INTRODUCTION

Double strand breaks in chromosomal DNA caused by ionizing radiation or other DNA damaging agents can compromise the genetic integrity of cells. To repair chromosomal breaks, complex mechanisms have been evolved which include both homologous and nonhomologous processes. Recombinational repair of chromosomal breaks in mammalian cells has been demonstrated by gene targeting, the homologous recombination of chromosomal DNA with extrachromosomal DNA (1, 2). Broken chromosomal ends in mammalian cells can also be rejoined by mechanisms that involve little or no homology at the DNA ends, frequently resulting in small deletions at the site of the break (1, 3, 4, 5). Homologous recombination and DNA end joining are also used to repair double strand breaks in extrachromosomal DNA, whether the breaks are introduced in vitro (6, 7) or in vivo (8).

Details of double strand break repair mechanisms are not well understood in mammalian systems. However, radiosensitive mutants have been isolated which are defective in double strand break repair, including the hamster xrs-6 cell line (9) and the scid mouse (10). The genes defective in xrs-6 and scid cells encode components of a DNA end-binding complex called DNA-PK,¹ for DNA-dependent protein kinase (11, 12, 13, 14). One of the components of this complex is the Ku autoantigen, a heterodimeric DNA-binding protein of 70- and 80-kDa subunits (15). Ku binds avidly to DNA ends, whether blunt or with 5 or 3 overhangs (16), as well as to other discontinuities in DNA structure (17).

The other component of the DNA end-binding complex is DNA-PK_cs, the 350-kDa catalytic subunit of the protein kinase which is activated upon binding to Ku (18). The DNA-PK complex phosphorylates a number of proteins, including p53, SV40 T antigen, and Ku itself (19). Recently, it has been demonstrated that the defect in xrs-6 cells can be complemented by a gene encoding the Ku80 component of the complex (12), whereas scid cells are complemented by the gene encoding the DNA-PK_cs component (13, 14).

Scid mice, originally identified by their immunodeficiency, have a defect in V(D)J recombination, the process by which antigen receptor genes are rearranged (10). Recent analysis has demonstrated that xrs-6 cells are also defective for V(D)J recombination (20), implicating both components of the DNA-PK complex in this site-specific recombination reaction. V(D)J recombination is initiated by the RAG1/RAG2 proteins at specific signal sequences (21). The intermediates in this process are unique for a site-specific recombination system, DNA hairpins at the ends of the antigen receptor coding sequences and blunt ends at the adjacent signal sequences. Completion of the reaction leading to signal joints and coding joints requires components of the generalized cellular double strand break repair machinery, presumably including the DNA-PK complex, although the details are not well understood.

To understand better the nature of the repair defect in xrs-6 and scid cells, we have analyzed the repair of double strand breaks in extrachromosomal substrates soon after their transfection into cells. Two plasmids are cleaved prior to transfection to measure double strand break-promoted homologous recombination and DNA end joining. A third plasmid is transfected to monitor DNA stability. Extrachromosomal DNA is then recovered from cells beginning 2 h after transfection. Although scid cells did not manifest defects in the repair of double strand breaks, Ku80-deficient xrs-6 cells showed enhanced degradation of transfected DNA and decreased precise end joining.

EXPERIMENTAL PROCEDURES

DNA manipulations were performed according to standard procedures (22). Plasmids Mneo, M5neo, and M3neo were described previously (23). pBRtet was generated from pBR322 by deleting the 311-bp ScaI/SspI fragment of the ampicillin resistance gene. Prior to transfection, M5neo was cleaved with SphI, M3neo was cleaved with PstI/AatII, and Mneo was cleaved with PstI.

Cell lines were cultured in 150-cm² tissue culture flasks and transfected by electroporation. Cells were harvested at subconfluence and resuspended in phosphate-buffered saline. 10⁸ cells were mixed with 150 µg of each plasmid in a final volume of 0.8 ml and electroporated, as described (23). Extrachromosomal DNA was recovered at the indicated times in 50 µl of H₂O, as described (23). Southern analysis was performed using 1 µl of the recovered DNA.

Escherichia coli DH10B was prepared for transformation by electroporation as recommended by the manufacturer (Life Technologies, Inc.). Bacteria (20 µl) were mixed with 0.5 µl of DNA and electroporated in a 0.1-cm cuvette using a Bio-Rad gene pulser with a setting of 1.7 kV. Prior to plating, they were incubated in 1 ml of SOC medium for 1 h at 37 °C.

Plasmid DNA was sequenced using the AmpliCycle^TM Sequencing Kit (Perkin-Elmer). The polymerase chain reaction sequencing primer (5-CGGCTATGACTGGGCACAACAG-3) lies 118 bp upstream of the PstI cleavage site in Mneo.

RESULTS

To delineate the nature of the double strand repair defect in the cell mutants, a three-plasmid assay based on bacterial transformation was utilized (Fig. 1). Plasmid pBRtet is introduced uncleaved into mammalian cells to serve as a transfection control. It confers tetracycline resistance (Tet^R) to bacteria when recovered from the transfected cells. Plasmids M5neo and M3neo are introduced as linear fragments and are used to measure extrachromosomal recombination and DNA end joining (23). Plasmid M5neo is cleaved with SphI (M5neo/S), and plasmid M3neo is cleaved with AatII and PstI (M3neo/AP). The plasmid backbones of M5neo/S and M3neo/AP are each capable of being recircularized in mammalian cells to measure DNA end joining (23). The SphI overhangs of M5neo/S can be precisely or imprecisely rejoined, whereas the two different 3 overhangs of M3neo/AP can only be imprecisely rejoined. When recovered from mammalian cells and reintroduced into bacteria, the recircularized plasmids confer ampicillin resistance (Amp^R), which serves as the measure of DNA end joining.

The M5neo and M3neo plasmids are also used to measure double strand break-promoted homologous recombination in mammalian cells (Fig. 1). M5neo contains the 5 end of the Tn5 neo gene, and M3neo contains the 3 end, with 352 bp of homology between them in the middle of the neo gene. The plasmid backbones are also homologous. When the two plasmids recombine within the neo homology region, a neo⁺ gene is recovered, and the recombined plasmid confers kanamycin resistance (Kan^R) to bacteria. Homologous recombination between two circular plasmids is inefficient in mammalian cells but is greatly stimulated when a double strand break is introduced next to the homology region (7). Thus, when M5neo is linearized with SphI and M3neo is cleaved with AatII and PstI to release the homologous neo fragment (Fig. 1), recombination is readily detected (23). This three plasmid assay is termed the TAK assay, since Tet^R measures transfection efficiency, Amp^R measures DNA end joining, and Kan^R measures homologous recombination.

To test the involvement of DNA-PK_cs in recombination and end joining, pBRtet, M5neo/S, and M3neo/AP were electroporated into SF-19 scid fibroblasts and a wild type control, CBP-9. After electroporation, cells were rinsed extensively and then incubated at 37 °C. Plasmid DNA was recovered from cells at various times after electroporation up to 24 h. The recovered DNA was electroporated into bacteria, and colonies were selected on tetracycline, ampicillin, and kanamycin plates (Fig. 2). At any one time point, the number of colonies obtained for both cell lines is similar on each of the antibiotic-containing plates, indicating that DNA uptake and stability, end joining, and homologous recombination are similar for both cell lines. Southern analysis confirms these results (data not shown). Thus, the double strand repair defect of scid cells is not manifested in the repair of these extrachromosomal substrates. Other extrachromosomal studies have yielded similar results, although these studies did not analyze these processes as directly and at such early times after transfection (24, 25, 26).

The results obtained with the pBRtet plasmid indicate that soon after entering cells, transfected DNA begins to be degraded in both cell lines (see Fig. 4). At later times, the rate of degradation slows, perhaps due to protection of undegraded DNA by chromatin proteins. Even with competing degradation, a clear signal for both end joining and recombination is obtained, as evidenced by the recovery of Amp^R and Kan^R colonies, respectively, beginning at the earliest 2-h time point. At this time point, the end joining products represent roughly 8% of the transfected DNA, and the recombination product represents about 1.5%. These results are similar to those obtained previously in monkey COS1 cells (23). Homologous recombination products are underrepresented relative to end joining products in this bacterial transformation assay, since only one repair event is needed to rejoin the plasmid backbones to confer Amp^R, whereas two repair events are required to reclose the plasmid to confer Kan^R. In addition to recombination within the neo gene, either a second recombination event or end joining must occur within the downstream plasmid sequences.

A similar experiment was performed in the Ku80-deficient xrs-6 cells and parental CHO-K1 cells. The three plasmids, pBRtet, M5neo/S, and M3neo/AP, were electroporated into the cells, and DNA was recovered at various time points after electroporation (Fig. 3). As seen in the mouse cells, the transfection control pBRtet is taken up by both the parental and mutant cells at a similar level but then begins to be nontransformable for bacteria soon after entering the cells, presumably due to degradation.

At 2 and 4 h after electroporation, the decrease in the number of transformants is similar for both cell lines (Fig. 3). However, a reproducible difference in the bacterial transformation frequency is noted at later times after transfection. By 8 h after transfection, the number of transformants obtained from DNA recovered from the xrs-6 cells has decreased about 50%. By 24 h after transfection, 6-fold fewer transformants are obtained. Although pBRtet is electroporated as supercoiled DNA, Southern analysis has indicated that most of DNA transfected into cells is nicked, relaxed, or linearized within the 1st h after transfection, even in wild type cells.² Thus, the decreased recovery of the input pBRtet DNA from the xrs-6 cells is likely due to greater degradation of DNA that has already suffered some damage.

Amp^R and Kan^R colonies were also selected after bacterial transformation to measure end joining and recombination, respectively. As with the Tet^R transfection control, xrs-6 and CHO-K1 cells yield similar numbers of Amp^R and Kan^R colonies for DNA recovered early after electroporation but exhibit differences at later time points (Fig. 3). The number of Amp^R colonies is similar at 2 and 4 h after electroporation, but by 8 h it decreases for the xrs-6 cells, such that by 24 h, 12-fold fewer transformants are obtained. The number of Kan^R colonies is nearly identical for both cell lines at 2 h after electroporation, a 20% decrease is obtained at 4 h for the xrs-6 cells, and by 24 h, a 24-fold decrease in the number of colonies is obtained. Thus, at early times after transfection, no obvious defect is manifested in the enzymatic machinery responsible for end joining and homologous recombination in the xrs-6 cells.

At late times after transfection, the decrease in the number of colonies is likely due to the overall higher amount of DNA degradation in the xrs-6 cells. The decrease in plasmid recovery is greater for the end joining (12-fold) and recombination (24-fold) products than for the transfection control plasmid (6-fold). Since the M5neo and M3neo plasmids are cleaved in vitro prior to transfection (whereas pBRtet is introduced as supercoiled DNA) the greater drop in plasmid recovery for the end joining products may be due to the double strand break acting as a lesion from which to initiate degradation. Since both plasmids are required for recombination, the drop in the recombinant product recovery is greatest.

To monitor directly the state of the transfected DNA, Southern analysis was performed. Recovered DNA was digested with HindIII and BamHI, which cleave at the 5 and 3 ends of the neo repeats, respectively. A neo fragment was used as probe to detect the input M3neo/AP and M5neo/S DNA, as well as neo end joining and recombination products.

Immediately after electroporation, bands from the input DNA are detected (Fig. 4). By 2 h, the 1.2-kb neo gene recombination product is readily detected (arrow). One of the end joining products is also detected, a 1.5-kb fragment derived from joining of the neo sequences in the M5neo/S plasmid with the neo sequences in the M3neo/AP fragment. (The major intramolecular end joining products are not detected in this Southern blot.) Both the 1.2- and 1.5-kb bands demonstrate repair of the double strand break in the neo gene. Based on plasmid recovery, these bands are primarily derived from linear plasmids in which the downstream double strand break is not repaired.

Degradation of the DNA is also apparent. Whereas the input DNA appears as two sharp bands immediately after transfection, by 2 and 4 h after electroporation, the bands appear as broad smears, beginning at their original sizes and trailing downward. At later times, much of the DNA has been completely degraded, in agreement with the observed decrease in bacterial transformation. The recombination and end joining products appear more stable over this time course than the input DNA. Since the input DNA has a double strand break adjacent to the neo sequences, an entry site for exonucleases may be provided, resulting in a faster rate of degradation. By contrast, when the double strand break within the neo gene has been healed by recombination or end joining, neo sequences can be protected from exonucleolytic digestion by the adjacent plasmid sequences. Alternatively, it is possible that DNA that has recombined or end joined is protected from nucleolytic digestion.

Overall, DNA recovered from xrs-6 cells gives a profile similar to that of DNA recovered from CHO-K1 cells. However, we consistently see enhanced smearing throughout the lanes for DNA recovered from xrs-6 cells. For example, at 8 h after transfection, fragments smaller than 0.6 kb are much more apparent in the xrs-6 cells. This may be due to enhanced degradation of a portion of the transfected DNA. The higher molecular weight smear of DNA may be single stranded since its mobility is retarded on the gel.

A difference is also noted in the recovery of the end joining and recombination products in the xrs-6 cells. At early time points after transfection, these products are similar in intensity between the two cells lines. However, by 24 h after transfection, these products have decreased in intensity for the xrs-6 cells. Thus, the smaller number of colonies recovered by ampicillin and kanamycin selection is reflected by direct analysis of DNA.

To analyze further DNA end joining in the Ku-deficient cells, only one of the two end joining substrates was transfected. M3neo/AP has two different 3 overhangs: an ACGT overhang derived from AatII cleavage and a TGCA overhang derived from PstI cleavage. These ends must be modified for the plasmid backbone to be recircularized. M3neo/AP and pBRtet were electroporated into CHO-K1 and xrs-6 cells, and DNA recovered from the cells was analyzed by bacterial transformation. Results are similar to those obtained in the three plasmid transfection. At 2 and 4 h after electroporation, the number of Amp^R colonies obtained from both cell lines is similar, whereas by 24 h, the number of colonies derived from the xrs-6 cells is reduced (Fig. 5). The amount of transformable pBRtet DNA also decreases in the xrs-6 cells by 24 h after electroporation. Thus, at early times after transfection, there is no apparent defect in imprecise end joining in the xrs-6 cells.

Since degradation of transfected DNA is increased in the Ku-deficient cells, defects in precise end joining would be expected if deletions occur more readily at the DNA ends. The Mneo plasmid, which contains an intact Tn5 neo gene, was utilized to test this (Fig. 6A). By cleaving within the neo coding sequence of Mneo, precise end joining could be determined by kanamycin selection of bacterial colonies as well as by reconstruction of a restriction site. Mneo was digested with PstI; the cleaved DNA, called Mneo/P, was purified by gel electrophoresis. It was electroporated with pBRtet into the CHO-K1 and xrs-6 cells. Extrachromosomal DNA was recovered at 0, 2, 4, 24, and 48 h after electroporation.

For both the tetracycline and ampicillin selections, the results are similar to those described in the previous experiments (Fig. 6B). The number of colonies obtained from both cell lines is similar early after transfection, whereas by 48 h, the number is much lower for the xrs-6 cells. A difference between the two experiments for the CHO-K1 cells is noted at 24 h, in that the number of Amp^R colonies obtained remains high for Mneo/P (Fig. 6B), whereas for M3neo/AP, the colony count has substantially decreased (Fig. 5). This difference may due to the greater distance of the double strand break in the Mneo/P plasmid from essential elements of the plasmid, e.g. the ampicillin resistance gene.

In assaying precise end joining, however, a different result is obtained for the two cell lines early after transfection. The number of Kan^R colonies is substantially lower for DNA recovered from the xrs-6 cells, even though the number of Amp^R colonies is the same. At both 2 and 4 h after transfection, the number of Kan^R colonies is 3-fold lower for the xrs-6 cells. Whereas 91% of plasmids recovered from the CHO-K1 cells are precisely rejoined 2 h after transfection, only 32% of the plasmids from the xrs-6 cells are precisely rejoined. At 4 h after transfection, the level of precise rejoining is reduced to 64% of the plasmids recovered from CHO-K1 cells, whereas the level of precise end joining is further reduced for plasmids recovered from the xrs-6 cells, to 21%. Thus, although the overall level of end joining is not reduced in the xrs-6 cells at the early time points, the percentage of products with precisely rejoined ends is lower relative to the CHO-K1 cells. At later time points, when the overall recovery of end joining products is reduced in xrs-6 cells, there is also a 3-5-fold lower recovery of products with precisely joined ends.

To examine the imprecise end joining products, Amp^R colonies were screened on kanamycin plates. As expected, a greater percentage of the colonies obtained from xrs-6 cells were Kan^S. Plasmid DNA was isolated from Kan^SAmp^R colonies derived from DNA recovered 48 h after transfection and analyzed by restriction digestion. Many of the plasmids contain deletions close to the maximum size possible to still result in bacterial transformation (approximately 2.0 kb). This was the case whether plasmids were recovered from CHO-K1 or xrs-6 cells (data not shown).

To examine products that might contain an intermediate amount of degradation, DNA recovered at an earlier time after transfection was analyzed. The 2-h time point was chosen, since the overall level of product recovery is the same for the two cell lines at this time. From CHO-K1 cells, 9 of 63 Amp^R colonies were determined to be Kan^S, whereas from xrs-6 cells, 12 of 20 Amp^R colonies were Kan^S. Plasmid DNA was isolated from the Kan^S colonies and analyzed by restriction digestion. One plasmid from CHO-K1 cells contains a deletion within the neo gene which is approximately 370 bp. Two plasmids from xrs-6 cells contain neo gene deletions of approximately 150 and 220 bp. Most of the plasmids, however, appear to contain small deletions (less than 50 bp) around the PstI site. These plasmids were subjected to polymerase chain reaction sequencing using a primer located 118 bp upstream of the PstI site (Fig. 6A).

Sequencing results demonstrate that most of these plasmids contain simple deletions within or flanking the PstI site (Fig. 7). For CHO-K1 cells, the deletions range from 2 to 14 bp. One of these plasmids underwent a more complex repair event since it contains an additional bp substitution (A to C) near the PstI cleavage site. For xrs-6 cells, the deletions cover a larger size range, from 3 to 62 bp. One of these contains an addition of 1 bp, a C, at the site of the deletion. The average size of the deletions in the plasmids from CHO-K1 cells is 6.6 bp, whereas for xrs-6 cells it is 23.2 bp. Thus, although the size of the deletions overlaps for the two cell lines, the average size of the deletions is 3-4-fold larger for the xrs-6 cells.

DISCUSSION

The stability, DNA end joining, and homologous recombination of extrachromosomal DNA have been assessed in Ku80- and DNA-PK_cs- deficient cell lines. We find that in the Ku-deficient xrs-6 cells, the stability of transfected DNA is reduced and that DNA degradation is enhanced. Prior to gross affects on DNA stability, the levels of double strand break-promoted homologous recombination as well as DNA end joining are unaffected in the xrs-6 cells. However, the recovery of products with precisely joined ends is reduced with a concomitant increase in the recovery of products containing deletions. Unlike the Ku-deficient cells, no reduction in the stability of transfected DNA was detected in DNA-PK_cs-deficient scid cells. Wild type levels of end joining and homologous recombination were also detected in the scid cells.

During the first few hours after transfection, the recovery of intact pBRtet DNA from transfected CHO-K1 and xrs-6 cells decreases at a similar rate but by 8 h after transfection begins to drop more rapidly for the xrs-6 cells. Although transfected as supercoiled DNA, the pBRtet plasmid is likely nicked or linearized soon after transfection. Since the Ku protein is able to bind nicks and single strand gaps as well as double strand breaks, it may protect multiple forms of DNA from degradative nucleases. Once the nucleases responsible for degradation of transfected DNA are identified, this question can be addressed directly in vitro. Although the xrs-6 cells have a primary defect in the gene encoding the Ku80 component of the DNA-PK complex (12, 27), they are also deficient in DNA-PK_cs kinase activity, since DNA-PK_cs is dependent upon Ku for activation (28). The decrease in DNA stability in the xrs-6 cells is likely due to the primary deficiency of the Ku heterodimer, rather than the indirect effect on DNA-PK_cs activity, since scid cells do not exhibit a decrease in DNA stability (see below).

Double strand break-promoted homologous recombination occurs at a normal level in the xrs-6 cell line before increased degradation of DNA is detectable. The reduction in recombinant product recovery at later time points is likely due to increased degradation. Homologous recombination of transfected substrates has been studied previously in xrs cells, with conflicting outcomes (29, 30). One report found little or no difference in the level of recombination (30), whereas the other found a 6-fold decrease in xrs cells (29). These studies relied on stable transformation of the cell lines by selectable markers to measure recombination, the gpt gene in one case (30) and the neo gene in the other (29). Assays involving stable transformation of DNA are much more indirect than the transient transfection assay utilized in the current study. The apparently lower amount of recombination in the one study may be related to enhanced degradation of DNA in the Ku-deficient cells.

Alternatively, it is possible that the different results are due to the particular xrs cell lines utilized in each experiment. The current study utilizes the xrs-6 cell line, whereas other studies analyzed recombination in the xrs-1 and xrs-7 (30) and xrs-5 (29) cell lines. Although these cell lines all belong to the same complementation group, they do not have identical responses to genotoxic agents. We have focused our analysis on the xrs-6 cell line because complementation of the repair defect by a Ku80 cDNA has been demonstrated directly in this cell line (12).

As with recombination, the level of DNA end joining was found to be unaffected in the Ku-deficient cells soon after transfection. However, the products recovered from the end joining reaction showed a striking difference in that there was a 3-fold reduction in plasmids having undergone precise end joining in the xrs-6 cells. This reduction was observed prior to any decrease in overall plasmid recovery. The plasmids recovered from xrs-6 cells contain more deletions at the site of cleavage, and the size of the deletions is on average larger, consistent with increased degradation from the ends. Since these deletions are relatively small in size, it is not surprising that recombination within the 352-bp homology region was not affected soon after transfection. Attempts to monitor precise end joining in Ku-deficient xrs cells have been reported previously using a stable transformation assay (30, 31). In both reports, it was concluded that there was no decrease. However, as mentioned above, stable transformation assays are much more indirect than the assays utilized in the present study, requiring integration of transfected DNA. It is not known when DNA stably integrates into genomic DNA after transfection, or by what mechanism. In addition, in both of the previous studies, correction factors were introduced to calculate precise end joining frequencies, due to decreased overall levels of stable transformation in the xrs cells.

The decreased recovery of precisely joined products we observe from the xrs-6 cells, as well as the overall increased DNA degradation, suggests that Ku plays a role in DNA end protection, preventing exonucleases from degrading DNA. Consistent with this model is the structure of rare signal joints recovered from xrs-6 cells in V(D)J recombination assays (20). Although signal joints are normally derived from the precise joining of blunt-ended intermediates, only 14% of signal joints from xrs-6 cells are precise, the remainder of which contain deletions (20). Ku has also been postulated to serve as an alignment factor in DNA repair, holding two ends together for rejoining. Our data do not support such a model, although we cannot rule out additional roles for Ku.

Since we do not observe increased DNA degradation in scid cells, DNA-PK_cs may not participate in such an end protection function. Although both xrs-6 and scid cells are defective in V(D)J recombination, the defect in xrs-6 cells is more severe, both quantitatively and qualitatively, than that in scid cells (20). Unlike xrs-6 cells, scid cells are able to form signal joints at normal levels, and 80% of these joints are precise ligations of the signal sequences. These results are consistent with the two subunits of the DNA-PK holoenzyme playing different roles in DNA repair, e.g. a structural role for Ku and a regulatory role for DNA-PK_cs. However, it has been observed that rare coding joints recovered from plasmid assays in scid cells have larger deletions than those from wild type cells and that signal joints are slightly more likely to contain deletions (20). In addition, scid mice can undergo a low frequency of normal V(D)J recombination (10) and can be induced to undergo V(D)J recombination upon irradiation (32). These results would be consistent with a leaky scid mutation that is induced upon DNA damage, rather than a distinct role from Ku in DNA repair. Alternatively, there may be functional redundancy such that another protein can partially substitute for DNA-PK_cs and that it is induced with DNA damage. The transfection of DNA ends could also potentially lead to such an induction, resulting in normal levels of repair of transfected DNA.

A role for Ku in protecting DNA ends from degradation leads to two related but distinct consequences in xrs-6 cells: loss of transfected plasmids and decreased precise end joining. In terms of chromosomal break repair, these results suggest that in addition to slower repair of breaks, larger deletions, as well as a greater number of deletions, at the site of breaks may lead to the increased sensitivity of xrs-6 cells to ionizing radiation. The rejoining of broken chromosome ends is critical for the proper replication and segregation of chromosomes. However, the faithful repair of a break, restoring the normal sequence, is necessary if the break occurs within critical coding or regulatory sequences. Studies on chromosomal break repair, for example using rare-cutting endonucleases (33), will allow a more definitive analysis of the consequences of Ku and DNA-PK_cs deficiency on genomic instability.