Hypothesis\n\nIn immune‑hot breast tumors, the spatial proximity of MIF‑high cancer cells to CD99‑high macrophages creates immunosuppressive niches that blunt PD‑1/CTLA4 blockade efficacy, independent of overall T‑cell infiltration.\n\n## Mechanistic Rationale\n\n- MIF secreted by cancer cells binds CD99 on stromal macrophages, triggering a shift toward an M2‑like phenotype.\n- These macrophages release IL‑10, TGF‑β and CCL2, which (a) upregulate PD‑L1 on neighboring tumor cells and (b) recruit regulatory T cells, fostering a local immune‑suppressive microdomain.\n- Spatial transcriptomics shows that cytokine gradients decay over ~50 µm; thus only tumor regions within this radius experience heightened checkpoint expression and T‑cell exhaustion.\n- This explains why bulk RNA‑seq detects high PD‑1/CTLA4 (immune‑hot signature) yet TIDE scores remain high: the immunosuppressive signal is spatially restricted and missed by bulk averaging.\n\n## Testable Predictions\n\n1. Multiplexed immunofluorescence or imaging mass cytometry will reveal a positive correlation between the shortest distance from MIF+ cancer cells to CD99+ macrophages and local PD‑L1/MHC‑II ratios within the same tumor section.\n2. Patients whose tumors have >30 % of MIF+ cancer cells within 30 µm of a CD99+ macrophage will exhibit significantly lower objective response rates to anti‑PD‑1/CTLA4 therapy than those with larger intercellular distances, even when overall CD8+ T‑cell density is high.\n3. Pharmacologic blockade of CSF1R (which reduces macrophage survival) or genetic knockdown of MIF in cancer cells will increase the average MIF‑to‑CD99 distance and sensitize tumors to checkpoint inhibition in xenograft models.\n\n## Experimental Approach\n\n- Cohort: Retrieve pretreatment biopsies from a published immune‑hot breast cancer cohort (n ≈ 120) with known response to PD‑1/CTLA4 inhibitors.\n- Imaging: Perform CODEX or multiplexed IF for MIF, CD99, PD‑L1, CD8, FoxP3, and cytokeratin; use open‑source tools (e.g., histocat) to compute nearest‑neighbor distances and local ligand‑receptor interaction scores.\n- Analysis: Logistic regression linking the proportion of MIF+ cells within 30 µm of a CD99+ macrophage to binary response, adjusting for TIDE score, CD8 density, and tumor mutational burden.\n- Validation: In orthotopic breast cancer models (e.g., 4T1) engineered to overexpress MIF, treat with CSF1R inhibitor (pexidartinib) ± anti‑PD‑1; measure tumor growth, macrophage spatial distribution (imaging), and survival.\n\nIf the spatial MIF/CD99 metric predicts resistance and its disruption restores sensitivity, the hypothesis is supported; lack of correlation or no effect of niche disruption would falsify it.\n\nReferences\n- Multi-omics integration analysis unveils heterogeneity in breast cancer Multi-omics integration analysis unveils heterogeneity in breast cancer\n- MOSAIC platform integrating spatial/single-cell transcriptomics with AI MOSAIC platform integrating spatial/single-cell transcriptomics with AI\n- Spatial tumor heterogeneity in pancreatic cancer facilitates subtype co‑existence and therapy response Spatial tumor heterogeneity in pancreatic cancer facilitates subtype co‑existence and therapy response