Diffuse large B-cell lymphomas are a genetically diverse group of diseases that constitute the largest group of non-Hodgkin lymphomas. Previous studies utilizing gene expression profiling clearly delineated three distinct subgroups of DLBCL patients: those with activated B-cell–like (ABC) disease; those with germinal-center B-cell–like (GCB) disease; those with “unclassified” disease (roughly between 10% and 20% of all DLBCL cases) (Nature 2000;403:503–511).
Recently, a team of researchers, including Louis Staudt, MD, PhD, Director of the Center for Cancer Genomics at the NCI, presented results from a study in which 574 tumor tissue samples from DLBCL patients were analyzed using array-based DNA copy number analysis, exome and transcriptome sequencing, and targeted amplicon resequencing of 372 genes to identify those genes having recurrent anomalies (N Engl J Med 2018;378:1396–1407).
Four distinct DLBCL genetic subtypes were identified in this study: MCD (presence of both MYD88L265P and CD79B mutations); BN2 (presence of BCL6 fusions and NOTCH2 mutations); N1 (presence of NOTCH1 mutations); EZB (based on EZH2 mutations and BCL2 translocations).
“We uncovered genetic subtypes of DLBCL with distinct genotypic, epigenetic, and clinical characteristics, providing a potential means of classification for guiding precision medicine strategies in patients with these disease states,” Staudt noted.
Analysis of the fresh-frozen DLBCL biopsy samples was accomplished using deep amplicon resequencing of 372 genes, exome and transcriptome sequencing, and DNA copy number analysis. The vast majority of tissue samples (96.5%) were obtained from the patients prior to treatment. For the patients tested in this study, the breakdown of the DLBCL gene expression profiling subtypes was as follows: ABC—51.4 percent; GCB—28.6 percent; unclassified—20.0 percent.
A tumor-only mutation-calling pipeline was developed because most samples lacked matching normal DNA. However, 48 patients had matched normal DNA samples, from which a random forest-based somatic mutation model was both created and validated. “This model predicted that roughly 94 percent of the mutations called by our tumor-only pipeline are somatic,” Staudt stated.
Regarding the process used to classify the different subtypes of DLBCL, he explained, “To identify genetic subtypes, we created an automated method that started with a set of seed classes and iteratively moved cases into and out of the classes to optimize a genetic distinctiveness metric.
“We chose four seeds as follows: CD79B–MYD88L265P double mutation, NOTCH2 mutation or BCL6 fusion in ABC or unclassified DLBCL, NOTCH1 mutation, and EZH2 mutation or BCL2 translocation. These seed selections, when plugged into the algorithm utilized in this study, gave rise to four different subtypes: MCD (MYD88L265P–CD79B seed), BN2 (BCL6–NOTCH2 seed), N1 (NOTCH1 seed), and EZB (EZH2–BCL2 seed).” The breakdown of the subtypes observed in this study included MCD–71 cases, BN2–98 cases, N1–19 cases, and EZB–69 cases.
Early Genetic DLBCL Subtypes
Genes that were altered at significantly different frequencies (p<0.01) in cases of ABC and GCB DLBCL present in this study's 574 samples were identified using multiplatform genomic analyses. Regarding the unclassified DLBCL cases, Staudt noted: “Because the genetic composition of unclassified DLBCL is unknown, we enriched for these cases: 20.0 percent in our cohort, as compared with 11.3 percent in a population-based cohort.”
One genetic feature frequently found in unclassified DLBCL is the co-occurrence of NOTCH2 mutations and BCL6 fusions (p=2.78×10−12), which distinguished unclassified from other DLBCL subtypes. Mutations of NOTCH-dependent gene expression inhibitor SPEN were enriched in unclassified DLBCL cases, with 30.4 percent of cases having SPEN or NOTCH2 mutations. While gain-of-function NOTCH1 mutations were noted in 19 cases, these were primarily in cases of ABC DLBCL (95%) and never co-occurred with NOTCH2 or SPEN mutations. “This suggests that NOTCH1 and NOTCH2 contribute to distinct pathogenetic pathways,” Staudt stated.
“We next investigated whether genetic aberrations were correlated with the ABC–GCB predictor score, a quantitative gene-expression metric with low values for the most GCB-like cases and high values for the most ABC-like,” he said. In the center of the distribution, NOTCH2 mutations and BCL6 fusions were concentrated, which is consistent with the observation that these variants are enriched in unclassified DLBCL. In contrast, cases at the far ABC end of the spectrum had enriched CD79B and MYD88L265P mutations, with significant co-occurrence of these aberrations (p=2.81×10−11).
“Although most NOTCH1 mutant cases were ABC-DLBCL, none had CD79B or MYD88L265P mutations; their predictor score values were also significantly lower than those of the CD79B–MYD88L265P double-mutant cases (p=0.006),” Staudt added. Significant co-occurrence of EZH2 mutations and BCL2 translocations (p=6.39×10−14) were observed in those cases having the lowest predictor scores (i.e., on the GCB end of the spectrum). “Taken together, these analyses suggest that the gene-expression subgroups may have distinct genetic subtypes,” he clarified.
Currently-Identified Genetic Subtypes
For the MCD subtype, 82 percent of these cases have a MYD88L265P or a CD79B abnormality (i.e., a mutation or amplification), with 42 percent having both anomalies co-occurring. “MCD frequently had gain or amplification of SPIB, which encodes a transcription factor that, along with IRF4, defines the ABC phenotype of DLBCL and promotes plasmacytic differentiation,” Staudt noted.
Mutations that inactivate BLIMP1 (PRDM1) and are present in the MCD subtype of DLBCL result in the blockage of full plasmacytic differentiation. “Interestingly,” Staudt commented, “mutations of the tumor suppressor TP53 were observed significantly less frequently in MCD than in other subtypes.
“Immune editing appears to be rather prominent in MCD genomes, with 76 percent acquiring a mutation or deletion of HLA-A, HLA-B, or HLA-C and 30 percent acquiring truncating mutations targeting the natural killer cell activator CD58,” he noted.
Many genetic changes that were enriched in the MCD subtype would have previously been ascribed to the ABC DLBCL subtype. Regarding the MYD88L265P and CD79B co-occurring mutations noted in some cases of MCD DLBCL, Staudt stated, “This genotype has been associated with response to the Bruton's tyrosine kinase (BTK) inhibitor ibrutinib in relapsed or refractory ABC DLBCL; this anomaly is also common in primary central nervous system lymphoma, which has an ABC phenotype and is also often noted to have a response to ibrutinib.”
Extensive extranodal involvement was noted in the tumors of patients with MCD-DLBCL; additionally, genes having acquired mutations in these tumors are also noted to be recurrently mutated in primary extranodal lymphomas. “Together, these observations would seem to indicate that the pathogenesis of nodal MCD DLBCL is related to that of primary extranodal lymphomas,” he explained.
“NOTCH pathway abnormalities were dominant in the BN2 subtype,” Staudt observed, “with 73 percent acquiring an amplification or mutation of NOTCH2, SPEN mutation, or mutation in DTX1, which is a NOTCH target gene.” Many BN2 cases having SPEN mutations (50%) lacked NOTCH2 anomalies, which implied NOTCH2 ligand-induced signaling has a role in BN2 disease. The other BN2-defining anomaly, BCL6 fusion, occurred in 73 percent of those cases. In cases with NOTCH2, SPEN, or DTX1 lesions, BCL6 fusions were noted to a significantly greater extent in BN2 cases than in non-BN2 ones (72% vs. 15%, p=2.31×10−10). This observation would seem to imply that oncogenic cooperation is occurring between these pathways in BN2 DLBCL.
One prominent feature of BN2 DLBCL was the presence of genetic abnormalities affecting regulators of the NF-κB pathway. “Lesions targeting the NF-κB negative regulator A20 (TNFAIP3) or its partner TNIP1 were observed in roughly 55 percent of these cases,” Staudt noted. “Additionally,” he continued, “two components of the B-cell receptor-dependent NF-κB pathway, protein kinase C beta (PRKCB) and BCL10, were altered by mutations or amplifications in 47 percent of BN2 cases.” Mutations for the genes encoding cyclin D3 and CXCR5 likely resulted in gain-of-function events, while immune escape for these tumors was thought to arise, in part, from inactivating mutations targeting the immune regulator CD70.
Concerning BN2 DLBCL, Staudt noted, “This subtype reveals some information about unclassified DLBCL, a previously obscure gene-expression subgroup; additionally, this disease is predicted to rely on B-cell receptor-dependent NF-κB activation, and thus be responsive to antagonists of B-cell receptor signaling such as BTK inhibitors.” NOTCH2 mutations are present in both BN2 DLBCL and marginal zone lymphoma, which is also responsive to ibrutinib. The second main anomaly associated with BN2 DLBCL, BCL6 fusions, are also frequently found in transformed marginal zone lymphoma, which implies that BN2 might arise from an occult marginal zone lymphoma.
The N1 subtype of DLBCL was characterized by the presence of NOTCH1 mutations and abnormalities affecting B-cell differentiation transcriptional regulators such as IRF4, ID3, and BCOR; this may contribute to the subtype's plasmacytic phenotype.
“TNFAIP3 mutations found in biopsies of patients with N1 DLBCL could reinforce this phenotype by fostering NF-κB–induced IRF4 expression,” Staudt elaborated. The highest signatures of T cells, myeloid cells, and follicular dendritic cells were observed in cases of N1 DLBCL, which would likely alter the tumor microenvironment.
“N1 DLBCL varies from that of the BN2 subtype clinically, genetically, and phenotypically; this is despite the functional similarities between NOTCH1 and NOTCH2, which implies a distinct pathogenesis,” he noted.
Genetic anomalies that had previously been ascribed to GCB DLBCL, including BCL2 translocation, EZH2 mutation, and REL amplification were enriched in EZB DLBCL. In addition, inactivation of the tumor suppressors CREBBP, EP300, KMT2D, and TNFRSF14 were also enhanced in EZB DLBCL. In 38 percent of EZB cases, the germinal center homing pathway, which involves S1PR2 and GNA1314, was disrupted. In 49 percent of cases, JAK-STAT signaling may have been promoted by mutation or amplification of STAT6 or by a mutation or deletion targeting the JAK-STAT negative regulator SOCS1.
MTOR mutations or amplification of MIR17HG, which encodes a microRNA targeting PTEN, may have been responsible for the activation of the PI3K/MTOR signaling noted in 23 percent of cases. It is also possible that EZB genomes may be altered by immune editing, as 39 percent acquired anomalies in the major histocompatibility complex class II pathway genes CIITA and HLA-DMA.
“In summary, most of the genetic lesions previously associated with GCB DLBCL were concentrated in the EZB subtype, reflecting a shared genetic pathogenesis,” Staudt stated, “which distinguishes them from other GCB tumors.”
“In our study, patients having different subtypes of DLBCL were noted to have different responses to immunochemotherapy. Those with BN2 or EZB disease tended to have more favorable responses to this treatment, while poorer outcomes were noted for those having N1 or MCD DLBCL,” Staudt explained. “Moreover, within ABC DLBCL, disparity in clinical outcomes can be traced, in part, to genetic heterogeneity, with inferior survival in the MCD and N1 subtypes and favorable survival in the BN2 subtype.” These findings should be considered when interpreting the results of clinical trials involving patients with ABC DLBCL who receive R-CHOP (rituximab plus cyclophosphamide-doxorubicin-vincristine-prednisone)–like therapy in the light of these genetic variations.
“The results of our studies suggest that, in clinical trials, targeted agents in DLBCL could be evaluated in the context of particular genetic subtypes or genetic aberrations that affect the targeted pathway,” he noted. “Drugs that target B-cell receptor–dependent NF-κB activation (e.g., inhibitors of BTK and protein kinase C beta) could be investigated in BN2 and MCD, given the enrichment for genetic aberrations in those subtypes that should activate or augment this signaling.”
When asked about potential future studies in this area, Staudt replied, “Our survival analysis tested a single hypothesis on the basis of a locked-down genetic-subtype algorithm that did not use the clinical data. However, evaluation of the relationship between these genetic subtypes and treatment response in additional cohorts will be crucial in order to confirm and extend these findings.
“Clinical trials evaluating B-cell receptor proximal signaling inhibitors (e.g., spleen tyrosine kinase inhibitors) or the downstream PI3K pathway could investigate whether there is a correlation between response and lesions that alter negative regulators of B-cell receptor signaling or the B-cell receptor subunits CD79a and cd79b; additionally, differing BCL2 expression levels could be considered in the assessment of response to BCL2 inhibitors.
“Finally, the use of immune checkpoint inhibitors could be studied in those having N1 DLBCL, given its prominent T-cell gene-expression signature and poor response to R-CHOP,” Staudt concluded.
Richard Simoneaux is a contributing writer.