Dissecting Diffuse Large B Cell Lymphoma Heterogeneity: New Insights from a State-Specific Molecular Characterization

Introduction.

Diffuse Large B cell Lymphoma (DLBCL) is the most frequent aggressive lymphoma, originating either from germinal-center (GCB) or post-germinal cells (ABC). Despite a 60% response rate to first-line therapy (R-CHOP) DLBCL remains a complex disease. Comprehensive evaluation is hindered by low input and artifact associated with FFPE material.

Aim.

We investigated the capacity of transcriptional and mutational profiling to dissect DLBCL B cell population heterogeneity on 178 FFPE DLBCL samples from patients treated with R-CHOP.

Methods.

Cell-of-Origin (COO) was determined using Lymph2Cx (Nanostring) (GCB=81, ABC=61, Unclassified=31, not determined=5). RNASeq-based gene expression profiling and digital deconvolution with EcoTyper were performed after normalizing data for transcript integrity number to account for FFPE-related fragmentation. Targeted sequencing of 169 DLBCL-associated genes (CappSeq panel) was used to identify (likely) pathogenic mutations (VAF ≥ 10%, by REVEL). Regulons identification was performed on public single-cell RNA sequencing dataset (EGAS00001004335) using pyscenic python package on the non-malignant component. Regulons enrichment and significance were assessed with decoupler mlm function.

Results.

EcoTyper analysis delineated five distinct malignant B cell states (S1-S5) representing different differentiation stages. State-associated event free survival curves showed the best outcome in S1 and the worst outcome in S5 B cell states, both in all cases (P=0.0029) and in R-IPI poor cases only (n=56; P<0.05). The B cell state (S1, S2-S4, S5 by recursive partitioning) remained independent prognostic marker in a multivariate model (n=143) along with COO and R-IPI (P<0.05). Mutational analysis identified a mean of 7 mutations per sample (range 0-28), with KMT2D, NOTCH2, TP53, MYD88, CD79B, CREBBP, EZH2, PIM1 among the most frequently mutated genes. LymphGen classification showed trends similar to literature, although FISH data were not available, without specific association with B cell states. An in-house pipeline based on Z-score was used to identify B cell state-specific mutations enrichment. Focusing on the S1 and S5 B cell states only (n=57), S1 was enriched for PAX5, H1-4, EZH2, IRF8 and SOCS1 mutations, while S5 mainly harbored CD79B, MYD88, TBL1XR1, IRF4, PIM1 and PRDM1 mutations. This mutational profile was confirmed by the re-analysis of an external fresh-frozen DLBCL cohort (n=96; PMID:33792219).
Regulon analysis on bulk RNASeq data was used to investigate the impact of S5 or S1 B cell state-specific gene mutations on transcription factors and regulatory pathways modulation. In particular, deregulated regulons in S5 B cell state included: i) regulons associated with the transcription factors CHD2, TBL1XR1, IRF4, involved in the block of B cell differentiation via alteration of the NCoR/SMRT complex and the modulation of B-cell receptor and NF-kB signaling pathways (all altered by IRF4, CD79B, TBL1XR1 mutations); ii) regulons associated with the transcription factors FOXP3 and PRDM1 (altered by PIM1, PRDM1 mutations) compatible with enhanced Th2 differentiation. Interestingly, the majority of these regulons resulted upregulated in a S5 vs S1 B cell state comparison, in keeping with the “gain of function” effect of the underlying mutations. Overall, these findings were consistent with the aggressiveness, the more frequent extranodal involvement and the poorer prognosis observed in S5 B cell state DLBCL patients.
On the contrary, the majority of regulons associated with S1 B cell state-specific mutations were overall downregulated, in keeping with the enrichment in “loss of function” mutations occurring in this B cell state. In detail, S1 B cell state-deregulated regulons included regulons associated with the transcription factors EZH2, BCL6, MYC involved in epigenetic regulation (altered by EZH2, IRF8 mutations), and with NFKB2, BCL3, REL involved in the NFkB-mediated microenvironment interactions (altered by SOCS1, IRF8 mutations). These observations align with the reported negative regulations of the EZH2 and BCL3-associated pathways in good prognosis subtypes.

Conclusion.

Our findings underscore the importance of B cell state-specific molecular characterization to better understand DLBCL pathogenesis and improve risk stratification identifying cases potentially eligible for new therapy approaches.