A model of oncogenic rearrangements: differences between chromosomal translocation mechanisms and simple double-strand break repair

Many hematologic and mesenchymal malignancies harbor reciprocal translocations that develop as early and essential events in oncogenesis.^1-3  Translocations likely result from contemporaneous DNA double-strand breaks (DSBs) generated by either endogenous (eg, V(D)J recombinase⁴ ) or exogenous (eg, topoisomerase II poisons⁵ ) agents. Mammalian cells have multiple pathways to repair DSBs, including mechanisms that involve little or no sequence homology, collectively termed nonhomologous end-joining (NHEJ).⁶  NHEJ is important for survival in response to clastogens and is critical for the repair of DSBs induced during V(D)J recombination.^6,7  Homology-directed repair, an essential pathway in mammalian cells,⁸  does not appear to be involved in the overwhelming majority of cancer-associated translocations, as breakpoint junctions lack extensive homology,^1-3  and translocations involving this pathway are not recovered in model systems.^9-12  However, another homologous repair pathway, single-strand annealing (SSA), efficiently generates translocations in model systems.^9,10 

Given that reciprocal translocations in human cancer cells appear to arise from NHEJ, we developed translocation reporters for use in mammalian cells in which significant sequence homology is absent from the DNA ends.

Targeting vectors, polymerase chain reaction (PCR) conditions, and primers are described in Document S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article). For translocation experiments, 2 × 10⁷  cells were electroporated with 50 μg pCBASce^9,13  or pCAGGS.^9,10  Fluorescence in situ hybridization (FISH) was performed as described.⁹ 

Two p5rE cell lines and 2 translocation clones were electroporated with 50

μg pCBASce. After G418 selection, pooled genomic DNA was isolated, digested with I-SceI, and PCR amplified. PCR products were TOPO-cloned (Invitrogen, Carlsbad, CA) and sequenced.

We modified reporters recently developed to study rearrangements at Alu elements⁹  (Figure 1A). Central to the translocation reporters is a neomycin phosphotransferase gene (neo) split by an intron to form neoSD and SAneo, each of which contains an I-SceI endonuclease cleavage site and is targeted to chromosomes 17 or 14, respectively, in mouse embryonic stem (ES) cells. DSBs generated by I-SceI followed by NHEJ to link neoSD and SAneo generates a neo⁺ gene on der(17). Repair within intronic sequences allows for the recovery of translocation junctions with a wide array of DNA end modifications.

We constructed 3 translocation reporters based on this design. With p5rE and p5rErev, chromosomal breakage by I-SceI followed by NHEJ with little sequence alteration results in a neo⁺ gene on der(17) with a 1.7-kb intron (Figure 1B), while formation of der(14) is not under selection. Der(14) can form by NHEJ or SSA, since a 210-bp repeat is located adjacent to the chromosome 14 breakpoint and 0.8 kb from the chromosome 17 breakpoint, allowing for a comparison of these 2 pathways when a repeat is distant from a breakpoint. To test whether compatible overhangs enhance translocation formation, the p5rE reporter contains the I-SceI sites in the same (compatible) orientation, in contrast with the p5rErev reporter, which contains the sites in opposite (incompatible) orientation. Reporter pCr15 contains 2 modifications, an expanded chromosome 17 intron, to accommodate larger deletions, and loss of the repeat, such that der(14) can only form by NHEJ (Figure 1C).

After I-SceI expression, neo⁺ clones were recovered using each reporter at a frequency of 3 to 4 × 10^-5, compared with less than 0.2 × 10^-7 without I-SceI. Thus, recovery of neo⁺ clones is not substantially affected by sequence homology in the vicinity of the breakpoints, complimentarity of DNA ends, or a larger segment for deletion. By FISH, 31 of 32 neo⁺ clones harbored the expected der(17) and der(14) chromosomes (Figure 1D; data not shown), indicating that our reporters faithfully score translocations.

To gain insight into the DSB repair mechanisms leading to translocations, we analyzed 120 neo⁺ clones (Figure 1E; data not shown). For der(17), 117 neo⁺ clones gave PCR products; for der(14), 110 neo⁺ clones gave PCR products, and 7 additional clones had deletions of at least 700 bp as detected by Southern blotting. Of the 234 breakpoint junctions, 172 involved NHEJ and fell into 3 classes: 138 (80%) simple deletions, 20 (12%) insertions, and 4 (2%) complex events, plus 10 unsequenced junctions (Figures S1-S2; data not shown). Although there was some variation in deletion size depending on the particular sequences being joined, the majority of deletions were less than 100 bp for all 3 reporters on both der(17) and der(14) (Figure 2A; Figure S2). Comparing data from a number of malignancies, deletions are observed in a majority of oncogenic translocation junctions as well, with a median deletion length similar to what we observed with our reporters (Table S1). Only 3 der(14) deletions extended more than 2.3 kb, the maximum intron for pCr15, suggesting that our approach recovers most translocations. No correlation was observed between deletion lengths on der(17) and der(14) for an individual neo⁺ clone (Figure 2B) or between deletion lengths from the centromeric and telomeric ends for an individual derivative chromosome (data not shown). Of the junctions with simple deletions, 117 (85%) exhibited microhomology (identical nucleotides at both ends) (Figure 2C), which is also common in oncogenic translocation junctions (Table S2).

In addition to deletions, insertions and duplications of genomic sequences are observed in a portion of oncogenic translocations (Tables S1-S2). Duplications could result from staggered nicks, followed by filling in of the resulting single-strand overhangs. Duplications of all or a portion of the 4 nucleotide I-SceI overhangs were observed in several neo⁺ clones (Table S1). Interestingly, although many of the duplications observed in tumors are also only a few bp, duplications reported in several leukemias are more than 100 bp, which would imply a mechanism in which the nicks are staggered at a substantial distance from each other.^14,15 

The median insertion length observed in oncogenic translocation junctions is a few base pair, although longer insertions also have been reported (Table S2). Of 20 insertions we obtained, 12 were less than 10 bp. Four others were more than 100 bp and were derived from a variety of sources (genomic DNA, mitochondrial DNA, I-SceI vector; Figure S1). The complex events contained duplications and inversions of nearby sequences, deletions, and insertions (Figure S1), resembling events reported in some malignancies.^14-17 

Reporter-specific junctions involving the I-SceI overhangs and sequence repeat also were noted. For p5rE, 24% (13 of 54) of der(17) junctions re-established the I-SceI site, presumably by annealing of the 4 nucleotide, 3′ overhangs. Thus, precise ligation can be used to generate translocation junctions. For p5rE and p5rErev, 65% (62 of 96) of der(14) junctions resulted from SSA compared with 35% (34 of 96) from NHEJ, demonstrating the predominance of SSA when repeats are on opposite sides of the breakpoints, even though one repeat is distant from the I-SceI site. This result suggests that the paucity of oncogenic translocation junctions within repetitive elements is not due to the distance of repetitive elements from breakpoints, but rather to other factors such as sequence divergence, which has recently been shown to substantially reduce Alu-Alu recombination events.⁹  It remains to be tested whether sequence repeats can influence translocation junction formation when both copies of the repeat are on the same side of the breakpoint, as has been proposed for some translocations.¹⁸ 

Using similar systems for DSB formation with I-SceI, we estimate that translocations occur approximately 10³-fold less frequently in our system than NHEJ involving a single DSB.^19,20  Translocations and intrachromosomal repair could occur via the same pathways, or they may arise via distinct pathways. To address this, we directly compared translocation and intrachromosomal junctions derived from DNA ends of the same sequence. To obtain translocation junctions, we expressed I-SceI in p5rE cells, while for intrachromosomal repair, we expressed I-SceI in p5rE translocation clones that had re-established the I-SceI site on der(17) (Figure 3A). For both, PCR was used to isolate breakpoint junctions that had lost the I-SceI site from pooled neo⁺ cells. Deletion lengths and microhomology obtained with this strategy for translocations did not differ significantly from translocation junctions isolated from p5rE neo⁺ clones (Figure 3B). The spectrum of translocation junctions differed, however, from the intrachromosomal repair junctions (Figure S2). Overall, only 13% of intrachromosomal repair junctions had deletions more than 30 bp, versus 53% of translocation junctions (P < .001; Figure 3B), and 35% of intrachromosomal junctions had more than 1 bp of microhomology compared to 74% of translocation junctions (P < .001; Figure 3C). Longer deletions and increased microhomology are characteristic of repair junctions from NHEJ-deficient cells (Ku70^-/-, Ku80^-/-, DNA-PKcs^-/-, XRCC4^-/-, Lig4^-/-),^21,22  raising the possibility that translocations occur through NHEJ pathways that are independent of one or more of these factors.

We have demonstrated that reciprocal translocations between nonhomologous sequences can be recovered in repair-proficient murine cells and that breakpoint junctions recapitulate those found in oncogenic translocations, with frequent deletions and microhomology, as well as insertions and more complex events. Moreover, in a direct comparison, we find that the spectrum of junctions for translocations differs from those derived from repair of a single DSB. Our reporters provide a system for determining the genetic requirements for translocation formation and identifying lineage-specific repair mechanisms in hematopoietic and mesenchymal cells.

We thank Margaret Leversha at the MSKCC Molecular Cytogenetics Core Facility for performing the FISH analysis, and members of the Jasin laboratory, especially Felipe Araujo and Jeremy Stark, for helpful discussions.