Ku70, an Essential Gene, Modulates the Frequency of Raav-Mediated Gene Targeting in Human Somatic Cells

Ku70, an essential gene, modulates the frequency of rAAV-mediated gene targeting in human somatic cells

Farjana J. Fattah, Natalie F. Lichter, Kazi R. Fattah, Sehyun Oh, and Eric A. Hendrickson*

Department of Biochemistry, Molecular Biology, and Biophysics, University of Minnesota Medical School, Minneapolis, MN 55455

Edited by Arthur Landy, Brown University, Providence, RI, and approved April 11, 2008 (received for review December 20, 2007) Gene targeting has two important applications. One is the inacti- because of the preferred usage of a competing pathway of DNA vation of genes (‘‘knockouts’’), and the second is the correction of DSB repair, nonhomologous end joining (NHEJ). a mutated allele back to wild-type (‘‘gene therapy’’). Central to NHEJ is an evolutionarily conserved process that joins nonho- these processes is the efficient introduction of the targeting DNA mologous DNA molecules together (5). In their seminal work on into the cells of interest. In humans, this targeting is often accom- gene targeting, Thomas and Capecchi (6) showed that although plished through the use of recombinant adeno-associated virus somatic mammalian cells can integrate a linear duplex DNA into (rAAV). rAAV is presumed to use a pathway of DNA double-strand corresponding homologous chromosomal sequences using HR, the break (DSB) repair termed homologous recombination (HR) to frequency with which recombination into nonhomologous se- mediate correct targeting; however, the specifics of this mecha- quences occurred via NHEJ was at least 1,000-fold greater. Al- nism remain unknown. In this work, we attempted to generate though not all of the details of NHEJ have been elucidated, much Ku70-null human somatic cells by using a rAAV-based gene knock- is known about the process. First, the heterodimeric Ku out strategy. Ku70 is the heterodimeric partner of Ku86, and (Ku86:Ku70) protein binds onto the ends of the donor DNA and together they constitute an end-binding activity that is required prevents the nucleolytic degradation that would otherwise shunt the for a pathway [nonhomologous end joining (NHEJ)] of DSB repair DNA into the HR pathway (see above and ref. 5). The binding of that is believed to compete with HR. Our data demonstrated that Ku to the ends of the DNA then recruits and activates the Ku70 is an essential gene in human somatic cells. More importantly, DNA-dependent protein kinase complex catalytic subunit (DNA- ؉/؊ however, in Ku70 cells, the frequency of gene targeting was 5- PKcs). This DNA-protein complex is then brought into contact with to 10-fold higher than in wild-type cells. RNA interference and a chromosome into which a DSB is introduced by a mechanism that short-hairpinned RNA strategies to deplete Ku70 phenocopied is poorly understood, although it correlates frequently with chro- these results in wild-type cells and greatly accentuated them in mosomal palindromic sequences (7). Regardless, the chromosomal ؉/؊ Ku70 cell lines. Thus, Ku70 protein levels significantly influ- ends are probably also occupied by Ku and DNA-PKcs, which enced the frequency of rAAV-mediated gene targeting in human facilitates the formation of a synaptic complex with the donor DNA somatic cells. Our data suggest that gene-targeting frequencies can (8). Once DNA-PKcs is properly assembled at the broken ends, it, be significantly improved in human cells by impairing the NHEJ in turn, recruits additional factors such as the nuclease, Artemis, to pathway, and we propose that Ku70 depletion can be used to trim the ends and a DNA ligase complex consisting of DNA ligase

facilitate both knockout and gene therapy approaches. IV (LIGIV)-x-ray cross-complementing group 4-XRCC-4-like fac- MEDICAL SCIENCES tor, to seal the break(s) (5). In summary, humans are different from gene therapy ͉ nonhomologous end joining (NHEJ) ͉ bacteria and lower eukaryotes in that DSB repair proceeds pri- recombinant adeno-associated virus (rAAV) ͉ Ku ͉ DNA protein kinase marily through a NHEJ recombinational pathway. Moreover, NHEJ must be overcome to facilitate gene targeting, which can only omatic gene targeting is defined as the intentional modification occur when the incoming DNA is shunted into the HR pathway. Sof a genetic locus in a living cell (1). This technology has two Adeno-associated virus (AAV) is a nonpathogenic parvovirus general applications of interest and importance. One is the inacti- with a natural tropism for human cells that depends on a helper vation of genes (‘‘knockouts’’), a process in which the two wild-type virus (usually adenovirus and hence the name) for a productive alleles of a gene are sequentially inactivated to determine the infection (9). In the intervening decade since it was demonstrated loss-of-function phenotype(s) of that particular gene. The second that recombinant AAV (rAAV) could be used as a vector for gene application is the clinically more relevant process of gene therapy, targeting in human cells (10), this methodology has gained wide which, in a strict sense, involves correcting a preexisting mutated acceptance (1). Nineteen different genes have been modified allele of a gene back to wild-type (‘‘knockins’’) to alleviate some (generally knocked out) in a plethora of immortalized and normal pathological phenotype associated with the mutation. Importantly, diploid human tissue culture lines (ref. 1 and F.J.F. and S.O., although these two processes are conceptually reciprocal opposites unpublished data). Moreover, human adult stem cell lines derived of one another, they are mechanistically identical because both from osteogenesis imperfecta patients afflicted with a dominant- require a form of DNA double-strand break (DSB) repair termed negative mutation in the COL1A1 gene have been corrected (a homologous recombination (HR). knockin) with rAAV vectors (11). Last, Ͼ20 clinical gene therapy In HR (2), the ends of the donor dsDNA are resected to yield trials using rAAV are currently in progress. There are, however, 3Ј-ssDNA overhangs, which are targets bound by RAD51 and RAD52. RAD51 is a potent strand exchange protein, and together Author contributions: F.J.F. and E.A.H. designed research; F.J.F., N.F.L., K.R.F., S.O., and with RAD54 (3), an ATPase that remodels chromatin, it facilitates E.A.H. performed research; S.O. contributed new reagents/analytic tools; F.J.F. and E.A.H. cross-over of the incoming donor DNA with its cognate chromo- analyzed data; and F.J.F. and E.A.H. wrote the paper. somal homologous sequences. This gene-targeting event generates The authors declare no conﬂict of interest. a complex structure that is identical to the linearized plasmid This article is a PNAS Direct Submission. ‘‘ends-out’’ recombination intermediates that have been defined in *To whom correspondence should be addressed at: 6-155 Jackson Hall, Department of yeast (4). Resolution of this structure requires the action of heli- Biochemistry, Molecular Biology, and Biophysics, 321 Church Street SE, University of cases, polymerases, resolvases, and ligases in a complicated process Minnesota Medical School, Minneapolis, MN 55455. E-mail: [email protected]. that has not yet been rigorously defined (2). Importantly, human This article contains supporting information online at www.pnas.org/cgi/content/full/ cells express all of the HR gene products needed to carry out gene 0712060105/DCSupplemental. targeting (1). These events occur, however, only at a low frequency © 2008 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0712060105 PNAS ͉ June 24, 2008 ͉ vol. 105 ͉ no. 25 ͉ 8703–8708 Downloaded by guest on September 30, 2021 Fig. 2. Disruption of the Ku70 locus in HCT116 cells by gene targeting. (A) PCR characterization of the Ku70ϩ/Ϫ clones 53 and 53-39. Ethidium bromide (EtBr)- Fig. 1. Experimental strategy and screening protocols for disruption of the stained agarose gels are shown. (Left) From left to right the lanes are: molecular Ku70 locus in human somatic cells. (A) Diagrams of the relevant portion (large weight markers, genomic DNA isolated from wild-type (WT) cells, and genomic rectangles: exons 3 and 4) of the human Ku70 genomic locus (horizontal line: DNA derived from clone 53 cells. The PCR used the primers RArmF and Ku70 3,4 genomic DNA) in the parental wild-type cells. The black arrows demarcate the R1 (Fig. 1), and the diagnostic Ϸ1.3-kb band was observed only in clone 53. (Right) ϩ/Ϫ indicated PCR primers. (B) Ku70 locus in Ku70 (clone 53) cells. One of the From left to right the lanes are: molecular weight markers, genomic DNA isolated endogenous alleles has been replaced by a homologous genomic sequence from clone 53 cells, and genomic DNA derived from clone 53-39 cells. The PCR (horizontal gray bars), LoxP sites (ﬁlled triangles), and the neomycin (Neo) resis- used the primer set 70 Cre F and 70 Cre R, and the diagnostic Ϸ350-bp PCR product tance gene (rectangle). Additional PCR primers (black arrows) are indicated. (C) was observed only in clone 53-39. (B) Identiﬁcation of correctly targeted clones ϩ/Ϫ Ku70 locus in Ku70 (clone 53-39) cells after Cre treatment. All symbols are as in after the second round of gene targeting. EtBr-stained agarose gels are shown. B. Additional PCR primers (black arrows) are also indicated. (D) Ku70 locus in From left to right the lanes are: molecular weight markers and then genomic DNA ϩ/Ϫ Ku70 (clone 53-39) cells after retargeting to the already inactive allele. All isolated from the indicated cell lines. (Upper) The PCR used the primer set Neo F2 symbols are as in B. Additional PCR primers (black arrows) are also indicated. (E) and Ku70 3,4 R1 (Fig. 1), and the diagnostic Ϸ1.3-kb band was observed only in ϩ/Ϫ Ku70 locus in Ku70 (clone 53-39) cells after correct second-round targeting to those clones that were correctly targeted and contained the Neo gene. (Lower) the functional allele to generate the null cell line. All symbols are as in B. The PCR used the primer set Ku70 Ex4 F1 and Ku70 3,4 R1 (Fig. 1), and the diagnostic Ϸ1.3-kb band, which indicated the retention of at least one copy of exon 4, was observed in all of the clones. (C) All correctly targeted clones still reports of potential difficulties with the clinical use of rAAV express Ku70 protein, albeit at reduced levels. Whole-cell extract of the indicated vectors. In particular, a recent report found that in a mouse model cell lines was subjected to immunoblot analyses to detect Ku70 (Upper), and for lysosomal storage disease mucopolysaccharidosis VII, the ran- tubulin (Lower) protein levels. dom integration of rAAV was associated with a high incidence of hepatocellular carcinoma (ref. 12; for review, see ref. 13). Thus, a better understanding of the mechanism of rAAV-mediated gene cell line that is diploid, has a stable karyotype, and is wild type for targeting and the factors that influence the frequency with which it almost all of the DNA DSB repair, checkpoint, and chromosome correctly targets (presumably HR-mediated) vs. those that influ- stability genes that have been examined (ref. 1 and Fig. 1A). ϩ Ϫ ence its random integration (presumably NHEJ-mediated) clearly Ku70 / HCT116 cells showed haploinsufficient deficits as they seems warranted. grew slower, were more sensitive to ionizing radiation, and had Using classic gene-targeting methodologies, our laboratory has shortened telomeres compared with the parental cell line (15). shown that the NHEJ gene Ku86 is essential in human somatic cells These phenotypes were not unexpected because heterozygous Ku86 ϩ Ϫ (14). To extend these studies to Ku70, we have used a rAAV vector (Ku86 / ) HCT116 cells had similar haploinsufficiencies (14, 16). approach. As expected, these experiments demonstrated that Ku70 Because we had also subsequently shown that human Ku86-null Ϫ Ϫ is also essential. Surprisingly, however, the frequency of correct (Ku86 / ) cell lines were not viable, we wanted to extend this gene targeting increased 5- to 10-fold in Ku70-heterozygous observation to human Ku70-deficient cells. To achieve this goal ϩ Ϫ (Ku70ϩ/Ϫ) cells. RNA interference and short-hairpinned RNA experimentally, one of the Ku70 / clones (clone 53; Figs. 1B and strategies to deplete Ku70 phenocopied these results in wild-type 2A) was transiently exposed to Cre recombinase. Cre should excise cells and greatly accentuated them in Ku70ϩ/Ϫ cell lines. To support the internal neomycin selection cassette, which is flanked by loxP the generality of these findings, we extended them to an additional sites, within the integrated targeting vector (Fig. 1B). Forty-eight two loci, the chemokine (C-C motif) receptor 5 (CCR5) gene and single cell clones were picked, duplicated into medium containing LIGIV and observed similar effects. Thus, Ku70 protein levels or lacking G418, and G418-sensitive clones were identified. Correct influenced the frequency of correct rAAV-mediated gene targeting excision of the neomycin gene was confirmed by the generation of in human somatic cells. Our data demonstrate that gene-targeting a diagnostic Ϸ350-bp PCR product with the primer set 70 Cre F and frequencies can be significantly improved by impairing the NHEJ 70 Cre R (Figs. 1C and 2A, 53-39 lane). Seven such clones were pathway, and we propose that Ku70 depletion can be used to obtained in this fashion, and one of them, clone 53-39, was used for facilitate both knockout and gene therapy approaches. a second round of gene targeting with the original targeting vector containing the neomycin selection cassette. Productive infection of Results clone 53-39 cells should produce three potential outcomes: (i) Use of Gene Targeting to Generate Ku70-Null HCT116 Cells. Using random targeting (the majority of the events); (ii) correct targeting, rAAV-based methodology, we have generated a Ku70ϩ/Ϫ HCT116 but of the already inactivated allele from the first round of targeting cell line (ref. 15 and Fig. 1B). HCT116 is a human colon carcinoma (i.e., ‘‘retargeting’’; Fig. 1D); or (iii) correct targeting of the

8704 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0712060105 Fattah et al. Downloaded by guest on September 30, 2021 Table 1. Summary of gene-targeting frequencies at the Ku70 locus No. of colonies No. of correctly Targeting Fold Cell line screened* targeted clones† frequency‡ increase§

WT HCT116 437 3 0.69 1.0 WT HCT116 ϩ Ku70 siRNA 173 10 5.78 8.4 WT HCT116 ϩ Ku70 shRNA 65 2 3.08 4.4 53-39 (Ku70ϩ/Ϫ) 97 4 4.12 6.0 53-39 (Ku70ϩ/Ϫ) ϩ Ku70 siRNA 111 23 20.72 30.0

*Drug-resistant clones that were also positive for the internal control PCR. †Drug-resistant clones that showed correct targeting by PCR. ‡Targeting frequency is the number of correctly targeted colonies per 100 drug-resistant colonies screened. §The WT targeting frequency is set at 1. The fold increase is the targeting frequency in a speciﬁc cell line divided by the targeting frequency in the WT background (0.69).

remaining functional allele to generate the desired null clone (Fig. parental cells. Moreover, the Ku86 protein level in 53-39 Ku70siRNA 1E). Several diagnostic PCR strategies were used to distinguish cells was also decreased, confirming previous observations suggest- these events (Fig. 1). In two separate screens (elaborated in detail ing that the stabilities of Ku70 and Ku86 are coordinately linked (5, below) a total of 27 correctly targeted clones were obtained. 14). Impressively, 23 correctly targeted (all retargeted) clones were Strikingly, all 27 clones were retargeted. This conclusion was recovered from 111 G418-resistant clones screened, for a targeting substantiated by the fact that although all of the clones were frequency of 20.72% (Table 1). This represented a 30-fold increase correctly targeted (Figs. 1D and 2B Upper, and data not shown), in gene targeting compared with the parental line. It is important they still retained exon 4 sequences (Fig. 2B Lower and data not to note that this was a minimum frequency because a statistically shown). Moreover, all 27 of the clones still expressed Ku70 protein equal number (i.e., 23) of null clones were presumably not recov- at levels that were Ϸ50% of that observed in the parental cell line ered because Ku70 is essential. In toto, these results demonstrated (Fig. 2C and data not shown). The large disequilibrium in gene that lowering the expression level of Ku70 significantly increased targeting in which 27 of 27 clones were retargeted and no null clones the gene-targeting frequency at the Ku70 locus in HCT116 cells. were obtained strongly suggests that Ku70, like Ku86, is an essential gene in human somatic cells. Diminished Ku70 Protein Levels Increase the Gene-Targeting Fre- quency at Other Loci in HCT116 Cells. To investigate the generality of Diminished Ku70 Protein Levels Increase the Gene-Targeting Fre- the enhancement of gene targeting by Ku70 depletion, the gene- quency of the Ku70 Locus in HCT116 Cells. The above results, although targeting frequency at another locus, chemokine (C-C motif) important, were rather expected. One surprising finding, however, receptor 5 (CCR5), was determined. CCR5 is a cell surface was observed. Our initial frequency for obtaining Ku70ϩ/Ϫ cell lines coreceptor for the human immune deficiency virus and has no was 0.69%, 3 correctly targeted clones identified from 437 G418- known role in DNA repair (18). Moreover, a rAAV targeting vector resistant, internal control PCR-positive clones screened (ref. 15 and for CCR5 and methodologies for identifying correct targeting MEDICAL SCIENCES Table 1). In the first attempt to obtain a null clone, 4 correctly events had already been well established by the laboratory of Bert targeted clones (all retargeted) were identified from a total of 97 Vogelstein (19). Using this vector and these methodologies, we G418-resistant colonies for a targeting frequency of 4.12% (Table determined that the targeting frequency for the parental cell line 1). This represented a 6-fold (4.12/0.69 ϭ 5.97) increase in the was 1.06% (Table 2). In four independent Ku70-reduced back- gene-targeting frequency in Ku70ϩ/Ϫ cells. To investigate whether grounds, we observed a significant increase in the correct gene- the reduced Ku70 expression level was facilitating a higher fre- targeting frequency (Table 2), which included Ku70shRNA cells quency of correct targeting, we attempted to phenocopy this effect. (5.2%; a 4.5-fold increase), Ku70siRNA cells (5.3%; a 4.6-fold Thus, the parental HCT116 cell line (Ku70ϩ/ϩ) was used to stably increase), 53-39 (Ku70ϩ/Ϫ) cells (6.6%; a 5.8-fold increase) and express a short hairpin RNA (shRNA), obtained from the 53-39siRNA cells (9.9%; a 8.6-fold increase). Thus, without exception MISSION TRC-Hs 1.0 (human) lentiviral shRNA library (17), and regardless of which technique was used, a reduction of Ku70 targeting Ku70. Five different shRNA sequences (608, 609, 610, expression increased the frequency of gene targeting 4- to 9-fold at 611, and 612) were tested for their ability to knock down the level the CCR5 locus. of Ku70 in HCT116 cells. Among these five sequences, 609 and 610 Finally, the effect of Ku70 depletion on gene-targeting frequency reduced Ku70 expression the best, to Ϸ10% of that observed in the at the LIGIV locus was analyzed. Recently, using a rAAV- parental line (Fig. 3A). The cell line expressing the shRNA se- mediated knockout approach (S.O. and E.A.H., unpublished re- quence 610 (Ku70shRNA) was used for Ku70 gene targeting, and 2 sults), we generated LigIVϩ/Ϫ cell lines in a HCT116 background by of 65 total clones were correctly targeted (Table 1). This repre- deleting part of exon 3 of this locus. Two correctly targeted LIGIV sented a 4.6-fold increase over the parental HCT116 line. Next, we cell lines were identified from 176 drug-resistant clones screens, for used the siGENOME SMARTPool on the parental HCT116 cells a targeting frequency of 1.13% (Table 3). When these studies were to silence Ku70 expression to Ϸ20% of that observed in the control repeated in the Ku70ϩ/Ϫ clone 53-39, 6 correctly targeted LIGIV transfected population (Fig. 3B). These cells (Ku70siRNA) had an cell lines from 148 drug-resistant clones were identified. This 8.4-fold increase in gene targeting (10 of 173; 5.78%; Table 1). To corresponds to a 4.05% targeting frequency and a 3.6-fold increase extend this line of experimentation to its logical conclusion, siRNA in gene targeting compared with the parental line. was used in an attempt to further knock down Ku70 protein In summary, multiple targeting experiments at three different expression in 53-39 Ku70ϩ/Ϫ cells that already contained reduced loci (Ku70, CCR5, and LIGIV) demonstrated that a reduction in levels of Ku70 because of genetic ablation. Thus, 53-39 Ku70ϩ/Ϫ Ku70 expression facilitates correct gene targeting in human somatic cells were transfected twice at 24-h intervals with the siGENOME cells. SMARTPool for Ku70. A significant reduction in Ku70 protein levels as assessed by Western blot analysis 48 h after the first siRNA Diminished Ku70 Protein Levels Do Not Decrease the Frequency of treatment was observed (Fig. 3B, 53-39 ϩ siRNA). 53-39 Ku70siRNA rAAV Random Integration. An explanation of the above results is that cells expressed Ϸ5% of the Ku70 protein compared with the in the presence of reduced levels of Ku70, the number of correct

Fattah et al. PNAS ͉ June 24, 2008 ͉ vol. 105 ͉ no. 25 ͉ 8705 Downloaded by guest on September 30, 2021 this latter possibility experimentally, the frequency of random integration events was measured. Equal numbers of the parental cells (WT HCT116), parental cells expressing shRNA (WT ϩ shRNA), Ku70ϩ/Ϫ cells (53-39), and Ku70ϩ/Ϫ cells expressing shRNA (53-39 ϩ shRNA) were independently infected with three different concentrations of two different rAAV viral stocks. Two to three weeks later, the total number of drug-resistant colonies was scored. There was no statistically significant difference between any of the cell lines for either virus at any concentration [supporting information (SI) Fig. S1]. Thus, we conclude that a reduction in Ku70 in human somatic cells does not result in a reduction of random integration of rAAV gene-targeting vectors. Discussion Ku70 Is Essential in Human Somatic Cells. The Ku heterodimer is a well conserved protein(s), with homologs known to exist in every species from bacteria to humans (5). In all organisms examined, mutations in either Ku subunit result in the expected deficits in DNA DSB repair, DNA recombination, and sensitivities to DNA- damaging agents. Importantly, in all organisms, with one glaring exception, Ku is nonetheless dispensable for viability. Intriguingly, humans appear to be unique in that Ku has evolved into an essential gene. This hypothesis is supported by the lack of documentation for even a single patient with a mutation in either Ku subunit. More- over, the targeted disruption of both alleles of the Ku86 gene in human HCT116 somatic cells was lethal (14). The reason that Ku should be uniquely essential in humans is not clear, although a role in telomere maintenance seems likely (16). Here, we corroborate the hypothesis that Ku is essential in humans by demonstrating that Ku70-null cells are not viable. This conclusion was based on 27 recovered second-round targeting events, which occurred solely on the already inactive allele (Fig. 2 and Table 1). There is overwhelm- ing evidence to support the view that targeting vectors target the paternal and maternal alleles without bias. In an experiment designed to disrupt ␤-catenin, 11 second-round targeting events occurred on the functional allele and 15 on the inactive allele (20). Similarly, in two independent studies involving Ki-ras,4of7(21) and 9 of 15 (22) targeting events occurred on the functional allele. Moreover, in rAAV-mediated targeting studies, 10 of 19 targeting events for COL1A1 occurred on the wild-type allele and 9 of 19 on the mutant allele (11). Because of this lack of bias in the targeting methodology, the striking disequilibrium of 27 of 27 retargeting events is strong evidence for the essential nature of Ku70 in HCT116 cells. A direct demonstration of the essential nature of Ku70 awaits the construction of a cell line expressing a conditionally null Ku70 allele.

Reduced Ku70 Expression Augments the Frequency of Correct Gene Targeting in Human Somatic Cells. Attempts to target the second allele of the Ku70 locus demonstrated that the gene-targeting Fig. 3. Disruption of Ku70 gene expression by RNA interference technolo- frequency was higher in Ku70ϩ/Ϫ cells compared with the parental gies. (A) Stable cell lines generated using shRNA expression show reduced cell line. The interpretation of this result, that a reduction in Ku expression of Ku. Wild-type HCT116 cells were infected with lentiviruses carrying short-hairpin sequences (609, 610, or both 609 and 610). Whole-cell protein levels increases the frequency of gene targeting, was con- extracts from stable puromycin-resistant clones were subjected to immuno- firmed by two independent methodologies: that of transient RNA blot analyses using extracts from wild-type (WT) and clone 53-39 as controls. interference and the stable use of short-hairpinned RNAs (Table 1) The extracts were probed sequentially for Ku70 expression and then Ku86 and and was confirmed at additional loci (Tables 2 and 3). It should be tubulin. For the gene-targeting experiments described in the text, the 610 noted, however, that there was not always a linear relationship shRNA-treated stable line was used. A PhosphorImager quantitation of this between the levels of Ku70 protein in a cell line and the frequency blot is shown below the figure. (B) Transient depletion of Ku70 protein of gene targeting. Thus, the Ku70shRNA cells expressed less Ku70 ϩ Ϫ expression. Wild-type and Ku70 / 53-39 cells were transfected with a Ku70 protein than the Ku70ϩ/Ϫ cells but had a lower, and not the siGENOME SMARTpool. All cells were transfected twice at 24-h intervals and predicted higher, frequency of gene targeting (Tables 1 and 2). This then harvested 72 h after the first siRNA treatment. A knockdown of gene shRNA expression was measured by immunoblotting with the indicated antibodies. was probably caused by the much slower growth of the Ku70 A PhosphorImager quantitation of this blot is shown below the figure. cells (F.J.F., unpublished observations), which may have a delete- rious effect on gene targeting because rAAV preferentially trans- duces actively dividing cells (23). Similarly, the Ku70siRNA cells ϩ Ϫ gene-targeting events increases. An alternative possibility is that the expressed less Ku70 protein than the Ku70 / cells, but the two cell absolute number of correct targeting events remains constant but lines often had about the same frequency of gene targeting (Tables that the number of random integration events is decreased. To test 1 and 2). This was likely because the Ku70siRNA cells represented a

8706 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0712060105 Fattah et al. Downloaded by guest on September 30, 2021 Table 2. Summary of gene-targeting frequencies at the CCR5 locus No. of colonies No. of correctly Targeting Fold Cell line screened* targeted clones† frequency‡ increase§

WT HCT116 262 3 1.15 1.0 WT HCT116 ϩ Ku70 siRNA 150 8 5.33 4.6 WT HCT116 ϩ Ku70 shRNA 96 5 5.20 4.5 53-39 (Ku70ϩ/Ϫ) 105 7 6.66 5.8 53-39 (Ku70ϩ/Ϫ) ϩ Ku70 siRNA 152 15 9.86 8.6

heterogeneous population of cells with some cells being effectively frequency than human somatic cells. Another possibility is that transfected and other cells not, which presumably translated itself there are locus-specific effects as, to date, only a single X-linked into heterogeneous targeting frequencies. These technical qualifi- locus (the hypoxanthine phosphoribosyltransferase locus) has been ers notwithstanding, a reduction in Ku70 protein expression was examined in the mouse (32, 33). In summary, additional experi- without exception always associated with an increase in the fre- ments are needed before it can be determined whether or not the quency of correct gene targeting. observations reported here for human somatic cells are evolution- The demonstration that Ku can regulate gene targeting is very arily conserved. well documented in fungal systems. Thus, deletion of Neurospora There are multiple indirect mechanisms whereby Ku could crassa Ku70 and Ku86 genes resulted in higher gene-targeting regulate gene targeting such as by modulating the conversion of frequencies (100% in the mutants compared with 20% in wild-type viral ssDNA into dsDNA. This process would be required if the cells) (24). This observation was then used in a robotics-driven, gene replacement aspect of rAAV gene targeting requires two whole-genome approach to expedite the functional inactivation of independent cross-over events (and thus two independent 3Ј ends 103 transcription factor genes (25) demonstrating the potent utility on the donor DNA). Hendrie and Russell, however, have argued of Ku-reduced strains. Moreover, there has been a spate of reports (albeit circumstantially) that rAAV-mediated gene targeting is of highly efficient correct gene targeting in Ku-deletion strains of more likely mediated by viral ssDNA rather than dsDNA (35). If five different species of Aspergillus [e.g., Aspergillus niger (26)], and this hypothesis is true, then we would instead favor a direct Sordaria macrospora (27), and Cryptococcus neoformans (28), dem- ‘‘competition’’ model that has been suggested by many laboratories onstrating that Ku-deficient strains are useful for gene targeting in (e.g., ref. 36) whereby Ku and Rad52 compete for the viral DNA filamentous fungi. Furthermore, the disruption of yKU70 (29) or ends and shunt the virus into either NHEJ or HR pathways, KlKU80 (30) also increased the frequency of gene targeting in the respectively. This model is consistent with the demonstration that budding yeasts Saccharomyces cerevisiae and Kluyveromyces lactis, both Ku and Rad52 physically bind to rAAV inverted terminal MEDICAL SCIENCES respectively. Thus, in fungi, where Ku is not essential, there is an repeats (ITRs) during a viral infection (37). In this model, a exceptionally good correlation between reduced Ku expression and reduction in Ku posits that Rad52 will statistically stand a better increased gene targeting. chance of binding the hairpin-shaped viral ITRs and funnel the An extrapolation of these observations to higher eukaryotes, DNA into the HR pathway. In a wild-type cell, however, where Ku however, has been lacking. Thus, deletion of Ku70 in chicken DT40 is more abundant than Rad52, the presence of normal amounts of cells (31) and Ku70 (32) or Ku86 (33) in mouse cells did not result Ku would favor random integration mediated by NHEJ. Mecha- in an increase in gene-targeting frequencies. The discrepancy nistically, the most likely role for Ku in rAAV random integration between human and chicken somatic cells can be reconciled given is to recruit DNA-PKcs to the viral ITRs. DNA-PKcs, would, in turn, that DT40 cells are known to employ HR at high frequency (34). recruit and activate the nuclease Artemis, which nicks open the viral An explanation for the difference between human and mouse cells, ITR hairpins (38), facilitating integration of the virus. The obser- however, is less obvious, albeit consistent with the more stringent vation that the vast majority of random rAAV integration events requirement for Ku/NHEJ in human cells (this work and ref. 14). that have been sequenced have a viral end point that maps to the One possibility may be that mouse embryonic stem or fibroblast viral ITRs supports this interpretation (39). The competition model cells, the cells in which the mouse targeting experiments were is not consistent with the conclusion that DNA-PK actually inhibits carried out, may, like DT40 cells, carry out HR at a higher rAAV integration in mouse muscle (40) nor with a report that shRNA-mediated silencing of DNA-PKcs gene expression in human MO59K somatic cells had no effect on rAAV-mediated gene Table 3. Summary of gene-targeting frequencies at the targeting (41). However, in preliminary studies with HCT116 cell LIGIV locus lines containing functional inactivation of either one or both No. of DNA-PKcs alleles, we have repeatedly seen large increases in the colonies No. of correctly Targeting Fold correct gene-targeting frequency that are comparable to those Cell line screened* targeted clones† frequency‡ increase§ reported here for Ku-reduced cell lines (unpublished data) sup- WT HCT116 176 2 1.13 1.0 porting both a role for DNA-PKcs in viral ITR nicking (38) and the 53-39 (Ku70ϩ/Ϫ) 148 6 4.05 3.5 competition model. Other investigators have suggested that in addition to Ku and *Drug-resistant clones that were also positive for the internal control PCR. † Rad52 competing for DNA ends, Ku may also actively suppress HR Drug-resistant clones that showed correct targeting by PCR. (42). We favor this hypothesis because it provides at least a partial ‡Targeting frequency is the number of correctly targeted colonies per 100 drug-resistant colonies screened. explanation for the finding that the frequency of random integra- §The WT targeting frequency is set at 1. The fold increase is the targeting tions was not lower in a Ku-reduced background (Fig. S1). Thus, if frequency in a speciﬁc cell line divided by the targeting frequency in the WT HR not only has less competition for the viral ends in the absence background (1.13). of Ku but is also more active, it may facilitate viral integrations at

Fattah et al. PNAS ͉ June 24, 2008 ͉ vol. 105 ͉ no. 25 ͉ 8707 Downloaded by guest on September 30, 2021 chromosomal sites that are quasi-homologous to the viral ITRs. Silencing of Ku70. The pLKO.1-puromycin-based lentiviral vectors containing Another potential explanation for the lack of an effect of a Ku sequence-verified shRNA targeting Ku70 (XRCC6; GenBank accession number deficiency on overall integrations is the recent description of an NM࿝001469) were obtained from the MISSION TRC-Hs 1.0 (Human) shRNA library alternative (Ku-independent) NHEJ pathway (A-NHEJ; reviewed through Sigma–Aldrich [TRCN0000039608 (608), TRCN0000039609 (609), TRCN0000039610 (610), TRCN0000039611 (611), TRCN0000039612 (612)]. For the in ref. 43). In contrast to the classical, Ku-dependent NHEJ, which RNAi experiments, predesigned, double-stranded siRNAs (SMARTPool) targeting works on virtually all DNA DSBs, A-NHEJ seems to only work on human Ku70 were purchased from Dharmacon. a subset of these and/or in specialized pathways because it can participate in the repair of switch recombination-induced DSBs but Targeting Vector Construction, Packaging, and Infection. The targeting vectors, not ionizing radiation-induced DSBs (43). If A-NHEJ can also work Ku70-Neo and LIGIV-Neo, were constructed by using the rAAV system as de- in the pathway of gene targeting to facilitate random integrations, scribed elsewhere (ref. 15 and S. O., unpublished data). The CCR5-Neo targeting it might explain why the frequency of correct gene targeting goes vector has been described in ref. 19. All virus packaging and infections were up (Tables 1–3) even when the frequency of random targeting does performed as described in ref. 19. not change (Fig. S1). In any case, it is important to emphasize that it has long been known that additional pathways for chromosomal Isolation of Genomic DNA and Genomic PCR. Genomic DNA for PCR screening was DNA integration are quite active in DNA-PK-defective cells (44). isolated by using phenol extraction followed by ethanol precipitation. Ku70- and LIGIV-targeting events were identified by PCR using the conditions described Regardless of the precise mechanism by which Ku regulates elsewhere (19). CCR5-targeting events were identified by PCR using the primers rAAV integration, we have demonstrated that a reduction in the and conditions described in ref. 19. levels of Ku70 in HCT116 human somatic cells greatly elevates the frequency of correct gene targeting. These observations have Antibodies and Immunoblotting. Ku70 and Ku86 (Santa Cruz Biotechnology) and significant practical implications for basic researchers interested in ␣-tubulin (Covance) antibodies were used for detection. Proteins were subjected gene-disruption strategies and for clinical researchers interested in to electrophoresis on a 4–20% gradient gel (Bio-Rad), electroblotted, and de- gene therapy. tected as described in ref. 15.

Methods ACKNOWLEDGMENTS. We are indebted to Dr. Bruce Vogelstein (The Johns Hopkins University, Baltimore, MD) and the many members of his laboratory who Cell Culture. The human colon cancer cell line HCT116 was obtained from were extremely generous with their reagents and advice. We also thank Brian American Type Culture Collection and maintained in McCoy’s 5A medium con- Ruis (University of Minnesota) for technical help in establishing and using the taining 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin, and 100 units/ml rAAV gene-targeting system. We thank Dr. A.-K. Bielinsky for helpful comments streptomycin. The cells were grown at 37°C in a humidiﬁed incubator with 5% on the manuscript. This work was supported in part by National Institutes of CO2. Cell lines harboring the targeting vector were grown in 1 mg/ml G418. Health Grants GM 069576 and HL079559 (to E.A.H.).

1. Hendrickson EA (2008) Gene targeting in human somatic cells, Sourcebook of Models 24. Ninomiya Y, Suzuki K, Ishii C, Inoue H (2004) Highly efficient gene replacements in for Biomedical Research, ed Conn PM (Humana, Totowa, NJ), pp 509–525. Neurospora strains deficient for nonhomologous end-joining. Proc Natl Acad Sci USA 2. van Veelen L, Wesoly J, Kanaar R (2006) Biochemical and cellular aspects of homolo- 101:12248–12253. gous recombination, DNA Damage Recognition, eds Seide W, Kow YW, Doetsch P 25. Colot HV, et al. (2006) A high-throughput gene knockout procedure for Neurospora reveals (Taylor and Francis, New York), pp 581–607. functions for multiple transcription factors. Proc Natl Acad Sci USA 103:10352–10357. 3. Heyer WD, Li X, Rolfsmeier M, Zhang XP (2006) Rad54: the Swiss Army knife of 26. Meyer V, et al. (2007) Highly efficient gene targeting in the Aspergillus niger kusA homologous recombination? Nucleic Acids Res 34:4115–4125. mutant. J Biotechnol 128:770–775. 4. Hastings PJ, McGill C, Shafer B, Strathern JN (1993) Ends-in vs. ends-out recombination 27. Poggeler S, Kuck U (2006) Highly efficient generation of signal transduction knockout in yeast. Genetics 135:973–980. mutants using a fungal strain deficient in the mammalian Ku70 ortholog. Gene 5. Hendrickson EA, Huffman JL, Tainer JA (2006) Structural aspects of Ku and the 378:1–10. DNA-dependent protein kinase complex, DNA Damage Recognition, eds Seide W, Kow 28. Goins CL, Gerik KJ, Lodge JK (2006) Improvements to gene deletion in the fungal YW, Doetsch P (Taylor and Francis, New York), pp 629–684. pathogen Cryptococcus neoformans: Absence of Ku proteins increases homologous 6. Thomas KR, Capecchi MR (1987) Site-directed mutagenesis by gene targeting in mouse recombination, and cotransformation of independent DNA molecules allows rapid embryo-derived stem cells. Cell 51:503–512. complementation of deletion phenotypes. Fungal Genet Biol 43:531–544. 7. Inagaki K, et al. (2007) DNA palindromes with a modest arm length of greater, similar 29. Yamana Y, et al. (2005) Regulation of homologous integration in yeast by the DNA 20 base pairs are a significant target for recombinant adeno-associated virus vector repair proteins Ku70 and RecQ. Mol Genet Genomics 273:167–176. integration in the liver, muscles, and heart in mice. J Virol 81:11290–11303. 30. Kooistra R, Hooykaas PJ, Steensma HY (2004) Efficient gene targeting in Kluyveromy- 8. Spagnolo L, Rivera-Calzada A, Pearl LH, Llorca O (2006) Three-dimensional structure of ces lactis. Yeast 21:781–792. the human DNA-PKcs/Ku70/Ku80 complex assembled on DNA and its implications for 31. Fukushima T, et al. (2001) Genetic analysis of the DNA-dependent protein kinase DNA DSB repair. Mol Cell 22:511–519. reveals an inhibitory role of Ku in late S–G2 phase DNA double-strand break repair. 9. Carter BJ (2004) Adeno-associated virus and the development of adeno-associated J Biol Chem 276:44413–44418. virus vectors: A historical perspective. Mol Ther 10:981–989. 32. Pierce AJ, Hu P, Han M, Ellis N, Jasin M (2001) Ku DNA end-binding protein modulates 10. Russell DW, Hirata RK (1998) Human gene targeting by viral vectors. Nat Genet homologous repair of double-strand breaks in mammalian cells. Genes Dev 15:3237–3242. 18:325–330. 33. Dominguez-Bendala J, Masutani M, McWhir J (2006) Down-regulation of PARP-1, but 11. Chamberlain JR, et al. (2004) Gene targeting in stem cells from individuals with not of Ku80 or DNA-PK , results in higher gene targeting efficiency. Cell Biol Int osteogenesis imperfecta. Science 303:1198–1201. cs 30:389–393. 12. Donsante A, et al. (2007) AAV vector integration sites in mouse hepatocellular carci- 34. Buerstedde JM, Takeda S (1991) Increased ratio of targeted to random integration noma. Science 317:477. after transfection of chicken B cell lines. Cell 67:179–188. 13. Russell DW (2007) AAV vectors, insertional mutagenesis, and cancer. Mol Ther 35. Hendrie PC, Russell DW (2005) Gene targeting with viral vectors. Mol Ther 12:9–17. 15:1740–1743. 14. Li G, Nelsen C, Hendrickson EA (2002) Ku86 is essential in human somatic cells. Proc Natl 36. Van Dyck E, Stasiak AZ, Stasiak A, West SC (1999) Binding of double-strand breaks in Acad Sci USA 99:832–837. DNA by human Rad52 protein. Nature 398:728–731. 15. Fattah KR, Ruis BL, Hendrickson EA (2008) Mutations to Ku reveal differences in human 37. Zentilin L, Marcello A, Giacca M (2001) Involvement of cellular double-stranded DNA somatic cell lines. DNA Repair 7:762–774. break binding proteins in processing of the recombinant adeno-associated virus 16. Myung K, et al. (2004) Regulation of telomere length and suppression of genomic genome. J Virol 75:12279–12287. instability in human somatic cells by Ku86. Mol Cell Biol 24:5050–5059. 38. Inagaki K, Ma C, Storm TA, Kay MA, Nakai H (2007) The role of DNA-PKcs and Artemis 17. Moffat J, et al. (2006) A lentiviral RNAi library for human and mouse genes applied to in opening viral DNA hairpin termini in various tissues in mice. J Virol 81:11304–11321. an arrayed viral high-content screen. Cell 124:1283–1298. 39. Miller DG, et al. (2005) Large-scale analysis of adeno-associated virus vector integration 18. Lederman MM, Penn-Nicholson A, Cho M, Mosier D (2006) Biology of CCR5 and its role sites in normal human cells. J Virol 79:11434–11442. in HIV infection and treatment. J Am Med Assoc 296:815–826. 40. Song S, et al. (2004) DNA-dependent PK inhibits adeno-associated virus DNA integra- 19. Kohli M, Rago C, Lengauer C, Kinzler KW, Vogelstein B (2004) Facile methods for tion. Proc Natl Acad Sci USA 101:2112–2116. generating human somatic cell gene knockouts using recombinant adeno-associated 41. Vasileva A, Linden RM, Jessberger R (2006) Homologous recombination is required for viruses. Nucleic Acids Res 32:e3. AAV-mediated gene targeting. Nucleic Acids Res 34:3345–3360. 20. Kim JS, et al. (2002) Oncogenic ␤-catenin is required for bone morphogenetic protein 42. Allen C, Kurimasa A, Brenneman MA, Chen DJ, Nickoloff JA (2002) DNA- 4 expression in human cancer cells. Cancer Res 62:2744–2748. dependent protein kinase suppresses double-strand break-induced and spontane- 21. Shirasawa S, Furuse M, Yokoyama N, Sasazuki T (1993) Altered growth of human colon ous homologous recombination. Proc Natl Acad Sci USA 99:3758– cancer cell lines disrupted at activated Ki-ras. Science 260:85–88. 3763. 22. Kim JS, Lee C, Foxworth A, Waldman T (2004) B-Raf is dispensable for K-Ras-mediated 43. Nussenzweig A, Nussenzweig MC (2007) A backup DNA repair pathway moves to the oncogenesis in human cancer cells. Cancer Res 64:1932–1937. forefront. Cell 131:223–225. 23. Russell DW, Miller AD, Alexander IE (1994) Adeno-associated virus vectors preferen- 44. Staunton JE, Weaver DT (1994) scid cells efficiently integrate hairpin and linear DNA tially transduce cells in S phase. Proc Natl Acad Sci USA 91:8915–8919. substrates. Mol Cell Biol 14:3876–3883.

8708 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0712060105 Fattah et al. Downloaded by guest on September 30, 2021