Genetic Subclones Reveal Phenotypic Heterogeneity in Diffuse Large B-Cell Lymphoma

Introduction

Genetic and gene expression subtypes of diffuse large B-cell lymphoma (DLBCL) capture biological differences between tumors that influence the response to immunochemotherapy. Single cell analysis of DLBCL could illuminate further genetic and phenotypic distinctions that influence tumor biology and therapeutic response.

Methods

We generated paired single-cell RNA (scRNA) and ATAC (scATAC) sequencing together with bulk RNA and exome analysis of 103 DLBCL patient biopsies (102 cases, 504,444 cells) and 3 tonsils (12,227 cells). We profiled tumors representing the landscape of gene expression subtypes (ABC=41, Unclassified=14, GCB=47) and genetic subtypes (MCD=8, BN2=13, N1=3, A53=8, EZB=23, ST2=3, Composite=4, Other=40). Sequencing analysis was performed using R and custom bioinformatic pipelines. Published tools were utilized to determine gene expression (Alizadeh et al 2000, Rosenwald et al 2002) and genetic subtypes (Wright et al 2020). Cell line viability was determined by FACS.

Results

We identified malignant B cells by their BCR clonotype and non-diploid copy number profiles, inferred from scRNA (310,509 malignant B cells). Most DLBCL tumors (82/103, 80%) contained two or more genetic subclones (median 2, range 1-5) based on DNA copy number differences. The genetic subclones had distinct phenotypes based on expression of six recurrent gene expression meta-signatures, herein termed themes. The germinal center (GC) B cell, memory B cell, plasma cell, and pan-B cell themes reflect B cell differentiation whereas two other themes – cell cycle and cell growth – reflect proliferative and metabolic states that are independent of the differentiation states. Surprisingly, 23% of DLBCL (24/103) harbored genetic subclones expressing B cell differentiation themes that distinguished them from other malignant cells in the same tumor. In a validation cohort (n=574), the GC B cell theme was associated with a favorable response to R-CHOP chemotherapy (p<0.02), as expected, while the cell growth theme (but not the cell cycle theme) was associated with adverse survival (p<0.02).

The above observations suggested that the genetic alterations that defined the subclones in DLBCL tumors influenced their biology. In support of this notion, we observed that tumor subclones that had deleted the INK4a/ARF tumor suppressor locus, which encodes the cell cycle inhibitor p16 (CDKN2A), had significantly higher expression of the cell cycle theme, as expected (n=3 cases with 13 subclones, p=0.01). Interestingly, subclones with amplification of the REL locus had significantly lower expression of the memory B cell theme (n=5 cases with 16 subclones, p=0.02), suggesting that this genetic abnormality may be selected to block the terminal differentiation of the malignant GC B cells into memory cells. Notably, in the independent validation set (n=574 cases), REL amplification was also associated with low expression of the memory B cell theme (p=5.3x10^-10). The REL locus encodes the NF-κB subunit c-REL and accordingly, scATAC data revealed enrichment for the NF-κB motif in open chromatin of subclones with REL amplification (p=3.1x10^-35). To explore the functional relationship between REL amplification and memory B cell differentiation, we expressed a dominant negative version of IκBα, which blocks nuclear translocation of c-REL, in DLBCL cell line models with REL amplification (WSU DLCL2, DOHH2) or REL WT (DLBCL1, LIB, Pfeiffer). Not only was this lethal for lines with REL amplification (p=4.9x10^-6), but transcriptome analysis revealed an acute upregulation of the memory B cell theme, supporting the notion that c-Rel actively blocks memory B cell differentiation.

Conclusions

Our study revealed an unanticipated role for tumor subclones in generating intra-tumoral biological diversity among the malignant cells. In particular, our analysis suggests a functional role for the highly recurrent REL amplification in GCB DLBCL, which may be selected for its ability to block terminal memory B cell differentiation. Our paired scRNA and scATAC sequencing represents a significant resource for the lymphoma research community that illuminates the pathobiology of these malignancies.