The cellular etiology of chromosome translocations

Abstract

Chromosome translocations are the most severe form of genome defect. Translocations represent the end product of a series of cellular mistakes and they form after cells suffer multiple DNA double strand breaks (DSBs), which evade the surveillance mechanisms that usually eliminate them. Rather than being accurately repaired, translocating DSBs are misjoined to form aberrant fusion chromosomes. Although translocations have been extensively characterized using cytological methods and their pathological relevance in cancer and numerous other diseases is well established, how translocations form in the context of the intact cell nucleus is poorly understood. A combination of imaging approaches and biochemical methods to probe genome architecture and chromatin structure suggest that the spatial organization of the genome and features of chromatin, including sequence properties, higher order chromatin structure and histone modifications, are key determinants of translocation formation.

Introduction

In 1973, Janet Rowley identified an aberrant chromosome as the cause of chronic myelogenous leukemia (CML) [1]. This momentous discovery changed the field of cancer biology forever by demonstrating that a chromosome defect could cause cancer. The now famous Philadelphia chromosome consists of a reciprocal fusion of parts of chromosome 9 and chromosome 22 and is a prototypical example of a cancer-causing chromosome translocation. The juxtaposition of the ABL gene, encoding a tyrosine kinase, on chromosome 9 and the BCR gene on chromosome 22, leads to the generation of a chimeric fusion protein with constitutive, oncogenic kinase activity [1,2]. Since the discovery of the Philadelphia chromosome, thousands of chromosome translocations have been characterized [3]. In addition to the generation of fusion proteins as in the case of BCR-ABL, translocations may lead to gene disruption or misregulation as seen in Burkitt’s lymphoma, where enhancer elements of one of three immunoglobulin loci (IGH, IGK, or IGL) become juxtaposed to the MYC gene leading to its constitutive activation [4,5]. While chromosome translocations have been particularly well characterized in blood cancers such as CML, they are similarly relevant and frequent in solid tumors as well as in non-cancerous diseases including infertility and schizophrenia [3].

The formation of a translocation requires three basic steps: 1) the occurrence of multiple DNA double strand breaks on distinct chromosomes, 2) the physical association of the broken ends, and 3) the rejoining of the broken partner chromosomes. Despite the undisputed pathological relevance of translocations, how these steps occur in vivo and in the context of the intact cell have remained largely unknown. Recent work is beginning to shed light on two key questions: What determines which chromosomes undergo translocation with each other and what determines where chromosomes break in the first place?

What determines the choice of translocation partners?

The genesis of a chromosome translocation requires that two DSBs come into physical contact to allow the illegitimate misjoining of the chromosome ends (Fig. 1). Since the physical interaction of the involved chromosomal loci is a fundamental step in the formation of translocations, their spatial arrangement is likely to directly contribute to translocation frequency [6]. Cytogenetic and biochemical evidence supports this notion.

Figure 1. The presence of breaks and the spatial arrangement of chromosomes influence translocation frequency.

Translocations cannot form in the absence of breaks. In the presence of breaks, the spatial positioning of the broken chromosomes affects translocation outcome. Proximal breaks translocate with high frequency. Distal breaks can also translocate, albeit at lower frequency.

Cytogenetic studies

Numerous cytogenetic studies have pointed to a strong correlation between spatial proximity of chromosomes or genes and their translocation frequencies by showing that proximal genome sites are more likely to form translocations than distal ones [7–12] (Fig. 1). As an example, in mouse lymphoma, translocations often involve chromosomes 12, 14, and 15, and these chromosomes are found with high frequency in a spatial cluster in normal mouse splenocytes when mapped by fluorescence in situ hybridization (FISH) [11]. Similarly, human chromosomes 4, 13 and 18, which all preferentially localize to the nuclear periphery, frequently translocate with each other, but they do not translocate with internally localized chromosomes, with which they are not in physical proximity [7]. Interestingly, translocation frequency also correlates with the degree of intermingling between chromosomes [8], strongly suggesting that the local arrangement of DSBs drives translocations. The same type of correlation applies to individual genes [9,12–14]. For example, the spatial proximity of the MYC gene, relative to its possible translocation partners IGH, IGK and IGL in Burkitts’ lymphoma, directly correlates with the observed frequency of these translocations in patients [12]. Many other similar examples exist [15–17].

Both the spatial arrangement of genomes and the occurrence of translocations is tissue- and cell-type specific. The comparison of translocation patterns and spatial genome organization amongst tissues further supports a role of genome organization in translocations [3,18,19]. In particular, tissue-specific translocation frequencies correlate with tissue-specific organization patterns [10,20]. In mice, for example, chromosomes 12 and 15 which frequently translocate in lymphomas are proximal in lymphocytes but not in hepatocytes, whereas chromosomes 5 and 6 which often translocate in hepatomas, are proximal in hepatocytes, but not in lymphocytes [20]. These correlations led to the proposal that tissue-specific genome organization is a major driver of chromosome translocations [18].

Genome-wide studies

While these studies provide evidence for the contribution of spatial genome organization as a determinant of the outcome of translocations, their correlative and retrospective analysis assumed that these regions form translocations, without, however, demonstrating it directly. Moreover, tumorigenic translocations are usually clonal and highly selected, and thus correlations may not accurately mirror the contribution of spatial organization to translocation frequency. Several recent studies overcame this limitation by capturing the genome-wide landscape of translocations in the absence of selection [21,22**]. These studies confirm the earlier morphological observations.

Sequencing of junctions of translocations, formed by experimentally induced single DSB at the c-myc or the Igh locus in primary B lymphocytes [21,22**], indicates that translocations occur most frequently on the same chromosome, whereas translocations with other chromosomes are much rarer [21,22**]. Considering that genome regions on the same chromosome are more proximal than loci on other chromosomes, these results highlight the notion that the relative distance of translocating partners determines formation of translocations. This interpretation is further supported by studies which used chromosome conformation capture techniques to map physical interactions on a genome-wide scale [23–25**]. In one approach, DSBs were created by integration of the ISceI restriction site in transformed pro-B cells expressing the RAG (recombination activated gene) endonuclease which cleaves the endogenous Igk locus and other sites. Translocation frequencies were then mapped and compared to the spatial arrangement of chromosomes [23**]. The most frequent translocation partners for the ISceI-induced DSBs were found within the endogenous Igk locus and within other RAG-target loci, suggesting that, as expected, the formation of DSBs is a pre-requisite for the formation of translocations. In support of this notion, the comparison of a genome-wide map of translocations involving Igh and Myc and the physical location of breaks throughout the genome in activated B lymphocytes showed that in the presence of the activation-induced cytidine deaminase AID, which triggers DSBs, the number and location of DNA breaks govern the rate of the chromosome translocations rather than the nuclear interactions [24**]. In contrast, using the same translocation capture sequence data set, a recent study arrived at the opposite conclusion, showing that the majority of AID-induced hotspots are found in domains that contact Igh at high frequency [25**]. However, when DSBs are not the limiting factor [23**], or in the absence of recurrent AID-induced DNA damage [24**], the frequency of translocations was directly related to the frequency of their pre-existing contacts. Taken together, these observations suggest that once DSBs have formed, the spatial arrangement of the chromosomal loci within the nuclear space determines their translocation frequency (Fig. 1).

Spatial proximity and translocations

Two models have been put forth for how translocations form within the nuclear 3D space [9,26,27]. The “breakage-first” model envisions that DSBs from distant locations are able to move towards each other over long distances and are then joined to form a permanent translocation. In an alternative “contact-first” model, joining of broken ends preferentially occurs between chromosomal loci that are found in close proximity before the formation of the breaks. The morphological and biochemical observations strongly support the contact- first model. Additional evidence comes from studies showing that DSBs have limited mobility within the mammalian nucleus [28–31]. Typically, in mammalian cells a DSB undergoes limited local motion with a mean squared displacement of ~1μm²/h, comparable to that of a locus on an intact chromatin fiber [28,29,32]. In contrast, similar experiments in yeast S. cerevisiae indicated increased chromosome mobility of persistent DSBs compared to intact chromosomal loci [33,34]. The observed increase was dependent on factors involved in steps of the homologous recombination (HR) repair pathway, presumably to facilitate homologous pairing during recombination [33,34]. Together with the finding that the mammalian DSB-repair protein 53BP1 promotes the end-joining of dysfunctional telomeres by increasing the local chromatin mobility [35], these studies indicate the involvement of key players of the DSB-repair pathways in controlling DSB mobility. However, even in the much smaller yeast nucleus, spatial proximity appears to play a role in determining recombination outcomes as illustrated by the fact that the MAT mating locus preferentially recombines with its most proximal potential partner rather than a distant potential partner [36].

Although cytogenetic observations, genome-wide mapping, and motion studies suggest that most translocations can be explained by the contact-first model, it is possible that distal breaks may also form translocations, but likely with reduced frequency (Fig. 1). One argument in favor of translocation formation from distal breaks is the observation of occasional long-range, apparently directed, motion of gene loci in living cells [37] and the observed ability of chromosome domains containing DSBs to move over several micrometers and cluster within the mammalian nucleus [38]. Moreover, in S. cerevisiae multiple DSBs coalesce into common repair centers [39]. While this focal assembly of repair proteins may increase their local concentration and affect the efficiency of repair [39], the spatial proximity of the involved breaks may also facilitate illegitimate misjoining. In mammalian cells, while clustering of few repair foci marked by 53BP1 has been observed in a limited number of cells [28], there is no indication that DSB clustering is the norm [28,31] and if it does happen it is likely reversible [29]. It would be interesting to assess whether congregation of repair foci in common focal centers contributes to the formation of chromosome translocations by clustering DSBs.

Why do chromosomes break where they break?

While spatial and temporal proximity of chromosomes is an essential determinant in the formation of translocations, it remains largely unclear what upstream factors predispose genomic regions to breakage and translocations in the first place. Circumstantial evidence suggests that DNA sequence features as well as chromatin properties may facilitate breakage susceptibility of genome regions (Fig. 2).

Figure 2. DNA and chromatin features in breakage susceptibility.

DNA sequence features (green), histone modifications (yellow), and chromatin structure (red) may facilitate breakage susceptibility and represent an important upstream event in translocation formation. (right) The combined effect of sequence and chromatin features is evident in prostate cancer, where liganded androgen receptor (AR) recruits AID and TOP2B to translocation breakage sites.

DNA sequence

DNA features that influence breakage may be sequence and/or structure related. In support of this notion, certain DNA sequences are recognized by endogenous nucleases leading to the formation of DSBs and translocations. RAG1/2 are endonucleases which create DSBs during V(D)J recombination in B- and T-cells. Translocations may form when RAG enzymes misrecognize sequences that resemble recombination signal sequences (RSS) normally found in V(D)J regions [41–43] (Fig. 2). In germinal center B cells, AID recognizes a single-stranded sequence motif during the transcription of regions involved in somatic hypermutation and class switch recombination and promotes DSBs to generate antibody diversity [44], however, misrecognition of non-Ig targets can lead to translocations [45]. AID-induced translocations were first observed in germinal center-derived B cell lymphomas, but have recently been discovered in other B cell lymphomas as well as in some solid tumors [46]. AID has also been suggested to contribute to translocations in other ways than misrecognition of RSS. In prostate cancer, AID is co-recruited with liganded androgen receptor (AR) to AR-binding DNA sequences, sensitizing them to DSB breaks and leading to the formation of translocations in the presence of genotoxic stress [47**] (Fig .2). Furthermore, several genome-wide studies have implicated off-target AID binding sites that may play a role in the formation of translocations [21,22,48–50]. These observations suggest that mis-recognition of sequences by cellular endonucleases promotes genome breakage.

What are the sequences most prone to breakage? CpG islands are one candidate. While representing only 1% of the human genome, CpG dinucleotides are present in 40–70% of bcl-2 and bcl-1 breakpoints in pro-B and pre-B lymphocytes [51], leading to the suggestion that CpGs are targeted by AID and RAG endonucleases. However, CpGs are not associated with translocations in other cell types including lymphoid-myeloid progenitors, mature B cells, and T cells [51], suggesting that if CpG islands do facilitate breakage, their presence is not sufficient to promote them and does not do so in all tissues.

Alu repeats which constitute an estimated 11% of the human genome have been proposed to serve as recombination hotspots for translocations by virtue of non-allelic homologous recombination [52]. However, in an engineered system to quantify translocations, the introduction of identical or divergent Alu repeats adjacent to induced DSB sites did not alter the frequency of translocations [53], suggesting that the presence of homology per se is not a driver of translocation frequency [53,54]. In support, the presence of Alu elements at sequenced translocation junctions in patient cases has been sparse and anecdotal, further suggesting that Alu elements are not universal markers of breakpoints [55].

Common fragile sites (CFS) have also been linked to translocations [56]. CFSs are cytologically defined regions of chromosomes containing gaps and constrictions in metaphase under partial replication stress, and these regions have been shown to be prone to breakage [57]. A recent large scale analysis of 746 cancer cell lines revealed extensive co-incidence of fragile sites with regions of cancer-causing homozygous deletions strongly supporting their role in tumorigenesis [58]. A potential link between fragile sites and translocation formation comes from the observation that exposure of thyroid cells to chemicals that induce fragile sites promotes RET/PTC translocations [59]. Although no single mechanism appears to account for the emergence of CFSs, some common sequence features have been identified. CFSs are enriched in strings of AT-dinucleotide repeats that give these regions high DNA helix flexibility and the ability to form stable non-B DNA secondary structures, which may inhibit DNA replication [60]. Indeed, translocations have been postulated to form at AT palindromic sequences through a mechanism involving cruciform DNA structures that may be prone to breakage [61], and computational analysis of five translocation genes (CBFB, HMGA1, LAMA4, MLL, and AFF4) revealed significantly higher AT content than control regions [62].

More direct evidence for DNA secondary structures in breakage and translocations was the discovery that the major breakpoint region of bcl-2 adopts a stable non-B DNA structure that is targeted by RAG in a sequence-independent manner [63**]. By containing stable regions of single-strandedness, this DNA structure promotes RAG-mediated cleavage of the bcl-2 locus and formation of the t(14;18) translocation in follicular lymphoma [63]. This secondary structure may be a “G-quadruplex,” a four-stranded DNA structure that can spontaneously form in G-rich sequences [64]. Further support that non-B DNA structures contribute to genomic instability and translocations comes from mouse studies in which the integration of sequences that form triplex H-DNA or left-handed Z-DNA increased chromosome breakage, deletion, and translocation events [65].

Topological features of DNA may also contribute to breakage susceptibility. Topoisomerase II (TOP2) generates a transient DSB to regulate under- and overwinding of DNA, for example in mitotic chromosomes and in replication, but also during transcription [66,67 ]. The normally beneficial function of TOP2 may at times, however, have detrimental effects. The TOP2 beta isoform has been shown to associate with androgen receptor upon transcriptional activation and to trigger DSBs at TMPRSS2 and ERG breakpoints in prostate cancer [68**] (Fig. 2). Interestingly, cancer patients who are treated with TOP2 poisons such as epipodophyllotoxins, which potentiate the DNA cleavage activity of TOP2, form therapy-related leukemias harboring characteristic balanced chromosome translocations most often involving the MLL gene and a partner gene [66,69].

Chromatin structure and histone modifications

Circumstantial evidence suggests that various aspects of chromatin may play a role in chromosome breakage susceptibility and translocations. Genome-wide mapping of translocating regions after a single DSB is introduced at the c-myc or Igh locus in primary B lymphocytes found that DSBs occur primarily in transcriptionally active regions [21,22**]. Along the same lines, two studies have documented breakpoints in, or near, transcriptionally active genome regions [47,70**]. In anaplastic large cell lymphoma, several genes in the vicinity of translocation breakpoints are highly expressed before translocations occur [70]. Similarly, liganded androgen receptor, a potent transcriptional activator, binds near the breakpoints of TMPRSS2, ERG, and ETV, which are involved in translocations in prostate cancer and under genotoxic stress, induces translocations [47**]. One interpretation of these observations is that chromatin remodeling and binding of transcription factors may predispose genomic regions to breakage and translocations [71].

Histone modifications modulate transcription, replication, DSB repair, and recombination, making them potential candidates in DSB susceptibility and translocation mechanisms [72]. H3K4me3 has been implicated in both RAG and AID-mediated DSB mechanisms. The RAG2 plant homeodomain finger binds to H3K4me3 at the Ig locus in V(D)J recombination, and mutation of this domain greatly diminishes the efficiency of recombination [73–75]. Furthermore, H3K4me3 stimulates RAG activity at sites other than its natural recognition site, especially at cryptic sites [76]. In T-cells, H3K4me3 peaks at cryptic RAG binding sites in certain translocation breakpoints and it has been proposed that this binding promotes translocations in T-cell leukemias [76]. Similarly, genome-wide analysis of AID-induced cleavage sites identified four non-immunoglobulin genes that accumulate high rates of mutations and participate in translocations [48]. Like the natural immunoglobulin target genes cleaved by AID [77], three of these four non-immunoglobulin genes featured enrichment of H3K4me3 at their breakage sites, as well as clusters of repeat DNA sequences. Genome-wide changes in H4K20 monomethylation in mice led to defective DSB repair, Ig class-switch recombination and to translocations involving the IgH locus [78]. H3K79 methylation has also been implicated in DNA recombination and possibly translocation formation. In prostate cancer cells, liganded androgen receptor induces enrichment of H3K79me2 at breakpoint regions, and overexpression of the H3K79 methyltranferase DOT1L increases the frequency of translocations in the presence of androgen and genotoxic stress [47**]. Genome-wide mapping of a set of histone modifications in a prostate cancer cell line has indicated possible enrichment of active chromatin marks, H3K4me3, H3K36me3, and acetylated H3, over the TMPRSS2-ERG translocation region [79]. In addition, the regions corresponding to prostate breakpoints other than TMPRSS2-ERG showed an entirely different pattern in that they were depleted of these active marks and instead enriched in the repressive mark H3K27me3, indicating that the relationship between histone marks and breakage susceptibility is complex [79]. Nevertheless, the sum of these observations points to the possibility of a combined effect of DNA sequence features, chromatin structure and histone modifications in chromosome breakage susceptibility.

Conclusions

Chromosome translocations are one of the most severe forms of genomic damage and their pathological relevance is undeniable. Yet, 40 years after the discovery of the first cancer-causing translocation, our understanding of their genesis is still rudimentary. A major reason for the lack of mechanistic insight into translocation formation has been the absence of sharp experimental tools to probe them. The recent application of chromosome conformation and deep-sequencing methods complements the traditionally used imaging approaches and extends these studies to a global, genome-wide level. But many questions remain; none maybe more pressing than what the temporal aspects of translocations are. How long does it take for a translocation to form? How rapidly do two unrepaired DSBs find each other? How do they move within the nuclear space? These questions will require the visualization of translocation formation in living cells. Such approaches, which are becoming technically feasible, will also allow probing of individual steps in translocation formation and to experimentally manipulate factors that affect translocation frequency, particularly chromatin structure and histone modifications.