Whole-Genome Sequencing of Primary Central Nervous System Lymphoma and Diffuse Large B-Cell Lymphoma

Introduction

Primary central nervous system lymphoma (PCNSL) is a rare subtype of non-Hodgkin's lymphoma. Although most cases (~95%) show histology of diffuse large B-cell lymphomas (DLBCLs), PCNSL shows very different biological and clinical characteristics from systemic DLBCL. Nevertheless, our knowledge about the molecular pathogenesis of PCNSL and genetic differences between both lymphomas are still incomplete.

Method

To obtain a comprehensive view of the genetic alterations, including mutations in non-coding regions as well as structural variants (SVs), we performed whole-genome sequencing (WGS) of 22 PCNSL cases. Subsequently, to unravel the genetic differences between PCNSL and systemic DLBCL, we re-analyzed WGS data from systemic DLBCL cases (N = 47) generated by the Cancer Genome Atlas Network (TCGA) and Cancer Genome Characterization Initiative (CGCI) using our in-house pipeline. The mean depth of WGS for tumor samples were 49X and 37X for PCNSL and DLBCL cases, respectively. Whole-exome sequencing (WES) was also performed for an additional 37 PCNSL cases to reliably capture driver alterations and also to analyze mutational signatures in PCNSL, which were compared to those obtained from the WES data for DLBCL from TCGA (N = 49).

Results

WGS identified 10.5 and 5.6 mutations per mega-base on average in PCNSL and DLBCL, respectively. We first explored the density of somatic mutations and identified 64 and 33 genomic loci showing significantly high mutation densities in PCNSL and DLBCL, respectively. In PCNSL, most of these loci corresponded to known targets of somatic hypermutations (SHMs) induced by activation-induced cytidine deaminase (AID), including those for IG genes (IGK, IGH and IGL), BCL6, and PIM1, as well as those for known driver genes, such as MYD88 and CD79B. Although most of the hypermutated regions were overlapped between PCNSL and DLBCL, some regions were differentially affected by hypermutations between both lymphoma types. For example, BCL2 and SGK1 loci were frequently affected by SHMs in germinal center B-cell (GCB) DLBCL, while not in PCNSL. In terms of non-coding driver mutations, we identified frequent mutations in a PAX5 enhancer region in 8/22 (36%) of PCNSL and 18/47 (38%) of DLBCL cases.

SVs were common in both lymphoma types, where 104 (PCNSL) and 57 (DLBCL) SVs were detected per sample. SV clusters were identified in 34 (PCNSL) and 13 (DLBCL) regions, of which several clusters were commonly seen in both PCNSL and DLBCL, and included IG loci, BCL6, FHIT, TOX and CDKN2A. In PCNSL, SVs were clustered within the loci for known targets of SHMs, such as BCL6, BTG2 and PIM1. As was the case with somatic mutations, the SV cluster corresponding to BCL2 was only seen in DLBCL. We then analyzed these clustered breakpoints for their proximity to known sequence motifs targeted by AID (CpG and WGCW). Breakpoints of SVs found in the targets of SHMs, including PIM1, BCL6, BTG2 and BCL2, showed an enrichment at or near the CpG, supporting the involvement of AID in the generation of these SVs. By analyzing these SV clusters, we identified several novel driver genes in PCNSL. For example, WGS and WES identified an enrichment of breakpoints of deletions (7/22) and loss-of-function mutations (6/37) in GRB2, strongly indicating its tumor suppressor role in PCNSL.

We also analyzed pentanucleotide signatures of mutations in coding sequences detected by WES of PCNSL and DLBCL, taking into consideration the two adjacent bases 3' and 5' of the substitutions as well as transcription strand biases. Two predominant mutational signatures were identified in PCNSL: the AID signature characterized by C>T mutations within the WRCY motif targeted by SHMs and the age-related signature involving C>T transition at CpG dinucleotides. For DLBCL, an additional signature (signature 17 according to Alexandrov et al.) was detected as well, which had been reported in DLBCL with an unknown mechanistic basis.

Conclusions

Comprehensive genomic analyses of a large cohort of PCNSL and DLBCL cases have revealed the major targets of somatic mutations and SVs, including novel driver genes. In both PCNSL and systemic DLBCL, an enhanced AID activity is thought to be associated with generation of both SHMs and SVs, although the activity and targets of AID seem to substantially differ between both lymphoma types, suggesting distinct pathogenesis therein.