RESEARCH ARTICLE

Transposition from a gutless adeno-transposon vector stabilizes transgene expression in vivo

Stephen R.Yant, Anja Ehrhardt, Jacob Giehm Mikkelsen, Leonard Meuse, Thao Pham, and Mark A. Kay*

Published online: 16 September 2002, doi:10.1038/nbt738

A major limitation of adenovirus-mediated therapy for inherited diseases is the instability of transgene expression in vivo, which originates at least in part from the loss of the linear, extrachromosomal vector . Herein we describe the production of a gene-deleted adenovirusÐtransposon vector that stably maintains -encoded transgenes in vivo through integration into host . This system uti- lizes a donor transposon vector that undergoes Flp-mediated recombination and excision of its therapeutic payload in the presence of the Flp and Sleeping Beauty recombinases. Systemic in vivo delivery of this sys- tem resulted in efficient generation of transposon circles and stable -mediated integration in mouse liver. Somatic integration was sufficient to maintain therapeutic levels of human coagulation Factor IX for more than six months in mice undergoing extensive liver proliferation. These vectors combine the versatili- ty of adenoviral vectors with the integration capabilities of a eukaryotic DNA transposon and should prove use- ful in the treatment of genetic diseases. http://www.nature.com/naturebiotechnology

Successful gene therapy for genetic diseases will require two major most active of these members is the Tc1-like element Sleeping attributes from a single vector: high-efficiency gene delivery and - Beauty, which was reconstructed from pieces of an ancient fish ele- long therapeutic . Among the vector technologies ment16,17. This transposon undergoes cut-and-paste transposition currently available, recombinant adenovirus (Ad) vectors remain through a DNA intermediate, a process that requires the binding of attractive vehicles for therapeutic gene transfer. These vectors can be the Sleeping Beauty transposase (SB) to short direct-repeat sequences grown to high titers, exhibit a broad tropism, can transduce dividing embedded in the terminal inverted repeats (IRs) of the element17. and nondividing cells, and are one of the most efficient vehicles for Transposition is catalyzed entirely by the SB transposase and always in vivo gene delivery1. However, early versions of Ad vectors were occurs into a TA target dinucleotide, which is duplicated upon inser- plagued by toxicity and immunogenicity due to in vivo synthesis of tion by cellular DNA repair pathways. Previously, we reported that viral antigens from still contained in the vector2–4. SB transposase expression allows -based transposons to inte- To overcome these problems, the Cre-loxP helper-dependent grate safely and stably into mouse liver DNA18. Importantly, transpo- © Group 2002 Nature Publishing (HD) system was developed to generate recombinant adenoviruses sition was sufficient to maintain long-term transgene expression in which all viral coding sequences have been deleted5. Because the in vivo, even during periods of extensive liver regeneration19. Despite HD vector requires only the inverted terminal repeats (ITRs) and its enormous potential, the inability to deliver this integrating system packaging signal for proper DNA replication and virus assembly, in vivo efficiently in a clinically relevant manner remains a formida- these vectors can accommodate up to 35 kilobases of foreign DNA. ble obstacle to its practical implementation in basic research and Importantly, these gene-deleted vectors exhibit little to no toxicity human gene therapy applications. in vivo and can produce therapeutic amounts of various in In this report, we have incorporated the SB integration machinery animals6–12. Nonetheless, transgene expression from these non- into Ad vectors to combine the major advantages of each system. In replicative HD vectors still remains unstable in vivo, declining by as the process, we have identified a new limitation in the SB system, much as 95% over a period of one year11. Recent in vitro studies indi- namely that to transpose, SB transposons need to circularize. cate that the frequency of genomic integration of Ad vectors is quite Consequently, we have generated gene-deleted Ad vectors that low13,14, suggesting that transgene instability originates at least in release circular transposons from the linear Ad through the part from the loss of the linear, extrachromosomal HD genome in activities of the Flp/FRT recombination system. These high-capacity transduced cells. adeno-transposon vectors were characterized in mice and could One approach to circumvent this obstacle is to incorporate a DNA facilitate the somatic integration of virus-encoded transgenes into transposon into the Ad vector to maintain Ad-encoded transgenes in host cell chromosomes, resulting in greatly improved longevity of dividing cells through genomic integration. Although integration Ad-based gene expression in vivo. itself does not always guarantee long-term gene expression, it is the most reliable way to maintain genetic material within a cell over Results time. Currently, the only DNA elements reported to function within Analysis of Ad-based transposition in vitro and in vivo. To determine the context of mammalian cells are members of the Tc1/mariner whether the Sleeping Beauty transposon system could function in the family of transposable elements15. Recent studies indicate that the context of an adenovirus, we studied Ad-based transposition in vivo

Departments of Pediatrics and , Stanford University School of Medicine, Stanford, CA 94305-5208. *Corresponding author ([email protected]).

www.nature.com/naturebiotechnology • OCTOBER 2002 • VOLUME 20 • nature biotechnology 999 RESEARCH ARTICLE

A A

B

B

C

Figure 1. Analysis of transposition from E1/E3-deleted adenoviruses and linear transposon DNA. (A) Structures of the E1/E3-deleted transposition Figure 2. The adeno-transposon system. (A) Overview of the strategy to vectors Ad-SB and Ad-ThAAT for in vivo delivery of the Sleeping Beauty facilitate transposition from a helper-dependent adenovirus vector. An Ad http://www.nature.com/naturebiotechnology transposaseÐtransposon system. 5’ITR and 3’ITR, Left and right termini vector containing a Sleeping Beauty transposon remains of adenovirus type 5, respectively; RSV, Rous sarcoma virus long terminal extrachromosomal in SB-expressing cells because the transposase cannot repeat promoter; SB, Sleeping Beauty transposase; pA, polyadenylation efficiently act upon linear DNA structures. When the transposon is flanked signal; E1/E3, adenovirus type 5 early regions 1 and 3, respectively; IR, by a pair of Flp recognition target (FRT) sequences, the Ad vector Sleeping Beauty sequences; hAAT, human α1-antitrypsin undergoes conditional rearrangement in cells co-expressing the Flp cDNA. (B) Long-term Ad-based transposon expression in mice in the recombinase. This results in excision of the transposon from the Ad presence and absence of transposase. C57Bl/6 mice (n = 5 mice per genome and its circularization by Flp-mediated recombination. In contrast group) were injected through the tail vein with 2 × 109 transducing units to their linear counterparts, these circular elements actively undergo DNA- (TU) AdThAAT together with 6 × 109 TU of either AdSB () encoding the mediated transposition, resulting in stable insertion of the transposon into SB transposase or Ad-null () as a control. (C) Transposition efficiency host cell chromosomes. Prom, Mammalian promoter. (B) Structure of from circular and linear transposable elements in cultured mammalian helper-dependent transposition vectors. We used a two-vector approach to cells. Cells were transfected with transposon DNA and, when applicable, analyze transposition from gutless Ad vectors in vivo. This strategy a helper plasmid encoding either wild-type (SB) transposase or a employs the use of one vector to provide the Flp and SB recombinases catalytically inactive mutant (mSB) transposase as a control. Transfected that act upon a second Ad vector containing an SB donor transposon. cells were growth-selected in G418 for 14 days, fixed, stained, and HD-SB-Flp and HD-mSB-Flp both encode the enhanced Flp recombinase counted. The number of G418-resistant (G418r) colonies (mean ± s.d.) (Flpe) necessary for conditional vector rearrangement, but HD-mSB-Flp obtained after three independent transfections is shown. SV40, Simian cannot support transposition because of an inactivating © Group 2002 Nature Publishing virus promoter; neo, neomycin-phosphotransferase gene. Left panel, introduced into the transposase gene. The donor vectors HD-FRT-Tnori, transposition of a supercoiled neo-marked transposon in HeLa cells; HD-Tnori, HD-FRT-TLacZ, HD-TLacZ, and HD-FRT-ThFIX contain donor middle panel, transposition frequency of a linearized neo-marked element transposons encoding kanamycin (3.4 kb), β-galactosidase (5.5 kb), or in HeLa cells; right panel, transposition of a linear neo-marked transposon human Factor IX (4.5 kb), respectively.The intron-containing transgenes with (+IR) and without (ÐIR) flanking inverted repeats in HeLa cells present in the LacZ and hFIX constructs are initially split and thus remain constitutively expressing the SB transposase (HeLa-SB)49. inactive until Flp-mediated circularization restores the correct reading frame. HD-Tnori and HD-TLacZ are control vectors that lack flanking FRT sites and thus cannot undergo the Flp-mediated recombination necessary for both circle formation and SB-mediated transposition. IgκMAR, two using an experimental strategy previously validated with plasmid- copies of the immunoglobulin κ matrix attachment region; stuffer DNA, a 18 based vectors . To do this, we injected C57Bl/6 mice with an E1/E3- 16.2 kb fragment of alphoid repeat DNA from human 17; deficient adenovirus encoding a human α1-antitrypsin transposon MTH, mouse metallothionein I gene promoter; EF1α, human elongation (Ad-ThAAT) together with a second adenovirus encoding either factor 1α gene enhancer-promoter; Tn5, bacterial promoter; kan, transposase (Ad-SB) or no transgene (Ad-null) as a control (Fig. 1A). kanamycin resistance gene; ori, p15A bacterial ; cam, chloramphenicol resistance gene; LacZ, recombinant split β-galactosidase We then monitored serum hAAT levels over time as a general measure gene containing intron 1 of the human chorionic gonadotropin (hCG) of transposition in these mice. Results of these studies are shown in 6 gene inserted between nucleotides 1,761 and 1,762 of the LacZ coding Figure 1B and illustrate a lack of any significant improvement in the sequence; ApoE/HCR, 1.2 kb fragment containing the hepatocyte control long-term persistence of transposon expression in Ad-SB-infected region (HCR) from the apolipoprotein E (ApoE) gene; hAATp, human α1-antitrypsin gene promoter; hFIX, split human Factor IX (hFIX) cDNA mice. Because in vitro studies using an adenovirus encoding a containing 1.4 kb truncated intron A from the hFIX gene. neomycin transposon also did not indicate transposition (data not shown), SB transposition from conventional first-generation Ad vec- tors appears to be very inefficient, if it occurs at all. behind this inhibitory pathway remains under investigation, these Uncontrolled transposase production and transposon lineariza- results suggest that the absence of Ad-based transposition might be tion both significantly impair transposition frequencies. While due, at least in part, to overproduction of the transposase in investigating possible cis- and/or trans-acting factor(s) that might transduced cells. Alternatively, it is possible that the linear viral inhibit transposition from Ad vectors, we learned that SB overex- genomes are in some way resistant to efficient SB transposition, even pression significantly impairs plasmid-based transposition in vitro in the absence of any associated proteins. For instance, in vitro stud- and in vivo (unpublished observations). Although the mechanism ies with the Tc1 transposon have shown that the frequency of trans-

1000 nature biotechnology • VOLUME 20 • OCTOBER 2002 • www.nature.com/naturebiotechnology RESEARCH ARTICLE

A Figure 3. Conditional rearrangement and chromosomal integration of HD vectors in vivo following expression of the Flp and SB recombinases. (A) Overview of the genetic approach used to recover transposons from mouse chromosomes. C57Bl/6-scid mice (n = 2 mice per group) were injected into the tail vein with 1 × 109 TU of each virus. We induced transposase expression in a stepwise manner during the five weeks immediately following vector administration by addition of increasing concentrations of ZnSO4 to the drinking water. Groups 1Ð3 were included as controls to demonstrate that transposon integration from an Ad vector requires both circle formation and transposase activity. (B) Structure of the predicted excised circular episome. Numbers represent HindIII fragment sizes in kilobases. H, HindIII. (C) Transposon DNA analysis by HindIII digestion and ethidium bromide gel electrophoresis. Lanes 1Ð8, DNA from eight independent kanr/cams clones recovered from the liver DNA of group 4 (HD-FRT-Tnori + HD-SB-Flp) mice. Arrows indicate internal transposon-specific fragments (2.4 kb and 0.4 kb). Size markers are in kilobases. (D) Transposon insertion site sequences. Target site duplications are shown in bold uppercase, novel flanking sequences are in lowercase, and transposon sequences are denoted by the central shaded box. The origin and identity of transposon flanking sequences, shown to the right, were identified through mouse genome database searches. B C requirement for the formation of a circular intermediate through the Flp/FRT recombination system. We also produced two gene-deleted Ad vectors that constitutively express the enhanced Flp recombinase and inducibly express either wild-type transposase (HD-SB-Flp) or catalytically inactive mutant SB (HD-mSB-Flp) in the presence of http://www.nature.com/naturebiotechnology zinc. Based on their potential ability to control transposase produc- D tion and to release circular transposon molecules, these low- immunogenic vectors are significantly more suitable for SB transpo- sition than their first-generation predecessors. Recovery and sequence analysis of vector integration sites. To determine whether Sleeping Beauty could function in the context of these adeno-transposon vectors, we administered HD-Tnori and HD-FRT-Tnori to mouse livers and induced transposition following co-injection with HD-SB-Flp or HD-mSB-Flp as a control. We then identified transposition events using a plasmid recovery strategy that facilitates the cloning and sequencing of cellular–transposon junc- position from a supercoiled, circular donor transposon is reduced tion fragments (Fig. 3A). Interestingly, when we analyzed total liver 20-fold upon linearization20. Indeed, when we studied transposition DNA from mice injected with HD-FRT-Tnori and HD-SB-Flp of circular and linear neo-marked elements in cultured mammalian (group 4), we obtained that were predominantly cams and © Group 2002 Nature Publishing cells, we found that circular SB transposons underwent transposi- kanr (62% or 21/34), suggesting efficient transposon excision from tion much more readily than their linearized counterparts (Fig. 1C). the camr donor (Fig. 3B). Plasmid DNA analysis from eight of these Therefore, our findings collectively indicate that the production of a cams/kanr bacteria showed that each clone contained two bands (2.4 functional adeno-transposon vector may require the use of regulato- and 0.4 kilobases) corresponding to internal transposon sequences, ry components and additional recombination systems to facilitate as well as a variety of novel bands indicative of transposase-mediated the formation of circular intermediates in the cell. integration (Fig. 3C). Sequence analyses showed that each end of the Experimental strategy to facilitate Ad-based transposition. transposon was flanked by TA dinucleotides, followed by sequences Herein we devised a adeno-transposon system (Fig. 2A) that we that were different from the Ad vector sequences originally flanking believed might overcome some of the limitations identified in the the transposon in HD-FRT-Tnori (Fig. 3D). These findings are con- previous section. This approach utilizes gene-deleted Ad vectors sistent with a cut-and-paste mode of transposon integration23.When containing Flp recognition target (FRT) site sequences flanking we screened the sequences flanking the transposon in these cams the Ad-encoded transposon. Recent reports demonstrating its clones for homologies in both GenBank and the Ensembl mouse activity in the context of both adenoviruses21 and mammals22 genome database by a BLAST search, we found that integration indicate that expression of the thermostably enhanced Flp recom- occurred into multiple different chromosomes. Cellular sequences binase should result in excision of the FRT-flanked transposon contained in one clone possessed 98% homology to intron 2 of the DNA from the Ad genome. Importantly, DNA excision through mouse mSP100-rs1 gene on chromosome 1 (clone 2; positivity, the Flp/FRT recombination system also results in DNA circular- 496/505; GenBank accession no. AF176311), whereas another clone ization, and thus should provide an improved substrate for SB exactly matched intron 3 of the mouse Mypt1 gene on chromosome transposition in transduced cells. 10 (clone 7; positivity, 906/906; ENSMUSG no. 19907). All other To investigate the potential utility of this system for stable in vivo clones showed significant homologies to either mouse repetitive ele- gene transfer, we constructed Ad vectors encoding either SB trans- ments or to currently undefined mouse sequences. Because chromo- posons or the necessary recombinases (Fig. 2B). In all cases, we pack- somal deletions and rearrangements have been observed at the site of aged these constructs into helper-dependent vectors to avoid the recombinant adeno-associated virus (AAV) integration24,25, problems associated with leaky expression of viral proteins. we carefully analyzed the genomic sequences at the point of transpo- Furthermore, transposon vectors were produced with and without son insertion but could find no alteration of these sites other than FRT sites flanking the transposon cassette to investigate in vivo the the duplication of the TA target dinucleotide.

www.nature.com/naturebiotechnology • OCTOBER 2002 • VOLUME 20 • nature biotechnology 1001 RESEARCH ARTICLE

To facilitate the loss of nonintegrated trans- poson forms in the liver, we induced additional hepatocellular cell cycling after the PH by repeatedly injecting these with carbon tetrachloride (CCl4) over the next 13 weeks and then analyzed liver sections for β-galactosidase expression. Results of these studies showed the presence of rare isolated X-gal+ hepatocytes in each of the control groups (Fig. 4E–G), and a maximum of three X-gal+ hepatocytes within a single cluster. Interestingly, cross-sectional analysis of liver tissue from mice that received HD-FRT-TLacZ and HD-SB-Flp showed the presence of many large X-gal+ clusters, each consisting of ∼10–70 hepatocytes (Fig. 4H, I), the formation of which is consistent with an integration event occurring sometime before the induction of hepatocellular regeneration. Moreover, even though the number of these foci cannot be used to determine the exact inte- Figure 4. Adenovirus-based β-galactosidase expression in mouse liver before and after gration frequency in vivo because CCl4 induces hepatocellular regeneration. C57Bl/6-scid mice were injected into the tail vein with 1 × 109 TU each of HD-TLacZ and HD-SB-Flp (A, E), HD-FRT-TLacZ alone (B, F), HD-FRT-TLacZ and HD-mSB-Flp a nonuniform proliferative response, we have (C, G), or HD-FRT-TLacZ and HD-SB-Flp (D, H). Transposase expression was induced in a stepwise found that SB-expressing mice formed manner during the five weeks immediately following vector administration by addition of increasing ∼11-fold more blue foci (arbitrarily defined as http://www.nature.com/naturebiotechnology concentrations of ZnSO4 to the animal drinking water. Representative sections are shown from liver ≥3 clustered blue hepatocytes) compared with removed during a surgical two-thirds partial hepatectomy (PH) 5 weeks postinjection (p.i.) (AÐD; mice expressing the mutant transposase. n = 5 mice per group) and again after an additional 16 weeks (EÐI; n = 3 mice per group). All animals Therefore, these data collectively demonstrate received intraperitoneal injections of CCl4 at 8, 10, 12, 15, 17, and 19 weeks after vector administration to further promote hepatocellular cell cycling. Arrows denote the presence of that transposase-mediated integration can β-galactosidase+ foci, examples of which are shown in more detail in the six lower (I) panels. maintain the expression of Ad-encoded trans- × × × Magnifications: AÐD, original 100; EÐH, original 40; I, originals 200. Bars, 500uM. Bar in D genes during cell cycling in vivo. applies to A-D and bar in H applies to E-H. Genomic integration stabilizes expression of a Factor IX transposon in mice undergo- In contrast to the injection of HD-FRT-Tnori and HD-SB-Flp ade- ing hepatic cell cycling. To explore its potential as a therapeutic novirus vectors, liver DNA from control mice (groups 1–3) produced agent, we tested whether integration using this adeno-transposon very few cams/kanr bacterial colonies, and DNA from each of these system could maintain long-term production of human coagula- clones did not contain any transposon-specific fragments. The tion Factor IX (hFIX) in adult mice. To do this, we injected HD- absence of detectable transgene integration in the absence of either an FRT-ThFIX into C57Bl/6 mice together with either HD-SB-Flp or excised transposon circle (groups 1–2) or transposase activity (groups HD-mSB-Flp as a control, and then monitored the production of © Group 2002 Nature Publishing 2–3) indicates that transposition from an adenovirus requires the human Factor IX after inducing transposase expression for three coordinated activities of the Flp and SB recombinases. Collectively, weeks. Based on the structural design of the HD-FRT-ThFIX vec- these data demonstrate that Sleeping Beauty can stably integrate Ad- tor (see Fig. 2B), Factor IX expression will only originate from cir- encoded transgenes into the genome of mouse hepatocytes in vivo. cular molecules that are excised from the Ad genome by the Flp Flp-Dependent reporter gene activation and persistence in recombinase. To facilitate the loss of nonintegrated transposon vivo. We studied the activity of the adeno-transposon system in forms, we repeatedly induced hepatocellular regeneration in these mouse hepatocytes by analyzing the persistence of expression mice by first doing a surgical two-thirds partial hepatectomy and from a β-galactosidase transposon in the presence and absence of then by injecting CCl4. Approximately six months after vector the Flp and SB recombinases. Importantly, we have physically administration, human FIX levels were nearly undetectable (1–2 separated the 5′ and 3′ portions of the LacZ gene in the parental ng/ml) in control mice injected with HD-mSB-Flp (n = 5), where- vectors to ensure that proper mRNA splicing and β-galactosidase as mice receiving HD-SB-Flp still had stable human FIX levels expression only occurs when a circular DNA molecule is formed in ranging from 83 ng/ml to 186 ng/ml (n = 5) (Fig. 5). This corre- the cell (see Fig. 2B). We injected immune-deficient mice with an sponds to ∼135-fold more Factor IX in the long term in SB- Ad vector encoding an inactive LacZ transposon with (HD-FRT- expressing mice compared with control mice receiving nontrans- TLacZ) or without (HD-TLacZ) flanking FRT sites together with posable vectors and is within a range that has been previously either HD-SB-Flp or HD-mSB-Flp as a control. We then induced shown to be therapeutic in a mouse model of hemophilia B18. transposase expression during the next five weeks and analyzed Collectively, these results suggest that the frequency of chromoso- the liver for β-galactosidase expression after its isolation during a mal integration through the adeno-transposon system is sufficient surgical two-thirds partial hepatectomy (PH). Under these experi- to provide significantly improved transgene longevity from gene- mental conditions, ∼45% of mouse hepatocytes were found to deleted Ad vectors. contain transposon circles at this time point (Fig. 4C, D), com- Toward the development of a single integrating adeno-transpo- pared with just 1–2% in control mice that either did not receive an son vector. To investigate whether this integrative system could be FRT-containing reporter gene (Fig. 4A) or did not receive a Flp safely packaged into a single high-capacity adenovirus, we produced recombinase gene (Fig. 4B). Therefore, transposase and transpo- the Ad vector HD-FRT-TNC-SB, which encodes a neo-marked son circles can be efficiently co-delivered in vivo using a two- transposon and a promoterless transposase gene that becomes acti- component Ad-based system. vated upon circle formation (Fig. 6A). This strategy prevents trans-

1002 nature biotechnology • VOLUME 20 • OCTOBER 2002 • www.nature.com/naturebiotechnology RESEARCH ARTICLE

poson29 vector systems, our integrative transposon system does not require a reverse- step that can produce truncated forms of the transgene. Furthermore, both of these Ad-retro sys- tems require a two-stage infection process in vivo to stably integrate the transgene in neighboring, actively replicating cells. Importantly, although our system requires an additional circular- ization step, we have found that cell cycling is not required for SB- mediated transposition in the liver (unpublished), indicating that the adeno-transposon system can function directly within the postmitotic tissues that it infects. Finally, despite all these different integrative systems that have been developed, the only reported in vivo studies thus far have been with an E1-deleted Ad vector encod- ing a luciferase gene flanked by retroviral elements32. Surprisingly, Figure 5. In vivo human Factor IX persistence in actively dividing mouse this vector (AdLTR) was found to integrate in rat brain tissue in the livers via the adeno-transposon system. C57Bl/6 mice (n = 5 mice per absence of exogenous . This suggests that the integrative group) were injected into the tail vein with 5 × 108 TU of HD-FRT-ThFIX process used by this vector may involve recombination with × 8 together with 5 10 TU of either HD-SB-Flp ( ) or HD-mSB-Flp ( ) as a endogenous retroviral sequences, which could result in the inser- control. Transposase expression was induced in treated animals during the first three weeks immediately following vector administration by tion of neighboring viral sequences as well. In contrast, we utilized addition of drinking water containing ZnSO4 (25 mM final). Approximately a low-immunogenic gene-deleted vector and have identified SB three weeks postinjection, the zinc was removed from the water and a transposition as the mechanism of transgene persistence in mouse surgical two-thirds partial hepatectomy (PH) was done to stimulate liver. In addition, we also followed the long-term production (more hepatocellular regeneration and loss of episomal vectors. Arrows indicate than six months) of a clinically relevant reporter gene (Factor IX) when animals received intraperitoneal injections of CCl4 to further promote hepatic cell cycling. in an immune-competent host and did so under conditions of extensive hepatocellular regeneration so as to test stringently its http://www.nature.com/naturebiotechnology ability to maintain transgenes over time. Based on these encourag- posase overproduction from the parental vector and should ensure ing results, we have begun developing new adeno-transposon vec- the elimination of transposase sequences over time. Although tors for stable transgene insertion using a single gene-deleted vec- methods to control Flp activity during vector production remain in tor. With further development, it might eventually be possible to development, we expressed Flp in trans to study the poten- control the level of DNA integration by regulating production of tial of this transposase activation approach. To do this, we injected both the Flp and SB recombinases. mice with HD-FRT-TNC-SB and tested for transposon integration Although recent studies indicate that high-level SB expression is by a plasmid recovery strategy two months after systemic delivery of not cytotoxic (unpublished), we have taken precautions to ensure a plasmid encoding the Flp recombinase. Although there were controlled transposase production from our integrative vector sys- instances in which the transposon jumped into the Ad vector itself tem. Furthermore, despite high transposition rates (∼20%) in the (n = 3/10), the majority of the insertions (70%) were found to be spermatids of transgenic mice16, the actual frequency in non- within mouse chromosomal DNA (Fig. 6B). These results demon- germ-cell tissues may be significantly lower (≤10–6) based on quan- strate the feasibility of using recombinase-activated transposase titative estimates in both embryonic stem (ES) cells37 and cultured expression to achieve stable integration in vivo from a single high- human cells (unpublished). Importantly, the theoretical risk of © Group 2002 Nature Publishing capacity Ad vector. germ-line transmission using our system may be quite small consid- ering that spermatids are highly refractory to infection by aden- Discussion ovirus38,39. Moreover, with the current vector design, expression of Here we have identified circular transposon forms as the optimal the transposase is transient because of degradation or loss of the substrate for SB-mediated transposition and have developed effi- transposase gene during . With that said, application of cient ways to deliver these circles in vivo by combining a site-specific this integrative system to transgenic mice ubiquitously expressing recombination system with a gene-deleted adenovirus. Although it is the Coxsackie adenovirus receptor40 might actually provide new still unclear how transposon circularization promotes transposition, means to produce stable transgenics. studies in other transposon systems suggest that supercoiling of Based on the 29 transposition events we have isolated thus far, closed circular donor DNA may promote structural features in the in vivo integration from the SB system appears to be random. DNA that favor either interactions with proteins and/or the ordered Therefore, it might be useful in future studies to investigate what assembly and activation of the transposase–transposon complex26–28. effect, if any, insulators41 have on long-term transposon expression, Importantly, we have shown that this adeno-transposon system considering that a subset of the total events probably occur at tran- could integrate transgenes into the liver chromosomes of adult mice, scriptionally inactive loci. Alternatively, it might be useful to employ likely resulting in lifelong gene maintenance in vivo. These findings alternative recombination systems in this system, such as the φC31 demonstrate DNA-mediated transposition from an adenovirus and phage recombinase that mediates site-specific integration in cul- provide an integrative strategy for high-efficiency gene transfer and tured mammalian cells42,43. Although genomic integration represents stable transduction in vivo. an important step toward lifelong gene expression using Ad vectors, This vector system offers a number of important advantages over ongoing research should help determine what role, if any, immune existing chimeric vector systems previously developed to maintain responses to the Ad and/or the encoded transgenes have on adenovirus transgenes through integration into host cell chromo- long-term transgene persistence in vivo. somes29–34. First, in marked contrast to Ad-based vectors that rely In summary, our data indicate that integration through the on the AAV integration machinery for stable transduction30,31,35, adeno-transposon system can considerably reduce the need for vec- transposase-mediated integration does not induce chromosomal tor readministration. With their ability to transduce dividing and deletions and/or rearrangements at the site of transgene integra- nondividing cells efficiently and stably, these gene-deleted vectors tion24,25. Second, unlike the Ad–retrovirus33,34,36 and Ad–retrotrans- provide a potential new means to treat genetic diseases.

www.nature.com/naturebiotechnology • OCTOBER 2002 • VOLUME 20 • nature biotechnology 1003 RESEARCH ARTICLE

A as described45,46. We collected mouse serum from the retro-orbital plexus and analyzed it by an ELISA for total hAAT or hFIX antigens as described18. Generation of E1/E3-deleted adenoviruses. Details of the construction of the adenoviral shuttle used for production of AdSB, AdThAAT (encodes a 2.8 kb hAAT transposon), and AdTnori (encodes a 3.4 kb neomycin-marked transposon) are available upon request. The E1/E3- deleted Ad vectors AdSB, AdThAAT, and AdTnori were prepared following co-transfection with pJM17 as described47. We amplified viral plaques in 293 cells, analyzed adenoviral DNA by HindIII digestion, and purified to high titer by two-step CsCl ultracentrifugation48. These viruses were titered by optical density (OD) measuring according to the estimation 12 that one OD260 = 1 × 10 transduction units (TU)/ml. Production of high-capacity Ad vectors. The contents of each Ad vector are B described in Figure 2. Details of the construction of all helper-dependent (HD) adenovirus vectors are available upon request. HD adenovirus vectors were produced as described11. We used a plaque-forming assay to determine the level of helper-virus contamination in our viral preps (typically ≤0.2%) and titered each HD vector by quantitative Southern blot analysis using viral DNA isolated from virally infected HeLa cells. Cell culture studies. To analyze transposition from supercoiled and linear substrates, we transfected 5 × 105 HeLa cells with 1.5 µg of pCMV-SB or pCMV-mSB18 together with either 1.5 µg pTnori-Cam or 1.5 µg of a gel- purified NdeI/AccI Tnori fragment using Superfect (Qiagen,Valencia,CA). In a follow-up study, we transfected 5 × 105 HeLa-SB cells49 with 3 µg of either Figure 6. Transposition in vivo from a single Ad vector encoding NdeI/XbaI-digested pTnori (retains both IRs) or NdeI/BamHI-digested http://www.nature.com/naturebiotechnology transposon and Flp-activated transposase activities. (A) Experimental pTnori (removes the 3′ IR/DR structure). In all cases, we trypsinized cells strategy for controlled transposase expression from a single Ad-based 48 h later and then growth-selected in DMEM containing G418 (500 µg/ml) transposition vector. A linear adenoviral vector (HD-FRT-TNC-SB) r containing the entire Sleeping Beauty transposon system flanked by a for 14 days. We counted the number of G418 colonies on each plate after fix- pair of FRT sites forms a circular intermediate in the presence of the ing in formaldehyde and staining with methylene blue. Flp recombinase. Circularization of the transposition cassette results in cis-activation of transposase expression and genomic transposition Transposon recovery from the mouse genome. We injected C57Bl/6-scid mice × 9 from the adenoviral circle. In this configuration, the episome-encoded with 1 10 TU each of HD-Tnori and HD-SB-Flp, HD-FRT-Tnori alone, transposase gene is either degraded after transposition or lost over HD-FRT-Tnori and HD-mSB-Flp, HD-FRT-Tnori and HD-SB-Flp, induced time during cell division. CMV, Minimal cytomegalovirus core promoter. transposase expression for five weeks by addition of increasing concentrations (B) Transposon insertion site sequences from mice injected with a of ZnSO4 to the animal drinking water, and then killed animals for liver DNA single integrating adenovirus vector. C57Bl/6-scid mice (n = 3) were isolation. For the single integrating vector, we injected C57Bl/6-scid mice first × 9 injected with 2.4 10 TU of HD-FRT-TNC-SB through the tail vein. with 2 × 109 TU of HD-FRT-TNC-SB and then with 25 µg pOG-Flpe22 one µ One week later, animals were injected with 25 g pCMV-Flpe to week later to stimulate vector rearrangement and transposase expression. promote Flp-mediated vector rearrangement, and seven weeks later, total liver DNA was isolated and used to recover integrated These mice were killed seven weeks later for DNA analysis. In all cases, we used 18 transposons through a plasmid recovery strategy.Target site a plasmid rescue strategy previously described to isolate flanking genomic

© Group 2002 Nature Publishing duplications are shown in bold uppercase, novel flanking sequences sequences from mouse chromosomes, and screened them against the mouse are in lowercase, and transposon sequences are denoted by the genome database (www.ensembl.org/Mus_musculus/) to identify homolo- central shaded box. gous mouse sequences. β-galactosidase expression. Mouse liver tissue was frozen in OCT buffer on µ β Experimental protocol dry ice. We stained liver sections (10 m) for -galactosidase expression using 5-bromo-4-chloro-3-indolyl β-D-galactoside (X-gal) and counted X-gal+- Animal studies. We obtained six-week-old C57Bl/6 and C57Bl/6-scid mice hepatic nuclei from two to three liver lobes using a light microscope. from Jackson Laboratory (Bar Harbor, ME). Animals were treated according to the NIH Guidelines for Animal Care and the guidelines of Stanford University. Mice were anesthetized with isoflurane (Abbott Laboratories, Acknowledgments Abbott Park, IL) for surgical partial hepatectomy, which uniformly induces We thank K. Ohashi for helpful discussions, J. Chamberlain for providing the one to two rounds of hepatic cell division throughout the liver over a C7-Cre cells, and S. Dymecki for providing Flp/FRT-based plasmids. This work was supported by NIH grant DK49022 (M.A.K.). A.E. is the recipient of a period of three weeks. We diluted CCl4 in mineral oil and injected it intraperitoneally into the abdomen of mice. This drug kills the cells in which Judith Pool National Hemophilia Fellowship. J.G.M. was funded by grants from it is metabolized, resulting in a compensatory proliferative response in the the Carlsberg Foundation and the Danish Medical Research Council. surrounding areas of the liver44. For transposase induction, we added ZnSO 4 Competing interests statement (25 mM final) to the animal drinking water following vector administra- The authors declare that they have no competing financial interests. tions41. Adenoviruses were diluted in PBS and injected into the tail vein. Plasmid were delivered to mice by hydrodynamics-based transfection Received 20 May 2002; accepted 16 July 2002

1004 nature biotechnology • VOLUME 20 • OCTOBER 2002 • www.nature.com/naturebiotechnology RESEARCH ARTICLE

1. Kozarsky, K.F. & Wilson, J.M. Gene therapy: adenovirus vectors. Curr. Opin. 26. Chalmers, R., Guhathakurta, A., Benjamin, H. & Kleckner, N. IHF modulation of Genet. Dev. 3, 499Ð503 (1993). Tn10 transposition: sensory transduction of supercoiling status via a proposed 2. Yang, Y. et al. Cellular immunity to viral antigens limits E1-deleted adenoviruses protein/DNA molecular spring. Cell 93, 897Ð908 (1998). for gene therapy. Proc. Natl. Acad. Sci. USA 91, 4407Ð4411 (1994). 27. Crellin, P. & Chalmers, R. ProteinÐDNA contacts and conformational changes in 3. Yang, Y., Li, Q., Ertl, H.C. & Wilson, J.M. Cellular and humoral immune responses the Tn10 transpososome during assembly and activation for cleavage. EMBO J. to viral antigens create barriers to lung-directed gene therapy with recombinant 20, 3882Ð3891 (2001). adenoviruses. J. Virol. 69, 2004Ð2015 (1995). 28. Wang, Z. & Harshey, R.M. Crucial role for DNA supercoiling in Mu transposition: a 4. Yang, Y., Jooss, K.U., Su, Q., Ertl, H.C. & Wilson, J.M. Immune responses to viral kinetic study. Proc. Natl. Acad. Sci. USA 91, 699Ð703 (1994). antigens versus transgene product in the elimination of recombinant adenovirus- 29. Soifer, H. et al. Stable integration of transgenes delivered by a infected hepatocytes in vivo. Gene Ther. 3, 137Ð144 (1996). retrotransposonÐadenovirus hybrid vector. Hum. Gene Ther. 12, 1417Ð1428 5. Parks, R.J. et al. A helper-dependent adenovirus vector system: removal of helper (2001). virus by Cre-mediated excision of the viral packaging signal. Proc. Natl. Acad. Sci. 30. Lieber, A., Steinwaerder, D.S., Carlson, C.A. & Kay, M.A. Integrating adenovirus- USA 93, 13565Ð13570 (1996). adeno-associated virus hybrid vectors devoid of all viral genes. J. Virol. 73, 6. Schiedner, G. et al. Genomic DNA transfer with a high-capacity adenovirus vector 9314Ð9324 (1999). results in improved in vivo gene expression and decreased toxicity. Nat. Genet. 31. Recchia, A. et al. Site-specific integration mediated by a hybrid adenovirus/adeno- 18, 180Ð183 (1998). associated virus vector. Proc. Natl. Acad. Sci. USA 96, 2615Ð2620 (1999). 7. Balague, C. et al. Sustained high-level expression of full-length human Factor VIII 32. Zheng, C., Baum, B.J., Iadarola, M.J. & O’Connell, B.C. Genomic integration and and restoration of clotting activity in hemophilic mice using a minimal adenovirus gene expression by a modified adenoviral vector. Nat. Biotechnol. 18, 176Ð180 vector. Blood 95, 820Ð828 (2000). (2000). 8. Clemens, P.R. et al. In vivo muscle gene transfer of full-length dystrophin with an 33. Murphy, S.J., Chong, H., Bell, S., Diaz, R.M. & Vile, R.G. Novel integrating aden- adenoviral vector that lacks all viral genes. Gene Ther. 3, 965Ð972 (1996). oviral/retroviral hybrid vector for gene therapy. Hum. Gene Ther. 13, 745Ð760 9. Morral, N. et al. High doses of a helper-dependent adenoviral vector yield supra- (2002). physiological levels of α1-antitrypsin with negligible toxicity. Hum. Gene Ther. 9, 34. Soifer, H. et al. A novel, helper-dependent, adenovirusÐretrovirus hybrid vec- 2709Ð2716 (1998). tor: stable transduction by a two-stage mechanism. Mol. Ther. 5, 599Ð608 10. Morral, N. et al. Administration of helper-dependent adenoviral vectors and (2002). sequential delivery of different vector serotype for long-term liver-directed gene 35. Shayakhmetov, D.M. et al. A high-capacity, capsid-modified hybrid transfer in baboons. Proc. Natl. Acad. Sci. USA 96, 12816Ð12821 (1999). adenovirus/adeno-associated virus vector for stable transduction of human 11. Ehrhardt, A. & Kay, M.A. A new adenoviral helper-dependent vector results in hematopoietic cells. J. Virol. 76, 1135Ð1143 (2002). long-term therapeutic levels of human coagulation Factor IX at low doses in vivo. 36. Feng, M. et al. Stable in vivo gene transduction via a novel adenoviral/retroviral Blood 99, 1Ð8 (2002). chimeric vector. Nat. Biotechnol. 15, 866Ð870 (1997). 12. Kim, I.H., Jozkowicz, A., Piedra, P.A., Oka, K. & Chan, L. Lifetime correction of 37. Luo, G., Ivics, Z., Izsvak, Z. & Bradley, A. Chromosomal transposition of a genetic deficiency in mice with a single injection of helper-dependent adenoviral Tc1/mariner-like element in mouse embryonic stem cells. Proc. Natl. Acad. Sci. vector. Proc. Natl. Acad. Sci. USA 98, 13282Ð13287 (2001). USA 95, 10769Ð10773 (1998). http://www.nature.com/naturebiotechnology 13. Harui, A., Suzuki, S., Kochanek, S. & Mitani, K. Frequency and stability of chromo- 38. Peters, A.H. et al. Absence of germline infection in male mice following intraven- somal integration of adenovirus vectors. J. Virol. 73, 6141Ð6146 (1999). tricular injection of adenovirus. Mol. Ther. 4, 603Ð613 (2001). 14. Hillgenberg, M., Tonnies, H. & Strauss, M. Chromosomal integration pattern of a 39. Kanatsu-Shinohara, M. et al. Adenovirus-mediated gene delivery and in vitro helper-dependent minimal adenovirus vector with a selectable marker inserted microinsemination produce offspring from infertile male mice. Proc. Natl. Acad. into a 27.4-kilobase genomic stuffer. J. Virol. 75, 9896Ð9908 (2001). Sci. USA 99, 1383Ð1388 (2002). 15. Plasterk, R.H., Izsvak, Z. & Ivics, Z. Resident aliens: the Tc1/mariner superfamily 40. Tallone, T. et al. A mouse model for adenovirus gene delivery. Proc. Natl. Acad. of transposable elements. Trends Genet. 15, 326Ð332 (1999). Sci. USA 98, 7910Ð7915 (2001). 16. Fischer, S.E., Wienholds, E. & Plasterk, R.H. Regulated transposition of a fish trans- 41. Steinwaerder, D.S. & Lieber, A. Insulation from viral transcriptional regulatory ele- poson in the mouse germ line. Proc. Natl. Acad. Sci. USA 98, 6759Ð6764 (2001). ments improves inducible transgene expression from adenovirus vectors in vitro 17. Ivics, Z., Hackett, P.B., Plasterk, R.H. & Izsvak, Z. Molecular reconstruction of and in vivo. Gene Ther. 7, 556Ð567 (2000). Sleeping Beauty, a Tc1-like transposon from fish, and its transposition in human 42. Groth, A.C., Olivares, E.C., Thyagarajan, B. & Calos, M.P. A phage integrase cells. Cell 91, 501Ð510 (1997). directs efficient site-specific integration in human cells. Proc. Natl. Acad. Sci. USA 18. Yant, S.R. et al. Somatic integration and long-term transgene expression in nor- 97, 5995Ð6000 (2000). mal and haemophilic mice using a DNA transposon system. Nat. Genet. 25, 43. Thyagarajan, B., Olivares, E.C., Hollis, R.P., Ginsburg, D.S. & Calos, M.P. Site- 35Ð41 (2000). specific genomic integration in mammalian cells mediated by phage phiC31 inte- 19. Nakai, H. et al. Extrachromosomal recombinant adeno-associated virus vector grase. Mol. Cell. Biol. 21, 3926Ð3934 (2001). genomes are primarily responsible for stable liver transduction in vivo. J. Virol. 75, 44. Burr, A.W., Carpenter, M.R., Hines, J.E., Gullick, W.J. & Burt, A.D. Intrahepatic dis- 6969Ð6976 (2001). tribution of transforming growth factor-α (TGFα) during liver regeneration follow- 20. Vos, J.C., De Baere, I. & Plasterk, R.H. Transposase is the only protein ing carbon tetrachloride-induced necrosis. J. Pathol. 170, 95Ð100 (1993).

© Group 2002 Nature Publishing required for in vitro transposition of Tc1. Genes Dev. 10, 755Ð761 (1996). 45. Liu, F., Song, Y. & Liu, D. Hydrodynamics-based transfection in animals by sys- 21. Ng, P., Beauchamp, C., Evelegh, C., Parks, R. & Graham, F.L. Development of a temic administration of plasmid DNA. Gene Ther. 6, 1258Ð1266 (1999). FLP/frt system for generating helper-dependent adenoviral vectors. Mol. Ther. 3, 46. Zhang, G., Budker, V. & Wolff, J.A. High levels of foreign gene expression in hepa- 809Ð815 (2001). tocytes after tail vein injections of naked plasmid DNA. Hum. Gene Ther. 10, 22. Rodriguez, C.I. et al. High-efficiency deleter mice show that FLPe is an alternative 1735Ð1737 (1999). to Cre-loxP. Nat. Genet. 25, 139Ð140 (2000). 47. McGrory, W.J., Bautista, D.S. & Graham, F.L.A simple technique for the rescue of 23. Plasterk, R.H. The Tc1/mariner transposon family. Curr. Top. Microbiol. Immunol. early region I into infectious human adenovirus type 5. Virology 163, 204, 125Ð143 (1996). 614Ð617 (1988). 24. Miller, D.G., Rutledge, E.A. & Russell, D.W. Chromosomal effects of adeno-asso- 48. Kay, M.A., Graham, F., Leland, F.& Woo, S.L.Therapeutic serum concentrations of ciated virus vector integration. Nat. Genet. 30, 147Ð148 (2002). human α1-antitrypsin after adenoviral-mediated gene transfer into mouse hepa- 25. Nakai, H., Iwaki, Y., Kay, M.A. & Couto, L.B. Isolation of recombinant adeno-asso- tocytes. Hepatology 21, 815Ð819 (1995). ciated virus vectorÐcellular DNA junctions from mouse liver. J. Virol. 73, 49. Izsvak, Z., Ivics, Z. & Plasterk, R.H. Sleeping Beauty, a wide host-range transposon 5438Ð5447 (1999). vector for genetic transformation in vertebrates. J. Mol. Biol. 302, 93Ð102 (2000).

www.nature.com/naturebiotechnology • OCTOBER 2002 • VOLUME 20 • nature biotechnology 1005