|
|||||||||
|
|
|
|
|
|
|
|
|
|
|
Blood, 1 December 2003, Vol. 102, No. 12, pp. 3871-3879. Prepublished online as a Blood First Edition Paper on August 21, 2003; DOI 10.1182/blood-2003-06-1841.
PLENARY PAPERS The molecular signature of mediastinal large B-cell lymphoma differs from that of other diffuse large B-cell lymphomas and shares features with classical Hodgkin lymphomaFrom the Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA; Whitehead Institute/Massachusetts Institute of Technology (MIT) Center for Genome Research, Cambridge, MA; Department of Pathology, Brigham and Women's Hospital, Boston, MA; Institute for Cancer Genetics, Columbia University, New York, NY; Department of Biostatistical Science, Dana-Farber Cancer Institute, Boston, MA; Department of Pathology, Massachusetts General Hospital, Boston, MA; Department of Pathology, Mayo Clinic, Rochester, MN; Hematology Department, Centre Hospitalier Lyon-Sud, Lyon, France; Department of Pathology, Centre Hospitalier Lyon-Sud, Lyon, France; Division of Hematology and Department of Medicine, Mayo Clinic, Rochester, MN; Department of Pediatric Oncology, Dana-Farber Cancer Institute and Howard Hughes Medical Institute, Boston, MA.
Mediastinal large B-cell lymphoma (MLBCL) is a recently identified subtype of diffuse large B-cell lymphoma (DLBCL) that characteristically presents as localized tumors in young female patients. Although MLBCL has distinctive pathologic features, it clinically resembles the nodular sclerosis subtype of classical Hodgkin lymphoma (cHL). To elucidate the molecular features of MLBCL, we compared the gene expression profiles of newly diagnosed MLBCL and DLBCL and developed a classifier of these diseases. MLBCLs had low levels of expression of multiple components of the B-cell receptor signaling cascade, a profile resembling that of Reed-Sternberg cells of cHL. Like cHLs, MLBCLs also had high levels of expression of the interleukin-13 (IL-13) receptor and downstream effectors of IL-13 signaling (Janus kinase-2 [JAK2] and signal transducer and activator of transcription-1 [STAT1]), tumor necrosis factor (TNF) family members, and TNF receptor-associated factor-1 (TRAF1). Increased expression of STAT1 and TRAF1 in MLBCL was confirmed by immunohisto-chemistry. Given the TRAF1 expression and known link to nuclear factor–  B (NF- B), MLBCLs were also evaluated for nuclear translocation of c-REL protein. In almost all cases, c-REL was localized to the nucleus, consistent with activation of the NF-B pathway. These studies identify a molecular link between MLBCL and cHL and a shared survival pathway.
Diffuse large B-cell lymphomas (DLBCLs) are heterogeneous tumors with variable clinical presentations and responses to therapy.1 At diagnosis, these aggressive B-cell malignancies involve a wide variety of nodal and extranodal disease sites; however, it is unclear whether tumors presenting in specific anatomic locations exhibit unique molecular features. One likely exception is the recently identified entity, primary mediastinal (thymic) large B-cell lymphoma (MLBCL), a subtype of DLBCL defined by a combination of clinical and pathologic features.2,3 Unlike DLBCL, which commonly arises in elderly patients of both sexes, MLBCL typically presents in younger women. These patients have bulky mediastinal masses with frequent invasion of adjacent structures.2 Although patients with MLBCL rarely have extrathoracic disease at diagnosis, extranodal sites including the central nervous system (CNS), liver, adrenals, and kidneys are often involved at relapse.4 MLBCLs typically consist of tumor cells with a pale cytoplasm and a diffuse growth pattern associated with variable degrees of sclerosis.3,5 However, there are no histologic features that reliably distinguish these tumors from DLBCLs that often involve mediastinal regional lymph nodes.
Although MLBCLs have rearranged immunoglobulin (Ig) genes and express B-cell lineage markers such as CD19, CD20, CD22, and CD79a (Ig MLBCLs have mutated, class-switched Ig genes without evidence of ongoing somatic mutation.9 These tumors also exhibit several characteristic genetic abnormalities, including gains of chromosomes 9p and 2p, and the associated JAK2 (9p24) and REL (2p16) loci.10,11 In contrast to DLBCLs, MLBCLs do not exhibit BCL2 rearrangements and rarely manifest translocations involving BCL6.12 Because MLBCL is currently considered a subtype of DLBCL (as opposed to a distinct disease), these tumors are treated similarly with empiric adriamycin-containing combination chemotherapy regimens. Involved-field radiation therapy is often added to the site of bulky, localized disease.13 However, combined modality therapy is only partially effective, because more than 40% of MLBCL patients die of their disease.13 Further, involved-field radiation therapy has been associated with long-term side effects including secondary malignancies and cardiac dysfunction.14 For these reasons, additional insights into the molecular signature of MLBCL and potential rational treatment targets are critically needed. To delineate unique features of MLBCL compared with DLBCL, we performed gene expression profiling of these 2 tumors and found them to be markedly different. Importantly, we also identified a molecular link between MLBCL and another lymphoid neoplasm with shared clinical features, classical Hodgkin lymphoma (cHL).
Case selection and histologic classification Frozen tumor specimens from newly diagnosed, previously untreated MLBCL patients (34) and DLBCL patients (176) were analyzed according to protocol approved by the institutional review board of the Dana Farber/Harvard Cancer Center. MLBCL tumor specimens were derived from mediastinal masses or contiguous nodal biopsies, and DLBCLs were all nodal tumor specimens. Primary MLBCLs were identified using clinical criteria (predominant mediastinal mass with or without local extension and no extrathoracic disease) and pathologic features. The histopathology and immunophenotype of each primary MBLCL were reviewed by expert hematopathologists to confirm the diagnosis. The diagnosis of each DLBCL was originally made by expert hematopathologists at the participating institutions. Target cRNAs of oligonucleotide microarrays Total RNA was extracted from each frozen tumor specimen, and biotinylated cRNAs were generated as previously described.15,16 Samples were hybridized overnight to Affymetrix U133A and U133B oligonucleotide microarrays (Affymetrix, Santa Clara, CA), which include probe sets from more than 44 000 genes. Arrays were subsequently developed with phycoerythrin-conjugated streptavidin (SAPE) and biotinylated antibody against streptavidin and scanned to obtain quantitative gene expression levels.15 The raw gene expression values were then scaled to account for differences in global chip intensity. Gene expression analysis
The top 15 000 genes from the U133A and U133B Affymetrix chips were selected based on their ranking as measured by a median absolute deviation (MAD) variation filter across all samples. From within this 15 000-gene pool, genes correlating with the class distinction of interest (mediastinal = 1 versus nonmediastinal = 2) were identified by ranking them according to their signal-to-noise ratio (SNR). For a given gene g, SNR (g) = ( MLBCL versus DLBCL classifier The discriminatory power of the gene expression signature was evaluated by building classifiers for the MLBCL versus DLBCL distinction. Naive Bayes (NB) and weighted voting (WV) classifiers that included 10 to 1000 genes were built and evaluated with regard to prediction errors using leave-one-out cross-validation (LOO-CV) (see Supplemental Document). The classifier with the lowest balanced error rate (proportion of samples wrongly classified, averaged within classes) was chosen for further analysis (see Supplemental Document). A more rigorous estimate of the error rate was also computed based on a 2-level CV procedure in which the choice of how many genes to use is automatically made within the CV loop (see Supplemental Document). Analysis of coregulated genes Genes that were coregulated (ie, genes whose expression values follow a similar pattern) with the top-ranked genes in MLBCL were identified using a 1 minus the Pearson correlation coefficient as the distance metric (see Supplemental Document). Enrichment test for cHL genes in MLBCL An enrichment test was used to evaluate the significance of the observed similarity between the MLBCL and cHL signatures. The cHL signature was defined using a set of genes independently identified by Schwering et al as differentially expressed in Hodgkin Reed-Sternberg (HRS) cell lines and normal purified B cells (centroblasts, centrocytes, and naive and memory B cells) using Affymetrix U95 oligonucleotide arrays.17 The similarity between the MLBCL and HRS signatures was assessed using the following recently described procedure.18 A total of 15 000 genes were selected from the U133A/B chips according to a MAD-based variation filter and ranked according to their SNR with respect to the "MLBCL versus DLBCL"class membership. The 294 genes with reduced expression in HRS cell lines17 were then located within our ranked list of 15 000 genes, and their proximity to the genes with lower levels of expression in MLBCL was measured by a Kolmogorov-Smirnoff (KS) score (with a higher score corresponding to a higher proximity). A similar procedure was used to place the 195 genes with increased expression in HRS cell lines17,19 within the 15 000 genes ranked in the opposite direction. Thereafter, permutation of the "MLBCL versus DLBCL" sample labels, associated reranking of the 15 000 genes, and computation of the corresponding KS scores were performed multiple times (n = 1000) so as to compare the observed KS scores with those that could be expected by chance under a random class labeling. Empirical P values were then computed to quantify the signifi-cance of the observed similarities between the genes with increased and decreased expression in the HRS cell lines and MLBCLs (pmax). An alternative, less stringent, method was also used to compute an empirical P value (pmin). This method was based on the computation of the KS score for multiple (n = 1000) random sets of 294 or 195 genes. This approach determines how likely it would be to obtain the observed KS score if a random set of genes were selected (see Supplemental Document). Immunohistochemistry Immunohistochemistry with anti–signal transducer and activator of transcription-1 (anti-STAT1) (9H2, Cell Signaling Technology, Beverly, MA) or anti–tumor necrosis factor (TNF) receptor-associated factor-1 (anti-TRAF1) (H3; Santa Cruz Biotechnology, Santa Cruz, CA) murine monoclonal antibodies was performed using 3 to 5 µ-thick formalin-fixed, paraffin-embedded specimens by standard immunohistochemical methods. Slides were deparaffinized and pretreated with 10 mM citrate, pH 6.0, (Zymed, South San Francisco, CA) in a steam pressure cooker (Decloaking Chamber, BioCare Medical, Walnut Creek, CA). Slides were then treated with Peroxidase Block (DAKO USA, Carpinteria, CA) for 5 minutes to quench endogenous peroxidase activity and incubated with a 1:5 dilution of goat serum in 50 mM Tris (tris(hydroxymethyl)aminomethane)–HCl, pH 7.4, for 20 minutes to block nonspecific binding sites. The anti-STAT1 antibody and anti-TRAF1 antibodies were diluted in 50 mM Tris-HCl, pH 7.4, with 3% goat serum (anti-STAT1 1:1000, anti-TRAF1 1:250) and applied to slides at room temperature for 1 hour. Goat antimouse horseradish peroxidase–conjugated antibody (Envision detection kit, DAKO USA) was applied for 30 minutes and developed using a diaminobenzidine (DAB) chromogen kit (DAKO USA).
c-REL immunohistochemistry was performed as outlined above with the following modifications. Formalin-fixed, paraffin-embedded sections were dewaxed and antigen retrieved in 1 mM EDTA (ethylenediaminetetraacetic acid) as published.20 Following blocking of nonspecific binding with 5% nonfat dried milk in Tris-buffered saline, pH 7.5, slides were incubated overnight with either affinity-purified rabbit anti–c-REL antibody at 1 µg/mL (PC139, Oncogene Research Products, Darmstadt, Germany) or pooled nonspecific rabbit Ig (Sigma, St Louis, MO). The slides were then washed and incubated with biotin-conjugated goat antirabbit secondary antibody (Vector, Burlingame, CA), followed by fluorescein isothiocyanate (FITC)–avidin (Molecular Probes, Eugene, OR). The sections were mounted and counterstained with propidium iodide (Molecular Probes). In separately reported studies, the association between nuclear c-REL immunostaining and nuclear factor– Fluorescence in situ hybridization (FISH) Air-dried touch preparations were prepared on Superfrost Plus slides from fresh frozen tumor specimens and stored in a desiccator at room temperature until use. Nuclei were hybridized to commercially available probes flanking or spanning the IgH, BCL2, and BCL6 loci (Vysis, Downer's Grove, IL) using conditions recommended by the manufacturer. After counterstaining with 4,6 diamidino-2-phenylindole (DAPI), interphase nuclei were scored for various chromosomal aberrations by fluorescence microscopy.
Clinical and pathologic features of MLBCL The clinical characteristics of the MLBCL and DLBCL patients included in this analysis are outlined in Table 1. The 34 MLBCL patients were predominantly young females (median age, 32 years) with localized and bulky disease (78% stage I/II). Those MLBCL patients with advanced-stage disease had contiguous involvement of the pleura and/or pericardium. In contrast, DLBCL patients were older (median age, 64 years) with an equal sex distribution and a higher incidence of advanced-stage disease (65% stage III/IV, Table 1). Pathologic review of the diagnostic MLBCL specimens confirmed the presence of sclerosis in 94% of cases; 77% of cases with available Ig immunohistochemistry had undetectable Ig, and 12% had equivocal Ig expression (Table 1). In the subset of MLBCLs with available genetic data, there was only one tumor with a BCL6 rearrangement (n = 19) and none with BCL2 translocations (n = 17) (Table 1), consistent with previous reports.12
MLBCLs have a unique transcriptional profile The distinctive clinical and pathologic characteristics of MLBCL—predominantly young women with Ig-negative, localized, and sclerotic tumors—suggested that this entity might also have a unique molecular signature. To address this possibility, diagnostic tumor specimens from the MBLCL and DLBCL cohorts were subjected to transcriptional profiling. The genes were sorted by their degree of correlation with the MLBCL versus DLBCL distinction according to the signal-to-noise metric (see "Patients, materials, and methods"). Permutation of the sample labels indicated that MLBCL had significantly lower expression of more than 1000 genes and significantly higher expression of more than 1000 additional genes, compared with DLBCLs (P < .01) (Figure 1A).
The top 50 genes with significantly higher or lower expression in MLBCL (as compared with DLBCL) are visually displayed in Figure 1B. The MLBCL transcriptional signature also included 3 genes previously reported to be expressed at high levels in this disease: the cell surface protein and lipid raft component, MAL; the recently described interleukin-4 (IL-4)–induced gene, FIG1; and the adhesion molecule, CD58/LFA3 (Table 2).8,21,22 These tumors also had lower levels of IgM, consistent with the previously noted absence of Ig in most cases of MLBCL (Figure 1B and Table 3.)40
Class prediction The discriminatory power of the gene expression signature was evaluated by building naive Bayes (NB) and weighted voting classifiers for the MLBCL versus DLBCL distinction. These classifiers were tested using a leave-one-out cross-validation strategy. A 100-gene NB model achieved the lowest balanced error rate (11%) (Figure 1B and Supplemental Document). Of interest, 6 patients whose DLBCL was classified by the NB model as MLBCL had predominant, although not exclusive, mediastinal disease (Figure 1B), and 2 patients whose tumors did not involve the mediastinum exhibited c-REL amplification (data not shown). In addition, 45 patients whose disease involved regional mediastinal lymph nodes as well as other nodal and extranodal sites were identified by the NB model as having DLBCL rather than primary MLBCL (Figure 1B). These data highlight the potential value of a molecular classifier in entities that are currently defined with only clinical and pathologic criteria. MLBCL transcriptional profile resembles that of cHL Inspection of the MLBCL transcriptional profile revealed striking similarities to that of cHL (Figure 1B; Tables 2, 3). Like Hodgkin Reed-Sternberg (HRS) cells, MLBCLs had low levels of expression of multiple B-cell signaling components and coreceptors (Table 3 and Figure 2).17,42-44 MLBCLs also had high levels of expression of cytokine pathway components, TNF family members, and extracellular matrix elements previously identified in cHL (Table 2).45 These observations are of particular interest because MLBCL and the most common subtype of cHL (nodular sclerosis) have similar clinical presentations—in younger patients with local/mediastinal tumors characterized by reactive fibrosis.
To determine the statistical significance of a potential MLBCL-cHL connection, we performed an enrichment test using an independently identified series of differentially expressed genes in HRS cell lines (Figure 3).17,19 When compared with random class labelings, the Kolmogorov-Smirnov (KS) statistic indicated that the observed MLBCL-cHL similarity was highly significant. Interestingly, this similarity was primarily attributable to genes with low levels of expression in HRS cell lines (versus normal B cells) and primary MLBCL (versus DLBCL) (pmax = .012, pmin < .001) (Figure 3). In contrast, the similarity between genes with high levels of expression in HRS cell lines and primary MLBCL was less significant (pmax = .213, pmin = .007). This may reflect the contribution of the tumor microenvironment to the gene signatures of primary MLBCL and/or the absence of this contribution from profiles of HRS cell lines maintained in vitro.19,45
Decreased expression of BCR signaling cascade components in MLBCL
Like cHL, MLBCLs had low levels of expression of multiple components of the BCR signaling cascade including the cell surface immunoglobulin receptor, IgM; the Ig The phosphatidylinositol-3 kinase (PI-3K)–regulated downstream kinase, AKT, and a major AKT target, FORKHEADP1 (FOXP1), were also expressed at low levels in MLBCL46,49 (Table 3 and Figure 2). Lymphoid transcription factors including Spi-B and Ikaros were also less abundant in MLBCL (Table 3 and Figure 2).50 In addition to the above-mentioned B-cell transcription factors and components of the BCR signaling cascade, MLBCLs, like HRS cells, had low levels of expression of the CD22 coreceptor and several major histocompatibility complex (MHC) class II molecules (Table 3 and Figure 2).46,51 MBLCLs also expressed low levels of the germinal center (GC) metalloendopeptidase, CD10 (Table 3). Increased expression of cytokine pathway components, TNF family members, and extracellular matrix elements in MLBCL
Like primary cHL, MLBCLs had increased expression of specific cytokine pathway members. For example, MLBCLs expressed high levels of the interleukin-13 receptor
MLBCLs also had high levels of expression of
Equally striking in MLBCL was the increased expression of TNF superfamily (SF) members previously implicated in the pathogenesis of cHL, including the OX40 ligand (TNFSF4) and the FAS ligand receptor (TNFRSF6) previously identified in MLBCL38 (Table 2). The herpes virus entry mediator (HVEM, TNFRSF14) and TRAF1 were also expressed at high levels in MLBCL. These results are of interest because HVEM associates with TRAF1 to stimulate NF-
Signaling through TNF receptors triggers distinct signaling pathways leading to either activation of NF-
Primary MLBCL and cHL often exhibit sclerosis, underscoring the potential importance of extracellular matrix components in both diseases. Consistent with these observations, MLBCLs had higher levels of expression of a TNF- Expression of STAT1 and TRAF1 proteins in MLBCL
The critical roles of IL-13 signaling and NF-
There was also uniform TRAF1 staining in the malignant cells of all MLBCLs, whereas the DLBCLs lacked TRAF1 expression (Figure 4A and Table 4). Consistent with prior reports of TRAF1 transcripts and protein in HRS cells of cHL,31,32 all of the cHL cases exhibited prominent TRAF1 staining (Table 4).
MLBCLs have immunohistochemical evidence of NF-
MLBCLs that lack Ig and associated BCR survival signals have likely developed alternative mechanisms for escaping cell death, similar to those described in cHL.45 In cHL, activation of the NF-
The unique profile of primary MLBCL—low levels of expression of BCR signaling pathway components, a distinctive cytokine pathway signature, and activation of NF- B—is strikingly similar to that of a clinically related disorder, cHL. Both MLBCL and cHL of the nodular sclerosis subtype commonly present in young patients as mediastinal tumors with absent surface Ig and prominent fibrosis. MLBCL and classical HRS cells also exhibit common genetic abnormalities including gains of chromosome 2p and 9p. Found in about 20% of MLBCL and up to 50% of cHL, gains in chromosome 2p are associated with amplification of the REL locus, one potential mechanism for increased NF-B activity and tumor cell resistance to apoptosis.11,25,68,69 Gains in chromosome 9p and the JAK2 locus are observed in about 75% of MLBCLs and 25% of cHLs and are unique to these diseases.10,11,26 In addition to these clinical, immunologic, and molecular similarities, there are also rare reported cases of composite cHL and MLBCL, further supporting a pathogenetic relationship between these tumors.70 The clonal relationship between HRS cells and non-Hodgkin lymphoma (NHL) cells in composite lymphomas was one of the strongest initial pieces of evidence that cHL was a B-cell lymphoma.71 In our own series, one patient with MLBCL subsequently relapsed with cHL (nodular sclerosis subtype).
The extent of similarity between the nodular sclerosis subtype of cHL and MLBCL is perhaps surprising given the degree to which the histopathologies and outcome of these 2 entities differ. Cells resembling HRS cells are usually not seen in MLBCL, and MLBCL cells retain the expression of multiple B-cell markers, including CD20, CD79a, BOB1, and OCT2,6,72 that are lost in HRS cells.17,73 In addition, nodular sclerosis cHL is characteristically associated with a polymorphous inflammatory infiltrate rich in plasma cells, neutrophils, and eosinophils, reactive cell types that are usually absent in MLBCL.45 The reactive infiltrate in cHL is recruited and maintained by additional chemokines and cytokines, such as IL-5, IL-6, and IL-10,45 that are not part of the MLBCL signature. Based on the existence of unusual patients with both nodular sclerosis cHL and MLBCL, we suggest these 2 entities arise from a common precursor cell with a growth and survival advantage stemming from genetic lesions that result in constitutive activation of NF- The reason for the lower levels of expression of Ig and BCR signaling components in MLBCL is not yet known. In a subset of cHLs, the lack of a functional BCR has been attributed to "crippling" somatic mutations of the rearranged immunoglobulin gene.74 In cHLs with functional gene rearrangements, the absence of Ig has been ascribed to deficiencies in transcription factors (OCT2, BOB1, PU1) necessary for Ig synthesis.73,75 In a recent study, 12 of 13 MLBCLs exhibited functional IgVH gene rearrangements, suggesting that destructive somatic mutations were not the primary cause of reduced surface Ig in this disease.9 More recently, abundant OCT2 and BOB1 transcription factors and Ig transcripts with a switched isotype were detected in MLBCL, suggesting that the transcriptional machinery for Ig synthesis was intact in these tumors.72 However, our MLBCL transcriptional profile reveals reduced levels of other critical B-cell transcription factors such as NFATc and Spi-B, which have known roles in regulating the expression of specific Ig isotypes.76,77
MLBCL has a prominent cytokine pathway signature, which likely reflects dynamic interactions between the tumor cells, infiltrating inflammatory cells, and the surrounding tumor stroma. Like cHLs, MLBCLs have increased abundance of high-affinity IL-13 receptor subunits and downstream effectors (JAK2 and STAT1, NF–IL-3), implicating this cytokine pathway in disease pathogenesis.45 Our studies also indicate that MLBCLs overexpress specific TNF family members that are known to interact with TRAF1 and subsequently activate NF-
The nuclear localization of c-REL and likely activation of NF-
Submitted June 9, 2003; accepted August 17, 2003.
Prepublished online as Blood First Edition Paper, August 21, 2003; DOI 10.1182/blood-2003-06-1841.
The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked "advertisement" in accordance with 18 U.S.C. section 1734.
K.J.S. and S.M. contributed equally to this study. T.R.G. and M.A.S. contributed equally to this study.
The online version of the article contains a data supplement.
Reprints: Margaret A. Shipp, Dana-Farber Cancer Institute, 44 Binney St, Boston, MA 02115; e-mail: margaret_shipp{at}dfci.harvard.edu.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Abstract
Introduction
), these tumors do not express Ig.
1–
-associated SRC-family protein tyrosine kinase (PTK), BLK; the SYK PTK-regulated scaffolding protein, BLNK; and the Bruton tyrosine kinase (BTK) binding protein, SAB (
2 (PLC