A Single Cell Atlas of Follicular Lymphoma across Clinical Stages Defines Tumor B Cell Archetypes, Immune Microenvironment Heterogeneity and Novel Cell States Associated with High-Risk FL

One of the challenges in follicular lymphoma (FL) treatment resides in the heterogeneous nature of the disease. Despite its largely indolent course, FL remains challenging, with high-risk groups defined as those who progress within 2 years of treatment initiation (POD24) or develop histologic transformation (tFL) towards an aggressive histology, both representing a leading cause of FL-related mortality. Single-cell omics studies of FL tumors have begun to describe the high degree of intratumor heterogeneity, the diversity of B cell states ranging from cycling to quiescent, as well as distinct tumor-infiltrating T-cell compositions linked to genetic patterns. However, high-resolution studies probing the extent of intratumor heterogeneity across large sample sizes and clinical stages (diagnosis, relapsed/refractory, tFL) are still needed to define biological determinants of high-risk vs. low-risk patients. Here, we generated a comprehensive single-cell atlas of 107 FL samples across clinical patterns to decipher recurrent B cell states and tumor microenvironment (TME) ecosystems, while identifying high-risk prognostic biomarkers.

We performed integrative single-cell RNA coupled with surface protein profiling, B- and T-cell receptor (BCR/TCR) sequencing of 107 FL biopsies, recovering 230,832 cells after stringent quality controls. FL samples were obtained from a real-world, clinically annotated, multicentric viable cell collection and comprised i) 60 recently diagnosed and untreated patients (7 who will remain untreated and observed and 53 who received frontline immuno-chemotherapy (IC), including 17 POD24) ii) 39 relapsed/refractory and 8 tFL. We distinguished malignant FL cells from non-malignant TME using canonical markers and BCR sequence, exploring heterogeneity in all components. We derived robust clusters using weighted-nearest-neighbor clustering informed by both RNA and protein expression. We characterized malignant B cell archetypes shared across patients based on rank-biased overlap metaclustering and canonical correlation analysis integration. We integrated TME cells to quantify immune subsets and clonally expanded T cell clones, and performed statistical analyses to identify clinically-relevant features linked to patient outcomes.

To discover novel features of FL cells and overcome interpatient heterogeneity, we searched for conserved archetypes across patients and clinical stages. Within the malignant B-cell compartment, we found substantial heterogeneity with 10 archetypes: Arch.1, 6 and 9 defined by 3 distinct memory-like identities, Arch.2 by IGHA expression, Arch.3 and 4 by GC-like dark and light zone markers, Arch.5 by MYC/NFKB pathways, Arch.7 by pre B like markers, Arch.8 by cell cycle and Arch.10 by plasma cell-like identity. Most archetypes coexisted within a given sample, albeit in varying proportions. Importantly, proteomic profiling allowed us to link FL-specific archetypes with surface marker expression, e.g. identifying Arch.7 as a CD20-negative cluster, allowing future integration in cytometry panels. We observed notable changes in archetype composition across the 4 clinical categories, culminating in tFL cases with higher proportions of Arch.3, 5, 8, 9 cells and reduced proportions of Arch.1, indicating cellular changes shaping clinical phenotypes. In newly diagnosed patients who received IC, the archetype composition in POD24 vs non-POD24 showed a significant enrichment of Arch.6 and decreased Arch.4 cells in high-risk patients; validation in larger bulk RNA-sequencing FL cohorts is ongoing. Unsupervised analysis of the TME composition distinguished two main ecosystems based on the proportions of T, NK and myeloid subsets: one characterized by T follicular helper, regulatory and effector cells enriched in diagnostic samples and one by naive T and myeloid subsets enriched in relapsed samples. We also identified expanded TCR clonotypes within CD8 subsets, which were enriched in POD24, suggesting that presence of expanded clones at diagnosis might affect IC efficacy.

Our results provide the largest comprehensive single-cell atlas of FL across clinical categories, from indolent to more aggressive subtypes, indicating transcriptional cellular shifts of tumor ecosystems along progression. We notably uncovered a novel FL archetype (Arch6) expressing stemness markers able to predict high-risk POD24 patients

Salles:Kite/Gilead: Consultancy; BMS/Celgene: Consultancy; Janssen: Consultancy, Research Funding; Genentech/Roche: Consultancy, Research Funding; Merck: Consultancy; BeiGene: Consultancy; Ipsen: Consultancy, Research Funding; Genmab: Consultancy, Research Funding; AbbVie: Consultancy, Research Funding; Molecular Partners: Consultancy; Incyte: Consultancy; Nurix: Research Funding. Brisou:Kite-Gilead: Honoraria. Ortiz Estevez:Bristol Myers Squibb: Current Employment, Current equity holder in publicly-traded company. Stokes:Bristol Myers Squibb: Current Employment, Current equity holder in publicly-traded company, Current holder of stock options in a privately-held company. Seth:BMS: Current Employment, Current equity holder in publicly-traded company. Huang:BMS: Current Employment. Kaplan:Bristol Myers Squibb: Current Employment, Current equity holder in publicly-traded company. Gandhi:Bristol Myers Squibb: Current Employment, Current equity holder in publicly-traded company. Nadel:Diatech Pharmaceuticals: Consultancy, Honoraria; Beigene: Consultancy, Honoraria; BMS: Research Funding; Institute for Follicular Lymphoma Innovation: Consultancy, Honoraria. Milpied:Innate Pharma: Research Funding; BMS: Research Funding. Roulland:BMS: Research Funding.