Cell-Free-DNA-Based Copy Number Index Score in Epithelial Ovarian Cancer—Impact for Diagnosis and Treatment Monitoring

Background: Chromosomal instability, a hallmark of cancer, results in changes in the copy number state. These deviant copy number states can be detected in the cell-free DNA (cfDNA) and provide a quantitative measure of the ctDNA levels by converting cfDNA next-generation sequencing results into a genome-wide copy number instability score (CNI-Score). Our aim was to determine the role of the CNI-Score in detecting epithelial ovarian cancer (EOC) and its role as a marker to monitor the response to treatment. Methods: Blood samples were prospectively collected from 109 patients with high-grade EOC. cfDNA was extracted and analyzed using a clinical-grade assay designed to calculate a genome-wide CNI-Score from low-coverage sequencing data. Stored data from 241 apparently healthy controls were used as a reference set. Results: Comparison of the CNI-Scores of primary EOC patients versus controls yielded sensitivities of 91% at a specificity of 95% to detect OC, respectively. Significantly elevated CNI-Scores were detected in primary (median: 87, IQR: 351) and recurrent (median: 346, IQR: 1891) blood samples. Substantially reduced CNI-Scores were detected after primary debulking surgery. Using a cut-off of 24, a diagnostic sensitivity of 87% for primary and recurrent EOC was determined at a specificity of 95%. CNI-Scores above this threshold were detected in 21/23 primary tumor (91%), 36/42 of platinum-eligible recurrent (85.7%), and 19/22 of non-platinum-eligible recurrent (86.3%) samples, respectively. Conclusion: ctDNA-quantification based on genomic instability determined by the CNI-Score was a biomarker with high diagnostic accuracy in high-grade EOC. The applied assay might be a promising tool for diagnostics and therapy monitoring, as it requires no a priori information about the tumor.

CLONAL HEMATOPOIESIS IN A CENTENARIAN COHORT

Mosaicism, the presence of two or more genotypically or karyotypically distinct populations of cells in a single individual, plays an important role in human disease. Mosaicism can result in mutations and/or chromosomal alterations such as loss, gain, or copy-number neutral loss of heterozygosity. Clonal mosaicism and its relationship to aging and cancer, has been previously studied, and earlier work suggests that clonal mosaicism tends to increase with age. The aim of our research is to use genotype data of centenarians to explore the relationship between extreme longevity and mosaic chromosomal alterations (mCAs). To this end, we analyzed genome-wide genotypes from blood-derived DNA of 338 individuals from the New England Centenarian Study. The participants in this dataset ranged from 45 to 112 years of age. For the detection of mCA events, we used MoChA (https://github.com/freeseek/mocha), a bcftools extension, that predicts mCAs based on B-allele frequency (BAF) and log2 intensity(R) ratio (LRR), and uses long-range phase information to increase sensitivity. Chromosomal alteration events, including whole chromosome events, were detected in 180 out of the 338 individuals. A total of 165 duplications, 97 deletions, and 9 copy-number neutral loss of heterozygosity were detected. Additionally, there were 42 events whose copy number state could not be determined with high confidence. 236 events out of the 313 were detected in individuals aged 100 and older. Our analysis of chromosomal alteration frequency by age indicates that, within centenarians, the proportion of individuals with mCAs significantly decreases with increased age (p < 0.05, correlation -0.73).

Algorithmic improvements for discovery of germline copy number variants in next-generation sequencing data

Copy number variants (CNVs) play a significant role in human heredity and disease, however sensitive and specific characterization of CNVs from NGS data has remained challenging. Detection is especially problematic for hybridization-capture data in which read counts are the sole source of copy number information. We describe two algorithmic adaptations that improve CNV detection accuracy in a Hidden Markov Model (HMM) context. First, we present a method for com puting target- and copy number state-specific emission distributions. Second, we demonstrate that the Pointwise Maximum a posteriori (PMAP) HMM decoding procedure yields improved sensitivity for small CNV calls compared to the more common Viterbi HMM decoder. We develop a prototype implementation, called Cobalt, and compare it to other CNV detection tools using sets of simulated and previously detected CNVs with sizes spanning a single exon up to a full chromosome. In both the simulation and previously detected CNV studies Cobalt shows similar sensitivity but significantly improved positive predictive value (PPV) compared to other callers. Overall sensitivity is 80%-90% for deletion CNVs spanning 1-4 targets and 90%-100% for larger deletion events, while sensitivity is somewhat lower for small duplication CNVs. Cobalt demonstrates significantly improved positive predictive value (PPV) compared to other callers with similar sensitivity, typically making 5X fewer total calls overall.

SplitThreader: Exploration and analysis of rearrangements in cancer genomes

Genomic rearrangements and associated copy number changes are important drivers in cancer as they can alter the expression of oncogenes and tumor suppressors, create gene fusions, and misregulate gene expression. Here we present SplitThreader (http://splitthreader.com), an open-source interactive web application for analysis and visualization of genomic rearrangements and copy number variation in cancer genomes. SplitThreader constructs a sequence graph of genomic rearrangements in the sample and uses a priority queue breadth-first search algorithm on the graph to search for novel interactions. This is applied to detect gene fusions and other novel sequences, as well as to evaluate distances in the rearranged genome between any genomic regions of interest, especially the repositioning of regulatory elements and their target genes. SplitThreader also analyzes each variant to categorize it by its relation to other variants and by its copy number concordance. This identifies balanced translocations, identifies simple and complex variants, and suggests likely false positives when copy number is not concordant across a candidate breakpoint. It also provides explanations when multiple variants affect the copy number state and obscure the contribution of a single variant, such as a deletion within a region that is overall amplified. Together, these categories triage the variants into groups and provide a starting point for further systematic analysis and manual curation. To demonstrate its utility, we apply SplitThreader to three cancer cell lines, MCF-7 and A549 with Illumina paired-end sequencing, and SK-BR-3, with long-read PacBio sequencing. Using SplitThreader, we examine the genomic rearrangements responsible for previously observed gene fusions in SK-BR-3 and MCF-7, and discover many of the fusions involved a complex series of multiple genomic rearrangements. We also find notable differences in the types of variants between the three cell lines, in particular a much higher proportion of reciprocal variants in SK-BR-3 and a distinct clustering of interchromosomal variants in SK-BR-3 and MCF-7 that is absent in A549.

Comparison of TP53 Alterations in Hematological Malignancies

Background: TP53 is altered in ~50% of human cancers. Alterations include mutations and deletions. Both frequently occur together, supporting the classical "two-hit" hypothesis for tumor-suppressor genes. Aim: Comparison of TP53 mutation/deletion patterns in different hematological malignancies, including AML, MDS, ALL, Burkitt lymphoma, CLL and T-PLL. We analyzed (i) the frequencies of TP53 mutations and deletions, (ii) the types of mutation, (iii) the mutation load, (iv) the correlations to cytogenetic aberrations, (v) the age dependency, and (vi) impact on survival. Patient cohort and methods: A total of 3383 cases (AML: n=858, MDS: n=943, ALL: n=358, Burkitt lymphoma: n=25, CLL: n=1148 and T-PLL: n=51) were analyzed for TP53 deletions by interphase FISH determining the copy number state and for TP53 mutations by next-generation amplicon deep sequencing. Karyotype data was available for all cases. Results: Overall, alterations in TP53 were detected in 361/3383 cases (11%; 186 cases with mutation only (mut only), 51 cases with deletion only (del only), 124 cases with mutation and deletion (mut+del)). Regarding the respective entities, the highest frequency of TP53 alterations was observed in patients with Burkitt lymphoma (total alteration frequency: 56%, mut+del: 12%, mut only: 44%, no case del only). Alterations in TP53 also occured with a high incidence in patients with T-PLL (total: 30%; mut+del: 10%; mut only: 4%; del only: 16%) followed by cases with ALL (total: 19%; mut+del: 6%; mut only: 8%; del only: 5%) and AML (total: 13%; mut+del: 5%; mut only: 7%; del only: 1%). By contrast, TP53 alterations occurred less frequently in patients with CLL (total: 8%; mut+del: 4%; mut only: 3%; del only: 1%) and MDS (total: 7%; mut+del: 1%; mut only: 5%; del only: 1%). Missense mutations were found to be the most abundant mutation type in all entities analyzed with a frequency ranging from 71% - 88%. In all entities mainly one mutation per case was detected; however, MDS cases were found to harbour a statistically increased proportion of cases with two mutations compared to the other entities (p = 0.003). High TP53 mutation loads were detected in T-PLL (median: 88%) and AML (47%), whereas the lower ones were found in ALL (28%), Burkitt lymphoma (39%), MDS (39%), and CLL (36%). A strong correlation of TP53 alterations with a complex karyotype was observed in AML (of patients with TP53 alteration: 5% with normal karyotype, 67% with complex karyotype, 28% with other aberrations), ALL (16% normal, 45% complex, 39% other), MDS (14% normal, 53% complex, 33% other), and T-PLL (20% normal, 47% complex, 33% other). By contrast, in CLL and Burkitt lymphoma, TP53 alterations were mainly correlated with other aberrations (CLL: 10% normal, 30% complex, 60% other; Burkitt: 29% normal, 0% complex, 71% other). TP53 mut and TP53 mut+del were significantly more frequent in patients ≥ 60 vs < 60 years in AML (9% vs. 2% for mut only, p < 0.001; 7% vs. 2% for mut+del, p = 0.001) and ALL (12% vs. 6% for mut only, p < 0.001; 13% vs. 3% for mut+del, p = 0.001). By contrast, no such differences were observed for patients with CLL, MDS, T-PLL and Burkitt lymphoma. Moreover, TP53 alterations (especially of TP53 mut+del) had a significant negative impact on OS in all entities except for T-PLL and Burkitt lymphoma, most probably due to their overall short OS or due the lower number of cases. Conclusion: The frequency of TP53 mutations and/or deletions as well as the mutation load clearly varied between different hematological malignancies with the highest incidence of TP53 mut in patients with Burkitt lymphoma (56%) and a rather low frequency in CLL (7%) and MDS (6%). TP53 del were frequent in patients with T-PLL (26%) and Burkitt lymphoma (12%) and are hardly found in MDS cases (2%). TP53 alterations are correlated to higher age in AML and ALL. Moreover, alterations in TP53 are correlated to a short OS and to a complex karyotype, with the exception of Burkitt lymphoma and CLL, were they were found to be associated to other cytogenetic aberrations. Thus, TP53 mutations and deletions need further investigation in the future, especially regarding their clinical impact in different hematologic entities. Disclosures Stengel: MLL Munich Leukemia Laboratory: Employment. Kern:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Schnittger:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Zenger:MLL Munich Leukemia Laboratory: Employment. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership.

A Comprehensive Cytogenetic and Molecular Genetic Characterization of Patients with T-PLL Revealed Two Distinct Genetic Subgroups and JAK3 Mutations As an Important Prognostic Marker

Background: T-cell prolymphocytic leukemia (T-PLL) is a rare, mature T-cell neoplasm with poor prognosis. Only few T-PLL cases have been analyzed with regard to cytogenetic and molecular genetic aberrations so far. Therefore, we performed a comprehensive characterization of patients with T-PLL, including the identification of potential correlations between the respective markers and their impact on prognosis. Methods: The cohort comprised 47 T-PLL cases (32 male, 15 female). Median age was 69.8 years (range: 32.7-86.6 years). Diagnosis of T-PLL was assigned by immunophenotyping and cytomorphology. All 47 patients were further investigated using (i) chromosome banding analysis (CBA), (ii) interphase FISH to determine the copy number state for TP53 and ATM and chromosomal rearrangements of TCRA/D and TCL1 and (iii) array CGH. Next-generation amplicon deep-sequencing was performed to analyze mutations in ATM,BCOR, TP53 (n=47, respectively); JAK1 (n=44) and JAK3 (n=45) were analyzed by Sanger sequencing. Clinical follow-up data was available for 43 patients. Results: In all 47 cases, chromosomal abnormalities and/or molecular mutations were detected. Combining CBA and FISH data, an inv(14)(q11q32)/t(14;14)(q11;q32) was observed in 37/47 (78.7%) cases, a t(X;14)(q27;q11) in 3 cases (6.4%) and an i(8)(q10) in 17/47 (36.2%) cases. ATM deletions were detected in 27/47 (57.5%), TP53 deletions in 11/47 (23.4%) patients. Array CGH analyses revealed additional gains and losses of specific chromosomal regions, mainly affecting 7q (deletions in region 7q34-7q36; n=16), 12p (deletions in 12p12-12p13; n=11) and 22q (deletions in 22q11-q12 with a concomitant gain of 22q12-q13; n=8). Regarding molecular analyses, the most frequently mutated gene was ATM (34/47; 72.3%). Mutations in TP53 were found in 7/47 (14.9%) and in BCOR in 4/47 (8.5%) patients. Mutations of JAK1 were found in 3/44 (6.8%), and of JAK3 in 8/45 (17.8%) cases. ATM and TP53 frequently carried a mutation of one allele and a deletion of the other: 23/34 (67.6%) cases with ATM mutation also showed an ATM deletion and in 5/7 (71.4%) cases with TP53 mutation also a TP53 deletion was detected. Regarding chromosomal aberrations, all cases with i(8)(q10) harbored a TCRA/D rearrangement and an ATM mutation, whereas TP53 mutations were not present in any case with i(8)(q10). ATM mutations were found to be correlated to TCRA/D rearrangements (33/40 TCRA/D+ cases, 82.5%; 1/7 TCRA/D- cases, 14.3%; p<0.001). In contrast, TP53 mutations were predominantly observed in patients without TCRA/D rearrangement (4/7 TCRA/D- cases, 57.1%; 3/40 TCRA/D+ cases, 7.5%; p=0.008). Additionally, all three JAK1 mutations were detected in cases with a TCRA/D rearrangement. When splitting the cohort into patients ≤60 years (n=13) and >60 years (n=34), JAK1 mutations (0/12 vs. 3/32) and mutations/deletions in the TP53 gene were detected exclusively in patients >60 years (TP53mut: 0/13 vs. 7/34; TP53del: 0/13 vs. 11/34). JAK3 mutations were also found predominantly in older patients (1/12; 8.3% vs. 7/33; 21.2%). Median overall survival (OS) was 27.4 months. No influence on OS was found for mutations and/or deletions of ATM, TP53, BCOR orJAK1 or aberrations of chromosomes 8 or 14. The age of patients was found to impact OS (median OS, ≤60 years: 29.0 months vs. >60 years: 15.9 months), although this was not significant (p=0.077). However, OS was found to be significantly shorter in patients with JAK3 mutation compared to patients without JAK3 mutation (median OS, 5.1 months vs. 29.1 months; p=0.009). Conclusions: Genetic abnormalities were revealed in all 47 cases with T-PLL. Two distinct genetic subgroups of T-PLL were identified: A large subset, comprising 81% of patients, showed abnormalities involving the TCRA/D locus activating the proto-oncogenes TCL1 (14q32) or MTCP1 (Xq28). This subgroup had higher frequencies of i(8)(q10) and of ATM mutations, while the second group was characterized by a higher frequency of TP53 mutations (figure). Further, JAK3 mutations were identified as an important prognostic marker, showing a significant negative impact on OS. Figure 1: Genetic abnormalities in T-PLL Figure 1:. Genetic abnormalities in T-PLL mut=mutation, del=deletion, TCRA/D=rearrangements involving TCRA/D, TCL1=rearrangements involving TCL1, MTCP1= rearrangements involving MTCP1 Disclosures Stengel: MLL Munich Leukemia Laboratory: Employment. Kern:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Zenger:MLL Munich Leukemia Laboratory: Employment. Perglerová:MLL2 s.r.o.: Employment. Schnittger:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach:MLL Munich Leukemia Laboratory: Employment, Equity Ownership.

DETECTING COPY NUMBER VARIATIONS FROM ARRAY CGH DATA BASED ON A CONDITIONAL RANDOM FIELD MODEL

Array comparative genomic hybridization (aCGH) allows identification of copy number alterations across genomes. The key computational challenge in analyzing copy number variations (CNVs) using aCGH data or other similar data generated by a variety of array technologies is the detection of segment boundaries of copy number changes and inference of the copy number state for each segment. We have developed a novel statistical model based on the framework of conditional random fields (CRFs) that can effectively combine data smoothing, segmentation and copy number state decoding into one unified framework. Our approach (termed CRF-CNV) provides great flexibilities in defining meaningful feature functions. Therefore, it can effectively integrate local spatial information of arbitrary sizes into the model. For model parameter estimations, we have adopted the conjugate gradient (CG) method for likelihood optimization and developed efficient forward/backward algorithms within the CG framework. The method is evaluated using real data with known copy numbers as well as simulated data with realistic assumptions, and compared with two popular publicly available programs. Experimental results have demonstrated that CRF-CNV outperforms a Bayesian Hidden Markov Model-based approach on both datasets in terms of copy number assignments. Comparing to a non-parametric approach, CRF-CNV has achieved much greater precision while maintaining the same level of recall on the real data, and their performance on the simulated data is comparable.

High-Resolution Genomic Profiling of Philadelphia Chromosome-Positive (Ph+) Acute Lymphoblastic Leukaemia (ALL) Identified Novel Recurrent Copy Number Variations Involved in Both Pathogenesis and Resistance to Tyrosine Kinase Inhibitors.

The Ph chromosome is the most frequent cytogenetic aberration associated with ALL and it represents the single most significant adverse prognostic marker. Despite the encouraging results achieved with imatinib, resistance develops rapidly and is quickly followed by disease progression. Some mechanisms of resistance have been widely described but the full knowledge of contributing factors driving both the disease and resistance remains to be defined. In order to identify at submicroscopic level genetic lesions driving leukemogenesis and resistance, we profiled until now the genomes of 18 patients, out of 55 Ph+ ALL patients treated in our institute, at diagnosis (n=11) or at the time of haematological relapse (n=7) during therapy with imatinib or dasatinib. 250 ng of genomic DNA were processed on 500K single nucleotide polymorphism (SNP) array according to protocols provided by the manufacturer (Affymetrix Inc., Santa Clara, CA, USA). The median SNP call rate of analysed samples was 96%. Raw signal data were analyzed by BRLMM algorithm and copy number state was calculated with respect to a set of 48 Hapmap normal individuals and a diploid reference set of samples obtained from acute leukaemia cases in remission. Regions of amplification and deletion were visualized by Integrated Genome Browser and mapped to RefSeq to identify the specific genes involved in the lesion. Our analysis identified multiple copy number alterations per case, with deletions outnumbering amplification almost 3:1. Lesions varied from loss or gain of complete chromosome arms (trisomy 4, monosomy 7, loss of 9p, 10q, 14q, 16q and gain of 1q and 17q) to microdeletions and microduplications targeting genomic intervals. The recurring microdeletions that we detected in at least 50% of patients (both at diagnosis and at relapse) included 1p36.21 (PRAMEF), 3q29 (TFCR), 7p14.1 (AMPH), 8p23 (DEFB105A), 14q11.2 (DAD1), 16p13.11 (PDXDC1, NTAN1, RRN3), 16p11.2 (SNP) and 19p13.2 (CARM1, SMARCA4). A common microamplification was 4q13.2 (TMPRSS11E) and 17q21.31. Some genomic alterations were identified in genes regulating B-lymphocyte differentiation, such as PAX5 (n=3), BLNK (n=1) and VPREB1 (n=6) and in genes with an established role in leukemogenesis, such as MDS, BTG1, MLLT3 and RUNX1. Furthermore, many of the deletions detected included genes encoded for phosphatase proteins (e.g. PTPRD, PPP1R9B, PTPN18) and for zinc-finger proteins without any difference between diagnosis and resistance. It is noteworthy that some lesions felt in regions lacking annotated genes (loss: 2p11.2, 3p12.3, 7q11.21 and 14q32.33; gain: 8q23.3 and 13q21.1). Using high-resolution genome wide approach we showed that Ph+ ALL is a more complex disease characterized by multiple genomic anomalies which may provide new insights into the mechanisms underlying leukemogenesis and may be used as targets for existing or novel drugs. Supported by: European LeukemiaNet, COFIN 2003, Novartis Oncology Clinical Development, AIL.