φC31 Integrase Confers Genomic Integration and Long-Term Transgene Expression in Rat Retina

Abstract

purpose. Gene therapy has shown promise in animal models of retinal disease, with the most success achieved to date with viral vectors used for gene delivery. Viral vectors, however, have side effects and limitations and are difficult to manufacture. The present study was conducted in an attempt to develop a novel system for long-term gene transfer in rat retinal pigment epithelium (RPE), by using nonviral transfection methods for gene transfer and the integrase from the bacteriophage φC31 to confer long-term gene expression by means of genomic integration.

methods. Efficient nonviral delivery of plasmid DNA to rat RPE in vivo was achieved by using subretinal injection of plasmid DNA, followed by in situ electroporation. Gene delivery was evaluated by analyzing enhanced green fluorescent protein (eGFP) expression in frozen sections. In subsequent experiments, a plasmid expressing luciferase, with or without a plasmid encoding the φC31 integrase, was delivered to rat RPE. Luciferase expression was followed over time by using in vivo luciferase imaging.

results. Subretinal injection followed by electroporation yielded abundant transgene expression in the rat RPE. Expression was strongest 48 hours after delivery. In the absence of φC31 integrase, transgene expression declined to near-background levels within 3 to 4 weeks after treatment. By contrast, coinjection of the integrase plasmid led to long-term stable transgene expression throughout the 4.5-month test period. Eyes injected with φC31 integrase showed ∼85-fold higher long-term transgene expression in the retina than eyes without integrase.

conclusions. Subretinal injection of DNA followed by electroporation affords abundant transfer of plasmid DNA in rat RPE. φC31 integrase confers robust long-term transgene expression by mediating genomic integration of the transgene. These findings suggest that φC31 integrase may be a simple and effective tool for nonviral long-term gene transfer in the eye.

Diseases of retinal degeneration include retinitis pigmentosa (RP) and related disorders, such as age-related macular degeneration (ARMD). There is no treatment for RP. Patients undergo a slow degeneration of the photoreceptor layer of the retina that leads eventually to blindness. ^{1 2} There are many genetic causes of RP, including known and unknown genes (summarized on RetNet: http://www.sph.uth.tmc.edu/Retnet/ provided in the public domain by the University of Texas Houston Health Science Center, Houston, TX) whose loss of function afflicts either photoreceptor cells directly or the retinal pigment epithelium (RPE) that maintains them. Deficiencies in RPE cells lead to a loss of photoreceptor support or phagocytosis and are responsible for many diseases of retinal degeneration in humans, ^{3 4} mice, ^{5 6} and the well-studied Royal College of Surgeons (RCS) rat. ^{7 8}  

Genetic diseases of the retina can theoretically be addressed by gene therapy. Indeed, progress has been made in treating animal models of retinal degeneration with gene therapy. ² Most of this success has been demonstrated by using viral vectors for transgene delivery, especially in treating RPE deficiencies. Delivery of the RPE65 gene via adeno-associated virus (AAV) has attenuated vision loss in mice ^{9 10 11} and has even reversed blindness in dogs. ^{12 13 14} The RCS rat has shown functional rescue after delivery of Mertk by adenovirus ¹⁵ and AAV ¹⁶ and after delivery of pigment epithelium derived factor (PEDF) by lentivirus. ¹⁷  

Despite these successes, viral vectors have many drawbacks. They can be immunogenic and toxic and have limited carrying capacity. ¹⁸ Moreover, they are difficult and expensive to manufacture. Whereas nonintegrating viruses tend to lose expression over time, randomly integrating viruses can cause insertional mutagenesis. ¹⁹ For these reasons, nonviral approaches to gene therapy in the eye merit further investigation. Nonviral methods have been used in few gene therapy studies in the eye, due to difficulties in obtaining gene delivery and long-term gene expression. The published studies have demonstrated limited transfection of the retina ²⁰ and transient gene expression. ^{20 21}  

In contrast, work in which the unidirectional integrase from bacteriophage φC31 was used has resulted in long-term gene expression after nonviral gene delivery. ^{22 23 24 25} φC31 integrase mediates recombination between a plasmid bearing the attB sequence and specific sequences in mammalian genomes termed pseudo-attP sites, which have partial identity with the phage attP site (Fig. 1) . This recombination results in genomic integration of the attB plasmid and is associated with long-term, stable expression of integrated transgenes in human cells in culture. ²² Moreover, φC31 integrase has been successful in in vivo studies, conferring long-term nonviral gene expression in mouse liver ²³ and in human skin grafted onto immune-deficient mice. ^{24 25}  

Recently, Matsuda and Cepko ²⁶ showed efficient nonviral delivery of plasmid DNA to neonatal mouse and neonatal rat photoreceptors by means of in vivo electroporation. In the present study, we adapted their electroporation method to deliver DNA to adult rat RPE cells in vivo. We chose a rat model for these experiments because rat models of retinal degeneration are readily available, ⁷ and the larger size of rat eyes compared with mouse eyes allows for greater technical ease. We employed in vivo luciferase imaging to record transgene expression over time and used the φC31 integrase to achieve integration into the rat genome and long-term transgene expression. 

Methods

DNA Constructs

The φC31 integrase expression plasmid pCMV-Int ²² and the negative control plasmid pCS ²³ have been described. The plasmid pNBL2 contains a CMV immediate-early promoter driving firefly luciferase, the SV40 promoter driving neo, and the φC31 integration sequence attB. ²⁷ To clone pEGFPLuc-attB(+), first the construct pCMVsGFPLuc was obtained as a gift from Richard N. Day (University of Virginia Health Sciences System, Charlottesville, VA). pCMVsGFPLuc was digested with SspI, treated with shrimp alkaline phosphatase (SAP; Roche Diagnostics, Indianapolis, IN), and heat inactivated according to the manufacturer’s instructions to prevent the vector from self-ligating. The plasmid pTA-attB ²⁸ was digested with HincII, and the ∼300-bp attB fragment was ligated into the digested pCMVsGFPLuc, to form pCMVsGFPLuc-attB. pCMVsGFPLuc-attB was digested with PstI, and the 3.0-kb fragment was gel isolated (MinElute Gel Extraction Kit; Qiagen, Hilden, Germany) and treated with SAP. The 4.0-kb fragment that included the cytomegalovirus (CMV) immediate-early promoter and the EGFPLuc gene was excised from the plasmid pEGFPLuc (BD-Clontech, Palo Alto, CA) by NsiI. This fragment was gel isolated and ligated into the vector fragment from pCMVsGFPLuc-attB. This reaction yielded plasmids in both possible orientations that were named pEGFPLuc-attB(+) and pEGFPLuc-attB(−), the former being the plasmid bearing attB in the same orientation as the EGFPLuc gene. Plasmids in both orientations were functional for GFP and luciferase activity, and the plasmid pEGFPLuc-attB(+) was arbitrarily chosen for use in experiments. 

Cell Culture, Transfection, and Selection

Rat2 cells (American Type Culture Collection, Manassas, VA) were grown according to supplier instructions in DMEM with 4 mM l-glutamine, adjusted to contain 1.5 g/L sodium bicarbonate and 4.5 g/L glucose (American Type Culture Collection), supplemented with 10% fetal bovine serum. Cells were transfected in six-well cell culture plates at 60% to 80% confluence with a commercial reagent (Fugene6; Roche Biosciences, Palo Alto, CA) used according to the manufacturer’s instructions. We used a 3:1 transfection agent-to-DNA charge-to-mass ratio, 20 ng of pL-attB, and 980 ng of pCS or pCMV-Int. After 48 hours, each well was divided equally to 16 × 10-cm cell culture plates, and 24 hours later selective medium was added containing 400 to 700 μg/mL geneticin (G418, a neomycin analogue; Invitrogen, Carlsbad, CA). Growth in selective medium was continued for 10 days. Colonies were trypsinized, pooled, and grown for an additional 3 days. 

Plasmid Rescue and PCR

Genomic DNA was prepared with a kit (Blood and Cell Culture DNA Maxi Kit; Qiagen). Genomic DNA (10 μg) was digested with 40 units of EcoRV and 40 units of XmnI for 16 hours. Digested DNA was extracted from the reaction by using 25:24:1 phenol-isochloroform-isoamyl alcohol (Invitrogen) and then ethanol precipitated and resuspended in TE buffer. DNA was ligated at 16°C for 16 hours with 400 units of ligase (New England BioLabs, Beverly, MA) in 300 μL total reaction volume, to promote self-ligation. Ligated DNA was extracted and ethanol precipitated as just described, resuspended in 5 μL TE, transformed into electroporation-competent cells (ElectroMax DH10B cells; Invitrogen), and plated on Luria-Bertani medium containing ampicillin (100 μg/mL) to select for transformants containing pNBL2 and integration junctions. Plasmid DNA was prepared (QiaPrep Spin Miniprep Kit; Qiagen) after growth in liquid culture and digested with HincII to check whether the plasmid contained integration junctions. Plasmids that contained integration junctions were sent for sequencing (Elim Biopharmaceuticals, Hayward, CA) with the primers GSPL2 (5′-TCGACGATGTAGGTCACGGTCTCGAAG-3′) and attB-R (5′-GTCGACATGCCCGCCCTGACCG-3′) for attL and attR junctions, respectively. Sequences from junctions were aligned with attB, and sequences adjacent to the aligned portions were assigned to their locations in the rat genome by using BLAT. ^{29 30}  

Primers were designed to PCR amplify the integration junctions identified by plasmid rescue. Each round of the nested PCR contained one vector-specific primer and another primer in the rat genome. Genomic primers were designed from the rescued sequence or, if enough sequence was not obtained during plasmid rescue, from sequence in the rat genome database. Four genomic primers were designed for each identified pseudo-attP site, a first-round primer and a second-round primer for each orientation of plasmid integration (plus strand or forward orientation and minus strand or reverse orientation). All primer sequences are given in Table 1 . 

First-round PCR reactions used 250 to 350 ng genomic DNA and PCR beads (puReTaq Ready-To-Go; Amersham Biosciences, Amersham, UK), with 2 minutes at 94°C, 35 cycles of 94°C for 30 seconds, 56°C for 30 seconds, and 72°C for 30 seconds, followed by 10 minutes at 72°C. First-round PCR products were diluted 50-fold with water, and 2 μL was used as template in the second-round PCR. Second-round PCR conditions were identical with first-round PCR conditions, except only 30 cycles were used. Fifteen microliters of product from the second round of PCR was subjected to agarose gel electrophoresis. Bands corresponding to expected product size were excised, purified (QiaQuick PCR Purification Kit; Qiagen), and cloned into pCR2.1-TOPO (TOPO Cloning Kit; Invitrogen). Colonies containing inserts were sequenced with standard sequencing primers. Bona fide recombination junctions were identified by sequence that matched attB, including sequence beyond the primers used for amplification, followed by a crossover to rat genomic sequence within 5 to 10 bp of the TT core. PCR conditions were identical for Rat2 genomic DNA and sclera/choroid/RPE genomic DNA from experimental animals. 

Animal Studies

All studies were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Review Board. The experiments were performed with approximately 1-month-old Fisher-334 male rats. 

Subretinal Injection and In Vivo Electroporation

Subretinal injections were performed essentially as described, ³¹ except that animals were injected in the superior hemisphere. Electroporation was performed on animals immediately after injection. Electrodes with diameter of 7 mm (Tweezertrode; BTX, San Diego, CA) were briefly soaked in PBS and applied to each cornea, with the negative electrode on the injected eye and 14 mm between electrodes (Fig. 2) . Five 140-V pulses (constituting a field strength of 10 V/mm) with a duration of 100 ms were applied using an electroporator (ECM 830; BTX) with 950 ms between pulses. 

In Vivo Bioluminescence Imaging

Animals were anesthetized with 2.5% isoflurane and received intraperitoneal injections of luciferin substrate (150 mg luciferin per kg body weight). After 10 minutes, bioluminescence was measured (IVIS 100 imaging system; Xenogen, Alameda, CA). For bioluminescence imaging, we used the camera settings field of view 15, f1 f-stop, high-resolution binning, and exposure times of 10 to 300 seconds. Signal was quantified and analyzed on computer (Living Image, ver. 2.20 software; Xenogen). Animals were imaged to measure transgene expression 2, 5, 8, 18, 30, 60, 90, and 138 days after injection. After the final time point, animals were killed, and their eyes were analyzed with either fluorescence microscopy of frozen sections, histology, or PCR query. 

Preparation of Frozen Retinal Sections

For frozen sections, eyeballs were fixed overnight at 4°C in a solution of 2.5% glutaraldehyde and 2% paraformaldehyde. After removal of the lens, tissues were rinsed in 0.1 M phosphate buffer and cryoprotected with 30% sucrose in 0.1 M phosphate buffer, embedded in optimal cutting temperature (OCT) compound, frozen at −80°C, and sectioned at 10 μm. 

RPE/Choroid/Sclera Preparation and Genomic DNA Preparation

Eyeballs were removed and placed in PBS during dissection. An incision was made around the ora serrata, and the anterior part of the eye was removed. The neural retina was then removed, and the posterior eyecup containing the sclera, choroid, and RPE was placed in a tube, frozen in a bath of 95% ethanol and dry ice, and stored at −80°C. Samples were later thawed, and genomic DNA was made (QIAamp DNA Micro Kit; Qiagen). 

Retinal Histology

Samples were prepared for histologic analysis as described elsewhere. ¹⁵  

Results

Gene Transfer to Rat RPE In Vivo

We wanted to achieve gene transfer to RPE cells in the rat retina in vivo. In vivo electroporation has been used to transfect photoreceptors in neonatal rats, ²⁶ and we adapted this method to target RPE cells in adult rats. DNA was injected subretinally, electrodes were placed on either cornea, and current was applied (Fig. 2) . 

To monitor transgene expression in vivo, we constructed the plasmid pEGFPLuc-attB(+). This plasmid encoded both EGFP and luciferase activity in the fusion protein EGFPLuc, as well as the attB sequence, for φC31 integrase-mediated integration. We used this plasmid, because it allowed us to monitor gene expression easily in living animals by in vivo bioluminescence imaging ³² and in retinal sections by fluorescence microscopy. 

With subretinal injection alone, we observed low levels of transgene expression by using bioluminescence imaging (Fig. 3A , rat on the left). Injection with subsequent electroporation resulted in ∼1000-fold higher transgene expression (Fig. 3A , rat on the right, Fig. 2B ). To localize the signal to a particular cell type, eyes were enucleated and frozen sections made. By examining the sections with fluorescence microscopy, we found that most of the detectable GFP expression was localized to the RPE (Fig. 3C) . 

Effect of φC31 Integrase on Long-Term Transgene Expression in Rat RPE

We next evaluated whether φC31 integrase confers long-term transgene expression in the rat retina. Fourteen animals received subretinal injections with subsequent electroporation in situ. Six animals received 2.5 μg of the EGFPLuc-attB plasmid with 2.5 μg of the empty vector plasmid pCS, six received 2.5 μg of the EGFPLuc-attB plasmid with 2.5 μg of the integrase-bearing plasmid pCMV-Int, and two animals were sham injected with TE and used as a negative control. Two animals remained untreated and acted as an additional negative control. Animals were imaged for luciferase activity at various time intervals after treatment. Two days after treatment, all animals that received the EGFPLuc-attB plasmid showed similar levels of luciferase activity, averaging ∼2.2 × 10⁸ photons/s, >3300-fold above background levels. Whereas luciferase levels in the group that did not receive integrase dropped rapidly, reaching near background levels by day 18, co-injection of integrase led to an initial decline followed by a high level of stable luciferase expression at all time points tested (Fig. 4) . The long-term expression observed with φC31 integrase was ∼85-fold higher than that without integrase, and this difference was statistically significant (P < 0.005). At 4.5 months posttreatment, animals were killed, and their eyes were taken for further analysis. A subset of eyes were sectioned and analyzed by fluorescence microscopy for localization of the long-term transgene expression. GFP signal was localized to RPE cells, suggesting that the long-term expression observed from bioluminescent imaging in vivo was due to integrase activity in RPE cells (data not shown). 

Integration in the Rat Genome In Vitro

We next sought to identify the locations of φC31 integrase-mediated integration in the rat genome. Integration site identification from in vivo transfection of retina was expected to be difficult for two reasons. There would be relatively few cells bearing integration sites due to the small size of the RPE tissue. Also, extrachromosomal plasmid DNA that persisted in the tissue would interfere with rescue techniques. For these reasons, we first used the readily cultured Rat2 cell line to identify endogenous sequences that served as φC31 integrase-mediated integration sites in the rat genome. 

Rat2 cells were cotransfected with a plasmid bearing attB (pNBL2) and either a plasmid bearing integrase (pCSI) or carrier DNA (pCS). Stably transfected cells were selected by resistance to G418. Cotransfection with φC31 integrase resulted in 2.1-fold more G418-resistant colonies after 10 days (P < 0.01; Student’s t-test). This colony number increase is consistent with previous studies describing integration in the human and mouse genomes ²² and provides evidence for elevated integration frequency in the presence of φC31 integrase. Colonies from the φC31 integrase group (∼200 total) were pooled and used for preparation of genomic DNA. We used a plasmid-rescue method to identify genomic locations of φC31 integrase-mediated integration into the rat genome (Fig. 5A) . Twenty-six bacterial colonies were chosen for sequencing. Of these, eight colonies contained good sequence bearing a crossover between attB and rat genomic DNA and were termed pseudo-attP sites. The remainder returned bad sequence, due to a colony containing no recombination junction or more than one recombination junction (8/26), or sequence that did not meet our criteria for bona fide recombination junctions (10/26). It is likely that many of these events represent artifacts resulting from analysis of events as pools or from recombination in bacteria during plasmid rescue (Chalberg TW, Olivares EC, Portlock JL, Thyagarajan B, Kirby PJ, Calos MP, unpublished data, 2005). 

Three sites that were later discovered to be integration hotspots—rps1, -2, and -3 (rat pseudo-attP site)—are shown with their chromosomal locations in Figure 5B . These pseudo-attP sites were aligned and, consistent with previous findings, ^{22 23} were found to have a limited level of sequence identity with each other and with φC31 attP. Similar to the native φC31 attP site, ³³ rat pseudo-attP sites also featured inverted repeats (Fig. 5B) . 

Identification of Integration Hotspots in the Rat Genome

Previous studies with the φC31 integrase in the human and mouse genomes have indicated that the enzyme preferentially mediates integration at a relatively small number of genomic hotspots. ^{22 23 24 25} It therefore seemed likely that some of the pseudo-attP sites we identified might represent integration hotspots in the rat genome. Integration hotspots can be identified by PCR analysis in which primers are designed to detect the presence of the integration at a given genomic site. Because of the sensitivity of nested PCR, integration sites could potentially be detected even in scarce tissue samples, such as rat RPE. 

Accordingly, PCR primers were designed that would detect integration if it occurred at any of the pseudo-attP sites identified. Nested primers were designed for integration into both possible strands (plus and minus) flanking the pseudo-attP site. Nested PCR analysis with the genomic DNA derived from four independent pools of Rat2 colonies (each pool consisting of ∼200 colonies) indicated that three pseudo-attP sites were detected in multiple separate pools, defining hotspots for φC31 integrase-mediated integration in the rat genome (Fig. 6A) . Two sites, rps1 and -2, were present in all pools tested (4/4). Integration in both orientations is typical at φC31 pseudo-attP sites (Chalberg TW, Olivares EC, Portlock JL, Thyagarajan B, Kirby PJ, Calos MP, unpublished data, 2005), presumably because the sites are partially palindromic. 

Integration in Rat RPE In Vivo

Having developed a sensitive nested PCR assay for integration into the rat genome, we next tested whether integration had occurred in the retinas of animals that received φC31 integrase. Genomic DNA was prepared from the posterior eye of a subset of treated animals. The neural retina was dissected, and DNA was made from the remaining posterior eyecup. The resultant genomic DNA, therefore, included material from RPE as well as sclera and choroid tissue. Genomic DNA was assayed by PCR for the presence of one of the rat hotspots from Figure 6A , rps1. Whereas the noninjected contralateral eye and animals treated with the empty vector pCS were negative, the eye injected with φC31 integrase showed evidence of integration at rps1 (Fig. 6B) . This PCR band was cloned, and sequence analysis demonstrated covalent linkage of attB vector sequence and rat genomic DNA at rps1. Taken together with the evidence of integrase-mediated long-term expression in rat RPE, this result suggests that φC31 integrase mediates integration at the genomic location rps1 in rat RPE in vivo. 

Procedural Damage and Histology

Treated animals were examined periodically after the procedure. Damage was observable in a subset of animals from the treated and sham-injected groups. Damage included cataracts (9/14), inflammation (1/14), and small or inset eyes (8/14). Damage was highly variable among individuals, with some animals incurring no observable damage and others having pronounced cataracts, inflammation, or small eyes. 

Eyes of some animals were prepared for histologic analysis. Whereas some sections showed apparently normal retina, sections from other areas of the eye showed retinal damage, including compromised photoreceptor layers and scar tissue formation (data not shown). 

Subsequent experiments were performed to identify the source of procedural damage. Animals received no treatment, subretinal injection alone, electroporation alone, or subretinal injection plus electroporation. Damage was similar among treated groups and again was highly variable within groups. Based on these simple experiments, therefore, it was not possible to identify the sources of procedural damage. 

Discussion

We demonstrated that DNA can be transferred to RPE cells in the adult rat retina in vivo by electroporation. Transgene expression was readily detectable by both in vivo bioluminescence imaging and fluorescence microscopy. Furthermore, co-delivery of the φC31 integrase greatly prolonged the robust expression of the transgene. 

In vivo bioluminescence imaging offered a convenient and quantitative method for observing transgene expression in the retina over time. To our knowledge, this is the first report of bioluminescence imaging in the retina, extending to the eye the use of this powerful tool for observing gene expression over time. Whereas in vivo bioluminescence imaging resolved luciferase expression to the entire eye area, analysis of frozen sections with GFP provided the opportunity for further localization of the signal. An alternative approach for localization would be to use tissue-specific promoters driving luciferase to assay expression from specific cell types. By using the bioluminescence method, we found that transgene expression was transient without φC31 integrase, declining to near-background levels after ∼18 days. Because extrachromosomal plasmid DNA is known to persist in nondividing tissues, ^{34 35} this decrease in gene expression was most likely caused by gene silencing. 

Our central finding is that cotransfection of φC31 integrase with a plasmid bearing attB leads to long-term gene expression in rat RPE. Expression levels experienced an initial decrease in the group that received integrase. This is consistent with previous data and represents a transition from initial high levels arising from unintegrated plasmid copies to steady state expression levels resulting from integrated plasmid. Because not every expressed copy of the transgene becomes integrated, an initial decline in expression is expected. The group receiving integrase maintained high expression levels over 4.5 months and had expression levels ∼85-fold higher than animals that did not receive integrase. We identified integrase-mediated recombination hotspots in the rat genome and found evidence for integration at a particular genomic position only in eyes treated with integrase. This result is consistent with previous reports that φC31 integrase confers long-term transgene expression by means of genomic integration. ^{22 23 24 25}  

We observed ocular damage in some animals as a result of treatment. Damage was highly variable and was not correlated with co-injection of φC31 integrase. Because the damage was variable, it was difficult to attribute the cause to a particular part of the procedure. It is possible that damage was caused by the needle during surgery, as has been documented, ³⁶ or as a result of the electroporation. By analyzing many more animals, we could perhaps identify the source of damage caused by the procedure, but we believe that improving the procedure to minimize the damage is a more important step. 

Matsuda and Cepko ²⁶ reported little damage when using a similar technique to transfect the neural retina. ²⁶ There are several experimental differences that may account for our observations. First, we used 1-month-old rats, whereas Matsuda and Cepko used newborn animals. It is possible that animals suffer less damage at an earlier stage of development or have a higher healing capacity. Second, we used an anterior approach for subretinal injection, whereas they injected transsclerally. Third, although the field strength for both experiments was 10 V/mm, the absolute voltages used were different. Whereas Matsuda and Cepko used 100 V, we used 140 V, owing to the larger distance between eyes in adult animals. Fourth, whereas Cepko and Matsuda monitored near-term results (≤50 days), we examined animals for many months after treatment. 

Having established nonviral permanent retinal gene transfer by genomic integration, a logical next step would be to apply the technology to animal models of retinal degeneration. Because the techniques used in the study resulted in ocular damage, it is first necessary to refine the method of DNA delivery. 

Although this study was limited to rat retina, φC31 integrase is also known to provide long-term expression in mice ^{22 23} and in human cells, ^{22 24 25} and it is likely to provide a similar effect in retinal cells in these other species. Although the pattern of integration into the rat genome described herein is instructive in understanding the behavior of phage integrases in mammalian genomes, integration into the human genome is of ultimate importance in a therapeutic context. For this reason, we have limited our investigation of rat pseudo-attP sites to a small number of sites. We have not taken an exhaustive account of the rat genome integration profile or investigated potential disease associated with integration into any of these particular sites, because the different integration pattern in human cells could lead to different results and is of ultimately greater relevance. Pseudo-attP sites in the human genome have been described previously, ^{22 24} and we continue to investigate the pattern of integration into the human genome. These studies suggest that there are 10² to 10³ pseudo-attP sites in the human genome, and that more than 40% of integrations occur in the nine most frequently used hotspots (Chalberg TW, Olivares EC, Portlock JL, Thyagarajan B, Kirby PJ, Calos MP, unpublished data, 2005). 

Despite the many advantages of nonviral gene delivery, including increased safety, decreased toxicity, decreased cost, and large transgene size capacity, interest in nonviral gene therapy has been limited, in part owing to transient gene expression. The demonstration that φC31 integrase confers long-term expression in the retina serves to bolster the prospects of nonviral gene therapy and may stimulate further interest in novel nonviral delivery methods. φC31 integrase also offers the additional feature of a more limited number of integration sites than random integration, making it a desirable tool for chromosome engineering. Because φC31 integrase-mediated integration provides strong, stable gene expression while reducing the probability of insertional mutagenesis, the φC31 integrase system has great potential as a novel approach to nonviral eye gene therapy that may be translatable to the clinic. 

 Supported by National Institutes of Health Grants DK58187 and HL68112 (MPC) and National Eye Institute Grant EY11405 (DV) and the Foundation Fighting Blindness (DV). TWC is the recipient of a predoctoral fellowship from the Howard Hughes Medical Institute.

 Submitted for publication October 22, 2004; revised January 24 and February 18, 2005; accepted February 28, 2005.

 Disclosure: T.W. Chalberg, None; H.L. Genise, None; D. Vollrath, None; M.P. Calos, (P), Poetic Genetics LLC (C)

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Integration mediated by φC31 integrase, as it operates in nature and as it is used in mammalian cells. (A) In nature, the φC31 integrase recombines the attP site in the phage genome into the attB locus of the Streptomyces genome. (B) In the context of mammalian cells, φC31 integrase is used to mediate integration of a plasmid bearing attB and a gene of interest into endogenous genomic pseudo-attP sites.

Table 1.

 

Primer Sequences Used to Query Integration at Integration Hotspots in the Rat Genome

	First-Round Primers			Second-Round Primers
		Name	Sequence (5′–3′)	Name	Sequence (5′–3′)
Vector
All sites	attBF2	atgtaggtcacggtctcgaagc	attBF3	cgaagccgcggtgcg
Genomic
rps1-F	rps1-A1	tccttacccgtcatcgc	rps1-A2	gtttgggagccttctggtgc
rps1-R	rps1-B1	tacggagatggtggttgcactg	rps1-B2	ggggacaggttggagtacagg
rps2-F	rps2-A1	gcccctcccatctatctcta	rps2-A2	tcttgtagaaatggggtcatac
rps2-R	rps2-B1	acgtgcccatgacacacctt	rps2-B2	tgtatggcccaagcagt
rps3-F	rps3-A1	gccccagagcaacacctt	rps3-A2	aagctttgtctgtgccgacac
rps3-R	rps3-B1	catacccacagtggtcagacgg	rps3-B2	ctccatggcaccactgttgtcc

Figure 2.

 

Arrangement of electrodes for in vivo electroporation for RPE transfection. (A) Tweezer-type electrodes were placed on the corneal surface of either eye of a 1-month-old Sprague-Dawley rat. (B) The current was applied with the positive electrode contralateral to the injected eye. After prior injection of plasmid DNA into the subretinal space of the right eye, this arrangement electrophoresed the negatively-charged DNA toward the RPE layer (arrowheads).

Figure 3.

 

Electroporation enhances nonviral gene transfer to rat RPE cells in vivo. (A) After intraperitoneal administration of luciferin substrate, luciferase activity was quantified by a cooled bioluminescence camera in vivo. Bioluminescence data was obtained from 1- to 5-minute exposures, and these images were overlaid on photographic reference images. Shown are typical animals after subretinal injection of DNA alone (left) or subretinal injection of DNA followed by electroporation in situ (right). (B) Quantification of these results showed a ∼1000-fold increase in luciferase activity with electroporation versus subretinal injection alone. Error bars, SE. (C) Frozen sections of rat retinas treated with in vivo subretinal injection of DNA and electroporation showed localization of EGFP expression to the RPE layer. A subretinal detachment was evident in these samples taken 48 hours after subretinal injection.

Figure 4.

 

φC31 integrase conferred increased and prolonged transgene expression in rat RPE. (A) Groups of six rats received right-eye subretinal injections of pEGFPLuc-attB with carrier DNA (♦) or a plasmid bearing φC31 integrase (▪), followed by subsequent electroporation (5 × 100-ms pulses of 10 V/mm). Animals were observed over time by using in vivo bioluminescence imaging. Similar levels of luciferase expression were measured in both groups 48 hours after injection. In the absence of integrase, transgene expression declined to near-background levels (▴) within 3 to 4 weeks after treatment. Co-injection of the integrase plasmid led to long-term stable transgene expression at all time points tested, throughout the 4.5 month length of the experiment (**P < 0.01, ***P < 0.005; two-sided Wilcoxon rank-sum test). Animals receiving integrase showed ∼85-fold higher expression after 4.5 months than animals receiving carrier DNA. (B) In vivo bioluminescence imaging for typical animals 4.5 months after injection without (left) and with (right) co-injection of φC31 integrase.

Figure 5.

 

φC31 pseudo-attP sites from the rat genome. (A) Plasmid rescue was used to recover integration junctions from genomic DNA. After genomic DNA was isolated from pools of G418-resistant colonies, DNA was digested with restriction enzymes that did not cut within the plasmid. DNA was self-ligated to form circles that consisted of the plasmid vector, plus genomic DNA adjacent to the site of integration. The ligation mixture was transformed into bacteria and grown for use in sequencing of integration junctions. (B) Pseudo-attP sites in the rat genome were identified by plasmid rescue in Rat2 cells and show limited sequence identity to attP. The sequences of three rat pseudo-attP sites are shown, along with their chromosomal locations. Solid lines: inverted repeats flanking the area where recombination occurred.

Figure 6.

 

φC31 pseudo-attP sites are integration hotspots in the rat genome. (A) PCR assays were developed to detect integration into the rat genome at known pseudo-attP sites. The pseudo-sites rps1 and -2 were present in both orientations in all four independent pools tested. A third site, rps3, was present in three of four independent pools in one orientation and two of four pools in the other orientation. (B) Genomic DNA was made from the posterior eyecup of treated animals and probed for integration at rps1 in the forward orientation. Genomic DNA from Rat2 cells was used as an in vitro positive control (IVP). Neither the animal injected with empty vector (pCS) nor the uninjected contralateral (CL) eyes showed evidence of integration. Only the eye injected (I) with integrase (pCSI) showed a positive result for integration at this site, suggesting that φC31 integrase mediates integration in the rat retina in vivo.

The authors thank Douglas Yasumura, Matthew M. LaVail, Seth Blackshaw, Constance Cepko, Marilyn Masek, Mitra Alizadeh, Wei Feng, Timothy C. Doyle, and Christopher H. Contag for helpful advice during this work.