahepatic cholangiocarcinoma (ICC) is a bile duct cancer arising within the liver. It carries a very poor prognosis because most patients present at an advanced stage, and existing chemotherapy offers limited benefit. Immunotherapy has transformed other cancers, but ICC responses have been disappointing.
Tumor microenvironment (TME) heterogeneity is believed to explain why only some ICC patients respond to immunotherapy. The mix of immune cells surrounding and infiltrating the tumor varies enormously between patients, and this variation likely determines whether a patient's immune system can be activated to fight the cancer.
The study goal was to use gene expression profiling across a large collection of ICC tumors to classify the immune landscape into distinct subtypes - much like how breast cancer is classified into molecular subtypes - with the aim of identifying which patients are most likely to benefit from checkpoint immunotherapy.
Clinical impact is direct: by knowing a patient's ICC immune subtype from a biopsy, oncologists could select patients most likely to respond to PD-1/PD-L1 blocking antibodies, avoiding ineffective treatment in non-responding subtypes.
The MCP-counter algorithm was used to estimate the abundance of eight immune and stromal cell types (T cells, cytotoxic T cells, NK cells, B cells, monocytes, myeloid dendritic cells, neutrophils, endothelial cells, and fibroblasts) directly from bulk RNA-seq gene expression data, without needing single-cell sequencing.
Multi-cohort analysis integrated 566 ICC tumor samples from multiple independent datasets spanning Asian and Western patient populations, providing statistical power and cross-validation across different genomic platforms.
Consensus clustering was applied to the cell composition profiles to identify recurring immune patterns. The algorithm identified four stable immune clusters that appeared consistently across multiple independent datasets.
Molecular and clinical annotation of each immune subtype included analysis of oncogenic mutations, copy number alterations, FGFR2 fusions, IDH1/2 mutations, PD-L1 expression, and patient survival outcomes.
Subtype I1 (Immune Desert) is characterized by very low infiltration of all immune cell types - tumors essentially exclude immune cells. This subtype has the worst prognosis because there are no immune cells available to attack the cancer, and checkpoint inhibitors have nothing to reactivate.
Subtype I2 (Immunogenic) shows high infiltration of cytotoxic T cells, NK cells, and activated immune responses alongside high PD-L1 expression. This is the most promising subtype for checkpoint immunotherapy - the immune system is engaged but being blocked by PD-L1, which an anti-PD-1 antibody could unblock.
Subtype I3 (Myeloid) is dominated by monocytes, macrophages, and other myeloid lineage cells rather than T cells. These tumors have an immunosuppressive environment driven by myeloid cells, and may respond better to strategies targeting macrophage reprogramming rather than T cell checkpoints.
Subtype I4 (Mesenchymal) is enriched in fibroblasts and features a highly fibrotic stroma alongside immune exclusion. This subtype is associated with the mesenchymal ICC molecular class and may require anti-fibrotic combination strategies to allow immune access to the tumor.
FGFR2 fusions are enriched in the I2 immunogenic subtype, which is clinically important because FGFR2 inhibitors (pemigatinib, infigratinib) are already approved for ICC. Patients in I2 might benefit from combining FGFR inhibitors with checkpoint immunotherapy.
IDH1/2 mutations are distributed differently across subtypes - IDH-mutated tumors tend to be enriched in subtypes with lower immune infiltration, consistent with reports that IDH-mutant tumors may suppress immune responses through epigenetic mechanisms.
TP53 mutations associate more strongly with immunosuppressive subtypes, suggesting that loss of p53 function may contribute to immune evasion mechanisms in ICC.
Tumor mutational burden (TMB) was highest in the I2 subtype, consistent with the known relationship between higher mutation burden and improved immunotherapy response across cancer types.
Patient selection for PD-1 trials should prioritize I2 subtype patients - this group has all the hallmarks of checkpoint-responsive tumors: high TILs, high PD-L1, high TMB, and existing immune engagement.
I3 subtype patients may be better served by clinical trials targeting myeloid cells, such as CSF1R inhibitors that deplete immunosuppressive macrophages, or CXCR2 inhibitors that block neutrophil recruitment.
I4 subtype intervention could focus on combination anti-fibrotic strategies (FAK inhibitors, TGF-beta blockers) to break down the stromal barrier before immunotherapy.
Immune subtyping in clinical practice could be performed on standard tumor biopsies using a gene expression signature panel, making this stratification accessible without complex single-cell technologies.
Prospective clinical validation is the critical next step - the four-subtype classification needs to be tested in ongoing ICC clinical trials to confirm that I2 patients actually respond better to anti-PD-1 agents.
Single-cell RNA sequencing of ICC tumors would provide higher resolution insight into the specific cell subpopulations within each immune subtype, potentially revealing additional therapeutic vulnerabilities.
Spatial transcriptomics could reveal how the immune cell geography within the tumor (cells at the invasive margin versus tumor core) differs across subtypes and correlates with clinical outcomes.
Longitudinal immune monitoring - tracking how a patient's immune subtype evolves during treatment - could reveal mechanisms of acquired resistance to checkpoint immunotherapy in ICC.