|Home | About | Journals | Submit | Contact Us | Français|
Gene expression profiling studies have distinguished diffuse large B-cell lymphomas (DLBCLs) by cell of origin, with distinct pathogenetic mechanisms and prognosis. We attempted to identify DLBCL molecular subtypes in an epidemiologic study of 214 DLBCL patients diagnosed during 1998-2000 with archival tissues to investigate etiology. Immunohistochemical staining for CD10, BCL6, LMO2, MUM1/IRF4, and BCL2 and fluorescence in situ hybridization for t(14;18) were conducted, with ≥93% blinded duplicate agreement. CD10, LMO2, and BCL2 expression was similar to previous reports (32%, 44%, and 44% of DLBCLs, respectively), but BCL6 and MUM1/IRF4 expression was lower than expected (29% and 5%, respectively). We classified 112/214 (52%) cases as germinal center B-cell-like DLBCL (GCB-DLBCL; Hans et al., Blood 2004; CD10+ or CD10-/BCL6+/MUM1-), with no difference in prognosis compared with non-GCB-DLBCL (Cox regression, P=0.48). Comparing other GCB correlates, LMO2 expression and t(14;18) were more common but not exclusive to GCB-DLBCL as defined in our study, whereas BCL2 expression did not differ between DLBCL molecular subtypes. We could not confidently identify patients with GCB-DLBCL using these immunohistochemistry-based markers on archival tissues.
Diffuse large B-cell lymphoma (DLBCL), the most common non-Hodgkin lymphoma (NHL) subtype, represents a heterogeneous group of malignancies . Although DLBCL is potentially curable, with approximately 50% average five-year overall survival, the clinical course varies widely, and the International Prognostic Index (IPI) remains the best clinical prognostic predictor available [2-3].
Studies of gene expression profiles have distinguished DLBCL subtypes by cell of origin, demonstrating that patients with germinal center B-cell-like DLBCL (GCB-DLBCL) have a substantially better prognosis than patients with non-GCB-DLBCL . Additional research has shown that the disease subtypes are characterized by distinct oncogenic events [e.g., t(14;18) in GCB-DLBCL and constitutive NFkB activation in non-GCB-DLBCL] [5-6], leading us to hypothesize that the disease subtypes may also have distinct etiologies.
Identification of GCB- versus non-GCB DLBCL in epidemiologic studies to investigate etiology requires a biomarker that can be applied to fixed archival tissue specimens, since few epidemiologic studies collect fresh tissues. Using immunohistochemical (IHC) staining, Hans et al. distinguished DLBCL subtypes based on expression of three key markers of B-cell differentiation (CD10, BCL6, and MUM1/IRF4), by directly comparing results from IHC and gene expression for the same cases . However, other IHC-based studies of protein expression and DLBCL prognosis using these or other combinations of markers have yielded inconsistent results [8-19], possibly due to small sample sizes, use of convenience clinical samples with variability in patient populations, methodological differences in IHC staining and interpretation, and varying ability of IHC to recapitulate the GEP signature [20-21].
In order to identify molecular subtypes of DLBCL in a population-based epidemiologic study with archival tissue, we systematically evaluated expression of CD10, BCL6, and MUM1/IRF4 as well as other GCB correlates [LMO2 and BCL2 expression, t(14;18)].
The study population derived from the National Cancer Institute-Surveillance Epidemiology and End Results (NCI-SEER) NHL population-based case-control and prognosis studies, described in detail previously [22-23]. Briefly, the population-based case-control study included 1321 NHL cases (participation rate 76%) diagnosed during 1998-2000, aged 20-74 years, without known HIV infection, identified using rapid case ascertainment among residents of four SEER registries (Iowa, Detroit, Los Angeles, Seattle). Demographic data were obtained from patient interviews, and clinical data (date of diagnosis, stage, presence of B-symptoms, first course of therapy, date of last follow-up, and vital status) were obtained for 99% of cases from linkage to the SEER registries most recently in early 2008. Institutional Review Boards at the National Cancer Institute and each SEER center approved the study protocol. Participants provided written, informed consent prior to completing the interview.
All 1321 NHL cases were initially histologically confirmed as NHL and coded according to the International Classification of Diseases for Oncology, 2nd Edition (ICD-O-2) by the local diagnosing pathologist. Updated ICD-O-2/ICD-O-3 codes based on local pathology review were received during the 2008 SEER record linkage. Pathology reports were obtained for 1215/1321 (92%) patients for review by an expert hematopathologist (Mohammad A. Vasef, University of New Mexico), who classified cases according to the World Health Organization classification for lymphoid neoplasms and assigned a confidence score to the subtype diagnosis (≥90% versus <90%). For all cases with low confidence in the NHL subtype classification as well as a 10% random sample of cases with high confidence in the NHL subtype classification, additional immunostaining was conducted to establish the NHL subtype for those patients with available pathology material (N=472). All cases were then assigned a final diagnosis based on the review of tumor specimens (N=472) or pathology reports (N=743), updated ICD-O-2 code from the SEER record linkage if pathology review data were not available (N=76), or original ICD-O-2 code from SEER at the time of case identification if updated data were not available (N=30).
Cases assigned a final diagnosis of DLBCL (ICD-O-2: 9680-84, 9688, 9712; ICD-O-3: 9678-80, 9684; N=417, of which 381 were confirmed by pathology report and/or specimen review) were eligible for this analysis. Sufficient archived, unstained slides (5-micron sections) from formalin-fixed, paraffin-embedded tumors were available for 240/417 (58%) DLBCL patients for laboratory analysis; tumor specimens were not available for the remaining patients (N=177, 42%).
Immunohistochemistry (IHC) staining was conducted for five markers (CD10, BCL6, MUM1/ IRF4, LMO2, and BCL2) according to previously published methods [7, 16]. Samples were evaluated by expert hematopathologists, and those with ≥30% tumor cells stained were considered positive. Successful staining was achieved for ≥94% of slides for all markers. The final analytic study population included 214 DLBCL patients with IHC data for all five markers and available follow-up and clinical data. Duplicate slides from 28 randomly selected individuals were interspersed and blinded from the laboratory and hematopathologists, with ≥93% agreement for all markers. Additional information on the antibodies and approach to interpretation is available upon request.
We also identified tumors with the t(14;18) chromosomal translocation, the most common cytogenetic abnormality in NHL, using fluorescence in situ hybridization according to previously published methods (Bhavana J. Dave, Smrati Jain, University of Nebraska) .
The prognostic significance of marker expression was evaluated using hazard ratios (HRs) and 95% confidence intervals (CIs) derived from multivariate Cox proportional hazards regression, adjusting for age, demographics (sex, race, study center, years of education), and clinical factors (stage, presence of B-symptoms, type of initial therapy) using a continuous risk score (SAS v9.1, SAS Institute, Inc., Cary, NC) . Characteristics of the DLBCL molecular subtypes were compared using the Pearson chi-square statistic.
The study population was predominantly non-Hispanic white (90%) and male (59%) with a median age at DLBCL diagnosis of 59 years (Table 1). Characteristics of the present DLBCL case subset were comparable to all DLBCL cases from the parent case-control study. Clinically, 28% of patients had B-symptoms, and 86% received initial chemotherapy (pre-rituximab era). During follow-up, 69 (32%) patients died. The median follow-up of living patients was 7.7 years (range, 2.3-9.0).
CD10, LMO2, and BCL2 were expressed in 32%, 44%, and 44% of DLBCLs, respectively, similar to previous reports, but expression of BCL6 and MUM1/IRF4 was lower than expected (29% and 5%, respectively). Of the five markers evaluated, the strongest individual predictors of overall survival were BCL2 (44% positive, HR=2.15, 95%CI 1.33-3.50, P=0.0019) and LMO2 (44% positive, HR=0.62, 95%CI 0.38-1.03, P=0.062; Table 2).
Based on the combination of markers recommended by Hans et al. , we classified 112 patients with GCB-DLBCL (CD10+ or CD10-/ BCL6+/MUM1-), and while these patients had somewhat better overall survival than patients with non-GCB-DLBCL (HR=0.84, 95%CI 0.52-1.35; Table 2), this was not statistically significant (P=0.48). To further evaluate the molecular classification of the DLBCLs, we compared expression of other GCB correlates. LMO2 expression and the t(14;18) were more common in GCB- than non-GCB-DLBCL [LMO2: 61% versus 26%, P<0.0001; t(14;18): 49% versus 36%, P=0.06), whereas BCL2 expression did not differ between the DLBCL molecular subtypes (51% versus 44%, P=0.32). Other clinical and demographic characteristics also were similar among patients with GCB- and non-GCB-DLBCL (Table 1).
We used IHC staining and fluorescence in situ hybridization for six key markers of B-cell differentiation to attempt to discriminate DLBCL molecular subtypes in a population-based epidemiologic study with archival tissues. Using Hans et al. , we classified 112 (52%) cases as GCB-DLBCL (CD10+ or CD10-/BCL6+/MUM1-), but patterns of additional molecular characteristics in our data suggested that we could not be confident in our classification. We observed the t(14;18) in 37% of non-GCB-DLBCL, whereas other series have reported a prevalence of <5% . This observation suggests misclassification of the GCB-DLBCL subtype in our cases because of the high reliability of using fluorescence in situ hybridization to identify the t(14;18). Further supporting the likelihood of misclassification, expression of the GC marker LMO2 was absent in 39% of GCB-DLBCL and present in 26% of non-GCB-DLBCL, whereas expression of the activated B-cell marker BCL2 was present in 51% of GCB-DLBCL and absent in 56% of non-GCB-DLBCL. Although both LMO2 and BCL2 are expressed in a subset of lymphocytes at other stages of differentiation [16, 25], their lack of clear clustering with the DLBCL molecular subtypes in our data suggests that we cannot rule out the possibility of misclassification of the GCB-DLBCL subtype in our cases.
Our experience is illustrative of the promise and the challenges of molecular subtyping in epidemiologic studies. Our duplicate quality control samples indicated ≥93% agreement for all markers, supporting the internal consistency of our results. However, previous research has demonstrated that variability in IHC staining and interpretation can markedly affect results, particularly for BCL6 and MUM1/IRF4 . This variability and degradation of the antigens on our stored slides may account for the lower than expected percentage of cases staining positive for each of these markers in our study and would contribute to misclassification of the DLBCL molecular subtypes. Specifically, a substantially larger proportion of our cases compared with Hans et al.  were classified as non -GCB-DLBCL (93% vs. 69%) due to a lack of BCL6 expression (Figure 1).
Although our study was population-based and used rapid reporting, we were unable to enroll cases with rapidly fatal disease. Observed survival in our study was therefore better than expected in the general population, but consistent with survival estimates for DLBCL patients surviving one year following diagnosis . Exclusion of cases with rapidly fatal disease may have limited our ability to assess the prognostic significance of certain markers and decreased the proportion of non-GCB-DLBCLs, which have worse prognosis, in our population. Additionally, we were unable to obtain tumor tissue for all cases, though patient characteristics were similar for those with and without tissue. Detailed treatment and clinical data were not available, but our demographic and clinical risk score has a level of predictability similar to the IPI [5, 23].
Finally, we did not evaluate GCET1 and FOXP1, which were recently shown to modestly improve identification of GCB-DLBCL compared with Hans et al..
Discrimination of DLBCL molecular subtypes in epidemiologic studies requires identification of valid and reliable markers that can be used on archival tissues, are robust to variability in patient populations (e.g., by treatment, race, or comorbid conditions such as AIDS),[12, 14-15, 17-19] and have high inter-laboratory agreement . Establishment of such markers will also facilitate widespread adoption of DLBCL molecular subtyping in routine clinical care. Although IHC staining is a standard approach, methodological differences in IHC staining and interpretation for some of the current key B-cell markers suggest the need for new markers, further optimization of current markers for archival tissues, or the pursuit of alternative approaches .
We gratefully acknowledge the contributions of the staff and scientists at the SEER centers of Iowa, Los Angeles, Detroit, and Seattle for the conduct of the study's field effort; Freda Selk (University of Iowa) for coordination of tumor specimens; and Matthew Butcher, Mary McAdams, Pete Hui, Lonn Irish, and Michael Stagner (Information Management Services, Inc.) for programming support.
The authors have no conflicts of interest to disclose.