Further Delineation of Chromosomal Consensus Regions in Primary

Further delineation of chromosomal consensus regions in primary mediastinal B-cell lymphomas: an analysis of 37 tumor samples using high-resolution genomic proﬁling (array-CGH)

S Wessendorf1,6, TFE Barth2,6, A Viardot1, A Mueller3, HA Kestler3, H Kohlhammer1, P Lichter4, M Bentz5,HDo¨hner1,PMo¨ller2 and C Schwaenen1

1Klinik fu¨r Innere Medizin III, Zentrum fu¨r Innere Medizin der Universita¨t Ulm, Ulm, Germany; 2Institut fu¨r Pathologie, Universita¨t Ulm, Ulm, Germany; 3Forschungsdozentur Bioinformatik, Universita¨t Ulm, Ulm, Germany; 4Abt. Molekulare Genetik, Deutsches Krebsforschungszentrum, Heidelberg, Germany and 5Sta¨dtisches Klinikum Karlsruhe, Karlsruhe, Germany

Primary mediastinal B-cell lymphoma (PMBL) is an aggressive the expression of BSAP, BOB1, OCT2, PAX5 and PU1 was extranodal B-cell non-Hodgkin’s lymphoma with specific clin- added to the spectrum typical of PMBL features.9 ical, histopathological and genomic features. To characterize Genetically, a pattern of highly recurrent karyotype alterations further the genotype of PMBL, we analyzed 37 tumor samples and PMBL cell lines Med-B1 and Karpas1106P using array- with the hallmark of chromosomal gains of the subtelomeric based comparative genomic hybridization (matrix- or array- region of chromosome 9 supported the concept of a unique CGH) to a 2.8k genomic microarray. Due to a higher genomic disease entity that distinguishes PMBL from other B-cell non- resolution, we identified altered chromosomal regions in much Hodgkin’s lymphomas.10,11 Together with less specific gains on higher frequencies compared with standard CGH: for example, 2p15 and frequent mutations of the SOCS1 gene, a notable þ 9p24 (68%), þ 2p15 (51%), þ 7q22 (32%), þ 9q34 (32%), genomic similarity to classical Hodgkin’s lymphoma was þ 11q23 (18%), þ 12q (30%) and þ 18q21 (24%). Moreover, 12–14 previously unknown small interstitial chromosomal low copy observed. The genetic overlap of the two entities has been number alterations (for example, À6p21, À11q13.3) and a total supported recently by the results of mRNA gene expressions 15–17 of 19 DNA amplifications were identified by array-CGH. For 17 studies. A number of critical genes, such as JAK2, REL, chromosomal localizations (10 gains and 7 losses), which were cMYC, MAL, PDL, SOCS1 and CDKN2A, have been discussed altered in more than 10% of the analyzed cases, we delineated to be involved in the pathogenesis of PMBL. However, the minimal consensus regions based on genomic base pair cascade of oncogenic events leading to PMBL is still elusive. The positions. These regions and selected immunohistochemistries detailed delineation of aberrant chromosomal regions using point to candidate genes that are discussed in the context of 18–23 NF-jB transcription activation, human leukocyte antigen class high-resolution, microarray-based genomic techniques I/II defects, impaired apoptosis and Janus kinase/signal allowing candidate gene identification provides a suitable tool transducer and activator of transcription (JAK/STAT) activation. for clarifying the genetic causes of this disease. Our data confirm the genomic uniqueness of this tumor and provide physically mapped genomic regions of interest for focused candidate gene analysis. Leukemia (2007) 21, 2463–2469; doi:10.1038/sj.leu.2404919; Materials and methods published online 30 August 2007 Keywords: mediastinal B-cell lymphoma; array-CGH; genomic Thirty-seven fresh, frozen tumor samples of PMBL were alterations included in the study. Histopathological diagnosis was based on morphologic and immunohistochemical criteria according to the WHO classification of malignant lymphomas1 by two expert hematopathologists (PM and TFEB). A total of 22 male and 15 female patients were analyzed. The median age was 36 years Introduction (for more details see Table 1). For all cases, chromosomal comparative genomic hybridization (cCGH) data along with Primary mediastinal B-cell lymphoma (PMBL) is a distinct selected fluorescent in situ hybridization and cytogenetic data clinicopathological entity in the WHO classification of non- were published earlier.11,12 However, for most cases of the 1 Hodgkin’s lymphomas (ICD-O: 9679/3). Due to the location analyzed series, sufficient material for systematic mRNA and morphologic characteristics, PMBL is a unique lymphoma expression analysis was not available any more. subset that probably derives primarily from asteroid thymic B For the construction of a 2800 targets comprising genomic 2–5 cells. PMBL is positive for CD19, CD20, CD45, and often microarray, a genome-wide distributed core set of DNA CD30, CD79a, CD11c, CD23 and negative for CD10 and fragments was selected out of the previously published, 6 CD21. Highly characteristic is a lack of immunoglobulin acknowledged ‘GoldenPath’ clone set. The targets were set at surface antigen and a missing expression of major histocompati- intervals of approximately 2 Mb across the entire human 7,8 bility complex (MHC) class I and II molecules. More recently, genome (alternating clones of the GoldenPath; in total 1500 clones). This compilation was supplemented with a set of 600 Correspondence: Dr S Wessendorf, Klinik fu¨r Innere Medizin III, DNA fragments, which focuses critical regions of B-cell Zentrum fu¨r Innere Medizin der Universita¨t Ulm, Robert-Koch-Strasse neoplasms as detected by previous cytogenetic studies, for 8, Ulm 89081, Germany. example, 2p13, 9p24 or 11q13. Finally, 699 DNA targets E-mail: [email protected] 6These authors contributed equally to this study. specific for, for example, proto-oncogenes or tumor suppressor Received 30 May 2007; revised 25 July 2007; accepted 26 July 2007; genes with discussed pathogenetic relevance in B-cell non- published online 30 August 2007 Hodgkin’s lymphoma were added. The final chip contained Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2464 Table 1 Patient characteristics

Clinical feature No. of patients Patients evaluable for characteristic Percentage

Gender, male/female 22/15 37 59.5/40.5 Stage, XIII/IV 6 29 21 LDH elevated 20 27 74 WHO performance 4 27 15 Age, 460 years 3 37 8

Figure 1 Ideogram-based synopsis of all identiﬁed chromosomal imbalances for 37 analyzed tumor samples for chromosomes 1–22. Chromosomal gains (low copy number and ampliﬁcations) are indicated by green bars on the right of each ideogram, losses are indicated in red. Sites of known copy number polymorphisms (CNPs) are indicated as light blue dots on the right side of each ideogram.

2799 clones. A list of all clones with respective physical clone on the array was compared with the ‘Database of mapping information is given in Supplementary Table S1. Genomic Variants’ (http://www.projects.tcag.ca/variation/).26,27 DNA preparation of all bacterial artificial chromosome and P1- Clones and sites located in regions with known copy number derived artificial chromosome clones and subsequent sequential polymorphisms are marked in Figure 1 and in Supplementary degenerative oligonucleotide primer-mediated DNA amplifica- Table S1, enabling a conservative evaluation of the array data. tion was performed as described by Fiegler et al.,24,25 with three Alterations, which entirely covered CNPs, were excluded from minor modifications of the original protocol. A detailed descrip- the evaluation (for example, 5p11, 7q35, 14q32 and 15q11). For tion of the applied PCR procedures is listed in the supplement. the present study, the Y chromosome was excluded from the The PCR products were spotted onto Corning CMT-Gaps II glass analysis and the X chromosome was only evaluated with regard slides using an Omnigrid microarrayer (Gene Machines, San to DNA amplifications, since in all experiments, gender Carlos, CA, USA). DNA labeling, hybridization and data difference between tumor DNA and control DNA was used to acquisition were performed as described previously.21–23,25 monitor the integrity of each experiment. Fluorescence ratios were obtained using an Axon 4000 B dual laser scanner, computed as log2 ratios and normalized based on the linear clone set of 1500 clones (for details, see Schwaenen Results et al.21), excluding X and Y clones for normalization. There- upon, the ratios of two reverse color experiments were In 33 of the 37 (89%) tumor samples, a total of 199 averaged. The diagnostic cutoff level for each individual chromosomal imbalances were detected by array-CGH. The experiment was determined by calculating the mean and median number of aberrations was 5 (range ¼ 0–19). A total of subsequently using plus/minus three times the standard devia- 130 chromosomal gains and 69 losses were identified. In four tion of all clones from chromosomes 1–22. Additionally, those patients, no aberrations were found. Chromosomal gains in regions were scored as aberrant, whose signal ratios narrowly more than 10% of all cases (nX4) were detected on 9p24 missed the threshold criteria but were deviated unidirectionally (n ¼ 24), 2p15 (n ¼ 19), 7q22 (n ¼ 12), 12q (n ¼ 11), 11q13 in at least three adjacent targets. This was done to take into (n ¼ 7), 12p (n ¼ 6), 3q27 (n ¼ 5), 1q and 21q (both n ¼ 4). account aberrations of tumors with lower tumor cell fractions Genomic losses affected preferably 15q (n ¼ 6), 4q and 13q14 (for more details see Schwaenen et al.21). The cutoff value for (both n ¼ 5), and 8p, 6p21, 11q13 and 14q32 (all n ¼ 4). the detection of DNA amplifications was determined by Nineteen high-level DNA amplifications were identified by repetitive hybridizations of normal male vs normal female array-CGH localizing to (i) the previously described regions DNA. Due to the ratios for clones mapping to chromosome X in 2p15 (n ¼ 4); 9p24 (n ¼ 4) and X (n ¼ 3; see also annotations for these experiments, the cutoff ratio for amplifications was 1.6 (for Table 4 in which an overview of all DNA amplifications is details see Wessendorf et al.20). All data sets were carefully given) and (ii) to the novel locations 1p, 2q, 5q and 7p, reviewed for frequently affected chromosomal sites of physio- however, all of the latter represented solitary aberrations. For logic copy number polymorphisms (CNP). Therefore every aberrations which were present in at least 10% of all cases

Leukemia Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2465 (nX4), a total of 17 consensus regions were delineated (for regions listed together with their respective ﬂanking DNA chip details see Kohlhammer et al.22). A detailed description of targets, is given in Tables 2–4. Data for solitary aberrations are recurrent imbalances (nX2), including physical consensus not shown. A graphical synopsis of chromosomal imbalances for

Table 2 Recurrent (nX2) chromosomal losses in PMBL detected by array-CGH

Chr. n % Cell MCR (patients) MCR ﬂanked by Genes Candidate miRNA loss line clone start/stop MCR genes Genomic Mb Band

1p 3 K 3p 5 13 K 3q 3 4q 5 6p 4 11 28.8–32.7 (4) 6p21 RP11-373N24; RP1-93N13 95 MHC class II 6q 2 7p 2 8p 4 11 11.6–30.8 (4) 8p22 RP11-148O21; RP11-363L24 87 TNFRSF10 hsa-mir383, 320 9p 2 9q 3 10q 2 M 11q 4 11 67.6–72.0 (4) 11q13.3 RP11-901A4; RP11-31L22 28 FADD hsa-mir139 13q 5 13 48.0–55.5 (5) 13q14.1 RP11-165D7; RP11-516G5 26 hsa-mir16-1, 15a 14q 4 11 95.2–ter (4) 14q32 RP11-1070N10; F 61a 47 annotationsb 15q 6 16 K 0–31.9 (5) 15q11 F; RP11-3D4 33 hsa-mir211 16q 3 17p 5 13 0–1.8 (4) 17p13 F; RP11-4F24 16 ABR, TUSC5 hsa-mir22 Abbreviations: CGH, comparative genomic hybridization; MCR, minimal common regions; MHC, major histocompatibility complex; PMBL, primary mediastinal B-cell lymphoma. For deletions which were detected in at least 10% of all cases (nX4), percentages and minimal consensus regions (MCR) together with their ﬂanking clones are listed. The number in parenthesis indicates the number of tumor samples forming up the MCR. Indicated is the number of known genes (OMIM annotated) within the MCR, only such genes are listed which are cited in the text (K ¼ Karpas 1106P; M ¼ Med-B1, cell line positive for the respective aberration). aIgVH genes were not counted individually. bFor a list of covered miRNAs on 14q32, see Supplementary Table S2.

Table 3 Recurrent (nX2) chromosomal gains (including DNA ampliﬁcations in square brackets) in PMBL detected by array-CGH

Chr. n % Cell line MCR (patients) MCR ﬂanked by Genes Candidate miRNA gain clone start/stop MCR genes Genomic Mb Band

1p 3 [1] K 1q 4 11 M 2p 19 [4] 51 K 59.2–60.9 (19) 2p15 RP11-260K8; RP11-498O5 1 BCL11A 2q 4 [1] 11 216.5–226.7 (4) 2q24 RP11-489P15; RP11-86O17 52 IL8RA/B hsa-mir215,194-1 3q 5 13 188.9–189.9 (4) 3q27 RP11-519M3; RP11-259F24 4 BCL6 hsa-mir28 5p 2 5q 2 6p 3 7p 2 7q 12 [2] 32 102.9–103.9 (11) 7q22 RP4-672O11; RP11-612L3 3 8p 2 8q 5 13 K 128.5–ter (4) 8q24 RP11-382A18; F 61 MYC hsa-mir30b, 30d, 151 9p 25 [4] 68 M,K 0–6.6 (25) 9p24 F; RP11-106A1 22 JAK2, PDL hsa-mir101-2 9q 12 32 M 126.4–132.6 (11) 9q34 RP11-205K6; RP11-295G24 47 11p 3 11q 7 18 108.6–113.7 (5) 11q23 RP11-402K1; RP11-212D19 25 hsa-mir34b, 34c 12p 6 16 Whole arm 12q 11 30 K 58.5–67.5 (9) 12q14 RP11-813P10; RP11-611O2 30 MDM2 hsa-let7i 15q 2 16q 2 18p 3 18q 9 24 K 54.7–64.9 18q21 RP11-108P20; RP11-342G3 23 MALT1, BCL2 19p 2 19q 2 21q 4 11 M Abbreviations: CGH, comparative genomic hybridization; MCR, minimal common regions; PMBL, primary mediastinal B-cell lymphoma. In analogy to Table 2, for events which were detected in at least 10% of all cases (nX4), percentages and consensus regions together with their ﬂanking clones are listed. The number in parenthesis indicates the number of tumor samples forming up the MCR. Indicated is the number of known genes (OMIM annotated) within the MCR, only such genes are listed which are cited in the text (K ¼ Karpas 1106P; M ¼ Med-B1, cell line positive for the respective aberration).

Leukemia Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2466 Table 4 Separate listing of high-level DNA ampliﬁcations (multiple events on one chromosome arm in parenthesis) in PMBL detected by array- CGH

Chr. n % Amplicon Amplicon ﬂanked by Genes Candidate miRNA amp clone start/stop amp genes Genomic Mb Band

1p 1 (2) 1p22.3 RP11-415P10; RP11-131L23 9 BCL10 91.9–92.8 1p22.1 RP11-324C23; RP11-330C7 5 TGFBR3 2p 4 11 59.2–62.6 2p15 RP11-260K8; RP11-52F10 7 BCL11A, REL 2q 1 216.5–226.8 2q24 RP11-489P15; RP11-86O17 44 IL8RA/B hsa-mir26b, 375, 153-1, 149 5p 1 25.3–49.9 5p15 RP11-192H6; RP11-269M20 44 7q 2 (4) 91.9–99.7 (2) 7q21 RP5-1099C19; RP11-44M6 37 CDK6 hsa-mir489, 25, 93, 106b 9p 4 11 5.0–6.6 9p24 RP11-509D8; RP11-106A1 JAK2, PDL2 Xpa 1 48.3 Xp11 RP11-8A2 Xqa 2 109.4 Xq22 RP4-557A17 153.5 Xq28 RP11-196H18 Abbreviations: CGH, comparative genomic hybridization; PMBL, primary mediastinal B-cell lymphoma. For the recurrent amplicons on 2p, 7q and 9p percentages and consensus regions and their flanking clones, for singular events only flanking clones are listed. aAlthough chromosome X was not included in the numerical analysis, for high-level amplifications, the localizations of the amplicon peaks in the cases with known amplicons (chromosomal CGH) was possible due to their circumscript high amplitudes (42.5). No amplicons were found in the cell lines Med-B1 and Karpas1106P.

all 37 cases according to chromosomal arm is given in Figure 1. Constitutive activation of NF-kB The genomic profiles of the tumor cell lines Med-B1 and A constitutive activation of this nuclear transcription factor in Karpas1106P were congruent to previously published re- PMBL has been demonstrated recently.29 In this context, for sults11,28 (data not shown). Their characteristic imbalances in 2p15, a markedly higher frequency of genomic gains/amplifica- comparison to aberrations of the primary tumor samples tions was identified using array-CGH when compared to cCGH including the parental tumor of Med-B1 are listed in Tables 2–4. (51 vs 20%). This is in line with recent data obtained by fluorescent in situ hybridization analysis demonstrating the presence of 2p15 gains up to 75%.30 In our investigation, we observed a continuous transition between high-level amplifica- Discussion tions and single copy gains. In all cases carrying amplifications (n ¼ 4), both REL and BCL11A were affected. For cases with Applying array-CGH, we aimed to identify previously unknown lower copy number gains BCL11A was always covered, whereas chromosomal imbalances and to further characterize those REL was not part of the minimal consensus region. To validate regions that have been described previously. Using the 2.8k further these findings, we compared the genomic status with chip, we found approximately three times as much genomic immunohistochemical stainings for BCL11A and REL, which losses in comparison to previous cCGH analysis (69 vs 25).10,11 have been analyzed previously by Weniger et al.30,31 Complete The higher frequency of these losses was caused by the presence data sets were available for 16/37 cases. In 10 of these 16 cases, of small-sized (o10 Mb) intrachromosomal deletions (for genomic gains at 2p15 were observed using array-CGH. In nine example, 6p21.3, 3p23–p25, 11q13.3) that were not detectable of these samples, a concomitant elevated (X30%) nuclear by the cCGH approach. The total number of genomic gains protein accumulation for REL and/or BCL11A were detected. In corresponded well to cCGH (130 vs 135). Imbalances affecting two cases with no aberrations on the 2p15 locus, immunohis- large sections (410 Mb) or the entire arm of chromosomes 1–15 tochemistry revealed a missing or only slight detection of the were detected by both techniques. Interestingly, for chromo- respective proteins and in the remaining four cases, protein somes 16 and 22, notably less chromosomal gains were found detection of 430% was found despite a diploid genomic status. by array-CGH (6 vs 18). In our hands, these discrepancies most Confirming reports by Weniger et al., there was no straight likely reflect a higher reliability of array-CGH for these correlation of the genomic status herein measured by array- chromosomes. This technique scores the signal ratios of solitary CGH and the relative mRNA transcript levels of BCL11A and DNA fragments that serve as an internal control for the REL (real-time quantitative PCR). aberration scoring of respective up- and downstream-located As a second putative chromosomal region involved in NF-kB targets. Thus, by separation of defined chromatin segments, pathway activation, we observed a DNA amplification/gain, array-CGH overcomes a diagnostic limitation of conventional affecting chromosomal band 1p22 covering the BCL10 gene. CGH, especially for those critical chromosomes on which This protein forms a caspase recruitment domain, and has been repeatedly false-positive cCGH signals have been observed in shown to activate NF-kB. BCL10 is reported to interact with the past. Moreover, taking advantage of the array’s resolution, other caspase recruitment domains containing proteins includ- an elevated number of distinct high-level DNA amplifications ing caspase recruitment domain9, 10, 11 and 14, which are (19 vs 8) were identified. Novel amplifications occurred as believed to function as upstream regulators in NF-kB signaling. small-sized amplicons of regions most of which have not been Moreover, BCL10 forms a complex with MALT1, which leads to discussed in lymphoma pathogenesis previously (see Table 4). a synergistic activation of NF-kB.32 In one of eight cases with For 17 genomic localizations, we were able to delineate MALT1 affecting gains on 18q21, a coincidental 1p22 gain was minimal common regions. Affected regions and genes will be found. We therefore propose to further consider these chromo- outlined in the context of molecular pathways with suspected somal regions in the analysis of the genetic causes of constitutive pathogenetic relevance. NF-kB activation in PMBL.

Leukemia Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2467 Defective surface expression of major histocompatibility investigations of this region. Apart from numeric 9p alterations, complex molecules coinciding with JAK2 mRNA overexpression, delayed degrada- Losses of the MHC class I/II genes on 6p are among the most tion of phosphorylated JAK2 caused by mutations or deletions of frequent deletions in aggressive B-cell non-Hodgkin’s lympho- the SOCS1 gene has recently been suggested to constitute an ma.33 Using array-CGH, in four cases, a distinct 3.9 Mb deletion alternative mechanism promoting cell proliferation in PMBL.35 on 6p21 was identified, which covers the majority of genes Additionally, homozygous deletions of approximately 0.7 Mb forming up the MHC (human leukocyte antigen class I and/or II). covering SOCS1 and PIG7/LITAF have been described in the Interestingly, a recent investigation in PMBL has shown a worse Karpas1106P cell line using array-CGH.36,37 In our array outcome in patients where MHC class II genes were down- analysis, interstitial deletions on the chromosomal band regulated on mRNA and protein expression level. However, a 16q13.13 have not been identified; however, the critical region deletion of the respective genomic region was not suspected to was not directly represented on the array (flanking DNA targets be causal of these gene dysregulations.34 In contrast to other B- were localized approximately 1 Mb distance up- and 0.7 Mb cell non-Hodgkin’s lymphoma entities expressing MHC class I/II downstream). surface molecules, alterations of these molecules might provide PMBL cells with an additional growth advantage by escape from Others the immune system, a mechanism which has been extensively For a number of previously analyzed regions, namely 8q24, studied in various solid tumor entities and cell culture 13,14 12q13, 13q14 and 18q21, we were able to suggest more models. Summarizing these data, it still remains controver- precisely genomic extensions on the basis of their respective sial whether MCH II gene deregulation in PMBL is attributable to flanking DNA fragments. However, these regions still cover genomic DNA copy number changes. Our data warrant further large chromatin segments (7–19 Mb) which are mostly in genomic analysis of 6p21 locus alterations in PMBL using, for accordance with previous findings using CGH and fluorescent example, oligonucleotide single nucleotide polymorphism array in situ hybridization. approaches in combination with gene expression data, as well In contrast, on 2q24, 3q27, 7q22 and 17p13 novel small- as analysis of molecules associated with function of MHC class sized aberrant regions were identified, each covering a very I/II such as b2-microglobulin and the invariant chain (li). limited number of coding sequences. A solitary DNA amplification has been identified on 2q24. Here, a consensus region of 10 Mb in size with co-amplification of the IL8RA/B genes has Impaired apoptosis by FAS-associated via death domain been detected in three additional cases. For these receptors, an dysregulation involvement in chemokine-mediated tumor cell growth has In eight cases, deletions on 8p or 11q13 affected genes involved in been reported. Moreover, this receptor causes an interleukin-8- FAS-associated via death domain (FADD)-mediated apoptosis. mediated angiogenetic effect. On 3q27, we delineated a 1 Mb Four cases had a loss of the 8p arm covering TNRSF that encodes consensus-gained region covering six genes including the a receptor which is activated by tumor necrosis factor-related transcriptional co-repressor BCL6, for which an important role apoptosis-inducing ligand and also transduces apoptosis signal via in the pathogenesis of DLBCL by suppression of TP53 gene FADD. Additionally, four cases harbored deletions on 11q13.3, expression has been described. For the remaining genes, no directly involving FADD, whose protein builds the death-inducing involvement in tumorigenesis has been described so far. Gains signaling complex together with procaspase 8. Deletions of this on the long arm of chromosome 7 belonged to the most frequent latter region have not been described in the context of aberrations (32%). A DNA amplification covering the CDK6 lymphomagenesis so far. Considering these data, a defective gene, which is important for cell-cycle G1 phase progression FADD-mediated apoptosis might add to an impaired physiologi- and G1/S transition, was identified proximally to a distinct cal balance for cell proliferation and apoptosis in PMBL. 1.0 Mb consensus region of low copy gains. Within this region, only three genes without involvement in malignancy have been annotated. Finally, a subtelomeric 1.8 Mb region was identified Altered T-cell response regulation and JAK/STAT on 17p13. This deletion contains the tumor suppressor gene pathway activation ABR, frequently deleted in medulloblastoma and TUSC5, Gain of the subtelomeric region of 9p was found to be the most identified in lung cancer cell lines. frequent aberration detected with array-CGH. In the present study, the precise extension of this hallmark region for PMBL was narrowed down to 6.6 Mb with an interspersed recurrent Conclusions high-level amplicon of approximately 1.6 Mb. This DNA amplicon covers 10 known genes and encompasses both, the Our data set gives detailed karyotypic features of PMBL and tyrosin kinase JAK2 and also the T-cell response regulators PDL1 genomically maps recurrent chromosomal consensus regions. and PDL2. A disease-specific expression dysregulation of 9p24- Due to the improved resolution of array-CGH, the frequency localized genes has been demonstrated in previous differential and the exact delineation of known aberrations, as well as new gene expression studies where PDL2 has proven to be the recurrent aberrations were performed more exactly than by strongest discriminator between PMBL and other diffuse large B- genomic screening methods in the past. Our findings of 15 cell lymphoma (DLBCL) entities. Our results confirm the recurrent intrachromosomal imbalances and DNA amplifica- involvement of these genes in a small commonly amplified tions point to genes and regions of genomic instability that ought region in PMBL. For the entire subtelomeric 6.6 Mb region on to be considered in molecular pathogenetic analysis in PMBL. 9p, a total of 22 gene annotations (OMIM) (plus 30 hypothetical genes or functional unknown proteins) have been identified. For eight of these, an involvement in transcriptional processes or Acknowledgements cell-cycle control (FOXD4, DMRT1, DMRT3, SMARCA2, GLIS, CDC37L1) or tumorigenesis (ANKRD15, UHRF2) has been This work was supported by the Deutsche Krebshilfe Grant 70- reported. Thus, these genes should also be considered for further 2840 Be3 and by the Deutsche Jose´ Carreras Stiftung Grant LR02/

Leukemia Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2468 19. We thankfully acknowledge Zoraja Keresman and Sandra Ruf 17 Savage KJ, Monti S, Kutok JL, Cattoretti G, Neuberg D, De Leval L for excellent technical assistance and Martin JS Dyer, MRC et al. The molecular signature of mediastinal large B-cell Toxicology Unit, Leicester University, UK for providing Kar- lymphoma differs from that of other diffuse large B-cell lymphomas pas1106P. PACs and BACs for this study were ordered from and shares features with classical Hodgkin lymphoma. Blood 2003; 102: 3871–3879. libraries provided by the Resource Centre/Primary Database of the 18 Solinas-Toldo S, Lampel S, Stilgenbauer S, Nickolenko J, Benner A, German Human Genome Project (Berlin, Germany). SW, CS and Dohner H et al. Matrix-based comparative genomic hybridization: PM are members of and supported by the German network project biochips to screen for genomic imbalances. Genes Chromosomes molecular mechanisms in malignant lymphoma (MMML). Cancer 1997; 20: 399–407. 19 Pinkel D, Segraves R, Sudar D, Clark S, Poole I, Kowbel D et al. High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet References 1998; 20: 207–211. 20 Wessendorf S, Schwaenen C, Kohlhammer H, Kienle D, Wrobel G, Barth TF et al. Hidden gene amplifications in aggressive 1 Jaffe ES, Harris NL, Stein H, Vardman JW (eds). Tumours of B-cell non-Hodgkin lymphomas detected by microarray-based Haematopoietic and Lymphoid Tissues. IARC Press: Lyon, 2001, comparative genomic hybridization. Oncogene 2003; 22: pp 174–176. 1425–1429. 2 Barth TF, Leithauser F, Joos S, Bentz M, Moller P. Mediastinal 21 Schwaenen C, Nessling M, Wessendorf S, Salvi T, Wrobel G, (thymic) large B-cell lymphoma: where do we stand? Lancet Oncol Radlwimmer B et al. Automated array-based genomic profiling in 2002; 3: 229–234. chronic lymphocytic leukemia: development of a clinical tool and 3 Moller P, Lammler B, Eberlein-Gonska M, Feichter GE, Hofmann discovery of recurrent genomic alterations. Proc Natl Acad Sci WJ, Schmitteckert H et al. Primary mediastinal clear cell USA 2004; 101: 1039–1044. lymphoma of B-cell type. Virchows Arch A Pathol Anat 22 Kohlhammer H, Schwaenen C, Wessendorf S, Holzmann K, Histopathol 1986; 409: 79–92. Kestler HA, Kienle D et al. Genomic DNA-chip hybridization in 4 Addis BJ, Isaacson PG. Large cell lymphoma of the mediastinum: a B-cell tumour of probable thymic origin. Histopathology 1986; 10: t(11;14)-positive mantle cell lymphomas shows a high frequency 379–390. of aberrations and allows a refined characterization of consensus 5 Hofmann WJ, Momburg F, Moller P. Thymic medullary cells regions. Blood 2004; 104: 795–801. expressing B lymphocyte antigens. Hum Pathol 1988; 19: 1280– 23 Holzmann K, Kohlhammer H, Schwaenen C, Wessendorf S, 1287. Kestler HA, Schwoerer A et al. Genomic DNA-chip hybridization 6 Moller P, Moldenhauer G, Momburg F, Lammler B, Eberlein- reveals a higher incidence of genomic amplifications in pancreatic Gonska M, Kiesel S et al. Mediastinal lymphoma of clear cell type cancer than conventional comparative genomic hybridization and is a tumor corresponding to terminal steps of B cell differentiation. leads to the identification of novel candidate genes. [published Blood 1987; 69: 1087–1095. erratum appears in Cancer Res 2004; 64: 6358]. Cancer Res 2004; 7 Kanavaros P, Gaulard P, Charlotte F, Martin N, Ducos C, Lebezu M 64: 4428–4433. et al. Discordant expression of immunoglobulin and its associated 24 Fiegler H, Carr P, Douglas EJ, Burford DC, Hunt S, Scott CE et al. molecule mb-1/CD79a is frequently found in mediastinal large B DNA microarrays for comparative genomic hybridization based on cell lymphomas. Am J Pathol 1995; 146: 735–741. DOP-PCR amplification of BAC and PAC clones. Genes Chromo- 8 Moller P, Lammler B, Herrmann B, Otto HF, Moldenhauer G, somes Cancer 2003; 36: 361–374. Momburg F. The primary mediastinal clear cell lymphoma of B- 25 Hummel M, Bentink S, Berger H, Klapper W, Wessendorf S, Barth cell type has variable defects in MHC antigen expression. TFE et al. A biologic definition of Burkitt’s lymphoma from Immunology 1986; 59: 411–417. transcriptional and genomic profiling. N Engl J Med 2006; 354: 9 Pileri SA, Gaidano G, Zinzani PL, Falini B, Gaulard P, Zucca E 2419–2430. et al. Primary mediastinal B-cell lymphoma: high frequency of 26 Sebat J, Lakshmi B, Troge J, Alexander J, Young J, Lundin P et al. BCL-6 mutations and consistent expression of the transcription Large-scale copy number polymorphism in the human genome. factors OCT-2, BOB.1, and PU.1 in the absence of immunoglo- Science 2004; 305: 525–528. bulins. Am J Pathol 2003; 162: 243–253. 27 Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, Qi Y 10 Joos S, Otano-Joos MI, Ziegler S, Bruderlein S, du Manoir S, Bentz et al. Detection of large-scale variation in the human genome. Nat M et al. Primary mediastinal (thymic) B-cell lymphoma is Genet 2004; 36: 949–951. characterized by gains of chromosomal material including 9p 28 Nacheva E, Dyer MJ, Metivier C, Jadayel D, Stranks G, Morilla R and amplification of the REL gene. Blood 1996; 87: 1571–1578. et al. B-cell non-Hodgkin’s lymphoma cell line (Karpas 1106) with 11 Bentz M, Barth TF, Bruderlein S, Bock D, Schwerer MJ, Baudis M complex translocation involving 18q21.3 but lacking BCL2 et al. Gain of chromosome arm 9p is characteristic of primary rearrangement and expression. Blood 1994; 84: 3422–3428. mediastinal B-cell lymphoma (MBL): comprehensive molecular 29 Feuerhake F, Kutok JL, Monti S, Chen W, Lacasce AS, Cattoretti G cytogenetic analysis and presentation of a novel MBL cell line. et al. NF{kappa}B activity, function, and target-gene signatures in Genes Chromosomes Cancer 2001; 30: 393–401. primary mediastinal large B-cell lymphoma and diffuse large B-cell 12 Joos S, Kupper M, Ohl S, von Bonin F, Mechtersheimer G, Bentz M lymphoma subtypes. Blood 2005; 106: 1392–1399. et al. Genomic imbalances including amplification of the tyrosine 30 Weniger MA, Gesk S, Ehrlich S, Martin-Subero JI, Dyer MJ, Siebert kinase gene JAK2 in CD30+ Hodgkin cells. Cancer Res 2000; 60: R et al. Gains of REL in primary mediastinal B-cell lymphoma 549–552. coincide with nuclear accumulation of REL protein. Genes 13 Moller P, Hammerling GJ. The role of surface HLA-A,B,C Chromosomes Cancer 2007; 46: 406–415. molecules in tumour immunity. Cancer Surv 1992; 13: 101–127. 31 Weniger MA, Pulford K, Gesk S, Ehrlich S, Banham AH, 14 Chang CC, Campoli M, Ferrone S. HLA class I defects in malignant Lyne L et al. Gains of the proto-oncogene BCL11A and lesions: what have we learned. Keio J Med 2003; 52: 220–229. nuclear accumulation of BCL11A(XL) protein are frequent 15 Rosenwald A, Wright G, Leroy K, Yu X, Gaulard P, Gascoyne RD in primary mediastinal B-cell lymphoma. Leukemia 2006; 20: et al. Molecular diagnosis of primary mediastinal B cell lymphoma 1880–1882. identifies a clinically favorable subgroup of diffuse large B cell 32 Nakagawa M, Hosokawa Y, Yonezumi M, Izumiyama K, Suzuki R, lymphoma related to Hodgkin lymphoma. J Exp Med 2003; 198: Tsuzuki S et al. MALT1 contains nuclear export signals and 851–862. regulates cytoplasmic localization of BCL10. Blood 2005; 106: 16 Calvo KR, Traverse-Glehen A, Pittaluga S, Jaffe ES. Molecular 4210–4216. profiling provides evidence of primary mediastinal large B-cell 33 Jordanova ES, Riemersma SA, Philippo K, Giphart-Gassler M, lymphoma as a distinct entity related to classic Hodgkin Schuuring E, Kluin PM. Hemizygous deletions in the HLA region lymphoma: implications for mediastinal gray zone lymphomas as account for loss of heterozygosity in the majority of diffuse large an intermediate form of B-cell lymphoma. Adv Anat Pathol 2004; B-cell lymphomas of the testis and the central nervous system. 11: 227–238. Genes Chromosomes Cancer 2002; 35: 38–48.

Leukemia Array-CGH in primary mediastinal B-cell lymphoma S Wessendorf et al 2469 34 Roberts RA, Wright G, Rosenwald AR, Jaramillo MA, Grogan TM, 36 Melzner I, Weniger MA, Bucur AJ, Bruderlein S, Dorsch K, Hasel C Miller TP et al. Loss of major histocompatibility class II gene and et al. Biallelic deletion within 16p13.13 including SOCS-1 in protein expression in primary mediastinal large B-cell lymphoma is Karpas1106P mediastinal B-cell lymphoma line is associated with highly coordinated and related to poor patient survival. Blood delayed degradation of JAK2 protein. Int J Cancer 2006; 118: 2006; 108: 311–318. 1941–1944. 35 Melzner I, Bucur AJ, Bruderlein S, Dorsch K, Hasel C, Barth TF 37 Mestre-Escorihuela C, Rubio-Moscardo F, Richter JA, Siebert R, et al. Biallelic mutation of SOCS-1 impairs JAK2 degradation and Climent J, Fresquet V et al. Homozygous deletions localize novel sustains phospho-JAK2 action in the MedB-1 mediastinal lympho- tumor suppressor genes in B-cell lymphomas. Blood 2007; 109: ma line. Blood 2005; 105: 2535–2542. 271–280.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu).

Leukemia