Studies Must Refine Genomic Classification Tools to Improve Use of Targeted Therapies in DLBCL Subgroups

September 30, 2020

A review was conducted to understand how precision medicine strategies can be successfully implemented in diffuse large B-cell lymphoma

Diffuse large B-cell lymphoma (DLBCL) is a clinically and molecularly heterogeneous disease in which efforts to individualize therapy remains limited. However, comprehensive multiplatform genomic analyses have sparked new knowledge that can be used to facilitate a major paradigm shift, allowing for the use of precision medicine-based approaches.

A review published in the Journal of Clinical Oncology was conducted to understand how precision medicine strategies can be successfully implemented in DLBCL, in which authors described recently proposed genomic categories of DLBCL and the potential ways to personalize treatment in these subgroups. Traditional classification tools used in DLBCL were examined, as well as prior unsuccessful efforts to personalize treatment based on the subtypes of disease.1

Tools for Classifying DLBCL Subtypes

About two-thirds of patients with DLBCL can experience prolonged survival outcomes with frontline chemoimmunotherapy, but the remaining population with relapsed/refractory disease are associated with poor outcomes. A variety of categorization systems are available to help understood the variability in sensitivity to this standard therapy, which is thought to be reflected in the underlying molecular heterogeneity of the disease.

The Issue With Frontline Chemoimmunotherapy

Initial studies evaluated the activate B-cell-like (ABC) and non-germinal center B-cell-like (GCB) subtypes of DLBCL. The proteasome inhibitor bortezomib (Velcade) was expected to overcome the poor prognosis associated with this population, but a randomized phase 2 study demonstrated no improvement with the addition of this agent to chemotherapy. The phase 3 REMoDL-B study also failed to show improvement in progression-free survival (PFS), regardless of cell-of-origin (COO) classification.2

In this study, the 20-month PFS rate was 70.1% (95% CI, 65.0%-74.7%) with R-CHOP alone versus 74.3% (95% CI, 69.3%-78.7%) with the addition of bortezomib (HR, 0.86; 95% CI, 0.65-1.13; P =.28). There were no differences observed in PFS or overall survival for either GCB or ABC subtypes of DLBCL.

The addition of Bruton tyrosine kinase (BTK) inhibitor ibrutinib (Imbruvica) to R-CHOP (rituximab [Rituxan], cyclophosphamide, doxorubicin, vincristine, prednisone) was also evaluated in this patient population. Improved responses were observed in a phase 1/2 clinical trial that prompted development of the phase 3 PHOENIX study that compared ibrutinib in combination with R-CHOP to R-CHOP alone. However, there were no improvements in the event-free survival with the addition of ibrutinib (HR, 0.934; 95% CI, 0.726-1.200; P=.5906).3 An exploratory analysis demonstrated a potential advantage of adding ibrutinib to R-CHOP in patients aged < 60 years and in those with DEL, but the benefit was not limited to those with true ABC disease.

Lenalidomide (Revlimid), an immunomodulatory drug, was also expected to have preferential activity in the ABC subgroup based on findings from preclinical models. The agent showed encouraging findings in a single-arm phase 2 study, but benefit was not observed when the agent was added to R-CHOP in the phase 3 randomized ROBUST clinical trial. However, there was no difference observed in PFS between the 2 arms (HR, 0.85; 95% CI, 0.63-1.14; P =.29).4

Both BCL-2 and MYC overexpression and gene rearrangements are under evaluation as well to determine potential targeted therapeutic approaches. The phase 1b/2 CAVALLI study evaluated the addition of the BCL-2 inhibitor venetoclax (Venclexta) to chemoimmunotherapy, in which the potential predictive impact of COO and DEL was explored. CAVALLI evaluated patients with high-risk disease and demonstrated a complete response rate of 69.2%, which was similar to the historical findings from the GOYA study. The outcomes appeared to improve compared with the historical data, but the improvement was not observed in COO tumors.5,6

The findings from studies incorporating targeted therapy into the maintenance setting have been disappointing. For example, the PRELUDE study demonstrated no benefit with the addition of the PKCβ inhibitor enzastaurin (DB102) after R-CHOP therapy, regardless of COO subtype, and the PILLAR-2 study saw no benefit with the addition of everolimus (Afinitor) after completion of frontline therapy in patients with high-risk DLBCL defined by International Prognostic Index (IPI).

Challenges With Large Studies in DLBCL

Overall, several large phase 3 studies have been conducted and failed to improve treatment with R-CHOP or failed to provide molecularly stratified therapies in DLBCL. This may be in part due to biased enrollment of healthier patients onto clinical trials. Many patients are excluded from patients due to their urgent need to initiate therapy, as well as the time required for trial screening. This leads to exclusion of many clinically aggressive cases in clinical trials, many of which have worse outcomes.

Another concern with these trials is the accuracy of COO classification by immunohistochemistry (IHC). The REMoDL-B study incorporated GEP for COO classification, but considering the challenge in incorporating molecular tools into patients urgently needing therapy, patients did not receive bortezomib treatment until cycle 2 after COO classification was determined. The PHOENIX trial potentially adopted a new solution, which was to enroll patients on the basis of IHC classification but perform GEP and retroactively analyze GEP classification results in prespecified analyses.

Molecular Classification Under Evaluation in DLBCL

IPI has been the most important and successful clinical risk stratification tool in DLBCL, according to the review. Several variants were derived from IPI, including the COO classification and various methods capturing double-hit lymphoma or related subtypes. However, despite its power and value as a reproducible model, it has not improved the stratification of patients as it does not align directly with the molecular heterogeneity of DLBCL. The genomic classification’s prognostic value is complementary to IPI but not sustainable.

The inventive application of gene expression profiling was the first major step toward understanding the genomic complexity of the disease by using hierarchical clustering algorithms on cDNA microarrays of tumors. Two principal subtypes were identified, which included the GCB and the ABC subtypes.

COO provides important prognostic information, but the Hans algorithm, which is based on a wieldier IHC platform, has become more broadly used. However, GEP remains the golden standard, whereas the sensitivity to IHC to assign COO is only ~70% for GCB and ~90% for non-GCB.

GEP technologies have advanced in recent years with the ability to use RNA from formalin-fixed paraffin-embedded tissue. These advancements have allowed for the development of assays that can be more reliably applied to patient samples. RNA-based approaches have become standard in assigning COO for research.

Another transcriptional profiling classification includes the comprehensive consensus clustering (CCC), which identified distinct variants of DLBCL. This is important for identifying distinctions in predominant field utilization pathways that are associated with either the presence or absence of B-cell receptor signaling and features of the tumor immune/inflammatory infiltrate. CCC identifies biologic heterogeneity in DLBCL similar to COO but has a more limited role in clinical practice currently. CCC has not yet been used to test individualized treatment approaches in DLBCL.

Deepening the Understanding of Genomics in DLBCL

Other biologic features that contribute to prognosis include MYC translocations, which tend to have unfavorable outcomes. Tumors harboring MYC and BCL-2/6 translocations, which are known as the double- or triple-hit lymphomas, have been shown to be more chemotherapy-refractory, in which intensified regimens have been associated with better outcomes retrospectively. For example, dose-adjusted R-EPOCH is the preferred intensified regimen in most centers.

Intensified regimens, however, are not targeted therapies, but they represent 1 of the few examples of genomically-stratified therapy in patients with DLBCL. Most double-hit tumors are considered GCB, and efforts to distinguish the remainder of GCB DLBCLs with RNA sequencing has led to the recognition of a distinct molecular subgroup that is characterized by a double-hit signature.

A distinct subgroup was also identified in a GEP of a large DLBCL sample cohort with a molecular high-grade (MHG) profile, which was associated with a significantly worse PFS. Patients with double-hit lymphoma that do have the MHG signature had similar outcomes to those with GCB DLBCL. This suggests that chromosomal rearrangements could be used to identify the more aggressive variant.

Chemotherapy-refractoriness could be better predicted in DLBCL with more sophisticated tools, the authors wrote, which could improve selection of patients for intensified chemotherapy approaches today and more targeted therapies tomorrow.

Identifying Novel Genomic Categories

The advancement of genomic characterization tools over the last decade has allowed for a deeper understanding of recurrent aberrations in DLBCL. For example, the COO transcriptional framework allows for identification of genetic alterations enriched in either GCB or ABC tumors. The inclusion of a broader range of genomic changes has also deepened this understanding considerably in conjunction with new computational tools, gene expression, and functional analyses.

One study conducted by the Harvard group used genomic information from 304 samples from newly diagnosed patients with DLBCL to create a novel classification system. The study included a range of genomic aberrations and classified tumors, which ultimately led to the identification of 5 genomic clusters with salient features.7

Clusters 1 and 5 were significantly enriched for ABC tumors, whereas clusters 3 and 4 were enriched for GCB tumors. Significant prognostic differences were observed in each case, with inferior outcomes in patients with cluster 3 and 5 tumors. These findings demonstrate the significant genomic variability beyond the COO classification, which can help explain the variability of the prognostic impact of COO classification across studies. More importantly, these findings underscore the difficulty of using COO classification to select patients for targeted therapies.

While MYC and BCL-2 alterations occurred together the most commonly in cluster 3, MYC and BCL-6translocations occurred most commonly in cluster 1, making this a biologically distinct group of patients from those in cluster 3, strengthening the idea that double-hit lymphomas may have prognostic relevance but not biologic homogeneity.

The National Cancer Institute (NCI) identified genes with recurrent aberrations among 574 samples from patients with DLBCL by using exome and transcriptome sequencing, array-based copy number analysis, and targeted amplicon resequencing. This analysis was layered onto the COO classification to define 4 distinct subgroups of DLBCL. The NCI more recently conducted an analysis to characterize the genomic underpinnings of these 4 subgroups more fully.8

In this analysis, the most common mutation was TP53, which was associated with high rates of aneuploidy. A fifth cluster was also defined in this analysis, which was termed A53 (aneuploid with TP53 inactivation). Another seed class was also created, the ST2, which includes SGK1 and TET2 mutations. The DHITSig cluster was further stratified, defined by EZH2 and BCL-2 aberrations into MYC+ and MYC– categories. Overall, the NCI was able to define 7 genomic subgroups, which can be used to provide a probabilistic classification of a tumor from an individual patient with DLBCL into a genetic subgroup.

The other novel genomic categories defined by the NCI are BN2, EZB, MCD, and NI. Key genomic characteristics in the BNq arm include alterations in NOTCH2 signaling, B-cell differentiation, regulators of NF-kB pathway, immune evasion or loss of CD70 loss, CCND3 mutations, and BCL-6 SVs. EZB includes characteristics like alterations in epigenetic regulators, mutational activation of EZH2, alterations affecting B-cell signaling, inactivation of S1PR2/GN13 pathway, PTEN deletions and mutations, BCL-2 SVs, RELamplification, perturbed interactions with T follicular helper cells, MYC+ and MYC–.MCD includes mutations in CD79B, MYD88L265P, CDKN2A deletions, BCL2 copy gain/amplification, and immune evasion. NI includes alterations in NOTCH1, mutations in B-0cell differentiation regulators, and IKBKB, altered B-cell differentiation.

Another study applied a targeted 293-gene panel to a large population-based cohort of 928 patients with DLBCL, in which 5 distinct molecular subgroups were also defined, termed MYD88, BCL2, TET2/SGK1, SOCS1/SGK1, and NOTCH2. About one-fourth of patients in this analysis failed to fall into 2 of these subgroup categories and were not classified. The subgroups of this study were generally similar to those described by both the NCI and Harvard groups.

The classifications found among these studies and others have important distinctions and nuances, providing a deep and broad understanding of the biology of DLBCL. However, the results of these studies appear broadly similar, and several subgroups identified across the studies have unmistakable resemblances to others. The overlapping conclusions of these studies support the presence of previously unrecognized distinct genomic subgroups with coordinate biology.

A more detailed genetic subclassification is expected to supersede the COO framework, or at least provide a finer way to assess prognosis in DLBCL. The promise of these studies is in their potential to successfully individualize therapy.

Novel classifications provide testable predictions for specific therapeutic vulnerabilities for select groups that can be tested in preclinical models. The goal of this research is to select treatment that selectively targets the vulnerabilities implied by a patient’s genomic classification and improve outcomes overall.

Reference

1. Crombie JL, Armand P. Diffuse large b-cell lymphoma’s new genomics: the bridge and the chasm. Published online August 19, 2020. J Clin Oncol. doi: 10.1200/JCO.20.01501

2. Davies A, Cummin TE, Barrans S, et al: Gene-expression profiling of bortezomib added to standard chemoimmunotherapy for diffuse large B-cell lymphoma (REMoDL-B): An open-label, randomised, phase 3 trial. Lancet Oncol 20:649-662, 2019. doi: 10.1016/S1470-2045(18)30935-5

3. Younes A, Sehn LH, Johnson P, et al: Randomized phase III trial of ibrutinib and rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone in non-germinal center B-cell diffuse large B-cell lymphoma. J Clin Oncol 37:1285-1295, 2019. doi: 10.1200/JCO.18.02403.

4. Vitolo U, Witzig TE, Gascoyne RD, et al: ROBUST: First report of phase III randomized study of lenalidomide/R-CHOP (R2-CHOP) vs placebo/R-CHOP in previously untreated ABC-type diffuse large B-cell lymphoma. Hematol Oncol 37:36-37, 2019. doi: 10.1002/hon.5_2629

5. Zelenetz AD, Salles G, Mason KD, et al: Venetoclax plus R- or G-CHOP in non-Hodgkin lymphoma: Results from the CAVALLI phase 1b trial. Blood 133: 1964-1976, 2019. doi: 10.1182/blood-2018-11-880526.

6. Morschhauser F, Feugier P, Flinn IW, et al: Venetoclax plus rituximab, cyclophosphamide, doxorubicin, vincristine and prednisolone (R-CHOP) improves outcomes in BCL2-positive first-line diffuse large B-cell lymphoma (DLBCL): First safety, efficacy and biomarker analyses from the phase II CAVALLI study. Blood 132, 2018 (suppl; abstr 782).

7. Chapuy B, Stewart C, Dunford AJ, et al: Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat Med 24:679-690, 2018. doi: 10.1038/s41591-018-0016-8.

8. Schmitz R, Wright GW, Huang DW, et al: Genetics and pathogenesis of diffuse large B-cell lymphoma. N Engl J Med 378:1396-1407, 2018. doi: 10.1056/NEJMoa1801445