A Hyperactive Transposase Promotes Persistent Gene Transfer of a piggyBac DNA Transposon.

Nonviral vector systems are used increasingly in gene targeting and gene transfer applications. The piggyBac transposon represents an alternative integrating vector for in vivo gene transfer. We hypothesized that this system could achieve persistent gene transfer to the liver when administered systemically. We report that a novel hyperactive transposase generated higher transposition efficiency than a codon-optimized transposase in a human liver cell line. Hyperactive transposase-mediated reporter gene expression persisted at levels twice that of codon-optimized transposase in the livers of mice for the 6-month study. Of note, expression persisted in mice following partial hepatectomy, consistent with expression from an integrated transgene. We also used the hyperactive transposase to deliver the human α(1)-antitrypsin gene and achieved stable expression in serum. To determine the integration pattern of insertions, we performed large-scale mapping in human cells and recovered 60,685 unique hyperactive transposase-mediated insertions. We found that a hyperactive piggyBac transposase conferred an altered pattern of integration from that of insect piggyBac transposase, with a decreased frequency of integration near transcription start sites than previously reported. Our results support that the piggyBac transposon combined with the hyperactive transposase is an efficient integrating vector system for in vitro and in vivo applications.Molecular Therapy - Nucleic Acids (2012) 1, e50; doi:10.1038/mtna.2012.12; published online 16 October 2012.

Introduction

Integrating vector systems show great potential in treating genetic diseases. Viral vectors can achieve stable, persistent transgene expression and functional correction of disease. However, oncoretroviral integration is increased near transcriptional start sites (TSS), increasing the risk of genotoxicity by insertional mutagenesis.[1,2]

An alternative to viral gene transfer is use of nonviral integrating vectors.[3,4] The well-studied DNA transposon Sleeping Beauty (SB) supports sufficient gene transfer for functional correction in several therapeutic applications.[4,5,6,7,8] Moreover, the integration site preference of the SB transposon appears to be very different from retroviral vectors and exhibits no preferential integration into genes.[9,10] Modifications to the SB transposase created a hyperactive protein that catalyzes significantly higher transposition in mammalian cells.[11,12,13,14]

SB can achieve therapeutically relevant gene transfer to the liver,[6] supporting the development of transposons for therapeutic gene transfer. The piggyBac DNA transposon (PB), originally isolated from the cabbage looper moth Trichoplusia ni, encodes a transposase (iPB) that is highly active in mammalian cells and integrates at the conserved TTAA tetranucleotide sequence.[15,16,17,18] Furthermore, Bradley and Cadiñanos developed a murine codon-optimized PB transposase (mPB) that was 20 times more efficient than iPB in mouse embryonic stem cells.[19] Recently, a hyperactive form of PB transposase (hypPB) with seven amino acid substitutions (I30V, S103P, G165S, M282V, S509G, N570S, N538K on mPB) was developed and screened for transposition in yeast and subsequently in mouse embryonic stem cells. The hypPB increased relative integration frequency by nine times over mPB in mouse embryonic stem cells.[20] Hyperactive PBs present an alternative to SB for use in gene transfer investigations.

Here we report that iPB with these seven amino acid substitutions (iPB7) generated higher transposition efficiency than mPB transposase in in vitro transposition assays in a human liver cell line. Furthermore, following hydrodynamic delivery in mice, firefly luciferase expression driven by the murine albumin enhancer/human α1 antitrypsin promoter in a PB transposon delivered by mPB persisted at least 8 months (duration of the study). In addition, animals transfected with iPB7 showed higher expression than those transfected with mPB for at least 6 months (duration of study). The human α1 antitrypsin (hAAT) gene under the control of a liver-specific promoter delivered with iPB7 to murine liver achieved stable and persistent hAAT expression. Additionally, we carried out large-scale mapping studies of transposon insertion patterns in human cells. We found an altered pattern of integration from that reported previously for iPB,[17,21,22] including a reduced frequency of integration near transcription start sites.

Results

iPB7 efficiently transposes in human hepatoma cells

In mammalian cells, iPB with these seven amino acid substitutions (iPB7) transposes with greater efficiency than the mPB ( and , ). To determine the activity of iPB7 in a human liver cell line, Huh-7 cells were co-transfected with a PB transposon cassette carrying the puromycin-resistant gene driven by the cytomegalovirus promoter (pXLBacII_GFPPuro) and the iPB7 transposase (). The transposon cassette was transfected alone or with mPB as controls. iPB7 supported twice the transposition efficiency as mPB in Huh-7 cells (). These data indicate iPB7 is active in a human liver cell line.

iPB7 exhibits efficient transposition at various ratios

To delineate the effect of varying the abundance of transposon and transposase on gene transfer efficiency, we transfected Huh-7 cells with a fixed concentration of transposon donor plasmid pXLBacII_GFPPuro (100 ng) and varied the amount of iPB7 transposase plasmid to obtain ratios of 1:5, 1:2, 1:1, 2:1, and 5:1 of transposon:transposase plasmids, pXLBacII_GFPPuro:iPB7, respectively. As a control, the transposon cassette alone was transfected. The number of puromycin-resistant colonies at each of 1:5, 1:2, 1:1, and 5:1 transposon:transposase ratios was similar. However, when transposon was added at two times the concentration of iPB7, the number of resistant colonies increased significantly compared to the other ratios tested (P < 0.05) ().

piggyBac-mediated gene transfer persists in vivo

To determine the efficiency and persistence of transgene expression supported by the PB transposases in vivo, 6-week-old Balb/c mice (n = 4 per condition) were transfected hydrodynamically with 25 µg of PB transposon plasmid encoding the firefly luciferase reporter gene driven by a liver-specific promoter and mPB (1:1). As controls, animals received Lactated Ringer's solution or the luciferase transgene PB cassette and a catalytically inactive transposase (mPBD268L). Bioluminescence was monitored at 1-month intervals for 8 months. After 1 month, luciferase expression in mice transfected with mPB was 10 times that of expression in mice transfected with the inactive transposase (). mPB-mediated expression was stable and persisted for the duration of the study (8 months). This difference was significant (P < 0.001) across all time points when compared to mice receiving inactive transposase. After partial hepatectomy at 5 and 1/2 months, luciferase activity in both groups decreased. When comparing the mean expression levels before (red bracket) and after (black bracket) liver resection, there was no change in expression in animals receiving mPB (P = 0.25). In contrast, there was a significant decrease in mean luciferase expression levels after partial hepatectomy in animals receiving the inactive control transposase (P < 0.005). Luciferase signal decreased tenfold in animals receiving the inactive transposase at the 6-month time point. We note that while there is an apparent increase in luciferase expression following partial hepatectomy in animals receiving the inactive transposase, this increase is not significant. These results are consistent with expression from integrated transgenes in the mPB-treated animals.

iPB7 transposase promotes higher expression and persistence than mPB transposase in vivo

To compare the activity of iPB7 to that of mPB in vivo we transfected, via hydrodynamic tail vein injection, 6-week-old Balb/c mice with transposon plasmid containing the firefly luciferase reporter (described previously) and transposase plasmid encoding either iPB7 or mPB. Luciferase expression in animals that received iPB7 was about twice that in animals receiving mPB (), which correlates well with our in vitro data. Expression persisted at these levels for the duration of the experiment (6 months). These results indicate for the first time that transgene expression from integrated PB transposon delivered with iPB7 transposase is persistent and higher than that of mPB in vivo.

iPB7 can direct secreted protein expression in vivo

As a proof-of-principle experiment to test the utility of iPB7 in stably expressing a therapeutically relevant protein, we delivered a transposon plasmid carrying the human α1 antitrypsin (hAAT) cDNA under the control of a liver-specific promoter with the iPB7 transposase plasmid hydrodynamically (). As a negative control, we delivered Lactated Ringer's Solution alone. At periodic intervals, we analyzed serum hAAT concentration by enzyme-linked immunosorbent assay (ELISA). To determine the effect of PB dose in vivo, 5 µg () or 25 µg () of transposon plasmid was administered to mice along with varying amounts of iPB7 transposase plasmid (1:1, 2:1, and 5:1 transposon:transposase ratios). Overall, increasing the amount of transposon from 5 µg to 25 µg increased the expression level of hAAT by twofold (P = 0.008). The ratio did not significantly affect the amount of secreted protein ().

Large scale iPB7-mediated integration analysis in human cells

To determine the overall integration pattern of iPB7, we isolated DNA from puromycin-selected HeLa cells containing PB chromosomal insertions and used ligation-mediated PCR to amplify transposon-genome junctions. Deep sequencing was performed using the Solexa Genome Analyzer II (Illumina). We mapped a total of 60,685 unique integrations in HeLa cells. Approximately 52.4% of integrations mapped to RefSeq genes (). We observed 11 and 13 percent of integration within 5 kb upstream or downstream of TSS and CpG islands, respectively, while 33.3% were found within repeat elements. As a control, we carried out large-scale mapping with wild-type insect PB transposase (17,214 unique integrations). The integration frequencies within genes (50.4%) and near TSS (10.4%) were slightly less than those of iPB7. These differences, though small, were significant (Chi-square test P = 2.97 × 10−06 and 2.78 × 10−04, respectively). iPB integration frequencies near CpG islands (11.5%) and within repeat elements (35.0%) were also significantly different from those of iPB7 (P = 1.30 × 10−04 and 4.60 × 10−05, respectively).

It seemed likely that the differences between iPB and iPB7 integration patterns reflected the amino acid changes in iPB7. To explore this, we carried out large-scale integration mapping studies with transposases containing various combinations of the seven amino acid substitutions in iPB7 (). We evaluated a mutant with four N-terminal mutations (iPB-N4). Two of the N4 mutations (I30V and S103P) lie within a nonconserved amino terminal domain of unknown function while the third (G165S) is located in a conserved domain of unknown function. The fourth N-terminal substitution (M282V) occurs in the highly conserved DDE/DDD catalytic domain.[23,24] We also tested a mutant with three C-terminal mutations (iPB-C3). Two of the C3 mutations (N570S, N538K) occur within a cysteine-rich domain well conserved among piggyBac family members hypothesized to be a novel zinc finger domain.[23,25] We also investigated transposase mutants containing the four N-terminal mutations and each of the N538K and N570S mutations individually (iPB-N4 + 538 and iPB-N4 + 570, respectively). The results are summarized in . The integration patterns for each mutant were not significantly different with respect to iPB7 with the exception of iPB-C3. We recovered 62,814 integrations from cells treated with iPB-C3. The overall integration pattern for iPB-C3 differed significantly from that of iPB7 (P = 2.93 × 10−29). In particular, the frequencies of integration within RefSeq genes, and near TSS and CpG islands were significantly lower than those of iPB7 ().

Interestingly, when we relaxed the requirement for integrations recovered at TTAA sequences such that other genomic junction sequences were accepted, we recovered an additional 1,438 integrations (63,318 total) from the iPB7-treated HeLa cells. Thus, about 97.7% of the unique integration junctions were at TTAAs whereas 2.3% were in sequences other than TTAA (). Moreover, the most common non-TTAA junction sequences recovered were CTAA and TTTA, as well as their complementary sequences. Analysis of the iPB control and various mutants with the relaxed requirement yielded similar results to that of iPB7 (1.5–3% non-TTAA unique integrations). This indicates a preference for TTAA as the site of integration, but not an absolute requirement.

Discussion

Here we report in vivo gene transfer using iPB7 transposase containing seven amino acid substitutions that yield significantly increased integration efficiency when introduced onto the mammalian codon-optimized (mPB) background (hypPB).[20] Use of nonviral integrating vector systems presents an attractive option for achieving long-term therapeutic gene transfer. Currently, the most commonly used DNA transposon in therapeutic gene transfer research is SB.[4,5,7,8] Several groups reported the generation of SB hyperactive mutants that increased transposition efficiency in human cells over that of wild-type transposase.[11,12,13,14] The size limit of SB previously reported[11] can be increased about twofold by using the sandwich configuration presented by Zayed et al.[14] These studies and others have established the utility of DNA transposons as gene therapy vectors, leading the way to employ other DNA transposons such as piggyBac. Others previously reported that the PB transposase has greater activity than SB transposase,[18,19,26,27] which may prove advantageous in gene transfer applications. Recently, Doherty and colleagues demonstrated that hypPB had increased function compared to SB100X, currently the most active form of SB transposase.[26] Studies show the efficient integration of cassettes up to 100 kb using PB with little to no decrease in integration efficiency.[16,27,28] These differences as well as the persistent, stable in vivo expression supported by iPB7 indicate this transposon/transposase combination offers a useful alternative to other integrating vector systems for therapeutic applications and genome engineering.

Our mPB transposition activity in Huh-7 cells () agree with those previously reported for iPB activity in HepG2 cells.[29] These in vitro data reveal that iPB7 transposase significantly increased the frequency of transposition (~twofold). We note the overall transposition efficiency represents ~0.1% of total cells transfected. This is likely due to the poor transfection efficiency of Huh7 cells, but similar to that obtained in previous liver cell line studies.[29] These results correlate well with those in vivo in which we observed about twice the luciferase expression in animals receiving iPB7 compared to those receiving mPB. We investigated various ratios and doses of transposon:transposase plasmids in vitro and in vivo. With respect to native iPB transposase, there are conflicting reports concerning the overproduction inhibition phenomenon, in other words, decrease in integration with increased transposase expression.[17,18,30] Wu et al. and Grabundzija et al. reported a decrease in efficiency with increasing amounts of transposase[18,30] whereas Wilson and colleagues found no evidence of overproduction inhibition under similar conditions.[17] Of note, Wu and colleagues reported the most efficient ratio as 2:1 transposon:transposase when using 200 ng transposon. Our results agree with Wu et al. when using 100 ng of transposon. However, we observed no difference in gene transfer efficiency at any ratio in vivo. When we delivered the transgene hAAT with the iPB7 transposase, we demonstrated that a 25 µg dose of PB transposon plasmid achieved twice the plasma levels of hAAT protein compared to the 5 µg dose. Manipulating the various ratios of transposon plasmid to transposase plasmid did not significantly affect the secreted protein abundance (). These results correlate well with our in vitro data. Our observations highlight the utility of iPB7 for use in therapeutically relevant applications of a secreted protein.

After delivering a firefly luciferase cDNA to mouse hepatocytes using mPB and monitoring bioluminescence over 8 months, we observed stable expression as noted by others.[29,31] At ~60 days post-transfection, Saridey and colleagues reported about fivefold greater luciferase gene expression.[31] The lower expression levels in our studies are likely due to the use of a liver-specific promoter as opposed to a cytomegalovirus promoter.[32] Additionally, expression levels remained stable over time in animals receiving mPB, even after partial resection of the liver. As liver tissue regeneration in the mouse occurs in 3–7 days,[33] persistent reporter gene expression likely arises from integrated form of the transgene. This is important for achieving long-term therapeutic correction.

There are limited reports investigating the integration profile of PB transposase. Previous data suggest that PB has a much lower preference for integration into genes compared to retroviral and lentiviral vectors.[34,35,36] To address this, we employed large-scale mapping using deep sequencing technology with iPB7 and native iPB for comparison. We were surprised to observe that the frequency of integration near transcription start sites of both iPB7 and iPB was lower than that reported in the literature. Previous reports indicate that iPB integrates within 5 kb of TSS at a frequency of 16–20% depending on cell type.[17,21,22] While our data indicate a lower percentage than previously reported, we acknowledge that a preference for integration near TSS remains. We note that our study mapped integration events in HeLa cells while other reports used HeLa, HEK293T, and human T-cells.[17,21,22] These differences could also be due to the methods used to sequence. Previous studies used plasmid recovery whereas we used ligation-mediated PCR followed by Illumina sequencing. The restriction sites used in our studies may have conferred some bias due to size constraints of amplicons generated for the Illumina sequencing platform. These previous data analyses included smaller numbers of unique integrations, thus the greater number of insertions that we analyzed may provide greater confidence in the insertion site preferences. The reduced frequency of PB insertion near transcription start sites suggests a safer profile for therapeutic applications, although further studies are warranted to investigate the safety profile of PB.

We observed significant differences in overall integration patterns between iPB and iPB7. Based on this observation, we hypothesized that the amino acid substitutions might contribute to target DNA integration specificity. Thus, we compared various transposase mutants in large-scale mapping studies. Interestingly, we found a significant difference in integration pattern between the iPB7 and the iPB-C3 mutant, whereas there were no significant differences between the other mutants (iPB-N4, iPB-N4 + 538, iPB-N4 + 570) and iPB7. The integration frequencies near TSS and within RefSeq genes may suggest a safer profile compared to iPB. Additionally, this result supports the observation that the three amino acid residues G509, N538, and N570 in the C-terminus might contribute to integration site preference. Indeed, the presence of a RING/PHD zinc finger domain in the C-terminus indicates that this domain may be involved in chromatin interactions.[23,25] These changes are subtle yet statistically significant; it will be interesting to determine their biological relevance by confirming these changes in a different tissue/cell type or through subsequent biochemical assays in vitro. Interestingly, our data also reveal for the first time that PB transposase does not have an absolute requirement for TTAA sequences at the sites of integration. This differs from previous reports indicating that the TTAA tetranucleotide sequence is completely necessary for integration.[16] We note that we cannot rule out the possibility that the non-TTAA junction sequences present in our analysis were recovered from transposase-independent integrations. Although given that the majority of such sequences differed in only one nucleotide, this explanation seems unlikely. We propose that these sequences represent activity similar to the nonspecific “star” activity exhibited by some restriction endonucleases.

While PB offers advantages, the reported integration preference near transcription start sites poses a potential hurdle to overcome in adopting this vector system for therapeutic gene transfer applications. Even though it has not been shown, the potential for genotoxicity from integration of the PB transposon near proto-oncogenes remains. It would be beneficial to develop methods of directing integration to a “safe” site(s) in the genome. To that end, it is encouraging that the PB transposase can be modified with DNA binding proteins to potentially redirect integration.[37,38] The feasibility of achieving clinically relevant in vivo gene transfer with the iPB7 transposon system is unknown. Our data support the hypothesis that the activity of the iPB7 vector system is sufficiently robust to attain phenotypic correction. To deliver this plasmid-based system in vivo, it may be feasible to administer local hydrodynamic delivery to specific tissues and organs[39,40] or pursue ex vivo applications.[41] Additionally, packaging the transposon in a viral vector such as an integration-defective lentiviral vector is one option to increase the in vivo delivery efficiency,[42,43,44,45] and this can be addressed in future studies. We propose the iPB7 vector system warrants further development for therapeutic applications.

Materials and Methods

Plasmid constructs. mPB was constructed by inserting murine codon-optimized piggyBac transposase cDNA (Gen Bank accession number: EF587698)[19] into the EcoRI/XbaI sites of pcDNA3.1 (Invitrogen, Grand Island, NY). All primer sequences used in cloning are listed in . mPBD268L was constructed by mutating two nucleotides in the mPB cDNA via site-directed mutagenesis (Quikchange Site-directed Mutagenesis Kit; Stratagene, Santa Clara, CA) using mPB as template. iPB7, iPB-N4, iPB-C3, iPB-N4 + 538, and iPB N4 + 570 were constructed as described previously.[26] To clone pXLBacII_GFPPuro, pXLBacII-Luc, pXLBacII-hAAT a piggyBac transposon plasmid with a multiple cloning site (pXL-BacII-MCS) was constructed by inserting the overlapping oligos nondirectionally into the BclI/PmlI sites of the pXLBacII_cassette.[18] The pXLBacII_cassette carries the minimal PB sequences of 308 bp and 238 bp of the 5′ and 3′ ends, respectively.[46] pXLBacII_GFPPuro was constructed by inserting a BglII/AgeI fragment containing the cytomegalovirus promoter and GFP-Puromycin resistance gene from the pGIPZ vector (Invitrogen, Grand Island, NY) into the linearized pXLBacIIMCS. To construct pXLBacII-Luc, the EcoRV/XbaI fragment of a mammalian expression vector containing firefly luciferase cDNA driven by the murine albumin enhancer/human α1 antitrypsin hybrid promoter[47] was cloned into the EcoRV/XbaI-linearized pXLBacIIMCS. To construct pXLBacII-hAAT, NcoI and AgeI flanking restriction sites were introduced into the human α1 antitrypsin cDNA by PCR amplification from p-T/hAAT, subcloned into pCR2.1-TOPO-TA (Invitrogen, Grand Island, NY). Subsequently, luciferase cDNA was replaced with the NcoI/AgeI fragment of the hAAT cDNA subclone by insertion into the NcoI/AgeI vector fragment of pXLBacII-Luc. As described in Mitra et al.,[25] the yeast transposon donor plasmid contained a mini-piggyBac transposon composed of 328 bp of the piggyBac 5′ and 361 bp of the piggyBac 3′ ends flanking a kanamycin resistance gene. The transposase was supplied by a second plasmid containing the piggyBac transposase gene under the galactose-inducible control of the GALS promoter.[48]

In vitro transposition assay. 200 ng pXLBacII_GFPPuro alone or with 200 ng pcDNA3.1-mPB, pcDNA3.1-iPB7, or pcDNA3.1-mPBD268L was transfected into Huh-7 cells (2 × 105 cells) using TransIT-LT1 reagent (Mirus, Madison, WI) according to the manufacturer's instructions and cultured under puromycin selection (2 µg/ml) for 3 weeks; 100 ng pXLBacII_GFPPuro alone or with varying amounts of pcDNA3.1-iPB7 were transfected into Huh-7 cells (2 × 105 cells) using TransIT-LT1 reagent as above and cultured under puromycin selection (2 µg/ml) for 3 weeks. Investigated ratios (transposon:transposase) included 5:1, 2:1, 1:1, 1:2, and 1:5 in the Huh-7 cell line. Total DNA was normalized to 600 ng between experimental conditions. For each experiment, cells were fixed in 4% paraformaldehyde and counted following 1 hour methylene blue staining.

In vivo gene transfer. All mice for this study were housed at the University of Iowa Animal Care Facilities. All animal procedures were previously approved by the Institutional Animal Care and Use Committee (IACUC) and in accordance with National Institutes of Health guidelines.

Luciferase. 25 µg pXLBacII- Luc in a 1:1 ratio with either pcDNA3.1-mPB or pcDNA3.1-mPBD268L () in Lactated Ringer's (2 ml) was delivered hydrodynamically, as previously described[39] to 6-week-old Balb/c mice. 5 µg pXLBacII- Luc in a 1:1 ratio with either pcDNA3.1-mPB or pcDNA3.1-iPB7 () in Lactated Ringer's (2 ml) was delivered hydrodynamically, as previously described[39] to 6-week-old Balb/c mice. In brief, plasmid DNA was prepared in 2 ml of sterile Lactated Ringer's solution at room temperature. Mice (n = 4 in each group) were restrained, and the lateral tail vein was accessed using a 27 gauge needle (Becton Dickinson, Franklin Lakes, NJ). The solution was administrated over 5–8 seconds.

hAAT. Either 5 or 25 µg pXLBacII-hAAT transposon was given with varying quantities of pcDNA3.1-iPB7 to generate an in vivo dose–response curve using ratios of 5:1, 2:1, or 1:1 (triangle, square, or diamond respectfully) for each of the 5 µg and 25 µg amounts. DNA was prepared in 2 ml Lactated Ringer's solution and delivered hydrodynamically to 8- to 10-week-old C57Bl/6 mice (4–6 in each group) as described above.

In vivo bioluminescent imaging. Luciferase expression was monitored noninvasively using bioluminescence imaging. Animals were injected intraperitoneally with 100 µl/10 g of body weight of D-luciferin (Xenogen, Alemeda, CA) (15 mg/ml in phosphate-buffered saline) and anesthetized by isoflurane inhalation. Approximately 5 minutes after luciferin injection, the animals were imaged using a Xenogen IVIS200 CCD camera. Imaging data were analyzed, and signal intensity was quantified using Xenogen Living Image software. Expression levels were normalized by subtracting those from naive (untreated) mice.

Partial hepatectomy. To verify luciferase expression was due to integrated transgenes, partial hepatectomies were performed 5 and 1/2 months after PB luciferase injection. Mice were injected subcutaneously with 25 cc/kg mouse normal saline and 2.5 µg/gm mouse Flunixin prior to start of the procedure. Mice were anesthetized with 2.5–4% isoflurane throughout the procedure. The skin was sterilized with 1% povidone-iodine followed by alcohol. Approximately 50% of the liver including the left lateral and median lobes was resected after a midventral laparotomy. Incisions were closed with 4–0 silk sutures. Four total drops/mouse of bupivacaine were placed on the muscle and skin after closure. Animal mortality following partial hepatectomy was 0%.

In vivo detection of hAAT. Serum hAAT concentrations were measured using ELISA. Microtiter plates were coated with 100 µl of goat anti-hAAT (MP Biomedical, Solon, CA) antibody at a 1/200 dilution in Voller's Buffer and incubated overnight at 4 °C. Plates were blocked in 100 µl of a blocking buffer (1% bovine serum albumin solution) for 2 hours at room temperature. Duplicate standard curves and 100 µl diluted plasma samples (diluted 1:4 in ELISA-phosphate-buffered saline) were incubated for 2 hours at 37 °C. Following sample incubation, plates were incubated with peroxidase-conjugated goat anti-hAAT antibody (Abcam, Cambridge, MA) at a 1/10,000 dilution for 2 hours at 37 °C. The peroxidase activity was expressed by incubation with 100 µl tetramethyl benzidine peroxidase solution. After 3 minutes of development time, the reaction was stopped with 2N H2SO4. The color generated was quantified using a microplate reader (Molecular Devices, Sunnyvale, CA) at 450 nm. Plates were washed in between each reaction with phosphate-buffered saline.

Integration site recovery and illumina sequencing. Integration sites were recovered as described.[49] Briefly, HeLa cells (5 × 106) were transfected with 10 µg pXLBacIIGFP_Puro transposon and 2 µg each transposase and selected with puromycin (0.5 µg/ml) for 3 weeks. Genomic DNA from three separate transfections was extracted from the integration library using the DNeasy tissue kit (Qiagen, Valencia, CA). Pooled DNA (2 µg) was digested overnight with ApoI or BstYI at 50 °C and 60 °C, respectively; DNA fragments were purified with the QIAquick PCR purification kit (Qiagen, Valencia, CA) and ligated to ApoI and BstYI linkers overnight at 16 °C. Nested PCR was carried out under stringent conditions using transposon end-specific primers complementary to transposon sequences and linker-specific primers complementary to the DNA linker sequence. Primers and linkers used in this study are listed in . DNA barcodes were included in the second-round PCR primers in order to track sample origin (). The PCR products were gel purified, pooled, and sequenced using the Illumina HiSeq2000 sequencing platform.

Sequence analysis and annotation

Read identification and trimming. Sequences were analyzed using a modified version of the Integration Analysis System (IAS).[50] The pipeline was run independently on each lane of sequencing with the following inputs, FASTA file with sequencing reads and a barcode file that contained the barcodes, sample IDs, and the expected flanking sequences. The sequences were split by barcode into sample-specific files. Crossmatch (P. Green, unpublished data) identifies the presence and location of the flanking sequences allowing for them to be trimmed leaving the genomic sequence present.

Mapping and sequence analysis. The samples were mapped back to the human genome reference assembly (GRCh37/HG19) using Bowtie.[51] The methods for mapping and annotating are presented in Brett et al. [50] with the following addition. The final annotated sites were collapsed down ± 5 bp from the most frequently seen sites, removing errors in mapping or incorrect trimming of the adaptor sequence. To determine if an integration occurred at a non-TTAA, for each integration site we evaluated the most frequently observed non-TTAA sequence and required that sequence (1) be present in greater than 50% of the reads and (2) match the reference genome.

Comparison of integration rates in genomic features. The location of the genomic features was determined using UCSC's (University of California, Santa Cruz) table browser. Transcription start sites and transcribed regions, CpG islands, and repetitive elements were identified from the refGene, cpgIslandExt, and rmsk tables, respectively (http://genome.ucsc.edu).[52] The list of cancer genes came from the Retrovirus Tagged Cancer Gene Database (RTCGD, http://RTCGD.ncifcrf.gov).[53] The genomic location of these cancer genes was determined using the refFlat table to determine the transcribed region of the genes. The CpG Islands and transcription start sites were widened to include 5 kb to either side. For each feature, overlapping positions were merged to avoid double counting sites.

Each category of features was evaluated to determine if the rate of integrations differed from expected. The expected rated of integration was determined based upon the relative number of TTAA sites found within the category. The number of integrations in each category was determined using BEDTools[54] by intersecting the set of integrations with the genomic loci for the category, as defined above. Only integration sites containing a TTAA were used. A Chi-squared test was used to determine if the integration rate of the samples differed from the expected integration rate. The samples were also compared to iPB7 using a Chi-squared test. A Bonferroni corrected P value of 0.00083 was used for significance.

Figure S1. Relative excision frequency of iPB7. Table S1. Primer sequences used in cloning. Table S2. Primers and oligos used for integration library generation. Table S3. Barcodes used in library generation for Illumina sequencing. Materials and Methods.

Table 1

Integration frequencies of PB transposase at TTAA target sites in HeLa cells (%)