Cell of Origin and Genomic Profile of Primary Central Nervous System Lymphoma Determined Using the Nanostring Lst Assay and Ultra Deep Targeted Next Generation Sequencing

Introduction: Primary central nervous system lymphoma (PCNSL) is characterized by a consistently inferior outcome compared to systemic diffuse large B-cell lymphoma (DLBCL) with a median overall survival of 2-4 years. Therefore, PCNSL represents a significant clinical challenge with an urgent need for better biomarkers and novel therapeutic targets. The distinction of DLBCL into cell of origin (COO) categories based on patterns of gene expression reminiscent of germinal center B-cell (GC-) type, activated B-cell (ABC-) type, with a small number of "unclassified" cases, has profound prognostic and potential therapeutic implications with many novel therapies displaying differential efficacy between the GC and ABC groups. Therefore, the precise determination of COO is becoming increasingly important for the appropriate patient stratification for novel targeted therapies. The NanoString LST assay is a digital gene-expression based test for COO assignment, representing a more accurate means of COO determination compared to standard immunohistochemistry (IHC) based algorithms, and was recently successfully applied in nodal DLBCLs. In the present study, we tested the LST assay on a cohort of PCNSL patients and performed a complementary targeted genomic profiling on a subset of these patients.

Methods: The COO of 77 PCNSL patients was determined with the NanoString LST assay using RNAs extracted from FFPE samples and compared with the results obtained by the Hans IHC algorithm. The genomic profiles of 62 PCNSL cases were determined using ultra-deep targeted next generation sequencing (NGS) of 13 recurrently mutated target genes, including CARD11, CCND3, CD79B, CSMD2, CSMD3, IRF4, C-MYC, MYD88, PAX5, PIM1, PRDM1, PTPRD and TP53 using the TruSeq Custom Amplicon approach (dual strand workflow) on a HiSeq4000 Instrument (Illumina) to an average depth of 30.000x. Only variants supported by more than 100 mutant reads with an allele frequency (VAF) of >5% were considered.

Results: The LST assay resulted in the following subtype calls: 80.5% ABC, 12.9% GC and 6.5% unclassified (UC). Using the Hans IHC algorithm, 94.8% (73/77) of the PCNSL cases were classified as ABC subtype with 5.2% (4/77) representing the GC subtype. Interestingly, the COO classification obtained using the LST assay showed a different outcome in 16.9% (13/77) of the cases: Out of the 12 cases classified as ABC subtype with the IHC algorithm, 7 cases presented as GC subtype and 5 showed an UC profile using the LST assay. On the other hand, one case characterized as GC subtype with the IHC assay was classified as ABC when analyzed using the LST assay. As for the mutation analysis, we detected a total of 402 somatic mutations in the 62 PCNSL cases across the 13 genes analyzed with individual cases harboring mutations in 4 genes on average (range: 1-11). The most frequently mutated genes were MYD88 (76%), PIM1 (55%) and PRMD1 (50%). The mutation frequencies in the remaining genes were as follows: CD79B (40%), PTPRD (32%), MYC (31%), PAX-5 (31%), CARD11 (26%), TP53 (26%), IRF4 (23%), CCND3 (19%), CSMD2 (19%) and CSMD3 (18%). Analyzing the mutation profiles of the ABC and GC subtypes, we observed an enrichment of CD79B (47% vs 20%) and PIM1 (59% vs 20%) in cases with ABC subtype, with IRF4 mutations being present in ABC cases only (29% vs 0%). None of the other genes demonstrated a considerable difference between the GC and ABC subtypes with regards to the mutation frequency.

Conclusions: We successfully applied, for the first time, the NanoString LST assay for COO determination using FFPE derived RNA samples from a cohort of 77 PCNSL patients. Interestingly, the LST assay revealed a lower proportion of patients with ABC subtype compared our IHC findings (80.5% vs 94.8%). This may reflect the genuine biology of these cases, as the NanoString LST assay represents the most reliable COO determination approach compared to the original gold standard GEP method. The frequent mutations revealed in components of different actionable pathways including the BCR pathway, cell cycle regulation and CNS development highlights the role of these variants in the pathogenesis of PCNSL. Considering the emergence of novel therapies with selective activity in the GC and ABC patient groups the precise assignment of COO and the complementary mutation analysis will most likely have an important role in guiding the appropriate patient management in PCNSL.