|Home | About | Journals | Submit | Contact Us | Français|
Double-strand breaks (DSBs) are particularly deleterious DNA lesions for which cells have developed multiple mechanisms of repair. One major mechanism of DSB repair in mammalian cells is homologous recombination (HR), whereby a homologous donor sequence is used as a template for repair. For this reason, HR repair of DSBs is also being exploited for gene modification in possible therapeutic approaches. HR is sensitive to sequence divergence, such that the cell has developed ways to suppress recombination between diverged (“homeologous”) sequences. In this report, we have examined several aspects of HR between homeologous sequences in mouse and human cells. We found that gene conversion tracts are similar for mouse and human cells and are generally ≤100 bp, even in Msh2−/− cells which fail to suppress homeologous recombination. Gene conversion tracts are mostly unidirectional, with no observed mutations. Additionally, no alterations were observed in the donor sequences. While both mouse and human cells suppress homeologous recombination, the suppression is substantially less in the transformed human cells, despite similarities in the gene conversion tracts. BLM-deficient mouse and human cells suppress homeologous recombination to a similar extent as wild-type cells, unlike Sgs1-deficient Saccharomyces cerevisiae.
The ability of a cell to repair DNA damage is integral to maintaining genome integrity. One common type of damage that is particularly detrimental is a double-strand break (DSB), where both strands of DNA are broken. If not accurately repaired, DSBs can lead to cell death, chromosomal rearrangements, and loss of genetic material (reviewed in references 14 and 19). One mechanism of DSB repair is homologous recombination (HR), in which an unbroken homologous sequence, the donor of genetic information, is used as a template for repair of the broken sequence, the recipient of genetic information. HR intermediates possess heteroduplex DNA (hDNA), where one strand of DNA is derived from the donor sequence, and the second strand is derived from the recipient sequence. Mismatches in hDNA are substrates of the mismatch repair machinery (MMR) (reviewed in reference 38), leading to gene conversion. HR is the preferred repair pathway of DSBs in Saccharomyces cerevisiae (reviewed in references 42 and 46), plays an important role in repair of DSBs in Drosophila (1, 32), and is a major repair pathway of DSBs that occur during S/G2 in mammalian cells (33, 54).
Two pathways appear to predominate for the repair of DSBs by HR, both of which can give rise to noncrossover products, which predominate in mitotic mammalian cells (Fig. (Fig.1)1) (29, 52, 60). In the DSB repair model proposed by Szostak et al. (61), double Holliday junctions are resolved to result in recombinant products (Fig. (Fig.1A).1A). More recent evidence suggests the existence of an alternative pathway, termed synthesis-dependent strand annealing (SDSA) (Fig. (Fig.1B)1B) (20, 40, 42, 52). One difference between these two pathways is that the DSB repair model requires capture of both DNA ends (Fig. (Fig.1A),1A), which can lead to bidirectional gene conversion tracts. In contrast, SDSA can involve only one end of the broken DNA followed by dissociation (Fig. (Fig.1B),1B), resulting in predominantly unidirectional gene conversion tracts. Another difference is that the donor sequence can be altered during DSB repair while it typically remains unchanged after SDSA.
HR repair is sensitive to differences between the recombining sequences, and cells have developed ways to suppress recombination between diverged sequences. This suppression of “homeologous” recombination reduces HR both between diverged repeats and with foreign DNA. Suppression of homeologous recombination is conserved across species and requires the MMR machinery (7, 10, 11, 49, 56). For example, MSH2 dramatically reduces both gene targeting (12) and DSB-induced HR (15) between sequences with >1% divergence in murine embryonic stem (ES) cells.
Another protein that has been proposed to suppress homeologous recombination is Sgs1, the budding yeast RecQ helicase, as sequence divergence has little effect on recombination frequencies in Sgs1 mutants (39, 59). Sgs1 mutants have other phenotypes as well; for example, they demonstrate a hyperrecombination phenotype associated with spontaneous repair (22, 65, 68). The mammalian homolog of Sgs1 is BLM, mutants of which also have a hyperrecombination phenotype, as evidenced by a high frequency of sister-chromatid exchange (SCE) in both human and mouse cells (18, 24, 34, 69). Evidence suggests that Drosophila BLM, like Sgs1, has a role in the suppression of homeologous recombination (30) although mammalian BLM has not been tested in this regard. Supporting a possible role for BLM in suppressing homeologous recombination is the observation that BLM associates with MMR factors in a large protein complex (64; reviewed in reference 21), and BLM directly interacts with two components of the MMR machinery, MLH1 (45) and MSH6 (44), which, like MSH2, is known to suppress homeologous recombination (13).
To gain more insight into mammalian HR mechanisms, as well as factors that control recombination between homeologous sequences, we examined recombination between homologous and homeologous sequences in both murine and human cells. By taking advantage of multiple, single base pair polymorphisms distributed along the donor in gene conversion substrates, we examined both the nature of gene conversion tracts and the fate of the donor sequence. Unidirectional tracts with a bias in conversion to one side of the DSB predominated in both mouse and human cells, supporting an SDSA mechanism of HR. Moreover, the donor remained unaltered after HR. Interestingly, while transformed human cells suppressed homeologous recombination, the degree of suppression was less than that observed in mouse cells. For either cell type, BLM deficiency did not alter this suppression, unlike what is observed in yeast Sgs1 mutants. Either other RecQ helicase family members play a role in the suppression of homeologous recombination, or mammalian RecQ helicases do not play a role in this process.
H-DR-WT (where DR is direct repeat and WT is wild type), pneo-WT, pneo-10mu, and S2neo were previously described (15, 16, 57). H-DR-10mu was constructed by placing an ApaI/PmlI fragment of pneo-10mu into ApaI/PmlI-linearized H-DR-8mu, increasing the single nucleotide polymorphisms in the 745-bp donor fragment from 8 (1.2% total divergence) to 10 (1.5% total divergence) (Fig. (Fig.2A;2A; see also Fig. S2 in the supplemental material).
H-DR-WT was previously targeted to the X-linked Hprt locus in E14 wild-type (12) and Msh2−/− male mouse ES cells (15). H-DR-10mu was targeted to wild-type and Msh2−/− cells using a similar approach. Cells were resuspended in phosphate-buffered saline (PBS) or OptiMem (Gibco, BRL) to 2 × 107 cells/ml, and 0.65 ml was used for each transfection, for a total of 1.3 × 107 cells per transfection. SacI/XhoI-linearized targeting vector (70 μg) was electroporated with the Bio-Rad Gene Pulser II in 0.4-cm cuvettes (800 V; 3 μF). Electroporated cells were split into four to five 10-cm plates, and 24 h later, puromycin was added to cells (1.6 μg/ml) to select for integration of the targeting vector. After 5 days, 6-thioguanine (4 μg/ml) was added to select for integration at the Hprt locus. Puro+ Hprt−/− clones were verified for Hprt targeting by Southern blotting (see Fig. S1A in the supplemental material).
S2neo was targeted to the Rb1 locus of Blmtet/tet ES cells (69). In these cells, the endogenous Blm alleles are modified to contain a tet operator and a tTA transactivator (tet-off). S2neo targeting was confirmed by Southern blotting, similar to the method of Stark and Jasin (60; also data not shown). To suppress BLM expression in Blmtet/tet cells, doxycycline was added to the medium for 48 h prior to transfection at a final concentration of 1 μg/ml and removed 24 h after transfection when G418 was added to the medium to select neo+ recombinants.
Human cell lines were simian virus 40 (SV40)-transformed cell lines GM00637 (wild type) and GM08505 from a Bloom's syndrome (BS) patient (Coriell Institute, Camden, NJ, and a kind gift from Nathan Ellis). For random integration of S2neo into GM00637 and GM08505 cells, the S2neo vector was linearized with HpaI, and 35 μg was electroporated into 1.3 × 107 cells of each line under the same conditions as described above. As the S2neo vector contains pgkhyg, cells were subjected to hygromycin selection (150 μg/ml) 48 h after electroporation. After 18 to 20 days, hygromycin-resistant clones were picked and expanded; single-copy integration of S2neo was confirmed by Southern blotting (see Fig. S1B in the supplemental material), and three clones of each cell line were used for HR analyses.
All murine ES cells were cultured on gelatin-coated dishes in standard medium supplemented with 833 U/ml of ESGRO leukemia inhibitory factor (Millipore, Netherlands), as previously described (53). Human fibroblasts were cultured in high-glucose Dulbecco's modified Eagle's medium (DME-HG) with 10% fetal bovine serum (Gemini; West Sacramento, CA) and penicillin/streptomycin (Gemini; West Sacramento, CA) in 5% CO2 at 37°C.
For intrachromosomal HR assays, 1.3 × 107 ES cells in 0.65 ml of PBS were electroporated (250 V; 950 μF) with 25 μg of either the I-SceI expression plasmid (pCBASce) or the empty vector (pCAGGS) (250 V; 950 μF). Transfected cells were split into four to five 10-cm dishes. After 24 h, medium containing G418 (200 μg/ml) was added to the cells for 10 to 12 days. G418-resistant colonies were fixed and stained with Giemsa for cell counts or picked for clonal analysis. The frequency of neo+ colonies was calculated by dividing the number of neo+ cells by the number of electroporated cells and correcting for 50% viability of cells after electroporation. Comparisons of the average neo+ frequencies between genotypes were analyzed using unpaired student t test.
For plasmid HR assays in ES cells and human fibroblasts, 1.3 × 107 cells were cotransfected with 25 μg of either H-DR-WT or H-DR-10mu plasmid DNA and 25 μg of either pCBASce or pCAGGS plasmid DNA (250 V; 950 μF). G418 (200 μg/ml) was added 24 h later, and colonies were fixed and stained with Giemsa 10 to 12 days later (ES cells) or 18 to 20 days later (human fibroblasts). neo+ frequencies were determined by dividing the number of neo+ colonies after selection by the number of cells surviving electroporation 24 h later. The numbers of neo+ clones in the absence of I-SceI expression were extremely low, as frequencies from transfection of the empty vector (pCAGGS) ranged from <2 × 10−6 to 10−7.
For S2neo DSB-induced HR gene targeting assays, cells containing S2neo were cotransfected by electroporation of 25 μg of either pneo-WT or pneo-10mu and 25 μg of either pCBASce or pCAGGS (250 V; 950 μF). G418 (200 μg/ml) was added 24 h later, and colonies were fixed and stained with Giemsa 10 to 12 days later (ES cells) or 18 to 20 days later (human fibroblasts). neo+ frequencies were determined as with the intrachromosomal HR and plasmid HR assays, except for experiments with Blmtet/tet cells. When we calculated the neo+ frequency with the Blmtet/tet cells, we also accounted for a 1.4-fold decrease in colony formation when BLM was suppressed.
An 825-bp fragment was amplified from the chromosomal neo gene in neo+ recombinants from H-DR-10mu cells or S2neo cells transfected with pneo-10mu. The primers were Neo1 (5′-GCCAATATGGGATCGGCCATTGAACAA) and Neo3 (5′-CCTCAGAAGAACTCGTCAAGA) (15) (Fig. (Fig.2A,2A, primers 1 and 2). Amplification was performed by denaturation at 94°C for 2 min, followed by 35 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 45 s, with extension at 72°C for 7 min. Amplified products were purified and sequenced for conversion of each polymorphism, as well as other sequence changes. hDNA was determined by overlapping double peaks in the sequencing chromatogram using 4Peaks software (version 1.7.2; Mekentosj, Amsterdam, Netherlands).
To analyze changes in donor sequence, a region specific to the pneo-10mu sequence of H-DR-10mu was amplified from neo+ recombinants from H-DR-10mu ES cells. PCR was preformed as described above, using primers H-DR-neo1. (5′-ATCAGCAGCCTCTGTTCCAC) and H-DR-neo1a (5′-GACATCTCTGTAGCCCCGTAA) (Fig. (Fig.2A,2A, primers 3 and 4). PCR products were purified and sequenced.
To confirm doxycycline suppression of BLM expression in Blmtet/tet ES cells after each transfection, 60 μg of protein lysates was loaded and analyzed by Western blotting using an antibody that recognizes mouse and human BLM (AB 476; AbCam, Cambridge, MA). BLM expression was also confirmed in GM00637 cells, and lack of expression was confirmed in GM08505 cells under the same conditions (data not shown).
To measure intrachromosomal homologous recombination, the H-DR-WT substrate was constructed and targeted to E14 (wild-type) and MMR-defective (Msh2−/−) murine ES cells (15). Briefly, the substrate contains two nonfunctional copies of the neomycin resistance gene, neo (Fig. (Fig.2A).2A). The first copy, S2neo, contains the full-length coding sequence and 5′ and 3′ untranslated regions (UTRs). However, an 18-bp I-SceI recognition sequence inserted at the original NcoI site includes an in-frame nonsense mutation, resulting in a nonfunctional gene. The downstream copy of neo, pneo-WT, contains the original uninterrupted NcoI site. However, truncations on both the 5′ and 3′ ends result in a nonfunctional product as well. Upon expression of I-SceI endonuclease by electroporation of the pCBASce expression plasmid (52), a DSB occurs at the I-SceI recognition sequence of S2neo. Subsequent recombination using the homologous pneo-WT donor can restore the NcoI site. Selection for neo+ clones identifies all homologous recombination products that resulted in a noncrossover gene conversion product.
Elliott and Jasin determined that the MMR machinery suppresses DSB-induced homeologous recombination in mouse ES cells by using a substrate H-DR-8mu with 1.2% sequence divergence in the downstream copy of neo, pneo-8mu (15). In this substrate, heterology is first encountered 46 bp to the left of the DSB (NruI) and 8 bp to the right of the DSB (NsiI). One of the limitations of the H-DR-8mu substrate was the lack of polymorphisms to the right of the DSB site, resulting in the potential to overlook gene conversion tracts that were between 8 bp (NsiI) and 106 bp (PmlI) to the right of the NcoI site; this is likely the peak region for conversion, as gene conversion tracts are typically <100 bp to the left of the NcoI site (16). Because of this limitation, we constructed an intrachromosomal repair substrate, H-DR-10mu, that contains 1.5% sequence divergence due to two additional restriction fragment polymorphisms to the right of the NcoI site in the downstream neo repeat, pneo-10mu, (Fig. (Fig.2A;2A; see also Fig. S2 in the supplemental material). H-DR-10mu was targeted to the Hprt locus of both E14 (wild-type) and Msh2−/− cells. Multiple independently targeted clones were isolated and used in subsequent analyses. Single-copy integration was confirmed by Southern blot analyses (Fig. S1A).
Recombination rates were determined by selection of neo+ recombinants after electroporation with pCBASce. As previously observed, neo+ frequencies were not statistically different in wild-type or Msh2−/− H-DR-WT cell lines (2.5 × 10−4 versus 1.5 × 10−4, respectively; P = 0.18) (Fig. (Fig.2B).2B). When the donor sequence divergence increased to 1.5% in the H-DR-10mu substrate, wild-type cells suppressed homeologous recombination 25-fold, decreasing the rate of recombination relative to H-DR-WT to 4% (P < 0.01) (Fig. (Fig.2B).2B). Previous studies support these findings as they suggest a trend where increases in heterology decrease DSB-induced gene targeting rates (15). In contrast, Msh2−/− cells failed to suppress recombination at the H-DR-10mu substrate, with levels of recombination similar to that of H-DR-WT (P = 0.13) (Fig. (Fig.2B2B).
With the additional polymorphisms to the right of the DSB, we analyzed the directionality of gene conversion tracts in both wild-type and MMR-defective cells. Conversion of polymorphisms was determined by DNA sequencing, both to simplify the analysis and, importantly, to determine if homeologous recombination is an error-free process. Using primers specific for the S2neo recipient (Fig. (Fig.2A,2A, primers 1 and 2), PCR products were sequenced for conversion at each polymorphism, with double peaks on sequencing chromatograms indicating unrepaired hDNA.
The average lengths of the gene conversion were similar in the wild-type and Msh2−/− cells (minimum, ~96 bp; maximum, ~193 bp) (see Fig. S2 in the supplemental material). Despite the similar lengths, we noted a shift in the distribution of the tracts, with 31% of the tracts in Msh2−/− cells extending to the most distant 3′ polymorphism, PmlI, compared with 10% in the wild type (P = 0.04, Fisher's exact test) (see Table Table22 and Fig. S2). As this polymorphism is proximal to the end of the homology, a nick may be generated by a flap endonuclease, biasing mismatch correction in favor of the recipient in the wild-type cells. Due to the MMR deficiency, a prevalence of unrepaired hDNA was observed in gene conversion tracts from Msh2−/− cells compared to the wild type (see Fig. S2), as previously observed (15). Only 2 of 40 (5%) wild-type tracts had at least one polymorphism suggestive of hDNA; in contrast, 18 of 35 (51%) tracts from Msh2−/− cells had evidence of hDNA. Similarly, almost half of the polymorphisms (45.1%) in the Msh2−/− tracts were “converted” through hDNA; the exception was the closest polymorphism to the DSB site (NsiI), which was fully converted in 75% of the gene conversion tracts, suggestive of double-strand gap formation close to the DSB site rather than of mismatch correction. Another four polymorphisms in the Msh2−/− tracts were not converted, creating discontinuous tracts, which were not observed in wild-type cells.
Excluding the NsiI polymorphism, the majority of the gene conversion tracts in both wild-type and Msh2−/− cells were unidirectional, with a bias to the right of the DSB (Fig. (Fig.2C;2C; see also Fig. S2 in the supplemental material). We observed that 9.3% of the tracts were bidirectional, raising the possibility that some gene conversion engages both DNA ends. When we multiply the percentage of gene conversion tracts that convert only to the left side of the break at or beyond NruI (11/75 or 15%) by the percentage of tracts that convert only to the right of the break at or beyond BspEI (29/75 or 39%) for both genotypes, we would expect ~6% bidirectional tracts. Thus, our low frequency of bidirectional tracts is consistent with two-ended invasion by synthesis-dependent strand annealing (SDSA) although other possibilities exist (see discussion).
The two existing models for the derivation of noncrossover gene conversions predict different outcomes for the donor sequence during repair. In the DSB repair model proposed by Szostak et al., the donor sequence can be altered due to its incorporation into a double Holliday junction (Fig. (Fig.1A,1A, site ii) (61). In our system, changes in the donor sequence can in principle be detected by conversion of polymorphisms in the homeologous donor sequence (pneo-10mu) to the sequence of the recipient (S2neo). However, in SDSA, the donor sequence serves only as a template for repair synthesis, and so it remains unaltered in most cases (2; reviewed in reference 43). Using primers specific to the pneo-10mu donor sequence (Fig. (Fig.2A),2A), the donor was sequenced from 40 neo+ recombinants from wild-type cells and 35 neo+ recombinants from Msh2−/− cells. Of the 75 events analyzed, none of the donor sequences was found converted to that of the recipient for any of the polymorphisms, which provides evidence for an SDSA pathway of gene conversion (Table (Table1).1). No other sequence changes were noted in the donor sequences.
To determine whether gene conversion is an error-free process, we also analyzed the recipient S2neo (now neo+) sequences. In the 75 neo+ recombinants from wild-type and Msh2−/− cells, the only observed sequence changes were the conversion of the I-SceI site and the coconverted restriction site polymorphisms from the pneo-10mu donor (192 total). No other base pair changes were noted in the sequence of the recipient from either wild-type or Msh2−/− cells (Table (Table1).1). Given that ~750 bp were sequenced from the recipient for each recombinant, the mutation rate in sequences flanking the DSB is <2 × 10−5 during gene conversion. If we consider that repair synthesis giving rise to gene conversion is likely much less than 750 bp since the average gene conversion tract length is only 96 to 193 bp, the detectable mutation rate of repair synthesis can be estimated to be less than 0.69 × 10−4 to 1.38 × 10−4. It is important to note that only mutations that preserve a neo+ gene will be detected in this assay.
We showed that recombination between sequences with 1.5% sequence divergence is suppressed 25-fold in wild-type murine ES cells. This suggests a tightly regulated system to prevent recombination between homeologous sequences. To determine if this regulation is conserved in human cells, we tested a transformed human fibroblast cell line for its ability to suppress recombination between diverged sequences. To simplify the analysis, we first tested recombination using a cotransfection approach. Thus, either H-DR-WT or H-DR-10mu was cotransfected with pCBASce, and neo+ clones were selected. In this way, recombination induced by I-SceI likely occurs in the H-DR plasmid; neo+ colonies arise by integration of the recombined plasmid. Using wild-type and Msh2−/− murine ES cells as a first test of this cotransfection approach, we found that homeologous recombination was suppressed 28-fold in wild-type cells and that this suppression was almost entirely MMR dependent (Fig. (Fig.3A)3A) (P = 0.78), similar to what we obtained with the chromosomally integrated H-DR substrates. Human fibroblasts were then tested for their ability to suppress homeologous recombination. As expected, SV40-transformed fibroblasts that were MMR proficient (GM00637) had a decrease in the frequency of neo+ colonies using the H-DR-10mu substrate relative to H-DR-WT (Fig. (Fig.3A).3A). Surprisingly, however, the reduction in recombination was only 2.3-fold with the diverged substrate (43% ± 8%; P < 0.01). Transformed human fibroblasts are therefore capable of suppressing recombination between sequences with 1.5% divergence although they appear to be less sensitive to sequence divergence than murine ES cells, at least using this plasmid assay.
Given the weak suppression of homeologous recombination in the human cells when the DSB was introduced in a plasmid, we sought to confirm these results with a system in which the DSB was introduced into the chromosome. Although in mouse cells we were able to efficiently target the H-DR substrates to the same genomic locus, in human cells we circumvented inefficient spontaneous targeting by creating cell lines with just the recipient S2neo gene integrated in the chromosome. In this way, the same locus could be used for DSB-induced gene conversion using either pneo-WT (homologous donor) or pneo-10mu (homeologous donor). When we previously applied this DSB-induced gene-targeting assay approach to mouse ES cells, we found that suppression of homeologous recombination was as strong as that observed with the H-DR substrates (17-fold reduction with pneo-10mu compared with pneo-WT) (15).
S2neo was randomly integrated into the genome of GM00637 cells (Fig. (Fig.3B),3B), and single-copy integrants were verified by Southern blotting (see Fig. S1B in the supplemental material). Three independent clones were expanded and independently tested. DSB-induced gene targeting was measured by cotransfecting pCBASce with either pneo-WT or pneo-10mu, and repair events were calculated from the rate of neo+ clones after G418 selection. We found that human cells suppressed DSB-induced gene targeting with the pneo-10mu donor sequence 2.7-fold (36% ± 3% relative to pneo-WT) (Fig. (Fig.3C;3C; see also Table S1), similar to the extent of suppression observed in the plasmid H-DR assay (Fig. (Fig.3A).3A). A similarly weak suppression of homeologous recombination was obtained with two other human cell lines (HeLa and U2OS; ~2-fold) although another mouse cell line, like the ES cells, gave a more robust suppression (transformed embryonic fibroblasts; ~7-fold) (data not shown). These results suggest that transformed human cells suppress recombination between diverged sequences but to a lesser extent than murine cells.
Human cells could be less sensitive to sequence divergence if polymorphisms between the recombining molecules were less likely to be encountered during gene conversion. This would be expected to be manifested by shorter gene conversion tracts. To test this, recombination events were analyzed from independent neo+ clones from S2neo gene targeting using pneo-10mu as a donor sequence. The neo+ gene was amplified from 58 independent clones and sequenced to determine the extent of gene conversion (Fig. (Fig.3D3D and Table Table2).2). Gene conversion tracts were generally short in human cells, such that 83% were <100 bp. Polymorphisms located 46 bp and 47 bp to the left and right of the DSB were incorporated in 17% and 41% of the recombinants, respectively. Wild-type murine ES cells gave similar results in terms of both intrachromosomal recombination with H-DR-10mu (78%) and DSB-induced gene targeting with pneo-8mu (83%) (16). As with the murine cells, a substantial percentage of tracts were unidirectional in the human cells (43% versus 37% for wild-type ES cells) (Fig. (Fig.3E).3E). Only 9% of neo+ clones converted polymorphisms on both sides of the DSB. In addition, gene conversion was biased to the right side of the DSB (34% versus 9%), similar to results in ES cells. Therefore, the similarity in conversion tracts between murine and human cells does not appear to account for the different degree of suppression of homeologous recombination.
Studies in yeast have demonstrated that, in addition to the MMR machinery, the RecQ helicase Sgs1 plays a role in suppressing spontaneous homeologous recombination (39, 59). Although mammalian cells contain five RecQ helicases, BLM is considered to be closest to Sgs1 (9). To test if BLM plays a role in suppressing homeologous recombination, we utilized SV40-transformed human fibroblasts from a Bloom syndrome (BS) patient, GM08505. This cell line carries a homozygous BLM mutation resulting in a frameshift and premature translation termination codon at amino acid 740 within the helicase domain (6-bp deletion; 7-bp insertion) (17). Using the H-DR plasmid cotransfection approach, the absolute recombination frequency was observed to be lower in BS cells than in wild-type cells although we also found that they transfected less well (see Table S1 in the supplemental material). However, in comparing the relative neo+ frequencies, a similar decrease was observed in the ratio of colonies for H-DR-10mu versus H-DR-WT (34.5% ± 10.5%) (Fig. (Fig.4A;4A; see also Table S1), indicating a similar suppression of homeologous recombination as in wild-type cells.
We extended these analyses by randomly integrating S2neo into BS human cells. Three independent clones with a single-copy integration of S2neo were identified (see Fig. S1B in the supplemental material). The recovery of neo+ clones was lower in the BS cells, even when considerable clonal variation in transfection efficiency is taken into account (see Table S1). As with the plasmid H-DR-10mu assay, DSB-induced gene targeting of pneo-10mu was suppressed to a similar extent in BS cells (31.5% ± 2.6%) (Fig. (Fig.4A;4A; see also Table S1) as in wild-type cells. Thus, BLM does not appear to affect homeologous recombination.
We analyzed 62 gene conversion tracts from neo+ BS clones derived from the S2neo gene-targeting assay to determine if BLM affects the extent of conversion (Fig. (Fig.4B4B and Table Table2).2). Similar to wild-type human cells, gene conversion tracts in BS cells were short (<100 bp; 72%), and tracts were primarily unidirectional (40% versus 8% bidirectional) (Fig. (Fig.4C),4C), with a bias to tracts on the right side of the DSB (35% versus 5%). These results indicate that BLM does not substantially alter gene conversion tracts.
Given the phenotype of sgs1Δ in yeast and that human cells appear to be less sensitive to homeology than murine cells, we also sought to examine the effect of BLM in murine cells. For this, we utilized the doxycycline-repressible murine ES cell line Blmtet/tet (69). In the presence of doxycycline, no BLM protein is detectable by Western blotting in these cells (Fig. (Fig.4D).4D). Like BS cells from patients, these murine cells have an elevated level of SCE when BLM is absent (18, 69).
We first examined overall recombination levels in the Blmtet/tet cells in the presence and absence of doxycycline using the S2neo DSB-induced gene-targeting assay. We found a small but significant decrease in the overall rate of gene targeting using the pneo-WT plasmid (1.4-fold; P = 0.015) (Fig. (Fig.4E).4E). This decrease was not due to nonspecific effects of doxycycline as recombination levels in wild-type ES cells are unaffected by doxycycline (Fig. (Fig.4E4E).
We next tested DSB-induced gene targeting with the homeologous pneo-10mu. We found that in the presence of BLM (without doxycycline), DSB-induced gene targeting with the pneo-10mu donor was suppressed 11-fold relative to gene targeting using the pneo-WT donor (Fig. (Fig.4E).4E). In the absence of BLM (with doxycycline), we found a similar decrease in gene targeting with the pneo-10mu donor (10-fold) (Fig. (Fig.4E).4E). These results suggest that, unlike with Sgs1, mammalian BLM does not play a role in suppressing recombination between diverged sequences.
Approaches are being developed to modify endogenous human genes using rare cutting endonucleases for purposes including correcting disease loci (50, 63). The approaches rely on accurate repair of the broken locus with an incoming template by gene conversion, yet little is known about mechanisms of gene conversion in human cells. The work presented here examines several aspects of HR in mouse and human cells and clarifies the role of mammalian BLM in recombination between homeologous sequences. Utilizing a repair substrate with 1.5% sequence divergence, we found that both wild-type mouse and human cells suppress homeologous recombination but that the suppression is substantially greater in mouse cells than in the human cell lines. Gene conversion tract lengths were nonetheless similar for mouse and human, with a large majority of gene conversion tracts of <100 bp. Thus, for the purpose of genome modification, efficient incorporation of polymorphisms requires them to be close to the DSB site (<50 bp). Most gene conversion tracts in both mouse and human cells extended from one end of the DSB. Additionally, donor sequences remained unchanged after gene conversion, and recipient sequences only showed conversion of polymorphisms, with no other sequence changes associated with HR. While MSH2-deficient mouse cells failed to suppress homeologous recombination, BLM-deficient mouse or human cells suppressed homeologous recombination to a similar extent as wild-type cells although we detected a small overall reduction in recombination in the BLM-deficient mouse cells. These results suggest that impairing BLM function will not enhance DSB-promoted gene modification in human cells.
The pneo-10mu donor sequence provided a more symmetrical placement of polymorphisms than our previously published donor sequence (15). Thus, beyond the closest polymorphism to the DSB site, which appears to be converted primarily by gap repair, polymorphisms were present at ~50 bp to the right and to the left of the DSB, allowing us to observe a substantially larger number of rightward conversions than previously observed (wild type, 16/40; Msh2−/−, 8/35). In fact, the two new polymorphisms (BspEI and NaeI) were the most frequently converted polymorphisms involving hDNA, such that a higher incidence of the conversions occurred to the right of the DSB (39%) than to the left of the DSB (15%). One explanation could be the asymmetry of the DSB in relation to the donor sequence homology. The length of homology is higher to the left of the DSB (527 bp) than to the right (213 bp) (Fig. (Fig.2A)2A) such that the resected left end may be more successful during homology search than the right end. Strand invasion from the left, followed by repair synthesis, would lead to conversion of the polymorphisms on the right.
Two major mechanisms for HR have been proposed, DSB repair (61) and SDSA (1, 20, 40, 42, 52) (Fig. (Fig.1).1). In DSB repair, both strands engage the donor while in SDSA one or both strands can engage the donor. The low frequency of bidirectional gene conversion tracts and the bias in conversion to one side of the DSB (the right), despite the symmetry of the polymorphisms, suggest that one-ended invasion is frequent, supporting an SDSA mechanism. The infrequent bidirectional tracts that do form could potentially arise from one-ended invasion during SDSA but with MMR correction of hDNA formed during strand invasion (42) (Fig. (Fig.1,1, site i); the presence of fewer bidirectional tracts in Msh2−/− cells, rather than more, would seem to argue for this mechanism. Alternatively, bidirectional tracts may form from engagement of both ends from either DSB repair (Fig. (Fig.1A)1A) or two-ended invasion during SDSA. In yeast, most mitotic gene conversions are also unidirectional (8, 41), with gap repair studies in Drosophila providing the best evidence for two-ended invasion during SDSA (1, 35).
Further support for an SDSA mechanism comes from sequencing the donor fragment. In DSB repair, the donor sequence as well as the recipient can undergo conversion as a consequence of hDNA repair (Fig. (Fig.1A,1A, sites i and ii, respectively), whereas in SDSA, the donor sequence only transiently participates in hDNA (Fig. (Fig.1,1, sites i) (20, 40, 42). We found that none of the polymorphisms in 75 donor sequences analyzed was converted to the recipient sequence, and none of the polymorphisms in 35 donor sequences analyzed specifically from Msh2−/− cells showed evidence of hDNA, implying a preference for the SDSA pathway during noncrossover gene conversion. Noncrossovers in yeast meiosis also do not show donor sequence alterations (25, 47), despite the physical presence of double Holliday junctions during meiosis (55), although, importantly, these intermediates give rise to crossovers with noncrossovers produced earlier by SDSA (2). We cannot rule out that a portion of recombination events use the pneo donor on the sister chromatid, which has segregated, rather than the same chromatid; however, it seems likely that the pneo repeat nearby on the same chromatid would be frequently used in HR.
Thus, the evidence supporting SDSA as the major HR pathway in mitotic mammalian cells comes from four directions: as reported here, most gene conversions are unidirectional, and none shows evidence of donor conversion; in addition, as reported previously, most mitotic conversions are not associated with crossing over (29, 52, 60), and some mitotic conversions can be coupled to nonhomologous end joining (NHEJ) (29, 51).
We along with others have found that sequence divergence substantially suppresses HR in rodent cells and that this suppression is relieved by mismatch repair deficiency (3, 15, 58, 62). For example, 1.5% sequence divergence suppressed HR 25-fold in mouse cells (Fig. 2B, H-DR-10mu), and this suppression was completely relieved by Msh2 mutation. However, the transformed human cell lines suppressed HR only 2.8-fold (pneo-10mu gene targeting) (Fig. (Fig.3C;3C; see also Table S1 in the supplemental material). This weaker suppression of homeologous recombination in human cells cannot simply be explained by a lower chance of encountering sequence divergence as the gene conversion tract lengths were similar in murine and human cells (Table (Table2).2). Two other transformed human cell lines (U2OS and HeLa) also gave a weak suppression of homeologous recombination while another murine cell line (transformed embryonic fibroblasts) gave a more robust suppression (J. R. LaRocque and M. Jasin, unpublished results). One interpretation is that mouse cells have a more robust MMR system although to our knowledge a systematic side-by-side analysis of mutation rates in mouse and human cells has not been done. Thus, why recombination in these human cell lines is not as sensitive to sequence divergence as in murine cells and whether primary human cells will also show a similar phenomenon warrant further investigation.
Studies in yeast suggest that while the MMR machinery plays a significant role in suppressing homeologous recombination, the RecQ helicase, Sgs1, also plays a role in this regulation (39, 59), and evidence in flies suggests a similar role for Drosophila BLM (30). It has been proposed that the Sgs1 helicase works together with MMR proteins to unwind recombination intermediates containing mismatches (26). The mammalian RecQ helicase BLM is considered to be the most likely homolog of Sgs1. Although it does not have a role in mismatch repair (31, 45), BLM has been reported to interact with components of the MMR machinery (44, 45), making it a good candidate to suppress homeologous recombination in concert with the MMR machinery. Contrary to results in yeast and flies, however, we found that BLM-deficient mammalian cells suppressed homeologous recombination to the same extent as wild-type cells. For example, we found that 1.5% sequence divergence reduced HR by ≥10-fold in both wild-type and BLM-deficient murine ES cells, with little effect on Msh2−/− cells; similar results were obtained in human BS cells. Although it is possible that another RecQ helicase performs this role in mammalian cells, the role of RecQ helicases in HR is complex and may reflect activities other than heteroduplex rejection.
Although homeologous recombination was unaffected, we did observe a small but significant decrease in DSB-induced HR in Blmtet/tet cells with the homologous donor when Blm expression was repressed. Human BS cells showed a similar trend. A characteristic feature of BS cells is their high rate of SCE, which would have suggested that they would be hyperrecombinogenic in our assays. Several alternatives can explain this paradox. First, increased SCE may be a result of increased lesion formation in BS cells, for example, arising from defects in DNA replication, rather than a hyperrecombination phenotype per se to repair a defined lesion. Phenotypes associated with BS cells support this, including increased spontaneous γH2AX foci (48), increased spontaneous chromatid breaks (23, 37), and increased DNA anaphase bridges (5, 6). Second, given the biochemical role of BLM in Holliday junction dissolution (67), gene conversion events may be directed toward a crossover pathway in the absence of BLM, lowering the frequency of noncrossover gene conversion. Although crossovers may be increased by BLM deficiency, it seems unlikely that this would account for the difference, given that most of our events appear to occur by an SDSA mechanism. In addition, we observed the same fold reduction in HR in a substrate with homologous sequences that can undergo crossing over (LaRocque and Jasin, unpublished). Finally, BLM may promote overall gene conversion rates by promoting strand resection (27) or even strand exchange itself (4). Although we did not observe a decrease in gene conversion tract lengths in BS cells that would indicate reduced resection, most gene conversion tracts are short and so do not rule out a role for BLM in end resection.
In yeast, sgs1Δ mutants demonstrate a hyperrecombination phenotype when spontaneous homologous recombination is analyzed (22, 65, 68), but HO-induced recombination (28) and plasmid gap repair (66) are slightly decreased, consistent with our results. In Drosophila, BLM deficiency is associated with a marked decrease in SDSA during gap repair (1, 36), which has been proposed to be related to a requirement for helicase activity to support multiple rounds of strand invasion. BLM-deficient chicken DT40 cells also show substantially reduced DSB-induced HR, especially when a donor which has a large block of heterology (55 bp) is used. Together, these results indicate that BLM promotes DSB-induced gene conversion although the extent may be small (yeast and mammals) or large (flies and chicken), depending on the organism.
This work was supported by grants R01GM54668 from the NIH and NSF0346354 from the National Science Foundation to M.J. J.R.L. was supported by NRSA Postdoctoral Fellowship F32GM084637 from the NIH.
We are grateful to the Jasin lab for helpful commentary regarding experimental design and analysis and to Kyoji Horie and Kosuke Yusa for providing the Blmtet/tet ES cells and Nathan Ellis and Karen Ouyang for providing human fibroblast cell lines.
Published ahead of print on 12 February 2010.
†Supplemental material for this article may be found at http://mcb.asm.org/.