Integrated single-cell and spatial analysis identifies T cell exhaustion and adhesion signatures in early follicular lymphoma transformation

Abstract INTRODUCTION. Germinal center (GC) lymphomas, including follicular lymphoma (FL) and diffuse large B cell lymphoma (DLBCL), arise from B cells at distinct GC stages and display molecular and clinical heterogeneity. FL transformation is associated with poor outcomes, yet predictive biomarkers at diagnosis remain unclear. We integrated single-cell and spatial transcriptomics to dissect tumor and microenvironmental heterogeneity and identify features associated with early FL transformation. METHODS. Five lymph node samples collected at diagnosis were analyzed using single-cell DNA sequencing (scDNA-seq, Mission Bio), single-cell RNA sequencing (scRNA-seq, 10x Genomics), and spatial transcriptomics (Visium HD, 10x Genomics). The cohort included 3 DLBCL cases (1 GCB, 2 ABC) and 2 FL cases: 1 non-transformed (ntFL) and 1 that underwent histological transformation 16 months after diagnosis (tFL), sampled during the indolent FL phase. RESULTS. Single-cell RNA-seq analysis revealed meaningful heterogeneity among malignant B cells across GC lymphoma subtypes. ntFL was enriched in light-zone GC B cells, characterized by overexpression of cytokine signaling genes. In contrast, malignant B cells from tFL and GCB DLBCL shared a transcriptional profile suggestive of a common ancestral precursor (intermediate dark zone/light zone GC B cells with strong BCR signaling). ABC DLBCL samples were enriched in pre-memory B and pre-plasma cells. Within the tumor microenvironment (TME), canonical exhaustion markers were more highly expressed in CD8+ Teff and CD4+ Treg cells in DLBCL and tFL compared to ntFL. CD4+ Tfh cells overexpressed adhesion-related genes, with significantly higher expression in DLBCL and tFL. Based on these findings, we defined two transcriptional signatures: an exhaustion signature (CD8+ Teff and CD4+ Treg) and an adhesion signature (CD4+ Tfh). Both signatures were significantly enriched in tFL versus ntFL (exhaustion: p<2.22x10-16; adhesion: p=0.0042). These findings were validated in an independent FL cohort (Sarkozy et al. 2024), where both signatures remained significantly elevated in tFL, particularly in early transformation cases (exhaustion: p=0.0045; adhesion: p=2.4x10-9). To investigate the mutational landscape underlying these transcriptomic findings, we performed scDNA-seq analysis. In 4 patients, we identified mutations in chromatin-modifying and oncogenic genes. In GCB and tFL, KMT2D mutations were clonal first hits, while EZH2 (tFL) and ATM (GCB) were found as secondary hits. A second scDNA-seq analysis identified clonal hematopoiesis-related mutations in non-B TME cells, with 2 or 3 mutations per sample in TET2, ASXL1, or DNMT3A. To assess whether spatial architecture contributes to these transcriptional profiles, we integrated previous data from a spatial perspective using Visium HD. We included FFPE lymph node blocks from ntFL, GCB DLBCL, and tFL at diagnosis (tFL-FL) and transformation (tFL-DLBCL). Using gene signatures derived from our paired scRNA-seq dataset, we calculated cell type scores for each 8x8 mm bin. Each bin was assigned to the cell type with the highest score. In ntFL, B cell bins were frequently surrounded by myeloid cell bins, whereas in tFL-FL, the neighborhood of B cell bins was enriched in CD8+ T and exhausted T cell bins. B cell bins in tFL-DLBCL and GCB samples showed broader interactions with TME bins (CD4+ T and exhausted T) and fewer B-B contacts, reflecting diffuse growth patterns. Exhausted CD4+ Treg and CD8+ Teff bins were more enriched around B cell bins in tFL-FL versus ntFL (Z-score: 0.39 vs. 0.27) and further increased after transformation (Z-score: 0.49 vs. 0.39). Ligand-receptor interaction analysis confirmed stronger interactions between CD4⁺ Tfh bins and B cell bins in tFL-FL versus ntFL (e.g., CD40LG-CD40: 0.40 vs. 0.26; CXCL13-CXCR5: 0.76 vs. 0.65; CD2-CD58: 0.27 vs. 0.22).CONCLUSIONS. These findings suggest that ntFL and tFL represent distinct biological entities, while tFL and GCB DLBCL share a common ancestral precursor. The combined expansion of exhausted T cells and CD4⁺ Tfh with an adhesion phenotype, together with their increased spatial interaction with malignant B cells, defines an immunosuppressive niche with potential prognostic value for identifying FL patients at high risk of early transformation. The detection of clonal hematopoiesis-associated mutations in non-B TME cells across all samples suggests a potential role in lymphomagenesis.