Mutation Impact of Targeted Genes in Diffuse Large B-Cell Lymphoma Patients Treated with Ibrutinib

Introduction: Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin's lymphoma, accounting for approximately 30% of newly diagnosed cases in the United States. DLBCL can progress quickly and advanced cases are inconsistently cured with current therapies. Ibrutinib, a first-in-class Bruton's tyrosine kinase (BTK) inhibitor, is approved in the US and EU for patients with previously-treated mantle cell lymphoma and chronic lymphocytic leukemia (CLL) who have had one prior treatment, CLL with 17p deletion, and Waldenström macroglobulinemia. The activated B-cell (ABC) subtype of DLBCL is considered especially high risk and is characterized by chronic active B-cell receptor signaling, which is blocked by ibrutinib.  Recent phase 2 trial results of ibrutinib as a single agent in DLBCL patients show an overall response rate of 41% in the ABC subtype (Wilson et al. ASH 2012; Wilson et al. Nat Med, 2015). Through targeted deep sequencing, we investigated the impact of baseline mutations of 317 targeted genes on clinical response of 51 DLBCL patients treated with ibrutinib. Based on our mutation impact analysis, we identified potential biomarkers for predicting DLBCL patient response to ibrutinib. In particular, we found sets of gene mutation patterns indicating poor (or good) clinical response across all subtypes (ABC, germinal-center B-cell–like [GCB], non-GCB) of DLBCL as well as uniquely within a subtype.

Methods: H&E-stained slides from DLBCL patients enrolled in either PCYC-04753 (NCT00849654) or PCYC-1106 (NCT01325701) were reviewed to ensure sufficient nucleated cellularity and tumor content. DNA and RNA were extracted from unstained sections of FFPE DLBCL tumor biopsies. Sequencing was performed using the FoundationOne™ Heme panel following the validated NGS-based protocol to interrogate complete DNA coding sequences of 405 genes, as well as selected introns of 31 genes involved in rearrangements. RNA sequences of 265 commonly rearranged genes were analyzed to identify gene fusions. A subgroup of samples were analyzed on earlier versions of the FoundationOne™ panels where only DNA was extracted and sequenced. Sequence data was processed and analyzed for base substitutions, insertions, deletions, copy-number alterations, and selected gene fusions. Mutation impact indices (MII) of 317 genes were calculated and plotted for overall gene mutation pattern recognition. Chi-square association tests were performed on cases where sufficient sample sizes were available to determine statistical significance of mutation impact. DLBCL subtype classifications by gene expression profiling (GEP) and Hans’ IHC were investigated and compared.  For GEP, we used OmicSoft ArrayStudio’s classification module to build linear discriminant analysis (LDA) model/classifier and neural networks with 5-fold cross validation procedure for model selection. The LDA was the best performing model and was selected for final GEP classification. Since only 29 (of 51) patients had central lab Hans’ IHC classification information, we compared trends of the mutation impact results based on Hans’ classification and GEP classification.

Results: Single or multiple gene MII were generated from baseline tumor biopsies from DLBCL patients treated with single agent ibrutinib. The MII were generally consistent between GEP or Hans’ IHC classification of tumor biopsies. We found novel baseline gene mutations associated with poor clinical response (SD or PD) to ibrutinib. These genes were involved in regulation of transcription (eg, mutations in EP300 in all DLBCL subtypes combined group [p=0.034], mutations in RB1 in ABC-DLBCL [p=0.031]), epigenetic modification (eg, mutations in MLL2 in ABC-DLBCL [p=0.053]), programmed cell death (mutations in BCL2 in all DLBCL subtypes [p=0.096]), and PI3K-AKT-mTOR pathway (eg, mutations in TSC2 in ABC-DLBCL [p=0.031]). Mutations identified as indicating good clinical response include mutations in CD79B (p=0.072) and MYD88 (p=0.024) in ABC-DLBCL. Co-existence of MYD88 and CD79B mutations (double-mutants) in ABC-DLBCL showed a stronger association to good clinical response (p=0.004) consistent with recent observations from Wilson et al. ASH 2012 & Nat Med, 2015.

Conclusions: Our investigation reveals unique mutation patterns that underlie DLBCL subtypes and highlights the need for personalized medicine approaches to treating DLBCL patients.