Frequent somatic transfer of mitochondrial DNA into the nuclear genome of human cancer cells

Frequent somatic transfer of mitochondrial DNA into the nuclear genome of human cancer cells Young Seok Ju, 1 Jose M.C. Tubio, 1,45 William Mifsud, 1,45 Beiyuan Fu, 2 Helen R. Davies, 1 Manasa Ramakrishna, 1 Yilong Li, 1 Lucy Yates, 1 Gunes Gundem, 1 Patrick S. Tarpey, 1 Sam Behjati, 1 Elli Papaemmanuil, 1 Sancha Martin, 1 Anthony Fullam, 1 Moritz Gerstung, 1 ICGC Prostate Cancer Working Group, 46 ICGC Bone Cancer Working Group, 46 ICGC Breast Cancer Working Group, 46 Jyoti Nangalia, 1,3,4 Anthony R. Green, 3,4 Carlos Caldas, 3,5 Åke Borg, 6,7,8 Andrew Tutt, 9 Ming Ta Michael Lee, 10,11 Laura J. van’t Veer, 12,13 Benita K.T. Tan, 14 Samuel Aparicio, 15 Paul N. Span, 16 John W.M. Martens, 17 Stian Knappskog, 18,19 Anne Vincent-Salomon, 20 Anne-Lise Børresen-Dale, 21,22 Jórunn Erla Eyfjörd, 23 Ola Myklebost, 24 Adrienne M. Flanagan, 25,26 Christopher Foster, 27 David E. Neal, 28,29 Colin Cooper, 30,31 Rosalind Eeles, 32,33 G. Steven Bova, 34 Sunil R. Lakhani, 35,36,37 Christine Desmedt, 38 Gilles Thomas, 39,44 Andrea L. Richardson, 40,41 Colin A. Purdie, 42 Alastair M. Thompson, 43 Ultan McDermott, 1 Fengtang Yang, 2 Serena Nik-Zainal, 1 Peter J. Campbell, 1 and Michael R. Stratton 1 1–43 [Author affiliations appear at end of paper.] Mitochondrial genomes are separated from the nuclear genome for most of the cell cycle by the nuclear double membrane, intervening cytoplasm, and the mitochondrial double membrane. Despite these physical barriers, we show that somatically acquired mitochondrial-nuclear genome fusion sequences are present in cancer cells. Most occur in conjunction with intranuclear genomic rearrangements, and the features of the fusion fragments indicate that nonhomologous end joining and/or replication-dependent DNA double-strand break repair are the dominant mechanisms involved. Remarkably, mitochondrial-nuclear genome fusions occur at a similar rate per base pair of DNA as interchromosomal nuclear rearrangements, indicating the presence of a high frequency of contact between mitochondrial and nuclear DNA in some somatic cells. Transmission of mitochondrial DNA to the nuclear genome occurs in neoplastically transformed cells, but we do not exclude the possibility that some mitochondrial-nuclear DNA fusions observed in cancer occurred years earlier in normal somatic cells. [Supplemental material is available for this article.] Somatically acquired structural rearrangements are common features of the nuclear genomes of cancer cells. These may range from simple chromosomal rearrangements (Campbell et al. 2008) to more complex, compound patterns, such as chromothripsis (Stephens et al. 2011) and chromoplexy (Baca et al. 2013), or mobilization of transposable elements (Lee et al. 2012; Tubio et al. 2014). Intrachromosomal rearrangements are generally more common than interchromosomal rearrangements, indicating a higher likelihood of joining a double-strand break in a chromosome to another break in the same chromosome despite the availabilityof a much larger quantityof nuclear DNA from other chromosomes (Stephens et al. 2009). In addition to the nuclear genome, human cells have a few hundred to a few thousand mitochondria, each carrying one or a few copies of the 16,569-bp-long circular mtDNA (Smeitink et al. 2001; Friedman and Nunnari 2014; Ju et al. 2014). During endosymbiotic co-evolution, most of the genetic information present in the ancestral mitochondrion has transferred to the nuclear genome (Gray et al. 1999; Adams and Palmer 2003; Timmis et al. 2004). An apparent burst of mtDNA transfer occurred during primate evolution ∼54 million years ago (Gherman et al. 2007) and occasional, probably more recent, transfer in humans has been observed in the germline (Turner et al. 2003; Goldin et al. 2004; Chen et al. 2005; Millar et al. 2010; Dayama et al. 2014). 44 Deceased. 45 These authors contributed equally to this work. 46 A full list of members is provided in the Supplemental Material. Corresponding author: [email protected] Article published online before print. Article, supplemental material, and publication date are at http://www.genome.org/cgi/doi/10.1101/gr.190470.115. Freely available online through the Genome Research Open Access option. © 2015 Ju et al. This article, published in Genome Research, is available under a Creative Commons License (Attribution 4.0 International), as described at http://creativecommons.org/licenses/by/4.0/. Research 814 Genome Research 25:814–824 Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/15; www.genome.org www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from Although mtDNA nuclear transfer in a HeLa cell line derivative, and thus occurring in vitro, has been reported (Shay et al. 1991), de novo nuclear transfer of mtDNA in animal somatic tissues has not previously been comprehensively studied to our knowledge. To investigate the possibility of somatic mitochondrial-nuclear DNA fusion, we analyzed next-generation paired-end DNA whole-genome sequencing data from 559 primary cancers, 28 cancer cell lines (referred as 587 cancer whole genome below) and normal DNAs from the same individuals (Supplemental Table 1). Results Discovery of somatic mtDNA transfers to cancer nuclear genomes From the 587 pairs of cancer and normal whole-genome sequencing data, we searched for cancer-specific clusters of discordant paired-end sequence reads in which one member of the read-pair mapped to the nuclear genome and the other to the mitochondrial genome, and then characterized the nuclearmitochondrial genome junctions to nucleotide resolution using individual sequence reads that bridged the junction (Fig. 1A). In 12 samples (overall positive rate 2.0%, 12 out of 587 samples), we observed 25 cancer-specific mitochondrial-nuclear DNA junctions (Table 1; Supplemental Figs. 1–6). Given that there are two junctions for a single integration event, we conclude that there are most likely 16 independent mtDNA insertions (Table 1). In addition to somatic transfers, we observed several novel rare germline (inherited) events that wereshared between cancerand paired normal samples (Supplemental Table 2; Supplemental Material). Breast cancer PD11372a showed a somatically acquired integration of almost the entire human mtDNA sequence (16,556 bp) into a highly amplified 2.75-Mb-long region of Chromosome 10q22.3. The integration event was strongly supported by both discordant and split read clusters (Fig. 1B–D) and was confirmed by shortand long-range PCR across the nuclear-mitochondrial genome junctions (Supplemental Figs. 7, 8; Supplemental Table 3). It was not found in normal tissue (blood) from the same individual or from all the other cases and did not match any known inherited nuclear mtDNA-like sequences (known as numts) (Gherman et al. 2007; Hazkani-Covo et al. 2010). Consistent with its somatic origin, the mtDNA fused to the nuclear genome harbored sequence polymorphisms identical to those present in the mitochondria of this individual (14,905 G > A; 15,028 C > A; 15,043 G > A; 15,326 A > G; 15,452 C > A, and 15,607 A > G). Fluorescence in situ hybridization (FISH) experimentsperformedonformalin-fixedparaffinembeddedtissueconfirmed that the fused DNA segment exists in the nuclei of cancer cells (Fig. 1E). In total, we found 10 primary cancers (1.8%, 10/559) and two cancer cell lines (7.1%, 2/28) with somatic mtDNA integrations into their nuclear genomes (Table 1; Supplemental Figs. 1–6). Of the 12 cancers, two (primary cancer PD13296a and cancer cell line NCI-H2087) had more than one mitochondrial-nuclear DNA translocation event. All integrations were supported by both discordantandsplitreadsandfurtherconfirmedbyPCRacrossthenuclear-mitochondrial genome junctions (Supplemental Fig. 7; SupplementalTable3).AllinheritedmtDNAsubstitutionpolymorphisms near these breakpoints were detected (Table 1). To further visualize the transfer events, we performed high-resolution FISH on stretched DNA fibers (fiber FISH) from the melanoma cell line, CP66-MEL (Fig. 2A). Somatic nuclear integration of mtDNA is frequently combined with other rearrangements of the nuclear genome The rate of somatic nuclear transfer of mtDNA may vary according to tumor type. Triple-negative breast cancer showed a fivefold higher frequency compared to estrogen-receptor (ER) positive breast cancers (6.2% and 1.2%, respectively; Fisher’s exact test P = 0.002). Triple-negative breast cancer genomes carry a higher number of chromosomal rearrangements than ER-positive breast (average 254 and 94, respectively, in our data set). As a result, there was a suggestive positive correlation between the number of chromosomal rearrangements and mtDNA transfers (Mann-Whitney Utest, one-sided P= 0.05) (Fig. 2B). The length of mtDNA fragments transferred ranged from 148 bp to entire mitochondrial genomes (16.5 kb) (Table 1). Interestingly, breakpoints in mtDNA were enriched near the mitochondrial genome heavy strand origin of replication (χ 2 test, P= 0.0005) (Fig. 2C). This suggests that the generation of mtDNA segments to be integrated into the nuclear genome is not random and may occur in a mtDNA replication-dependent manner (Lenglez et al. 2010). Of the 25 mitochondrial-nuclear DNA junctions, at least 17 (68.0%) were clearly associated with other nuclear chromosomal rearrangements (e.g., inversions, translocations, and large deletions) in the vicinity (Table 1; Supplemental Figs. 1–6). For example, with respect to PD11372a described earlier, genomic fragmentsfrom Chromosomes 10, 11, and mtDNA generated complex derivative chromosomes (Fig. 3A). In PD6047a, an mtDNA fragment was involved in chains of complex genomic translocations involving Chromosomes 6, 7, 11, 22, and X (Fig. 3B). In PD10014a, a local inversion was combined with the mtDNA integration event (Fig. 3C), and in PD4252a, a 16.5-kb mtDNA integration was found in a position on the X Chromosome from which ∼20 kb of nuclear DNA had been somatically deleted (Fig. 3D). Thus, mtDNA is often integrated into nuclear genomes in the vicinity of, or as part of, complex rearrangements. Although germline numts tend to occur near transposable elements such as LINEs and SINEs (Mishmar et al. 2004), we do not observe this association for somatic events (χ 2 test, two-sided P= 0.33) (Supplemental Table 4). The mechanism and timing of somatic nuclear transfer of mtDNA There was overlapping sequence microhomology (from 1 to 4 bp) in 20/25 breakpoints (80%) (Fig. 4A,B; Table 1; Supplemental Figs. 1–6), substantially more than expected by chance (χ 2 test, P=5× 10 −26 ). Thus, DNA sequence microhomology plays an important role in mitochondrial-nuclear DNA integration events, although blunt-end DNA repair was also observed. In two breakpoints, we also found nontemplated short-nucleotide insertions (1 and 4 bp long) (Fig. 4A; Table 1). Overall, these features are characteristic of DNA double-strand break repair by nonhomologous end joining (NHEJ) (Hastings et al. 2009). However, they do not rule out replication-based mechanisms switching template between nuclear and mtDNA, such as microhomology-mediated break-induced replication (MMBIR) (Liu et al. 2011). We investigated the timing of somatic mtDNA integration into the nuclear genome by assessing cases in which a metastatic sample had been sequenced in addition to the primary tumor. One such case (PD4252a) showed the mitochondrial-nuclear integration event in the primary but not in the metastasis (Fig. 4C), indicating that mtDNA transfer to the nucleus can occur after Nuclear integration of mitochondrial DNA in cancer Genome Research 815 www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from neoplastic transformation and during the course of subclonal evolution of the cancer. The other (PD6728b) showed it in both the primary and metastasis (Fig. 4C), suggesting that this event occurred in the common ancestral cancerclone or in normal somatic cells prior to neoplastic change. Nuclear transfer of mtDNA is unexpectedly frequent in human somatic cells To obtain a perspective on the frequency of mitochondrial-nuclear DNA translocation, we compared its rate to that of intranuclear 1 0 20 40 60 80 100 120 140 160 180 200 220 240 2 0 20 40 60 80 100 120 140 160 180 200 220 240 3 0 20 40 60 80 100 120 140 160 180 4 0 20 40 60 80 100 120 140 160 180 5 0 20 40 60 80 100 120 140 160 180 6 0 20 40 60 80 100 120 140 160 7 0 20 40 60 80 100 120 140 8 0 20 40 60 80 100 120 140 9 0 20 40 60 80 100 120 140 10 0 20 40 60 80 100 120 11 0 20 40 60 80 100 120 12 0 20 40 60 80 100 120 13 0 20 40 60 80 100 14 0 20 40 60 80 100 15 0 20 40 60 80 100 16 0 20 40 60 80 17 0 20 40 60 80 18 0 20 40 60 19 0 20 40 20 0 20 40 60 21 0 20 40 22 0 20 40 x 0 20 40 60 80 100 120 140 tumour blood 1 Mb 10 kb 100 bp 1 bp 1 bp 100 bp 10 kb 1 Mb PD11372a CTCCTGGGTG AGAAA CTCCTGGGTG TTGGCCTCAC GATAT TTGGCCTCAC Fusion Chr10 mtDNA TACTGTGGC CC AGACCTCTT ACACT GC AGACCTCTT TACTGTGGC CC CTCAG x 60 x 27 x 25 x 76 chr10 (+) 81,670,932 MT (-) 15,157 MT (-) 15,171 chr10 (+) 78,920,385 x 24 x 28 16,556 bp * * * * * * * * * * * * * * * * * * * * * * * ** chr10 position (Mb) 0 20406080100120 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 0 2 4 6 8 10 Copy number * ** * ** ** * * * * * * * * * * * * * * ** * ** * * ** ** * * * * ******* **** * * * * ** * * ** ** * * * * ** * ******* **** *** * * * * ** * * * * * * * * * * * ** * * **** * * **** *** * * ** ** **** **** *** ** ** * * * * ** * *** **** * * * * * * * * * * * * * * * * * * * * * * ** * ** **** * * * * * * * * * * * * * * * * * * * * * * * ** * * ** **** * * * * * ** * * ** * ** ** **** * * ** * * * ** ** * ** * ** * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * ** * * *** * ** * ** * * * * * * * * ** * * * * * * * * * * * * * * * * * * * * * * * * * * * * * ** * * * * * ** * * * * * * * ** ** * * **** **** * * * * * * * * * * * ** ** ** ** ** * * * * ** ** * ** * *** * ** ******** * ** ** **** ** *** *** * * * * * * * * * * * **** * * * ***** * *** * ********** * ** * * * * ** * * * * * * * * ** * * * * mtDNA PD11372a deletion type tandem duplication type inversion type (head-head) inversion type (tail-tail) mtDNA chr10: 80Mb chr10: 78Mb merged E DC BA mitochondrial DNA (inserted segment) (2) DRs (1) SRs (Nu) nuclear DNA nuclear DNA Mapping to the reference genome (using BWA) cancer genome (3) SRs (MT) nuclear (Nu) genome mitochondrial (MT) genome (2) DRs (1) SRs (Nu) mapped in Nu. genome as mate-unmmaped (3) SRs (MT) mapped in MT genome as mate-unmmaped Figure 1. Discovery of somatic nuclear mtDNA transfer from PD11372a. (A) The strategy for detection of nuclear mtDNA transfer events. See Methods for a detailed description. (SRs) Split-reads, (DRs) discordant reads, (Nu) nucleus, (MT) mitochondria. (B) Graphical representation of discordant read clusters in PD11372a and its paired-normal tissue (PD11372b). The red arrow indicates tumor-specific discordant-read clusters in Chr 10. Chromosome ideograms are shown in the outer layer. The distance between each discordant read and one prior to it (the inter-read distance) is plotted on the vertical axis on a log-scale in the middle (tumor) and inner layer (blood). Blue dots shown in the middle layer represent known numts. (C) mtDNA integration in PD11372a. Breakpoint sequences are shown.Red rectanglehighlights microhomology. Numbers of discordant split reads arepresented. Inherited mtDNA substitution polymorphisms are shown by red asterisks. (D) Rearrangement architectures of Chromosome 10 of PD11372a. DNA copy numbers are shown by black dots. The copy number for 2.75-Mb-long region fused with mtDNA is colored in red. Reads supporting rearrangements (large deletions, tandem duplications, tail-tail and head-head inversions) are shown by arcs and vertical lines. Chr 10-mtDNA fusions are shown with red arrows. (E) Nuclear FISH confirms the mitochondrial-nuclear DNA fusion in the nucleus. (Red) Chr 10 (80 Mb), (blue) Chr 10 (78 Mb), and (green) mtDNA. Ju et al. 816 Genome Research www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from Table 1. Summary of somatic mitochondrial-nuclear DNA fusions identified from 12 cancer samples Tissue Sample Left junction Right junction Frag. size (bp) Microhomology (bp,bp) Variants (#D/#P) a Context of rearrangement Nuclear MT MT Nuclear Primary PD11372a 10+:81,670,932] [M−:15,157 M−:15,171] [10+:78,920,385 16,556 (0,1) 6/6 mtDNA insertion with complex rearrangements PD4252a X+:45,631,665] [M+:14,450 M+:14,496] [X+:45,652,120 16,616 (2,1) 2/2 mtDNA insertion with large chr. deletion PD6047a X+:14,944,764] [M−:12,735 M−:16,128] [7+:96,923,229 13,177 (1,1) 6/6 Multiple interchromosomal translocations PD10014a 17−:75,618,348] [M−:13,365 M−:9055] [17+:75,688,733 4311 (2,3) 0/0 mtDNA insertion with Chr 17 inversion PD13296a 4+:102,463,870] [M−:14,705 M−:13,235] [4+:102,464,084 1471 (4,0) 0/0 mtDNA insertion with large chr. deletion 6+:103,639,248] [M+:14,692 M+:14,972]TA AT [6+:103,690,941 281 (2,0) 2/2 mtDNA insertion with large chr. deletion PD6728b 2+:138,664,890] [M−:13,199 M−:13,052] [2−:139,012,040 148 (4,2) 1/1 mtDNA insertion with complex rearrangements PD11397a 19−:12,650,382] [M+:16,233 M+:96] [17+:40,005,738 433 (0,2) 1/1 Multiple interchromosomal translocations PD7404a 1+:44,914,376] [M+:3732 –– >200 (1,–) 0/0 – PD6733b 6−:45,823,498] [M+:16,107 –– >200 (0,–) 1/1 – PD11768a 1−:144,944,326] [M+:16,104 –– >200 (4,–) 1/1 – Cell line CP66-MEL 3+:47,419,506] [M−:7048 M−:16,193] [3+:47,419,447 7425 (1,1) 1/1 mtDNA insertion NCI-H2087 10+:26,775,605] [M+:1690 –– >200 (1,–) 1/1 – 20−:33,836,717] [M−:5666 –– >200 (1,–) 1/1 – 17−:7,481,787] T [M−:3452 –– >200 (1,–) 1/1 b Multiple interchromosomal translocations 17−:31,744,235] [M+:4346 –– >200 (3,–) 1/1 – a Inherited mtDNA polymorphisms in the vicinity of breakpoints. (#D) Number of detected, (#P) number of present. b A somatically acquired heteroplasmic mutation in mitochondria. Nuclear integration of mitochondrial DNA in cancer Genome Research 817 www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from interchromosomal translocation, taking into account the sizes and copy numbers of the mitochondrial and nuclear genomes. Our sequencing data suggest that each cancer cell carries ∼500 copies of circular mtDNA (median value 495) (Fig. 5A), amounting in aggregate to ∼8 million base pairs (bp) of mtDNA (500 copies × 16.5 kb) enclosed by the mitochondrial double membrane in the cytoplasm of each cancer cell. The average frequency in the cancers analyzed of mitochondrial-nuclear DNA fusion was 5.1 × 10 −3 junctions per million bp of mtDNA, only half the average rate of intranuclear interchromosomal translocation (1.2 × 10 −2 junctions per million bp) and similar to that of Chromosomes 2, 4, and 13 (Fig. 5B). Given the multiple physical barriers to contact between the two genomes, the results indicate remarkably high rates of mtDNA escape, contact, and/or integration with nuclear DNA in human cancer cells. These appear to be considerably higher than in the germline across human evolution but comparable to those observed in Saccharomyces cerevisiae (Thorsness and Fox 1990) and for chloroplast DNA migration into the nucleus in tobacco plants (Methods; Supplemental Material; Huang et al. 2003). Discussion Despite multiple physical barriers, there are plausible mechanisms by which mtDNA and nuclear DNA could come into contact (Fig. 5C). Free mtDNA can be released into the cytoplasm from degradingmitochondriaoraftermitophagy (Zhang et al. 2008; Eiyama et al. 2013; Higgins and Coughlan 2014). Degradation of mitochondria may be accelerated in cancer cells due to hypoxia and increased energy demands (Zhang et al. 2008; Eiyama et al. 2013; Higgins and Coughlan 2014). Even without a bespoke molecular process for transportation, mtDNA could then, in principle, migrate to the nucleus during mitotic metaphase oranaphasewhenthenuclearmembrane has broken down. When these events are coupled with concurrent double-strand breaks (DSBs) and/or replication fork stalling of nuclear chromosomal DNA, mtDNA could be picked up and integrated into the nuclear genome as part of the process of rejoining DSBs (NHEJ) (Hastings et al. 2009) or used as an alternative DNA template in replication (MMBIR) (Liu et al. 2011). Micronuclei in cancer cells, which can be generated byerrorsinsegregationofmitoticnuclear chromosomes, may contribute to the events.Chromosomesinmicronucleifrequently undergo defective and delayed DNA replication, resulting in extensive fragmentation with subsequent jumbled rejoiningcomparedtotheiroriginalorder and orientation (Crasta et al. 2012; Forment et al. 2012). Thus, mtDNA fragments incorporated into micronuclei could end up fused to shattered nuclear chromosomes. It is worthy of note that mtDNA escapingto the nucleus canbe activelyused for DNA repair in Saccharomyces cerevisiae (Ricchetti et al. 1999; Yu and Gabriel 1999), particularly when error-free DSB DNA repair is not possible. Whether this applies in mammalian cells is unknown. Some of the somatic nuclear mtDNA integrations we identified are directly adjacent to nuclear genes. For example, nuclearmtDNA fusion in PD11372a occurred in the fifth intron of the KCNMA1 gene, a potassium channel frequently amplified in prostate and breast cancers (Oeggerli et al. 2012). However, we do not find obvious enrichment of the nuclear-mtDNA fusion breakpoints near human nuclear genes. RNA-seq from the NCI-H2087 cell-line indicates that mtDNA fragments in the nucleus of the cell line are not expressed as parts of mitochondrial-nuclear fusion transcripts.Thus, themajorityofthenuclearmtDNAtranslocation events are likely to be passenger events, similar to mutations of all other types in most cancer genomes. However, we do not excludethepossibilitythatsomeofthese eventsmayhavefunctional consequences in human cancer by generating fusion mRNA transcripts (Shay et al. 1991) and/or truncating cancer genes by mtDNA insertion within exons. chr3 (downstream)chr3 (upstream) mtDNA (7.4kb) A CP66-MEL (melanoma cell-line) 250 500 1000 B 78 133 125 Cancer tissue types Breast (ER +ve) Other breast Breast (triple −ve) Other types mtDNA transfers -+ cancer samples 0 1 2 3 4 0 2 4 6 8 10 12 14 0.5-3kb 3-5.5kb 5.5-8kb 8-11.5kb 11.5-14kb 14-0.5kb Ratio (obs/exp) Frequency of events Location of breakpoints in mitochondrial genome Expected Observed Ratio (obs/exp) P = 0.00052 * D-Loop rRNAs 16s12s CO1 ND5ND4 ND6 Replication origin ( H strand) Replication origin ( L strand) C CYB Figure 2. Features of somatic mtDNA nuclear transfer in 12 cancer samples. (A) Fiber FISH visualizes the mitochondrial-nuclear DNA fusion from the CP66-MEL cell line. (B) Positive correlation between mtDNA transfer and numbers of nuclear chromosomal rearrangements (large deletion, tandem duplication, inversion, and translocation) in cancer genomes. Median values are shown. (C) mtDNA breakpoints are enriched in the 14 kbto 500-bp region of the MT genome. (Top) Blue and red bars represent the expected and observed numbers of breakpoints in each interval of MT genome, respectively. Green line shows ratio between observed and expected numbers. A χ 2 test was applied to test enrichment. (Bottom) Schematic structural features of the MT genome corresponding to the intervals are shown. Ju et al. 818 Genome Research www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from This study has shown that fusion of mtDNA to nuclear DNA occurs in human somatic cells at a rate similar to that of translocation between nuclearchromosomes. Physical migration of mtDNA into the nucleus may be much more frequent in stem cells than ones in a terminally differentiated stage (Schneider et al. 2014). Further studies will need to address the mechanisms by which the apparent physical barriers to contact between mitochondrial and nuclear DNA are so effectively overcome. Methods Samples and sequencing data We analyzed 559 primary tumors and 28 cancer cell-lines in this study. Paired-normal samples for all the cancers were also included in this study in parallel. Whole-genome sequencesusedinthisstudyweregenerated by Illumina platforms (either Genome Analyzer or HiSeq 2000). Cancer genomesweresequencedtoatleast25×coverage. With respect to TCGA data, we downloaded aligned BAM files through UCSC CGHub (http://cghub.ucsc.edu). Sequencing reads were aligned on the human reference genome build 37 (GRCh37) and human reference mtDNA sequence (revised Cambridge reference sequence, rCRS) (Andrews et al. 1999), mainly by the BWA alignment tool (Li and Durbin 2009). SAMtools (Li et al. 2009) was used for manipulating sequence reads. Calling mitochondrial-nuclear DNA fusion events We employed a pipeline for identification of putative mtDNA translocation to chromosomal DNA (Fig. 1A). From paired-end whole-genome sequencing data of tumors, we extracted discordant reads (DRs), where one end aligned uniquely to mtDNA and the other end to nuclear DNA. In all cases, both ends must have a mapping quality greater than zero. Those discordant reads are clustered together using the following criteria: reads sharing (1) close alignment positions (<500 nucleotides) for both ends on nuclear and mtDNA, and (2) the same orientations.In order to remove false positives, we removed clusters supported by less than five discordant reads. In order to remove potential germline calls, several filters are applied to the tumor candidate cluster. The clusters from tumor cells were removed if they overlap with clusters identified from matched and/or unmatched normal tissues by more tolerable criteria (supported by more than one discordant read) from (1) its paired-normal tissue, and (2) from the other 586 unmatched normals. Filtered clusters were further refined with known germline human numts, a combined set from the human reference genome (hg19) detected by BLAT (Kent 2002) (n= 123) and from Simone et al. (2011) (n= 766). Finally, 25 clusters were selected as somatic candidates. Nucleotide-resolution breakpoints for the translocation junctions To obtain nucleotide-resolution breakpoints, we searched for splitreads(SRs) withoneof the endsspanning thejunctionof thetranslocation. We extracted “orphan”or “mate-unmapped”reads (one end of a read is unmapped by the BWA aligner) in the vicinity (<1000bp)ofdiscordant-readclustersonnuclearandmitochondrial genome sequences. Sequences from the unmapped end are then re-aligned by BLAT (Kent 2002), which enablessplit-read mapping. A 10 0 5 5 10 15 20 25 30 35 3 40 45 4 50 55 60 65 6 70 75 80 85 8 90 95 100 105 110 115 1 120 125 130 135 11 0 5 5 10 15 5 20 25 5 30 35 40 45 5 50 55 5 60 65 5 70 75 5 80 85 5 90 95 5 100 105 110 115 5 120 125 130 135 5 MT 0 PD11372a D B chrX(+):45,631,665 chrX(+):45,652,120 mtDNA insertion (16.5kb) nuclear DNA deletion (20 kb)  chromosome X 0 20 40 60 80 100 120 140 p11.3 Copy number 45.61 45.62 45.63 45.64 45.65 45.66 45.67 024 o oooo o oooooooo o ooooooooooooooo o o o o oo oo o ooo o oooooooooo oooooooooooooo o ooooooooooooooooooooo oo oooooooooooooo ooooooo ooo ooooooooooooooo o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o 0 8 aberrant read clusters PD4252a PD6047a 2 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 3 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 4 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 5 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 6 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 7 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 8 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 11 0 10 20 30 40 50 60 70 80 90 100 110 120 130 12 0 10 20 30 40 50 60 70 80 90 100 110 120 130 13 0 10 20 30 40 50 60 70 80 90 100 110 16 0 10 20 30 40 50 60 70 80 90 22 0 10 20 30 40 50 x 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 MT 0 mtDNA (-) chr7(+):96,923,229 chrX(+):14,944,764 13.1 kb mtDNA(-): 13,365 - 9,055 (4.3kb) chr17(+):75,688,733 local inversion C chr17(+):75,564,373 chr17(-):75,655,898 - 75,618,348 (37.6kb) mtDNA insertion PD10014a Figure 3. Concurrence of somatic mtDNA nuclear transfers with other structural variations. The complex web of rearrangements in the vicinity of mitochondrial-nuclear DNA fusions from four examples. (A) In PD11372a, mtDNA integration with complex rearrangements between Chr 10 and 11. (B)In PD6047a, mtDNA integration with complex rearrangements among Chr 6, 7, 11, 22, and X. (A,B) DNA copy numbers are shown by black dots with a log scale. Red lines represent translocations involving mtDNA. (C) In PD10014a, mtDNA integration combined with a local inversion (yellow). (D) In PD4252a, mtDNA integration with a local deletion. DNA copy numbers are shown with blue dots and lines. Aberrant read clusters (discordant and split reads) are shown by green and red arrows, respectively. Nuclear integration of mitochondrial DNA in cancer Genome Research 819 www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from Validation by PCR A PCR validation assay of the somatic mtDNA transfer was performed using genomic DNA from both cancer and paired-normal tissues. Primers were designed to amplify all the breakpoints (Supplemental Table 3). The short-fragment PCR reactions were performed as previously described (Tubio et al. 2014). With respect to long-range PCR, elongation time was increased 1 min per 1 kb. Generation of FISH probes Human bacterial artificial chromosomes (BAC) and fosmid clones used in this study were obtained from the clone archive team of the Wellcome Trust Sanger Institute. Plasmid DNA was prepared using the PhasePrep BAC DNA kit (Sigma-Aldrich). Human mtDNA was isolated from lymphoblastoid cells using a Mitochondrial DNA Isolation kit (Abcam). Probes for use in FISH were made as described before (Gribble et al. 2013). Purified mtDNA and plasmid DNA were first amplified using a GenomePlex Whole Genome Amplification (WGA) kit (Sigma-Aldrich) following the manufacturer’s protocols, then labeled using a WGA reamplification kit (SigmaAldrich) with a custom-made dNTP mix. Probes for interphase FISH were labeled directly with Aminoallyl-dUTPs - ATTO488, -Cy3, -Texas Red, and -Cy5 (Jena Bioscience); probes for fiber-FISH were labeledwithBiotin-16-dUTP,Digoxigenin11-dUTP (Roche), and DNP-11-dUTP (PerkinElmer). Validation by fiber-FISH with singlemolecule DNA fibers generated by molecular combing Single-molecule DNA fibers from the cancer cell line, CP66-MEL, were prepared by molecular combing (Michalet et al. 1997) following the manufacturer’s instructions (Genomic Vision). Briefly, the cells were embedded in a low-meltpoint agarose plug (1 million cells per plug), followed by proteinase K digestion, washing in 1 × TE (10 mM Tris, 1 mM EDTA, pH 8.0) and beta-agarose digestion steps. The DNA fibers were mechanically stretched onto saline-coated coverslips using a Molecular Combing System (Genomic Vision). For fiber-FISH, ∼500 ng of labeled DNA from each probe and 4 μg of human Cot-1 DNA (Invitrogen) were precipitated using ethanol, then resuspended in a mix (1:1) of hybridization buffer (containing 2 × SSC, 10% sarkosyl, 2 M NaCl, 10% SDS, and blocking aid [Invitrogen]) and deionized formamide (final concentration 50%). Coverslips coated with combed DNA fibers were dehydrated through a 70%, 90%, and 100% ethanol series and aged at 65°C for 30 sec, followed by denaturation in an alkaline denature solution (0.5 M NaOH, 1.5 M NaCl) for 1–3 min, three washes with 1×PBS (Invitrogen), and dehydration through a 70%, 90%, and 100% ethanol series. The probe mix was denatured at 65°C for 10 min before being applied onto the coverslips, and the hybridization was carried out in a 37° C incubator overnight. The post-hybridization washes consisted of two rounds of washes in 50% formamide/2 × SSC (v/v), followed by two additional washes in 2 × SSC. All post-hybridization washes were done at 25°C, 5 min each time. Digoxigenin-11-dUTP (Roche) labeled probes were detected using a 1:100 dilution of monoclonal chr4: GGCGA AACCC CATT TCTACT Fusion: GGCGA AACCC CATT GGTCGT mtDNA: TTTTT CATAT CATT GGTCGT TAAT A chr4(+)102,463,870 mtDNA(-):14,705 PD13296a (2 mtDNA nuclear integration events) mtDNA(-):13,235 chr4(+):102,464,084 4bp microhomology blunt-end DNA joining GCTACGATTT CTTTTGATGT : mtDNA GCTACGATTT AAATAACCAC : Fusion GTTATCTTCA AAATAACCAC : chr4 Integration #1 Integration #2 chr6(+):103,639,248 mtDNA(+) 14,692 mtDNA(+) 14,972 chr6(+):103,690,941 chr6: TTGTAAGA AC TAATAGAATG Fusion: TTGTAAGA AC AACCACGACC mtDNA:CACGGACT AC AACCACGACC GTAAATTATG GCTGAATCAT : mtDNA GTAAATTATG TAAT AAAATATTTG : Fusion CTGGGTCCTA AAAATATTTG : chr6 2bp microhomology non-template 4bp insertion PD6728b B chr2(+):138,664,890 mtDNA(-) 13,199 mtDNA(-) 13,052 chr2(-):139,012,040 chr2: TCATCT TGCT TGCGTTTTGC Fusion: TCATCT TGCT GCGAACAGAG mtDNA:GCAGAC TGCT GCGAACAGAG 4bp microhomology GGGGTGGGGC CT TCTATGGC : mtDNA GGGGTGGGGC CT GACTGCAG : Fusion CTTGGTCTTG CT GACTGCAG : chr2 2bp microhomology C PD4252 Fertilized egg Blood (PD4252b) MRCA LN metastasis (PD4252c) Primary locus (PD4252a; subclonal) mtDNA transfer PD6728 Fertilized egg Blood (PD6728a) MRCA LN metastasis (PD6728c; clonal) Primary locus (PD6728b; clonal) mtDNA transfer Transformation Transformation Figure 4. Nucleotide-resolution breakpoint sequences and the timing of somatic mtDNA nuclear integration. (A) Breakpoint sequences of nuclear-mtDNA fusions in PD13296a. Red rectangles highlight sequence microhomology and nontemplate nucleotides insertion. (B) Breakpoint sequences of nuclearmtDNA fusions in PD6728b. Red rectangles highlight sequence microhomology. (C) Phylogenetic trees showing the timing of somatic mtDNA nuclear transfers in PD4252 and PD6728 samples. (MRCA) Most recent common ancestor cell. Ju et al. 820 Genome Research www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from mouse anti-dig antibody (Sigma-Aldrich) and a 1:100 of Texas RedX-conjugated goat anti-mouse IgG (Molecular Probes/Invitrogen); DNP-11-dUTP (PerkinElmer) labeled probes were detected using a 1:100 dilution of Alexa 488-conjugated rabbit anti-DNP IgG and 1:100 Alexa 488-conjugated donkey anti-rabbit IgG (Molecular Probes/Invitrogen); biotin-16-dUTP (Roche) labeled probes were detected with one layer 1:100 of Cy3-avidin (Sigma-Aldrich). After detection, slides were mounted with SlowFade Gold mounting solution containing 4′,6-diamidino-2-phenylindole (Molecular Probes/Invitrogen). Images were visualized on a Zeiss AxioImager D1 microscope. Digital image capture and processing were carried out using the SmartCapture software (Digital Scientific UK). Nuclear interphase FISH Nuclei extraction from paraffin-embedded tissue of patient PD11372a and interphase-FISH followed Paternoster et al. (2002), with the exception that 60-μmthick sections were used in our study. The post-hybridization washes consisted of two rounds of washes in 50% formamide/2 × SSC (v/v), followed by two additional washes in 2 × SSC. Slides were mounted with SlowFade Gold mounting solution containing 4′,6-diamidino-2-phenylindole (Molecular Probes/ Invitrogen). Images were captured and processed as described above. Correlation between somatic mtDNA integration site and transposable elements We performed a study similar to the previous report (Mishmar et al. 2004). We calculated the distance between each mtDNA-insertion site (breakpoint) and its nearest transposable elements (either of SINE, LINE, LTR, simple repeat, or DNA transposon by RepeatMasker, downloaded from the UCSC Genome Browser, June 6, 2013). Then, each mtDNA-insertion site was categorized into one of four groups: (A) breakpoint within a transposable element; (B) breakpoint within 15 bp from a transposable element; (C) within 15–150 bp; and (D), >150 bp. In order to understand the positional enrichment of breakpoints from transposable elements, we randomly generated in silico breakpoint positions 40 times as many (total n= 1000) as we observed from each chromosome in the real data set. In silico breakpoints located within gaps of the human reference genome were removed and replaced by newly generated insertions. For these in silico-generated breakpoints, the distances from the nearest transposable elements were calculated and then categorized into one of the four groups (A, B,C, and D). Finally, the difference in the frequency of breakpoints in each group between the observed and in silico-generated data set was compared using a χ 2 test. Assessment of mtDNA copy numbers To understand mtDNA copy numbers in a cancer cell, we compared average read depth of coverage between 22 autosomes and mtDNA. With respect to the tumor sequences by whole-genome sequencing, average haploid autosomal coverage (RD autosome ) was obtained from the read depth of 2.685-Gb-long autosomal regions (excluding chromosomal gaps). Likewise, average mtDNA coverage (RD mtDNA ) was obtained from the read depth of the 500 1000 2000 Cancer tissue types Breast Osteosarcoma Other types Prostate A Estimated circular mtDNA copy numbers (in cytoplasm) per cancer cell 0.00 0.01 0.02 0.03 chromosomes Translocation rate (# of events per Mb) chr2 chr17 chr19 mtDNA chr4 chr13 B mitophagy mitochondrial degradation escape of mtDNA nuclear membrane breaks down beginning of mitosis DNA double-strand breaks (chromosome shattering) and/or replication fork stalling mtDNA migration to the nucleus (micronucleus) (rate> 2x10 pcpg) cell membrane -4 nucleus (micronucleus) mtDNA integration DSB repair (NHEJ, MMBIR) C Figure 5. Frequency and potential mechanisms of somatic mtDNA nuclear transfer in human cancer. (A) Estimated circular mtDNA copy numbers (in the cytoplasm) per cancer cell from 587 cancer tissues sequenced. The ratio of read depths between autosomes and mtDNA was used (see Methods). (B) Similar frequency of somatic nuclear mtDNA integrations compared to the frequency between autosomes (chromosomal translocation). (C) A model of somatic mtDNA transfer to the nuclear genomes. Nuclear integration of mitochondrial DNA in cancer Genome Research 821 www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from 16.5-kb mitochondrial genome. Finally, mtDNA copy number in a diploid cell (C mt ) is calculated as shown below: Cmt =2×RDmtDNA RDautosome . Assessment of translocation rate for autosomes and mitochondria We identified structural variations among nuclear chromosomes (large deletions, tandem duplications, inversions, and interchromosomal translocations) using the BRASS II algorithm (NikZainal et al. 2012), which identifies rearrangements by clustering discordant read pairs that point to the same junction and confirms breakpoints by local assembly of unmapped reads. The sensitivity and specificity of the BRASS II algorithm is equivalent to those values of the algorithm used for mitochondrial-nuclear DNA fusions (data not shown). We extracted interchromosomal translocations to calculate the rate of such events. The rate of each haploid autosome (Rtr,ch) is calculated as shown below: Rtr,ch(events per megabase)=Ntr,ch/(2×Lch)/Nsam, where Ntr,ch is the total number of somatic interchromosomal translocation junctions involving a specific chromosome, Lch is thelengthofthenonredundantregionofthechromosomeinmegabases, and Nsam is the total number of samples analyzed. To obtain the unique region length (Lch), we excluded redundant (or highly repetitive) sequence lengths from the ungapped length of each chromosome. Genomic regions classified in one or more of the three criteria shown below were defined as redundant, where translocation events could not be easilydetected due to ambiguous read alignment: (1) simple repeats, located by Tandem Repeats Finder (Benson 1999); (2) segmental duplications with moderate to high sequence similarity (≥95%) (Bailey et al. 2002), or (3) repetitive sequences including up to 10 different classes of repeats (such as SINE, LINE, LTR, DNA transposons, andmicrosatellites), located by the RepeatMasker program (http://www.repeatmasker.org), with a low divergence level (divergence < 5%). These nonredundant sequence regions were downloaded from the UCSC Genome Browser (http://genome.ucsc.edu). Similarly, the rate of mitochondrial-nuclear DNA translocations (Rtr,mt) was calculated as below: Rtr,mt(events per megabase)=Ntr,mt/(Cmt ×LmtDNA)/Nsam, where Ntr,mt is the total number of junctions of somatic mitochondrial-nuclear DNA fusions identified, C mt is the median value of mitochondrial genome copy numbers in a diploid cancer cell calculated above (495 copies), and L mtDNA is the length of the mitochondrial genome in megabases (0.016569 Mb). Assessment of the rates of nuclear mtDNA fusion and mtDNA escape to the nucleus Fusion of mtDNA to the nuclear genome requires at least two events, each of which could influence the rate of mitochondrialnuclearDNAfusion.TheseincludeescapeofmtDNAtothenucleus and integration to nuclear DNA. According to this model, the overall number of such fusion events can be calculated using the rates for these processes (ρ escape and ρ integration , respectively): Ntr =Nsam ×Ngen × r escape × r integration, where Ntr is the number of total somatic mitochondrial-nuclear DNA fusion events (n= 12), Nsam is the total number of cancer tissues (n= 587), and Ngen is the number of average cell generation from the fertilized egg. Using a reasonable assumption that Ngen= 1000, we obtain the rate of somatic mtDNA fusion to the nuclear genome (ρ escape ×ρ integration )tobe2×10 −5 per cell per cell generation (pcpg). With one more very conservative assumption that ρ integration is 0.1, we obtain ρ escape to be 2 × 10 −4 pcpg, or at least one escape event per 5000 cell generations. We hypothesize that the real ρ integration value is thought to be much lower than 0.1, which results in a higher ρ escape . For example, during the generation of knockout mice, homologous recombination allows one fixation event per 1000–10,000 microinjected DNA copies (Brinster et al. 1985). The integration rate may, however, be higher than the rate in cancer cells with defective homologous recombination-based repair and increased availability of nuclear double-strand breaks, which can be joined to by NHEJ or MMBIR. The mtDNA fusion to the nuclear genome in the germline (the rate of numts insertion) is around 5 × 10 −6 per germ cell per individual generation in previous phylogenetic studies (HazkaniCovo et al. 2010). The rate is equivalent to ∼5×10 −8 pcpg, given that the number of germ cell divisions per human generation is ∼100 (401 in males and 31 in females [Drost and Lee 1995]). Data access Sequence data for sample pairs with positive mtDNA nuclear transfer have been submitted to the European Genome-phenome Archive (EGA; https://www.ebi.ac.uk/ega/home). The study accession number is EGAS00001001234. Sample accession numbers are available in Supplemental Table 1. List of affiliations 1 Cancer Genome Project, Wellcome Trust Sanger Institute, Hinxton, Cambridge CB10 1SA, United Kingdom; 2 Cytogenetics Facility, Wellcome Trust Sanger Institute, Hinxton, Cambridge CB10 1SA, United Kingdom; 3 Cambridge University Hospitals NHS Foundation Trust, Cambridge CB2 0QQ, United Kingdom; 4 Department of Haematology, University of Cambridge, Cambridge CB2 0XY, United Kingdom; 5 Cancer Research UK (CRUK) Cambridge Institute, University of Cambridge, Cambridge CB2 0RE, United Kingdom; 6 BioCare, Strategic Cancer Research Program, SE-223 81 Lund, Sweden; 7 CREATE Health, Strategic Centre for Translational Cancer Research, SE-221 00 Lund, Sweden; 8 Department of Oncology and Pathology, Lund University Cancer Center, SE-221 85 Lund, Sweden; 9 Breakthrough Breast Cancer Research Unit, Research Oncology, King’s College London, Guy’s Hospital, London SE1 9RT, United Kingdom; 10 Laboratory for International Alliance on Genomic Research, RIKEN Center for Integrative Medical Sciences, 230-0045 Yokohama, Japan; 11 National Center for Genome Medicine, Institute of Biomedical Sciences, Academia Sinica, Taipei 115, Taiwan; 12 Department of Laboratory Medicine, Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, California 94158, USA; 13 Netherlands Cancer Institute, 1066 CX Amsterdam, Netherlands; 14 Department of General Surgery, Singapore General Hospital, Singapore 169608; 15 Department of Molecular Oncology, British Columbia Cancer Agency, Vancouver V5Z 1L3, Canada; 16 Department of Radiation Oncology and Department of Laboratory Medicine, Radboud University Medical Center, 6525 HP Nijmegen, Netherlands; 17 Department of Medical Oncology, Erasmus MC Cancer Institute, Erasmus Ju et al. 822 Genome Research www.genome.org Cold Spring Harbor Laboratory Press on October 16, 2016 - Published by genome.cshlp.orgDownloaded from