Integrated spatial transcriptomics, single-nucleus RNA sequencing, and histopathology to reveal immune-suppressive niches in colorectal cancer liver metastases.

e15544 Background: Progression of colorectal cancer (CRC) is shaped by interactions between tumor, immune, and stromal cells, particularly within the liver metastases (LM) microenvironment. Spatial transcriptomics enables in situ dissection of these interactions within intact tissue architecture. We analyzed matched primary and LM CRC samples to identify liver-specific spatial programs associated with immune suppression and clinical outcome. Methods: Patients (pts) undergoing resection of primary CRC and LM were retrospectively selected and profiled with the spatial transcriptomic (ST) platform CosMx Spatial Molecular Imager with a 1k plex panel (Bruker Spatial) and single-nucleus RNA sequencing (snRNA-seq) with the Flex kit (10x Genomics) within a partner-of-choice framework. Tissue sections were annotated to define reference mucosa, premalignant lesions, and tumor regions. Rigorous QC removed low-quality cells, segmentation artifacts, retaining high-confidence spatially resolved cells. Spatial domains and neighbourhood interactions were defined using graph-based approaches and analyzed by cell-type enrichment, molecular status, and clinical prognosis (px) (classified as good or bad based on OS ( > or 6 months). Results: Eighteen pts were included; 14 pts had stage IV disease (77.8%); 11 pts (61.1%) were classified as good px (Table). Cohort-level snRNA-seq analysis identified site-specific transcriptional programs, with LM enriched for IFN-low states and EMT, squamous, and fetal progenitor programs. Bad px pts clustered exclusively within IFN-low groups, whereas IFN-high profiles, characterized by higher CD74 expression, were restricted to good px pts. ST profiling revealed macrophage-rich, perivascular immunosuppressive niches in LM, preferentially localized at tumor borders in bad px pts. Increased MAPK pathway activity was observed in RAS-mut metastatic tumor cells, with spatial immune and stromal organization varying by px and treatment. Conclusions: Integrated snRNA-seq and ST analyses show that CRC LM are characterized by IFN-low, EMT/fetal-like transcriptional programs and spatially organized, macrophage-rich perivascular immune suppressive niches associated with bad prognosis and RAS status, providing biological insight into immune evasion mechanisms in CRC pts. Pt characteristics (n=18). Good Prognostic 11 (61) Bad Prognostic7 (39) Sex Fem 4 (36) 4 (57) Male 7 (64) 3 (43) Stage I 1 (9) 0 II 1 (9) 1 (14) III 1 (9) 0 IV 8 (73) 6 (86) Age Median (range) 61 (30-74) 60 (52-75) Molecular Status RAS mut 2 (18) 3 (43) BRAF mut 1 (9) 1 (14) wt 8 (73) 3 (43) MMR MSS 11 6 (86) MSI 0 1 (14) Site Right 2 (18) 4 (57) Left 9 (82) 3 (43) Treatment before surgery LM CT<