Analysis of the Human CD5-Positive B-Cell Compartment As Cell-of-Origin of Chronic Lymphocytic Leukemia at Single-Cell Level

CLL is defined by clonal expansion of mature CD5-positive B cells with low expression of CD79 and CD20. CLL is subclassified into cases expressing a mutated BCR (M-CLL) and a more aggressive subtype with an unmutated BCR (UM-CLL). Bulk gene expression profiling with oligonucleotide microarrays has suggested that UM-CLL originates from naïve CD5-positive B cells regularly present in healthy humans. Rare memory CD5⁺CD27⁺ B cells have been proposed as the cell-of-origin (COO) of M-CLL (Seifert, J Exp Med 2012).

To revisit and to refine potential COO for CLL, we performed combined gene expression profiling and BCR sequencing on single normal and CLL-phenotype CD5⁺ B cells in healthy sibling stem cell donors (SCD) and single CLL cells of their sibling patient undergoing allogeneic stem cell transplantation (SCT). Flow cytometry of peripheral blood detected a distinct population of CD5⁺CD20^lowCD79^low B cells of 5 of 16 screened SCD. These CLL-phenotype cells were isolated by cell sorting and mixed in equal cellular numbers with normal CD5⁺CD20^hiCD79^hi cells sorted in parallel. These CD5⁺ B-cell mixes were multiplexed in 2 pools with CLL cells from the related patients and 2 unrelated high-count MBL samples (hcMBL). Both cell pools were analyzed for 5’ gene expression profiling and BCR sequencing on the 10X Genomics platform. Single-cell data were processed using Seurat using default filtering settings, demultiplexed based on expressed SNP, and assigned to individual samples based on abundantly present HLA transcripts.

All cells of any CLL and of 1 hcMBL sample expressed completely identical BCR; 4 CLL expressed an unmutated BCR corresponding to their unfavorable clinical course requiring SCT. SCD CD5⁺CD20^hiCD79^hi cells expressed unique, mostly completely unmutated BCR (n=2003; median identity to germ-line IGHV = 100%) without any evidence for clonal expansion. In contrast, SCD CD5⁺CD20^lowCD79^low populations contained 436 singleton cells and 1391 cells belonging to 1 to 8 unique expanded clones with fraction sizes of 0.03-0.89 per case. In the 2nd hcMBL, 104 of 1184 cells belonged to 8 minor clones unrelated to the dominant hcMBL clone. Despite a median 100% IGHV identity to germ-line, BCR of unexpanded CD5⁺CD20^lowCD79^low cells carried significantly more somatic mutations than CD5⁺CD20^hiCD79^hi cells (p<0.0001). BCR expressed by expanded CD5⁺CD20^lowCD79^low SCD cells had a median IGHV germline identity of 94.5% (range 89.9-100%; p<0.0001). Comparison of BCR of both single and expanded CD5⁺CD20^lowCD79^low cells with paired in silico generated germline counterparts and a publicly available BCR repertoire showed significantly increased C>T transitions and preferential mutations of AID motifs, strongly suggesting that BCR were altered by somatic hypermutation during physiological antigen-driven immune responses. Only 0.9% of CD5⁺CD20^hiCD79^hi and 2.2% of unexpanded CD5⁺CD20^lowCD79^low cells expressed CLL subset-defining stereotypic BCR. In expanded CD5⁺CD20^lowCD79^low SCD clones this percentage increased to 8.3% but no major subset BCR stereotypes were present.

Combined gene expression clustering of all cells showed a single CD5⁺CD20^hiCD79^hi cell cluster from all 5 SCD and scattered CD5⁺CD20^lowCD79^low cell populations that partially overlapped with hcMBL and CLL clusters. Using CD20 expression as anchor, pseudotime analysis demonstrated a continuum from normal CD5⁺CD20^highCD79^high cells over non-expanded to expanded CD5⁺CD20^lowCD79^low cells and to hcMBL. Unmutated CLL branched from the non-expanded CD5⁺CD20^lowCD79^low in opposite direction. Pathway analysis showed decreased levels of genes involved in cytokine signaling for both malignant and nonmalignant CD5⁺CD20^lowCD79l^ow cells. CLL cells had increased expression of the oxidative phosphorylation gene set. Increased oxidative consumption rates of CLL compared to hcMBL and normal CD5^neg B cells was confirmed by metabolic analyses.

In conclusion, single-cell analysis fails to reveal distinct naïve and memory states within the normal human CD5⁺ B-cell compartment as COO for CLL. Rather, gradual evolution occurs from naïve CD5⁺ cells over nonexpanded CLL-phenotype cells to eventual clonal expansion, presumably during antigen-driven immune responses. Gene expression profiles suggest a continuous further development to hcMBL and presumably M-CLL. In contrast, UM-CLL emerges via a different evolutionary trajectory.