ALBERT

Description: Abstract 3364 Introduction Chimeric fusion genes generated by chromosomal translocations are highly prevalent in childhood acute lymphoblastic leukemia (ALL) and mainly represent early, prenatal events. The t(12;21)(p13;q22) translocation generates the ETV6/RUNX1 fusion gene and is the most frequent gene recombination in childhood B-lineage ALL, occurring in approximately 25% of the cases. It is associated with favorable prognosis even though a substantial proportion of the cases relapse. The fusion gene itself initiates the pre-leukemic process but additional genetic hits are needed to trigger a full blown leukemia. Being one of the best-characterized childhood leukemias, both nature and mechanisms of the events that cooperate with the chimeric protein are still poorly understood. Furthermore, studies addressing the genetic origin of relapse demonstrated a clonal relationship between relapse and diagnostic sample, assuming the existence of an ancestral, pre-leukemic clone. Objective Second-generation sequencing of both ends of huge numbers of DNA fragments allows comprehensive characterization of patterns of somatic rearrangements on an unprecedented, high-resolution level. By using the Illumina mate-pair massively parallel sequencing technology and intra-individual side-by-side comparison of leukemic and normal germline DNA, we aim to elucidate the cooperating genetic events in leukemogenesis in ETV6/RUNX1-ALLs as well as the clonal relationship between relapse and diagnostic sample. Methods We investigated diagnostic and relapse samples as well as non-leukemic germline material from one pediatric patient, diagnosed with ETV6/RUNX1-ALL in Germany. Mate-pair genomic sequencing libraries with an insert size of approximately 2-kb were constructed and paired-end sequence reads of 36-bp each were generated on the Illumina Genome Analyzer IIx from randomly created ~500-bp DNA fragments. Data were filtered and aligned to the human reference genome (GRCh37) using BWA. Reads considered PCR duplicates were removed and detection as well as clustering of structural variants (translocations, deletions, inversions) was subsequently carried out with GASV. Discordantly mapping read pairs defined potential structural variations and cluster sizes of at least 4 uniquely and correctly mapping read pairs were included in further analyses. In order to confirm breakpoints and resolve them to base-pair level, areas of putative chromosomal rearrangements were amplified from genomic DNA of tumor and matched normal sample and were conventionally sequenced. Results An average of ~73,000,000 read pairs were generated for each sample and after alignment, the whole genome was sequenced with a mean fragment coverage of 18.8X. A substantial variation in prevalence of structural variants could be detected between paired diagnostic and relapse samples. Within the diagnostic sample we could in total observe 739 deletions, 66 inversions and 107 translocations while the relapse sample exhibited 26 deletions, 14 inversions and 240 translocations. In a first analysis we focused on translocations, presuming the high impact of chromosomal rearrangements on leukemogenesis. Subtracting translocations being of germline origin, 73 translocations at diagnosis and 207 translocations in relapse could be detected, of which 183 (74%) were identified being intragenic. Remarkably, both samples shared only 16 translocations (6%), while 57 (22%) uniquely appear in the diagnostic sample and 191 (72%) could only be observed in the relapse sample. Intragenic, shared translocations in diagnostic and relapse samples include the t(12;24) PDE3A/RN18S1, the t(2;17) PID1/UNC45B as well as the t(1;6) HFM1/EYA4 fusion gene products. Detection of the known ETV6/RUNX1 translocation in diagnostic and relapse sample as well as confirmation of selected breakpoints via Sanger sequencing validated our methodological approach. Conclusion Mate-pair sequencing of leukemic samples in comparison to germline material provides a powerful tool to identify genome-wide chromosomal structural variations and will allow analysis of clonality between diagnostic and relapse samples. The low percentage of shared translocations gives a first hint probably objecting the thesis of a common pre-leukemic clone, but will only be elucidated by analysis of further patient samples. Disclosures: No relevant conflicts of interest to declare.

Description: Introduction The ETV6-RUNX1 fusion gene,the most common subtype of childhood pB-ALL, is acquired in utero, producing a persistent and hidden preleukemic clone. However, the underlying mechanism explaining how the preleukemic clone evolves to pB-ALL remains to be identified. The lack of genetically engineered human-like ETV6-RUNX1 pB-ALL models has hampered our understanding of the pathogenesis of this disease. Methods We have used a novel experimental approach to generate a murine strain that mimics the human ETV6-RUNX1 pB-ALL. We expressed ETV6-RUNX1 specifically in hematopoietic stem cells (HSC) of C57BL/6 x CBA mice by placing ETV6-RUNX1 under the control of the Sca1 promoter. Two founder mice were obtained for the Sca1-ETV6-RUNX1 transgene, which had normal gestation, were viable and developed normally. Sca1-ETV6-RUNX1 transgenic mice were characterized with respect to clinical, immunephenotypic and genetic characteristics. For the detection of shared secondary genomic alterations we analyzed three murine Sca1-ETV6-RUNX1 and 11 ETV6-RUNX1 positive human pB-ALL and corresponding germline by whole-exome (WES) and whole-genome sequencing using a HiSeq 2500 (Illumina) platform. Results In our transgenic murine model Sca1-ETV6-RUNX1 transgene expression was detected in HSCs, while there was no detectable expression in pro B cells or later stages of B-cell development, which mimics human ETV6-RUNX1 preleukemic biology. Sca1-ETV6-RUNX1 mice developed exclusively pB-ALL at a low penetrance (7.5%; 3 out of 40) with a CD19+ B220+ IgM- cell surface phenotype. Overall survival was not significantly reduced compared to wild-type mice (P value = 0.7901). pB-ALL in Sca1-ETV6-RUNX1 mice manifested with splenomegaly, disruption of splenic architecture, and appearance of blast cells in the peripheral blood (PB). All leukemic cells displayed clonal immature BCR rearrangement. Tumor pro B cells grew independent of IL-7 and were able to propagate the disease when transplanted into sub-lethally irradiated syngeneic recipient mice. Whole-exome sequencing of murine pB-ALL revealed in one mouse a deletion of three amino acids in the B-cell differentiation factor EBF1, which is well known in the context of human ETV6-RUNX1 leukemia. Additionally we found mutations in genesimplicated in histone modification, i.e. in KDM5C causing a premature translation stop. We compared the genomic alterations detected in the mouse model to published genomic data of pediatric ETV6 -RUNX1 pB-ALL and identified multiple copy number variations, which are shared between the murine and human ETV6 -RUNX1 pB-ALL. Among them were copy number gains and losses including i.e. the tumorsuppressor locus CDKN2A/B with a well-known role in human and mouse pB-ALL. A high proportion of genes implicated in histone modification was also mutated in published data of human ETV6-RUNX1 positive pB-ALL. We validated this novel finding of recurrent alterations of histone modifying genes in both the murine model and the human disease using an independent human ETV6-RUNX1 cohort of 11 patients. In this cohort were able to reproduce this finding. Similar to the murine model, we also detected a missense mutation in the methyltransferase KDM5C in one patient of our cohort of ETV6-RUNX1 positive patients. Conclusion In summary, we have characterized a new Sca1-ETV6-RUNX1 mouse model and this is, to our knowledge the first model, which represents a phenocopy of the human pB-ALL. Sca1-ETV6-RUNX1 mice develop exclusively pB-ALL at a very low penetrance as it is the case in human ETV6-RUNX1 positive pB-ALL. The acquisition of secondary mutations in pB-ALL with a high proportion in histone modifying genes confers the second hit for the conversion of a preleukemic clone into the clinically overt ETV6-RUNX1 positive pB-ALL disease. These findings are important for encouraging novel interventions that might help to prevent or treat ETV6-RUNX1 positive childhood leukemias. Disclosures No relevant conflicts of interest to declare.

Description: Abstract 401 Introduction: High hyperdiploidy (51–67 chromosomes) is the most frequent numerical cytogenetic alteration found in pediatric B-cell precursor acute lymphoblastic leukemia (ALL), occurring in 25–30% of patients. It is characterized by nonrandom gains of chromosomes X, 4, 6, 10, 14, 17, 18, or 21. Children suffering from high hyperdiploid ALL have a good prognosis, nevertheless in 15–20% of cases the disease will recur. The mechanisms involved in the pathogenesis of primary and relapsing high hyperdiploid ALL are poorly understood. In some cases, IGH rearrangements arise in utero, indicating an early formation of pre-leukemic clones. However, the cellular origin of these pre-leukemic clones, as well as the molecular mechanism underlying the formation of high hyperdiploid cells, remains to be determined. Further genetic changes assisting in the development of ALL and recurrent disease are still unknown. Objective: By using massive parallel genome-wide next generation sequencing (Illumina/Solexa), we intended to identify specific cytogenetic structural variations (SVs) of high hyperdiploid ALL and possible clonal relationships between paired diagnostic and relapse ALL samples. Method: Paired-end sequencing libraries were generated from genomic DNA of diagnostic and relapse leukemic samples as well as germline DNA from the same patient. Libraries of two patients and one high hyperdiploid ALL cell line (MHH-CALL-2) with insert sizes of 350–400 bp were sequenced with paired end reads. Read lengths of 36 bp (Genome analyzer IIx) or 51 bp (HiSeq 2000) were sequenced, respectively. Sequencing raw data were aligned to the human reference genome hg19 (GRCh 37) by Burrows-Wheeler Aligner (BWA) and duplicate reads were removed. Copy number variants (CNVs), deletions, intrachromosomal inversions and interchromosomal translocations were analyzed by FREEC and GASV. After subtraction of germline SVs, putative leukemia-specific SVs were obtained. These were validated by PCR performed on genomic DNA. Specific breakpoints of SVs at single base resolution were identified by capillary sequencing of the PCR products. Results: Sequencing of different libraries yielded 95–279 million unique reads that mapped with both ends to the reference genome. Sequence coverages of 57–87% and fragment coverages of 4.9–12.3x were achieved (Table 1). CNV profiles with 10 kb resolution were generated. A comparison of the CNVs of diagnosis and relapse ALL samples demonstrated a high degree of conformity with only few additional alterations present mainly, but not exclusively, in the relapse samples. In one of the patients, a large gain of chromosome 1q was only observed in the relapse sample (Figure 1). SV analysis of all samples resulted in a total of 375 intragenic deletions, 16 intergenic inversions and 83 translocations (Table 1). PCR validation identified 2 previously unknown somatic translocations in the MHH-CALL-2 cell line concerning chromosomes 3 and 7 as well as chromosomes 15 and 18. Furthermore, 6 novel translocations present at diagnosis and relapse could be validated in patient samples. They were concerning chromosomes 3, 11, 12 and 20. One unique new relapse-specific translocation t(4;7) was identified. Conclusion: Paired-end sequencing of leukemia samples and matched non-tumor materials provides a robust tool for the discovery of genome-wide structural rearrangements. The high degree of conformity of CNVs and SVs detected in paired diagnosis/relapse samples indicate a common origin and a close relationship of the leukemic clones at diagnosis and relapse. The observation of few additional alterations in both diagnostic and relapse samples suggests the presence of different subclones at the time of diagnosis and the evolution of the relapse clone from either the diagnostic clone or a minor subclone. Disclosures: No relevant conflicts of interest to declare.

Description: Abstract 1437 Introduction: Genetic heterogeneity is common not only in solid tumors, but also in leukemias. The analysis of genetic heterogeneity among single cancer cells is vital for a better understanding of cancer evolution and therapeutic failure of systemic cancer therapy. So far, comprehensive genome-wide single cell studies were limited by many technical difficulties. Here, we present a novel approach, combining adapter-linker PCR based whole genome amplification (WGA) with 2nd generation sequencing, that enables comprehensive and comparative genome-wide analysis of single leukemic cells. Methods: WGA, based on adapter-linker PCR (Klein et al PNAS 1999, Stoecklein et al Cancer Cell 2008), of three individually picked cells of the permanent leukemia cell line REH was performed. WGA products, subsequently fragmented to 100 bp or 250 bp, were used for library preparation. After loading one amplified single cell genome per flowcell, DNA was sequenced with paired end (PE) reads (2× 75bp or 2× 100 bp respectively) on a Genome Analyzer IIx or a HiSeq 2000 (Illumina). After alignment with Burrows-Wheeler Aligner (BWA), removal of duplicate read pairs, and identification of SNPs by the Genome Analysis Toolkit (GATK), copy number variants (CNV), loss of heterozygosity (LOH) and allele dropout rates were analyzed, based on the human reference genome (hg19/GRCh37). Results were compared to data obtained by hybridizing pooled gDNA of REH cells of the same passage to a SNP 6.0 array (Affymetrix). Interchromosomal translocations were determined in single cells of the same passage of REH cells by spectral karyotyping (SKY) and compared to sequencing data, analyzed by Geometric Analysis of Structural Variants (GASV). Results: With our approach we obtained up to 600 mio mappable reads per run, evenly spread over the genome, which led to a sequence coverage of up to 67%, with an even higher coverage of coding sequence (76%) and a sequence depth of 16x. Comparison of SNP arraydata with PE sequencing data showed, that they are highly overlapping (99,3%) regarding the detection of normal copy numbers. But also for copy number alterations, consistency between both methods was observed in detecting losses (94.1%) or gains (77.1%) of genomic material (figure 1). Up to 97% of regions of LOH detected by sequencing, were also detected by the SNP array, when analyzed in a resolution of 500K bp. By analyzing the data with higher resolutions of up to 10K bp, an increasing amount of regions of LOH could be detected. However, decreased correlation between SNP array and sequencing data (max. 74.5%) was observed, with high correlation between the sequencing runs (85%). This indicates increased detection of false positive LOH regions by the SNP array and the sequencing approach to be superior in this high resolution. To assess the allele dropout rate as a quality control for the PCR based WGA method, the heterozygous SNPs detected by PE sequencing were compared to those called by the SNP array. High consistency (95%) indicates an allele dropout rate of only 5%. To analyze the accuracy of our approach in detecting genetic heterogeneity between single cells, we assessed the variability in the SNP profile between the three individual cells. As they are derived from a permanent cell line, they are expected to be highly similar. In fact, the SNPs, that were covered in all three sequencing runs showed a variation of less than 0,1% among the single REH cells. As the SNP array is not applicable to asses copy number neutral variations as translocations, the karyotype of REH cells was assessed by SKY, confirming the predescribed translocations t(4;12), t(4;16), t(5;12), t(16;21) and t(12;21). Breakpoint regions comparable to those defined by SKY, were identified for all 5 translocations by analysis of discordant read pairs with GASV. The detection of additional, exclusively by sequencing identified breakpoints, is currently under intensified investigation, to confirm potentially newly discovered breakpoints and reliably rule out false positive results. Conclusion: Our approach provides a powerful tool to achieve an unprecedented genome-wide overview on genomic variations of single cells. The robustness of our single cell approach in comparison to the data acquired with pooled gDNA and the homogeneity of our results in the permanent REH cell line clearly shows the reliability of our approach to assess single cell heterogeneity in primary leukemic samples. Disclosures: No relevant conflicts of interest to declare.

Description: High hyperdiploid acute lymphoblastic leukemia (HeH-ALL) is characterized by 51-67 chromosomes and nonrandom gains of specific chromosomes (X, 4, 6, 10, 14, 17, 18, and 21). It presents the most frequent numerical cytogenetic alteration in childhood pre B-cell ALL occurring in 25-30% of cases. Recurrent disease will affect 15-20%. Pre-leukemic HeH clones are generated in utero, but cooperating oncogenic lesions are necessary for overt leukemia and remain to be determined. Recently, a phenomenon termed chromothripsis has been described in which massive structural variations occur in a single aberrant mitosis. Whole or partial chromosomes are shattered and some fragments are lost in the process of rejoining. Thus, characteristically, chromosomal copy numbers oscillate between two copy number states. Chromothripsis has been suggested to be a tumor-driving alteration that may be present in 2-3% of all human cancers. Its role as a potential cooperating or initiating lesion in HeH-ALL has not been determined. We applied state-of-the-art whole-genome next-generation-sequencing to analyze structural variations in six pediatric patients with recurrent HeH-ALL. Matched sample sets taken at diagnosis, remission and/or relapse were compared. Paired end sequencing was carried out on a Genome Analyzer IIx or a HiSeq 2000 (Illumina), respectively. Reads were aligned against the human reference genome (GRCh37) using BWA. Translocations were detected by GASV. Copy number variations were analyzed by FREEC. Structural variations were validated by PCR/Sanger sequencing and FISH. Of the six patients analyzed, five harbored on average one interchromosomal translocation or intrachromosomal inversion, but one patient presented with massive genomic rearrangements (Figure). These affected chromosome 3, 11, 12 and 20. Ten copy number shifts on chromosome 3 oscillating between two copy number states (2 and 3) indicated that these rearrangements were caused by chromothripsis. Breakpoint sequencing revealed that one of the identified translocations (t(12;20)(p13.1;p12.3)), was indeed a three-loci-rearrangement composed of small fragments derived from chromosomes 3, 12 and 20. Characteristically for chromothripsis, the breakpoints clustered closely. Three breakpoints separated by 224 bp and 64 kb were located in the transducin (beta)-like X-linked receptor 1 (TBL1XR1) gene. Other genes repeatedly targeted included the MACRO domain-containing protein 2 (MACROD2) gene (a deacetylase involved in deacetylation of lysine residues in histones and other proteins), the KIAA1467 gene (a transmembrane protein of the integrin alpha FG-GAP repeat containing 3 (ITFG3) family), and a novel regulatory lincRNA (ENSG00000243276). MACROD2 was previously observed as a target of chromothripsis in a colorectal carcinoma. Thus, the characterized breakpoints may identify fragile genomic sites prone to chromothriptic rearrangement. DNA repair was effectuated by non-homologous-end-joining as typical addition of non-template nucleotides with microhomologies of two to four nucleotides at the breakpoints demonstrated. Copy number profiles of this patient showed that at least two distinct leukemic clones could be identified at diagnosis. One had acquired chromothriptic alterations and presented the dominant clone at relapse indicating chemotherapy resistance and tumor-driving potential. Prior whole-exome sequencing did not reveal mutations in known oncogenes or tumor suppressor genes. Therefore, loss of function or expression of genes affected by chromosomal rearrangements, such as TBL1XR1 that is recurrently mutated in childhood ALL with ETV6-RUNX1 translocation, may account for the tumor-driving effect. All leukemic cells at diagnosis showed conformity concerning number and pattern of whole chromosome gains demonstrating that chromothripsis was not an initiating oncogenic event, but occurred secondary to high hyperdiploidy. Further aberrations (t(4;7), loss of 4q) were gained by the chromothriptic clone and could be detected by FISH in minor subclones pointing at ongoing clonal evolution. Taken together, our study reveals chromothripsis as a novel assisting and tumor-driving lesion in HeH ALL. Chromothripsis in HeH-ALL. Copy number variations and translocations at diagnosis (left) and relapse (right). (magenta: chromothriptic translocations; green: other translocations) Disclosures: No relevant conflicts of interest to declare.

Description: Abstract 518 Introduction The chimeric fusion gene ETV6/RUNX1 generated by the interchromosomal translocation t(12;21) presents the most frequent chromosomal aberration in childhood acute lymphoblastic leukemia (ALL), occurring in approximately 25% of all patients. This ALL subtype is associated with an overall favorable prognosis, nevertheless 10–20% of children will suffer from relapse. ETV6/RUNX1-positive preleukemic clones arise already in utero, but require additional cooperating oncogenic lesions for the development of overt leukemia. The nature of the assisting genetic alterations and the mechanisms driving the development of leukemia and recurrent disease are still not well understood. Methods We applied state-of-the-art whole genome and whole exome next-generation-sequencing to comprehensively analyze the assisting oncogenic alterations in pediatric patients with initial and/or recurrent ETV6/RUNX1+ ALL (primary disease n=11, recurrent disease n=7). Matched sample sets taken at initial diagnosis, remission and relapse were compared. Mate pair and/or paired end sequencing was carried out for whole genome analysis with inserts spanning 2 kb or 500 kb, respectively. Constructed libraries were sequenced from both ends with 36- or 50-bp reads on a Genome Analyzer IIx or a HiSeq 2000 (Illumina/Solexa), respectively. Reads were aligned against the human reference genome (GRCh37) using BWA. Duplicate reads were removed. Unique reads with high mapping quality (q〉35) served as input for GASV which detected translocations and inversions based on the mapping coordinates, insert sizes and read orientation. Variations covered by at least three reads in the tumor sample and not detected in the remission sample or the Database of Genomic Variants were reported. For detection of copy number variations, the program FREEC carried out coverage normalization, computation of copy number ratios between paired leukemia and remission samples (with up to 10 kb resolution) and subsequent segmentation. A subset of six selected patients was further investigated by targeted enrichment of whole exomic regions employing SeqCap EZ libraries (Roche) and 100 bp single read next-generation-sequencing on a HighSeq 2000. Mutations were called using GATK and further processed by an in-house bioinformatic pipeline. Putative somatically acquired mutations were validated by PCR, Sanger sequencing and FISH analysis. Results Genomes were sequenced to at least 13× physical coverage (mate pair) and 6.7× sequence coverage (paired-end). Exome sequencing achieved a minimum of 25× sequence coverage. In silico we detected 155 tumor-specific intragenic translocations. On average each tumor harbored 9 acquired translocations. With the exception of one case (13 translocations at diagnosis, 9 at relapse), the number of translocations was higher in relapse than in the matched diagnosis sample (additional 3 translocations on average). Ongoing validation studies confirmed the defining ETV6/RUNX1 translocation t(12;21) and identified 13 novel translocations. The genes affected are involved in essential signaling pathways, such as cytokine signaling (LIFR), calcium signaling (RCAN2), insulin and anti-apoptotic signaling (PHIP). Interestingly, also a factor essential for pre-mRNA splicing (IBP160) and genes encoding regulatory RNAs (miRNAs, lincRNAs and RNAs involved in splicing) were rearranged. A validated intragenic deletion of 836 bp leading to a frameshift and premature stop affected a calcium ion sensor of the ferlin protein family. Recurrent deletions in 9 of 11 cases (82%) ranging from 5 to 200 kb were detected in the immunoglobulin lambda variable gene cluster (IGLV) on chromosome 22q11. Some of the deletions were extending into the pre-B lymphocyte 1 gene (VPREB1) locus. In silico the probabilty of illegitimate RAG-mediated recombination at the breakpoint sites was determined by evaluation of RIC scores. RIC scores indicated that aberrant V(D)J rearrangements involving cryptic recombination sequence signals had caused the deletions on chromosome 22q11. Conclusion We present somatic mutations that are promising novel candicate genes (e.g. LIFR, RCAN2, PHIP, IBP160) for cooperating secondary mutations in ETV6/RUNX1+ ALL and discuss their impact on the molecular pathology of primary and recurrent disease. Disclosures: No relevant conflicts of interest to declare.