Chromosomal translocations induced at specified loci in human stem cells

Abstract

The precise genetic manipulation of stem and precursor cells offers extraordinary potential for the analysis, prevention, and treatment of human malignancies. Chromosomal translocations are hallmarks of several tumor types where they are thought to have arisen in stem or precursor cells. Although approaches exist to study factors involved in translocation formation in mouse cells, approaches in human cells have been lacking, especially in relevant cell types. The technology of zinc finger nucleases (ZFNs) allows DNA double-strand breaks (DSBs) to be introduced into specified chromosomal loci. We harnessed this technology to induce chromosomal translocations in human cells by generating concurrent DSBs at 2 endogenous loci, the PPP1R12C/p84 gene on chromosome 19 and the IL2Rγ gene on the X chromosome. Translocation breakpoint junctions for t(19;X) were detected with nested quantitative PCR in a high throughput 96-well format using denaturation curves and DNA sequencing in a variety of human cell types, including embryonic stem (hES) cells and hES cell-derived mesenchymal precursor cells. Although readily detected, translocations were less frequent than repair of a single DSB by gene targeting or nonhomologous end-joining, neither of which leads to gross chromosomal rearrangements. While previous studies have relied on laborious genetic modification of cells and extensive growth in culture, the approach described in this report is readily applicable to primary human cells, including mutipotent and pluripotent cells, to uncover both the underlying mechanisms and phenotypic consequences of targeted translocations and other genomic rearrangements.

Recurrent chromosomal translocations are associated with many cancers where they are considered to be the initiating event for tumorigenic transformation. As many as half of hematological malignancies have a specific translocation signature, as do a number of tumors of mesenchymal origin, including Ewing's sarcoma, rhabdomyosarcoma, and synovial sarcoma (1, 2). Recurrent oncogenic chromosomal rearrangements have also recently been identified in some carcinomas, including tumors of the prostate (3) and small cell lung cancer (4), raising the possibility that they have a more widespread contribution to the etiology of solid tumors of epithelial origin than was previously recognized (5).

Given the prevalence of chromosomal translocations in human malignancy, understanding how translocations are formed in human cells and the factors involved in their formation could lead to measures to prevent their occurrence. The initiating lesions in most cases are likely to be contemporaneous DNA double-strand breaks (DSBs) on heterologous chromosomes that are misjoined (2, 6). Sequencing of numerous breakpoint junctions from human translocations indicates that a nonhomologous end-joining (NHEJ) pathway of DSB repair gives rise to the misjoining events, since little or no homology is found at the ends of the translocating chromosomes. A set of NHEJ factors have been identified in mammalian cells, including the Ku DNA end-binding protein and an associated kinase DNA-PKcs, DNA ligase IV and an associated binding partner XRCC4, and a number of DNA end-processing factors (7). Because some level of NHEJ can occur in the absence of these canonical factors, “alternative” NHEJ pathway(s) have also been proposed.

Mechanisms and factors affecting chromosomal translocations have been investigated in mouse cells in several contexts. Embryonic stem cells have provided a genetically tractable system to target DSBs to 2 chromosomal loci to induce translocations (6). These studies have investigated the contribution of DSB repair pathways to translocation formation and have surprisingly shown that NHEJ-based translocations are increased in the absence of the canonical NHEJ factor Ku, suggesting that an alternative NHEJ pathway is involved in translocation formation (8). Moreover, the immune system has provided a unique system to investigate translocations arising from DNA damage induced by the activation-induced cytidine deaminase or RAG recombinase (8–12). Translocations increase in the absence of the canonical NHEJ factor XRCC4, again suggesting the involvement of a noncanonical NHEJ pathway in mouse lymphocyte translocations (12), although junctions were not analyzed.

While valuable, these approaches in mouse cells have limited application to human cells which are more difficult to clone and genetically manipulate. And although basic mechanisms of DSB repair are shared between mammalian species, differences have been noted between mouse and humans. For example, human cells appear to have higher levels of DNA-PK activity compared with mouse cells (13). In addition, sequence elements that participate in oncogenic translocations are typically located in noncoding regions which are not well-conserved between mouse and human. Thus, the development of a human system to investigate translocation formation has been an imperative.

In this report, we present an approach to induce and recover translocations in human cells at specified loci. DSBs are introduced contemporaneously into 2 loci using endonucleases with long recognition sites (“meganucleases”), one of which is the commonly used I-SceI homing endonuclease (14, 15). Key to this approach is the technology of zinc finger nucleases (ZFNs) in which zinc finger DNA-binding modules are assembled to recognize specified sites in the genome and then fused to the FokI endonuclease domain to produce a site-specific nuclease (16, 17). To identify and quantify translocations, we used a 96-well quantitative PCR format. With this approach, translocation breakpoint junctions were obtained in 3 cell types: a human embryonic kidney 293 derivative cell line (TOS4A), human embryonic stem (hES) cells, and hES-derived mesenchymal precursor (hES-MP) cells. The translocation frequency was lower than that of repair of a single-break by other DSB repair pathways, and translocation breakpoint junctions were found to recapitulate characteristics of patient-derived junctions. With this approach, genetically unmodified human cells can be used to determine how DSBs give rise to translocations in human cells and the factors involved. Further, our results suggest an approach to derive translocations in human cells relevant to tumor formation.

Results

Recovery of Induced Chromosomal Translocations from Human Embryonic Kidney 293 Cells (TOS4A).

Our approach for inducing defined chromosomal translocations in human cells relies on the formation of 2 concurrent DSBs in heterologous chromosomes using site-specific endonucleases. To determine the feasibility of our approach, we used the 293-derived cell line TOS4A (18). TOS4A cells contain a unique 18 bp I-SceI endonuclease recognition site randomly integrated into their genome in the context of a single copy of the DR-GFP reporter (Fig. 1A; see also below). We performed fluorescence in situ hybridization (FISH) to determine the chromosomal location of DR-GFP, which revealed that DR-GFP integrated very close to the centromere in 1 of the 3 chromosomes 6 present in the 293 cell line (Fig. S1A).

Fig. 1.

Induction of chromosomal translocations in TOS4A cells. (A) Design of the translocation system. DSBs are induced at the I-SceI and ZFN^IL2Rγ sites in the genome of TOS4A cells. After concurrent DSB formation and misjoining, translocations can in principle lead to either monocentric derivative chromosomes, der(6) and der(X), or dicentric and acentric chromosomes. (B) PCR approach to identify translocation breakpoint junctions for der(6) and der(X). Fragment sizes are calculated with the overhangs filled in (black). (C) Recovery of t(6;X) translocation after 2 rounds of PCR enrichment for the der(X) breakpoint junction. FISH using whole chromosome paints to chromosome 6 (green) and the X chromosome (red) verified the translocation. Breakpoint junction sequences are shown. (D) Translocation NHEJ, single-break NHEJ, and HR in TOS4A cells after expression of I-SceI and ZFN^IL2Rγ. Translocation frequency is evaluated with 2 primer sets. Single-break NHEJ is measured with primers that flank the I-SceI site. To differentiate imprecise NHEJ from HR and SSA, PCR fragments are cleaved with I-SceI and/or LweI. Imprecise NHEJ products are resistant to both (*), whereas HR and SSA products are cleaved by LweI but not I-SceI. (See also Fig. S2C.) Z, ZFN^IL2Rγ; S, I-SceI^hi. HR is the percent GFP+ cells. Results from 3 independent experiments are shown with 1 standard deviation from the mean.

The second site in the genome targeted for DSB formation is cleaved by a ZFN. The ZFN cleavage site is within exon 5 of the IL2Rγ gene on the X chromosome (Xq13) (19). The 2 subunits of the ZFN^IL2Rγ are designed to bind two 12-bp sequences spaced 5 bp apart; cleavage occurs within the spacer, typically producing a 5-base 5′ overhang (J. C. Miller, personal communication). TOS4A cells contain a normal X chromosome with a ZFN^IL2Rγ site and a rearranged X chromosome that apparently also has a ZFN^IL2Rγ site (Fig. S1 A and B). After concurrent DSB formation at the I-SceI and ZFN^IL2Rγ sites, DNA ends on chromosomes 6 and X can in principle be misjoined to produce chromosomal translocations with either monocentric or dicentric/acentric derivative chromosomes (Fig. 1A). PCR products consistent with translocation breakpoint junctions have been previously reported (20), although cells harboring translocations were not recovered and conditions for quantification of translocations were not established.

To verify translocation formation, we set out to recover cells harboring t(6;X) chromosomes. TOS4A cells were transfected with the I-SceI and ZFN^IL2Rγ expression vectors and sib-selection was performed over several cell generations using PCR screening for breakpoint junctions (SI Materials and Methods). Two sets of PCR primers were used to screen for alternative translocation outcomes (Figs. 1B and S1B), and FISH was used to detect the translocation chromosomes. One set of primers led to the identification of a reciprocal translocation between chromosomes 6 and X, as detected by whole chromosome painting (Fig. 1C). The chromosomes were monocentric, allowing us to infer that DR-GFP had integrated oriented toward the centromere during the generation of the TOS4A cell line. Screening with the second set of primers led to the identification of what appeared to be a dicentric chromosome involving the rearranged X chromosome (Fig. S1B), indicating that the genetically compromised 293 cells are able to stabilize what would otherwise be a mitotically unstable chromosome. The reciprocal acentric chromosome was not detected by either PCR or FISH. These cells may not be able to propagate an acentric chromosome, or, alternatively, the reciprocal chromosome may never have been formed.

Sequences of the breakpoint junctions in both translocation configurations showed joining by NHEJ with surprisingly little modification of the DNA ends. For the monocentric der(6) and der(X) chromosomes, a net gain of 3 bp was observed as a result of partial fill-in of the 5′ and 3′ overhangs at the chromosome ends (Fig. 1C). For the apparently dicentric chromosome, the breakpoint junction was also formed by fill-in of the 5′ and 3′ overhangs (Fig. S1B).

Development of a Quantitative Assay for Chromosomal Translocation Formation.

To quantify translocations, we developed a high throughput 96-well format screen for breakpoint junctions (Fig. S1C). After transfection with the I-SceI and ZFN^IL2Rγ expression vectors, TOS4A cells were seeded at 10⁴ cells per well. Forty-eight hours later, cells in each well were lysed and PCR was performed to detect the breakpoint junctions. Nested PCR was performed to amplify the translocations, with the second round of PCR performed in the presence of SYBR Green. A well was considered to be positive when it gave a PCR product with the expected melting temperature (≥85 °C) for a breakpoint junction (Fig. S1C). Sequencing confirmed that these were bona fide breakpoint junctions (see below). The translocation frequency was then calculated as the ratio of the number of positive wells to the total number of transfected cells; if the number of positive wells was 14–30, statistical analysis was performed following a beta cumulative distribution function to correct for wells with more than 1 translocation (SI Materials and Methods).

Using this method, we compared the recovery of the 4 possible breakpoint junctions, reproducibly obtaining translocation frequencies of approximately 2 × 10⁻⁵ per cell for each junction (see Fig. 1D). These results confirm that the 96-well format is a valid method for quantifying translocation formation. Further, these results suggest that all 4 chromosome configurations [monocentric der(6) and der(X), dicentric, and acentric] are equally likely, and that the 48-h time frame of the assay permits the detection of unstable chromosomes. However, it should be noted that the presence of rearranged X chromosomes in the parental 293 cells complicates the certain assignment of a breakpoint junction to a particular chromosome configuration without supporting FISH analysis. We also analyzed each well of a 96-well plate for all 4 translocation breakpoint junctions. Reciprocal translocation products were obtained in some but not all wells (Fig. S1D), raising the possibility that nonreciprocal translocations are frequent in these cells, even for the monocentric chromosomes. Although the reciprocal chromosome product may have failed to amplify, PCR products were reproducibly obtained from positive wells, suggesting that PCR failure is not typical.

Translocation Breakpoint Junctions Have Characteristics of Oncogenic Translocations.

About 90% of patient-derived translocation breakpoint junctions exhibit modifications to the DNA ends before joining. Such modifications are typical of an NHEJ-based repair mechanism (6). We sequenced 57 translocation breakpoint junctions from the TOS4A cells to determine what types of DNA end modifications occur in our induced translocations (Fig. S2A). A portion of the translocations (14/57, 25%) had junctions in which the DNA ends were joined without modification, as defined by retention of the DNA overhangs without further alteration (Fig. S2A). Thus, human cells have the capability to completely “fill-in” both 5′ and 3′ overhangs to maintain DNA end sequences. The remaining junctions had deletions at one or both ends. Of these, 14% of junctions (8/57) had only lost bases from the overhangs without further modification. Altogether, the deletions were short, with 93% of junctions (53/57) having lost ≤ 35 bp from the DNA ends (Fig. S2B). We estimate the limit of detection for deletions is ≤ 350 bp; thus, while translocations with large deletions will not be scored in this assay, the majority of translocation breakpoint junctions appear to be recoverable from these cells. Nine of the breakpoint junctions had insertions (Fig. S2A). Five of the insertions were small (1–13 bp), while the remaining 4 were larger (41–117 bp) and included sequences duplicated from the IL2Rγ gene that were originally located either adjacent to the breakpoint or at some distance.

Microhomologies are frequently observed at translocation breakpoint junctions in oncogenic translocations (6, 21), and they were also observed at the breakpoint junctions we recovered. Overall, 44% of junctions (21/48) had microhomologies which ranged from 1–4 bp, with 27% of the junctions (13/48) showing greater than or equal to 2 bp of microhomology (Fig. S2B). It is notable that although 4 base repeats flank 2 of the breakpoints, only 2 of 52 breakpoint junctions used this microhomology (Fig. S2A). Thus, while microhomology is commonly observed, moderate stretches of microhomology (e.g., 4 bp) do not appear to drive human translocations.

Simultaneous Evaluation of Multiple DSB Repair Pathways in TOS4A Cells.

Translocations are infrequent outcomes of DSB repair, as most DSB repair events maintain overall genomic integrity. Typically, the 2 ends of a single chromosome break are simply rejoined through an NHEJ pathway or repaired by precise homologous recombination (HR). The TOS4A system allows simultaneous quantification of all 3 DSB repair pathways (translocation NHEJ, single-break NHEJ, and HR). As previously described, the DR-GFP reporter quantifies HR by reconstitution of a functional GFP gene, since DSB formation by I-SceI induces gene conversion at the GFP repeats (Fig. 1D) (22). Single-break NHEJ is measured at the DR-GFP reporter by modification of the I-SceI site during rejoining (I-SceI “site loss”), as NHEJ is frequently imprecise, producing nucleotide modifications at DNA ends similar to translocation NHEJ. Imprecise NHEJ is distinguished from HR in the I-SceI site-loss assay by the restriction enzyme LweI, because HR results in mutation of the I-SceI site with concomitant gain of an LweI site from the downstream GFP repeat while NHEJ does not (Fig. S2C).

To quantify all 3 DSB repair pathways in the same experiment, the I-SceI expression vector was transfected into the TOS4A cells at a higher concentration than the ZFN^IL2Rγ expression vector (I-SceI^hi). Transfected cells were then divided into 2 plates: a 96-well plate for translocation formation and a 10-cm dish for HR and imprecise NHEJ quantification. Under these defined conditions, HR was determined to have occurred at a frequency of 2.8% using flow cytometry for GFP+ cells, imprecise NHEJ at 29% using I-SceI site-loss PCR (Fig. S2C), and translocations at approximately 2 × 10⁻⁵ assaying for 2 breakpoint junctions by PCR (Fig. 1D). If both the I-SceI and ZFN^IL2Rγ expression vectors were transfected at high concentrations, translocations increased to approximately 10⁻⁴ and so were more difficult to quantify. With the ZFN^IL2Rγ/I-SceI^hi combination of vectors, each of the DSB repair pathways can be simultaneously quantified, allowing the assessment of factors that may change the balance of these pathways.

Inducing Defined Chromosomal Translocations in Genetically Unmodified hES and hES-MP Cells.

In addition to ZFN^IL2Rγ, ZFNs have recently been described that cleave other endogenous loci in the human genome (17, 23). This development allows DSB formation at 2 endogenous loci in human cells, abrogating the need for prior genetic modification of cells. This approach is particularly powerful for nontransformed human cell types which are difficult to clone. We made use of a zinc finger nuclease, ZFN^p84, that cleaves the PPP1R12C/p84 locus on chromosome 19 (Fig. 2A). The site of DSB induction by ZFN^p84 is close to the AAV integration site (24), which can be used as a universal site for gene insertion. Co-expression of ZFN^p84 and ZFN^IL2Rγ would lead to concurrent formation of DSBs on chromosomes 19 and X; misjoining of the DSBs would lead to t(19;X) translocations (Fig. 2A). We focused on the formation of the monocentric derivative chromosomes, der(19) and der(X), although events corresponding to the formation of the dicentric chromosome were also detected. A PCR method similar to that described above was used to detect the appearance of chromosomal translocation breakpoint junctions (Fig. 2B). Using TOS4A cells, we recovered both der(19) and der(X) translocation junctions in a manner dependent on expression of both ZFN^p84 and ZFN^IL2Rγ (Fig. 2B). We obtained similar translocation frequencies for t(19;X) as with t(6;X) (10⁻⁴–10⁻⁵). In addition to TOS4A cells, we were also able to efficiently recover translocation breakpoint junctions from SV40-transformed human fibroblast cell lines.

Fig. 2.

Induction of chromosomal translocations at endogenous human loci. (A) Design of the translocation system. DSBs are induced at the ZFN^IL2Rγ and ZFN^p84 sites, as indicated. Only der(19) and der(X) monocentric chromosomes are shown. The PCR approach to identify translocation breakpoint junctions for der(6) and der(X) is shown. Fragment sizes are calculated with the overhangs filled in (black). (B) PCR for der(19) and der(X) breakpoint junctions in TOS4A cells after ZFN^p84 and ZFN^IL2Rγ expression. (C) Sequences of der(X) breakpoint junction sequences from hES-MP cells. Microhomologies are underlined. The boxed C indicates that ZFN^p84 cleavage may sometimes result in an alternative overhang. The 231-bp insertion is described in Fig. S3C legend.

We next applied this approach to genetically unmodified hES cells and hES-MP cells which are proficient at multilineage differentiation into fat, cartilage, bone, and skeletal muscle (25). Multipotent mesenchymal cells are likely to be precursors for some sarcomas harboring characteristic translocations (26, 27). Using nucleofection, hES and hES-MP cells can be efficiently co-transfected, as measured by GFP and DsRed expression. For example, approximately 80% of hES-MP cells express either protein, and 70% of successfully transfected cells express both proteins (Fig. S3A). To assay translocations, cells were cotransfected with expression vectors for ZFN^p84 and ZFN^IL2Rγ and then plated into 96-well plates at 4 × 10⁴ cells per well (Fig. S3B). Four days later, cells in each well were lysed and quantification by PCR was performed using primers for der(X), as the PCR was more robust for this derivative chromosome, although in some experiments, der(19) was also characterized (Fig. S3C). In hES cells, the frequency of translocations cells was 2.2 ± 0.7 × 10⁻⁶; in hES-MP cells, it was 7.5 ± 1.0 × 10⁻⁶. These results indicate that translocations can readily be detected in non-transformed cells types at specified endogenous loci. The frequency of translocations did not reach the level found in TOS4A cells, suggesting that hES or hES-MP cells may be less prone to genomic rearrangements; however, we cannot exclude that differences in transfection or DSB efficiency effect translocation formation.

We sequenced 27 translocation breakpoint junctions from the hES-MP cells and 19 from the hES cells for der(X) (Fig. 2D and Fig. S3C). Because the results were similar, we combined the junctions from both cell types for analysis. Compared with the I-SceI/ZFN^ILR2γ induced translocations, the breakpoint junctions from ZFN^p84/ZFN^IL2Rγ induced translocations were notable in that fewer junctions were found without end modifications (3/46, 7% vs. 25%, P = 0.0167, Fisher's exact test), as defined by fill-in of the two 5′ overhangs, but significantly more had junctions that had deleted bases only from the overhangs (34/46, 74% vs. 19%, P < 0.0001). As a result, overall the deletions were smaller, such that 93% of junctions (43/46) had deleted ≤ 17 bp (vs. ≤ 35 bp) (Fig. S3D). This fine structure junctional difference may be due to the different type of overhangs and particular sequence of the overhangs, since bases from the two 5′ overhangs can anneal. For example, the junction using the “CC” or “CCA” microhomology in the overhangs appears to be overrepresented in comparison to other junctions. It is worth noting that the primary cleavage of the ZFN^p84 is expected to result in a 4-base 5′ overhang as shown (Fig. 2A); however, ZFNs in which the zinc finger binding sites are spaced 6 bp apart, as for the ZFN^p84, can generate minor cleavage products (28), which in this case would result in an additional “C” to the overhang for a “CCAC” microhomology.

Insertions were also observed in 11% (5/46) of junctions. Three breakpoint junctions had small insertions of 1–3 bp, while the remaining 2 were larger (67 and 231 bp). The largest insertion involved a duplication of p84 sequences located adjacent to the breakpoint as well as unidentified DNA; insertions of multiple segments of DNA in tumor cell rearrangements have been termed “genomic shards” (29). Microhomologies were also observed at the breakpoint junctions. Overall, 49% of junctions (20/41) had microhomologies, ranging from 1–3 bp, with 27% of the junctions (11/41) showing greater than or equal to 2 bp of microhomology (Fig. S3D). These results are comparable to those obtained in the t(6;X) junctions.

DSB Repair Pathways in Human Multipotent Stem Cells.

Human multipotent stem cells offer the potential for diverse studies, including oncogenesis and lineage analysis, yet approaches to genetic modification and understanding DNA instability are limited. Using ZFNs, HR can be assayed by DSB-mediated gene targeting, as has been used in mouse cells (14). Specifically, DSBs promote integration of a marker gene flanked by DNA sequences homologous to the target gene. We used a promoter-less GFP gene flanked by p84 sequences to target p84. The donor plasmid consists of two 750-bp sequences that flank the ZFN^p84 site interrupted by a splice acceptor and a GFP ORF followed by a polyadenylation signal sequence (pGFP^p84 donor) (Fig. 3A). When ZFN^p84 is expressed in cells transfected with the pGFP^p84 donor, DSB-promoted gene targeting will result in GFP expressed from the endogenous p84 promoter. Random integration may also fortuitously lead to GFP expression, but at a much lower frequency.

Fig. 3.

Comparison of DSB repair pathways in multipotent hES-MP cells. (A) DSB-mediated gene targeting strategy to quantify HR. A promoterless GFP donor (GFP^p84) can target the p84 locus upon ZFN^p84 cleavage and be expressed from the p84 promoter. Homology arms (blue) are each approximately 750 bp. (B) Flow cytometric analysis of hES-MP cells 2 weeks after transfection with GFP^p84 and the indicated ZFNs. Cells transfected with ZFN^p84 (Right) contain a significant GFP+ population which was sorted for further analysis. (C) PCR analysis of genomic DNA from sorted GFP+ cells to verify DSB-mediated gene targeting. PCR primers and fragment sizes are shown in A. (D) Translocation NHEJ, single-break NHEJ, and HR after ZFN^IL2Rγ and ZFN^p84 cleavage. Translocation frequency is quantified for the der(X) breakpoint junction. For single-break NHEJ, imprecise repair products were detected by PCR amplification across the ZFN^p84 site and colony hybridization (Fig. S4). HR is quantified by DSB-mediated gene targeting. Results from 3 independent experiments are shown with 1 standard deviation from the mean.

In conjunction with the translocation assays, we co-transfected hES-MP cells with the pGFP^p84 donor plasmid and the ZFN expression vectors and incubated cells for 4 days. In experiments that included ZFN^p84, we obtained 2.9 ± 0.2% GFP+ cells, while in the absence of ZFN^p84, we obtained only 0.07 ± 0.04% GFP+ cells (Fig. 3B), indicating that a DSB stimulates targeted integration of GFP in the hES-MP cells. The level of GFP+ cells was stable over 2-week incubation at which time GFP+ cells were sorted by flow cytometry for genomic DNA isolation. Targeted integration was confirmed by PCR of the GFP+ sorted cells (Fig. 3C).

To complete the DSB repair analysis, single-break NHEJ was also examined. PCR amplification was performed with primers that flanked the ZFN^p84 site; PCR fragments were purified, cloned, and sequenced. From cells transfected with ZFN^p84, 3 of 67 sequences were modified, whereas in the absence of ZFN^p84, none of 86 sequences were modified. Interestingly, the 3 NHEJ events were simple fill-in reactions to the 5′ overhangs (Fig. 3D). Overhang fill-in followed by blunt end ligation was also observed in DSB repair studies in worms (30) and human T cells (23). Given that fill-in of the 5′ overhangs appeared to be the most frequent single-break NHEJ event, we performed colony hybridization with probes to the fill-in products (Fig. S4). Using this approach, we obtained 33 fill-in products from 1,655 colonies analyzed, which is 1.9% p84 allele modification; assuming that most products represent a modification of just one of the two p84 alleles in a cell, the frequency of p84 site modification for cells is 3.8% for the fill-in product.

More recently, ZFN^IL2Rγ has been modified to promote a higher specificity of cleavage, termed _HiFiZFN^IL2Rγ (31). We tested _HiFiZFN^IL2Rγ with ZFN^p84 and obtained an approximate 2-fold higher translocation frequency (Fig. S4), suggesting that cleavage may be somewhat higher than with ZFN^IL2Rγ due to increased specificity. We also measured single-break NHEJ at the IL2Rγ site for the common fill-in product using colony hybridization; similar to p84, we obtained a frequency of IL2Rγ site modification for cells of 3.4% for the fill-in product (Fig. S4).

Discussion

The analytic and therapeutic modification of human primary, precursor, or stem cells requires the least possible genetic manipulation and the shortest possible duration of in vitro culture. To overcome the need for genetic modification of cells before induced chromosomal translocation, 2 sets of ZFNs that each target an endogenous locus were expressed in human cells. Translocation formation was quantified using nested PCR, and screening for denaturation temperature in a 96-well format in conjunction with sequencing of translocation breakpoint junctions. With this approach, t(19;X) translocations were induced within 96-h post-transfection in multipotent hES-MP cells and pluripotent hES cells at frequencies of 10⁻⁵ to 10⁻⁶. By comparison, HR and single-break NHEJ were much more efficient, as measured by DSB-mediated gene targeting and imprecise NHEJ in the hES-MP cells, respectively; this greater reliance on HR and single-break NHEJ serves to maintain genomic integrity in these multipotential cells. Concurrent analysis of HR, single-break NHEJ, and translocation frequency will allow a determination of how the balance of these pathways is perturbed when various DSB repair pathway components are altered. This is an important question given that these pathways are known to impinge on each other [e.g., (32)].

Quantification of translocation frequency using the 96-well format was initially developed using the 293-derived cell line TOS4A. Although 293 cells are transformed and contain numerous chromosome aberrations, they are easy to culture and transfect and are a commonly used human cell line for DSB repair analysis (e.g., ref. 18). With the TOS4A cells, we confirmed that breakpoint junctions obtained by PCR represent translocations on the chromosome level. Because TOS4A cells contain an integrated DR-GFP reporter, we also simultaneously assayed HR using a DSB-induced gene conversion assay. An advantage to this assay is that the background of GFP+ cells is lower than that obtained with the DSB-mediated gene targeting assay (<0.01% vs. 0.07%). In the latter case, random integrations of the GFP^p84 targeting vector near a promoter can restore a GFP+ gene at a low frequency.

Breakpoint junctions for the t(19;X) translocations recovered in these experiments have a range of DNA-end modifications, including deletions, insertions, and microhomology, which recapitulate characteristics of translocation breakpoint junctions in human tumors (2, 6). Deletions were frequent and typically short (<20 bp) with only a few having deleted >100 bp. The current primer sets limit detection of deletions to approximately 400 bp. Based on our previous experiments in mouse cells (6), we expect that the majority of translocations are being captured with these primers; additional primers set further from the breakpoints would test this. Insertions were less frequent and were often just a few bp, although one insertion was composed of 2 unrelated DNA segments, similar to genomic shards found inserted at some breakpoint junctions in some tumor cell rearrangements (29). Microhomology of 1–4 bp was found in about half of the junctions.

A key question is what factors promote chromosomal translocation formation. Clearly, NHEJ is central to this process, but the role of the canonical NHEJ pathway is uncertain, since, when tested, they are not required for, but rather suppress, translocations (8, 12). Even for single-break NHEJ, the pathways involved appear to be complex and not fully dependent on canonical NHEJ components (7, 33). One hypothesis is that the NHEJ pathway(s) that mediates translocation transformation may still use the same components used for single-break NHEJ, just less efficiently. Occasional failure of single-break NHEJ could lead to persistence of DSBs, allowing for loss of DNA-end synapsis and hence the possibility of 2 DSBs joining to each other. These persistent DSBs could be more prone to end modification. The alternative to this long-lived DSB hypothesis is that translocations (and other 2-break events) arise from similar pathways and components as single-break repair, but that the contribution of the canonical NHEJ pathway is greater for single-break repair than are alternative NHEJ pathways. Greater use of the canonical NHEJ pathway would suppress illegitimate events such as translocations, whereas greater use of alternative pathways would increase their occurrence. A clear understanding of alternative NHEJ will shed light on these hypotheses.

Clinically relevant chromosomal translocations found in some sarcomas have been hypothesized to occur in mesenchymal stem cells (26, 27). The use of site-specific nucleases to induce chromosomal translocations as demonstrated here in human cells provides the proof of principle to consider designing custom nucleases directed toward genomic regions implicated in oncogenic translocations in multipotent cells. Creating physiologically relevant translocations at endogenous loci, rather than ectopically expressing fusion proteins that arise from translocations, would reproduce the cellular milieu in which fusion proteins typically arise: fusion proteins created during translocation formation would be expressed from the endogenous promoter, and the copy number of the nontranslocated alleles would be reduced from 2 to 1. Effects on proliferation and other cellular phenotypes arising from fusion gene expression can, therefore, be assessed with a much greater precision.

Multipotent mesenchymal stem cells are used in differentiation studies of various lineages, including chondrocytes and osteoblasts, which have the potential for use in tissue engineering and regenerative medicine (25). Thus, the ability to genetically manipulate these cells at defined loci has impact beyond the study of DSB repair mechanisms or tumorigenesis. The experiments presented here with the p84 gene provide a model to target a gene for modification in these cells. Modification of other genes relevant to the biology of mesenchymal precursors or their derivatives can be achieved by designing ZFNs to cleave other sites of interest.

In addition to IL2Rγ and p84, ZFNs have been successfully designed for other loci in human cells, such as CCR5 (23), CHK2 (34), and VEGF-A (17, 23). Moreover, improvements to ZFN design have resulted in ZFNs with increased locus specificity. For example, ZFNs have been modified to efficiently cleave DNA only when paired together as a heterodimer (31). ZFNs have also been fused to destablization domains to regulate their levels (35). Like the homing endonuclease I-SceI, ZFNs can be considered ‘meganucleases’ because of their long recognition sequences. Engineering of homing endonucleases to recognize an endogenous site in the genome, while challenging, has recently proved to be successful (36), providing an alternative approach to ZFNs for human genome modification. Synthetic chemical reagents have also recently been shown to lead to efficient DSB formation in the human genome (37, 38). Our results, as well as advances in site-specific DSB technology, suggest that modification of the human genome for a variety of purposes, including translocation formation, will become increasingly possible and efficient in the near future.

Materials and Methods

Additional materials and methods are found in SI Materials and Methods.

TOS4A cells (18) were transfected with pCBASce (39) and ZFN^IL2Rγ (19) using Lipofectamine and lysed for analysis 48 h later. Human ES and hES-MP cells (25) were transfected with Amaxa technology (Lonzo) nucleofector kit V (program B-16). A total of 5 × 10⁶ cells were transfected with 2.5 μg of each ZFN^IL2Rγ and 2.5 μg ZFN^p84 (±30 μg GFP^p84). After transfection, 4 × 10⁶ cells were plated in a 96-well plate and 10⁶ cells used for HR or NHEJ analysis. After 4 days, cells were lysed for analysis. The first round of PCR was performed with 4–7 μL cell lysate from each well in a total of 50 μL per well (23 cycles, annealing temperature of 60 °C). Then 0.5–1 μL of the first PCR was used in a second nested PCR in a total of 20 μL per well (40 cycles, annealing temperature 60 °C) with SYBR Green for PCR (Stratagene MX3005). PCR products were purified by the PCR gel Purification Kit (Invitrogen) and directly sequenced or first cloned into StrataClone PCR Cloning Kit (Stratagene).