Genomic landscape and chronological reconstruction of driver events in multiple myeloma

ARTICLE https://doi.org/10.1038/s41467-019-11680-1 OPEN 1234567890():,; Genomic landscape and chronological reconstruction of driver events in multiple myeloma Francesco Maura1,2,3,14, Niccoló Bolli3,4,14, Nicos Angelopoulos2,5, Kevin J. Dawson2, Daniel Leongamornlert2, Inigo Martincorena2, Thomas J. Mitchell 2, Anthony Fullam2, Santiago Gonzalez 6, Raphael Szalat7, Federico Abascal 2, Bernardo Rodriguez-Martin8, Mehmet Kemal Samur7, Dominik Glodzik 2,9, Marco Roncador2, Mariateresa Fulciniti7, Yu Tzu Tai7, Stephane Minvielle10, Florence Magrangeas10, Philippe Moreau10, Paolo Corradini3,4, Kenneth C. Anderson7, Jose M.C. Tubio2,8, David C. Wedge11, Moritz Gerstung 6, Hervé Avet-Loiseau12, Nikhil Munshi7,13,15 & Peter J. Campbell2,15 The multiple myeloma (MM) genome is heterogeneous and evolves through preclinical and post-diagnosis phases. Here we report a catalog and hierarchy of driver lesions using sequences from 67 MM genomes serially collected from 30 patients together with public exome datasets. Bayesian clustering defines at least 7 genomic subgroups with distinct sets of co-operating events. Focusing on whole genome sequencing data, complex structural events emerge as major drivers, including chromothripsis and a novel replication-based mechanism of templated insertions, which typically occur early. Hyperdiploidy also occurs early, with individual trisomies often acquired in different chronological windows during evolution, and with a preferred order of acquisition. Conversely, positively selected point mutations, whole genome duplication and chromoplexy events occur in later disease phases. Thus, initiating driver events, drawn from a limited repertoire of structural and numerical chromosomal changes, shape preferred trajectories of evolution that are biologically relevant but heterogeneous across patients. 1 Myeloma Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 2 The Cancer, Ageing and Somatic Mutation Programme, Wellcome Sanger Institute, Hinxton, Cambridgeshire CB10 1SA, UK. 3 Department of Medical Oncology and Hemato-Oncology, University of Milan, Milan, Italy. 4 Department of Medical Oncology and Hematology, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy. 5 School of Computer Science and Electronic Engineering, University of Essex, Colchester, UK. 6 European Bioinformatics Institute, European Molecular Biology Laboratory (EMBLEBI), Hinxton, UK. 7 Jerome Lipper Multiple Myeloma Center, Dana–Farber Cancer Institute, Harvard Medical School, Boston, MA, USA. 8 CIMUS - Molecular Medicine and Chronic Diseases Research Centre, University of Santiago de Compostela, Santiago de Compostela, Spain. 9 Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, NY, USA. 10 CRCINA, INSERM, CNRS, Université d’Angers, Université de Nantes, Nantes, France. 11 University of Oxford, Big Data Institute, Oxford, UK. 12 IUC-Oncopole, and CRCT INSERM U1037, 31100 Toulouse, France. 13 Veterans Administration Boston Healthcare System, West Roxbury, MA, USA. 14These authors contributed equally: Francesco Maura, Niccoló Bolli. 15These authors jointly supervised this work: Nikhil Munshi and Peter J. Campbell. Correspondence and requests for materials should be addressed to N.M. (email: ) or to P.J.C. (email: ) NATURE COMMUNICATIONS | (2019)10:3835 | https://doi.org/10.1038/s41467-019-11680-1 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-11680-1 T he genome of multiple myeloma (MM) is complex and heterogeneous, with a high frequency of structural variants (SVs) and copy-number abnormalities (CNAs)1–3. Translocations between the immunoglobulin heavy chain (IGH) locus and recurrent oncogenes are found in ~40% of patients. Cases without IGH translocations often have a distinctive pattern of hyperdiploidy affecting odd-numbered chromosomes, where the underlying target genes remain mysterious. These SVs and recurrent CNAs are considered early drivers, being detectable also in premalignant stages of the disease1–3. Cancer genes are also frequently altered by driver point mutations, with mitogenactivated protein kinase (MAPK) and nuclear factor kappa-lightchain-enhancer of activated B cells (NF-KB) signaling as major targets4–8. Many blood cancers develop along preferred evolutionary trajectories. Early driver events, drawn from a restricted set of possible events, differ in which subsequent cancer genes confer clonal advantage, leading to considerable substructures of cooperativity and mutual exclusivity among cancer genes. These subtypes vary in chemosensitivity and survival, suggesting that although patients share a common histological and clinical phenotype, the underlying biology is distinctly heterogeneous. Preliminary studies have suggested that these patterns exist in MM as well5–12, but have not yet been systematically defined in large cohorts with broad sequencing coverage. There have been recent reports of dependencies among MM driver mutations using either targeted or exome-based approaches9,13, but we lack a comprehensive characterization of MM genomic subgroups based on the complete catalog of driver mutations, copy-number changes and recurrent SVs. In addition, since the first MM whole genome sequencing study6, the landscape of nonrecurrent SVs and complex events has not been systematically explored. In this study, we combine a large cohort of serial MM samples analyzed by whole-genome sequencing (WGS) with a publicly available dataset to define driver events and how they group across patients, with implications for disease classification. Furthermore, we describe the temporal evolution of the disease in preclinical phases, highlighting the unexpected dynamism of genomic changes, often in the form of private, complex structural events. Results Landscape of driver mutations in MM. We performed whole WGS of 67 tumor samples collected at different time points from 30 MM patients, together with matched germline controls (Supplementary Data 1 and 2, “Methods”). We also included in our analyses published whole exome data from 804 patients within the CoMMpass trial (NCT01454297)14. To discover MM driver genes, we analyzed the ratio of nonsynonymous to synonymous mutations, correcting for mutational spectrum and covariates of mutation density across the genome with the dNdScv algorithm15–17. Overall, 55 genes were significantly mutated with a false discovery rate of 1% (Fig. 1a and Supplementary Data 3). Our shortlist of genes showed a 65% overlap with a recently published study from the Myeloma Genome Project (MGP) (Supplementary Data 3)13, with less frequently mutated genes accounting for the discordant calls. This is expected given the random sampling and differences in statistical approaches and power between studies (Supplementary Data 3). To confirm this, we restricted the multiple hypotheses testing pvalue correction to the set of (...truncated)