Mechanisms of Recombination between Diverged Sequences in Wild-Type and BLM-Deficient Mouse and Human Cells

Plasmid constructs and DNA manipulations.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. 2A; see also Fig. S2 in the supplemental material).

• Open in new tab
• Download powerpoint

FIG. 2.

Homologous and homeologous recombination in wild-type and Msh2^−/− mouse ES cells. (A) Intrachromosomal recombination substrates measure homologous (H-DR-WT) and homeologous recombination (H-DR-10mu). S2neo is nonfunctional due to an insertion of the I-SceI recognition sequence at the endogenous NcoI site. pneo-WT is nonfunctional as both the 5′ and 3′ ends are truncated; pneo-10mu is similar but contains 10 silent restriction site polymorphisms that increase the divergence with S2neo to ∼1.5%. After I-SceI cleavage, noncrossover gene conversion between the two repeats results in neo⁺ recombinants, with the I-SceI converted to the NcoI site (striped). The numbers of base pairs of homologous sequences on each side of the DSB are shown. For H-DR-10mu, gene conversion may also result in the incorporation of the polymorphisms (A, ApaI; L, ApaLI; P, PstI; B, BamHI; X, XbaI; Nr, NruI; Ns, NsiI; Bs, BspEI; Na, NaeI; Pm, PmlI). Primers 1 and 2 were used to sequence the neo⁺ gene, and primers 3 and 4 were used to sequence pneo-10mu. (B) Msh2^−/− cells fail to suppress recombination between diverged sequences. The H-DR substrates were targeted to the Hprt locus in wild-type and Msh2^−/− ES cells. After I-SceI expression, the HR rate was determined by dividing the total number of neo⁺ colonies by the number of cells surviving electroporation. Rates relative to H-DR-WT are graphically represented. Data represent the mean and standard error of the mean (SEM) of four independent experiments. (C) Classes of gene conversion tracts from murine ES cells. Each tract was grouped into one of five classes (represented graphically in descending order): conversion of only the NcoI site, conversion of NcoI and NsiI (8 bp to right of DSB), conversion to right of the break (>8 bp; unidirectional), conversion to left of the break (unidirectional), and conversion to both sides of the break (bidirectional).

Cell lines and integration of repair substrates.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 Blm^tet/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 Blm^tet/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.

HR assays.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 Blm^tet/tet cells. When we calculated the neo⁺ frequency with the Blm^tet/tet cells, we also accounted for a 1.4-fold decrease in colony formation when BLM was suppressed.

Molecular analyses.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. 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. 2A, primers 3 and 4). PCR products were purified and sequenced.

Western analyses.To confirm doxycycline suppression of BLM expression in Blm^tet/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).

H-DR substrates measure homologous and homeologous recombination in murine ES cells.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. 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. 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).

Mismatch repair machinery is required for suppression of recombination between sequences with 1.5% divergence.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. 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. 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. 2B).

Unidirectional tracts are observed more frequently than bidirectional tracts.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. 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 2 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. 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 donor sequence was unaltered during gene conversion.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. 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. 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 1). No other sequence changes were noted in the donor sequences.

Mutations were not observed in the recipient during gene conversion.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 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.

Suppression of homeologous recombination is conserved in human cells although to a lesser extent than in murine cells.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. 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. 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.

• Open in new tab
• Download powerpoint

FIG. 3.

Homologous and homeologous recombination in wild-type human cells. (A) Human fibroblasts suppress recombination between diverged sequences in plasmid substrates. Recombination was determined in human fibroblasts (GM00637) by cotransfecting the H-DR-WT or H-DR-10mu substrate with the I-SceI expression plasmid and scoring for neo⁺ recombinants. As controls, wild-type and Msh2^−/− murine ES cells were similarly analyzed. Rates relative to H-DR-WT are graphically represented. Data represent the mean and SEM of four independent experiments. (B) HR by DSB-induced gene targeting. S2neo was randomly integrated into the chromosome of human fibroblasts (GM00637), and three single-copy integrants were identified (see Fig. S1B in the supplemental material). pneo-WT or pneo-10mu plasmids were then cotransfected with pCBASce. S2neo receives the break and can repair off the pneo-WT or pneo-10mu plasmids, giving rise to noncrossover neo⁺ gene conversion recombinants. (C) Human fibroblasts suppress recombination between diverged sequences during DSB-induced gene targeting. Rates relative to pneo-WT are graphically represented. Data represent the mean and SEM of three independent experiments, each containing at least two of three clones. Individual clones had similar relative rates (see Table S1 for values of each clone). (D) Gene conversion tracts in human fibroblasts from DSB-induced gene targeting. Fifty-eight neo⁺ clones derived from cells cotransfected with pneo-10mu and pCBASce were analyzed to determine the minimal extent of gene conversion. The average minimum and maximum gene conversion tracts are indicated (± SEM), where the minimum tract length is calculated to the last polymorphism converted in both directions and where the maximum tract length includes the distance to the next polymorphism in both directions. The numbers at the top of the graph are the percentages of recombinants with conversion at each polymorphism. NcoI (bold) was converted in 100% of recombinants. The number of clones represented by each tract is indicated on the right. (E) Classes of gene conversion tracts from human fibroblasts. Each tract from panel D was placed into one of five classes, as described in Fig. 2C.

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. 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. 3C; see also Table S1), similar to the extent of suppression observed in the plasmid H-DR assay (Fig. 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.

Gene conversion tracts in human cells are similar to those of murine ES 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. 3D and Table 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. 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.

The ratio of homeologous to homologous recombination is unaffected in BLM-deficient human cells.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. 4A; see also Table S1), indicating a similar suppression of homeologous recombination as in wild-type cells.

• Open in new tab
• Download powerpoint

FIG. 4.

Homologous and homeologous recombination in BLM-deficient mammalian cells. (A) Fibroblasts from BS patients were analyzed for their ability to suppress homeologous recombination utilizing the H-DR plasmid assay (left) and the DSB-induced gene-targeting assay (right). Rates relative to H-DR-WT (left) or pneo-WT (right) are represented. Data represent the mean and SEM of four independent experiments (left) or three independent experiments, each containing two of three clones (right). Individual clones had similar relative rates (see Table S1 in the supplemental material for values of each clone). (B) Gene conversion tracts in BS cells from DSB-induced gene targeting. Minimal gene conversion tracts from 61 neo⁺ clones are shown. The average minimum and maximum gene conversion tracts are indicated (± SEM), as described in the legend of Fig. 3D. The numbers at the top of the graph are the percentages of recombinants with conversion at each polymorphism. NcoI (bold) was converted in 100% of recombinants. The number of clones represented by each tract is indicated on the right. (C) Classes of gene conversion from BS cells. Each tract from panel B was grouped into one of five classes, as described in the legend of Fig. 2C. (D) BLM expression in Blm^tet/tetS2neo murine ES cells is suppressed by doxycycline (1 μg/ml for 48 h, the time of electroporation) to levels undetectable by Western blotting. (E) DSB-induced gene targeting in Blm^tet/tet and wild-type murine cells in the presence or absence of doxycycline. Homologous recombination between diverged sequences is suppressed in the presence or absence of BLM. The overall HR rate, however, is decreased about 1.4-fold in the absence of BLM. HR rates were determined by dividing the total number of neo⁺ colonies by the number of cells surviving 24 h after electroporation; the small decrease in Blm^tet/tet clonogenic formation in the presence of doxycycline was also factored in. Rates relative to pneo-WT donor sequence are graphically represented. Data represent the mean and SEM from six independent experiments.

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. 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. 4B and Table 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. 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.

DSB-induced gene targeting is decreased in BLM-deficient murine ES cells.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 Blm^tet/tet (69). In the presence of doxycycline, no BLM protein is detectable by Western blotting in these cells (Fig. 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 Blm^tet/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. 4E). This decrease was not due to nonspecific effects of doxycycline as recombination levels in wild-type ES cells are unaffected by doxycycline (Fig. 4E).

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. 4E). In the absence of BLM (with doxycycline), we found a similar decrease in gene targeting with the pneo-10mu donor (10-fold) (Fig. 4E). These results suggest that, unlike with Sgs1, mammalian BLM does not play a role in suppressing recombination between diverged sequences.

Gene conversion mechanisms in mammalian 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. 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. 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. 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. 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. 1A, sites i and ii, respectively), whereas in SDSA, the donor sequence only transiently participates in hDNA (Fig. 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. 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 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.

BLM and homeologous recombination.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 Blm^tet/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.