Abstract
Over 95% of primary central nervous system lymphoma (PCNSL) cases are histologically classified as diffuse large B-cell lymphoma (DLBCL), but present markedly poorer prognosis than systemic DLBCL. This discrepancy may stem from its high molecular heterogeneity, immune privilege, inhibitory tumor microenvironment (TME) and dilemma in drug delivery. Current understanding in the tumor ecosystem of PCNSL remains limited. This study employed single-cell RNA sequencing (scRNA-seq) and spatial transcriptomic (ST) analyses to decode tumor microenvironmental features of PCNSL.
We performed scRNA-seq of PCNSL tumor samples and reactive lymphoid hyperplasia samples, and additionally conducted ST on PCNSL tumor samples. Unsupervised clustering of B cells identified 4 malignant B cell (mBc) subpopulations with apparently transcriptional and spatial discrepancy. Pseudotime trajectory analysis revealed that mBc3 represented the origin, which evolved into mBc4 and mBc1 and ultimately differentiated into mBc2. To assess the clinical relevance of these subsets, we applied BayesPrism for deconvolution of bulk RNA-seq data from DLBCL cohorts, presenting mBc2 subset was significantly correlated to poor prognosis (p< 0.001). Kyoto Encyclopedia of Genes and Genomes pathway analysis and scMetabolism analyses indicated that mBc2 exhibited a strong oxidative phosphorylation (OXPHOS) metabolic signature. ST analyses of tumor specimens further confirmed the overlap of mBc2 localization and the regions with elevated OXPHOS activity.
Based on its high OXPHOS characteristics, we further explored its molecular expression features in an attempt to identify potential therapeutic targets. COX7B, a structural subunit of mitochondrial respiratory chain complex IV, was identified as an mBc2 marker through Venn analysis, specifically overexpressed in the mBc2 subset, the terminal stage of developmental trajectory. In vitroassays demonstrated that COX7B expression is upregulated in activated B-cell-like DLBCL cell lines, and its knockdown significantly reduced cell viability and induced apoptosis, supporting its potential as a therapeutic target.
The previous studies have shown that clonally expanded CD8+T cells in PCNSL undergo a dynamic transition from a pre-exhausted state to an exhausted state, exhibiting more pronounced T cell exhaustion features compared to systemic DLBCL, suggesting the presence of an immunosuppressive TME in PCNSL. However, the specific mechanisms driving T cell exhaustion and immune suppression remain incompletely understood. Recognizing the critical role of T cells in the TME and in determining therapeutic response, we further investigated the immunosuppressive landscape of T cells and regulatory mechanisms mediated by tumor cells. T cell cluster analysis demonstrated a substantial enrichment of exhausted T cells (exhTc) and regulatory T cells (TregCD4) in PCNSL. Key exhausted molecules (TIGIT, HAVCR2/TIM-3, LAG3, CTLA4, and PDCD1/PD-1) were notably upregulated. ST confirmed high expression of HAVCR2/TIM-3 in T-cell-enriched regions. Importantly, the ligands for HAVCR2/TIM-3, CTLA4 and CEACAM1 were found to be upregulated in malignant B cells, with CD80 (CTLA4 ligands) particularly enriched in B-cell regions of spatial slices. These findings suggested active involvement of mBc in shaping the immunosuppressive TME. To explore intercellular signaling, we used CellChat to construct the communication network within the PCNSL microenvironment. mBc subsets showed stronger interactions with exhTc and TregCD4 compared to other TME cell types, primarily mediated by the macrophage migration inhibitory factor (MIF) signaling pathway. CD74+CXCR4 and CD74+CD44 were identified as key ligand-receptor pairs in this pathway. Analysis of spatial slices revealed elevated expression of CXCR4 and CD74 in tumor regions, further indicating widespread activation of the MIF pathway. These findings suggest that mBc subsets mediate the immunosuppression of exhTc and TregCD4 via MIF signaling.
In conclusion, this study reveals the intratumoral heterogeneity and immunosuppressive microenvironment of PCNSL, identifies the high-risk COX7B⁺ mBc2 subset, and indicates COX7B as a potential therapeutic target. Additionally, we clarify the key role of MIF signaling in immune suppression, offering new directions for targeted and immune therapies in PCNSL.
