Genomic insertion of lentiviral DNA circles directed by the yeast Flp recombinase

Integration of foreign DNA plays a pivotal role in genetic engineering, animal transgenesis, and therapeutic gene transfer. Lentiviral vectors (LV) efficiently insert genetic cargo in both dividing and non-dividing cells and have attracted, therefore, major attention as a gene transfer tool. During lentiviral transduction, linear vector DNA is associated with viral and cellular proteins in the preintegration complex (PIC) which is transported across the nuclear membrane and facilitates tethering of viral genomic DNA to chromatin. As part of the PIC, cellular LEDGF/p75 is thought to anchor viral DNA on chromatin, allowing the viral integrase to insert viral DNA in a fashion that supposedly favors integration in actively transcribed genes [1–3]. Great interest is attracted to ways of altering the lentiviral integration profile, allowing gene insertion in predetermined and safe vector landing sites. It remains unknown, however, whether lentiviral integration can be directed by alternative integration machineries, despite the integrity of the PIC and the exquisite involvement of cellular factors in nuclear entry and chromatin association.

During lentivirus infection circular forms of the viral genomic DNA are generated by either non-homologous end joining (NHEJ) of the full-length linear viral DNA (creating 2-LTR circles) [4], or by homologous recombination between the two LTRs of the episomal viral DNA (creating 1-LTR circles) [5]. Such episomal DNA circles have traditionally been considered dead-end products of reverse transcription [5] and are eventually lost together with episomal linear vector forms as a result of host cell division. However, integration-defective lentiviral (IDLV) vectors carrying an inactive integrase protein that abolish the normal viral integration pathway have recently emerged as novel efficient gene carriers, facilitating high levels of transient expression from linear and circular DNA forms [6–9]. Increased persistence of transgene expression in dividing cells has been demonstrated by allowing circles harboring the simian virus 40 (SV40) origin of replication to replicate episomally in cells containing the SV40 large T antigen [9]. Moreover, non-integrating lentiviral vectors have been shown to facilitate stable in vivo therapeutic levels of transgene expression in non-dividing neuronal cells [8] and muscle [6]. Recently, the episomal nature of the integration-defective vectors was further exploited as a template source for high-efficient gene correction by homologous recombination in human cell lines and in human embryonic stem cells [10, 11].

High stability and accessibility of DNA circles by cellular proteins have led us to suggest that stable circular DNA intermediates can be engineered to act as putative substrates for gene insertion by nonviral gene-inserting proteins. We test this hypothesis here by using a model system based on the actions of the site-directed Flp recombinase and demonstrate for the first time that LV-derived DNA circles can serve as a substrate for gene insertion facilitated by exogenous nonviral recombinases. Our findings provide evidence that trans-acting integrases are able to gain access to and insert transduced lentiviral DNA with possible application in (i) viral vector manipulation for improved viral gene transfer and (ii) Flp-based cell engineering methods focusing on hard-to-transfect cells and/or creation of site-directed gene insertions that do not contain bacterial remnants of plasmid DNA.

To provide proof-of-principle that trans-acting nonviral recombinases are able to gain access to lentivirally delivered substrates and facilitate their insertion, we designed an LV-derived hybrid integration system based upon the integrating properties of the yeast Flp recombinase. First, we constructed an LV substrate vector, pLV/FRT-hygro, containing an ATG-deficient FRT-hygro fusion cassette in the context of a self-inactivating (SIN) LV vector (Figure 1A). The Flp recombination target (FRT) sequence is recognized by the Flp recombinase which mediates recombination between two identical FRT sites. By including the FRT sequence in this vector, we reasoned that LV DNA circles generated during vector transduction would serve as substrates for Flp-dependent site-directed insertion of viral DNA into FRT sites engineered into the genome of transduced cells. Integration of the ATG-deficient FRT-hygro fusion gene into an engineered FRT site flanked upstream by a promoter and an ATG sequence would regenerate a functional hygromycin B expression cassette, allowing selection of cells containing site-directed vector insertion (Figure 1B). Hence, by using a site-directed Flp-based approach, we were able to score circle insertions selectively but also suspected a low circle insertion rate due to strong limitations on available target sites in the transduced cells.

The conditions for the Flp-based integration machinery were optimized by packaging vector RNA in virus particles carrying an inactive viral integrase protein harboring the class I D64V mutation [15]. By blocking the normal viral integration machinery, we were able to increase the copy number of 2-LTR circles from 3.8 to 22 circles per cell (Figure 2) corresponding to a 3.9-fold increase of circular DNA (correlated to the total amount of HIV-1 DNA) available for Flp-based recombination in transduced cells (Figure 2). To rule out possible overestimation of the total vector DNA due to plasmid contamination, we carried out a quantitative PCR analysis using a plasmid-specific primer set amplifying an amplicon within the Amp resistance gene in the plasmid backbone. Although within the detection limit of the assay, the level of contaminating plasmids was neglectable relative to measured copy numbers of total vector DNA (data not shown) and considered, therefore, to be without notable influence on measurements of total vector DNA. Importantly, the integrase mutation did not influence production or transduction capacity of the viral particles, since p24 values of the virus preparations and the total amount of episomal vector DNA present in the transduced cells were comparable between vector preparations carrying active and inactive viral integrase (Figure 2). In accordance with previous reports [6, 16] the titer of integrase-deficient vectors was reduced approximately 1000-fold (from ~10⁸ to ~10⁵ CFU/ml) compared to vectors carrying the active integrase (data not shown).

To tag chromosomal DNA with recognition sequences for the Flp recombinase, we generated a Sleeping Beauty [14] (SB) transposon-derived docking vector, pSBT/RSV-FGIP, containing the Flp recombination target (FRT) sequence as part of a new active fusion variant of the eGFP reporter gene (Figure 1A). The eGFP fusion gene, containing an FRT sequence between the start codon and the remaining part of the eGFP coding sequence, was flanked downstream by an IRES element and the puromycin resistance gene. The docking SB transposon was inserted into the genomic DNA of HEK-293 cells by co-transfection of pSBT/RSV-FGIP with pCMV-SB, a plasmid encoding the SB transposase. To provide optimal conditions for subsequent Flp-based gene insertion, we utilized conditions which in our hands consistently allow insertion of >1 vector in transfected HEK-293 cells. The rationale was to create a cell line tagged with at least two docking sites. Among several eGFP-positive HEK-293 cell lines that were generated by puromycin selection, one cell line, HEK/FGIP1, was utilized for further studies. This particular cell line was found to contain at least 3 vector insertions, as determined by Southern blot analysis (data not shown). To confirm functionality of the FRT sequence in the context of the eGFP fusion gene, HEK/FGIP1 cells were co-transfected with 2 μg plasmid carrying an ATG-deficient FRT-hygro cassette (pLV/FRT-hygro) and 10 μg plasmid encoding Flp recombinase. Subsequent hygromycin B selection demonstrated efficient insertion into the engineered FRT site (Figure 3A). Analysis of the transfected HEK-SB/FGIP1 cells by quantitative PCR (data not shown) showed that each cell on average contained 1.3 × 10⁴ ± 2.3 × 10³ FRT-tagged donor plasmids 24 hours after transfection, leading to an estimated insertion rate of 2.8 × 10^-7 Flp-based insertions per single copy of transfected FRT-hygro-containing plasmid in the culture. Notably, co-transfection of FRT-hygro substrate plasmid with a negative control plasmid did not result in colony formation, demonstrating the absolute lack of false positives in the system. As expected by the presence of >1 docking insertion in this cell line, eGFP expression was not turned off in hygromycin B resistant cells due to the fact that not all docking sites had been targeted by Flp recombination. However, subsequent PCR and sequence analysis of multiple hygromycin B-resistant clones confirmed correct Flp-based plasmid insertion in the eGFP gene of the integrated docking vector (data not shown). In summary, these data provide evidence that the Flp docking construction containing the novel FRT-eGFP fusion gene is functional in context of an integrated SB vector.

During lentiviral transduction double-stranded viral DNA is transported through the nuclear membrane as part of the PIC consisting of both viral and cellular proteins. Attempts to access the viral DNA by nonviral recombinases may therefore be hampered by the reduced DNA accessibility within the context of the PIC. To demonstrate that Flp indeed can gain access to the circular viral DNA and facilitate genomic insertion of LV circular substrates, we transfected the HEK/FGIP1 cell line with a Flp expression plasmid one day prior to transduction with the integration-defective (ID) LV/FRT-hygro vector (see Figure 1A). After hygromycin B selection 97 ± 15 resistant colonies were obtained (Figure 3B), whereas resistant colonies could not be detected in cells transfected with an empty control plasmid prior to IDLV/FRT-hygro transduction.

To examine whether IDLVs, co-transduced with DNA circle donor IDLVs, could serve as a source of Flp recombinase we created an LV SIN vector, pLV/PGK-Flpx9, containing a PGK-driven Flp gene (Figure 1A). First, HEK/FGIP1 cells were transduced with IDLV/PGK-Flp and on the following day transfected with an FRT-tagged plasmid substrate (pLV/FRT-hygro). After two weeks of hygromycin B selection, we detected 35 ± 9 drug-resistant colonies as a result of Flp-mediated plasmid insertion (Figure 3C), whereas transduction with a viral vector, IDLV/PGK-eGFP, lacking the Flp expression cassette did not result in any colony formation. These findings demonstrated that Flp recombinase, transiently expressed from integration-defective LV vectors, was able to confer substrate recombination and site-specific gene insertion. We finally co-transduced the HEK/FGIP1 cell line with IDLV/FRT-hygro and IDLV/PGK-Flp and subsequently selected transduced cells for hygromycin B resistance. In this setup, we obtained 58 ± 6 colonies per co-transduction (Figure 3D). To examine whether the efficiency of Flp-directed gene insertion varied among clones tagged with the docking vector, we generated five additional FRT-tagged HEK clones (HEK/FGIP2-6). All these cell lines demonstrated levels of eGFP expression that were comparable with the expression of eGFP in HEK/FGIP1 as measured by FACS analysis (data not shown). By co-transduction of these cell lines with IDLV/FRT-hygro and IDLV/PGK-Flp we obtained from 15 to 35 hygromycin B-resistant colonies (Figure 3E), indicating that efficiency of LV circle insertion varied only to a small degree between the tested FRT-tagged clones.

To demonstrate that the DNA circles were precisely inserted by Flp into the FRT docking site, we first analyzed five representative clones by southern blot analysis of genomic DNA digested with XbaI, which cuts within the duplicated FRT sequences (Figure 4A). This analysis verified that full-length circles had been site-specifically inserted. To further characterize the inserted DNA circles 22 clones (10 and 12 clones from figure 3B and 3D, respectively) were isolated and analyzed. PCR amplification of genomic DNA using primer sets flanking the upstream and downstream FRT junction sites, respectively, resulted in PCR fragments indicative of site-specific circle insertion into the SB-tagged locus (Figure 4B, panel I–II and V–VI). Sequencing of these PCR products confirmed precise recombination between the genomic FRT site and the circle-associated FRT site, except for one clone (clone 14) in which part of the downstream region of the vector had been deleted. Although this clone carried marks of an inaccurate circle insertion, it could alternatively be the result of Flp-based recombination between a linear substrate and the genomic FRT site. In theory, linear substrates may be as efficient substrates as their circular counterparts but the outcome of vector incorporation into the genomic FRT site may differ. Whereas circular forms are resolved by the Flp recombinase into the genomic site without generating DNA breaks, linear forms containing free LTR ends would be expected to generate potentially harmful double-strand DNA breaks. Such complications would suggest that LV hybrid systems based upon cut-and-paste transposases acting on circular and linear substrates may be advantageous and perhaps safer than recombinase-based systems.

Next, we carried out PCR amplification and sequencing of fragments containing the LTRs of the site-directed insertions. Eight 1-LTR and fourteen 2-LTR circle integrations, created by Flp-based insertion of circles generated by homologous recombination and NHEJ, respectively, were detected (Figure 4B, panel III and VII) confirming that the integrated DNA in each clone was indeed derived from circular substrates. We analyzed 33 additional clones and from a total of 55 clones, 30 1-LTR and 25 2-LTR insertions were identified (data not shown). According to previous findings claiming a 9:1 ratio of 1-LTR to 2-LTR circles present in transduced cells [17], there appears to be a preference for insertion of 2-LTR circles in our system. Although this finding could reflect differences in the cellular location of circle species and/or their accessibility for proteins in trans, possible explanations remain speculative. Altogether, our data demonstrate that DNA circles generated during lentiviral transduction are indeed accessible for nonviral integrases in trans and may serve as substrates for genomic recombination by Flp protein transiently delivered by an integration-defective vector.

Based upon 2-LTR copy numbers, measured by q-PCR, we calculated an estimated insertion rate of 1.7 × 10^-5 Flp-based insertions per single copy of transduced 2-LTR circle (number of drug-resistant colonies containing a 2-LTR insertion/(number of 2-LTR circles per cell × total number of transduced cells)), leading us to suggest that 2-LTR circles, in comparison to transfected plasmid DNA, might be more efficient substrates for Flp-based recombination.

The D64V mutation in the HIV-1 integrase protein removes all enzymatic activity of the protein and has been reported to reduce integration of vector DNA 1.000–10.000-fold [6, 15, 16]. We reproducibly measured a 1000-fold reduction by colony-forming assays and, hence, could not formally rule out that integrase-independent insertion of linear IDLV/FRT-hygro or IDLV/PGK-Flp could have occurred in hygromycin B-selected clones. We therefore verified by PCR analysis that unspecific integration of linear vector DNA had not occurred in any of the hygromycin B-resistant clones (Figure 4B, IV and VII–IX), providing proof that the cell clones carrying a site-directed insertion did not carry additional background vector insertions.

Our findings provide, to our knowledge, the first proof-of-principle that integration of lentiviral DNA can be facilitated by a nonviral recombinase, thereby altering the integration profile of LV vectors – in this case towards a site-directed profile using DNA circles as a substrate for gene insertion. The normal catalytic activities of the lentiviral integrase were replaced in a drug-selective approach with the site-directed properties of the yeast Flp recombinase. As Flp recognition sites are not present in the human genome, an eGFP fusion gene containing the FRT sequences was inserted by a novel transposon-based FRT docking vector prior to IDLV transduction. The site-directed approach allowing insertion only in the genomic engineered sites resulted, as expected, in a fairly low insertion rate. As more potent alternatives, LV-hybrids with integration machineries derived from the phage ΦC31 integrase [18], hyperactive Sleeping Beauty transposases [19] or the AAV Rep protein [20] will have direct relevance in human cells, although cut-and-paste transposon systems would be preferred to mimimize the risk of generating double-strand DNA breaks as a possible outcome of recombinase-based insertion of linear substrates. Site-directed integration technologies based on the recombination capabilities of tyrosine recombinases as Cre or Flp have been widely used to direct targeted insertion of nonviral DNA plasmid substrates into genomic recognition sites. Flp-mediated plasmid insertion routinely allows genetic engineering of easy-to-transfect cell lines and has become pivotal in comparative gene expression studies in which integrated transgenes of interest may otherwise be variably influenced by the surrounding DNA. Our data describe a lentivirus-based technology with possible implication for Flp-mediated gene insertion in hard-to-transfect cell lines or potentially in FRT-tagged tissues or primary cells that are not easily transfected. Recent evidence has supported the fact that episomal lentiviral circles are highly stable in non-dividing cells and may provide persistent transgene expression in vivo [8]. For many applications, the stability of lentiviral circles in non-dividing cells does not call for genomic integration of the transgene. Nevertheless, stable viral episomes may represent permanent targets for recombinase-directed insertion of lentiviral circles in non-dividing cells.

The obvious difference in efficiency (as measured by the number of hygromycin B-resistant colonies) between plasmid- and the LV circle-based Flp integration systems (Figure 3A versus Figure 3B and 3D) may reflect differences in the availability of circular substrates in the two systems (plasmid DNA versus LV circles) or, alternatively, that plasmid DNA is somehow better suited for Flp-based insertion. Quantitative PCR analyses comparing the number of plasmids in transfected cells with the number of episomal 2-LTR DNA circles in LV-transduced cells indicated that about 600 times more episomal DNA circles are available in the plasmid-based system compared to the LV-based system (1.3 × 10⁴ plasmids per cell compared to 22 2-LTR DNA circles per cell). However, as we found that 50% of the hygromycin B-resistant clones were the result of 2-LTR circle insertions, we estimate that the high number of plasmids available in transfected cells produced only 52 times more hygromycin B-resistant colonies than transduced 2-LTR episomes (2546 versus 48.5 colonies), leading to the assumption that LV-derived 2-LTR circular DNA is more efficiently integrated than plasmid-based substrates. Possible explanations for this finding, including varying substrate concentrations in different cellular compartments and differences in accessibility caused by structural substrate constraints (relaxed versus supercoiled substrates), are currently subjects for further scrutiny. Comparable levels of colony formation in experiments using IDLVs as a source of Flp recombinase and/or circular DNA substrates (Figure 3B, C, and 3D) indicate that both virus-derived Flp and DNA substrates are limiting factors for efficient Flp-based gene insertion.

By integrating viral circles rather than plasmid DNA, we obtained Flp-based insertions that are not potentially harnessed by bacterial sequences derived from the plasmid backbone. A recent in vivo study has shown an increase in heterochromatin-like histone modifications correlating with lower levels of transgene expression from transfected plasmids containing bacterial sequences compared to plasmids devoid of bacteria-derived DNA [21]. Together, our findings provide a novel tool for Flp-based genetic engineering and pave the way for future applications of lentiviral DNA circles as potential substrates for more therapeutic relevant nonviral integration machineries in somatic gene transfer.

Site-specific integration systems as Cre and Flp are important tools in genetic engineering and animal transgenesis but can in some instances be hampered by the requirement for transfection of plasmid DNA. Lentiviral vectors efficiently transduce both proliferating and quiescent cells but have a preference for integration into transcriptional units thereby representing a potential risk of insertional mutagenesis. A portion of the lentiviral DNA in transduced cells is converted into episomal circlular forms. We have herein demonstrated for the first time that an exogenous recombinase has access to lentiviral DNA as substrate for DNA insertion with the potential of altering the integration profile of HIV-1-derived vectors. In the present setup we combined lentiviral delivery with the site-specific integration properties of the Flp recombinase in a drug-selective approach. By packaging the viral vector with the inactive D64V integrase mutant we were able to increase the amount of circular substrates 4-fold and at the same time abolish the viral integration machinery. We demonstrate that both 1- and 2-LTR circles were inserted into engineered FRT acceptor sites in the genome of human cells by trans-acting Flp recombinase delivered either by Flp-encoding transfected plasmid DNA or by co-transduced integrase-defective lentiviral vectors carrying a Flp expression cassette. Among several advantages of this approach, genetic cargo inserted as viral DNA circles is not potentially harnessed by bacterial sequences derived from the plasmid backbone, the latter which is believed to negatively affect the long-term stability of transgene expression. Moreover, this hybrid lenti-Flp technology may be usefull in studies that require Flp-mediated gene insertion in hard-to-transfect cell lines. Our findings provide evidence that trans-acting integrases are able to gain access to and insert transduced lentiviral DNA and open up for the development of new hybrid systems which do not rely on an engineered genomic target sequence and which are based on more therapeutically relevant recombinases or transposases for potential use in gene transfer.