|Home | About | Journals | Submit | Contact Us | Français|
Sister chromatid recombination (SCR) is a potentially error-free pathway for the repair of double-strand breaks arising during replication and is thought to be important for the prevention of genomic instability and cancer. Analysis of sister chromatid recombination at a molecular level has been limited by the difficulty of selecting specifically for these events. To overcome this, we have developed a novel “nested intron” reporter that allows the positive selection in mammalian cells of “long tract” gene conversion events arising between sister chromatids. We show that these events arise spontaneously in cycling cells and are strongly induced by a site-specific double-strand break (DSB) caused by the restriction endonuclease, I-SceI. Notably, some I-SceI-induced sister chromatid recombination events entailed multiple rounds of gene amplification within the reporter, with the generation of a concatemer of amplified gene segments. Thus, there is an intimate relationship between sister chromatid recombination control and certain types of gene amplification. Dysregulated sister chromatid recombination may contribute to cancer progression, in part, by promoting gene amplification.
Double strand breaks (DSB) arising during the S or G2 phases of the cell cycle can be repaired by sister chromatid recombination (SCR), a homologous recombination pathway that uses the neighboring intact sister chromatid as a template for repair [1–4]. A major physiological trigger to SCR during replication is thought to be the stalling of a DNA polymerase complex on abnormal DNA structure [5–10]. Because of the sequence identity of the two sister chromatids, SCR could be mutationally silent and error-free. Current models propose a hierarchical relationship between SCR and error-prone mechanisms of double-strand break repair (DSBR), including non-homologous endjoining (NHEJ) and single strand annealing [11–13]. If SCR is not used, a DSB arising at a stalled fork may recombine or rejoin with a heterologous locus resulting in a chromosomal translocation or deletion. SCR is therefore an important mechanism for preventing genomic instability.
Since SCR is likely to occur at sites of DNA polymerase stalling, it is notable that several recombination proteins accumulate at sites of replication arrest in mammalian cells. These include BRCA1, BARD1, Rad51 , BRCA2/FANCD1  and the product of the Bloom’s syndrome gene, BLM . This suggests a role for these gene products in SCR . Consistent with this, primary cells lacking BRCA1, BRCA2/FANCD1, other Fanconi anemia genes or BLM develop spontaneous chromosome aberrations with predominant “chromatid-type” errors—lesions that reflect a failure of recombination during replication [7,17–20]. Other cancer genes potentially involved in SCR include H2AX [21–23], ATR , the ataxia telangiectasia gene, ATM, the Nijmegen breakage syndrome gene, NBS1, and the MRE11 gene [25–29]. S phase checkpoint proteins that function with ATR might also regulate SCR[30–32]. Thus, a number of distinct protein complexes appear to cooperate to control SCR, and the failure of this control seems to be a potent trigger to cancer.
SCR in mammalian cells has been studied predominantly by use of the “sister chromatid exchange” (SCE) assay. This allows the cytological identification of crossover events between sister chromatids, but does not provide a molecular picture of the repair event. Further, crossover events represent a minority of recombination events in somatic cells [2,33,34]. Although many ways exist to quantify homologous recombination in somatic cells [11,35–37], few address specifically the SCR pathway. A major reason for this is the difficulty of selecting for some SCR outcomes to the exclusion of other recombination pathways. Two previous molecular studies of SCR used random screening of clones to identify a number of I-SceI restriction endonuclease-induced SCR events in which “long tract gene conversion” (LTGC) between sister chromatids produced a characteristic expansion of the recombination reporter [2,4]. This approach was of limited power, since it employed Southern blot screening of clones, not selection of clones in which SCR had occurred. Spontaneously arising SCR events could not be detected by this method.
In an effort to improve our understanding of SCR control, we have developed a novel recombination reporter, termed a “nested intron” reporter, which allows the positive, specific selection of SCR events in mammalian cells. We have used this to analyze spontaneously arising SCR/LTGC events in mammalian cells, as well as I-SceI/DSB-induced events. We find that the molecular outcomes of SCR, both spontaneous and I-SceI-induced, vary qualitatively on a clone-to-clone basis. We report here a novel outcome of SCR, characterized by multiple rounds of gene amplification within the “nested intron” reporter.
The “nested intron” recombination reporter HRsub (Fig. 1) was constructed in several steps using PCR and standard cloning procedures; the resulting construct was confirmed correct by DNA sequencing. Modified gfp copies were produced by PCR from the enhanced green fluorescent gene (GFP, Clontech). The donor gfp copy (first repeat) has no promoter and lacks the first 12 amino acids of the ORF. The recipient gfp copy (second repeat) is driven by the human CMV promoter, but is inactivated by an insertion of the 18 bp I-SceI recognition site 5′-TAGGGATAACAGGGTAAT-3′, replacing 4 bp in the middle of the natural GFP sequence and creating a premature stop codon (underlined). The blasticidin-S-deaminase gene (BsdR) expression cassette was constructed by PCR from the pUB6/V5-HisC plasmid (BsdR cDNA; Invitrogen) and from the pSG5 vector (rabbit β-globin intron II; Stratagene). The SV40 early promoter driving BsdR expression was amplified from pUB6/V5-HisC and the BGH polyadenylation signal was amplified from vector pcDNA3 (Invitrogen). The HRsub reporter was assembled in pBlue-script II SK(−) (Stratagene), then subcloned into a modified version of the pPUR vector (Clontech), pPURO’, in which the promoter driving expression of the puromycin resistance gene was replaced by the PGK promoter from pMSCVpuro (Clontech) and the polyadenylation signal replaced by a PCR fragment amplified from pEGFP-N3 (Clontech) containing the HSV TK polyA. Test constructs described in Fig. 2 were produced from the same PCR fragments and similar procedures, and were subcloned into vector pcDNA3 (Invitrogen). The I-SceI expression vector, pcDNA3βmycNLS-I-SceI, is a modified version of pCMV-I-SceI  containing the I-SceI coding sequence fused at its 5′ end to a triple myc tag and a nuclear localization signal, in vector pcDNA3β .
Genomic DNA (gDNA) was extracted from ~5 × 106 cells using the Puregene DNA Isolation Kit (Gentra Systems). PCR analysis of BsdR+ clones was carried out using 200 ng of gDNA and primers F (Bsd gene-mut-Fw 5′-atggccaagccattgtctcaa) and R (Bsd gene-Rev 5′-tagccctcccacacataacc) to amplify a 2.6 kb fragment from BsdR exon A to BsdR exon B (Expand Long Template PCR System, Roche). The PCR products were digested with I-SceI (New England Biolabs) and analyzed by 1% agarose gel electrophoresis (Fig. 3).
Southern blotting analysis was performed using 5 µg of gDNA as described previously . A 0.7 kb PCR fragment containing the complete GFP coding sequence was used as a probe (Random-primed DNA labeling kit; Roche). This product was amplified from pcDNA3–wtGFP (Fig. 2) using primers EGFP gene-Fw 5′-atggtgagcaagggcgagg and EGFP gene-Rev 5′-ctttacttgtacagctcgtcc.
Total RNA was prepared from 2 × 106 to 5 × 106 electroporated cells, selected in either G418 (400µg/ml) or blasticidin (5 µg/ml), by using the RNeasy Mini kit (Qiagen). Northern hybridization was performed with a 0.4 kb PCR fragment probe containing the complete BsdR coding sequence. This product was amplified from pcDNA3–wtBsdR (see Fig. 2) using primers F and R (Fig. 4) and labeled using the BrightStar Psoralen-Biotin labeling kit (Ambion). Detection was performed using the BrightStar Biodetect kit (Ambion).
Cell lines were cultivated in DMEM medium supplemented with 10% Fetal Clone I (Hyclone) at 37 °C and 6% CO2. To generate stable cell lines, 6 × 106 U2OS cells were electroporated with 1 µg of linearized plasmid DNA in a 0.4 cm-electrode gap cuvette (BioRad Gene Pulser; 960µF/250 V). The constructs were linearized with AflIII prior to electroporation. To construct the U2OS reporter cell lines #18 and #24, puromycin (2 µg/ml) was added to the medium 48 h after electroporation. After 2 weeks under continuous selection, puromycin-resistant colonies were isolated and screened by Southern-blotting with the GFP probe. The U2OS Clones #18 and #24 were determined to contain a single, complete, randomly integrated copy of the 7.8 kb HRsub reporter. Seven different restriction enzymes were used for Southern analysis: BamHI, EcoRI, PstI, EcoRV, HindIII, SpeI and SmaI (data not shown). In subsequent transfections (0.5–1.0) × 106 cells were transfected with 25 µg of pcDNA3βmycNLS-I-SceI, using a modified calcium phosphate-mediated transfection procedure.
Recombination was induced by transient transfection of cells with pcDNA3βmycNLS-I-SceI. To assay GFP expression, cells were analyzed 4–5 days post-transfection using a Becton–Dickinson FACScan on a green (FL1) versus red (FL2) fluorescence plot with 25% FL2–FL1 compensation to visualize GFP+ cells . Transfection efficiency was measured by flow cytometry for a calcium phosphate-mediated co-transfection of pcDNA3βmycNLS-ISceI with pcDNA3–wtGFP (ratio 10:1), the percentage of green cells representing in this case the percentage of transfected cells. On average, the transfection efficiency of I-SceI or control plasmid was >75%. To assay SCR events, 24 h post-transfection, cells were trypsinized, counted and replated at 50,000, 100,000 and 200,000 per 10 cm dish (duplicates). Blasticidin (5 µg/ml, Invitrogen) was added to the medium 4 days after plating. After 2 weeks under continuous selection, BsdR+ colonies were either stained and counted to determine frequencies, or isolated and expanded for molecular analysis of genomic DNA. Staining of colonies was performed using fixation in 10% methanol/10% acetic acid, followed by incubation in 0.4% crystal violet/20% ethanol. Viability and plating efficiency were determined by seeding 200 cells per 6 cm dish (quadruplicates) into non-selective medium, incubation for 2 weeks and colony staining. On average, the plating efficiency of each clone was ~30–40%.
We based the “nested intron” reporter on a standard recombination reporter containing tandem mutant copies of the gene encoding enhanced green fluorescent protein (GFP, Fig. 1). The first GFP copy is truncated at the 5′-end; the second is full-length, with an interruption by an 18 bp sequence recognized by the rare-cutting restriction endonuclease, I-SceI . Because of the orientation of the GFP direct repeats, homologous recombination between these two copies, but not single strand annealing, could generate a wild-type GFP gene. “Short tract” gene conversion (STGC) could arise from either inter-chromatid recombination (SCR) or intrachromatid recombination (Fig. 1A). LTGC or crossing over between sister chromatids could each generate an expansion of the reporter resulting in triplication of the GFP copies (Fig. 1B; ). Since LTGC and crossing over are both SCR events, “GFP triplication” is diagnostic of SCR. In yeast and mammals, crossing over is suppressed in somatic cells [2,33,41]. Hence, LTGC is the major pathway leading to “GFP triplication” and we have accordingly termed this outcome “SCR/LTGC”. It should be noted that a small proportion of “GFP” triplication events might in fact represent crossing over associated with SCR.
In the “GFP triplication” outcome, the region between the GFP copies (indicated by a red star, Fig. 1B) is duplicated. To select for this duplication event, we inserted into this region an expression cassette containing two artificial exons (A and B) of the blasticidin-S-deaminase gene (BsdR; Fig. 1C). BsdR exons A and B were placed “head-to-toe”, so that wild-type BsdR is not expressed when the reporter is in its original, unrearranged form. Intron sequences were placed to create a splice donor site 3′ of exon A and a splice acceptor site 5′ of exon B. An SV40 promoter was inserted 5′ of exon A and a polyadenylation signal 3′ of exon B (Fig. 1C). Cells containing the unrearranged reporter or the products of STGC are sensitive to blasticidin (BsdR−) since the two BsdR exons are not properly orientated with respect to each other. In contrast, an LTGC or cross over event could duplicate the BsdR cassette, resulting in the positioning of one BsdR exon A in a correct orientation with respect to exon B of the neighboring, duplicated BsdR cassette (Fig. 1D). The corresponding transcript (initiated at promoter α, Fig. 1D) could then be spliced to generate a wild-type BsdR mRNA, making the cell blasticidin resistant (BsdR+). The BsdR intron would contain a “nested” wild-type GFP gene (Fig. 1D). Through analysis of a series of test constructs mimicking the “nested intron” rearrangement (Fig. 2 and Table 1), we found that optimal expression of BsdRwas obtained when the BsdR andGFPgenes were transcribed in similar orientations, and we designed the final reporter, termed “HR sub”, accordingly (Fig. 1C). Other test constructs revealed no detectable deleterious effect of the GFP polyadenylation signal upon expression of BsdR (data not shown). This is consistent with in vitro data showing that a polyadenylation signal encountered within an intron is suppressed pending completion of the splicing event [42,43].
We electroporated the recombination-proficient human osteosarcoma cell line, U2OS, with the linearized HR sub reporter and used Southern blotting and PCR to identify two clones, #18 and #24, out of a total of ~60 independent clones examined, which contained only one full-length copy of the reporter (data not shown, see also Figs. 5 and and6).6). We used flow cytometry for GFP to quantify gene conversion in the reporter, in each of the clones. In addition, we quantified SCR/LTGC by counting BsdR+ colonies after selection of cells in blasticidin. In the same assays, parental U2OS cells, lacking any recombination reporter, generated no GFP+ and no BsdR+ events.
Passage of Clones #18 and #24 in tissue culture revealed low frequencies of spontaneously arising GFP+ cells at ~45 × 10−6 (Clone #18) and ~30 × 10−6 (Clone #24, Fig. 3A) and of BsdR+ cells at ~32 × 10−6 (Clone #18) and ~3 × 10−6 (Clone #24, Fig. 3B). Thus, some spontaneous recombination events at this locus (as measured by the generation of a GFP+ cell) occurred by a SCR/LTGC mechanism (giving rise to a BsdR+ cell). However, as discussed in detail below, the majority (>95%) of spontaneous recombination events produced BsdR+ clones that were GFP−. Because of this, it is difficult to estimate accurately the ratio of spontaneous SCR/LTGC:GC events. These measurements of spontaneous recombination represent cumulative frequencies of stochastic events—parameters only indirectly related to the actual spontaneous recombination rates at the loci in question. The measured frequencies of spontaneous GFP+ events varied ~2-fold between U2OS Clones #18 and #24. For each of U2OS #18 and #24, the frequencies of spontaneous BsdR+ events varied up to 10-fold. This wider variability is probably a reflection of the very low actual numbers of BsdR+ colonies counted in such assays (typically, between 3 and 30 per million cells).
The endonuclease I-SceI can be used to generate a site-specific DSB within the recombination reporter and thereby promote gene conversion at the reporter locus. Transfection of an I-SceI-encoding plasmid into Clones #18 or #24 strongly induced both GFP+ and BsdR+ events (Fig. 3A and B). To obtain accurate values, the frequencies noted here have been corrected for transfection efficiency and (for BsdR+ events) plating efficiency of the cells (see Section 2). More than 95% of I-SceI-induced BsdR+ colonies were found to be GFP+ (as predicted for the SCR/LTGC outcome; Fig. 1D). This high percentage of GFP+ clones is to be expected, since the DSB induced by I-SceI in the second copy of the GFP gene is repaired using the wild-type sequence of the first GFP copy on the sister chromatid. Results of four independent experiments using U2OS Clone #18 are summarized in Table 2. In these experiments, on average ~7.0% of I-SceI-transfected cells became GFP+ and ~1.1% of transfected cells became BsdR+. In one experiment using U2OS Clone #24, ~3.8% of I-SceI-transfected cells became GFP+ and ~1.3% became BsdR+. Thus, ~16% (Clone #18) and ~35% (Clone #24) of I-SceI-induced GFP+ cells were also BsdR+. These results suggest that between 1/6 and 1/3 of all I-SceI-induced gene conversion events that lead to the formation of wtGFP engage a SCR/LTGC mechanism. In summary, therefore, by examining two independent U2OS clones carrying single copies of randomly integrated “nested intron” reporters, we found a 2-fold variation in I-SceI-induced GC events (measured as GFP+), and a 2-fold variation in the ratio of I-SceI-induced SCR/LTGC:GC events (measured as BsdR+:GFP+).
To confirm that the expected splicing events had occurred within the BsdR transcript in BsdR+ cells, we analyzed Northern blots from pooled I-SceI-induced BsdR+ colonies, using a BsdR probe. This revealed a processed BsdR transcript that co-migrated with wild-type BsdR mRNA (pcDNA3–wtBsdR; Fig. 2B and Fig. 3C), as well as with the processed BsdR mRNA expressed from a series of test constructs containing appropriately orientated BsdR exons A and B separated either by a simple intron (p Test BsdR A+B; Fig. 2C and Fig. 3C) or by an intron containing a functional expressed GFP gene (pTest SAMOR; Fig. 2D and Fig. 3C). The insertion of a simple, intact intron within the BsdR gene dramatically increased expression of BsdR (compare constructs depicted in Fig. 2B and C; BsdR expression levels shown in Fig. 3C). The presence of a functional GFP gene within the “nested intron” reduced expression of the BsdR gene (compare constructs depicted in Fig. 2C and D; BsdR expression levels shown in Fig. 3C). The cause of this “knock-down” effect is not known, but may reflect promoter interference within the “nested intron” construct (pTest SAMOR; Fig. 2D).
To determine the mechanism(s) responsible for generating BsdR+ events, we prepared genomic DNA from 88 individual BsdR+ clones (27 spontaneously arising, 61 I-SceI-induced), and analyzed the recombination breakpoint within the “nested” intron by PCR, using a forward primer specific for BsdR exon A and a reverse primer specific for BsdR exon B (Fig. 4). By this arrangement, only those clones that contained the duplicated BsdR cassette should generate a PCR product; this was in fact the case (Fig. 4 and data not shown). If the breakpoint leading to duplication of the BsdR cassette were homologous (reflecting recombination between GFP repeats), the PCR product generated across the “nested” intron should be 2.6 kb (Fig. 4A). Conversely, a non-homologous mechanism of BsdR cassette duplication might generate PCR products across the “nested” intron of sizes other than 2.6 kb. We found that all PCR products analyzed from 88 distinct BsdR+ clones (whether spontaneously arising or I-SceI-induced) were 2.6 kb in size (Fig. 4B and data not shown), consistent with a mechanism of BsdR cassette duplication caused by homologous recombination between two GFP repeats.
Surprisingly, all but one spontaneously arising BsdR+ clones (26/27; 96%) were GFP− and gave rise to PCR products that contained an intact I-SceI site (Fig. 4B, S1 and S2). The one remaining BsdR+ clone was GFP+ and gave rise to a PCR product lacking an I-SceI site (data not shown). Sequencing of a number of PCR products from different BsdR+ GFP− clones showed that the GFP copy “nested” within the wt BsdR intron was identical in sequence to the I-SceI-containing GFP copy of the original, unrearranged reporter (data not shown). This indicates that, in these clones, the site for recombination within the I-SceI-containing GFP copy lay 3′ of the I-SceI restriction site, thus preserving the I-SceI site in the final recombination product and leading to a GFP− clone. The fact that there was a bias in favor of GFP− status for spontaneously-arising BsdR+ events implies that the preferred site for spontaneous recombination within the I-SceI-containing GFP copy lies 3′ of the I-SceI site. The reasons for this observed preference are unclear. However, this might be a reflection of the available sequence homology either side of the I-SceI site, the 5′ homology being 353 bp and the 3′ homology 569 bp. In studies of spontaneous recombination between episomal plasmids, the efficiency of recombination was found to increase as a function of available sequence homology . The shortest tract of homology sufficient to support recombination between episomal plasmidswas found to be ~200 bp . Perhaps the 5′ tract of 353 bp is less efficient in supporting recombination than the more extensive 3′ tract of 569 bp.
In the analysis of I-SceI-induced BsdR+ clones, PCR products derived from GFP+ clones were found to lack an I-SceI site (Fig. 4B, I1 and I2). We also analyzed the rare GFP− BsdR+ clones present in I-SceI-transfected cultures. All but one PCR product from these GFP− clones were found to contain an intact I-SceI site, suggesting that the restriction site had not been cut by I-SceI. These clones therefore represent “background” spontaneously arising SCR/LTGCevents, more than 95% of which would be expected to retain an I-SceI site and be GFP− (as noted above). One exception to this pattern was a single BsdR+ GFP− clone, derived from an I-SceI-transfected cell culture, which generated a PCR product of apparently unaltered size (2.6 kb) that paradoxically lacked an I-SceI site. Sequencing of the “nested” GFP copy of this clone revealed an 8 bp deletion within the I-SceI site, indicating that I-SceI had cut this site and suggesting that the DSB was subsequently repaired by NHEJ. It is possible that this represents a novel, NHEJ-mediated fusion event between sister chromatids. A more parsimonious interpretation is that it represents non-homologous end-joining of a cleaved I-SceI site, occurring in a spontaneously arising BsdR+ GFP− cell that was already present at the time of I-SceI transfection. Given the existence of “background” spontaneously arising BsdR+ GFP− clones in I-SceI transfected cultures, the preferential retention of an intact I-SceI site in the “nested” GFP copy in such clones and the high efficiency of end-joining in mammalian cells, the latter interpretation seems more likely.
To determine the rearrangements associated with BsdR+ status, we first analyzed the structure of the reporter locus in BsdR+ colonies, using Southern blotting with GFP as a probe (Fig. 5A). The entire reporter can be retrieved as one GFP-hybridizing fragment in genomic DNA digested with PstI. In this case, the unrearranged reporter locus of the parental Clones #18 and #24 produced a GFP-hybridizing fragment of 5.9 kb (Fig. 5A; Fig. 5B and C, parental lane “P”). Conversely, any expansion in the size of the reporter should cause an increase in the size of the PstI restriction fragment. The predicted size of this fragment in the “GFP triplication” outcome is 9.1 kb (Fig. 5A). The BsdR cassette contains a HindIII site (Fig. 5A). By digesting genomic DNA with HindIII, we therefore expected to separate the two GFP copies of the original unrearranged reporter into two GFP-hybridizing fragments—one fragment of 3.05 kb and one larger fragment, the size of the latter depending upon the location of the nearest HindIII site on the chromosome 3′ of the reporter locus. This was the case (Fig. 5D and E, parental lane “P”). The larger fragment varied between Clones #18 and #24, since the reporters were randomly integrated. If a clone had undergone “GFP triplication”, this would entail duplication of the BsdR cassette and should therefore introduce a second HindIII site within the reporter, resulting in the separation of all three GFP copies onto different sized genomic DNA fragments (Fig. 5A). Digestion of genomic DNA with HindIII should now generate a new GFP-hybridizing fragment of 3.2 kb in addition to the pre-existing 3.05 kb and larger fragments.
We first examined the structure of the reporter in spontaneously arising BsdR+ clones. In both Clone #18 and Clone #24, the majority of BsdR+ clones examined revealed a Southern-blot GFP hybridization pattern consistent with “GFP triplication” (Fig. 5B–E). Although most of these BsdR+ clones contained the expected three GFP copies, a minority revealed aberrant SCR/LTGC events (Fig. 5B and D, Clones #2 and #4), in which only two GFP copies were detected in the expanded reporter. PCR analysis of the “nested intron” of these clones, using the above-noted BsdR primers (Fig. 4), revealed the expected 2.6 kb PCR product, indicating that a SCR/LTGC event had indeed occurred. Repair events of this type must have entailed a gene conversion tract sufficiently long (>1.03 kb) to duplicate BsdR exon B, but not long enough to triplicate the GFP genes (i.e., <3.22 kb).
Southern analysis of some spontaneously arising BsdR+ clones in U2OS #24 cells revealed a retained, apparently un-rearranged reporter locus in addition to the expanded reporter locus (Fig. 5C and E, #3, #5 and #6). These hybridization patterns may reflect non-dysjunction of sister chromatids occurring in the same cell cycle as the spontaneous SCR/LTGC event. The significance of these observations is unclear at present.
Next, we analyzed the structure of the rearranged reporter in I-SceI-induced BsdR+ GFP+ colonies by Southern blotting, with use of a GFP probe (Fig. 6). For all 61 clones examined, the PstI digest revealed a GFP-hybridizing Southern blot fragment larger than the 5.9 kb of the original parental reporter locus. As was noted for spontaneously arising SCR/LTGC events, most I-SceI-induced BsdR+ clones examined revealed “GFP triplication” within the reporter. However, as was noted in spontaneously arising BsdR+ clones, some I-SceI-induced clones contained rearrangements suggestive of early termination of the “long tract” (Fig. 6A and C, Clone #5; Fig. 6B and D, Clones #3, 4, 5, 7). In total, 23/27 (85%) I-SceI-induced BsdR+ GFP+ clones of U2OS #18 revealed the typical “GFP triplication” outcome (see Fig. 1D), and 4/27 (15%) revealed aberrant patterns of LTGC/SCR. Equivalent frequencies for U2OS #24 were 28/34 (82%) and 6/34 (18%), respectively.
Southern analysis of some I-SceI-induced BsdR+ colonies revealed two copies of the reporter—both the “GFP triplication” outcome and the parental reporter (data not shown). Invariably, a proportion of cells in these colonies were GFP−. Recloning of these colonies, initially in the absence and later in the presence of blasticidin, followed by assessment of GFP status and Southern analysis, revealed two distinct types of sub-clone. The first was GFP+ BsdR+, and contained only the “GFP triplication” reporter structure. The second was GFP−BsdR−, and contained only the unrearranged parental reporter structure (data not shown). The existence of these colonies of mixed reporter structure, and the separation of reporter types by recloning, was also noted by Johnson and Jasin . Perhaps the initial survival in blasticidin of those cells that contained the parental reporter structure (and were therefore genetically BsdR−) reflects the retention of wild-type BsdR mRNA by both daughter cells at the mitosis following the recombination event.
Surprisingly, a small proportion (~3%) of I-SceI-induced BsdR+ GFP+ clones (1/27 for U2OS #18; 1/34 for U2OS #24) revealed unexpected gene amplification events within the reporter. In Southern blots of these clones, the GFP probe hybridized intensely with only one greatly expanded fragment of PstI-digested genomic DNA (Fig. 6A, U2OS #18 lane 4; Fig. 6B, U2OS #24 lane 6). In the same clones, the HindIII digest, which cuts within the BsdR cassette, revealed an amplified fragment of 3.2 kb (Fig. 6C, U2OS #18 lane 4; Fig. 5D, U2OS #24 lane 6)—the size predicted for SCR-mediated duplication of the BsdR cassette (Fig. 5A). In one clone (U2OS #18, Fig. 6A and C, lane 4), densitometric analysis suggested that the BsdR cassette had been quadruplicated (Fig. 6E); in another (U2OS #24, Fig. 6B and D, lane 6), it had been triplicated. PCR performed on genomic DNA from these clones, using primers spanning the “nested” intron (Fig. 4A), revealed a 2.6 kb product lacking an I-SceI site in each case. These rearrangements therefore arose from multiple rounds of gene amplification within the reporter, mediated by homologous recombination between GFP repeats of two sister chromatids.
This work describes a novel molecular analysis of sister chromatid recombination in mammalian cells. By developing a “nested intron” reporter for the positive selection of “long tract” gene conversion events between sister chromatids, we have made quantitative analysis of this pathway technically straightforward. This has allowed us to extend the analysis of SCR beyond the scope of previous work, which relied upon random screening of clones to identify and analyze a small number of I-SceI-induced SCR events . Here, we have identified spontaneously arising SCR/LTGC events for the first time, proving that this pathway can be engaged during normal cell growth and that SCR/LTGC is not unique to I-SceI-mediated DSBs. We have also analyzed I-SceI-induced SCR/LTGC events in detail. This has produced the surprising finding that SCR can initiate gene amplification and mediate the subsequent concatemerization of the amplified locus.
Johnson and Jasin  estimated the frequency of I-SceI-induced SCR/LTGC events in the Chinese hamster cell lines V79 and AA8 as between 16 and 33% of all measurable I-SceI-induced recombination events. Using the human osteosarcoma cell line, U2OS, we estimate this ratio to be ~16 and ~35% for two distinct reporter lines with different (random) sites of integration. Despite the fact that we used a cell line different to those used by Johnson and Jasin and that the reporters were integrated at random loci, our estimate of this ratio agrees closely with that made by Johnson and Jasin. The “nested intron” reporter provides a means to study the genetic basis of the “choice” between short and long tract gene conversion pathways. Notably, U2OS cells are hypertriploid and, being a cancer cell line, have numerous genetic alterations. To study the genetic control of SCR, it may be advantageous in the future to study SCR “nested intron” reporters in genetically defined cells. Our current work suggests that the reporter functions in a similar fashion in at least one other cell type (N.P., unpublished observations).
It is not clear to what extent spontaneous SCR will resemble I-SceI-induced LTGC events. I-SceI could induce SCR at any cell cycle stage subsequent to replication of the reporter locus, including the G2 phase of the cell cycle. Even during S phase, DSBs induced by I-SceI need not be generated at or close to the replication fork. In contrast, spontaneously arising SCR in somatic cells may be triggered predominantly by DNA polymerase stalling, which would restrict the initiation of SCR to stalled replication forks in S phase. Further, although replication-related DSBs can arise in pathological situations [27,46–48], it is not clear that DSBs are necessary physiological triggers to SCR. Potentially, single-stranded DNA at stalled forks might stimulate SCR [49–51].
Whether spontaneously arising or I-SceI-induced, SCR/LTGC generates gene conversion tracts of variable length. In the I-SceI-induced SCR/LTGC events reported here, a majority (82–85%) of the tracts entailed triplication of the GFP copies. Other tracts terminated earlier, indicating some heterogeneity in the control of termination of the long gene conversion tract. Conceivably, factors that influence heteroduplex stability, branch migration or Holliday junction resolution might influence the extent of LTGC. Candidate regulators of these processes include, at a minimum, the Rad51 paralogs XRCC3 and Rad51C [52,53], BLM  and potentially mismatch repair genes. Short gene conversion tracts in mammalian cells commonly extend less than 100 bp [35,55]. In contrast, “GFP triplication” implies a gene conversion tract of at least 3.22 kb. These long tracts presumably require a negotiation between the branch migration process and chromatin neighboring the DSB. Notably, chromatin is modified extensively around DSBs, with the phosphorylation of the C terminal tail of histone variant, H2AX, and recruitment of associated proteins such as 53BP1, MDC1 and BRCA1/BARD1 [56–58]. In yeast, SIR silencing proteins relocalize to chromatin near a DSB [59,60]. Perhaps one function of these chromatin responses is to control resolution of the gene conversion tract. If so, “long tract” gene conversion might represent a failure of this regulation, perhaps loosely analogous to break-induced replication in yeast . The “GFP triplication” outcome specifies a minimum gene conversion tract length (3.22 kb) but not a maximum tract length. Some “GFP triplication” events may therefore entail gene conversion tracts longer than 3.22 kb.
A surprising feature of some SCR events described here is the amplification and concatemerization of the BsdR cassette within the reporter. These events were rare, constituting only ~3% of all I-SceI-induced BsdR+ events. Analysis of the amplified segment showed that it had been generated by homologous recombination between GFP repeats. Possibly, these events reflect repeated DSB induction by I-SceI and repeated engagement of an SCR pathway (whether LTGC or crossing over). However, each clone containing concatemers of BsdR cassettes revealed only one expansion form, rather than a mixture of intermediate forms containing fewer expansions. This raises the possibility that these amplification outcomes arose from single recombination “events”, rather than by repetitive breakage induced by I-SceI. Whatever the mechanisms involved, these findings indicate that SCR can both initiate gene amplification and mediate the further concatemerization of the amplified locus.
Rare examples of gene amplification have been reported in the normal development of certain model organisms, such as “magnification” in Drosophila rDNA genes [62,63]. Gene amplification can also arise pathologically in oncogene loci of cancer cells. Early gene amplification is intrachromosomal, indicating that dysregulated sister chromatid interactions initiate gene amplification . Interestingly, mismatch repair genes have been implicated in suppressing gene amplification , a finding consistent with the role for recombination identified here.
The boundaries of some oncogene amplicons have been found to coincide with chromosomal “fragile sites” [46,66]—loci that are prone to persistent breakage in the presence of agents that arrest DNA synthesis. Fragile site expression is controlled by the S phase signaling kinase, Atr, a key mediator of the cell’s response to replication arrest . Together, these observations suggest that fragile site breakage is initiated at or close to stalled replication forks. If so, fragile site expression and subsequent gene amplification might 160 result from dysregulated SCR control. “Nested intron”—type reporters will serve as useful tools to test this hypothesis, and to determine the specific roles of cancer genes such as BRCA1, BRCA2, Fanconi anemia genes and BLM in this process.
We thank members of the Scully lab for helpful discussions. This work was supported by Grants R01CA95175 and K01CA79576 and by a Pew Scholars Award to R.S.