ARTICLES https://doi.org/10.1038/s41590-018-0181-4 1 Aix Marseille Univ, CNRS, INSERM, CIML, Marseille, France. 2 Institut Carnot CALYM, Pierre-Bénite, France. 3 Hospices Civils de Lyon, Université Claude Bernard Lyon 1, INSERM 1052, Pierre-Bénite, France. *e-mail: milpied@ciml.univ-mrs.fr; nadel@ciml.univ-mrs.fr G erminal centers (GCs) are micro-anatomical structures in the center of B cell follicles in which antigen-activated B cells proliferate and mutate their B cell antigen receptor (BCR) and high-affinity variants are selected for export as memory B cells and plasma cells 1 . Within the GC, B cells transit from the dark zone (DZ), where they proliferate as centroblasts and undergo somatic hypermutation (SHM), to the light zone (LZ), where they make contacts with follicular dendritic cells and follicular helper T cells (T FH cells) for selection as centrocytes 2,3 . Some high-affinity cen- trocytes are selected to reenter the cell cycle and migrate back to the DZ for a new round of mutation and selection 4 . Although GC B cells transit through multiple states (for example, proliferation, migration, antigen presentation, signaling and differentiation), GC B cells are classically categorized into two subsets according to their location (DZ or LZ) and associated phenotype 4,5 . Most B cell lymphomas derive from GC-experienced B cells 6 . Follicular lymphoma (FL), the second most common adult lym- phoma in Western countries, is considered a prototypical GC B cell lymphoma 7 , with several features that identify the cell of origin of FL as a GC B cell 8 . First, FL is organized into follicular structures in which FL B cells are present as either small centrocyte-like lym- phocytes or centroblast-like proliferating cells. Second, FL B cells express markers of GC B cells, such as the surface proteins CD20 and CD10 and the transcription factors BCL-6 and LMO2. Third, in most cases of FL, the cytidine deaminase AID, which drives SHM in GC B cells, is expressed, and the BCR of FL B cells is targeted by ongoing SHM 9 . Finally, bulk transcriptomes of FL are closely related to those of GC B cells 10 and, more specifically, have a LZ-like gene- expression profile 5 . In preclinical stages, FL precursor B cells iteratively re-enter GC reactions during which they can accumulate additional onco- genic hits, which leads to more-advanced lymphoma lesions 8,11,12 . It remains to be determined whether FL B cells in clinical stages of the disease are similarly involved in GC-like functions or whether they are blocked at a given stage of the GC. Gene-expression profiling of bulk samples of tumor cells or normal B cells has not answered that question because this yields little insight into cellular heterogeneity. Integrative single-cell analysis of gene expression, surface phenotype and BCR sequence has been useful for studying the heterogeneous GC B cell subset 13 . Here we applied that approach to characterize B cell transcriptional heterogeneity in overt FL and compare it with the various states observed in normal (non- cancerous) GC B cells. We modeled the continuum of human GC B cell transitional states and identified characteristic gene clus- ters synchronously expressed in single GC B cells. We revealed a major desynchronization of such GC-specific gene-expression programs in FL. We validated and extended our findings with analyses of human GC and FL B cells by single-cell RNA sequenc- ing (scRNAseq). Our data reveal that GC-derived lymphoma B cells are not ‘frozen’ at a particular stage of GC maturation and might have distinct gene-expression profiles that are independent of subclonal divergence. Results Single-cell gene-expression analysis of human B cells. We assessed normal B cell subsets from human lymphoid organs (spleen or tonsil) by single-cell gene-expression analysis of a panel of 91 pre- selected genes (Supplementary Fig. 1a). Phenotypic index sorting linked the gene expression to the surface abundance of markers included in our flow cytometry sorting panels (Supplementary Fig. 1b,c). The qPCR assays were robustly quantitative and linear (Supplementary Fig. 1d,e). Hierarchical clustering of the single-cell gene-expression matrix segregated antibody-producing cells, mem- ory B cells and GC B cells (Supplementary Fig. 1f). Cells did not cluster by sample origin (P = 0.64 (χ 2 test)). We used dimensionality reduction to represent the gene-expression combinatorial diversity of B cells in two-dimensional maps. Principal-component analy- sis (PCA) (Supplementary Fig. 1g,h) and t-distributed stochastic neighbor embedding (t-SNE) analysis (Fig. 1a) generated maps in which the antibody-producing, memory and GC B cell types were well separated. The expression of known marker genes of memory B cells (NOTCH2 and JAM3), antibody-producing B cells (PRDM1 and SDC1) and GC B cells (BCL6 and AICDA) was highly restricted to the corresponding cluster (Fig. 1b). Phenotypically defined early Human germinal center transcriptional programs are de-synchronized in B cell lymphoma Pierre Milpied  1 *, Iñaki Cervera-Marzal 1 , Marie-Laure Mollichella 1 , Bruno Tesson 2 , Gabriel Brisou 1 , Alexandra Traverse-Glehen 3 , Gilles Salles  3 , Lionel Spinelli  1 and Bertrand Nadel  1 * Most adult B cell lymphomas originate from germinal center (GC) B cells, but it is unclear to what extent B cells in overt lymphoma retain the functional dynamics of GC B cells or are blocked at a particular stage of the GC reaction. Here we used inte- grative single-cell analysis of phenotype, gene expression and variable-region sequence of the immunoglobulin heavy-chain locus to track the characteristic human GC B cell program in follicular lymphoma B cells. By modeling the cyclic continuum of GC B cell transitional states, we identified characteristic patterns of synchronously expressed gene clusters. GC-specific gene- expression synchrony was lost in single lymphoma B cells. However, distinct follicular lymphoma–specific cell states co-existed within single patient biopsies. Our data show that lymphoma B cells are not blocked in a GC B cell state but might adopt new dynamic modes of functional diversity, which opens the possibility of novel definitions of lymphoma identity. NATURE IMMUNOLOGY | www.nature.com/natureimmunology