Golub, Todd R.

Abstract In an effort to identify novel genetic abnormalities in multiple myeloma (MM) we analyzed an initial set, as part of the Multiple Myeloma Research Consortium (MMRC) Genomics Initiative, of 94 MM patient samples on high-resolution array-based comparative genomic hybridization (aCGH) and gene expression profiling (GEP) technologies. The MMRC Genomics Initiative is a three-year research program spanning a full-spectrum of genomic technologies with the goal of analyzing 250 MM patient samples by 2009. Raw data from the Initiative are released in near real time through the MMRC Multiple Myeloma Genomics Portal. This initial aCGH dataset from the Agilent 244K platform represents the highest resolution gene copy number analysis with corresponding GEP released to date in MM. It has confirmed the cytogenetic definitions of MM with the data set being split between patients with hyperdiploid and non-hyperdiploid MM. Furthermore, the 244K platform has helped to refine minimal regions of change present on 1p, 1q, 6q, 8p, 13q, 14q, and 16q. A comprehensive screen of the dataset for genes residing in regions of homozygous deletion identified 105 genes. The most commonly affected gene was CDKN2C(p18) (8.3%) while 12 other regions were recurrently targeted. Several of these regions were recently implicated in the pathogenesis of MM but the large majority represents novel observations. The aCGH and GEP data were integrated using two different programs developed by the core institutions (The Broad Institute of Harvard & MIT and Translational Genomics Research Institute). First, Genomic Identification of Significant Targets in Cancer (GISTIC), which identifies highly significant regions of change and a limited set of candidate genes within these regions, was used to identify 285 genes associated with DNA content alterations. Second, Breakpoint Expression Correlation (BEC), which identifies dysregulated genes on either side of a copy number alteration, was used to identify 194 genes with expression changes associated with a breakpoint. This latter gene list includes all known IgH translocation target genes identified to date in myeloma. Overlapping the gene lists generated by GISTIC, BEC, and the homozygous deletion screen identified a number of genes previously implicated in MM such as CDKN2C(p18), CYLD, BIRC2/BIRC3, and two novel genes, FAM46C and RNF6. Ongoing data integration with RNAi and high content sequencing projects of the current Initiative holds great promise in furthering our understanding of MM. The MMRC Genomics Initiative and resulting efforts from the myeloma research community will increase our understanding of MM at the molecular level, help to identify new targets, and ultimately lead to the development of better, more effective therapies for myeloma.

Abstract The Multiple Myeloma Research Consortium (MMRC) Genomics Initiative was instigated to harness the power of multiple genomic approaches to further the understanding of multiple myeloma. To date, 137 samples from patients with newly diagnosed and relapsed myeloma (out of an estimated final total of 250 by the end of the year) have been subjected to expression profiling and array comparative genomic hybridization. To identify regions of recurrent copy number alteration with a high degree of confidence, we have used the Genomic Identification of Significant Targets In Cancer (GISTIC) algorithm, which detects such regions and assigns a probability to each. Application of GISTIC to the MMRC collection identified 14 significant regions of amplification and 15 significant regions of deletion. The algorithm further detects peaks of copy number change that contribute to each region’s significance. Of genes that were expressed in our dataset, a total of 64 across the genome were found to lie within the boundaries of significant amplification peaks and 30 were found to lie within significant deletion peaks. Given the likelihood that this gene collection is highly enriched for genes important in the pathogenesis of myeloma and potential therapeutic targets, it has been prioritized for further validation with the amplified genes submitted for arrayed RNAi and the deleted genes expedited for re-sequencing. In order to define a poor prognosis group, we developed and applied a model based on the gene expression dataset of Shaughnessy and colleagues. To identify pathways activated in poor prognosis disease, we applied Gene Set Enrichment Analysis with the Molecular Signatures Database. This demonstrated enrichment of multiple canonical pathways within the poor prognosis group, including those associated with proliferation, cell cycle progression and DNA repair. A search for recurrent copy number events associated with poor prognosis disease revealed that by far the most significant event, and the only one surviving correction for multiple hypothesis testing, was mono- or bi-allelic deletion of CDKN2C (p18). Interestingly, the poor prognosis group also demonstrated higher expression of p18 in those samples without bi-allelic deletion and of CDKN2A (p16). Deletion and increased expression of p18 in high proliferation index myeloma samples has previously been described. However, our analyses demonstrate the pre-eminence of p18 loss for defining poor prognosis disease compared to other recurrent copy number changes. Furthermore, it expands on work in other tumors demonstrating feedback mechanisms between cyclin dependent kinase inhibitor pathways, suggesting a more complex model of interplay between the pathways than previously described. In conclusion, the MMRC reference collection is proving an important tool in the fight to understand key genetic events in multiple myeloma and would ultimately be anticipated to contribute towards development of improved therapy for the disease.

Abstract The Multiple Myeloma Research Consortium (MMRC) Genomics Initiative is a three-year program to analyze tumor tissue from hundreds of multiple myeloma (MM) patients via gene expression profiling (GEP), comparative genomic hybridization (aCGH), and exon re-sequencing. In addition, RNAi knockdown of selected genes in MM tumor cell lines is being evaluated to identify potential new targets. All genomic data generated is scheduled for placement in an open-access Multiple Myeloma Genomics Portal pre-publication and in near real-time (www.broad.mit.edu/mmgp). Additionally, samples are also destined for drug validation and correlative science on clinical protocols as this study moves forward. This comprehensive project is spearheaded by the MMRC and conducted via collaboration with the Eli and Edythe L. Broad Institute of MIT and Harvard, the Translational Genomics Research Institute (TGen), Mayo Clinic Arizona, and The Dana-Farber Cancer Center. The study is supported by the collection from member institutions of the MMRC of bone marrow aspirates and matched peripheral blood samples from over 1000 patients. Specific genomic technologies that are currently being employed across this sample set include GEP using Affymetrix Human Genome U133A 2.0 Plus Arrays, and, in parallel, efforts to identify regions of genomic gain and loss are using Agilent Human Genome CGH arrays. In contrast to other large-scale genomic projects based on exon-sequencing of targeted gene sets, this project will be the first to perform genome-scale single molecule sequencing (SMS) of DNA from patient specimens. Results will be targeted against candidate classes of genes (e.g. kinases, phosphatases, known oncogenes and tumor suppressors), and genes from GEP or within candidate regions of copy gain or loss identified by the aCGH experiments. Mutations will be further validated in an independent set of patient specimens. Finally we will attempt to identify points of vulnerability of MM through systematic loss-of-function screens in myeloma cell lines using high-throughput RNA interference (using both shRNA and siRNA platforms). Importantly, data generated from this genomics initiative will ultimately be made public pre-publication through the established MMRC Multiple Myeloma Genomics Portal. Data from all aspects of this project (sample collection and analyte isolation, GEP, aCGH, SMS, RNAi and bioinformatics) will be described in this presentation. The power of this study is the comprehensive collection of gene expression, CGH, and genome sequencing on a single reference set of clinically annotated samples. The addition of RNAi screens makes this a very important and unique data resource, which we hope will help expedite the discovery of novel targeted agents for MM scientific community.

Abstract Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease with recognized variability in clinical outcome, genetic features, and cells of origin. To date, transcriptional profiling has been used to highlight similarities between DLBCL tumor cells and normal B-cell subtypes and associate genes and pathways with unfavorable outcome. Given the genetic heterogeneity in DLBCL, there are likely to be subsets of tumors with different pathogenetic mechanisms and possible treatment targets. To identify DLBCL subtypes that were sufficiently robust to be captured by multiple methods, we analyzed the profiles of 176 newly diagnosed DLBCLs using three different clustering algorithms (hierarchical clustering (HC), self-organizing maps (SOM), and probabilistic clustering (PC)), the top 5% of genes with the highest reproducibility across duplicate samples and largest variation across patient tumors, and a resampling-based method (consensus clustering) that automatically selects the most stable numbers of clusters with each algorithm. Three discrete subsets of DLBCLs -- “Oxidative Phosphorylation” (OxPhos), “B-cell Receptor/Proliferation” and “Host Response” (HR) -- were identified, characterized using gene set enrichment analysis and confirmed in an independent series of newly diagnosed DLBCLs with available array data. There was an association between cluster membership and examined genetic abnormalities in DLBCL. BCL2 translocations were more common in the OxPhos cluster whereas BCL6 translocations were more frequent in the BCR/proliferation cluster. Translocations of either type were uncommon in the HR cluster. Patients with HR DLBCLs were also significantly younger than those with OxPhos or BCR/proliferation tumors; HR patients also had a significantly higher incidence of splenic and BM involvement. The unique characteristics of HR tumors - fewer known genetic abnormalities and prominent host immune and inflammatory cell transcripts -prompted us to assess host immune cells in study tumors using morphologic and immunohistochemical approaches. HR DLBCLs contained significantly higher numbers of morphologically distinct CD2+/CD3+ tumor-infiltrating lymphocytes and interdigitating S100+/GILT+/CD1a-/CD123- dendritic cells. The HR cluster shared features of histologically defined T-cell/histiocyte-richBCL, including fewer known genetic abnormalities, younger age at presentation and frequent splenic and bone marrow involvement. The current study identifies tumor microenvironment and host inflammatory response as defining features in DLBCL, provides insights into the nature of the host immune response in a major DLBCL subtype and suggests rational treatment targets in newly identified tumor groups.

Abstract Primary mediastinal large B-cell lymphoma (MLBCL) is a clinically distinct entity that typically presents as localized, sclerotic disease in young, female patients. We previously characterized the transcriptional profiles of MLBCLs and identified important shared features with a clinically related disorder, classical Hodgkin lymphoma (cHL) (Blood 102:3871, 2003). Given the documented role of the NFkB survival pathway in Hodgkin Reed-Sternberg cells, we previously assessed NFkB activation in MLBCL by determining the subcellular location of the c-REL subunit of the NFkB heterodimer with a 2-color immunofluoresence assay. In a small pilot MLBCL series, c-REL was localized to the nucleus in the majority of examined cases, consistent with NFkB activation. In the current study, we evaluated c-REL subcellular localization in an additional series of MLBCLs and DLBCLs using a broadly applicable immunoperoxidase method. 100% of MLBCLs exhibited nuclear c-REL staining whereas DLBCL c-REL subcellular localization was more variable. Thereafter, we analyzed the transcription profiles of the 34 MLBCLs and 176 DLBCLs for coordinate expression of NFkB target genes, using literature-curated NFkB target gene lists from three independent sources and gene set enrichment analysis (GSEA). MLBCL signatures exhibited significant enrichment of 2 of the 3 NFkB target gene sets. In addition, 32 NFkB target genes from the combined set were significantly more abundant in MLBCLs than DLBCLs (> 30% more abundant and > 99th percentile in permutation analysis). Similar results were obtained in an independent series of MLBCLs and DLBCLs with available gene expression profiles (J. Exp. Med. 198:851, 2003). To assess the role of c-REL amplification in NFkB activation in our lymphoma series, we compared c-REL amplification, c-REL subcellular localization and coordinate expression of the identified NFkB target genes and classified the DLBCLs according to putative cell of origin. The majority of c-REL amplifications (67%) were found in DLBCLs of germinal center (GC) subtype, consistent with the observation that c-REL is part of the described GC signature. However, most (71%) of the examined GC DLBCLs had cytoplasmic c-REL expression and the GC DLBCLs did not have increased expression of NFkB target genes. Taken together with the MLBCL analyses, these studies indicate that: 1) NFkB is consistently activated in MLBCL; 2) c-REL amplification is not closely associated with NFkB activation in large cell lymphomas (LCLs); and 3) NFkB activation in LCL subtypes does not require amplification of the c-REL locus.

Abstract Abstract 259 Despite recent progress, our understanding of the genomic basis of diffuse large B cell lymphoma (DLBCL) is incomplete. We have performed massively parallel whole exome sequencing with deep coverage of 57 primary tumor samples from patients with DLBCL and compared them to matched normal tissue to gain insight into the genetic defects leading to DLBCL. We define genes that are significantly enriched for somatic mutations, and thus, may be drivers of the disease. In addition to identifying genes that are well established to be functionally relevant in DLBCL, such as MyD88, CARD11, EZH2, and BCL2, we also discovered new genes that are significantly mutated that have not been described to play a role in DLBCL previously, such as MLL2, 3 and4 genes and PCLO. One of the challenges for the understanding of DLBCL biology is to determine the relative contribution of different genes with somatic mutations to the pathogenesis of the disease in an individual patient. Identification of functionally relevant mutated genes that only occur in a small number of patients based on prevalence alone may require thousands of patient samples. We therefore devise several strategies to discover genes with a high likelihood of functionally relevant mutations independent of their prevalence. Genes that do not co-occur with other mutated genes in the same patients, i.e. are mutually exclusive, are more likely to represent strong drivers, and to occur early in the development of DLBCL. We determined that TP53, BCL2, PIM1, and CD79B fall into this group. We apply pathway-focused analysis to identify mutated genes with low prevalence that are highly likely to have a functional impact. For instance, we identified G13D mutations in KRAS in two patients, and mutations in BRAF, NOTCH1, SOCS1, and PTEN in other patients that were previously described in other malignancies. These mutated genes are identified as part of significantly enriched gene sets, including the B cell receptor signaling/NFkB pathway, JAK/STAT, and antigen processing gene sets. In addition to these pathways that are important for physiological B cell activation, we find that many histone methyltransferases are enriched as a geneset in addition to MLL2. We identified a number of hotspots that are subject to somatic mutation. Our data imply that many of these hotspots may be targets of activation-induced cytidine deaminase (AID), which is usually involved in somatic hypermutation, suggesting a possible mechanism of how somatic mutation in DLBCL is mediated. In summary, our data define novel genetic defects in DLBCL by whole exome sequencing of 57 patients. Our data suggest that driver mutations may not be randomly distributed across the genome, but may occur in a directed fashion. We identified new genes not previously implicated in DLBCL, and have developed strategies to prioritize mutations that are most likely to be drivers of the disease. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 1534 + equal contributions Diffuse large B-cell lymphoma (DLBCL) is a clinically and biologically heterogeneous disease with a high proliferation rate and infrequent somatic mutations of TP53 and genes encoding cell cycle pathway components. Despite recent advances in the molecular understanding of DLBCL pathogenesis, clinical models such as the International Prognostic Index (IPI) are still used to identify high-risk patients. By integrating high-resolution high-density SNP array data with transcriptional profiles and performing pathway analysis in 180 primary DLBCLs, we identified a comprehensive set of copy number alterations (CNAs) that decreased p53 activity and perturbed cell cycle regulation. Primary DLBCLs with single copy loss of 17p13.1 (TP53/RPL26/KDM6B) often had CNAs of an additional p53 modifier – 9p21.3 (CDKN2A/ARF), 19q13.42 (BCL2L12), 12q15 (MDM2) or 1q23.3 (MDM4/RFWD2). CNAs of the respective p53 modifiers, CDKN2A (ARF, 9p21.3), MDM2 (12q15) and MDM4/RFWD2 (1q23.3), occurred in largely separate groups of tumors. DLBCLs with CNAs of p53 pathway members frequently exhibited concurrent alterations of additional cell cycle components such as CCND3 (6p21.32), CDK6 (7q22.1), CDK2/CDK4 (12q15) and/or RB1 (13q14.2) or RBL2 (16q12.2). When the primary DLBCLs were clustered in the space of CNAs that perturb p53 pathway and cell cycle components, 66% of tumors had multiple alterations (“complex” pattern) whereas the remaining 34% of tumors lacked these lesions (“clean” signature). Tumors with “complex” alterations of p53 pathway and cell cycle components also had more total CNAs and more frequent TP53 mutations, highlighting the association between p53 deficiency, cell cycle deregulation and increased genomic instability in DLBCL. The patterns of “complex” vs. “clean” CNAs of p53 pathway and cell cycle components and the association between the “complex” signature and total CNAs were confirmed in an independent series of primary DLBCLs. To further characterize DLBCLs with “complex” vs. “clean” CNA patterns, we performed gene set enrichment analyses with publicly available series of p53 target genes and a RB-deficiency gene set which included multiple E2F targets. The p53 target transcripts were significantly less abundant in “complex” DLBCLs, directly linking their genetic signature of p53 deficiency with decreased p53 activity. Furthermore, the RB-deficiency gene set was significantly enriched in “complex” DLBCLs suggesting that these tumors had increased E2F-mediated cell cycle progression. Consistent with these observations, DLBCLs with “complex” CNA patterns also had significantly higher proliferation indices as determined by Ki67 immunostaining. We next assessed the prognostic significance of the “complex” CNA pattern in the subset of patients who were treated with R-CHOP (rituxan, cyclophosphamide, adriamycin, oncovin, prednisone) and had long-term follow-up. R-CHOP treated patients with “complex” CNA patterns had a 5-year overall survival of only 62%; in contrast, those with “clean” CNA signatures had an 100% overall survival (p =.001). The association between CN complexity and outcome was independent of transcriptionally defined categories and additive to the clinical IPI risk model. Taken together, these data provide a structural basis for deregulated cell cycle, increased cellular proliferation and unfavorable outcome in DLBCL and suggest targeted treatment strategies. Disclosures: No relevant conflicts of interest to declare.

Abstract Diffuse large B-cell lymphoma (DLBCL) is a genetically heterogeneous disease characterized by multiple low-frequency alterations including somatic mutations, copy number alterations (CNAs) and chromosomal rearrangements. We sought to identify previously unrecognized low-frequency genetic events, integrate recurrent alterations into comprehensive signatures and associate these signatures with clinical parameters. For these reasons, our multi-institutional international group assembled a cohort of 304 primary DLBCLs from newly diagnosed patients, 87% of whom were uniformly treated with state-of-the-art therapy (rituximab-containing CHOP regimen) and had long term followup. Tumors were subjected to whole exome sequencing with an extended bait set that included custom probes designed to capture recurrent chromosomal rearrangements. In this cohort, 47% of samples had available transcriptional profiling and assignment to associated disease subtypes. Analytical pipelines developed at the Broad Institute were used to detect mutations (MuTect), CNAs (Recapseq+Allelic Capseq) and chromosomal rearrangements (dRanger+Breakpointer) and assess clonality (Absolute). To analyze formalin-fixed paraffin-embedded tumors without paired normals we developed a method which utilized 8334 unrelated normal samples to stringently filter recurrent germline events and artifacts. Significant mutational drivers were identified using the MutSig2CV algorithm and recurrent CNAs were assessed with GISTIC2.0. In addition, we utilized a recently developed algorithm, CLUMPS2, to prioritize somatic mutations which cluster in 3-dimensional protein structure. With this approach, we identified > 90 recurrently mutated genes, 34 focal amplifications and 41 focal deletions, 20 arm-level events and > 200 chromosomal rearrangements in the DLBCL series. Of note, 33% of the mutational drivers were also perturbed by chromosomal rearrangements or CNAs, underscoring the importance of a comprehensive genetic analysis. In the large DLBCL series, we identified several previously unrecognized but potentially targetable alterations including mutations in NOTCH2 (8%) and TET2 (5%). The majority of identified chromosomal rearrangements involved translocations of potent regulatory regions to intact gene coding sequences. The most frequently rearrangements involved Ig regulatory elements which were translocated to BCL2, MYC, BCL6 and several additional genes with known roles in germinal center B-cell biology. After identifying recurrent somatic mutations, CNAs and chromosomal rearrangements, we performed hierarchical clustering and identified subsets of DLBCLs with comprehensive signatures comprised of specific alterations. A large subset of tumors shared recurrent alterations previously associated with follicular lymphoma including mutations of chromatin modifiers such as CREBBP, MLL2, and EZH2 in association with alterations of TNFRSF14 and GNA13 and translocations of BCL2. This cluster was enriched in GCB-type DLBCLs and contained a subset with select genetic alterations associated with an unfavorable outcome. An additional cohort of tumors was characterized by alterations perturbing B-cell differentiation including recurrent BCL6 translocations or alterations of PRDM1. A subset of these DLBCLs had alterations of NOTCH2 and additional pathway components or mutations of MYD88 in association with TNFAIP3, CD70 and EBF1, a master regulator of B-cell differentiation. An additional group of DLBCLs exhibited frequent MYD88 mutations in association with alterations of CD79B, PIM1, TBL1XR1 and ETV6 and BCL2 copy gain; these tumors were highly enriched for ABC-type DLBCLs. This coordinate signature and additional alterations of p53 pathway components were associated with outcome. We explored bases for the identified genetic alterations in DLBCL by performing an in silico mutational signature analysis. The most frequent mutational signatures were those of spontaneous deamination (aging) and AID with rare cases of microsatellite instability. We also assessed the clonality of identified genetic features to define cancer cell fraction and establish the timing of specific genetic events. The comprehensive genetic signatures of clinically annotated DLBCLs provide new insights regarding approaches to targeted therapy. Disclosures Link: Kite Pharma: Research Funding; Genentech: Consultancy, Research Funding. Rodig:Perkin Elmer: Membership on an entity's Board of Directors or advisory committees; BMS: Research Funding. Pfreundschuh:Boehringer Ingelheim, Celegene, Roche, Spectrum: Other: Advisory board; Roche: Honoraria; Amgen, Roche, Spectrum: Research Funding. Shipp:Gilead: Consultancy; Sanofi: Research Funding; BMS: Membership on an entity's Board of Directors or advisory committees, Research Funding; Merck: Membership on an entity's Board of Directors or advisory committees; Bayer: Membership on an entity's Board of Directors or advisory committees, Research Funding.

Abstract Abstract 832 We hypothesized that new therapeutic targets for multiple myeloma (MM) could be discovered through the integrative computational analysis of genomic data. Accordingly, we generated gene expression profiling and copy number data on 250 clinically-annotated MM patient samples. Utilizing an outlier statistical approach, we identified HOXA9 as the top candidate gene for further investigation. HOXA9 expression was particularly high in patients lacking canonical MM chromosomal translocations, and allele-specific expression analysis suggested that this overexpression was mono-allelic. Indeed, focal copy number amplifications at the HOXA locus were observed in some patients. Outlier HOXA9 expression was further validated in both a collection of 52 MM cell lines and 414 primary patient samples previously described. To further verify the aberrant expression of HOXA9 in MM, we performed quantitative RT-PCR, which confirmed expression in all MM patients and cell lines tested, with high-level expression in a subset. To further investigate the mechanism of aberrant HOXA9 expression, we interrogated the pattern of histone modification at the HOXA locus because HOXA gene expression is particularly regulated by such chromatin marks. Accordingly, immunoprecipitation studies showed an aberrantly low level of histone 3 lysine 27 trimethylation marks (H3K27me3) at the HOXA9 locus. H3K27me3 modification is normally associated with silencing of HOXA9 in normal B-cell development. As such, it appears likely that the aberrant expression of HOXA9 in MM is due at least in part to defects in histone modification at this locus. To determine the functional consequences of HOXA9 expression in MM, we performed RNAi-mediated knock-down experiments in MM cell lines. Seven independent HOXA9 shRNAs that diminished HOXA9 expression resulted in growth inhibition of 12/14 MM cell lines tested. Taken together, these experiments indicate that HOXA9 is essential for survival of MM cells, and that the mechanism of HOXA9 expression relates to aberrant histone modification at the HOXA9 locus. The data thus suggest that HOXA9 is an attractive new therapeutic target for MM. Disclosures: No relevant conflicts of interest to declare.

To gain insight into the genomic basis of diffuse large B-cell lymphoma (DLBCL), we performed massively parallel whole-exome sequencing of 55 primary tumor samples from patients with DLBCL and matched normal tissue. We identified recurrent mutations in genes that are well known to be functionally relevant in DLBCL, including MYD88 , CARD11 , EZH2 , and CREBBP . We also identified somatic mutations in genes for which a functional role in DLBCL has not been previously suspected. These genes include MEF2B , MLL2 , BTG1 , GNA13 , ACTB , P2RY8 , PCLO , and TNFRSF14 . Further, we show that BCL2 mutations commonly occur in patients with BCL2 / IgH rearrangements as a result of somatic hypermutation normally occurring at the IgH locus. The BCL2 point mutations are primarily synonymous, and likely caused by activation-induced cytidine deaminase–mediated somatic hypermutation, as shown by comprehensive analysis of enrichment of mutations in WRCY target motifs. Those nonsynonymous mutations that are observed tend to be found outside of the functionally important BH domains of the protein, suggesting that strong negative selection against BCL2 loss-of-function mutations is at play. Last, by using an algorithm designed to identify likely functionally relevant but infrequent mutations, we identify KRAS , BRAF , and NOTCH1 as likely drivers of DLBCL pathogenesis in some patients. Our data provide an unbiased view of the landscape of mutations in DLBCL, and this in turn may point toward new therapeutic strategies for the disease.