A Multiomic Analysis of Waldenstrom's Macroglobulinemia Identifies Three Subtypes of Disease Based on Impaired Plasma Cell Differentiation

Introduction

Waldenstrom’s macroglobulinemia (WM) is a lymphoplasmacytic lymphoma which recent DNA methylation studies have shown to exhibit multiple phenotypes. To enhance disease classification and explore the features and potential mechanisms underlying these subtypes, we performed a single-cell (sc) multiomic analysis on a series of MYD88 mutated WM cases, complemented by bulk RNA-seq and whole genome sequencing (WGS).

Methods

Single-cell multiomic analysis was performed on flow sorted CD19+/CD3- mature B-cells from 13 MYD88-mutated WM patients and was analyzed alongside reference B-cell populations derived from healthy tonsils. Patient-matched scRNA-seq and scATAC-seq data were preprocessed in Seurat and ArchR prior to modality integration in ArchR. We employed a combined automated and manual cell type annotation approach using the CellTypist and TRUST4 packages to identify cell types and assess clonality. Enrichment of transcription factor motifs at the single-cell level was inferred using the chromVAR package and pseudotime analyses were performed in ArchR. A second series of WM patients (n = 36) underwent bulk RNA-seq and a subset of both sc and bulk patients (n = 32) also had concordant WGS data. WGS data were preprocessed and analyzed for somatic variants using our genomics pipeline, the MGP1000.

Results

We show that the dominant feature of each patient’s tumor phenotype was that of a clonally expanded memory B-cell (MBC) population. The analysis revealed three disease subtypes—MBC-like, PC-like, and intermediate—based on the presence of plasma cells (PC) and gene expression patterns of PC and MBC markers. Comparing the clonal MBC to healthy MBC showed that each subtype differentially expressed a set of unique genes but shared the bulk of differentially enriched transcription factor motifs, which were mostly related to cell proliferation and aberrant B-cell transcriptional regulation. Broadly, the transcriptional landscape of the MBC-like points to an excess of BCR signaling, the PC-like exhibit enhanced canonical NF-kB signaling and unfolded protein response, and the intermediate shared features of both. Pseudotime trajectory analysis using ArchR shows that WM is characterized by clonally expanded MBC with variably blocked capacities for plasma cell differentiation, with the MBC-like unable to differentiate into PC and the PC-like unable to complete terminal differentiation into a normal mature PC. Genes most correlated with pseudotime include SPI1, SPIB, BCL11A on the MBC end and XBP1, POU2AF1 (OCT1), and NFKB1 on the PC end. We validated the existence of the three disease subtypes using hierarchical clustering of bulk RNA-seq data. A novel WGS analysis of patients with expression data identified mutations in MYD88 (97%), IGLL5 (52%), CXCR4 (24%), HIST1H1E (18%), ARID1A (15%), and MAP3K14 (NIK) (12%), among others. The MBC-like patients carried all of the NIK mutations in addition to the majority of CXCR4, HIST1H1E, and ARID1A mutations (7/8, 5/6, and 4/5, respectively). For somatic copy number abnormalities, we identified del 6q (27%), del 22p (27%), trisomy 4 (21%), and del 13p (21%), among others. The majority of del 6q were found in the PC-like and intermediate subtypes (7/9).

Conclusions

We show the existence of 3 subtypes of WM with distinct transcriptional and chromatin accessibility profiles, which demonstrates that WM is a disease characterized by a failure to complete normal differentiation from an MBC to a PC. These subtypes were validated in a second series of WM patients using hierarchical clustering of bulk RNA-seq data. Compared to healthy populations, all subtypes are enriched in genes and open transcription factor motifs pointing to aberrant B-cell transcriptional regulation, enhanced proliferative ability, and immunoglobulin production. The MBC-like subtype has mutated CXCR4, HIST1HE, ARID1A, and NIK, demonstrates a unique gene expression profile suggestive of chronic BCR signaling, and exhibits activity of transcription factor motifs reflecting blocked PC differentiation including SPI1, SPIB, and BCL11A. In contrast, the PC-like and intermediate subtypes are enriched in del 6q, have a transcriptional landscape pointing to upregulated canonical NF-kB and an unfolded protein response, and are able to differentiate into early PC but not normal mature PC via enrichment of the transcription factors XBP1, POU2AF1, and NFKB1.