Double-hit Signature with TP53 Abnormalities Predicts Poor Survival in Patients with Germinal Center Type Diffuse Large B-cell Lymphoma Treated with R-CHOP

Diffuse large B-cell lymphoma (DLBCL) is the most common type of non-Hodgkin lymphoma, but is a very heterogeneous disease characterized by recurrent chromosomal abnormalities and somatic mutations. The addition of rituximab to CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) chemotherapy (R-CHOP) has led to marked improvement in the survival of patients with DLBCL (1). The prognosis of patients with DLBCL has also been associated with biological characteristics of the tumor, such as the cell-of-origin (COO) and genetic aberrations. The two major subtypes of DLBCL based on COO are the germinal center B-cell–like (GCB) and activated B-cell–like (ABC) subtypes, as determined by gene expression profiling (GEP; refs. 2, 3). The presence of MYC and BCL2 and/or BCL6 translocations, so-called double/triple-hit (DH) lymphoma, has also been shown to predict for an aggressive clinical course in DLBCL (4–6). This observation led the 2017 World Health Organization classification to separate the DH cases from DLBCL, not otherwise specified (NOS; ref. 7). Recently, it was reported that GEP could also identify cases with the biological and clinical characteristics of DH lymphoma, including some without the requisite translocations (DHITsig-positive cases; ref. 8). In addition, cases with high MYC and BCL2 protein expression as assessed by IHC, so-called dual protein expressors, were shown to have an inferior survival (4, 9). Next-generation sequencing has also demonstrated the molecular heterogeneity of this disease, and has provided insights into the pathogenesis and biology of the GCB and ABC subtypes (10–14). Despite the generally good prognosis of GCB DLBCL, a useful prognostic model that could be easily applied in the clinical setting is lacking.

The purpose of this study was to develop a molecular subtyping schema for GCB DLBCL using genomic studies, such as FISH cytogenetic analysis, GEP, and mutation analysis to risk-stratify patients with GCB DLBCL.

We identified 253 patients diagnosed with de novo DLBCL in the years 2000–2016 at the City of Hope National Medical Center (Duarte, CA) and the University of Manitoba (CancerCare Manitoba, Winnipeg, Manitoba, Canada) and treated with R-CHOP. There were 87 GCB DLBCL cases with sufficient material and adequate clinical follow-up. Patients with primary mediastinal B-cell lymphoma, T-cell/histiocyte-rich large B-cell lymphoma, prior low-grade B-cell lymphoma, or an immunocompromised state (including human immunodeficiency virus infection) were excluded. Patients were included in the final cohort after the following criteria were met: complete clinical and laboratory data (Supplementary Table S1), GCB DLBCL by COO GEP, and adequate diagnostic biopsy material available for review and multiparameter analysis. This study was approved by the Institutional Review Boards at the City of Hope National Medical Center (Duarte, CA) and University of Manitoba (Winnipeg, Manitoba, Canada) in accordance with the Declaration of Helsinki with a waiver of informed consent obtained for retrospective samples.

Cases with formalin-fixed, paraffin-embedded tissue (FFPET) were stained with hematoxylin and eosin and re-reviewed to confirm the diagnosis of DLBCL, NOS. IHC was performed on 3- to 4-μm-thick sections of FFPET microarrays using antibodies to CD20, CD3, CD10, BCL6, MUM1, MYC, and BCL2. The slides were stained on the Ventana Discovery XT Platform (Ventana) for MYC, and on a Leica Bond III Instrument (Leica Biosystems) for all other stains. Cases were reviewed and scored independently by three expert hematopathologists (J.Y. Song, A.M. Perry, and M.R. Nasr). FISH cytogenetic analysis for MYC, BCL2, and BCL6 gene rearrangements was performed using the LSI Dual Color Break-apart Probes (Abbott Molecular). At least 100 nuclei were scored and rearrangement was defined as the presence of break-apart signals in ≥10% of the nuclei. Double/triple-hit lymphoma (DH) had concurrent rearrangements of MYC and BCL2 and/or BCL6 genes.

Tissue blocks in which ≥60% of the surface area consisted of tumor were selected for RNA and DNA extraction using the Qiagen Allprep RNA/DNA FFPET Kit (Qiagen), following the manufacturer's recommended protocol. We used a custom targeted panel of 334 genes (designed by Q. Gong and W.C. Chan), which includes the most frequently mutated genes in B-cell lymphoma, and performed DNA sequencing on an Illumina HiSeq 2500 (Illumina; see Supplementary Materials and Methods; Supplementary Table S2).

The Oncoscan Copy-number Variation Assay (Thermo Fisher Scientific) was performed according to the manufacturer's directions using 80 ng of DNA. The data files were analyzed with the Chromosome Analysis Suite (Thermo Fisher Scientific). Briefly, the generation of these files entails calculating the log₂ ratio, allelic difference, and B-allele frequency (BAF), and then identifying normal diploid regions. On the basis of the normal diploid regions, the log₂ ratio, allelic difference, and BAF are recomputed, if necessary. Segmentation and visualization of the copy-number abnormalities were performed with Nexus Copy Number 10.0 Software (Bio Discovery) using the SNP-FASST2 algorithm. The percentage aberrant genome per case was calculated by taking the total size of the aberrant regions divided by the total size of the genome for chromosomes 1–22.

We used 200 ng of RNA on the nCounter Platform (NanoString Technologies) to determine the COO using the Lymphoma/Leukemia Molecular Profiling Project (LLMPP) code set Lymph2Cx (3). Briefly, the RNA was hybridized to custom code sets overnight at 65°C and processed on the nCounter prep station, and gene expression data were acquired on the nCounter digital analyzer. We then uploaded the data to the LLMPP website (https://llmpp.nih.gov), which generated the COO using the nSolver (NanoString Technologies). In addition, we also ran the DLBCL90 double-hit gene expression signature (DHITsig) (8) on the nSolver. The DHITsig score was defined as DHITsig positive (greater than −6.3), DHITsig negative (less than −15.6), or DHITsig indeterminant (−6.3 to −15.6).

The grouping of GCB DLBCL was performed as follows in a step-wise manner (from GCB1 to GCB4):

• GCB1: cases that had TP53 abnormalities (homozygous TP53 loss or TP53 copy-number loss and/or TP53 mutation) and were DHITsig positive were categorized into GCB1. TP53 abnormalities are frequently seen in DHITsig-positive cases (8) and, because they are a strong independent prognostic factor in DLBCL (15, 16), we investigated TP53 abnormalities in conjunction with DHITsig GEP status.

• GCB2: the remaining cases that were DHITsig positive, but lacked TP53 abnormalities were categorized as GCB2.

• GCB3: cases that had an EZH2 mutation and/or BCL2 translocation, similar to the EZB group (17), but were DHITsig negative, were categorized in GCB3.

• GCB4: the remaining cases that were not DHITsig positive and lacked EZB (lacked EZH2 mutation and/or BCL2 translocation; Fig. 1).

We used data from the study by Wright and colleagues (14) for 188 patients diagnosed with GCB DLBCL who were treated with R-CHOP. We recategorized these cases with our new grouping method as follows: cases that had TP53 abnormalities and were DHITsig positive as GCB1, cases that were negative for TP53 abnormalities, but DHITsig positive as GCB2, cases with EZH2 mutation and/or BCL2 translocation as GCB3, and the remaining cases as GCB4. We also used the data from 221 patients diagnosed with GCB DLBCL and treated with R-CHOP from the study of Sha and colleagues (18) to determine whether TP53-mutated cases with double-hit MYC/BCL2 by FISH analysis also had a poor prognosis.

Overall survival (OS) and progression-free survival (PFS) were calculated as the time from initial diagnosis to death due to any cause (OS), or to the earliest of refractory disease, relapse, or death due to any cause (PFS). Patients alive at the last contact (OS) or patients alive without refractory/relapsed disease (PFS) were censored at the last contact. OS and PFS were estimated using the Kaplan–Meier product-limit method (19). HRs and P values were determined using the Cox model. All P values were two-sided and analyses were performed using SAS 9.4. Survival curves were generated using MedCalc Statistical Software version 19.1.5 (MedCalc Software bv).

A total of 87 patients with GCB DLBCL (median age, 60 years and male-to-female ratio, 1.7:1) had complete clinical, IHC, FISH, GEP, and sequencing data (Supplementary Tables S1 and S3). There was no difference in OS between the two centers (P = 0.50). There were a total of 11 DH/TH lymphomas identified by FISH analysis: four MYC/BCL2 double-hit lymphomas, five MYC/BCL2/BCL6 triple-hit lymphomas, and two MYC/BCL6 double-hit lymphomas. Fifteen cases had the double-hit signature (DHITsig positive), eight were indeterminant, and 64 cases were DHITsig negative using the DLBCL90 gene expression signature of Ennishi and colleagues (ref. 8; Fig. 1; Supplementary Fig. S1). The eight indeterminant cases had survival outcomes similar to those who were DHITsig negative and were analyzed together as one group for OS and PFS.

We also analyzed the cases for TP53 abnormalities (TP53 copy-number loss and/or TP53 mutations) independent of DHITsig, as well as EZH2 and/or BCL2 translocations (EZB), and found there was no significant difference in OS (P = 0.2 and P = 0.09, respectively; Supplementary Fig. S2). We, therefore, investigated whether combining these factors could improve prognostication.

DLBCL with TP53 mutations and/or deletions has a poor prognosis in patients treated with R-CHOP (15, 16). We found seven cases (8% of all cases) of GCB DLBCL that were DHITsig positive with TP53 abnormalities. By FISH analysis, two cases had a triple-hit, one was double-hit with MYC/BCL2, two cases had an MYC translocation only, one case had a BCL2 translocation only, and the other case had no such translocation. Cases in GCB1 had the worst OS (HR, 9.2; P = 0.0018) and shortest PFS (HR, 6.1; P = 0.002) compared with other groups (Fig. 2A and B). Patients with TP53 abnormalities that were DHITsig negative did not have the same poor survival (Fig. 2C and D).

The other eight cases (9%) that were DHITsig positive, but lacked TP53 abnormalities showed a predilection (88%) for having an EZH2 mutation and/or BCL2 translocation (EZB of Schmitz and colleagues; ref. 17). These cases also had a high frequency of MYC mutations (63%), but lacked mutations in SGK1 (0%), and had a low frequency of mutations in linker histone genes (e.g., HIST1H1E; Fig. 1). By FISH analysis, three cases were double-hit lymphoma with MYC/BCL2, two cases were triple-hit lymphoma with MYC/BCL2/BCL6, one case had an MYC translocation only, one case had a BCL2 translocation only, and the other one had no such translocation. Typically, DHITsig-positive cases have a poor OS when compared with DHITsig-negative cases (8), but this group had a good survival in our study after removing the cases with TP53 abnormalities (Fig. 2A and B).

There were 28 cases (32%) that were DHITsig negative and had an EZH2 mutation and/or BCL2 translocation. These were categorized as EZB like with some overlapping features with the DLBCL in cluster 3 of Chapuy and colleagues (20). These cases showed frequent mutations of BCL2 (50%), KMT2D (57%), TNFRSF14 (36%), SGK1 (36%), and histone modifying genes (39%), but mutations of MYC were infrequent (4%; Fig. 1). The survival of this group was intermediate compared with the other groups (Fig. 2A and B).

The largest group of cases (51%) was DHITsig negative and lacked EZH2 mutations and/or BCL2 translocations. These cases had frequent mutations in SGK1 (16%) and histone modifying genes (50%), as well as TET2 mutations (25%; Fig. 1). These cases had similarities to cluster 4 of Chapuy and colleagues (20). The survival of this group was excellent (Fig. 2A and B).

Using an independent cohort of 188 patients, with GCB DLBCL treated with R-CHOP, from the study by Wright and colleagues (14), we found that cases that were DHITsig positive with any TP53 abnormality (biallelic, copy-number loss, and/or mutation) had a poor OS and PFS (Fig. 3A and B; Supplementary Fig. S3), whereas those with a TP53 abnormality that were DHITsig negative did not show such poor survival. We then grouped the validation cohort cases using our GCB grouping strategy and observed that the GCB1 (DHITsig positive with TP53 abnormalities) had very poor OS (HR, 8.4; P < 0.0001) and PFS (HR, 3.8; P < 0.0001), just as we saw in our training cohort (Fig. 3C and D; Supplementary Fig. S4). We also observed that within particular international prognostic index (IPI)groups, GCB1 identified patients with the poorest survival (Supplementary Fig. S5). For GCB2 (HR, 2.2; P = 0.12) and GCB3 (HR, 1.4; P = 0.43), the OS was similar to GCB4, but there was a trend for a shorter PFS for GCB2 (HR, 1.7; P = 0.12) and GCB3 (HR, 1.6; P = 0.1; Fig. 3C and D). In addition, we found that cases that were DH/TH by FISH analysis (MYC/BCL2 or MYC/BCL2/BCL6) combined with TP53 abnormalities had the poorest OS (P < 0.0001) and PFS (P = 0.0076; Fig. 4A and B).

We also used a second independent cohort of 221 patients, with GCB DLBCL treated with R-CHOP, from the REMoDL-B trial (18, 21). However, for this cohort, we were only able to use the TP53 mutation status as the sole TP53 abnormality in our analysis because copy-number analysis was not available. We also did not have GEP for the DHITsig, so instead we used the FISH analysis as a surrogate for the DHITsig. Although there is not a direct correlation with the DHITsig, we found that cases that were double-hit with MYC/BCL2 by FISH and had a TP53 mutation, had a much shorter OS (HR, 8.5; P = 0.001) and PFS (HR, 9.5; P < 0.0001; Fig. 4C and D). These two studies support our finding that having a double-hit profile by GEP or FISH analysis in conjunction with a TP53 abnormality results in very poor survival in patients with GCB DLBCL.

Several recent studies have elucidated the genomic diversity of DLBCL and reported various subgroups with distinct biologic pathways (8, 13, 14, 17, 18, 20). However, this is the first study of its kind to incorporate FISH cytogenetic analysis, GEP using the DLBCL90 signature (8), targeted sequencing, and copy-number analysis to risk-stratify patients with GCB DLBCL treated with R-CHOP (Fig. 5). By combining these modalities, we found that cases with TP53 abnormalities that are DHITsig positive (GCB1) have a very poor prognosis, whereas cases that are DHITsig negative and lack EZH2 mutations and BCL2 translocations (GCB4) have an excellent prognosis when treated with R-CHOP. Interestingly, patients who are DHITsig positive, but lack TP53 abnormalities also have a good prognosis (GCB2), whereas those in the EZB-like group (GCB3) have an intermediate prognosis. We believe that these methods can be used to better risk-stratify patients with GCB DLBCL for clinical studies.

Recently, six studies looking at the biology of DLBCL have demonstrated various genetic profiles important in lymphomagenesis (8, 13, 14, 17, 18, 20). Wright and colleagues (14) and Schmitz and colleagues (17) have identified a subgroup of GCB DLBCL with EZH2 mutations and/or BCL2 translocations (EZB group) that had a less favorable outcome than patients without these lesions. Chapuy and colleagues (20) have recently identified five distinct clusters of DLBCL. They divided GCB DLBCL into two clusters, one similar to the EZB group (cluster 3) and the other enriched in mutations affecting core and linker histones and mutations in the GNA13/RhoA, JAK/STAT, and BRAF pathways (cluster 4), with a more favorable outcome than cluster 3. They also found a cluster characterized by biallelic inactivation of TP53 and/or CDKN2A loss with an unfavorable prognosis (cluster 2). The studies by Wright and colleagues (14), Schmitz and colleagues (17), and Chapuy and colleagues (20) all demonstrate differences in the biologic groupings, but there was clearly extensive overlap for some of the groups. Recent studies by Ennishi and colleagues (8) and Sha and colleagues (18) used GEP to develop a signature that is sensitive and specific for identifying high-grade B-cell lymphomas with features of DH lymphoma regardless of the status of MYC and BCL2 translocations (DHITsig positive and molecular high grade, respectively). These cases had poor survival, similar to those with ABC DLBCL, and had a significant association with the EZB group proposed by Schmitz and colleagues (17). Interestingly, cases that were DHITsig positive had a higher frequency of TP53 mutations compared with those that were DHITsig negative (8). It is well known that TP53 abnormalities in DLBCL are associated with a poor prognosis (15, 16) and, because we saw that a subset of the DHITsig-positive cases had TP53 abnormalities, it was of interest to investigate this issue. In addition, both Chapuy and colleagues (20) and Schmitz and colleagues (17) showed that the EZB-like cases have an intermediate prognosis in GCB DLBCL. We believe that our algorithm is an improvement with prognostic implications.

We found that our cases with TP53 abnormalities that also were DHITsig positive (GCB1, DHIT + TP53) had a very poor prognosis (Fig. 2). This finding was validated in an independent cohort (14) of cases showing poor survival for those who were DHITsig positive with TP53 abnormalities (Fig. 3). In both validation cohorts (14, 18), we also found that double-hit cases with MYC/BCL2 by FISH analysis with TP53 mutations had a poor survival (Fig. 4). Other studies of DLBCL have also shown that mutations and copy-number loss of TP53 are associated with a poor prognosis (15, 16). TP53 mutation in combination with copy-number loss results in p53 deficiency and the inability to eliminate damaged cells (15). Our cases that were DHITsig positive with TP53 abnormalities had a dismal prognosis, which confirms that the TP53 abnormality is key to the aggressiveness of these lymphomas. This is also consistent with recent findings that not all double- or triple-hit lymphomas have a poor prognosis (22, 23). In fact, our GCB2 group, which are the cases that were DHITsig positive, but lacked TP53 abnormalities, had a good prognosis despite the fact that many of these cases were double- or triple-hit lymphomas by FISH analysis (Figs. 2 and 3). This emphasizes the need to test for the DHITsig by GEP in conjunction with TP53 mutation and copy-number analysis to identify these high-risk patients. Cases that were DHITsig negative with TP53 abnormalities did not have a similar poor prognosis, likely due to residual TP53 function (Figs. 2 and 3). We believe the DHITsig is more robust and identifies more cases with a double-hit–like biology and clinical behavior, but FISH analysis for DH/TH may be used if GEP is not readily available.

We found that the DHITsig-positive cases without TP53 abnormalities (GCB2) had a relatively good survival, thus supporting this separation as significant from both the molecular and biological standpoints. Although our GCB2 and GCB3 groups had similarity to the EZB group of Schmitz and colleagues (17), we divided them according to the presence or absence of the DHITsig, with the positive ones assigned to GCB2. With the removal of the DHITsig-positive cases, the EZB-like group (GCB3) had an intermediate survival, indicating that these groups may be important to distinguish.

The cases in GCB3 (EZB-like) and GCB4 groups showed a higher proportion of mutations in SGK1, CARD11, and histone linker genes (e.g., HIST1H1E), and GCB4 also showed a higher proportion of TET2 mutations (25%) compared with GCB3 (11%). We found that GCB4 had a mutation pattern similar to cluster 4 of Chapuy and colleagues (20) with a good prognosis, whereas GCB3 was similar to cluster 3 or the EZB group, with a high frequency of BCL2 translocations and BCL2, CREBBP, and KMT2D mutations. However, cluster 3 also includes a proportion of DHITsig-positive cases, which we separated out in our study (GCB2) as a distinct biologic subgroup. We found that GCB3 had an intermediate prognosis when compared with GCB4 in our training cohort, but saw less of a difference in the validation cohort. These differences may be due to sample size, and a larger cohort is necessary to resolve these findings.

By using this risk stratification of GCB DLBCL, R-CHOP is sufficient to treat patients with GCB2 and GCB4, but patients with GCB3 may benefit from novel approaches, such as inclusion of an EZH2 or BCL2 inhibitor (Fig. 6). Patients with GCB1 should be enrolled in clinical trials evaluating novel therapeutic approaches, such as CD19-directed chimeric antigen receptor T cells, which are being evaluated as part of first-line therapy in patients with double- or triple-hit DLBCL (NCT03761056; https:clinicaltrials.gov/ct2/show/NCT03761056).

There are some limitations to our study, which include the retrospective nature of the study and the long time to accrue the cases, which may have led to an inherent selection bias. We tried to minimize this by including consecutive cases from the general population (Canadian territories, such as Manitoba in the training cohort, and British Columbia in the validation cohort). We also realize that the sample size in the study and validation cohorts is small and acknowledge that more cases need to be studied in the future to confirm our findings.

In conclusion, we found that patients who are DHITsig positive with TP53 abnormalities have a dismal prognosis and represent the highest risk subgroup of patients with GCB DLBCL. In contrast, DHITsig-positive cases without TP53 abnormalities (GCB2) have a good prognosis, EZB-like cases (GCB3) without the DHITsig have an intermediate prognosis, and cases that are DHITsig negative and lack an EZH2 mutation and/or BCL2 translocation (GCB4) have an excellent prognosis when treated with R-CHOP. Although analysis of more cases is necessary to confirm our findings, we believe that using just the DLBCL90 GEP, mutation profiling for TP53 and EZH2, and copy-number analysis for TP53, simplifies the prognostication of these patients and identifies those who may benefit from novel therapies or enrollment in a clinical trial.

A.F. Herrera reports grants and personal fees from Bristol Myers Squibb, Merck, Seattle Genetics, Genentech, and Kite/Gilead, personal fees from Karyopharm and Adaptive Biotechnologies, and grants from ADC Therapeutics outside the submitted work. P. Skrabek reports other from Novartis and Bristol Myers Squibb outside the submitted work. L. Popplewell reports other from Pfizer, Inc., Novartis Pharmaceuticals Corporation, and F. Hoffman-La Roche AG outside the submitted work. L.W. Kwak reports grants, personal fees, and other from Pepromene Bio, InnoLifes, personal fees and other from Theratest, personal fees from Enzychem LifeSciences, CJ Healthcare, and SELLAS Life Sciences, and personal fees and nonfinancial support from Celltrion outside the submitted work. D.W. Scott reports personal fees from AbbVie, AstraZeneca, grants and personal fees from Janssen, and grants from NanoString outside the submitted work, as well as has a patent for molecular subtyping of DLBCL pending and licensed to NanoString and a patent for identifying a gene expression signature for double hit lymphoma pending. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

J.Y. Song: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. A.M. Perry: Conceptualization, resources, data curation, validation, writing-review and editing. A.F. Herrera: Conceptualization, data curation, formal analysis, methodology, writing-review and editing. L. Chen: Data curation, software, validation, visualization, methodology, writing-original draft, writing-review and editing. P. Skrabek: Data curation, formal analysis, supervision, validation, writing-review and editing. M.R. Nasr: Data curation, validation, writing-review and editing. R.A. Ottesen: Data curation, software, formal analysis, validation, methodology, writing-review and editing. J. Nikowitz: Data curation, software, formal analysis, validation, writing-review and editing. V. Bedell: Data curation, methodology, writing-review and editing. J. Murata-Collins: Data curation, methodology, writing-review and editing. Y. Li: Data curation, investigation, writing-review and editing. C. McCarthy: Data curation, software, formal analysis, validation, investigation, writing-review and editing. R. Pillai: Formal analysis, writing-review and editing. J. Wang: Data curation, software, validation, investigation, writing-review and editing. X. Wu: Data curation, software, formal analysis, writing-review and editing. J. Zain: Resources, investigation, writing-review and editing. L. Popplewell: Resources, investigation, writing-review and editing. L.W. Kwak: Resources, funding acquisition, writing-review and editing. A.P. Nademanee: Data curation, formal analysis, investigation, writing-review and editing. J.C. Niland: Resources, data curation, software, supervision, writing-review and editing. D.W. Scott: Resources, data curation, software, validation, writing-review and editing. Q. Gong: Data curation, software, formal analysis, validation, investigation, methodology, writing-review and editing. W.C. Chan: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing. D.D. Weisenburger: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing-original draft, project administration, writing-review and editing.

This research was supported by the City of Hope National Medical Center Department of Pathology and the Toni Stephenson Lymphoma Center. Research reported in this article included work performed in the Pathology Core and Integrated Genomics Core supported by the NCI of the NIH under grant number P30CA033572.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.