Predictive biomarkers for inhibition of B-cell receptor signaling pathway in activated B-cell (ABC) diffuse large B-cell lymphoma (DLBCL)

Abstract

Background: Gene expression profiling (GEP) has been used to distinguish molecular subtypes of diffuse large B-cell lymphoma (DLBCL) based on their cell-of-origin (COO); germinal center B-cell (GCB) and activated B-cell (ABC). ABC DLBCLs have distinct genetic features and are associated with relatively poor prognosis. ABC subtyping is typically performed using a combination of immunohistochemistry (IHC) with CD10, BCL6, and MUM1, with the introduction of digital GEP potentially offering a more accurate and convenient method. In this study, I performed digital GEP on high risk, non-GCB type DLBCL samples, to better understand the clinical importance of COO, as well as to identify potential genetic biomarkers in these high risk patients. Methods: For digital GEP, NanoString analyses were performed on formalin-fixed paraffin-embedded (FFPE) DLBCL tissues. Clinical samples were collected from patients who were diagnosed with DLBCL between January 2008 and December 2015 at Severance Hospital, based on a search of medical records. Inclusion was limited to transplant-eligible patients with high risk clinical parameters, and predefined as non-GCB types by IHC. COO was reanalyzed by Lymph2Cx algorithm with 20-gene panel. Additional 13 genes related to ABC-specific signaling pathways were included as part of the NanoString assay. Results: A total of 69 non-GCB DLBCLs were selected for inclusion. They designated as high risk based on stage (95.7%), International Prognostic Index (IPI) score ≥3 (63.8%), and elevated lactate dehydrogenase (LDH) (65.2%). Patients were treated with 6-8 cycles of standard chemoimmunotherapy, with a median follow up duration of 52.1 months. Upfront autologous stem cell transplantation (ASCT) was performed in 27 responders. Serum LDH level, treatment response, and upfront ASCT were identified as important factors for survival. The benefit of upfront ASCT was greater in males than in females. The NanoString assay was performed in 59 cases, of which 19 discrepancies (27.3%) were identified, resulting in reclassification from non-GCB type into GCB or unclassifiable. COO itself did not affect survival outcomes. The prognostic value of gene expression levels was screened using a COX proportional hazard model. Significant associations (p<0.1) included CCDC50 for PFS; CYB5R2, LIMD1, ITPKB, Aiolos, CRBN, and IKKbeta for OS; and RAB7L1, ASB13, S1PR2, and IRAK4 for both. Among these, low RAB7L1 (<641.25) was predictive of worse PFS (5-year rate, 37.7% vs. 78.0%, p=0.003) and OS (5-year rate, 45.5% vs. 87.0%, p=0.004). Low S1PR2 (<78.85) was also associated with poor PFS (5-year rates, 50.0% vs. 68.7%, p=0.026) and OS (5-year rates, 51.4% vs. 77.3%, p=0.018). In multivariate analyses, low RAB7L1 was identified as an independent biomarker for PFS and OS with a hazard ratio (HR) of 4.53 (95%CI 1.33-15.39, p=0.015) and 8.05 (95%CI 1.42-45.65, p=0.018) in high-risk/ABC DLBCL. Low S1PR2 was an independent predictor for PFS with a HR of 5.42 (95% CI 1.01-28.97, p=0.048), but not OS, in high-risk/ABC DLBCL. Conclusion: Digital GEP using the Lymph2Cx was able to accurately distinguish patients based on ABC subtype, however, COO alone was not a prognostic factor in transplant-eligible patients with high-risk DLBCL. Upfront ASCT may improve survival in such a high-risk group. In addition, low RAB7L1 and S1PR2 were associated with poor prognosis, but this finding requires further validation.