Cytokine-Linked Remodeling of the Immune Microenvironment Is a Hallmark of Transformation in Follicular Lymphoma

Histologic transformation of follicular lymphoma (FL) to an aggressive lymphoma (transformed FL [tFL]) is a common disease complication associated with poor clinical outcomes. However, understanding of the mechanisms that drive FL transformation and the associated lymph node (LN) tumor microenvironment (TME) reorganization remains limited. To define the remodeling that occurs at tFL, we performed single-cell and spatial dissection of clinical samples over the process of transformation.

To identify immune cell populations and malignant B cells, single-cell RNA sequencing (scRNA-seq) was applied to >47,300 malignant and 24,000 non-malignant cells across 10 LN biopsies (4 tFL, 6 FL). tFL samples displayed reduced plasmacytoid (pDCs) and follicular dendritic cells (FDCs) (FDR-corrected Wilcoxon rank sum q=0.13). Bulk RNA-seq analysis of biopsy samples from 18 independent patients with relapsed/refractory disease (5 FL, 13 tFL) confirmed FL and tFL transcriptomes were distinct by principal components (PC) analysis (Hotelling t² test on first two PCs p=0.0018), with high expression of myeloid genes (CD163, FCGR3A, CES1) and low expression of the FDC gene FDCSP as the top tFL-enriched PC. Decreased FDC and pDCs in tFL, as well as increased fibroblastic reticular cells (FRCs) and monocytes (FDR-corrected Wilcoxon rank sum q<0.1), were detected by CIBERSORTx deconvolution.

As orthogonal validation of our findings in a larger independent cohort, we performed spatial transcriptomics (GeoMx) and proteomics (CODEX, Co-detection by Indexing) across 2 tissue microarrays (TMAs). We localized cell populations from our discovery analysis in the LN TME through acquiring 70 unique GeoMx regions of interest derived from 52 biopsies (33 FL,19 tFL; 9 were paired FL-tFL). Each was subjected to 4 samplings (CD3+ T cells, PAX5+ tumor cells, CD68+ macrophages and other cells). tFL was confirmed to display alterations in LN cell types that establish follicle organization through cell-interactions and chemokine secretion. Consistent with our discovery findings and the expected loss of FDC meshworks with transformation, tFL demonstrated reduced expression of FDC-associated genes and the FDC-secreted CXCL13 (log₂ fold-change [L₂FC] tFL/FL=-0.87, FDR-corrected Limma moderated t-test q=0.011) compared to FL. While FRC gene expression (e.g. COL1A1 L₂FC=1.1, q=1.5x10^-5) and expression of matrix remodeling genes (e.g. POSTN and TIMP1, q<0.01) were increased in tFL, we observed reduced expression of secreted chemokines CCL19 and CCL21 (L₂FC=-0.82 and q=0.0035, L₂FC=-1.3 and q=2.2x10^-5, respectively).

From our TMA analyses, we identified differences in monocyte subpopulations between FL and tFL. Our scRNA-seq data identified 3 subsets of monocytes with distinct chemokine expression. One subset was IL1-B^high, while another was APOE^highGPNMB^high; the latter were enriched in TMA samplings of tFL to FL (APOE L₂FC=0.89, q=1.2x10^-6, GPNMB L₂FC=1.2, q=3.3x10^-9). A final subset was MRC1(CD206)^highCXCL9^high, which were found to co-localize with cytotoxic T lymphocytes (CTLs); CXCL9 expression correlated with CTL genes CXCR3, NKG7 and GZMB (p<0.007) across TMA samples. Moreover, high expression of the CXCL9-receptor CXCR3 was identified by receptor-ligand inference in CD8+ CTLs. Spatial analysis of cases with co-existing FL and tFL histology (Visium HD and CODEX) demonstrated that MRC1(CD206)^highCXCL9^high cells were prominent at boundary regions separating tFL from FL.

Regarding the malignant B cells, we applied non-negative matrix factorization to our scRNA-seq data to identify signatures shared by FL cells across patients, including plasmablast-like (XPB1, MZB1, PRDM1), memory (KLF2, CD69) and CXCR4^high. Analysis of TMA samplings identified memory B cell genes depleted in tFL compared to FL (e.g. KLF2 log₂ FC=-1.1, q=9.2x10^-14), whereas cell cycle gene signature scores increased in tFL (t-test p=8.0x10^-5).

Altogether, our data suggests the chemokine and stromal cell driven spatial organization of indolent FL is disrupted upon transformation, associated with a combination of FDC loss, altered FRC function and increased monocytes. Ongoing work is focused on integrative analysis of spatial transcriptomic and proteomic datasets and investigation of the predictive potential of TME evolution for tFL risk.