Abstract BACKGROUND Leptomeningeal metastasis (LM) is a fatal neurological complication of cancer characterized by the entry of metastatic cancer cells originated from solid or hematologic tumors into the cerebrospinal fluid (CSF)-filled leptomeningeal compartment encasing the brain and the spinal cord. LM occurs in 5-20% in patients with solid tumors, especially those with breast cancer, lung cancer, melanoma. Untreated LM patients succumb within 4-6 weeks after diagnosis, or 2-5 months when they are treated with the current available therapies. To date, the only efficacious life-prolonging intervention was the application of proton craniospinal irradiation (pCSI) as demonstrated in the recent clinical trial, NCT04343573, with an average overall survival of 9 months. While a cohort of pCSI-treated patients survived over 18 months (responders), another cohort did not respond to pCSI and succumbed to the disease within 3 months post-treatment (non-responders). In this project, we aimed to characterize the molecular basis of pCSI response to turn the non-responders into responders. MATERIAL AND METHODS We performed proteomic analyses on serially collected CSF from LM patients at baseline (before pCSI), and multiple time points post-treatment. We further established syngeneic mouse models of LM-CSI to define the molecular basis of pCSI resistance. RESULTS We found that pCSI resulted in a significant drop in the levels of the chemokine CXCL1, a CXC-motif chemokine with elevated baseline levels after the onset of LM. Patients with significantly higher CSF CXCL1 levels prior to pCSI had worse central nervous system (CNS) progression-free survival. Using our preclinical LM mouse models, we found that both the cancer and host cells, namely macrophages, choroid plexus epithelial cells, and leptomeningeal fibroblasts, were a source of CXCL1. Genetic manipulation of Cxcl1expression in cancer cells or host revealed that the CXCL1 produced by the cancer cells was an essential factor for their growth in the leptomeninges. Moreover, we found that a subset of cancer cells expressed a receptor for CXCL1 - CXCR2 - and transcriptomic profiling of this rare population revealed an enrichment in pathways associated with cell cycle progression. Finally, intrathecally delivered CXCR2 antagonist hampered the LM growth. CONCLUSION Our results indicate that CXCL1/CXCR2 signaling axis regulates LM growth and suggests interruption of this signaling axis as a possible actionable therapeutic intervention to halt LM progression.