Should we Use Cell of Origin and Dual-Protein Expression in Treating DLBCL?

Treatment outcomes in diffuse large B-cell lymphoma (DLBCL) following standard R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) therapy is highly variable and dependent on a number of clinical, biologic, and genetic features. The identification of molecular heterogeneity via gene expression profiling dichotomizes patients based on the cell of origin (COO) model into germinal center B-cell like (GCB) and activated B-cell like (ABC) subsets, with ABC-DLBCL having a worse outcome. Along with the COO classification, other molecular phenotypes have also been identified, further highlighting the clinical and biologic complexity of this disease. Double-hit lymphomas, with concurrent chromosomal translocations of the MYC and BLC2 genes, or less commonly MYC and BCL6 genes, are associated with an aggressive clinical course and adverse outcomes when treated with R-CHOP. Furthermore, dual overexpression of MYC and BCL2 proteins has emerged as an important adverse prognostic factor, can be present through different mechanisms in both GCB and ABC subsets, and further complicates treatment considerations. Studies investigating the biologic underpinnings of these diverse subtypes have revealed a number of novel targets, which may provide therapeutic benefit. Moving forward, clinical trials focusing on molecular subsets of DLBCL, and incorporating rational targeted agents, will ideally lead to improved outcomes and allow a more personalized treatment approach. This review will focus on emerging data regarding DLBCL management based on either COO or dual overexpression of MYC/BCL2 proteins.

Diffuse large B-cell lymphoma (DLBCL) is the most common type of non-Hodgkin lymphoma in the Western hemisphere, with approximately 30,000 new diagnoses each year in the United States. Increasingly recognized as a very heterogeneous disease, DLBCL can be characterized by different pathologic subtypes, morphologic variants, and gene expression profiles.(1, 2) Standard of care treatment for patients with DLBCL is R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) which results in cure rates of approximately 70%.(3, 4) Unfortunately, for approximately 40% of patients, R-CHOP based therapy is ineffective and does not lead to cure. The curability of DLBCL is heavily influenced by the presence of risk factors including: elderly age, high international prognostic index (IPI) score, non-germinal center phenotype, and double-hit and dual- protein expression variants. Given the disappointing results with R-CHOP based therapy, recent efforts have focused on critically examining these higher risk subtypes and exploring alternative therapeutic strategies. Herein, we review the molecular classification of DLBCL, the impact of dual MYC and BCL2 protein overexpression, as well as current and future treatment approaches.

The IPI model was originally developed to risk stratify patients into good-, intermediate-, and poor-risk categories. Although widely used, it has not proven useful in guiding therapeutic decision-making at an individual level. Consequently, recent efforts have focused on defining molecular subtypes of DLBCL. Utilizing gene expression profiling (GEP), distinct subtypes of DLBCL were originally identified over 15 years ago.(5) Through this model, DLBCL can be categorized based on similarities to its phase of B-cell development from which the malignancy derives. DLBCL subtypes harboring a gene expression profile similar to normal germinal center B-cells are defined as the germinal center B-cell (GCB) subset. Conversely, lymphoma deriving from a more activated B-cell, or plasmablast, have been defined as the activated B-cell (ABC), or non-germinal center (non-GCB) phenotype.(6)The gold standard method initially used to assign COO was GEP using RNA microarray analysis. This method was fraught with limitations though, including long turnaround times, the need for fresh-frozen tissue, and the associated high cost. Accordingly, GEP has seen limited practical application.(7) Subsequently, efforts focused on simplifying the COO determination through the use of immunohistochemistry on commonly available formalin-fixed paraffin-embedded (FFPE) tissue. The algorithm by Hans et al. assigns COO based on CD10, BCL6, and MUM1 expression, and is the most widely used algorithm in the United States. Despite its wide use, it has its own limitations including a lack of standardization, issues with antibody performance, and also inter-observer scoring variability. Furthermore, when compared to GEP based techniques, the Hans algorithm was shown to have a concordance rate of approximately 80%.(8, 9)Recently, platforms have been developed which allow GEP to be performed on FFPE tissue. Specifically, the Lymph2CX assay utilizes GEP of 20 candidate genes in order to differentiate GCB from ABC subtypes of DLBCL. This technology is beneficial given its rapid turnaround time, accurate results compared to GEP, high concordance rates between independent laboratories, along with its use of FFPE tissue.(10, 11) Although still largely employed in the research setting, the Lymph2CX assay is commercially available and is a promising alternative to more accurately and rapidly determine COO status in DLBCL patients.

Interest in COO identification increased considerably after initial studies revealed differences in outcomes based on COO.(12) The study by Lenz and colleagues demonstrated that patients having a gene expression profile similar to germinal center B-cells had a more favorable 3-year progression-free survival (PFS) compared to the ABC subtype (74% vs 40%).(13) Following this landmark study, numerous other groups have reported similar findings showing inferior outcomes in the non-GCB/ABC subset.(5, 9, 10, 14, 15) Thought the body of evidence suggests inferior outcomes for the non-GCB/ABC subset, it should be noted that some prospective clinical studies, namely the RICOVER60 and the RCHOP14 vs RCHOP21 trial, have not uniformly demonstrated this trend.(16, 17)Similarly, COO appears to have a prognostic impact even in the relapsed/refractory setting. In the bio- CORAL study, Thieblemont and colleagues reported the outcomes of patients with relapsed/refractory DLBCL treated with either R-DHAP (rituximab, dexamethasone, high-dose cytarabine, and cisplatin) or R- ICE (rituximab, ifosfamide, carboplatin, and etoposide) salvage therapy followed by autologous stem cell transplant (ASCT). When patients samples were analyzed for COO based on the Hans algorithm, the GCB subtype was associated with a significantly superior PFS in the R-DHAP treated group.(18) Again, this data helps to underscore the poor prognostic impact of the non-GCB/ABC subset of DLBCL, in both the upfront and relapsed/refractory setting.In addition to the varying prognosis associated with each distinct COO phenotype, different molecular drivers appear to promote the development of each subtype. ABC DLBCL has been associated with BCL2 amplification, MYD88 mutations, increased nuclear factor kappa B (NF-kB) signaling, and CARD11 mutations, among others. Conversely, the GCB subtype is more likely to harbor BCL2 translocations, EZH2 mutations, PTEN deletions, and mTOR pathway activation.(19-22)Given the poor prognosis of the ABC subtype, initial efforts focused on investigating and designing clinical trials in this subgroup. Table 1 summarizes the current clinical studies focusing on novel treatments based on COO.

NF-kB signaling is thought to be integral in the pathogenesis of the ABC subtype of DLBCL, and consequently, initial efforts aimed at modulating this pathway in an effort to improve outcomes. In a study by Ruan et al., the proteasome inhibitor, bortezomib, was used in combination with R-CHOP in untreated patients with aggressive NHL, including 40 patients with DLBCL. In this study, the overall response rate (ORR) was 88%, with a 2-year PFS of 64% (95% CI, 47% to 77%) and 2-year overall survival (OS) of 70% (95% CI, 53% to 82%). Importantly, when the authors evaluated outcomes based on COO, the GCB and ABC phenotypes had similar outcomes in terms of progression-free and OS, suggesting that the addition of bortezomib improved the poor prognosis associated with the ABC phenotype.(23) Based on these encouraging results, the phase II pyramid study was conducted to further evaluate the safety and efficacy of this combination in non-GCB DLBCL. In this study, patients were randomized to R-CHOP or R-CHOP + bortezomib with the primary endpoint of PFS. With a median follow-up of 31.5 months, there was unfortunately no improvement in PFS or OS with the addition of bortezomib to the R-CHOP backbone.(3)Similarly, the REMoDL-B study was a randomized phase III trial evaluating the addition of bortezomib to R-CHOP therapy. In this study, FFPE tissue was utilized for GEP to allocate patients into ABC and GCB subgroups. All patients were initially treated with one cycle of R-CHOP therapy while concurrently undergoing GEP. Prior to the second cycle of therapy, patients were then randomized to either R-CHOP, or R-CHOP plus bortezomib. Again, despite encouraging preliminary data, this study failed to show that the addition of bortezomib significantly improved PFS at 30 months for either the ABC (HR=0.79, p=0.309) or GCB (HR=0.87, p=0.458) subsets.(24)

Lenalidomide is an immunomodulatory agent that has shown antitumor activity in various subtypes of NHL.(25) Preclinical data reveals that lenalidomide can block NF-kB signaling and also downregulates some of the B-cell receptor activation pathways thought to be important in the pathogenesis of non- GCB DLBCL.(26) In relapsed/refractory DLBCL, lenalidomide has been associated with response rates ranging from 19% to 28%.(27, 28) When evaluating the activity based on COO, lenalidomide was found to have an ORR of 53% in the non-GCB subtype versus 9% in the GCB subtype of DLBCL.(29) This work provided the foundation for further phase II studies combining lenalidomide with R-CHOP therapy in the upfront treatment of DLBCL.In the study by Nowakowski and colleagues, lenalidomide was combined with R-CHOP therapy in 64 patients with newly diagnosed DLBCL, yielding an ORR of 98% with 80% achieving a CR. Furthermore, there was no difference in the 2-year progression-free (60% vs 59%; p=0.83) or OS (83% vs 75%; p=0.61) between non-GCB and GCB subtypes.(30) This study suggests that the negative prognostic value of the non-GCB phenotype may potentially be mitigated by the addition of lenalidomide to standard R-CHOP therapy. Currently, two prospective trials (NCT01856192, NCT02285062) are evaluating R-CHOP vs R- CHOP plus lenalidomide in non-GCB DLBCL. Ideally these randomized studies will provide further clarity on the role of lenalidomide in the non-GCB subset.

B-cell receptor (BCR) signaling has been recognized as a key growth pathway in B-cell malignancies. Within this pathway, the Bruton’s tyrosine kinase (BTK) enzyme has been shown to interact with NF-kB, ultimately leading to proliferation and survival.(31) In the ABC subtype of DLBCL, mutations targeting the BCR pathway result in chronic active BCR signaling leading to recent efforts to pharmacologically inhibit many key enzymes in this pathway.(32) In ABC subtype DLBCL, ibrutinib, an irreversible BTK inhibitor, demonstrated an ORR of 40% compared to 5% in patients with the GCB subtype.(33) Furthermore, genomic alterations within ABC DLBCL, including mutations in CD79B and MYD88, have been found to be associated with sensitivity to BTK inhibition.(34) The combination of ibrutinib and R- CHOP was evaluated in a phase 1b study in a range of subtypes of NHL, including DLBCL. In 11 patients with DLCBL in which COO data was available, the combination resulted in a CR rate of 71% in the GCB subtype, with 100% of patients with the non-GCB subtype achieving a CR.(35) These results have promoted an ongoing phase III randomized, double blind, placebo-controlled trial, which compares R- CHOP to R-CHOP plus ibrutinib in non-GCB DLBCL (NCT01855750).In an effort to improve on the efficacy of the R-CHOP backbone, investigators at the National Cancer Institute developed the DA-EPOCH-R (dose adjusted etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin, and rituximab) regimen. This regimen utilizes a 96-hour infusion of etoposide, vincristine and doxorubicin; with rituximab, prednisone, and cyclophosphamide administered on a bolus schedule. The rationale behind this regimen lies in preclinical data showing continuous low- dose drug exposure enhances tumor kill. In a multicenter phase II study, the authors demonstrated a 5- year time to progression of 81% which was a marketed improvement when compared to the 5-year PFS of approximately 60% seen in previous R-CHOP studies.(36-38) These results prompted a randomized phase III study of R-CHOP vs DA-EPOCH-R in untreated DLBCL. With a median follow-up of 5 years, the more intensive DA-EPOCH-R regimen failed to improve event-free survival (EFS) (p=0.438) or OS (p=0.42) when compared to R-CHOP.(39) However, no current subset analysis has been presented. Therefore, conclusions regarding the efficacy of DA-EPOCH-R on ABC vs GCB subtypes is unknown.

Anti-CD20 monoclonal antibody therapy has led to a significant improvement in outcomes in DLBCL.(40) Rituximab, a chimeric anti-CD20 monoclonal antibody relies primarily on antibody-dependent cellular cytotoxicity (ADCC) and complement-mediated cytotoxicity. Efforts to improve on its immunogenicity through glycoengineering have resulted in newer generation anti-CD20 antibodies, such as obinutuzumab, which is thought to demonstrate more potent ADCC.(41) In the GOYA study, untreated DLBCL patients were randomized to CHOP in combination with either rituximab or obinutuzumab. With a median follow-up of 29 months, there was no significant difference in 3-year PFS (66.9% vs 69.9%; p=0.386) or OS (81.4% vs 81.2%; p=0.998) between the two arms. When analyzed based on COO, there was no statistically significant improvement in 3-year PFS for the obinutuzumab plus CHOP combination over R-CHOP for either the GCB or ABC subsets.(42)Alterations of the MYC proto-oncogene, classically associated with Burkitt lymphoma, occur in up to 15% of patients with DLBCL. In contrast to Burkitt lymphoma where MYC rearrangements are mainly diagnostic, the presence of MYC alterations in DLBCL are adversely prognostic. This is likely due to a distinct set of target genes for the MYC protein, a transcription factor, between the two diseases. Early studies showed that presence of the MYC rearrangement led to long-term disease control in less than one-third of patients, significantly inferior to expected outcomes for this disease. However, while the poor outcomes of DLBCL harboring MYC was initially attributed to increased proliferation due to MYC’s target genes, it is now clear that it is the dual aberration of MYC and BCL2 (or BCL6) that confers the poor prognosis.

In DLBCL, there are two common scenarios leading to increased MYC and BCL2 overexpression: 1) The co-occurrence of MYC oncogene rearrangements in concert with BCL2 and/or BLC6 rearrangements; and 2) The overexpression of the MYC and BCL2 proteins by post-translational processes not involving a chromosomal translocation. On the proteomic level, both of these types of alterations result in overexpression of the respective proteins and a more aggressive clinical course.(43, 44) In the updated WHO classification of lymphoid malignancies, patients harboring a MYC along with a BCL2 and/or BCL6 rearrangement are a new diagnostic entity termed high-grade B-cell lymphoma, and are commonly referred to as “double-hit” or “triple-hit” lymphoma, respectively. Similarly, co-expression of the MYC and BCL2 proteins without underlying genomic rearrangements should now be considered a new adverse prognostic indicator (though not a separate diagnostic category) in DLBCL, termed dual-protein expression lymphoma (DEL). In defining DEL, the WHO suggests an IHC cutoff point of ≥40% for MYC and >50% for BCL2.(45)Compared to double-hit lymphoma (DHL), DEL is encountered more frequently in clinical practice. Johnson et al. demonstrated that 18% of patients analyzed displayed dual overexpression of BCL2 and MYC. Similarly, in an analysis by Hu et al., 34% of patients demonstrated these aberrations.(46, 47) Conversely, DHL is estimated to make up approximately 5-12% of DLBCL cases. (48, 49) The presence of the dual-protein expression phenotype has been shown to have a strong impact on clinical outcomes in DLBCL. In a paper by Green et al., the outcomes of R-CHOP treated patients with DEL were compared to patients with either MYC or BCL2 overexpression, and to patients with DLBCL without these aberrations. In the analysis, patients with either MYC or BCL2 overexpression, (but not both), demonstrated similar outcomes to those without such abnormalities. Conversely, patients with DEL demonstrated an inferior 3-year PFS (39% vs 75%; p<0.001) and 3-year OS (43% vs 86%; p<0.001) compared to those with either single or no protein overexpression.(50) These data underscore that it is the co-expression of MYC and BCL2 proteins that is key to an adverse outcome. In the relapsed or refractory setting, both the DHL and DEL phenotype appear to confer a poor prognosis. In a retrospective study by Herrera and colleagues, 117 patients with relapsed/refractory DLBCL underwent ASCT. Within this study, DEL patients demonstrated a 4-year PFS of 48% compared to 59% in non-DEL patients (p=0.049). A similar, though more dramatic trend was demonstrated in the DHL patients, whereby the 4-year PFS was 28% in DHL patients compared to 57% in non-DHL patients (p=0.013).(51) Based on such data, it appears that the prognostic impact of both DEL and DHL holds true in both the upfront and relapse/refractory setting.Despite efforts to precisely subdivide DLBCL into distinct risk groups, there is considerable overlap between COO, dual-protein expression, and double-hit subsets. While approximately 80-90% of double- hit/triple-hit lymphoma cases also demonstrate increased expression of MYC and BCL2, it is important to recognize that these entities are not equivalent.(47, 50) The dual-protein expression phenotype appears to highly correlate with the ABC/non-GCB subset of DLBCL, with prior reports revealing that 63% of non- GCB DLBCL demonstrated dual-protein expression compared to 37% of GCB DLBCL. Similarly, DHL is known to be more common in GCB DLBCL.(50) When examining cases of DLBCL without concurrent MYC/BCL2 rearrangements, patients harboring the DEL phenotype demonstrated both an inferior PFS and OS.(47) DEL also has a distinct clinical phenotype in which patients tend to be older, with a median age of 71 years compared to a median age of 62 years in patients without double-hit or dual-protein expression DLBCL. Additionally, DEL patients are more likely to have advanced stage disease, a poor performance status, a higher Ki-67 proliferative index, intermediate/high- to high-risk IPI scores, multiple extranodal sites of disease, and a lower complete response rate to R-CHOP chemotherapy.(50)When assessing the impact of dual-protein expression on IPI score, a worse PFS and OS was observed in DEL patients with both low- and high-risk IPI scores. In terms of COO, patients exhibiting co-expression of MYC and BCL2 demonstrated a significantly worse PFS and OS within both COO subtypes. Furthermore, when stratifying the data based on COO and excluding those patients with DEL, Hu et al noted similar outcomes for the ABC and GCB subtypes of DLBCL (PFS: p=0.4291; OS: p=0.3163). Although controversial, based on these findings, the authors argue that the inferior outcomes associated with ABC DLBCL are a consequence of the increased frequency of dual MYC and BCL2 overexpression in this subset and not due to the cell-of-origin.(47) In contrast, a recent post-hoc analysis of patients enrolled in two German trials did not find dual-protein expression to supercede the prognostic significance of ABC DLBCL, suggesting a more complex interaction between these biologic features. Of note, the method of COO assessment was different between the two studies as one used IHC and the other used Lymph2CX.(52) Recent efforts have attempted to integrate the various factors including COO, dual-protein expression/double-hit, and IPI score. In the study by Scott et al., 344 pre-treatment FFPE tissue biopsies from patients with DLBCL were analyzed using the Lymph2CX assay for COO, in addition to undergoing FISH and IHC analysis for double-hit and dual-protein expression status. In the analysis, the authors found that COO, dual-protein expression/double-hit, and IPI score all retained their prognostic impact in DLBCL. As demonstrated in the above study, dual-protein expression was associated with worse outcomes compared to n`on-MYC/BCL2 patients. Furthermore, COO was found to be most impactful on the prognosis of those with intermediate-risk IPI (ABC vs GCB : 5-year time to progression, 53% vs 74%; p=0.003).(10) In contrast to the results attained by Hu et al., Scott and colleagues argue that COO and dual-protein expression status provide independent prognostic value.(10, 47) Specifically, in the non- MYC/BCL2 group, COO impacted outcomes with the GCB subset demonstrating a superior freedom from progression compared to the ABC subset.(10)With a more thorough understanding of the poor prognosis of patients with DHL and DEL, retrospective studies have attempted to evaluate the impact of various treatment regimens on prognosis in these patients. Given the rarity of this entity in the literature, and the lack of adequate reporting, there is a scarcity of treatment data specifically evaluating DEL. As a consequence, treatment strategies generally mirror DHL. Two recent studies have specifically attempted to address this question, though, given the small numbers, firm conclusions cannot be made. In a retrospective study, Aggarwal and colleagues evaluated the outcomes of patients with DEL treated with either R-CHOP or DA-EPOCH-R therapy. Although small patient numbers (n=16), the authors noted a significantly higher relapse rate of 80% in the R-CHOP treated group, compared to 18% in the DA-EPOCH-R group (p=0.042).(53) Similarly, investigators at the MD Anderson Cancer Center also retrospectively evaluated the outcomes of 16 patients with DEL treated with DA-EPOCH-R, demonstrating a 1-year PFS of 65% and 1-year OS of 86%. When comparing these outcomes to patients with DHL and DLBCL without the dual-protein expression/double-hit phenotype, there were no significant difference in outcomes.(54) Much of the data influencing treatment patterns in DEL is based on retrospective studies in DHL. Specifically, the largest retrospective analysis by Petrich et al. examined 311 patients with DHL treated with numerous chemotherapy regimens including standard R-CHOP, along with various intensive induction regimens including DA-EPOCH-R, among others. In the analysis, the authors found a significantly inferior median PFS for R-CHOP therapy (7.8 vs 21.6 months; p=0.001) when compared to the composite group of intensive induction regimens, though no OS benefit was noted (p=0.564).(55) Likewise, investigators from the MD Anderson Cancer Center retrospectively evaluated their outcomes in 129 patients treated for DHL. In the analysis, outcomes were stratified by treatment regimen with those treated with DA-EPOCH-R showing superior EFS (p=0.004) and OS (p=0.057) when compared to patients treated with R-CHOP.(56) Based on these studies, more intensified induction strategies have been considered in both DHL and DEL, though this is largely based on the poor outcomes seen with R- CHOP, rather than quality data to suggest that other combinations yield superior outcomes.In DEL, the risk of central nervous system (CNS) progression or recurrence appears higher than in patients without DEL.(57) Depsite this fact, given the paucity of data, no firm recommendations regarding CNS directed therapy can be made.Given the poor prognosis for many patients with DEL, numerous rationale therapeutic strategies are being explored in an effort to improve outcomes. Table 1 summarizes the current clinical studies evaluating novel agents in DHL and DEL.One of the strategies to modulate the MYC oncoprotein has focused on targeting its transcription with the use of bromodomain and extraterminal (BET) inhibitors. BET protein have been shown to act as regulatory factors for MYC, and in preclinical studies BET inhibition results in downregulation of both MYC and BCL2 transcription.(58, 59) Currently there are a number of active, early phase, clinical studies evaluating BET inhibitors (NCT01949883, NCT01943851, NCT02431260, NCT02711137). The BCL2 protein plays a crucial role in regulating apoptosis and promoting cell survival, and in concert with the MYC protein, contributes to tumorigenesis in double-hit and DEL.(49, 60) Venetoclax is a first- in-class, selective oral BCL2 inhibitor which has shown promising activity in a range of NHL subtypes.(61) Venetoclax is currently being combined with DA-EPOCH-R therapy in a phase I/II intergroup trial in patients with DHL or DEL (NCT03036904).As discussed above, lenalidomide has been shown to have preferential activity in non-GCB DLBCL. Furthermore, given that this COO subset is enriched in the dual-protein expression phenotype, a phase I/II study is currently underway to explore the combination of lenalidomide plus DA-EPOCH-R in MYC- associated B-cell lymphomas (NCT02213913). Conclusions COO and dual-protein expression status are two biologic features used to further classify DLBCL, each conferring an impact on biology, clinical prognosis, and response to standard R-CHOP therapy. At initial presentation, it is paramount to classify new DLBCL cases based on COO, and furthermore, evaluate for the presence of dual-protein expression and the double-hit phenotype. To this end, adequate tissue sampling is crucial to facilitate such testing, which ultimately allows clinicians to gain a more thorough understanding of the biology of DLBCL on an individual level. New testing modalities such as the Lymph2CX assay hold the promise of more accurately and rapidly determining the COO in patients, and will likely be increasingly used in the clinical setting.Despite progress in elucidating novel subsets of DLBCL, factors such as COO and dual-protein expression have yet to lead to changes in our upfront treatment. The reasons for this are varied and speak to the inherent bias and conflicting results in the available retrospective data, and the lack of prospective studies showing a consistent benefit with alternative treatment strategies. Alhough upfront therapy has not been impacted by such advances, in some treatment settings, it does appear though that disease biology may be useful in clinical decision making. To date, R-CHOP therapy is still considered the standard front-line therapy for ABC and GCB DLBCL, though more mature clinical trial follow-up may lead to changes in our current practice. Similarly, for patients with a double-hit phenotype, augmented regimens such as DA-EPOCH-R are commonly utilized based largely on retrospective data. It is important to recognize that there is a paucity of data, prospective or retrospective, guiding treatment selection in lymphomas with dual expression without DHL biology; consequently, intensified treatment is a less widely accepted practice in this subset. Hopefully, results from prospective studies will help further clarify the role of intensified regimens. Novel agents targeting key molecular pathways in DLBCL hold the promise of altering the management of this disease, but at the present time, should not be considered outside of a clinical trial. Future research efforts should focus on rationally designed clinical studies aimed at these specific subsets of BI-3812 DLBCL.