Spatial immune landscape in metastatic colorectal cancer

Microsatellite stable (MSS) metastatic colorectal cancer (mCRC) does not respond well to immunotherapy, but clinical trials have shown promise with immunotherapy in MSS mCRC without liver metastases, especially lung only metastatic disease. To understand the underlying reasons for this site-specific difference in response, we investigated the spatial immune landscape in liver and lung metastases using spatial transcriptomics (Xenium) and single nuclear RNA-sequencing (snRNA seq) on immunotherapy naive, MSS tumors (9 primary, 9 liver and 6 lung samples for Xenium and 17 primary, 21 liver and 5 lung samples for snRNA seq). We defined three spatial regions in Xenium samples- tumor region (contains tumor cell nests), peritumor region (stromal and immune cells that separate tumor region from adjacent parenchyma) and adjacent parenchyma (adjacent normal-appearing tissue in the same section). Expressed as a proportion of all immune cells, liver tumor regions had higher macrophage infiltration (p<0.01), while lung tumor regions had higher CD8 T cell infiltration (p=0.03). CD8 T cell proportion was not different between liver and lung metastases in peritumor regions or adjacent parenchyma, or in the sn-RNA seq data, highlighting the spatial gradient of CD8 T cells in different organ sites. CD8 T cell density (per mm2) in liver tumor regions (but not peritumor region or adjacent parenchyma) was lower than paired primary colon (p=0.03) and lung metastasis (not significant). Tumor regions had a higher proportion of exhausted CD8 T cells (expressing at least one exhaustion marker- PDCD1, HAVCR2, TIGIT, CTLA4, LAG3 ) than the peritumor region or adjacent parenchyma in all three organs. Among T cells that infiltrated tumor regions, liver had a higher Treg proportion than lung (p=0.02), this did not hold true in peritumor region and adjacent parenchyma. Differential gene expression between Tregs of liver and lung tumor regions revealed upregulation of ENTPD1 (codes for CD39, which suppresses CD8 T cells) in liver Tregs; this finding was also present in our sn-RNA seq data. To investigate the reasons for organ-specific variability in the immune landscape of mCRC, we compared the expression of chemokine genes in non-tumor cells of liver and lung metastases. Liver samples had markedly higher levels of CXCL12 (highest in the adjacent parenchyma) than lung metastases. Tregs and CD8 T cells in the liver had higher levels of CXCR4 (CXCL12 receptor) than the lung. Spatial gradients in CXCL12 (highest in adjacent parenchyma), may impair CD8 T cell infiltration in the tumor. In summary, we identified an immunosuppressive immune microenvironment in liver than lung mCRC with unique spatial patterns of CD8 T cell and Treg infiltration, and chemokine expression. This may explain the differential response of immunotherapy between these two sites. Inhibition of the CXCL12-CXCR4 axis and CD39 may be potential strategies to augment immunotherapy in patients with liver mCRC.