Targeted Genome Modifications Using Integrase ... - CyberLeninka

9 downloads 0 Views 261KB Size Report
Dec 12, 2007 - We thank Eva Guhl (Charité Medical School, Berlin, Germany) for ... Bohne (Hannover Medical School, Hannover, Germany), Frank Radecke.
original article

© The American Society of Gene Therapy

Targeted Genome Modifications Using Integrase-deficient Lentiviral Vectors Tatjana I Cornu1 and Toni Cathomen1 Charité Medical School, Institute of Virology (CBF), Berlin, Germany

1

Gene correction aims at repairing a defective gene directly in the cellular genome, which warrants tissue-specific and sustained expression of the repaired gene through its endogenous promoter. We have developed a novel system based on integrase-deficient ­ lentiviral vectors (IDLVs) that allows us to correct an ­endogenous mutation using a strategy based on homologous ­ recombination (HR). In a proof-of-concept approach, an IDLV encoding a repair template was co-delivered with an I-SceI nuclease expression vector to rescue a ­ defective enhanced green fluorescent protein (EGFP) gene. Expression of the nuclease created a double-strand break within the target locus, which was crucial for ­stimulating IDLV-based gene repair. Stable gene ­correction was ­realized in up to 12% of the cells, depending on the vector dose, the nuclease expression levels, and the cell type. Genotypic analyses confirmed that gene correction was the result of genuine HR between the target locus and the IDLV repair template. This study presents IDLVs as valuable tools for introducing precise and permanent genetic modifications in human cells. Received 25 January 2007; accepted 28 September 2007; published online 13 November 2007. doi:10.1038/sj.mt.6300345

Introduction The goal of gene-addition-type protocols is to complement a defective genome with a therapeutic DNA. While this concept is relatively simple, and early clinical trials have been ­successful,1–4 the approach has some drawbacks, including the ­ difficulty in regulating transgene expression and the risk of insertional ­mutagenesis. In contrast, correcting a mutation directly at the endogenous chromosomal locus by gene targeting preserves both tissue-specific and temporal control of gene expression. Gene correction is a rather complex operation involving homologous ­recombination (HR) between the target locus and an exogenous DNA repair template, and the molecular details have not yet been fully resolved. Until recently, the low frequency of HR in mammalian cells prevented strategies aiming at gene correction from reaching therapeutically reasonable levels. During the past few years, however, studies carried out in various laboratories have established that the creation of a targeted DNA double-strand break (DSB) stimulates

HR several hundred-fold by activating the cellular DNA repair machinery. Proof-of-principle approaches involving the meganuclease I-SceI, which binds to an 18 bp recognition site, demonstrated that the insertion of a DSB in the target locus stimulates recombination with an exogenous DNA substrate.5–7 Importantly, the stimulation of HR by I-SceI has been accomplished in various cell lines, including mouse embryonic stem cells,7 thereby indicating that DSB-induced stimulation of HR is a universal cellular phenomenon. Recent advances in generating artificial nucleases such as zinc-finger nucleases, to insert a DSB at pre-selected sites in the human genome, have further paved the way for HR-based strategies in gene therapy.8 On the basis of the achievements and setbacks in some recent trials,1–4 HR-based genome editing in hematopoietic stem cells (HSCs) would appear to be an especially attractive alternative. Although vectors based on adeno-associated virus (AAV) have proved successful in achieving gene targeting at high frequencies in cultured cells,9–11 attempts to transduce HSCs have shown only limited success.12 γ-Retroviral vectors are better suited for transducing HSCs,13 and attempts have been made to correct genes using such vectors. Probably because of the fast integration ­ kinetics of reverse-transcribed vector genomes, the frequency of gene targeting was very low, and the concept of using γ-retroviral vectors for therapeutic gene correction was therefore abandoned.14 In this study, we explored the use of a novel vector system for gene targeting approaches. Integrase-deficient lentiviral vectors (IDLVs) combine high infectivity with the ability to provide an episomal DNA substrate for HR. The viral integrase (IN) is a key component in lentiviral integration. Because IN is also involved in reverse transcription of human immunodeficiency virus type 1 RNA and nuclear import of the pre-integration complex, it cannot be deleted completely. However, recent studies have shown that IN mutants harboring mutations in the catalytic core permit the formation of episomal long-terminal repeat (LTR) circles and efficient expression of a transgene.15–21 In this study, we tested five different IN mutants for their ability to mediate transient transgene expression and to reduce the frequency of vector integration. Using a cell-based enhanced green fluorescent protein (EGFP)-rescue assay, we demonstrated that, upon stimulation of HR through a targeted DSB, IDLVs have the capacity to serve as templates for HR and to mediate gene correction at high frequencies. These results present IDLVs as promising tools for introducing precise and permanent genetic modifications in mammalian cells.

Correspondence: Toni Cathomen, Charité Medical School, Institute of Virology (CBF), Hindenburgdamm 27, D-12203 Berlin, Germany. E-mail: [email protected] Molecular Therapy vol. 15 no. 12, 2107–2113 dec. 2007

2107

© The American Society of Gene Therapy

IDLV-based Gene Targeting

Results IDLV as a template for gene expression Transgene expression from IDLVs probably depends on both the number and the structure of the extrachromosomal viral DNA.21 We chose to modify various key residues in the IN gene to find the best-suited vector for transient gene expression and targeted genome modification. Mutation H12A dissolves a N-terminal zincbinding motif that promotes proper folding and tetramerization of IN,22,23 whereas mutations D64A, D116A, and F185A are expected to interfere directly with the catalytic activity of IN.24,25,18 Mutation K264E is located in the DNA-binding domain and should result in reduced strand transfer activity.26,27 As a surrogate marker for the intracellular presence of vector genomes, the expression kinetics of EGFP after transduction with the various IDLVs was assessed over a period of 6 weeks. The IDLVs expressed EGFP from an internal cytomegalovirus (CMV) promoter and were used for transducing HEK293T cells with various vector doses. Flow cytometry was performed at 1 week intervals to determine the percentage of EGFP-positive cells (Figure 1a–c). As expected, the number of EGFP-positive cells decreased rapidly over time for all vector doses

1

K264E

F185A

D116A

D64A

e

6 ng

10

1

0.1

wt

K264E

F185A

D116A

D64A

% EGFP-positive cells

10

1

0.1

0.01

K264E

F185A

D116A

D64A

H12A

K264E F185A D116A D64A

H12A

60 ng 6 ng

4,000

2,000

wt

K264E F185A D116A D64A

H12A

10,000

2 day 7 day

7,500 5,000 2,500 0

wt

wt

6,000

0

f

2d 1w 2w 3w 4w 5w 6w

1

8,000

H12A

0.6 ng

100

60 ng 6 ng

10

0.1

H12A

Mean fluorescense intensity

% EGFP-positive cells

c

wt

100

0.01

100

% Integration

10

0.1

b

d

60 ng

100

Mean fluorescense intensity

% EGFP-positive cells

a

and remained stable after 5 weeks. The residual integration frequency of the various IDLVs was calculated by comparing the percentage of EGFP-positive cells after 6 weeks with the transduction rate at day 2 after infection (Figure 1d). The analysis confirmed the reduced tendency of IDLV vectors to integrate. Depending on the IN mutation and vector dose, the integration frequencies ranged between 0.35 and 2.3%. Assessment of the EGFP expression levels at day 2 after transduction revealed significant differences (Figure 1e). While IN mutations affecting the catalytic core (D64A, D116A) supported high transgene expression, EGFP levels were significantly (P < 0.01) lower after transduction with IDLVs harboring mutations H12A, F185A, and K264E. Moreover, comparison of the wildtype vector with IDLV-D64A and IDLV-D116A revealed that a tenfold higher vector dose (60 as against 6 ng) had to be used in order to attain the same transgene expression levels. Importantly, when compared with the wild-type vector, IDLV-mediated EGFP expression dropped quickly within the first week after transduction (Figure 1f), thereby confirming that cell proliferation leads to rapid dilution of the LTR circles. On the basis of this

60

6 wt

60 6 D116A

60 6 D64A

ng p24 IN

Figure 1 Characterization of integrase-deficient lentiviral vectors. (a–c) Kinetics of enhanced green fluorescent protein (EGFP) expression from various integrase-deficient lentiviral vectors (IDLVs). HEK293T cells were transduced with different amounts of the indicated vectors (equivalent of 60 ng, 6 ng, and 0.6 ng of p24/50,000 cells). Cells were harvested at 1 week intervals and the percentage of EGFP-positive cells was determined by flow cytometry. (d) Integration frequency. The frequency of IDLV integration for the indicated vector doses (ng of p24/50,000 cells) is shown as the proportion of EGFP-positive cells at week 6 in relation to those at day 2. (e, f) EGFP expression level. The EGFP expression levels are displayed as the mean fluorescence intensity as determined by flow cytometry 2 or 7 days after transduction with the indicated vector doses. The respective mutations in integrase (IN) are indicated; wt refers to wild-type IN.

2108

www.moleculartherapy.org vol. 15 no. 12 dec. 2007

© The American Society of Gene Therapy

Cell-type dependency of IDLV-mediated gene targeting In order to investigate whether the efficiency of IDLV-mediated gene targeting is cell type–dependent, human polyclonal target cell lines were produced by transduction with a lentiviral ­vector harboring the target locus under the control of an internal CMV promoter. After transduction of these cells with IDLV-RM and LV-SceI, or with IDLV-RM and IDLV-SceI, the frequency Molecular Therapy vol. 15 no. 12 dec. 2007

∂LTR

∂LacZ

EGFP ∂GFP

LacZ

LacZ

b % EGFP-positive cells

IDLV as a template for gene targeting In order to explore whether IDLVs can serve as a template for targeted genome modifications, gene targeting was evaluated in a cellular assay based on the rescue of a defective EGFP gene.28 A polyclonal U2-OS-based cell line containing a single copy target locus under the control of the LTR promoter was established by retroviral transduction (Figure 2a). In order to ensure that no functional EGFP is expressed from these cells, the LacZ gene is followed by stop codons, and the EGFP gene was truncated at the 5′-end (∂GFP). The binding site for the I-SceI nuclease was placed between the open reading frames for LacZ and ∂GFP. The repair matrix harbors a non-transcribed 5′-truncated LacZ–EGFP fusion gene, and was based on IN mutant D64A (IDLV-RM). Using this setup, rescue of EGFP expression can occur exclusively through HR between the LacZ-∂GFP target locus and the ∂LacZ-EGFP repair matrix. The U2-OS target cell line was co-transduced with IDLVRM and various amounts of a lentiviral I-SceI expression vector containing either wild-type IN (LV-SceI) or IN mutant D64A (IDLV-SceI). As determined by flow cytometry, gene correction was achieved in up to 5% of the cells co-transduced with IDLV-RM and LV-SceI (Figure 2b), and up to 0.8% of the cells transduced with IDLV-RM and IDLV-SceI (Figure 2c). In both cases, the correction frequencies were dependent on the amount of I-SceI expression vector. In the absence of I-SceI, only very few cells underwent HR at the target locus (Figure 2b). Also, when a repair matrix with a wild-type IN (LV-RM) was used, EGFP­positive cells were hardly detectable (data not shown), which indicates that EGFP expression stemming from a repair vector integrating ­in-frame with an active gene is negligible. Vector-mediated I-SceI expression levels were assessed by immunoblotting 2 days after transduction with LV-SceI or IDLVSceI. While I-SceI expression from LV-SceI was readily detectable, protein expression from IDLV-SceI was not (Figure 2d). This is in accordance with the flow cytometric analysis of EGFP expression (Figure 1e), which showed significantly lower transgene expression levels from the D64A-based IDLV. We assume that the lower level of IDLV-mediated I-SceI expression is the main reason for the observed lower frequency of gene correction. In summary, these experiments clearly demonstrate that IDLVs can mediate stable genome modifications in a dose-­dependent manner. They further establish that efficient IDLV-mediated gene targeting is dependent on the presence of a DSB in the target locus, and identify nuclease expression as the rate-limiting step.

a

EGFP

10

IRES IRES

IRES

Neo Neo

Neo

wpre

∂LTR

RM TL

wpre

cTL

wpre

LV-Scel (wt IN) 300

1

100 30

0.1 0 0.01

c % EGFP-positive cells

c­ haracterization, IDLVs harboring IN mutation D64A were chosen for all further experiments, because they combine high initial EGFP expression with a low integration frequency.

IDLV-based Gene Targeting

1

2

3

4

5

6

7

Week 10

IDLV-Scel (D64A IN)

1 300 100 30

0.1

0.01

1

2

3

4

5

6

7

Week

d

IN

wt

ng p24

30

I-Scel

D64A 150

50

cto 30





Figure 2  Integrase-deficient lentiviral vector (IDLV)–mediated gene correction. (a) Schematic of the genomic target locus (TL) and the IDLVbased repair matrix (RM). The U2-OS-based target cell line was established by retroviral transduction. Internal ribosome entry site–based expression of the neomycin resistance gene permitted selection of TLcontaining cells in the presence of geneticin. The TL consists of a LacZ gene under the control of the long terminal repeat (LTR) promoter, followed by stop codons, the I-SceI recognition site, and a 5′-truncated enhanced green fluorescent protein (EGFP) (∂GFP). In order to ensure that EGFP-rescue is strictly dependent on homologous recombination (HR), the RM is devoid of any promoter and contains a 5′-truncated LacZ gene fused in-frame to the EGFP gene. ∂LTR, lentiviral LTR in the SIN configuration; cTL, corrected target locus. (b, c) Gene targeting in U2OS cells. Target cells were transduced with constant amounts of IDLV-RM (200 ng) and variable amounts of either b LV-SceI (wild-type integrase) or c IDLV-SceI (D64A integrase). The cells were subjected to flow cytometry at 1 week intervals to determine the fraction of EGFP-positive cells. Numbers refer to vector dose (ng of p24/50,000 cells). (d) I-SceI expression levels. HEK293T cells were transduced with either LV-SceI or IDLVSceI. After 2 days cells were harvested and analyzed by immunoblotting using an anti-HA tag specific antibody. Numbers refer to vector dose (ng of p24/105 cells).

of EGFP rescue was determined by flow cytometry at 1 week intervals (Figure 3a–c). Depending on vector dose, gene correction was achieved in 12 or 8% of U2-OS cells co-transduced with IDLV-RM and LV-SceI, and in 1.9 or 1.1% of cells transduced with IDLV-RM and IDLV-SceI (Figure 3a). In the absence of I-SceI expression, rescue of EGFP expression was barely ­detectable 2109

© The American Society of Gene Therapy

IDLV-based Gene Targeting

∂GFP

I-Scel

wt

D64

35

36

37

3

5

8

22

24 H2O

2

3

5

6

9

10

11

21

24 H2O

H2O

cto

29

cto

Mock

Clone# 26

wt

DSB

HT-1080

D64

U2-OS H2O

HEK293

CMV

RI DSB

0.1 w/o 0.01

GDH 1

2

3

4

Week

b

∂LTR

cTL

D64/H D64/L

1

LacZ

cTL

cto

wt/H wt/L

EGFP

Mock

b

U2-OS

LacZ

Mock

10

a

D64

100

Molecular characterization of transduced cells by genotyping In order to confirm that EGFP expression in the transduced target cells was the result of a genuine HR event, we evaluated the genotype of these cells. To this end, a nested polymerase chain reaction (PCR) was performed with two sets of primers that amplify a product only from the corrected LacZ-EGFP target locus (Figure  4a). For all cell lines, a PCR product was detected from samples co-transduced with IDLV-RM and either LV-SceI or IDLV-SceI (Figure 4b). Sham infected cells (mock) or the parental cell lines without target locus (cto) remained negative. This demonstrates that EGFP-positive cells arise as a result of legitimate HR between the target locus and the IDLV-based repair matrix. In order to characterize the potential for mutagenesis, we assessed whether

wt

a % EGFP-positive cells

(0.03%). Interestingly, the expression level of the target locus appears to be an essential determinant of the gene targeting frequency, as shown by the fact that gene correction at the CMVdriven LacZ-∂GFP locus was more efficient than correction at the LTR-controlled target locus (Figure 2b and c). Similar gene correction frequencies were accomplished in HEK293-based target cells (Figure 3b), whereas IDLV-mediated gene targeting in HT-1080 cells was considerably lower (Figure 3c). With the higher vector dose, 2.3% of cells co-transduced with IDLV-RM and LV-SceI and 0.4% of cells infected with IDLV-RM and IDLVSceI were EGFP-positive. These results confirm that IDLVs can serve as templates for gene targeting, and that nuclease expression is the rate-limiting step. In view of the fact that we did not observe significant differences in the number of EGFP-positive cells when transduced with an EGFP control lentivirus (data not shown), these experiments suggest that the frequency of IDLV-mediated gene targeting is cell type–dependent.

c

HEK293

100

1

0.01

RI GDH

D64/H D64/L

0.1

d 2

3

Clone# cTL

4

Week

% EGFP-positive cells

c

RI

HT-1080

10

wt/H wt/L

1

D64/H D64/L

0.1

0.01

1

2

3

4

Week

Figure 3 Cell type dependency of integrase-deficient lentiviral ­vector (IDLV)–mediated gene targeting. Gene correction was performed in (a) U2-OS, (b) HEK293, and (c) HT-1080 cells. Target cells were co-transduced with a constant amount of IDLV-RM (150 ng) and variable amounts of either LV-SceI (wt) or IDLV-SceI (D64) using a vector dose of 120 ng (H) or 40 ng of p24/50,000 cells (L), respectively. The absence of an I-SceI expression vector is indicated by w/o. The fraction of enhanced green fluorescent protein (EGFP)–positive cells was determined at 1 week intervals by flow cytometry. RM, repair matrix; wt, wild type.

2110

1

IDLV-RM IDLV-Sce l

1

IDLV-RM LV-Scel

% EGFP-positive cells

cTL wt/H wt/L

10

GDH

Figure 4  Genotypic analysis. (a) Schematic illustrations of the corrected target locus (cTL) and an integration event into the I-SceI induced DNA double strand break (DSB). The positions of the primers used for the nested polymerase chain reaction protocols are shown. (b) Genotyping of polyclonal cell cultures. Genomic DNA was isolated from the indicated target cells after co-transduction with IDLV-RM and LV-SceI (wt) or IDLVRM and IDLV-SceI (D64), respectively. PCRs were performed to detect the cTL, random integration events of the I-SceI expression vectors (RI), integration of the I-SceI vector into the nuclease induced DSB, or a control reaction amplifying a fraction of the glyceraldehyde 3-phosphate dehydrogenase gene (GDH). Sham-transduced target cells (mock), the respective parental cell lines (cto), and water (H2O) served as controls. (c, d) Genotyping of clonal cell populations. Genomic DNA was isolated from HEK293 target cells after co-transduction with IDLV-RM and LV-SceI c or IDLV-RM and IDLV-SceI d. PCRs were performed to detect the cTL, RI events, or GDH. IDLV, integrase-deficient lentiviral vector; RM, repair matrix; wt, wild type.

www.moleculartherapy.org vol. 15 no. 12 dec. 2007

© The American Society of Gene Therapy

we could detect the I-SceI expression cassette in the extracted genomic DNA. As expected, in all cell lines transduced with the LV-SceI vector which harbors the wild-type IN, an I-SceI-specific product was amplified. In contrast, stable integration of IDLV-SceI was detectable only in the U2-OS target cells. Given that episomal DNA seems to integrate preferentially into pre-existing DSBs,29 we also assessed whether we could detect integration of the I-­SceIexpression vectors into the DSB created by the I-SceI nuclease. After performing a nested PCR with primers that bind within the I-SceI gene and the uncorrected ∂GFP target locus (Figure 4a), integration into the DSB was detected only in U2-OS target cells with the IN-deficient I-SceI-expression vector (Figure 4b). In order to delineate further IDLV-mediated genome editing, single HEK293 cell clones were characterized. In an assessment of cells co-transduced with IDLV-RM and LV-SceI, 9 out of 10 clones revealed integration of the LV-SceI vector (Figure 4c). Interestingly, clone #22 was EGFP-positive and harbored a corrected target locus, but integration of the LV-SceI vector could not be detected. Given that non-stimulated (i.e., in the absence of a DSB) gene targeting is a rare event, we speculate that the lentiviral LV-SceI vector was lost as a result of either chromosomal instability during proliferation of these cells or I-SceI-induced chromosomal rearrangements. In contrast, integration of the IDLV-SceI vector could not be detected in any of the 10 clones analyzed (Figure 4d). In summary, the genotypic analyses validated the results obtained by flow cytometry. Furthermore, they verified that rescue of EGFP expression in all the target cell lines analyzed was the result of genuine HR between the chromosomal target locus and the IDLV-based repair matrix. With regard to insertional mutagenesis, the results confirmed the non-integrative character of IDLVs but indicated that LTR circles can integrate into pre­existing DSBs.

Discussion Non-integrating lentiviral vectors retain the high transduction efficiency of their integrating counterparts in the absence of the potential problems relating to insertional mutagenesis. This type of vector can be generated by either altering the function of IN or by changing the two conserved CA residues in the attachment site of the viral LTRs (reviewed in ref. 21). As a result of intramolecular recombination, IDLVs form circular DNA episomes, and sustained transgene expression has been reported after transduction of non-dividing cells both in vivo and in vitro.15–21 Because these LTR circles lack replication signals, they are diluted in proliferating cells and transgene expression is short-lived. In this study, we present the use of IDLV as a tool for achieving precise and stable genome modifications in human-derived cells. Our results confirm that some IDLVs mediate transient transgene expression in dividing cells, and establish IDLVs as efficient templates for legitimate HR. Several laboratories have reported the use of viral vectors to accomplish precise modifications of the genome (reviewed in ref. 30). Some promising studies reported a targeting frequency of up to 0.2% when using helper-dependent adenoviral vectors in mouse embryonic stem cells.31–33 Given the cytotoxic and/or immunogenic nature of these vectors,34 however, it remains to be Molecular Therapy vol. 15 no. 12 dec. 2007

IDLV-based Gene Targeting

determined whether they are ideal tools for achieving sustained gene correction. In contrast, the use of AAV vectors for gene correction proved quite successful in cultured cells. In initial experiments aiming at the correction of single copy selection markers, the targeting frequency approached 1%.9,35 In the absence of any selection pressure, however, the frequency of gene targeting was considerably lower, reaching frequencies of 0.006–0.1%, depending on the cell type.36 Nonetheless, a recent in vivo study demonstrated that AAV-mediated gene correction could also be accomplished in the mouse liver, at frequencies of ~10–4 (ref. 37). Early studies in yeast and vertebrate cells indicated that the process of HR between a chromosomal locus and an episomal DNA substrate can be stimulated up to 1,000-fold by inserting a DSB in the genomic target locus (reviewed in ref. 38). This principle holds true for AAV-mediated gene targeting, and it has been shown that correction of a marker gene could be enhanced >100fold when a DSB is inserted in the target locus using I-SceI.10,11 In both studies, gene correction frequencies of ~1% were reached. Our studies point out that, as in AAV, IDLV-mediated gene targeting reaches clinically relevant levels only in the presence of a DSB. Depending upon the cell type and the nuclease expression level, targeted genome modifications with IDLVs were realized in up to 12% of the cells. The cell type dependency of IDLV-mediated gene targeting could reflect variations in the availability of cellular factors required for DSB-stimulated gene correction, especially DNA repair factors involved in homology-directed repair. Moreover, factors mediating non-homologous end-joining seem to be differentially available in the cells used in this study. Integration of IDLVs, which probably depends on non-homologous end-­joining, was detected only in U2-OS cells. It will be interesting to find out whether exogenous stimulation of the homology-directed repair pathway and suppression of non-homologous end-joining will enhance HR-based gene repair. Vector-induced insertional mutagenesis and genotoxicity have been under scrutiny for the last couple of years.39 Although our data confirmed the low frequency of integration associated with IDLVs, they nonetheless verified a residual integration activity, as reported earlier.40 Moreover, the genotypic characterization revealed that IDLVs integrate into the I-SceI-induced DSB. This is in good agreement with observations made for AAV vectors, which were shown to integrate preferentially into pre-existing DSBs.29 We have generated several IN mutants and compared the level of transgene expression and the residual integration frequency. Only the IN class I mutants D64A and D116A combined high transgene expression with low residual integration. It will be important to screen additional IN mutants and possibly combine them with attachment-site mutants in order to enhance the transgene expression levels and further reduce residual integration activity. Even so, the use of IN-deficient vectors combined with the SIN configuration (which self-inactivates the LTR ­promoters)41,42 should significantly minimize the risk of insertional oncogenesis.43,44 Given the complexity of a mammalian genome, precise manipulations might be most promising in an ex vivo approach aimed at the genetic correction of stem cells. On the basis of the existing large body of knowledge and several successful clinical trials,1–4 IDLV-based genetic correction of HSCs could be an 2111

© The American Society of Gene Therapy

IDLV-based Gene Targeting

especially promising approach. A gene-targeting frequency of 10% might be sufficient for clinical applications, especially if corrected cells have a selective growth advantage, as reported for corrected HSCs in severe combined immunodeficiency-X1 patients.2 Several hurdles will have to be overcome before this system can be applied in a pre-clinical context in somatic stem cells. In order to correct a mutation that triggers a genetic disorder, artificial nucleases have to be developed that create a DSB close to the mutation. The design and use of specific zinc-finger nucleases will be especially promising for that purpose, in view of the fact that several laboratories have demonstrated the feasibility of stimulating targeted genome modifications in human cells by the use of such zinc-finger nucleases.28,45–48 Additional optimization of the design of the IDLVs with respect to integration activity, nuclease expression levels, and homology to the target locus will be crucial to advance IDLV-based gene correction as a true tool for accomplishing precise genetic modifications in somatic stem cells with a minimum of genotoxic side effects.

Materials and Methods Plasmids. In order to generate plasmid pLV-CMV.SceI, the I-SceI gene

was amplified by PCR from plasmid pRK5.LHA-SceI49 and ligated into p156RRLsinPPTcmv.GFP.pre (pLV-CMV.EGFP; gift of Alexander Pfeifer, University of Bonn, Bonn, Germany). The repair matrix pLV.∂LacZGFPiNwpre is based on vector pLV-CMV.EGFP and was assembled by a multi-step PCR strategy. Plasmids pS11.LacZs31∂GFPiNwpre and pLV-CMV.LacZs31∂GFPiNwpre, which encode the target locus, are based on plasmids S11.EGiN (gift of Helmut Hanenberg, HeinrichHeine-University, Düsseldorf, Germany) and pLV-CMV.EGFP, respectively, and generated by inserting the LacZs31∂GFP cassette from pCMV. LacZs31∂GFP.28 The packaging plasmids encoding the mutant INs (H12A, D64A, D116A, F185A, and K264E) were created by site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA), using plasmid pMDL.gpRRE (gift of Inder Verma, The Salk Institute, La Jolla, CA) as a template. Maps and sequences of plasmids are available upon request. Generation of viral vectors. Retroviral vectors for establishing the U2-OS

target cell line were obtained by the following method: Phoenix cells were transfected in a 10 cm vessel by the calcium phosphate method with 3 µg of pMD-G (gift of Inder Verma, La Jolla, CA), 5 µg pM57 (gift of Christopher Baum, Hannover Medical School, Hannover, Germany) and 10 µg of pS11.LacZs31∂GFPiNwpre. Viral supernatants were collected 48 hours later and passed through a 45 µm filter. Lentiviral vector production was based on the method described in ref. 41. Briefly, 7.8 × 106 HEK293T cells were transfected by calcium phosphate in a 150 mm dish with 22 µg of a pLV-genomic plasmid, 14.6 µg of pMDL.gpRRE or mutant pMDL.gpRRE, 5.6 µg of pRSV-Rev (gift of Inder Verma, La Jolla, CA) and 7.9 µg of pMD-G. Vectors were harvested 48 and 72 hours later, passed through a 45 µm filter, and concentrated by ultracentrifugation (2 hours at 67,000g at 10 °C). Pellets were resuspended at room temperature for 30 minutes in Dulbecco’s modified Eagle’s medium. The lentiviral vector titer was determined based on p24 quantification using the VIDAS human immunodeficiency virus p24 II kit (bioMérieux, Nürtingen, Germany). Target cell lines and gene correction assay. U2-OS, HEK293, HT-1080,

and HEK293T cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum. The U2-OS, HEK293, and HT-1080-based target cell lines were generated by transduction with either retroviral vector S11.LacZs31∂GFPiNwpre or lentiviral vector LVCMV.LacZs31∂GFPiNwpre. In order to ensure that the target cell lines contained a single copy target locus, the parental cells were infected with a pre-defined volume of vector that rendered