Systematic Analysis of Cell-of-Origin, Sequencing and Genomic Imbalances Identifies a Distinct Subset of Rituximab Treated Diffuse Large B-Cell Lymphoma with an Inferior Survival

Introduction: Diffuse Large B-cell Lymphoma (DLBCL), exhibits a wide range of clinical and biological heterogeneity. While initially responsive to rituximab based therapy, more than 70% of patients relapse within 5 years.  Two distinct subtypes of DLBCLs were previously identified by gene expression profiling, that correlated to the cell of origin (COO); currently an immunohistochemistry(IHC) surrogate (Hans algorithm) is used to identify this COO. However, the value of COO identification by IHC is being questioned in patients treated with rituximab based regimen.  As such there is a need for the identification of additional prognostication markers. The purpose of the present study is to systematically characterize a cohort of patients with DLBCL using a variety of methods including COO by IHC, copy number changes by array-based comparative genomic hybridization, and single gene mutations by next generation sequencing.  The goal is to identify prognostic molecular biomarkers that predict survival better than current methods such as COO in DLBCL patients treated with rituximab.

Methods: Patients diagnosed between 2003 and 2011 and treated with R-CHOP/R-EPOCH at BIDMC were identified. IPI/R-IPI, ECOG performance status, overall survival data were collected from retrospective chart review. Molecular and immunohistochemical studies were performed on formalin-fixed paraffin embedded tissue (FFPE), which was obtained prior to initiation of therapy (at the time of diagnosis).

aCGH: Using a previously developed assay for aCGH to detect genomic gain/loss from archival FFPE, we characterized each DLBCL sample for the presence or absence of 50 copy number variations (CNVs) from 32 common regions of overlapping genomic imbalances comprising 36 minimal common regions.  The calling criteria were based on GISTIC defined peaks based on copy number data from three publicly available datasets: IS-172, IS-51HR, EMEXP-3463.

Gene panels: Next Generation Sequencing (NGS) was performed using a targeted hybrid capture panel and run on a Miseq (Illumina, Inc.). Gene selection for the panel was based on frequently mutated genes reported in DLBCL, Follicular lymphoma (FL), and Mantle Cell Lymphoma (MCL). Also selected for the panel were genes involved in known dysregulated pathways, therapeutic targets, and genes mapped to sites of genomic gains or losses.

Cell of Origin: Immunostains were performed for BCL6, CD10, MUM1, FOXP1, GCET and LMO2.  The immunostains were blindly scored by three different pathologists. For this study, GCB vs ABC determination was made using the Hans algorithm (CD10, BCL6 and MUM1 expression).

Statistics: Kaplan-Meier (KM) survival analysis was performed using R (version 3.2.1).

Results:  Data on an initial subset of 49 patients with comprehensive molecular, COO and clinical information is presented (and an additional ~100 case analysis is in progress).  Average age of patients at the time of diagnosis was 63, with 23 males and 26 females. By univariate analysis, the COO was not significantly associated with survival in the patients treated with rituximab based regimen (KM p=0.678).   By contrast, three of the CNVs were associated with survival by Kaplan Meier analysis (Table 1).  Also four, gene mutations were significantly associated with poor survival by univariate analysis (Table 1).

Gene / CR	p value (KM)
Ab29 (D1p13.1)	0.00787
PIK3C2G	0.00126
PIM1	0.00499
CD79B	0.00567
DTX1	0.0102
Ab3(A1q21.1-q25.1(3))	0.0193
Ab28 (D1p36.32-p36.31)	0.0211

Table 1. Markers significantly associated with survival in rituximab treated patients with DLBCL. Ab28 predicted for better survival, whereas the presence of the other mutations or CNVs predicted for poor survival.

Conclusions:  Our systematic analysis of DLBCLs using multiple immunohistochemical and molecular methods identified mutational and copy number biomarkers predictive of survival in DLBCL patients, treated with rituximab in univariate analysis.  Multivariate analysis and data from additional cases will reveal whether these molecular biomarkers can better predict patient survival, compared to current methods (such as COO by IHC).