Hybrid Lentivirus-phiC31-int-NLS Vector Allows Site-Specific Recombination in Murine and Human Cells but Induces DNA Damage

Gene transfer allows transient or permanent genetic modifications of cells for experimental or therapeutic purposes. Gene delivery by HIV-derived lentiviral vector (LV) is highly effective but the risk of insertional mutagenesis is important and the random/uncontrollable integration of the DNA vector can deregulate the cell transcriptional activity. Non Integrative Lentiviral Vectors (NILVs) solve this issue in non-dividing cells, but they do not allow long term expression in dividing cells. In this context, obtaining stable expression while avoiding the problems inherent to unpredictable DNA vector integration requires the ability to control the integration site. One possibility is to use the integrase of phage phiC31 (phiC31-int) which catalyzes efficient site-specific recombination between the attP site in the phage genome and the chromosomal attB site of its Streptomyces host. Previous studies showed that phiC31-int is active in many eukaryotic cells, such as murine or human cells, and directs the integration of a DNA substrate into pseudo attP sites (pattP) which are homologous to the native attP site. In this study, we combined the efficiency of NILV for gene delivery and the specificity of phiC31-int for DNA substrate integration to engineer a hybrid tool for gene transfer with the aim of allowing long term expression in dividing and non-dividing cells preventing genotoxicity. We demonstrated the feasibility to target NILV integration in human and murine pattP sites with a dual NILV vectors system: one which delivers phiC31-int, the other which constitute the substrate containing an attB site in its DNA sequence. These promising results are however alleviated by the occurrence of significant DNA damages. Further improvements are thus required to prevent chromosomal rearrangements for a therapeutic use of the system. However, its use as a tool for experimental applications such as transgenesis is already applicable.


Background
Gene transfer technologies are essential for genetics studies and gene therapies. However, major challenges remain to be addressed. A major issue is the lack of control over the site of DNA integration in the host genome which leads to unpredictable gene expression level and potentially undesirable mutagenesis of important cellular genes [1]. Recent strategies to tackle this challenge are relying on the use of genome editing tools such as ZFNs [2][3][4][5][6][7], TALENs [8][9][10][11][12][13] or more recently CRISPR-Cas system [14][15][16][17]. However, the vectorization of these tools into viral vectors to optimize their use ex vivo or in vivo raises several problems. Indeed, ZFNs function as dimers and generally require cotransduction of three vectors (one for each dimer and one for the recombinating substrate) [18][19][20]. Moreover ZFNs may induce cellular toxicity due to off target activity [21][22][23]. TALENs have an important size with repeat domains hampering their vectorization [24]. CRISPR-Cas is a very recent tool and its vectorization has not yet been described. One may however expect its vectorization into viral vectors will be challenging as the system is based on the concomitant use of a chimeric DNA displaying hairpin structures [25] and of Caspase 9 which induces apoptosis when over-expressed [26]. These features will undoubtedly represent challenges for the vectorization of CRISPR-Cas into viral vectors for targeted integration.
Site-specific recombinases such as Cre [27][28][29][30][31][32][33][34] or FLP [35][36][37][38] of the tyrosine recombinases family are other genome editing tools more easily vectorizable and widely used for the purpose of site specific integration. However, the use of these recombinases is limited by the absence of endogenous recognition site in mammalian cells and by the bidirectionality of the recombination reaction they mediate. Within the superfamily of site-specific recombinases, phage integrases catalyse unidirectional recombination events [39]. Among these the PhiC31 phage integrase (phiC31-int), of the large serine recombinases family, is the most commonly used site-specific integrase for gene transfer purposes [40,41]. In its natural context, phiC31-int mediates efficient recombination between the phage attachment site (attP) and the bacterial attachment site (attB). The recombination of these two sites results in the unidirectional and site-specific integration of the phage genome into the bacterial chromosome (reviewed in [39]) leading to an integrated phage genome flanked by the recombinant attL (left) and attR (right) sites ( Figure 1). When used for gene transfer into eukaryotic cells, phiC31-int can catalyse integration of a plasmid containing an attB sequence into endogenous pseudo attP sites (pattP) displaying a high degree of homology with the wild type attP site [42]. Hence, associated with transfection techniques, phiC31-int has been successfully exploited to stably modify the genome into particular genomic sites of many types of eukaryotic cells in vitro [43][44][45][46][47][48][49][50] for transgenesis [48,[51][52][53][54][55] and gene therapy applications [56][57][58][59][60][61][62][63][64][65]. The use of phiC31-int presents several advantages. First, the recombinase can be used to generate conservative recombination between attB and pseudo attP sites [42]. Second, Chalberg et al demonstrated that the majority of phiC31-int mediated recombination events in the human genome occur in intergenic regions [66]. However, in most of these studies the vectorization of phiC31-int relied on cotransfection (or nucleofection) of plasmids for both the delivery of phiC31-int and of the transgene, thus limiting this technology to in vitro or ex vivo applications.
A strategy to increase the efficiency of DNA delivery in vivo is to use viral-derived vectors. As the expression of the genome editing tool must be transitory to avoid genotoxicity, it could therefore be delivered by a transient viral vector [67]. Even though phiC31-int has already been delivered by an adenoviral vector system [68,69] the use of such vectors is limited by their cell toxicity and immunogenicity [70][71][72]. In contrast, lentiviral vectors (LV) have the advantage to be non-immunogenic [73,74] and offers the possibility to be pseudotyped by different envelops, allowing a high degree of flexibility regarding the tropism of the particles and the type of cell(s) they transduce (for review see Cronin J. et al [75]). Most importantly, it was shown that LV integrase can be modified to obtain non integrating lentiviral vectors (NILVs) [76][77][78], which act as episomal vectors. Hence, NILVs are vectors of choice to deliver genome editing tools and have been successfully used to deliver transposases [79,80], FLP [81] or ZFNs [82,83]. However, the use of NILVs has never been described for the vectorization of a serine recombinase.
In the present study, we combine the unidirectional site-specific recombination capability of phiC31-int with the efficiency of NILVs for gene transfer. For targeted integration,two different NILVs, one delivering the DNA sequence to be integrated and containing the natural attB site, the other expressing the phiC31int are used. Through a step by step approach, we demonstrate for the first time that phiC31-int can be vectorized in NILV and we provide clues to further improve the system. However, analysis of integration events reveals that significant DNA damages can result from phiC31-int mediated recombination. In conclusion, the vectorization of serine recombinases in NILVs is feasible and constitutes a promising tool for basic research; however, one should remain cautious about the chromosomal aberrations that can be induced by these recombinases, particularly for clinical uses.

Results and Discussion
NILV genomes can be used as a substrate for site-specific integration mediated by phiC31-int into human genome We first assessed the ability of a NILV DNA genome to be used as a substrate for phiC31-int. We therefore generated a Hela cell line constitutively expressing phiC31-int thanks to an integrative lentiviral vector expressing the phiC31-int under the control of the CMV promoter (LV CMV-phiC31-int) to. Hela cells were transduced and clonal populations were isolated and analyzed by RT-PCR to estimate the phiC31-int expression level ( Figure 2A). The clone Hi16 was selected for its robust constitutive expression of phiC31-int.
The ability of the constitutively expressed phiC31-int to mediate recombination between a NILV bearing an attB site with genomic pattP sites was then tested. Hi16 cells were transduced with a nonintegrative lentiviral vector expressing the Neomycine (Neo) resistance gene under the control of the CMV promoter and containing an attB site (NILV attB-CMV-Neo). After two weeks of G418 selection we obtained four cell clones which genomes were analyzed. Theoretically Neo integration is expected to arise from 3 distinct mechanisms: (i) phiC31-int-specific integration, (ii) residual integration mediated by a residual activity of the mutant HIV integrase (review about this field [84]) or (iii) illegitimate integration due to recombination of the episomal DNA molecule with the cellular genome by host cell mechanisms ( Figure 3A). If the analysis is realized with clonal populations, these three different mechanisms should be discriminated by performing 2 PCR assays, one amplifying the LTR and the other amplifying the attB site. A positive LTR PCR reveals the presence of a LTR-LTR junction, indicating that the integration has occurred through a LTR-independent mechanism, either involving attB site-specific recombination or illegitimate recombination. In contrast, a positive attB PCR reveals that the attB site is intact, indicating an attB-independent integration, either involving LTR dependent (residual) integration or illegitimate recombination. In summary, cells are analyzed without knowing whether their genome contains one or several integration of NILV. A positive result for LTR PCR only or for attB PCR only allows to identify the mechanism of integration without ambiguities (ie respectively phic31-int specific integration or residual integration). As the LTR PCR does not amplify the 1-LTR region generated when the linear genome circularized through homologous recombination of both LTR regions, recombination involving 1-LTR circles would always result in LTR2 in the PCR test. In cells where a single integration of a 1-LTR circle event arose, attB+/LTR2 or attB2/LTR2 profiles may be detected, respectively when the recombination is independent or dependent on phiC31-int. In all our experiments, we never observed attB2/LTR2 clones. On the contrary, if both PCR are positive integration could result either from illegitimate recombination or from double independent recombination events (ie: phiC31-int site specific recombination and HIV integrase mediated residual integration). We collected the four clones and checked integration pattern of NILV with PCR. The PCR analysis revealed that all clones were positive for both LTR and attB PCRs ( Figure 2B). Thus this result does not allow to conclude about the nature of the integration events.
To clear up this ambiguity, genomic DNA of isolated clones was further analyzed by inverse PCR (iPCR) to isolate potential specific integration events. attB based primers (P1/P19) were used to amplify junctions of recombination. To avoid the PCR background generated by the non-recombined attB site, an enzymatic cocktail including an enzyme which cuts only once in the vector sequence, in 59 to the attB site ( Figure 3B) was used. In this way, we were able to isolate pseudo attR junctions with iPCR, (but not the pseudo attL junction). Using this adapted strategy, a lentiviral genome integrated by phiC31-int at the human locus Xq22.1 was detected ( Figure 4). Interestingly this locus had already been described as a preferential pattP site in the human genome [66]. Moreover, the core sequence is exactly in the same position, suggesting that the recombination event occurred very precisely at this locus.
Although NILV integration also occurred through other means than phiC31-int recombination, our results clearly demonstrate that a NILV is a suitable substrate for phiC31-int mediated recombination in human cells. We therefore investigated whether phiC31-int could function when vectorized in a NILV.
NILVs allow adequate expression of phiC31-int to mediate recombination into a reporter system containing an attP site artificially introduced in human cells The ability of a NILV to express phiC31-int and mediate recombination between another NILV genome carrying an attB site and a wild-type attP site artificially introduced in a human cell genome was further tested. First, a clonal Hela cell line containing the wild-type attP site inserted in its genome (HDsred line) was generated using an integrative LV (CMV-attP-DsRED2). HDsred cells were then transduced with two NILVs, one allowing expression of phiC31-int (NILV CMV-phiC31-int), and the other expressing Neo and containing the attB site (NILV attB-CMV-Neo). After cotransduction cells were grown with G418 to select integration events. The genomic DNA extracted from Neo resistant clones was analyzed by PCR to detect attL recombination junction ( Figure 5A). Results showed only background signal generated by non-recombined attP site (figure 5B). To prevent this amplification, the genomic DNA was digested by a restriction enzyme that cuts both attP and attR sites but not attL site ( Figure 5A) prior to PCR amplification. PCR results obtained after the enzymatic treatment revealed an attL junction in the population transduced with the highest dose of phiC31-int vector ( Figure 5C). These results have been further confirmed by nested PCR ( Figure 5D) and PCR product sequencing. Taken together, these results demonstrate that a NILV can deliver a functional phiC31-int capable to integrate an episomal lentiviral substrate containing an attB site into an attP site artificially introduced into the human genome.
Although the used PCR strategy does not allow to estimate the targeting efficiency of the attP site with the double vector system, the need of a nuclease digestion before PCR to reveal specific integration events suggests that the efficiency of the two NILVs system is low. As this assay involved the two natural recombination sites of PhiC31-int, further reduced efficiency would be expected when targeting endogenous pattP sites. Consequently, we next focused on the improvement of the two NILVs system efficiency.
Modification of the phiC31-int sequence to improve the efficiency of the NILV phiC31-int to allow target integration in pseudosites attP It was previously shown that C-terminal addition of a nuclear localization system (NLS) to phiC31-Int improves its efficiency in eukaryotic cells [85]. We therefore tested this improved integrase in NILV vectorization strategy. To compare the two versions of phiC31-int, Hela cells were cotransduced with the following NILVs vectors: CMV-Neo with or without attB site and CMV-phiC31-int with or without NLS. Cells were grown with G418 and the resistant clones were quantified. The results from experiments in which CMV-phiC31-int was cotransduced with either CMV-Neo or attB-CMV-Neo are presented in Figure 6A. The CMV-Neo and the attB-CMV-Neo conditions did not display significant differences, indicating that no significant PhiC31-int recombinase activity occurred. In contrast, when the cells are transduced with CMV-phiC31-int-NLS instead of CMV-phiC31-Int ( Figure 6B) the presence of the NLS sequence induced significant differences between the CMV-Neo and the attB-CMV-Neo conditions. These results show that addition of a C-terminal NLS to phiC31-int significantly increased its recombination efficiency with pattP sites. Indeed the phiC31-int-NLS mediated integration was 2 to 2.5 fold above the background level produced by NILV residual integration ( Figure 6B).
The use of the phiC31-int-NLS vectorized in a NILV allows to significantly increase the efficiency of recombination. We therefore further tested this hybrid lentivirus phiC31-int-NLS vector to target genomic pattP site into murine and human cells.

The two NILVs system allows site specific recombination in murine and human cells but induced aberrant chromosomal rearrangements
To target pattP site in the murine and human genome with the two vectors system, Hela cells and NIH-3T3 cells were cotransduced with the NILV attB-CMV-Neo and the hybrid lentivirus phiC31-int-NLS vector. After two weeks of selection, Neo resistant cells were collected and several clonal populations were isolated to facilitate interpretation of PCR analyses. The clones were analyzed with the LTR and attB PCRs assay ( Figure 3A) to determine the proportion of specific integration events compared to NILV residual integration and illegitimate recombination events. We categorized clones in the 3 following groups: LTR+/ attB2 clones (group I, ie phiC31-int-NLS integration), LTR2/ attB+ clones (group II, ie residual integration) and LTR+/attB+ clones (group III, ie illegitimate integration or mixed integration profile). The results are presented in Table 1.
We obtained 108 clones for murine cells and 28 for human cells. 7.5% of murine clones are in group I and 20.3% in group III (Table 1A). Consequently the integration mediated by the hybrid lentivirus phiC31-int-NLS vector is comprised between 7.5% and 27.8%. As no human clone corresponded to group III, the proportion of the hybrid lentivirus phiC31-int-NLS vector specific integration corresponds to the proportion of group I, ie 46.4% (Table 1B).
To determine which integration sites were targeted and confirm the type of integration of clonal groups I and III, we performed an analysis by iPCR as previously described ( Figure 3B). The sequencing of iPCR products demonstrates that all human and murine clones tested contain in their genome an integrated vector with a recombinant pattern at the attB site. Therefore, the two vectors system that we developed allows targeting pattP sites in the murine and human genomes ( Table 2).
Twelve murine integration sites were isolated, three from group I with the two junctions (pattL and pattR) and nine from group III with only the pattR junction (Table 2). For two of the three group I clones both flanking regions were sequenced but the isolated junctions were too short to allow identification of the integration locus. Surprisingly, these two clones of group I present abnormal pattL and pattR junctions where several bases were missing, probably due to a deletion mechanism. Similarly, the sequence analysis of group III clones shows that recombination between the attB site and pattP site is not as precise as expected. Indeed, 6 out of 9 clones have missing bases in the pattR junctions. However, because we isolated only one flanking region for clones of group III, we cannot conclude about deletion events, as the missing bases could result from a gap of the attB core region involved in recombination.
Five human integration sites were isolated, all from group I. Both junctions were isolated for three sites and only the pattR junction for two sites ( Table 2). As for the murine integration site analysis, missing bases in the recombined pseudo-sites were detected, including in clones for which both L and R junction could be determined. This further confirms that missing bases indeed reveal a deletion mechanism, probably occurring during the phiC31-int mediated recombination between a natural attB site and a pattP site. Furthermore, the integration site of the three sites for which the two junctions were isolated could be localized exactly (Table 2). Interestingly, for these three sites chromosomal gaps were observed between the pattR site and the pattL site. The size gaps are 13 bp, 795 pb and 4.8 kbp. The two first gaps could result from a mechanism of deletion which could occur through NHEJ pathway. Nevertheless, these hypotheses cannot explain the gap of 4.8 kbp. Indeed, in this case, the two flanking regions isolated during iPCR are in the same chromosomal orientation, which is not the case when normal recombination occurs. We hypothesize that the observed aberrant recombination results from two successive recombination events involving two pattP sites located 4.8 kb from each other that led to an inversion of the 4.8 kbp sequence ( figure 7). This mechanistic model was previously proposed to explain chromosomal rearrangements in mammalian cells resulting from aberrant recombinations mediated by phiC31-int [86].
In conclusion the hybrid lentivirus-phiC31-int-NLS vector that we developed allows targeted integration in pattP sites in murine and human genomes but seem to induce frequent deletions of base pairs in the attB site present into the vector or into the endogenous pattP site. Moreover, it may induce chromosomal deletions and translocations.

Conclusions
Our work establishes the ability of the hybrid lentivirus-phiC31int-NLS to integrate a NILV substrate into attP or pattP sites in murine and human cell lines. Although the number of integration sites isolated in this study does not allow to determine preferential genome recognition sites for integration mediated by phiC31-int vectorized in NILV, a human pattP site already described could be isolated [66]. Most importantly, we demonstrated that the use of hybrid lentiviral phiC31-int-NLS vector can induce DNA damages, probably due to the activity of the recombinase. Indeed, other reports already described similar results using non-viral transfection of PhiC31-int. Anja Ehrhardt et al have shown that 15% of transgenes integrated by phiC31-int were flanked by chromosomal DNA sequence from different chromosomes [86] [87,88]. In addition, Ehrard et al. have shown that phiC31-int is competent to integrate linear DNA fragments. They did not characterize the mechanisms and the consequences of this type of event are unknown, but one may hypothesize that such event would induce chromosome break. Considering that the cycle of NILVs involves linear intermediates, it would be of particular interest to further investigate the ability of phiC31-int to integrate these linear forms and the consequences of such events. Taken together, these observations limit the use of the hybrid lentiviral phiC31-int-NLS vector for clinical applications, it remains useful in transgenesis contexts where non aberrant recombination events can be selected. Alternatively, the hybrid lentiviral system may be used to vectorize other recombinases or genome editing tools with higher safety features. Indeed other serine recombinases have been shown to function in human cells [89][90][91] and could be valuable candidates for vectorization in NILVs. Moreover, the modification of recombinases and other genome editing tools by directed evolution techniques [92,93] could be used to render them hyperspecific and hyper-efficient in order to improve their safety features. For instance, directed evolution has proven efficient to modify the efficiency and/or specify of a variety of molecular tools, including Sleeping Beauty [94] Cre recombinase [95], FLP recombinase [96], ZFNs [97] or phiC31-int [98].

Plasmids
The encapsidation plasmid expressing a functional integrase (p8.91 IN WT ) has been described previously [77]. The encapsidation plasmid expressing a deficient integrase (p8. 91  The envelope expression plasmid pMDG(VSV) was used to express the VSV-G from the human CMV immediate early promoter [74].

Cell culture
Human epithelial HeLa, Hi16, HeLa-DsRED2 and 293T cells and murine NIH 3T3 cells were grown in Dulbecco's modified medium (Invitrogen) supplemented with antibiotics (100 U/mL penicillin and 100 mg/mL streptomycin) and 10% heat inactivated fetal calf serum (Eurobio). The cells were plated and cultured in a humidified incubator at 37uC in a 5% CO2 and 90% air atmosphere.

Genomic DNA extraction
Genomic DNA extractions were performed with a lysis buffer composed of TrisHCl 10 mM (pH 7.5), EDTA 10 mM, SDS 0.6%, RNase A (Qiagen) 100 mg/mL and proteinase K (Eurobio) 100 mg/mL. The lysates were purified by phenol/chloroform and precipitated using sodium acetate and ethanol.

Generation of a reporter cell line HeLa-DsRED2
The HeLa-DsRED2 cell line containing an attP site and expressing DsRED2 fluorescent protein was generated using the lentiviral vector CMV-attP-DsRED2. HeLa cells were transduced with LV CMV-attP-DsRED2 (unconcentrated supernatant). Cells were grown for 3 days and seeded at clonal density in 96-well plates (0.3 cell per well) to generate single cell derived colonies. Clones were analyzed for DsRED2 expression by flow cytometry and PCR amplification of the vector genome.

Transduction
Cell suspensions were incubated for 3 hours with required vectors in medium supplemented with 1 mM DEAE-dextran. After 3 hours of incubation, cells were seed at desired density in fresh medium and grown for the purpose of the experiment.
Hi16 cells (2.10 6 cells/mL) were transduced with NILV-phic31int (300 ng of p24) and NILV-Neo (100 ng of p24). Cells were seeded in 10 cm plates and grown in medium supplemented with 1 mg/mL of G418 for 12 days. Cells were then seeded in 96 well plates at low density (0.3 cell per well) to generate single cell derived colonies. HeLa DsRED2 cells (2.10 6 cells/mL) were transduced with NILV CMV-phic31-int (50,150, 300 ng of p24) and NILV CMV-Neo (100 ng of p24). Cells were seeded in 10 cm plates and grown in medium supplemented with 1 mg/mL of G418 (renewed every 3 days) for 12 days before extraction and analysis of genomic DNA.
Hela and NIH-3T3 cells (5.10 5 cells/mL) were transduced with NILV CMV-phic31-int (300 ng of p24) and NILV CMV-Neo (100 ng of p24). Cells were seeded in 10 cm plates and grown in medium supplemented with 1 mg/mL of G418 for 12 days. Cells were then seeded in 96 well plates at low density (0.3 cell per well) to generate single cell derived colonies.

Evaluation of Recombination Frequency
HeLa were directly transduced in suspension (8.10 4 cells/mL) with NILV CMV-phic31-int (3, 6, 18 and 36 ng of p24) and NILV CMV-Neo (12 ng of p24) during 3 hours in 150 mL of medium supplemented with 1 mM DEAE-dextran. Cells were then seeded in 6-wells plates with 2 mL of fresh medium. The day after, medium was removed and replaced with fresh medium supplemented with 1 mg/mL of G418. The medium was replaced every 3 days. Cells were grown 12 days, until clones developed and were then fixed with PFA 4% and stained with trypan blue. Clones on each well were counted. We transduced cells in three replicate tubes for each condition, and results are expressed as the mean of three measurements.

PCR reactions
RT-PCR. To analyze phic31-int expression by RT-PCR, total RNAs were extracted from HeLa cells using the RNeasy minikit (Qiagen), according to the manufacturer's instructions. Then, RNAs were reverse transcribed using the Superscript First Strand Synthesis kit (Invitrogen), according to the manufacturer's instructions. Phic31-int cDNA was amplified using the primers 59- Inverse PCR. Genomic DNA (10 mg) was primarily submitted to enzymatic restriction. The enzymatic cocktail used was: XbaI (100 U), StuI (100 U) BsrGI (30 U) and BstXI (30 U). After 16 hours incubation, the enzymes were heat inactived at 65uC for 30 minutes. Cohesive ends were filled in with Klenow (15 U) and dNTPs (2 mM) at 25uC for 20 minutes. Klenow was inactived with 1 mM EDTA. The products of digestion were purified by phenol/chloroform and precipitated using sodium acetate and ethanol. The ligation of the digestion products was performed by ligase (1.000 U, NEB) within ligation buffer supplemented with ATP (1 mM). The products of ligation were purified by phenol/ chloroform and precipitated using sodium acetate and ethanol. PCR was performed with 100 ng of DNA with primers allowing the amplification of the attB region.
Adapted inverse PCR was performed using the same protocol with the addition of BmtI (30 U) in the enzymatic restriction cocktail.
The products of inverse PCR were visualized on 0.8% agarose gel with ethidium bromide staining. These products were extracted and purified using the Wizard SV Gel and PCR Clean-up System (Promega) according to the manufacturer's instructions, then cloned in a plasmid using pGEM-T Easy Vector System I according to the manufacturer's instructions. Next, inverse PCR products were sequenced using T7 and/or Sp6 primers.

Sequence analysis
All sequencing was performed by Eurofins genomics. Sequences were aligned with vector and genomic DNA, and recombination junctions were identified by sequence matching to attB. Human and murine blasts were performed using the NCBI and ensembl genome databases. The chromosomal localization of pseudosites attP has been performed using the Genebank GRCm38.p2 C57BL/6J assembly in mouse and the Primary Assembly GRCh38 for Human.