Spatially resolved profiling reveals the perifollicular border zone as a key immunologic hub in follicular lymphoma undergoing mosunetuzumab therapy

Abstract Introduction. The tumor microenvironment (TME) plays a central role in the efficacy of T cell-engaging immunotherapies, including CD20×CD3 bispecific antibodies (BsAbs) in lymphoma. In follicular lymphoma (FL), multiplex immunofluorescence (mxIF) and single-cell RNAseq show disruption of nodal architecture and T cell subtype content (Radtke et al., Cancer Cell 2024). Spatially resolved transcriptomics (ST) offers a powerful approach to characterize the TME and cell interactions in situ. We applied ST and mxIF to analyze the TME in FL and nodal marginal zone lymphoma (MZL) tumors from patients receiving first-line CD20×CD3 BsAb mosunetuzumab (Mosun) in our investigator-sponsored phase 2 trial (NCT04792502). Methods. Trial subjects (n=52) had FL or MZL requiring first-line therapy due to symptoms or tumor burden. They received Mosun for 4 cycles, followed by Mosun alone in case of interim complete response (CR) or Mosun + lenalidomide otherwise. We performed 10x Visium ST on formalin-fixed biopsies collected pre-treatment (n=12) or after 4-6 cycles of Mosun (paired n=2 biopsies), as well as single-cell resolved mxIF (Lunaphore COMET; n=4 pre-treatment biopsies). Spatial domains were defined by unsupervised clustering of integrated ST data, with further differential gene expression (via DESeq2), pathway enrichment (MSigDB and curated immune gene sets from Tumuluru et al. Blood 2025), and cell-cell interaction analyses (via CellChat). TME composition was inferred by spatial deconvolution (CARD) and CytoSPACE using scRNA-seq references. mxIF images were processed by QuPath, with InstanSeg segmentation and cell-type assignment supported by a random-forest classifier. Spatial domains were manually annotated and cellular neighborhoods defined based on each cell's 100 nearest neighbors. Results. Across all baseline FL (n=10; 8 with CR, 2 refractory) and MZL (n=2, with CR) samples, four consistent spatial domains were identified: neoplastic follicle/germinal center (GC), border zone (BZ), interfollicular (IF), and stromal. The BZ showed the highest density of CD8⁺ T cells and LAMP3⁺ dendritic cells (DC; P0.001), and intense CCR7–CCL19/21 signaling. LAMP3+ DC were the major source of CCL19–CCR7 interactions targeting naïve/central memory CD8+ T cells, and Galectin 9–TIM-3 interactions targeting exhausted CD8+ T cells. Enriched BZ gene sets were related to T cell activation, differentiation, migration, and cytotoxicity, including upregulation of IL7R, GZMB, CD69, CCR7, SELL, ITGAX, CXCL9, and CCL19/21. Stromal signatures showed activated and CXCL10⁺ high endothelial venules in the BZ, suggesting active T cell recruitment. Notably, cytotoxic CD8⁺ T cells were excluded from the GC (P0.001), and many exhibited an exhausted transcriptional program. On mxIF images (n=4 FL tumors encompassing 500K cells), BZ was significantly enriched with CD8+ T cells (compared with GC), as well as DCs and Tregs (compared with either GC or IF;P.001). CD4+ and CD8+ T cells overexpressed PD-1 in GC and BZ compared with other domains (P.001). In neighborhood analysis, BZ showed the highest entropy (P0.001 vs. other zones), indicating greater cellular diversity. A specimen from a FL refractory to Mosun showed disruption of GC/BZ morphology with expansion of BZ cellular neighborhood rich in Tregs, macrophages, and LAG3+CD8+ T cells. Paired FL specimens pre- and on-treatment (n=2 pairs) showed absence of B-cells in one sample on treatment, but residual GC-like areas in another. Expanded BZ areas were depleted of B cells, showing abundance of CD8+ T cells and macrophages, with overexpression of T cell activation and exhaustion signatures. Upregulated T cell-related genes included LCK, ZAP70, CCR7, TCF7, LEF1, as well as the inhibitory receptors LAG3 and VSIR. There were increased interactions involving CD47 and Galectin 9. Residual GC areas remained devoid of T cells but showed increased macrophage and cDC1 content and increased frequency of interactions involving CTLA4-CD80 and CD70-CD27 compared with pre-treatment. Conclusions. This spatial analysis of TME identifies the perifollicular BZ as a hub for T cell recruitment and activity in FL, expanding during mosunetuzumab therapy. Increased macrophage content and T cell regulatory mechanisms occurring on therapy support exploration of rational BsAb-based combinations with agents that modulate T cell function, macrophage activity, or immune checkpoints to enhance therapeutic efficacy.