Biomarkers Informed by Single-Cell and Spatial Transcriptomics-Biomarkers for Grade 3 Follicular Lymphoma

Follicular lymphoma (FL) patients have variable outcomes, underscoring the need for biomarkers for improved risk stratification. Current FL grading systems, based on subjective centroblast counts, suffer from poor reproducibility, despite evidence linking grade 3 FL to worse prognosis. We aimed to identify objective biomarkers for centroblasts and centrocytes to improve FL prognostication. We reanalyzed publicly available spatial and single-cell transcriptomic data from normal germinal centers and FL samples. Reanalysis revealed distinct gene expression profiles: AICDA (AID) and CXCR4 were highly expressed in germinal center dark zone cells (centroblasts) and CD40 and TFRC (CD71) in light zone cells (centrocytes). Single-cell RNA sequencing of FL samples further showed AID and CXCR4 overexpression in malignant grade 3A cells and CD40 and CD71 in grades 1 to 2 cells. We validated these findings using immunohistochemistry (single and multiplex) on tonsils and 59 FL specimens (42 grades 1 to 2, 17 grade 3). Grade 3 FL showed significantly higher expression of AID, CD71, and Ki67 compared to grades 1 to 2, with CXCR4 approaching significance. Receiver operating characteristic curve analysis identified optimal cutoffs for AID (1.54%), CXCR4 (21.9%), Ki67 (21.6%), and CD71 (7.57%) to distinguish grade 3 from grades 1 to 2 FL, with AID showing the best discriminatory ability. Crucially, AID expression evaluation showed reproducibility across 2 different digital algorithms and 2 independent visual observers. Furthermore, we observed a trend toward shorter disease-specific survival in patients with both FL grade 3 and high AID expression. This prognostic observation held true regardless of whether AID overexpression was assessed via digital evaluation (cutoff: 1.54%) or visual estimation (cutoff: 2%). In conclusion, AID, CXCR4, CD71, and Ki67 are promising biomarkers for objectively identifying FL grade 3, potentially enhancing the reproducibility of grading and serving as independent prognostic tools. Further clinical validation in uniformly treated FL cohorts is warranted.