Search

Programmed death ligands (PDLs) are immune-regulatory molecules that are frequently affected by chromosomal alterations in B-cell lymphomas. Although PDL copy-number variations are well characterized, a detailed and comprehensive analysis of structural rearrangements (SRs) and associated phenotypic consequences is largely lacking. Here, we used oligonucleotide capture sequencing of 67 formalin-fixed paraffin-embedded tissues derived from primary B-cell lymphomas and 1 cell line to detect and characterize, at base-pair resolution, SRs of the PDL locus (9p24.1; harboring PDL1/CD274 and PDL2/PDCD1LG2). We describe 36 novel PDL SRs, including 17 intrachromosomal events (inversions, duplications, deletions) and 19 translocations involving BZRAP-AS1, CD44, GET4, IL4R, KIAA0226L, MID1, RCC1, PTPN1 and segments of the immunoglobulin loci. Moreover, analysis of the precise chromosomal breakpoints reveals 2 distinct cluster breakpoint regions (CBRs) within either CD274 (CBR1) or PDCD1LG2 (CBR2). To determine the phenotypic consequences of these SRs, we performed immunohistochemistry for CD274 and PDCD1LG2 on primary pretreatment biopsies and found that PDL SRs are significantly associated with PDL protein expression. Finally, stable ectopic expression of wild-type PDCD1LG2 and the PDCD1LG2-IGHV7-81 fusion showed, in coculture, significantly reduced T-cell activation. Taken together, our data demonstrate the complementary utility of fluorescence in situ hybridization and capture sequencing approaches and provide a classification scheme for PDL SRs with implications for future studies using PDL immune-checkpoint inhibitors in B-cell lymphomas.

The clustered homeobox proteins play crucial roles in development, hematopoiesis, and leukemia, yet the targets they regulate and their mechanisms of action are poorly understood. Here, we identified the binding sites for Hoxa9 and the Hox cofactor Meis1 on a genome-wide level and profiled their associated epigenetic modifications and transcriptional targets. Hoxa9 and the Hox cofactor Meis1 cobind at hundreds of highly evolutionarily conserved sites, most of which are distant from transcription start sites. These sites show high levels of histone H3K4 monomethylation and CBP/P300 binding characteristic of enhancers. Furthermore, a subset of these sites shows enhancer activity in transient transfection assays. Many Hoxa9 and Meis1 binding sites are also bound by PU.1 and other lineage-restricted transcription factors previously implicated in establishment of myeloid enhancers. Conditional Hoxa9 activation is associated with CBP/P300 recruitment, histone acetylation, and transcriptional activation of a network of proto-oncogenes, including Erg, Flt3, Lmo2, Myb, and Sox4. Collectively, this work suggests that Hoxa9 regulates transcription by interacting with enhancers of genes important for hematopoiesis and leukemia.

Hematopoietic stem cells (HSCs) are generally defined by their dual properties of pluripotency and extensive self-renewal capacity. However, a lack of experimental clarity as to what constitutes extensive self-renewal capacity coupled with an absence of methods to prospectively isolate long-term repopulating cells with defined self-renewal activities has made it difficult to identify the essential components of the self-renewal machinery and investigate their regulation. We now show that cells capable of repopulating irradiated congenic hosts for 4 months and producing clones of cells that can be serially transplanted are selectively and highly enriched in the CD150+ subset of the EPCR+CD48−CD45+ fraction of mouse fetal liver and adult bone marrow cells. In contrast, cells that repopulate primary hosts for the same period but show more limited self-renewal activity are enriched in the CD150− subset. Comparative transcriptome analyses of these 2 subsets with each other and with HSCs whose self-renewal activity has been rapidly extinguished in vitro revealed 3 new genes (VWF, Rhob, Pld3) whose elevated expression is a consistent and selective feature of the long-term repopulating cells with durable self-renewal capacity. These findings establish the identity of a phenotypically and molecularly distinct class of pluripotent hematopoietic cells with lifelong self-renewal capacity.

Introduction: Malignant lymphomas account for 5% of all newly diagnosed cancer cases per year and affect patients of all ages. Genomic rearrangements represent a pathogenic hallmark of most B-cell lymphoma entities and some are associated with an unfavorable clinical outcome. Recurrent structural genomic aberrations involving the MHC class II transactivator CIITA (located on chromosome 16p13) have been identified in multiple lymphoma subtypes in which they contribute to an immune escape phenotype. Moreover, inactivating mutations of the tumor suppressor gene SOCS1, located in close proximity to CIITA on chromosome 16, result in constitutively active JAK-STAT-signaling across a spectrum of lymphomas. Preliminary data from our group indicate that SOCS1 is also involved in recurrent rearrangements. However, more detailed study of these rearrangements was hampered due to the lack of analytic methods and platforms applicable to archival formalin-fixed and paraffin-embedded tissue (FFPET) specimens. Here, using FFPET biopsies, we sought to characterize the comprehensive landscape of CIITA and SOCS1rearrangement partner genes, determine the exact breakpoint anatomy, and study the functional impact of individual alterations. Methods: In order to select cases for DNA extraction and subsequent sequencing analysis we revisited the results of previously conducted fluorescence in-situ hybridization (FISH) experiments performed on lymphoma cases arranged on tissue microarrays. Cases were included based on the presence of rearrangements in CIITA, SOCS1 and the 9p24.1 locus. 92 specimens met these criteria and DNA was extracted using the Qiagen AllPrep DNA/RNA FFPE kit. An optimized 96-well-plate protocol was used for whole genome library construction (FFPET small gap protocol) and 16 to 18 libraries were multiplexed and pooled prior to capture using a custom Agilent SureSelect design targeting the aforementioned loci. Sequencing was performed on an Illumina HiSeq 2500. Predicted structural variants (SV) were generated with the computational tools DELLY and deStruct, and subsequently filtered based on read support. High confidence predictions were validated by either PCR-amplification of the genomic DNA spanning the breakpoint sequence followed by Sanger Sequencing, or customized FISH assays. Results: Library construction was successful in 68 cases (74%) and the mean coverage across the target regions for these libraries was above 600x with 97.7% of bases having an average depth of > 100x. After filtering, a list of 82 rearrangement events was generated consisting of 16 translocations (12 affecting CIITA and 4 affecting SOCS1), 44 deletions, 19 inversions and 3 duplications. Mapping of the translocation breakpoints revealed two distinct breakpoint cluster regions within intron 1 of CIITA and exon 2 of the SOCS1 gene. None of the exact breakpoints were recurrent and identification of rearrangement partner genes confirmed that CIITA and SOCS1 rearrange promiscuously. Furthermore, we demonstrate that some translocation events result in the generation of novel in-frame fusion transcripts (e.g. CIITA-PRDM16), warranting further functional characterization. Consistent with previous studies, intra-chromosomal alterations were detected in intron 1 of CIITA in48% of the cases while SOCS1 was shown to be frequently deleted (20% of cases).Overall, the concordance between FISH and targeted capture sequencing was 63%, reflecting the differences in sensitivity to detect large- and small-scale genomic events. Conclusions: Targeted capture sequencing, in conjunction with FISH, represents a valuable tool to explore the spectrum of genomic structural variants and rearrangement partner genes in archival lymphoma FFPET specimens. Future studies are required to address the functional impact of specific alterations and the utility of capture-sequencing-based assays for clinical decision-making. Disclosures Scott: NanoString Technologies: Patents & Royalties: named inventor on a patent for molecular subtyping of DLBCL that has been licensed to NanoString Technologies.

INTRODUCTION Acute myeloid leukemia (AML) is a hematopoietic malignancy that comprises almost 25% of pediatricleukemiasand is characterized by genetic and epigenetic alterations that lead to impairment of differentiation and clonal expansion. Despite intensive chemotherapy, more than half of children with AML either fail to achieve a complete remission (CR) or relapse after initial response.As such, the availability of a predictor of treatment failure at diagnosis may allow early institution of alternative therapies to improve outcome. MicroRNAs (miRNAs), a class of small non-coding RNAs that regulate the translation of their target mRNAs, are frequentlydysregulatedin cancers, and thus may serve as robust biomarkers of patient outcome. METHODS To identifymiRNAbiomarkers associated with treatment failure and candidate therapeutic targets, we performed a comprehensive sequence-based characterization of the pediatric AMLmiRNAexpression landscape usingmiRNA-seqdata from 637 primary samples. AmiRNAexpression-based model for EFS separating patients into low, intermediate, andhigh riskgroups was produced using penalized Cox regression. The model was designed usingmiRNAexpression data obtained from a training cohort, which consisted of two-thirds of our study cohort (n=425), and then tested on the remaining one-third of our study cohort (n=212). The training and test cohorts were derived by random selection. RESULTS A 36-miRNA EFS predictive model was generated. This model was comprised of16miRNAsthat were over-expressed and 20 that were under-expressed in patients who experienced an event (death, relapse or IF). Among the 36miRNAtranscripts were miR-155, miR-335, miR-139 and miR-375, which have been previously individually associated with survival in pediatric AML. To demonstrate the potential clinical utility of the model, we determined 2 miRNAmodel score thresholds in the training cohort to separate patients into low, intermediate and high miRNAmodel score risk groups. The miRNAmodel score groupings were significantly associated with EFS in both the training cohort and test cohort (Figure 1A; P Furthermore, to demonstrate the strength of our predictive model, we evaluated the clinical significance of the model in each of the low, standard and high CM risk cohorts. In this analysis, our model was capable of further stratifying patients in each of the 3 CM risk cohorts into 3 distinct miRNAmodel score-based risk categories (Figure 1B, P CONCLUSIONS We present amiRNAexpression-based predictor of outcome in pediatric AML, comprised of 36miRNAtranscripts. Our predictive model was developed and tested on a large cohort of primary patient samples (n=637), and demonstrated that diagnosticmiRNAexpression profiles can identify risk groups in patients independent of other CM risk factors. Moreover, this model is applicable to RNA from samples that are routinely obtained for diagnosis, and thus has the potential to impact clinical practice. Figure 1. Stratifying patients using the miRNA-based EFS predictive model. A)Kaplan-Meier plots displaying EFS differences between patients in variousmiRNA model score groups. Patients with highmiRNA model scores had the poorest outcomes, while patients with the lowmiRNA model scores had superior outcomes. B)Kaplan-Meier curves depicting EFS of patients when classified using ourmiRNA model score groupings. ThemiRNA model score groupings were effective in further stratifying patients within the low, standard and high conventional CM risk group classifications. Figure 1. Stratifying patients using the miRNA-based EFS predictive model. A)Kaplan-Meier plots displaying EFS differences between patients in variousmiRNA model score groups. Patients with highmiRNA model scores had the poorest outcomes, while patients with the lowmiRNA model scores had superior outcomes. B)Kaplan-Meier curves depicting EFS of patients when classified using ourmiRNA model score groupings. ThemiRNA model score groupings were effective in further stratifying patients within the low, standard and high conventional CM risk group classifications. Disclosures No relevant conflicts of interest to declare.

Introduction Induction chemotherapy results in complete remission in 80% of children with acute myeloid leukemia (AML). However, many patients either fail to achieve a remission, or relapse after an initial response and subsequently die of their disease. Although large numbers of somatic karyotypic and molecular alterations have been identified, the majority of them do not indicate a specific target or distinct pathway that can be readily exploited for therapeutic intervention. Materials & Methods As part of a genome-scale approach to identify prognostic markers and therapeutic targets, we provide a comprehensive characterization of the pediatric AML transcriptome, detailing miRNA & mRNA expression patterns and miRNA:mRNA interactions that are characteristic of the disease. A total of 676 patients were considered for this study. Our discovery cohort consisted of miRNA-seq from 259 primary, 22 refractory and 38 relapse samples, and mRNA-seq from 158 primary, 12 refractory and 47 relapse samples. We confirmed our survival analyses on a validation cohort that consisted of miRNA-seq and mRNA-seq from 378 and 87 primary samples, respectively. Unsupervised non-negative matrix factorization (NMF) was used to identify patient subgroups based on miRNA/mRNA expression. To identify miRNA/mRNA expression associated with patient survival, Cox proportional hazards analysis was performed. Wilcoxon tests were performed to identify differentially expressed miRNAs/mRNAs between samples. To screen for functional miRNA:mRNA interactions, we identified miRNA and mRNA pairs with anti-correlated expression profiles and miRNA binding site predictions consistent with miRNA:mRNA interaction. Results Survival analysis of both the discovery and validation cohorts revealed that 6 miRNAs were associated with overall survival (OS) and event free survival (EFS) (p-val NMF clustering of miRNA expression profiles revealed 2 groups of patients, with each group characterized by particular genomic alterations: Group 1 cases were enriched for NPM1 mutation and FLT3 -ITD, while Group 2 cases were enriched for t(8;21), inv(16), MLL rearrangements and CEBPA mutation (Fisher's exact test p-val NMF clustering of mRNA expression revealed 5 groups of patients, in which the group with abundant expression of ribosomal genes was further distinguished by superior OS and EFS (log-rank p-val Conclusions Through a detailed analysis of the transcriptome (Figure A), we identified miRNAs whose expression levels were significantly associated with clinical outcome. In addition, we showed that abundant expression of miR-106a-363 might contribute to treatment resistance by modulating genes involved in energy metabolism. We also demonstrated that reduced expression of ribosomal genes is associated with inferior outcomes, suggesting a dysregulation of protein translation in treatment resistance. Overall, our transcriptome profiles provide clinically meaningful data for risk and response identification and define novel pathways that may be amenable to therapeutic targeting. Figure 1. Summary of the pediatric AML transcriptome Lagadinou ED, et al. Cell Stem Cell. 2013. Figure 1. Summary of the pediatric AML transcriptome. / Lagadinou ED, et al. Cell Stem Cell. 2013. Disclosures No relevant conflicts of interest to declare.

Abstract 28 Hematopoietic stem cells (HSC) exhibit heterogeneity in self-renewal and differentiation activity, but the extent to which this is intrinsically determined and extrinsically regulated is still poorly understood. In the mouse, purities of HSCs can now be achieved to allow such questions to be addressed directly. Interestingly, tracking the outputs of large numbers of serial transplantable clones produced from single-cell transplants, or the clonal progenies of vector-marked/barcoded cells indicate the existence in mice of 2 subsets of HSCs with durable self-renewal ability. These 2 subsets are characterized by distinct lineage output programs that are maintained through the many HSC self-renewal divisions required to serially propagate a clone in vivo. To begin to ask whether similar subsets of human HSCs exist, we have created a diverse lentiviral library encoding an estimated >105 different barcode sequences and GFP, and then used this library to track the in vivo clonal outputs of transduced human CD34+ cord blood cells in xenografted mice. For this experiment, CD34+ cells isolated immunomagnetically to a purity of >80% were exposed to virus for 6 hours in the presence of growth factors and then immediately injected intravenously into 2 sublethally irradiated NOD/SCID-IL2Rγ−/− mice (1.2 × 105 cells per mouse; 30% GFP+ cells after 3 days in vitro). Different subsets of human cells were then isolated by FACS from immunostained bone marrow cells aspirated sequentially from the femurs of the mice at intervals from 4–27 weeks post-transplant and the identity, number and size of clones in each established by next generation sequencing of barcoded amplicons derived from each sample. To identify barcodes arising from PCR and sequencing errors and calibrate clone sizes, we included 3 controls of 20, 100 and 500 cells with a known barcode at each datapoint. The data from these controls allowed a threshold of 20 cells per clone to be established with >95% confidence. We then compared the representation of clones among all samples from each mouse to derive the number and size of all clones detected, assuming a mouse contains 2×108 bone marrow cells. This analysis revealed a total of 154 uniquely barcoded clones containing up to 2×108 human hematopoietic cells in the 2 mice (8–30 × 106 in one and 4–165 × 106 in the other at any single time point). Analysis of the representation of each clone over time showed successive waves of repopulation from different clones with lineage output profiles consistent with those obtained by transplanting separate fractions of CD34+ cord blood cells distinguished by their surface phenotypes. Specifically, we detected 50 clones (32% of all clones) that were not sustained at detectable levels beyond 9 weeks post-transplantation and were characterized by robust myeloid differentiation with variable B cell outputs at 4 weeks. Another 30 clones (19%) showed significant but also transient outputs of either or both the myeloid and B cell lineage, disappearing between week 9 and 16 post-transplant. Mature cell output was detected from a total of 74 clones (48%) at the 27 week time point, among which 36 (23%) were not evident during the first 4 months post-transplant. These late-appearing clones were mostly small (contributing up to 3 × 105 total hematopoietic cells at week 27) and made a significantly higher contribution to the total human myeloid population than to the total human B cell population. Notably, the 12 long term clones that showed robust mature cell output detectable in all 3 sites sampled at week 27 when the mice were sacrificed (left leg vs right leg vs pelvis) contained both myeloid and lymphoid cells but with large (>100-fold) variations in their representation in the 3 different sites. This latter finding suggests less trafficking of human cells between sites than expected from parabiotic mouse experiments or substantial differences in the differentiation control exerted in different locations. Additionally, from one of the mice, we obtained the first direct evidence of a large output of human T cells (>9 × 106) that was part of a long term multi-lineage clone detectable at 27 weeks post-transplant. This first use of a barcoding strategy to analyze the clonal dynamics of normal human CD34+ cells with in vivo repopulating activity demonstrates the power of this approach to analyze their lineage outputs and sets the stage for novel applications to expanded and transformed populations. Disclosures: No relevant conflicts of interest to declare.

Abstract 632 Introduction: Non-Hodgkin lymphomas (NHL) are the most common type of lymphoma and can be broadly classified as indolent (slow-growing) diseases, progressing over many years; and aggressive (fast-growing) diseases, which progress rapidly. The latter class includes diffuse large B-cell lymphoma (DLBCL), which accounts for approximately 30% of all NHL diagnoses. Three DLBCL subtypes have been identified based on gene expression profiling, namely: germinal center B-cell (GCB), activated B-cell (ABC) and primary mediastinal B-cell lymphoma (PMBCL). These subtypes show substantial differences in response to treatment and ultimate disease outcome, suggesting that molecular subtyping is an important prognostic indicator and that each subtype may benefit from a distinct treatment regimen. Despite recent advances in cancer genomics revealing molecular and mutational differences between these subtypes, further studies focused on the common NHL subtypes are required to identify critical players in the pathogenesis of DLBCL that may be targeted by pharmacological intervention to improve patient outcome. Methods: Using ultra-high throughput whole genome shotgun sequencing (WGSS) and whole transcriptome shotgun sequencing (WTSS/RNA-seq) we have discovered protein-coding mutations in NHL genomes. With a focus on recurrent and likely gain-of-function mutations we have established procedures to model the three-dimensional structures of mutant proteins and using a computational “molecular docking” pipeline have identified candidate molecules with specificity for the mutant protein. These small molecule compounds are acquired and tested in cell proliferation assays against a suite of DLBCL cell lines characterized for target mutations. Results: Mutations affecting a single key tyrosine in the catalytic site of enhancer of zeste, homolog 2 (EZH2), a member of the Polycomb-group family involved in transcriptional repression were identified (Morin, R. et al. 2010 Nature Genetics 42(2):181-5). This mutation, in a gene previously unknown to be mutated in cancer, is restricted to the GCB subtype of lymphomas and is highly prevalent in patient samples and DLBCL cell lines. Mutations have also been observed in other proteins involved in epigenetic regulation and thus afford potentially novel therapeutic targets. In proof-of-principle experiments small molecule inhibitors were identified using molecular docking approaches to target the effect of EZH2 mutations in both mutant and wild-type DLBCL cell lines. We identified and imported 96 compounds from the Developmental Therapeutic Program NCI/NIH repository. These compounds were tested in alamarBlue cell proliferation assays revealing three with activity at 10uM concentration in EZH2 mutant but not wild-type cells. Computational optimization of these compounds is underway to identify related compounds with improved activities at reduced concentrations. Conclusions: High-throughput sequencing platforms have enabled the identification of recurrent, non-synonymous protein mutations in tumor genomes and transcriptomes. Such a catalogue of mutations provides new avenues of exploration for targeted therapy including small molecule inhibitors. Despite intensive efforts launched in recent years to determine the crystal structure for every human protein, many (including EZH2) do not currently have three dimensional structures. This poses a challenge to novel drug discovery but can be overcome using homology modeling and/or targeting other members of a pathway. Our observations also demonstrate the importance of epigenetic regulation in NHL tumorigenesis and thus provide potential new therapeutic targets. Disclosures: No relevant conflicts of interest to declare.

Abstract 4187 Background. Diffuse large B-cell lymphomas (DLBCL), which comprise the most common subtype of non-Hodgkin lymphomas, can be further classified at the molecular level into activated B-cell (ABC) type and germinal centre B-cell (GCB) type. The t(14;18) translocation, which juxtaposes BCL2 to the immunoglobulin heavy gene, is frequent in the GCB subtype, while BCL2 can be up-regulated by gene amplifications and activation of the NF-kB pathway in the ABC subtype. We have recently recently re-sequenced DLBCL transcriptomes and have reported that EZH2 is recurrently mutated in the germinal center lymphomas (Morin et al., Nature Genetics 2010). This project also revealed that the gene containing the most variants was BCL2. The aim of this project was to determine the incidence and clinical significance of BCL2 mutations in a larger cohort of de novo DLBCL samples and matched germline in a subset of cases. In addition, the specificity of BCL2 to DLBCL was determined by sequencing the BCL2 gene in other lymphoma subtypes. Methods. 31 cases of DLBCL were subjected to whole transcriptome shotgun sequencing as part of the EZH2 study. The presence of BCL2 variants was determined using the Bayesian variant caller “SNVmix”. The BCL2 gene (exons 1–3) was re-sequenced using the Sanger method in a total of 514 samples, of which 349 cases were primary DLBCL, 30 were small lymphocytic lymphoma (SLL), 26 were follicular lymphoma (FL), 25 were mantle cell lymphoma (MCL), 25 were peripheral T cell lymphoma, 8 were purified centroblasts and 51 were germline DNA from patients who had DLBCL or FL in this study. The presence of the translocation t(14;18) was determined by fluorescence in situ hybridization in 145 DLBCL samples and by karyotype in 26 FL samples. The presence of BCL2 protein expression was determined using clone 124 (Dako) in 284 cases. The association between the presence of mutations and other variables was determined using the Fisher's exact test and the association to overall survival (OS) was determined using the log Rank test. Results. In whole transcriptome shotgun sequencing of 31 DLBCL samples, we found BCL2 to contain expressed point mutations in 11/31 cases, including a total of 19 non-synonymous mutations and 30 synonymous mutations. Sanger sequencing revealed BCL2 mutations in 117/349 DLBCL, 22/26 FL, 3/25 MCL, 1/30 SLL and 2/25 PTCL and no mutations were identified in the germline samples or purified centroblasts. The presence of BCL2 mutations was associated with the GCB phenotype (p = 7.6×10-7) and was enriched in t(14;18)-positive cases. The ratio of transitions to transversions was 1.47, suggestive that these mutations were inserted as a consequence of somatic hypermutation (p = 6.28 × 10-21). Non-synonymous mutations occurred in 41 DLBCL cases and were predominantly located in the flexible loop domain and between the BH1 and BH3 domains while sparing all the BH domains, which are responsible for anti-apoptotic function. Despite this, BCL2 protein was present in 238/284 cases using the Dako antibody. Non-synonymous mutations in BCL2 were not associated with an inferior OS or PFS, even within the GCB subtype. Conclusions. BCL2 mutations in DLBCL and FL are much more prevalent than previously anticipated and appear to be a consequence of on-going somatic hypermutation in the germinal center. The functional and clinical significance of these mutations should be studied further, given that BH3 mimetics may eventually be considered as therapeutic options in these patients. Disclosures: No relevant conflicts of interest to declare.

Abstract 704 Relapsed ALL is a leading cause of childhood cancer death, and the biologic factors responsible for relapse are poorly understood, particularly in cases lacking sentinel chromosomal alterations. Recent studies from the Children's Oncology Group high risk ALL TARGET (Therapeutically Applicable Research to Generate Effective Targets) project that used genome-wide profiling of DNA copy number alterations and candidate gene resequencing have identified novel biomarkers of relapse (IKZF1 alteration) and therapeutic targets (JAK mutation). As a complementary approach to identify novel genomic alterations, we used second generation sequencing technology to sequence the tumor transcriptome of three cases from the COG P9906 high risk (HR) B-precursor ALL trial. The selected cases had previously been profiled by high resolution SNP and gene expression arrays and candidate gene resequencing, and lacked known sentinel chromosomal rearrangements. Each case bore features previously associated with poor treatment outcome: a gene expression profile (GEP) similar to that of BCR-ABL1 positive ALL (all cases), deletion or mutation of IZKF1 (two cases), and JAK mutation (JAK2 R867Q, one case). cDNA libraries were generated from poly-A enriched RNA and 36-50 base paired-end sequencing performed using the Illumina Genome Analyzer. Sequence alignment, variant detection and fusion transcript identification were performed using custom scripts and multiple published reference alignment and de-novo assembly algorithms. A total of 115-127 million total and 93-97 million mapped, unique reads were obtained per case. The average depth of coverage of Refseq exons ranged from 25- to 39-fold. A minimum of 5 putative fusion transcripts were identified per case, some of which were known from prior transcriptome sequencing to be recurring false positives. However, a novel transcript with an in-frame fusion of exon 9 of the striatin gene STRN3 to exon 18 of JAK2 (STRN3-JAK2) was identified in one case, and confirmed by RT-PCR and direct Sanger sequencing. Fusion of NUP214 to ABL1 was identified in a second case and also confirmed by direct sequencing. The NUP214-ABL1 rearrangement has previously only been identified in T-lineage ALL. In this case, the translocation was accompanied by amplification of the NUP214-ABL1 region at 9q. RT-PCR screening of an additional 60 high-risk ALL cases with GEP data suggestive of kinase alteration identified an additional two cases with NUP214-ABL1 fusion, each of which was accompanied by NUP214-ABL1 amplification. These two novel fusion transcripts are predicted to result in aberrant kinase signaling, and are candidates for novel therapeutic intervention. Both occurred in ALLs with a BCR-ABL1-like GEP that lacked known JAK mutations, suggesting that additional novel activating kinase mutations can be discovered via detailed sequence analysis of the 50% of BCR-ABL1-like ALLs that do not have JAK mutations. Aberrant splice variants and truncated isoforms arising from DNA copy number alterations, including internal deletions of PAX5 and truncating deletions of BTG1 were also identified using the transcriptome sequencing data. In addition, these data identified over 400 candidate non-synonymous single nucleotide and insertion/deletion variations in each patient. Known mutations involving PAX5, IKZF1 and JAK2 were robustly identified. Whole genome sequencing of matched normal DNA is underway to remove germline variation from the list of putative variants, and transcriptomic sequencing of additional cases of HR childhood ALL are being performed. Together, these data indicate that transcriptomic sequencing is a powerful method to identify novel genetic alterations in ALL, and may be used to identify novel targets for therapeutic intervention. Disclosures: No relevant conflicts of interest to declare.

Abstract 603 Background: Acquired drug resistance often limits treatment efficacy in multiple myeloma, and may in part be caused by selection for cells bearing mutations conferring drug resistance that were either present in an extant subset of myeloma cells prior to treatment, or were acquired during therapy. We are applying whole transcriptome shotgun sequencing (WTSS) to myeloma BM at diagnosis and drug-resistant relapse, in order to identify gene expression and mutational signatures of drug resistance. We have an archive of frozen myeloma total bone marrow mononuclear cell fractions, as well as a smaller archive of CD138+ enriched specimens. The enriched specimens can be expected to contain purer populations of malignant cells, but may under-represent the heterogeneity of the disease. We are therefore interested in a comparison of the results of WTSS on these two different types of specimens. Methods: Bone marrow from a myeloma patient who responded to front line drug therapy and who subsequently experienced drug resistant disease relapse was identified from our biorepository. RNA was isolated from the total bone marrow mononuclear cell population and from a CD138+ enriched population at diagnosis and at relapse, yielding four samples for analysis (pre-treatment unprocessed, pre-treatment enriched, relapse unprocessed, relapse enriched). Poly-adenylated RNA from each of the four samples was converted to cDNA using a random primed approach, and paired-end sequences collected using Illumina instruments. Results: We sequenced the transcriptomes of the four samples by generating a total of 694,076,590 36- and 50-mer sequences (30,681,959,896 base-pairs). Of these, 325,998,000 (14,063,396,516 base-pairs) were mapped to reference genome or transcriptome databases using MAQ. Each base in 3,094 transcripts was represented at a minimum of 6 fold redundant coverage. 68.57 % (223,555,599) of the reads matched perfectly to reference sequences. Reads containing single nucleotide mismatches to the reference sequence were evaluated using a Bayesian mixture model to estimate the presence of single nucleotide variants (SNVs) which yielded 89,944 high confidence predictions divided amongst the four samples. Of these, 74,704 (83.1%) were identical to known SNVs present in dbSNP and other sequence datasets, and 13,073 represented non-coding or synonymous variants. The remaining 2,167 mismatches most likely represent undiscovered variants also present in the patient's normal DNA or somatic mutations. These two possibilities will be distinguished by sequencing the candidate variants using DNA purified from the patient's non-malignant cells. For candidate sequence variant positions well represented at the position and gene level in each of the two sample pairs we were comparing (pre- and post-treatment unprocessed, and pre- and post-treatment enriched) we found 85 unique to the pre-treatment enriched sample, 145 unique to the post-treatment enriched sample, 250 unique to the pre-treatment unprocessed sample and 51 unique tunique o the post-treatment unprocessed sample. Some of the genes affected by these candidate mutations include genes that have previously been observed in cancer, such as PIM2, SHC1 and SCRIB. Gene expression levels were estimated using the previously published RPKM (reads per kilobase of exon model per million mapped reads) metric. For the unprocessed samples, 108 genes were found to be up-regulated pre-treatment and 95 post-treatment. The enriched samples presented 80 genes with higher expression before treatment and 54 after treatment. Because we are using a paired end sequencing approach, the potential exists to identify candidate expressed gene fusions. Initial searches for candidate gene fusion events have resulted in the detection of a fusion involving IgH and MMSET, corresponding to a t(4;14) translocation which had previously been detected with standard methods. Discussion: It is possible to sequence and compare the expressed myeloma genome of patients at diagnosis and at relapse with WTSS, in order to gain insight into the evolution of myeloma under drug selection and its role in the development of treatment resistance. With the study of a larger cohort of patients and subsequent functional characterization of the candidate sequence variants, we hope to identify recurrent mechanisms of myeloma drug resistance which might be amenable to therapeutic intervention. Disclosures: No relevant conflicts of interest to declare.

Abstract 3967 Poster Board III-903 Background Follicular lymphoma (FL) is considered an indolent but incurable lymphoma. Treatment with cyclophosphamide, vincristine, prednisone and rituximab (CVP-R) provides complete or partial responses in most patients. The BCL2 translocation t(14;18) is present in 85% of the cases, but additional genomic alterations must occur to induce overt FL. We have sequenced the genome and transcriptome of a cytogenetically normal FL sample taken from a patient (pt 1) that had an unusually aggressive clinical course with the aim of identifying genomic alterations that could contribute to FL pathogenesis. We identified a mutation in FAS/CD95, a key component of the extrinsic apoptotic pathway. We hypothesized that FAS mutations may contribute to treatment resistance in FL by inhibiting apoptosis. This study investigates the prevalence and clinical outcome of patients with FL harbouring mutations in the exons coding for the functional “death domain” of the FAS gene. Methods The initial FL sample was subjected to cytogenetic analysis and whole genome tiling array comparative genomic hybridization followed by next-generation sequencing of the tumour genome and tumor transcriptome following the manufacturer's protocol (Illumina). FAS emerged as a potential candidate that could explain the unusually aggressive FL. We performed PCR amplification of exons 7,8,9 and the 3'UTR of the FAS gene with universal M13F(-21) and M13 primer extensions on pre-treatment FL samples derived from 214 patients including 33 diffuse large B cell lymphoma samples that had evolved from a prior FL (i.e. paired FL and DLBCL). PCR products were purified using AMPpure magnetic beads and bi-directionally sequenced using BigDye® Terminator v3.1 and an ABI 3730 XL sequencer. Analysis was performed using Mutation Surveyor. Mutations were considered present if they were observed in both forward and reverse reads. Results Patient 1 had a t(14;18) negative, grade 1 FL, that progressed rapidly despite CVP-R and second line chemotherapy but is now in complete remission following an allogeneic stem cell transplant. The FL immunophenotype was CD19+, CD10+, BCL2+, BCL6+ and lambda clonal. Minimal genomic gains and losses were observed by aCGH. After filtering known single nucleotide polymorphisms (SNPs), a total of 320 candidate novel protein-altering changes were identified (affecting 298 genes) in the tumor genome and transcriptome sequence data. Validation of the mutated genes by Sanger sequencing revealed a novel, somatic and coding mutation in FAS at genomic position chr10:90764005, changing C to T, resulting in a premature truncation of the protein. We then sequenced exons 7, 8, 9 and the 3'UTR of the FAS gene from 214 FL patients. Ten novel FAS mutations were detected, of these six were coding, three of which produced a truncated protein. Of the six coding mutations, two were observed in the transformed DLBCL sample. Coding mutations in FAS appeared to be associated with an aggressive clinical course (median time to progression=12 months for coding mutations (n=6) vs 34 months for non-coding or wild type (n=208), P=0.06). 2 of the 6 patients developed early transformation to DLBCL, 2 had treatment resistant FL that required allogeneic bone marrow transplant and 2 have already died due to progressive treatment refractory FL. Conclusion Next generation sequencing technology revealed novel somatic mutations in a cytogenetically normal FL sample, one of which was a mutation in the FAS gene. Sanger sequencing of a large cohort of FL samples revealed that 5% of cases harboured novel mutations in the death domain of FAS (2% coding, 3% non-coding). Coding mutations were rare but when present were associated with atypically aggressive disease. Disclosures: Connors: Roche Canada: Research Funding. Gascoyne:Roche Canada: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding.

Significant advances have been made in the development of methods for purifying murine hematopoietic cells with longterm (>4 months) in vivo reconstituting ability although these longterm repopulating cells (LTRCs) remain heterogeneous with regard to the self-renewal (SR) activity they display when transplanted into irradiated hosts. Furthermore our group has also identified cell culture conditions that differentially alter LTRC activity without immediate effects on their proliferation or survival. Here, we show that highly purified LTRCs with high and low SR properties can be prospectively isolated from normal adult mouse bone marrow (ABM) as 2 separate populations according to their expression of CD150 within the EPCR ++ CD48 − CD45 mid fraction of cells: 56% total LTRCs and 43% of the high SR type in the CD150 + subset vs. 39% total LTRCs and 32% of the low SR type in the CD150 − subset (as determined from 62 and 28 single cell transplants, respectively). As a first test of whether these populations would likely be useful to search for new molecular differences associated with their different SR properties, we compared the level of expression in these 2 populations of a small set of genes previously reported to regulate LTRC SR activity: c-Kit , Bmi1 , Gata3, Rae28, Ezh2 and Lnk by quantitative real-time PCR (Q-RT-PCR). This exercise revealed transcript levels of the first 4 of these genes to be significantly higher in the CD150 + subset that is selectively enriched in high SR LTRCs, thus validating the concept that they have a distinct molecular signature. Previous evidence shows that high SR LTRCs are present in both FL LTRCs and ABM LTRCs but they differ in some properties (i.e.: cell cycle status, regeneration kinetics). We therefore began a search for ontogeny-independent components of the SR machinery by comparing tags present in 2 LongSAGE libraries produced from CD45 mid lin − Rho − SP ABM cells and from lin − Sca1 + CD43 + Mac1 + embryonic day 14.5 fetal liver (FL) cells (each 20–30% total LTRCs and 12–20% of the high SR type, as determined by 132 (FL) and 352 (ABM) single cell transplants, respectively). From these comparisons and additional data in other publicly available datasets for primitive murine hematopoietic cells, we identified 28 genes not previously shown to have a functional role in LTRC SR control. We then compared the level of expression of these 28 genes between the CD150 + subsets of EPCR ++ CD48 − CD45 mid ABM cells and FL cells (24% total LTRCs and 12% high SR LTRCs in the FL subset) and their respective downstream lin − progeny. This comparison revealed 10 of these genes to be down-regulated in the lin − populations of both ABM and FL. Further comparison of the expression of these 10 genes between the high vs. low SR LTRCs (found in the CD150 + and CD150 − subsets of EPCR ++ CD48 − CD45 mid ) ABM cells showed the expression of 5 ( Vwf , Rhob , Pld3, Prnp and Smarcc2 ) to be downregulated in the CD150 − (low SR LTRC) subset. Interestingly, the first 4 of these genes, as well as 2 of the preliminary set of SR regulators ( Bmi1 and Gata3 ), were also selectively down-regulated in EPCR ++ CD150 + CD48 − CD45 mid ABM cells that had been incubated for 16 hours in 1 or 10 ng/ml Steel factor + 20 ng/ml IL-11 (conditions that decrease LTRC activity in vivo 4–5-fold before any of these divide or die). Taken together, these results point to the existence of more, although a rather small number of additional genes, including Vwf , Rhob , Pld3, and Prnp , whose products may be involved in controlling the SR potential of normal mouse LTRCs.

The homeobox transcription factors Hoxa9 and Meis1 are critical mediators of transformation by MLL fusion proteins and have also been associated with poor prognosis in other AML subtypes. Despite their pivotal role in leukemia, their mechanism of cooperation and their direct target genes are largely unknown. We have established an experimental model to address these questions by transforming murine hematopoietic progenitors with epitope-tagged forms of Hoxa9 and Meis1. Chromatin immunoprecipitation and massively parallel Illumina sequencing (ChIP-seq) was used to identify binding sites for Hoxa9 and Meis1 across the genome in leukemic cells. We generated 1.7 Gb of sequence data for immunoprecipitated DNA and an untagged control, using two biological replicates, and filtered out (1) regions with nonspecific binding, (2) enriched regions that overlapped certain types of repetitive elements, and (3) nonreproducible enriched regions. We validated a subset of enriched regions by ChIP-chip and ChIP-PCR. De novo motif discovery verified that the resulting set of regions contained a motif that corresponded (E value = 6E-16) to a published Hoxa9 binding motif (Shen et al.; Mol. Cell. Biol. 1999, 19:3051–61). Overlapping or immediately adjacent binding of Hoxa9 and Meis1 was seen for 52% of Hoxa9 and 33% of Meis1 enriched regions, suggesting that these two transcription factors cooperate directly in oncogenesis rather than acting through parallel pathways. Only 7% of enriched regions fell within 2 Kb of known transcriptional start sites, while 46% of regions were within gene boundaries; the remaining peaks were intergenic. Remarkably, given the low overall level of intergenic sequence conservation, over 95% of enriched regions overlapped evolutionarily highly conserved regions (Phastcons for 17 mammalian species >50%; P value2-fold change, respectively. Phenotypic changes are evident by 96 h; therefore, the 72 h timepoint is most informative of direct Hoxa9 targets. We compared these data to genes containing ChIP-seq peaks and genes with transcription start sites closest to intergenic peaks. ~10% of genes altered at 72 h were associated with a ChIP-seq peak. Another group of investigators identified targets of Meis1 through comparison of Hoxa9- and Hoxa9+Meis1-immortalized cells (Wang et al.; Blood 2005, 106:254–63); ~14% of reported Meis1 targets were associated with a ChIP-seq peak. Hoxa9 and Meis1 binding sites in selected target genes, such the tyrosine kinase Flt3 and the surface marker Cd34, have been confirmed by ChIP-PCR/ChIP-chip. Our current data support a model in which Hoxa9 and Meis1 directly modulate a substantial number of downstream effectors through co-occupancy of regulatory motifs within proximal promoters, intragenic elements, and distally located regions. Studies are underway to determine whether Hoxa9/Meis1 binding influences gene expression and epigenetic modification at promoters and particularly at HCNEs.

Recent improvements in the development of methods for isolating functionally validated populations of nearly pure (>20%) murine hematopoietic stem cells (HSCs) have made it possible to analyze the molecular basis of the properties of these cells with increased precision. One intriguing feature of HSCs is the change they undergo in many of their key properties during development - a change that affects the control of their self-renewal, cycling status, differentiated progeny output and steel factor sensitivity. To investigate how these differences are mediated, we undertook a genome-wide analysis of the transcripts present in highly purified fetal and adult HSCs using an adaptation of the LongSAGE methodology that allows its application to small numbers of cells (10 ng of RNA) by inclusion of an initial PCR amplification step that preserves the transcript repertoire while excluding less than 0.25% of the transcripts. The LongSAGE methodology was adopted because it is sequence-based and thus quantitative and independent of prior knowledge of expressed genes or variations in their hybridization to matching or related cDNAs, ESTs or derived oligos. A suspension of 10,000 lin-Sca-1+CD43+Mac1+ fetal liver (FL) cells (∼20% pure HSCs as determined by 16-week limiting dilution and single cell transplantation experiments) was obtained from embryonic day 14.5 fetuses. From these cells, we generated a 160,000-tag LongSAGE library containing 7865 tags that map uniquely to the mouse genome (using the RefSeq database through DiscoverySpace; www.bcgsc.ca/DiscoverySpace). A suspension of 3700 CD45midlin-Rho-SP cells (∼30% pure HSCs) was isolated from adult mouse bone marrow (BM) and then used to generate a 37,000-tag LongSAGE library (956 uniquely mapped tags). Both of these libraries contained tags identifying transcripts that have been previously reported to be associated with HSCs from FL and/or adult BM, including c-kit, pbx-1, tgf-β, cul-4a, PrP, c-myc, robo1, sox17, as well as a number of Smarc transcripts, ubiquitin ligase transcripts and TNF-related transcripts. As a first test of the utility of the libraries, we looked for differences in the expression of genes that have been broadly associated with differences in cellular proliferative activity. This comparison identified many such tags in the FL HSC library that were absent from the adult BM HSC library, including multiple cyclins (A2, B2, C, D1, D2, D3, E1, F, H, I, J L1, L2), cdc20, cdc5b, and plk, as well as 34 of the top 50 “proliferation” genes identified by Venezia et al (PLoS Biology, 2004) to be selectively expressed in adult BM HSCs that had been stimulated to proliferate. Other transcripts that were present at significantly higher levels in the FL HSC library (95% C.I. using Audic Claverie statistics) included msl2, rbx1, lmo2, pfn1, and 16 members of the tripartite motif protein (trim) family. Conversely, many transcripts for components of the proteosome, involved in nucleic acid binding, and transcripts coding for proteins with receptor activity were present at higher levels (or uniquely) in the adult BM HSC library. Taken together, these findings establish the validity and potential of these permanent HSC transcriptome resources for further investigation of mechanisms that determine the different biology of fetal and adult HSCs.