Cell of Origin (COO) determination
The cell of origin of DLBCL explains part of the heterogeneity of the disease. Using gene expression profiling, DLBCL NOS was initially classified into germinal center B cell (GCB), activated B-cell (ABC), and non-classifiable types (20). GCB type DLBCL is thought to derive from centroblasts while the ABC type has features reminiscent of a B-cell committed to terminal B-cell differentiation (21) (see Figure 1).
The COO classification could predict overall survival of DLBCL patients and response to R-CHOP therapy. The GCB subtype has a more favorable prognosis compared to the ABC subtype. (22,23,24,25). Determination of cell of origin also has implications on drug therapy in relapsed/refractory disease. Recently, ibrutinib, a BCR inhibitor was found to be effective in relapsed and refractory ABC type DLBCL (26).
The gold standard method for determining COO has been gene expression profiling (GEP) (27). This modality, however, is not available at all centers and many assays require fresh tissue, which may not always be available. In routine practice, various immunohistochemistry (IHC) based algorithms (see Figure 2) such as Hans, Tally, Choi and Visco-Young are used to determine COO for DLBCL (28,29,30,31). IHC methods for COO assignment are more practical and widely available, but are plagued by inter-observer variability that can result in discordant classification compared with the gold standard. Gutiérrez-García et al demonstrated that when compared with GEP, different IHC algorithms; Colomo, Hans, Muris, Choi, and Tally, misclassified cases at a higher rate when defining the GCB subset: 41%, 48%, 30%, 60%, and 40%, respectively (32). Gene-expression profiling and not immunophenotypic algorithms predicts prognosis in patients with diffuse large B-cell lymphoma treated with immunochemotherapy. In this study, while the GEP-defined groups showed significantly different 5-year progression-free survival (76% vs 31% for GCB- and ABC- DLBCL) and overall survival (80% vs 45%), none of the IHC algorithms retained the prognostic impact of the COO groups (GCB vs non-GCB). Other studies have also suggested that IHC algorithms may not have the same prognostic impact as other methods of COO determination (33,34,35). These results underscore the unmet clinical need for a methodology that uses readily available biopsy material for accurate COO classification, concordant with GEP while maintaining prognostic utility.
The COO classification is not without limitations however. COO does not explain all the heterogeneity in the behavior and prognosis of DLBCL. Using various approaches, subgroups within ABC showing favorable prognosis have been identified and adverse prognostic groups have also been identified for the GCB type. In a study involving 574 DLBCL biopsy samples using exome and transcriptome sequencing, array-based DNA copy-number analysis, and targeted amplicon resequencing of 372 genes to identify genes with recurrent aberrations, Schmitz et al. identified four DLBCL genetic subtypes: BN2, EZB, N1, and MCD, with the former two determined to have favorable prognosis and the latter two associated with poor prognosis (36). Progression free survival and overall survival varied significantly within these groups. Of note, heterogeneity in behavior within the ABC group could be identified, with inferior survival in the MCD and N1 subtypes of ABC and favorable survival in the BN2 subtype. In GCBs, EZB subtype had a worse predicted 5-year survival compared with other GCBs non-classifiable by this system. Significant independent and additive contributions to survival by gene expression profiling (ABC vs GCB) were noted in this genetic subtyping model.
Chapuy et al used a combination of recurrent mutations, somatic copy number alterations
and structural variants (SV) to identify five distinct groups of DLBCL including a hitherto unappreciated group of ABC-DLBCL with favorable prognosis (37). They also identified a subtype of GCB-DLBCL with poor prognosis characterized by mutations and structural variants ofBCL2, mutations in PTEN and chromatin modifiers such asKMT2DCREBBP , and EZH2 and focal10q23.31/PTEN  loss. These subgroups within the different COO types provide insight not only for prognosis but also suggest possible mechanisms for therapeutic intervention. In spite of the usefulness of these multiparametric approaches for classification, it will be challenging to apply these genetic classification models in the clinical settings due to limitations in the amount of diagnostic material available, resources and bioinformatics expertise.
Disease Monitoring During Therapy
The rationale of using multiple agent chemotherapies in the treatment of cancers is to avoid development of resistance, but resistance does develop regardless of the type and combination of therapy. The neoplastic cells may develop adaptive resistance to chemotherapy in the course of treatment. Serial tumor profiling can provide insight into the pathways mediating adaptive resistance and identify targets for novel therapeutic drug development to overcome the resistance mechanisms (38). Real time analysis may also allow prompt detection of resistance and suggest alteration of therapy if signatures of resistance portending refractoriness or relapse are detected. In DLBCL patients, it is not usually feasible to obtain tissue regularly for such profiling without invasive measures, which cause patient discomfort and increase the risk of infections from multiple biopsy procedures. Thus, a non-invasive method for tissue sampling at multiple time points would be of great benefit.
Post Treatment Surveillance
The role of post-therapy or remission imaging surveillance in the management of DLBCL patients is controversial. While the National Comprehensive Cancer Network (NCCN) and European Society of Medical Oncology (ESMO) guidelines recommend surveillance imaging for stage III/IV disease for the first two years after completion of front-line therapy (39), the 2014 Lugano classification system advises against routine surveillance imaging (40). The positive predictive value of post treatment PET is low, resulting in patient anxiety and increased costs for these unnecessary medical procedures (41,42). Studies have also failed to establish a survival advantage in imaging detected disease relapse (43,44). Realization of the lack of survival benefit of surveillance imaging has resulted in decreased rates of surveillance, although this practice is still quite frequent, with over half of DLBCL patients diagnosed in 2014-2016 undergoing surveillance imaging (45). Thus, there is a pressing need for more specific and sensitive methods for detecting recurrence of disease while also avoiding risks of radiation exposure (46).
Novel Methodologies in DLBCL Management- Liquid Biopsy Based Approaches
Liquid biopsy techniques which involve assessment of cancer related biomarkers in bodily fluid samples, have potential to resolve the unmet clinical needs in management of DLBCL. Common liquid biopsy techniques include circulating tumor cells (CTC), circulating tumor DNA (ctDNA) and exosomes.