Endothelial ICAM-1 Polarization Traps T Cells at Melanoma Tumor Edges

Scientists have uncovered a vascular mechanism that helps explain why some melanomas keep immune cells out of the tumor interior despite being crowded with them at the margins. In a study published in EMBO Molecular Medicine, researchers report that a single adhesion molecule on tumor blood vessels, ICAM-1, becomes spatially concentrated at the tumor periphery as melanomas grow, and that this polarization acts as a checkpoint governing where CD8+ T cells travel within the tumor. Blocking that molecule in mice redistributed T cells toward the tumor core, boosted their cytotoxic activity, slowed tumor growth, and made an otherwise treatment-resistant melanoma respond to anti-PD-1 immunotherapy.

The work began with a question that has long frustrated oncologists: why do tumors that appear inflamed with T cells sometimes fail to respond to immune checkpoint blockade? The answer increasingly lies not just in how many T cells are present, but in where they sit relative to malignant cells. Tumors are conventionally classified into three spatial immune phenotypes: inflamed, where cytotoxic T cells intermingle with tumor cells throughout the tissue; immune-desert, where effector T cells are scarce everywhere; and immune-excluded, where T cells pile up at the invasive margin but never penetrate the core. The excluded state is associated with ineffective tumor control, and the team behind the new study had previously shown that disrupted vascular integrity at the melanoma periphery contributes to this exclusion.

To map the phenomenon in patients, the researchers profiled 37 archived pretreatment human melanomas, both primary and metastatic, using CD8 immunohistochemistry. Roughly 38 percent of the tumors were classified as infiltrated, about 22 percent as excluded, and about 41 percent as desert. Among the 20 patients who had received immune checkpoint inhibitors targeting PD-1 or CTLA-4, response rates followed a clear gradient: 75 percent of patients with infiltrated tumors responded, compared with 50 percent of those with excluded tumors and just 17 percent of those with desert tumors. Patients with excluded tumors also showed a trend toward poorer overall survival than those with infiltrated tumors, with a median overall survival of 5.6 years versus not reached and a hazard ratio of 2.54, although that difference did not reach statistical significance.

To trace how these spatial patterns develop over time, the team turned to mouse models. They studied YUMMER1.7 melanomas, which carry BrafV600E, Pten loss, and Cdkn2a loss and are densely T cell-inflamed thanks to UV-induced mutations, along with the immunologically cold YUMM1.7 line and the NRAS-driven, anti-PD-1-resistant OSUMMER.13 line. Using spatial transcriptomics on the Xenium platform, with tissue microarrays built from cores sampled at the tumor periphery, defined as within 500 micrometers of the margin, and the core, they quantified immune-cell distributions in small tumors of roughly 100 cubic millimeters and large tumors of roughly 900 cubic millimeters. In small YUMMER1.7 tumors, immune populations were largely comparable between periphery and core. In large tumors, a marked shift toward peripheral enrichment emerged, and among lymphocytes CD8+ T cells showed pronounced exclusion from the core. Flow cytometry on dissected regions confirmed that CD8+ and CD4+ T cells were significantly more abundant at the periphery, while regulatory T cells, NK cells, monocytes, macrophages, neutrophils, and B cells showed no significant regional differences. Necrosis was minimal in both models, averaging between roughly 1 and 2.4 percent of tumor area, ruling out core necrosis as the explanation.

With the exclusion pattern established, the researchers searched for vascular determinants. They isolated endothelial cells, defined as CD45-negative and CD31-positive, from the periphery and core of large YUMMER1.7 tumors and performed bulk RNA sequencing. Gene Ontology analysis showed that peripheral endothelium was enriched for pathways involving cell adhesion, leukocyte migration, extracellular matrix organization, chemotaxis, inflammatory response, and endothelial migration, painting a picture of an activated, pro-adhesive vascular state at the tumor margin. Among adhesion and trafficking molecules, Icam1 emerged as one of the most significantly induced transcripts in peripheral endothelial cells, alongside chemokines such as Cxcl12, Cxcl13, and Cxcl14 and matrix-remodeling genes. The tumor core endothelium, by contrast, expressed a comparatively quiescent program.

Flow cytometry validated the transcriptomic finding at the protein level. Among the adhesion molecules tested, ICAM-1 showed the highest surface expression on tumor endothelial cells and was significantly enriched at the periphery. Immunofluorescence revealed that in small tumors, endothelial ICAM-1 staining was relatively uniform, but in large tumors it became strongly polarized to the periphery, with a marked reduction in the core. Quantification confirmed a significant periphery-core difference in large tumors, with a P value of 0.0003, and a pronounced decline in core endothelial ICAM-1 as tumors grew from small to large, with a P value below 0.0001. Notably, the T cells most strongly excluded from the core, CD8+ and CD4+ cells, carried the highest levels of LFA-1, the integrin ligand partner of ICAM-1, supporting a functional ICAM-1/LFA-1 axis in intratumoral positioning. P-selectin was also elevated peripherally, but VCAM-1 and E-selectin showed no regional differences.

The ICAM-1 polarization coincided with vascular destabilization. Fibrinogen deposition, a readout of vascular leakiness, was minimal in small tumors but pronounced at the periphery of large tumors relative to the core, with a P value below 0.0001, and dextran extravasation assays corroborated the finding. Pericyte coverage, marked by desmin staining, declined with tumor progression and was lowest at the periphery of large tumors. The researchers also checked whether hypoxia might be driving ICAM-1 expression, since hypoxia is known to regulate the molecule, but pimonidazole staining detected no robust hypoxic signal in viable regions of either small or large tumors. In human melanoma single-nucleus RNA sequencing data from 25 tumors, the fraction of ICAM-1-positive endothelial cells showed a modest positive association with overall T and NK cell abundance, with a Spearman correlation of 0.38 and a two-sided P value of 0.058, though endothelial ICAM-1 was not significantly linked to specific T cell states or TCR clonality after correction for multiple testing.

The therapeutic test followed. Mice bearing established YUMMER1.7 tumors received anti-ICAM-1 antibody or a rat IgG control every other day for seven doses starting when tumors reached about 70 cubic millimeters. Anti-ICAM-1 treatment significantly slowed tumor growth pooled across three independent experiments with ten mice per group. Immunofluorescence showed that control tumors had CD8+ T cells confined largely to the periphery, whereas blockade produced a more uniform distribution, with CD8+ counts rising in both regions but more steeply in the core. Flow cytometry showed the frequency of CD8+ T cells among live cells increased from 2.75 to 4.13 percent, roughly 1.5-fold, with a P value of 0.008, and the proportion of granzyme B-positive cytotoxic cells rose. CD45+ immune cells, CD4+ T cells, and regulatory T cells were unchanged. Strikingly, ICAM-1 blockade also reduced vascular leakage and increased pericyte coverage in both tumor regions, suggesting the treatment simultaneously relieved peripheral T-cell sequestration and improved vascular integrity, without altering P-selectin expression.

The most clinically provocative result came from the immune-refractory YUMM1.7 model, which has sparse baseline T-cell infiltration. Neither anti-ICAM-1 nor anti-PD-1 alone delayed tumor growth, but the combination markedly slowed progression compared with all other groups. Baseline CD8+ T-cell frequency in IgG-treated YUMM1.7 tumors was extremely low at 0.14 percent of live cells; monotherapies nudged it to 0.24 and 0.60 percent, while the combination raised it roughly eightfold to 1.18 percent. Even at that low abundance, the infiltrating CD8+ cells displayed a more activated phenotype, with higher proportions of granzyme B-positive cells, CD62L-negative CD44-positive effector memory cells, and CD69-positive cells. The authors caution that overall CD8+ abundance remained low even with combination therapy, and that both a minimal threshold of intratumoral T cells and improved function appear necessary for tumor control in cold tumors.

The study carries important caveats. The syngeneic models were each examined in a single sex, so sex-specific effects could not be evaluated, and systemic blockade was used rather than endothelial-specific targeting, leaving open the relative contributions of endothelial versus tumor-cell ICAM-1. The authors also note that ICAM-1 biology is context-dependent: on tumor cells it can promote metastasis or support antitumor immunity, and whether ICAM-1 inhibition adds benefit to dual checkpoint blockade with anti-PD-1 and anti-CTLA-4 remains untested. They further emphasize that careful optimization of dosing and timing will be needed, since endothelial ICAM-1 is critical for leukocyte trafficking generally, and indiscriminate inhibition could impede immune access rather than enhance it. Still, the findings add a spatial dimension to ICAM-1 biology and suggest that the ICAM-1/LFA-1 axis is a modifiable vascular checkpoint that, in principle, could be paired with immunotherapy to give T cells deeper access to tumors they are currently kept from reaching.

Subject of Research: Spatial polarization of endothelial ICAM-1 as a vascular mechanism of T-cell exclusion in melanoma

Article Title: Spatial polarization of endothelial ICAM-1 governs T-cell exclusion in melanoma

Article References: Park, H.-R., Kim, S. J., Zaman, N., Kozlov, N. P., Tagore, S., Wu, L. W., Izar, B., & Kim, M. (2026). Spatial polarization of endothelial ICAM-1 governs T-cell exclusion in melanoma. EMBO Molecular Medicine. https://doi.org/10.1038/s44321-026-00542-1

Image Credits: AI Generated

DOI: 10.1038/s44321-026-00542-1

Keywords: melanoma, ICAM-1, LFA-1, T-cell exclusion, tumor vasculature, immune checkpoint blockade, anti-PD-1, spatial transcriptomics, vascular leakage, pericyte coverage, tumor microenvironment, CD8+ T cells