Longitudinal Single Cell Analyses Reveal the Co-Evolutionary Dynamics of the Tumor and Microenvironment Accompanying Follicular Lymphoma Transformation

Introduction: Transformation of follicular lymphoma (FL) from an indolent to aggressive high-grade diffuse large B-cell lymphoma (tFL) is associated with poor outcomes compared to that of patients that do not transform, and is a leading cause of FL-related mortality. We and others have previously reported the evolution of the genetic landscape from FL to aggressive tFL from bulk DNA-sequencing. However, a detailed understanding of the biological determinants of this high-risk phenotype to identify both lymphoma-intrinsic and -extrinsic vulnerabilities are lacking, but are needed to inform rational therapeutic approaches. Methods: To characterize the co-evolution of both the lymphoma and its tumor microenvironment (TME), we enriched specific populations by flow sorting lymph node cell suspensions to provide a higher cellular resolution of 3 distinct populations: malignant B-cells, T-cells and other CD45+ immune cells. Our cohort comprised 3 FL patients with no transformation (ntFL) and 4 patients with transformation (paired pre-transformed (pFL) and later tFL samples). Within these populations we simultaneously interrogated: single cell transcriptomic, B-cell receptor (BCR) and T-cell receptor (TCR) analyses (10X Genomics) on approximately 5000 cells/population and parallel bulk DNA and RNA-sequencing. We compared our observations to single cell transcriptomic data from reactive lymph nodes (RLNs), pediatric tonsils and additional validation cohorts of tFLs and de novo DLBCLs. Results: We profiled >110,000 cells from our discovery cohort (malignant B-cells >50,000 and TME cells >60,000 cells). Combined analyses of somatic copy number alterations (CNAs) derived by InferCNV and BCR clonotypes indicated that these genetic alterations were not the dominant drivers of transcriptional heterogeneity, in either indolent FL or aggressive tFL. Instead, by using non-negative matrix factorization (NMF) to probe the common lymphoma-specific metaprograms (MPs), we found malignant B-cells were composed of a broader spectrum of cellular states beyond the classical germinal center (GC) cell-of-origin, ranging from naïve-, GC- to memory-like states. Strikingly, specific lymphoma MPs mapped with a high degree of semblance to signatures identified within RLN and tonsils, suggesting that part of the normal B-cell transcriptional architecture is preserved in both FL and tFL. We identified fluctuations in the B-cell states and programs across ntFLs and paired pFL-tFLs, for example an increase in interferon-memory and plasma-like states and a decrease in MHC class I expression in tFL malignant B-cells indicating that the balance of these states contribute to shaping the different clinical phenotypes. Integrated analyses of the TME revealed substantial heterogeneity in the tumor-infiltrating T-cell (CD4, CD8), natural killer (NK) and myeloid subpopulations. We observed higher tumor-infiltrating CD4+ cells in patients without transformation (ntFL) compared to higher proportions of CD8+ sub-populations accompanying transformation (paired pFL-tFL). Many T-follicular helper (Tfh) populations including activated (CD69, TNFRSF4, TNFRSF18) were mostly enriched in ntFL and pFL but significantly reduced in tFL. In contrast, there was a progressive shift towards dysfunctional CD8+ populations in those with transformation. Using a computed 12-gene exhaustion score to quantify this activity in CD8+ populations, we demonstrated a gradient from effector to exhausted clusters, with the most marked enrichment in the tFL state (ntFL vs tFL; p <2.2x10-16), and a degree of exhaustion already pre-existing in the pFL. We identified significantly expanded TCR clonotypes at transformation, many of which were not detected prior to transformation. Notably, many of these unique clonotypes resided within the CD8+ exhausted populations. Similarly, there was a divergence in NK cell states between non-transformed (cytotoxic NK) and transformed FL patients (inflamed and exhausted). Overall, the immune axis differs in non-transformed patients and becomes progressively dysfunctional from pFL to tFL. Conclusions: In summary, we present detailed single cell multi-omic analyses providing novel insights into the co-evolutionary B-cell and TME dynamics of FL towards transformation, with distinct composition and states in non-transformed, pre- and transformed FL patients.