a Institut National de la Santé et de la Recherche Médicale (INSERM), U613; Etablissement Français du Sang (EFS) - Bretagne; Faculté de Médecine et des Sciences de la Santé, Université de Bretagne Occidentale (UBO), Brest, Franceb Institute of Medical Genetics, School of Medicine, Cardiff University, Heath Park, Cardiff CF14 4XN, UK
*Corresponding author: Jian-Min Chen INSERM U613 and EFS - Bretagne 46 rue Félix Le Dantec 29218 Brest, France Tel.: +33-2-9801-8102 Fax: +33-2-9843-0555 E-mail: jian-min.chen@univ-brest.fr
November 2011
Introduction Conventional models of carcinogenesis are invariably based upon the cumulative acquisition of point mutations and chromosomal rearrangements in the genome(s) of the tumour cells allowing them to override the natural restraints on their growth (Stratton et al., 2009). This notwithstanding, cancer may also evolve through the sudden acquisition of multiple concurrent or quasi-concurrent mutations provoked by some catastrophic intracellular event (Meyerson and Pellman, 2011). For example, a successive breakage-fusion-bridge cycle triggered by critical telomere attrition can result in the gene amplification often observed in different types of cancer (McClintock, 1941; Sahin and Depinho, 2010). The failure of cytokinesis can also lead to the formation of extensive aneuploidy which may also promote tumorigenesis (Fujiwara et al., 2005).
Chromothripsis: a new catastrophic phenomenon in cancer developmentHumpty Dumpty sat on a wall, Humpty Dumpty had a great fall. All the king's horses and all the king's men Couldn't put Humpty together again. [Traditional nursery rhyme.]
Using massively parallel paired-end sequencing, Stephens and colleagues have described a new catastrophic phenomenon: the presence of tens or even hundreds of structural rearrangements, involving spatially localized genomic regions, in primary cancer samples as well as cancer cell lines (Stephens et al., 2011). These alterations primarily affected a single chromosome, although in some cases, multiple apparently concomitant alterations involved several different chromosomes. Stephens et al. (2011) convincingly concluded that the massive yet spatially localized genomic rearrangements must have resulted from a single catastrophic event [termed chromothripsis (Greek, chromos for chromosome, thripsis for shattered into pieces)] rather than from a series of independent and progressive alterations. This conclusion was based upon the following three lines of evidence:
- Many positions across the rearranged chromosome exhibited copy number changes that alternated between just two states, namely one (indicating heterozygous deletion) or two copies (indicating no loss or gain). Monte Carlo simulations suggested that under the progressive model, multiple copy number states would have been expected. - Heterozygosity was retained in regions with high copy number. Under the progressive model, a deletion occurring early on in the process would have permanently removed heterozygosity. - Chromosomal breakpoints exhibited a significantly higher degree of clustering along the chromosome or chromosomal arm than would have been expected under a model of independently acquired alterations.
It would therefore seem that in a single catastrophic event, "the chromosome or chromosomal region shatters into tens to hundreds of pieces, some (but not all) of which are then stitched together by the DNA repair machinery in a mosaic patchwork of genomic rearrangements" (Stephens et al., 2011). The catastrophe model was given further support by two additional observations. First, sequencing of a relapse specimen from a chronic lymphocytic leukemia patient after alemtuzumab treatment, collected some 31 months after the initial sample, revealed that all rearrangements identified in the initial sample were also present in the relapse sample and, more importantly, no new rearrangements were apparent in the relapse sample. This finding suggested that the process generating the massive rearrangements was not ongoing in the patient and had been resolved before the first diagnosis was made (Stephens et al., 2011). Second, there existed a significant overlap between the somatic genomic rearrangements evident in primary colorectal cancers and those in their corresponding metastases, suggesting that many rearrangements occurred in the primary cancers (Kloosterman et al., 2011b).
The initiating cause and repair mechanism of chromothripsis Although both ionizing radiation (Stephens et al., 2011) and premature chromosome compaction during mitosis (Meyerson and Pellman, 2011) have been suggested as initiating causes, the primary cause of chromothripsis remains to be established. Irrespective of the precise cause, the massive genomic rearrangements resembling "random stitch of shattered pieces", together with the highly characteristic features of the junction sequences (i.e., predominantly blunt junctions or junctions with terminal microhomologies of <4-bp, as well as the presence of short non-templated sequences at some of the junctions), point to non-homologous end joining (NHEJ) as being the most plausible mutational mechanism (Kloosterman et al., 2011a; Kloosterman et al., 2011b; Stephens et al., 2011). NHEJ involves the ligation of any two broken DNA ends, the enzymatic components (i.e., nuclease, DNA polymerases and ligase) of its machinery being the most mechanistically flexible in their respective classes (Lieber, 2010). Indeed, as opined by Chen and colleagues (Chen et al., 2010), "NHEJ of ends from simultaneous double-strand breaks has the potential to account for a diverse range of genomic rearrangements". Moreover, NHEJ can occur at any time during the cell cycle. It should also be emphasized that, in principle, NHEJ does not involve the synthesis of long templated sequence tracts that will be detected as duplications. In short, NHEJ often results in the loss of genetic material yielding deletions, balanced rearrangements such as inversions and translocations, or a combination of any of these (Chen et al., 2011).
Not all components of the observed massive rearrangements have necessarily been derived from a single catastrophic event The above notwithstanding, Stephens and colleagues did not argue that every component of the massive yet spatially localized genomic rearrangements would have necessarily been generated in one single mutational event but simply that the majority of these rearrangements probably occurred in a single event. For example, some regions of the affected chromosomes in some samples appeared to have been duplicated (i.e., copy number = 3), a finding which was assumed to have resulted from the later partial duplication of the derivative chromosomes (Stephens et al., 2011). Alternatively, such a duplication may have indeed resulted from NHEJ repair of simultaneously generated double-strand breaks, provided that the sister chromatid or homologous chromosome was also involved (Chen et al., 2011). Nonetheless, on its own, NHEJ cannot readily explain rearranged regions with a copy number of ≥4 (Chen et al., 2011). In this regard, Magrangeas and colleagues recently claimed that genomic rearrangements in some 1.3% (10/764) of their primary multiple myeloma samples recapitulated all the hallmark characteristics of chromothripsis (Magrangeas et al., 2011). However, these samples were often found to contain regions with a copy number of 3, 4 or 6. For samples exhibiting a copy number higher than 4, additional independent alterations may have occurred in the chromothripsis-derived chromosomes. Alternatively, the complex rearrangements in some of these samples may have resulted from the breakage-fusion-bridge cycle or alternatively could be explicable in terms of replication-based mechanisms (see below). Characterization of the breakpoint junctions in these samples may help to clarify the underlying generative mechanisms.
Replication-based mechanisms in the generation of chromosome catastrophes Since 2005, replication-based models including serial replication slippage (SRS) (Chen et al., 2005a; Chen et al., 2005b; Chen et al., 2005c), fork stalling and template switching (FoSTeS) (Lee et al., 2007) and microhomology-mediated break-induced replication (MMBIR) (Hastings et al., 2009; Sheen et al., 2007) have been used increasingly to account for the generation of diverse complex genomic rearrangements involving duplications, triplications or more copy number gains (Chen et al., 2011). The hallmark characteristic of these replication-based models in generating complex genomic rearrangements is SRS or serial template switching during a single cell cycle. All the steps of template switching are thought to be microhomology-dependent. However, whereas SRS or FoSTeS relies upon a stalled replication fork (i.e., the original templated strand remains intact during the process), MMBIR is predicated upon a collapsed replication fork (i.e., one-ended DSB). Given that nicks, which will be transformed to DSBs when encountered by a replication fork, are common in all living cells, MMBIR is likely to have considerably greater explanatory potential than either SRS or FosTes (Chauvin et al., 2009). The potential contribution of replication-based mechanisms to genome instability received a new impetus from a recent study, in which Liu and colleagues investigated 17 subjects with various development abnormalities by means of high-resolution genome analysis (Liu et al., 2011). Constitutional multiple copy number changes, including deletions, duplications and/or triplications, as well as inversions were observed in all cases. In each case, all rearrangements occurred within a single chromosome; in 15 of the 17 cases, the rearrangements were localized to the distal half of the affected chromosomal arms. In particular, genomic rearrangements in four cases turned out to be extremely complex. For example, patient BAB3105 had a total of 18 copy number changes on one or other of the chromosome 9 homologs; the rearrangement pattern was nml-dup-nml-dup-nml-dup-nml-dup-nml-dup-nml-dup-nml-dup-nml-dup-nml-dup-trp-dup-nml-dup-nml-trp-dup-nml-dup-nml-dup-nml-dup-nml-dup-nml (nml, normal copy; dup, duplication; del, deletion; tri, triplication). FISH and breakpoint junction data suggested that all additional copies of the duplicated and triplicated segments were randomly joined, forming a large "breakpoint junction cluster" on 9q21. By analogy with the phenomenon of chromothripsis, the observation of these extremely complex rearrangements in a single chromosome was also described as a chromosome catastrophe event (Liu et al., 2011). However, this kind of chromosomal change cannot be easily explained by the previously described NHEJ repair of simultaneously generated double-strand breaks. Sequencing of several breakpoint junctions in patient BAB3105 revealed the frequent occurrence of relatively long templated insertions (54-1542 bp), microhomologies and inversions at these junctions. Taking this and the aforementioned observations together, Liu and colleagues envisaged the involvement of a replicative mechanism in the generation of this complex chromosome catastrophe event which comprised multiple duplications and/or triplications; they regarded MMBIR as the most likely underlying mechanism. They further suggested that a potential replication fork collapse at 9q21 could account for the breakpoint clustering therein (Liu et al., 2011). All the extremely complex rearrangements on chromosome 9 were confirmed, by parental array comparative genomic hybridization and chromosome analyses, to have occurred de novo in patient BAB3105. All the additional copies of the duplicated and triplicated genomic segments could then have been derived from the paternal allele that was not transmitted by high-density informative single-nucleotide polymorphism analysis in the case-parent trio (Liu et al., 2011). Therefore, the rearrangements must have occurred in the father, either during early development as a postzygotic event or in a germline during spermatogenesis; further, the generation of this complex rearrangement event must have involved the two paternal chromosome 9 homologues (Liu et al., 2011). Liu and colleagues opined that chromothripsis is more likely to constitute an inherent cellular DNA replicative/repair process designed to maintain genome stability rather than simply constituting the "blowing apart" of a chromosome followed by putting the pieces of the puzzle together. Consequently, they proposed that the phenomenon termed chromothripsis might be more appropriately referred to as "chromoanasynthesis" (chromosome reconstitution or chromosome reassortment). In our view, there is a subtle distinction between the two concepts; whereas chromothripsis emphasizes the initial cause of genome instability, chromoanasynthesis emphasizes the organism's capability to cope with genome instability (Chen et al., 2011). The Liu finding has important implications for cancer development; "on a cellular level, the mechanisms underlying these DNA rearrangements occurring in cells at different stages of the human life (i.e., germline, postzygotic development, somatic differentiated cells) are likely to be the same" (Liu et al., 2011).
The role of chromothripsis should not however be overestimated Whereas the Stephen et al. study (2011) has established an entirely new model for cancer development, the role of chromothripsis should not however be overestimated. The simple reason is that the vast majority of the cells harboring such extensive genomic rearrangements would be likely to die. In other words, only those cells that acquired a newly-rearranged chromosome that happened to confer some growth advantages e.g. loss of tumour suppressor genes, amplification of oncogenes or a combination of these (Stephens et al., 2011), could survive. Indeed, in the Stephen et al. study, although chromothripsis was observed in up to ~25% of primary bone cancers, its detection frequency in many other cancer subtypes was only 2-3%. Moreover, the pairwise comparison of structural changes in colorectal tumors has demonstrated that primary and metastatic colorectal cancer genomes harbor distinct patterns of structural variation (Kloosterman et al., 2011b). Furthermore, different cancers may have different rearrangement profiles. For example, rearrangements observed in breast or pancreatic cancer have a tendency to be either distributed fairly randomly over the genome or, if localized, to be associated with substantial genomic amplification (Stephens et al., 2011).
Conclusions and perspectives The Stephen et al. study has not only significantly improved our understanding of cancer development but has also provided competing evidence that tens or even hundreds of double-strand breaks could be simultaneously repaired by the organism. With the wide application of next-generation sequencing, we may reasonably expect to see many more examples that are compatible with the chromothripsis phenomenon as well as various novel mechanisms of mutagenesis. It would also be worthwhile to attempt to experimentally replicate chromothripsis as a means to determine the exact initiating causes and repair mechanisms. Finally, the concept of chromothripsis also appears to be capable of explaining the generation of some complex de novo structural rearrangements in the germline (Kloosterman et al., 2011a) and could be helpful in understanding the mutational mechanisms underlying some previously reported germline complex rearrangements (Chen et al., 2011).
Acknowledgement This article is based partly on a review article by the authors, entitled "Transient hypermutability, chromothripsis and replication-based mechanisms in the generation of concurrent clustered mutations", published in Mutation Research - Reviews in Mutation Research (2011) doi:10.1016/j.mrrev.2011.10.002.