This PhD thesis establishes cerebrospinal fluid (CSF) and pleural effusions as minimally invasive liquid biopsy sources to dissect the tumor microenvironment (TME) in leptomeningeal disease (LMD) and pleural effusions (PE) using single-cell RNA sequencing (scRNA-seq), T cell receptor (TCR) profiling, spatial transcriptomics, and multi-omic integration.

 

We demonstrate that CSF faithfully recapitulates disease-specific immune landscapes from primary central nervous system inflammatory diseases, lymphoma, glioblastoma, and brain metastases, revealing compartmentalized T cell activation, myeloid metabolic reprogramming linked to altered lactate levels, and clonal trafficking between blood, CSF, and parenchymal lesions, validated by Xenium spatial transcriptomics.

 

In parallel, multi-omic profiling of >200,000 cells from lung, breast, and ovarian cancer PEs versus heart failure controls uncovers tumor heterogeneity mirroring primary lesions, disease-specific exhausted CD8+ T cell and tumor-associated macrophage states predictive of treatment response, and a validated 50-gene diagnostic signature outperforming cytology for malignancy detection.

 

These findings position proximal liquid biopsies as dynamic windows into evolving tumor-immune ecosystems, enabling real-time monitoring of clonal dynamics, immune stratification, and resistance emergence to guide precision immunotherapy while highlighting computational challenges in sparse, multi-modal data integration across compartments.