Fax: (917) 432-2310
Bcl-6 predicts improved prognosis in primary central nervous system lymphoma
Article first published online: 12 NOV 2007
Copyright © 2007 American Cancer Society
Volume 112, Issue 1, pages 151–156, 1 January 2008
How to Cite
Levy, O., DeAngelis, L. M., Filippa, D. A., Panageas, K. S. and Abrey, L. E. (2008), Bcl-6 predicts improved prognosis in primary central nervous system lymphoma. Cancer, 112: 151–156. doi: 10.1002/cncr.23149
- Issue published online: 17 DEC 2007
- Article first published online: 12 NOV 2007
- Manuscript Accepted: 30 JUL 2007
- Manuscript Revised: 12 JUL 2007
- Manuscript Received: 22 MAY 2007
- primary central nervous system lymphoma;
- molecular markers;
- B-cell differentiation;
- prognostic factors
In systemic non-Hodgkin lymphoma (NHL), tumors that resemble germinal center B-cells are less aggressive than tumors that resemble postgerminal center B-cells. However, the value of B-cell differentiation markers in primary central nervous system lymphoma (PCNSL) is less clear. The objectives of this study were to characterize the immunophenotypes of PCNSL samples and to determine their utility in predicting clinical outcomes.
The immunohistochemical staining profile of PCNSL was determined from 66 immunocompetent patients. Then, the authors evaluated whether theexpression of these proteins was associated with progression-free or overall survival.
Only 8% of PCNSL samples were positive for the cluster designation (CD) germinal center marker CD10. Another germinal center marker, bcl-6, was positive in 47% of samples. Ninety-four percent of samples expressed significant levels of the postgerminal center marker melanoma-associated antigen (MUM-1). In univariate analyses, only bcl-6 staining had a significant effect on progression-free survival (median, 20.5 months in bcl-6-positive patients vs 10.1 months in bcl-6-negative patients; P = .02). There was a nonsignificant trend toward improved overall survival in patients who had bcl-6 expressing tumors. Older age and poorer performance status, as observed previously, were associated with reduced survival.
Bcl-6 expression was associated with a better prognosis in patients with PCNSL. Cancer 2008. © 2007 American Cancer Society.
Primary central nervous system (CNS) lymphoma (PCNSL) is an uncommon form of non-Hodgkin lymphoma (NHL) that is restricted to the CNS. There is a clear distinction between PCNSL in immunocompromised and immunocompetent patients. PCNSL in patients with normal immune status rarely is associated with Epstein-Barr virus (EBV) infection, and they have a better prognosis than immunosuppressed patients.1, 2 Nearly all PCNSLs in immunocompetent patients have the pathologic characteristics of diffuse large B-cell lymphoma (DLBCL), which is the most common type of systemic NHL.2, 3 Numerous studies have examined the prognostic value of biomarkers in systemic NHL. In systemic DLBCL, higher levels of the antiapoptotic protein bcl-2 correlate with worse clinical outcomes.4–6 Studies have yielded mixed results regarding the prognostic utility of other tumor-related proteins, such as p537, 8 and Ki-67.9, 10 Markers related to different stages of B-cell differentiation have robust prognostic significance. Specifically, tumors resembling germinal center B-cells (that express markers such as cluster designation [CD] CD10 and bcl-6) behave less aggressively and have a better prognosis than tumors resembling postgerminal center-activated B-cells (that express markers such as melanoma-associated antigen [mutated] 1 [MUM-1]).11, 12 This classification scheme was derived originally from assessments of messenger RNA expression, but immunohistochemical analysis of protein expression also has been used to divide DLBCL in a similar fashion and retains prognostic value.13–15
In patients with PCNSL, older age and worse performance status at diagnosis consistently predict worse overall survival (OS).16, 17 However, there is no clear association between the expression of any tumor biomarker and clinical outcome. Specifically, several studies have failed to demonstrate any correlation between bcl-2 expression and OS3, 18, 19 compared with systemic DLBCL. Similarly, p53 and Ki-67 levels have no prognostic significance.3, 18 The value of B-cell differentiation markers in PCNSL is unclear. In particular, bcl-6 expression has been associated with better outcome in some studies20, 21 but not in others.22, 23 In the current study, our objective was to characterize the expression patterns of traditional tumor-associated proteins and B-cell differentiation markers in PCNSL samples. We also sought to determine the value of these immunophenotypes for predicting clinical outcomes in PCNSL.
MATERIALS AND METHODS
Patients were diagnosed between 1986 and 2003 and were evaluated and treated at Memorial Sloan-Kettering Cancer Center. After obtaining Internal Review Board approval, clinical data were obtained through a registered database and chart review. Inclusion criteria included a biopsy-proven CNS lymphoma without evidence of systemic lymphoma; a negative human immunodeficiency test; and sufficient pathologic material for immunohistochemical studies.
Paraffin-embedded samples were processed in a standard fashion and were subjected to antigen-retrieval protocols using either citric acid or ethylenediamine tetraacetic acid. The following antibodies and dilutions were used: CD20 (Dako), 1:1000 dilution; CD5 (Dako), 1:200 dilution; CD3 (Dako), 1:10 dilution; P53 (Dako), 1:500 dilution; Ki-67 (Dako), 1:200 dilution; bcl-2 (Dako), 1:200 dilution; CD10 (Novocastra), 1:400 dilution; MUM-1 (Dako), 1:100 dilution; and bcl-6 (Novocastra), 1:100 dilution. All of these are monoclonal antibodies except for anti-CD3, which is polyclonal. Antibody binding was visualized with peroxidase-3,3′diaminobenzidine tetrahydrochloride staining. EBV-encoded RNA (EBER) was analyzed using an in situ hybridization probe (Ventana). Slides were analyzed by a single pathologist (D.A.F.) who was blinded to all clinical information. Based on prior studies in systemic DLBCL, the following cutoff values were used in characterizing a sample as positive or negative: 20% positive cells for p53 and 60% positive cells for Ki-67. For the other markers, a sample was considered positive if >50% of neoplastic cells were stained with the antibody.
Correlations between immunostaining of markers and clinical data were assessed using contingency tables and chi-square tests. To achieve a more homogenous population, patients who did not receive methotrexate as part of their treatment regimen were excluded from the survival analysis. OS was defined as the time from diagnosis to death or to last follow-up for censored observations. Progression-free survival (PFS) was defined as the time from diagnosis to disease recurrence, disease progression, or death; the time of last follow-up was used for censored observations. Survival curves were constructed using the Kaplan-Meier method. Univariate analyses of survival curves were performed using the log-rank test; only variables with at least 5 patients in each dichotomous category were assessed. Variables that were statistically significant in the univariate analysis (P <.05) were included in a multivariate analysis using the Cox proportional hazards regression model.
Baseline characteristics, treatment information, and clinical responses are shown in Table 1. There were 66 patients included in this study, their median age was 65 years, and their median Karnofsky performance status (KPS) was 60. There were approximately equal numbers of men and women. Approximately 12% of patients had ocular involvement, and 15% had positive cerebrospinal fluid cytology. With regard to treatment, 75% of patients received regimens that contained high-dose methotrexate, and 57% received whole-brain radiotherapy.
|Characteristic||Percent of patients (no. /Total)|
|Age at diagnosis, y|
|KPS at diagnosis|
|Treatment with methotrexate|
|Treatment with radiation therapy|
Histologic analysis of the samples revealed that 85% (50 of 59 samples) had centroblastic morphology, and 10% (6 of 59 samples) had immunoblastic morphology. In all samples, the large neoplastic cells were positive for the B-cell marker CD20 and negative for the T-cell marker CD3, confirming that all tumors were B-cell lymphomas. There were variable numbers of T-cells infiltrating the tumors, a feature that is observed commonly in systemic DLBCL. Only 1 of 48 samples (2%) was positive for EBER, verifying that our patient population was immunocompetent. We analyzed the expression of bcl-2, Ki-67, and p53(Table 2). These proteins are involved in processes important for tumor behavior (apoptosis, proliferation, DNA repair) and have been studied in systemic DLBCL as well as other tumor types. Fifty-four of 61 samples tested (89%) were positive for bcl-2,32 of 61samples (52%) were positive for p53, and 28 of 60 samples (47%) were positive for Ki-67. Corresponding values for systemic DLBCL are shown for comparison.
|Marker*||Percent positive (absolute no.)||Percent positive in systemic DLBCL†|
|Germinal center(CD10+/bcl-6+[n = 4]; CD10+/bcl-6− [n = 1])‡||13 (5/38)||42|
|Nongerminal center (CD10−/bcl-6− [n = 30]; CD10−/bcl-6+/MUM1+ [n = 3])‡||87 (33/38)||52–58|
We measured the expression of 3 proteins that are expressed during different stages of B-cell differentiation: CD10, bcl-6, and MUM-1. Nearly all tumors (55 of 60 tumors; 92%) had low levels of CD10, which is expressed normally by germinal center B-cells. Bcl-6, another germinal center marker, was expressed in 26 of 57 tumors (46%). The postgerminal center marker MUM1 was identified in most tumor cells in 17 of 18 samples (94%) that were assayed. Unfortunately, sufficient material to perform MUM1 staining was available for only 18 samples. Using a classification scheme that was developed for systemic DLBCL,15 we had sufficient information to categorize 38 samples. Of these, 5 of 38 samples (13%) had a germinal center phenotype, whereas 33 of 38 samples (87%) had a nongerminal center phenotype.
Response and Survival Analysis
In the total group, 64% achieved a complete response with initial therapy. The median PFS was 15.7 months, and the median OS was 34.6 months in patients who received methotrexate. In prior studies, younger age at diagnosis and higher performance status were established as positive prognostic variables. In our study, age ≤60 years was associated with a longer OS (P = .006) (Table 3) and higher likelihood of a complete response (P = .005). Younger patients also demonstrated a trend toward longer PFS that nearly reached statistical significance (P = .053). Patients with a KPS ≥70 tended to have better PFS and OS, although these associations did not reach statistical significance (P = .12 and P = .11, respectively). Among all of the protein markers that we analyzed, only positive bcl-6 status was correlated with better PFS (P = .02) (Fig. 1). There was also a trend toward improved OS in patients who had bcl-6-positive tumors (P = .18). In a multivariate analysis that included age, KPS, and bcl-6 status, patients with bcl-6-positive tumors tended to have longer PFS (P = .16). In this analysis, younger age retained its significant association with better OS (P = .003).
|Variable||Median PFS, mo||P||Median OS, mo||P|
In this report, we have described the expression profiles of traditional tumor-associated proteins and B-cell differentiation markers in PCNSL samples from 66 patients. In addition, we assessed the association between expression of these proteins and clinical outcomes. We observed that, as a group, patients with PCNSL have high levels of bcl-2, Ki-67, and the postgerminal center protein MUM1, and they have low levels of the germinal center marker CD10. Therefore, the majority of the tumors in our sample would be classified as a nongerminal center phenotype according to the most widely used classification scheme for systemic DLBCL. Similar to what was reported in several prior studies, younger patients with better performance status tended to have superior clinical outcomes. Expression of bcl-6, which is considered a germinal center marker, was associated with longer PFS.
Patients with PCNSL have a shorter OS than patients with systemic DLBCL. Our data indicate that lymphomas arising in the CNS have more aggressive biologic characteristics relative to systemic DLBCL. Compared with the values reported for systemic DLBCL, we observed higher levels of the tumor-associated proteins bcl-2 and Ki-67 in PCNSL. These proteins most likely enhance tumor survival and proliferation, leading to worse clinical outcomes. Our analysis of B-cell differentiation markers also suggested a more aggressive phenotype. We observed that 87% of samples had a nongerminal center phenotype, which is associated with a poor prognosis in patients with systemic DLBCL. Other groups have generated similar findings. In contrast, only 52% to 58% of systemic DLBCL are classified as nongerminal center. Other factors also may contribute to the inferior clinical outcomes reported in patients with PCNSL, including local factors in CNS tissue and the blood-brain barrier.
There has been considerable debate regarding the origin of PCNSL. In our series, the overwhelming majority of tumors were CD10-negative and MUM1-positive. CD10 typically is expressed in germinal center B-cells, whereas MUM1 is expressed in B-cells that have left the germinal center. Using the classification developed by Hans et al.,15 we observed that almost all of the tumors had a nongerminal center phenotype. Other studies have yielded conflicting evidence regarding the stage of B-cell differentiation that corresponds most closely to PCNSL. Similar to MUM1, short heterodimer partner 1 (SHP1) is a transcription factor that is up-regulated in postgerminal center B cells. SHP1 is expressed in nearly all PCNSL.22 Conversely, CD138, a marker of later stage postgerminal center B cells, is observed rarely in PCNSL.20–23 Furthermore, immunoglobulin gene analysis in PCNSL demonstrated hypermutated but polyclonal gene rearrangements, most consistent with germinal center B cells.24–26 Taken together, the available data suggest that the majority of PCNSL corresponds to the transition between germinal center and postgerminal center B cells.
Bcl-6 was the only protein with expression that had significant prognostic value. Prior studies have generated conflicting results regarding an association between bcl-6 and survival. In a series of 31 patients with PCNSL, Braaten et al.20 reported that bcl-6 expression was associated with a significant increase in OS (101 months vs 14.7 months). However, other investigators observed no difference in survival between patients based on bcl-6 expression.21, 22 In 1 study, bcl-6 expression was more common in patients who were older and had a worse performance status,22 which may have biased the survival analysis. In the current study, multivariate analysis demonstrated a nonsignificant trend toward improved PFS in patients with bcl-6-positive tumors (P = .16); whereas younger age retained its significance in this analysis, suggesting that bcl-6 status may not be an independent prognostic factor but, rather, reflects younger age. However, we did not observe any association between bcl-6 status and age or performance status (data not shown). It is possible that our sample size was too small for bcl-6 status to retain statistical significance in a more rigorous multivariate analysis.
Bcl-6 is considered a marker of germinal center B cells. Almost half (46%) of PCNSL samples were bcl-6-positive, despite the finding that most PCNSLs were classified as the nongerminal center phenotype according to the criteria of Hans et al. Given the prognostic significance of bcl-6 expression, bcl-6 status most likely reflects a biologically meaningful feature of PCNSL. We believe that, for PCNSL, the nongerminal center category should be subdivided further into early and late nongerminal center phenotypes. This scheme is similar to that developed by Chang et al.13 for systemic DLBCL. The early nongerminal center phenotype is characterized by bcl-6 expression and a more favorable prognosis. Late nongerminal center PCNSL lacks bcl-6 expression and has a poorer prognosis.
The current study was limited by several factors: There were relatively small numbers of PCNSL samples; the analysis was performed retrospectively; all patients were from a single center; multiple markers were studied, and most did not have prognostic value; and few samples were available for MUM1 staining. Therefore, this study should be viewed as hypothesis-generating. Further prospective studies in uniformly treated patients should be performed to confirm our findings and the value of bcl-6 status in determining the prognosis of patients with PCNSL.