Lentiviral Protein Transduction with Genome-Modifying HIV-1 Integrase-I-PpoI Fusion Proteins: Studies on Specificity and Cytotoxicity

Rare-cutting endonucleases, such as the I-PpoI, can be used for the induction of double strand breaks (DSBs) in genome editing and targeted integration based on homologous recombination. For therapeutic approaches, the specificity and the pattern of off-target effects are of high importance in these techniques. For its applications, the endonuclease needs to be transported into the target cell nucleus, where the mechanism of transport may affect its function. Here, we have studied the lentiviral protein transduction of the integrase (IN)-PpoI fusion protein using the cis-packaging method. In genome-wide interaction studies, IN-fusion proteins were verified to bind their target sequence containing 28S ribosomal RNA (rRNA) genes with a 100-fold enrichment, despite the well-documented behavior of IN to be tethered into various genomic areas by host-cell factors. In addition, to estimate the applicability of the method, DSB-induced cytotoxic effects with different vector endonuclease configurations were studied in a panel of cells. Varying the amount and activity of endonuclease enabled the adjustment of ratio between the induced DSBs and transported DNA. In cell studies, certain cancerous cell lines were especially prone to DSBs in rRNA genes, which led us to test the protein transduction in a tumour environment in an in vivo study. In summary, the results highlight the potential of lentiviral vectors (LVVs) for the nuclear delivery of endonucleases.


Introduction
Delivery of proteins instead of their cDNAs into target cells is a feasible option in certain therapeutic and experimental approaches where the sustained presence of heterologous proteins is not required or desired. Protein transduction is traditionally achieved by the use of polypeptides or protein transduction domains fused to the protein of interest. There are some commonly used tools for such strategies, for example, the human immunodeficiency virus-1 (HIV-1) Tat protein, the Drosophila melanogaster antennapedia peptide (Ant), and the VP22 protein of the herpes simplex virus. Although these approaches are efficient in vitro, they suffer from the lack of appropriate vectorization, which hampers efficient protein delivery into specific tissues, cell types, or distinct cell compartments [1][2][3][4]. LVVs and vector derived virus-like particles (VLPs) have been used to transport therapeutic proteins into targeted cells as HIV-1 Vpr fusion proteins [5]. This so-called trans-packaging strategy [6] has been used in most studies where proteins have been packaged into lentiviral VLPs and LVVs. The cDNA of a foreign protein is cloned in-frame to vpr-gene in a separate expression plasmid, and the Vpr-fusion protein becomes incorporated into the newly formed vectors and VLPs through its interaction with the p6 protein of the Gag. However, the utility of the trans-packaging method may be subject to limitations in certain applications owing to the proapoptotic and cytotoxic properties of Vpr [7].
We have previously described a cis-packaging method, which is based on generating fusion proteins with the HIV-1 IN, the protein responsible for transgene integration [8].
Proteins of interest are cloned to the C-terminus of IN in the pol-gene of the vector packaging plasmid. Pol becomes translated as a Gag-Pol precursor polyprotein through a ribosomal frameshift that occurs at a frequency of 5-10% [9]. In addition to IN, pol contains the genes for the viral enzymes reverse transcriptase (RT) and protease (PR), the latter being responsible for the timely order of precursor protein processing that leads to virion maturation. Processing of the Pol to its subunits occurs after virion budding. INfusion proteins thus become incorporated into new vector particles as a part of the large Gag-Pol polyprotein, and they are released from Pol only after the vector particle has left the producer cell, thereby enabling the packaging of both toxic and inert proteins into LVVs and VLPs. Such particles are devoid of Vpr and retain their transgene-transferring capability and integration proficiency, with the latter property depending on the vector-contained IN-protein composition [8,10].
I-PpoI is a homing endonuclease from the slime mold Physarum polycephalum [11]. It has a 15 bp cleavage site present in the 28S ribosomal RNA (rRNA) gene, which is highly conserved throughout the eukarya. The rRNA genes are found as hundreds of copies in the ribosomal DNA (rDNA) areas of the short arms of the acrocentric chromosomes 13, 14, 15, 21, and 22 [12][13][14]. In addition to numerous rDNA sites, I-PpoI recognition sites can be found also elsewhere in the genome.
In this work, we have characterized the genomic DNAbinding specificity of the IN-I-PpoI fusion protein after its delivery into cells by LVVs. It is well-documented that IN takes advantage of cellular transcription factor LEDGF/p75 in order to gain access to chromatin [15,16]. Therefore characterizing the specificity of IN-I-PpoI's target binding is important for the applicability of cis-packaging. In addition to binding specificity, we wanted to characterize the cytotoxicity of IN-I-PpoI protein transduction. For any nuclease used, extensive amount of genomic DSBs in target sites and offtarget activity elsewhere in the genome leads to cytotoxicity. To be able to work with tolerable level of DSBs, for example in homologous-recombination based strategies one needs to adjust the ratio between the endonuclease-induced DSBs and transported proviral DNA. To achieve this, we used an IN-I-PpoI derivative with the H78A mutation in the I-PpoI sequence. This protein was generated to decrease I-PpoI's full enzymatic activity. Unlike the noncutting N119A mutant described before, H78A exhibits reduced catalytic activity [17]. The results confirm that the lentiviral delivery with cis-packaging method per se is not a determinant of endonuclease protein transduction cytotoxicity but that the characteristics of both the used endonuclease and the target cells play important roles.

Results
We wanted to ask the following research questions: to what extent the endonuclease binding into host cell DNA is specific, what the amount of off-target interactions is, how cells do respond to different levels of I-PpoI-derived DSBs, and whether the DNA-cleavage can be used successfully in totally different application-in an in vivo tumour model. For the experiments, vector batches containing either IN-I-PpoI or IN-I-PpoI H78A were produced (Table 1)  Mixed multimer or trans-complemented vectors were generated by mixing equal amounts of the fusion proteincontaining and either wild-type (wt) IN or inactive IN D64Vcoding packaging plasmids. The H78A mutation was introduced into I-PpoI in the fusion-IN to investigate the effects of a less active endonuclease. Replacing the histidine in the catalytic site of I-PpoI with an alanine has previously been reported to decrease the enzyme's activity to 48% of wild-type level [17,18].

ChIP-Analysis Confirms That I-PpoI Is Capable of Undergoing an Interaction with Its Target Sequence.
Before the large-scale studies, the interaction of IN-I-PpoI H78Acontaining vectors with a single I-PpoI target site on chromosome 1 (1p32.2) was confirmed using chromatin immunoprecipitation (ChIP) analysis. A statistically significant interaction between the vector-carried proteins and the target site was observed, whereas no positive qPCR signals were detected from nontransduced or LVV IN wt transduced cells ( Figure 1).
After verifying the target sequence binding by IN-I-PpoI H78A , the IN-endonucleases' chromatin interactions were studied using ChIP-sequencing, which allows mapping of all protein contacts with cellular DNA. Studies were conducted using the same vectors as used in the cytotoxicity study comprising HeLa cells and MRC-5s: the transcomplemented LVVs IN wt +IN-I-PpoI H78A and IN D64V +IN-I-PpoI H78A . Since the rDNA is not included in the chromosomal DNA sequences of the latest human genome version GRCh37/Hg19, hits in the unplaced contig ChrUn gl000220 were counted and regarded as rDNA interactions [10]. The unplaced contig contains one full-length and one partial rDNA repeat, in addition to an unknown area to which no rRNA gene sequences could be mapped.
All of the studied LVVs carrying IN-fusion proteins exhibited an increased interaction with the rDNA repeat, when compared to the nonmodified LVV IN wt (Figure 2(a)). Inside the repeat, interactions occurring in the 28S rRNA gene were further studied, since this structure harbors the 15bp I-PpoI target site (Figure 2(b), % of all rDNA interactions). After transduction with fusion protein containing LVVs, on average, the majority (∼52%) of the interaction sites within rDNA were localized to the 28S rRNA (2.30% of total interactions). In the case of IN wt control, only 2.1% of   Figure 2(c)). Taken together, it is concluded that the differences in cytotoxicity between the IN-I-PpoI with native endonuclease activity and its mutated form IN-I-PpoI H78A are not due to a reduced target DNA-binding ability of the latter, as verified by ChIP sequencing.

The Cytotoxicity of IN-I-PpoI Protein Transduction Is Dependent on the Enzyme's Activity and Amount of Protein
Packaged into Lentiviral Particles. I-PpoI has at least eight perfect full-length recognition sites in the human genome [10] in addition to approximately 400-600 sites found in the rDNA [19]. The high number of potential cleavage sites poses a remarkable challenge to the DNA damage repair capabilities of the transduced cells. An excess of endonucleases can result in genomic instability and genotoxicity, as observed with zinc finger nucleases [20][21][22].  e)). With the 10 ng dose, a cytotoxic effect was observed, confirming that the endonuclease was still packaged into the vectors. As expected, reducing the content and activity of the INendonuclease in vector particles proved to be a feasible way of modulating the cytotoxicity. In addition, the characteristics of the target cells affect the cytotoxicity encountered with this approach, the cis-packaging method itself being well suited for nuclear delivery of the endonucleases. Differences in cytotoxicity between vectors carrying the IN-endonuclease proteins can result from their different abilities to recognize I-PpoI sites, resulting in off-target effects. The ability of the IN-I-PpoI forms to bind to their genomic target sites was therefore next studied using chromatin ChIP-techniques.

Cell Culture Studies Indicate Increased Cytotoxicity in
Tumour-Derived Cell Lines. Although excessive DNA double strand break (DSB) formation is cytotoxic, site-specific cleavage can be exploited for therapeutic purposes in the form of genome editing and gene insertion through enhanced homologous recombination (HR [23,24]). To characterize whether cancerous cells in addition to the tested HeLa cells would be more sensitive to IN-I-PpoI-originated cytotoxicity,   Figures S3 and S4), although the impact on HUVEC cells was difficult to interpret due to the deviant viability at the last time point. However, overall differences between the cell lines were moderate. With a high LVV dose, especially, the A549 and BT4C cells, which likely also suffer from cancer-specific defects in their DSB repair pathways, perished or stopped dividing. Similarly, 293T cells, despite not being a cancer cell line, exhibited extensive cytotoxicity in response to LVV IN-I-PpoI transduction. HEK293-derived cell lines, such as 293T, are not suitable models for healthy cells in DSB experiments, since they are known to express the adenoviral oncoprotein E1B55K, which disrupts the functionality of the DNA damage response pathways [25].  were detected in any of the groups (Figure 6(c)). These results suggest that LVV protein transduction can be successfully used also in more difficult-to-transfect cellular environments.

Discussion
We have previously shown that LVV IN-I-PpoI can induce targeted DSBs and its cleavage-impaired mutant can increase transgene integration into rDNA [10]. However, there was no direct proof of a protein interaction with or close to the aimed I-PpoI recognition sites. Here, we have addressed this open question by using the ChIP sequencing with IN-I-PpoI and its catalytically impaired version IN-I-PpoI H78A . The fusion protein containing vectors exhibited increased chromatin interactions involving rDNA repeats and the 28S rRNA genes within these regions. The result highlights the feasibility of fusing chromatin-interacting proteins to IN. In addition, the non-rDNA-related I-PpoI recognition sites were more frequently present in the IN-fusion protein data sets as compared to control, demonstrating IN-I-PpoI's ability to interact with its target sites also outside rDNA.
The number of IN molecules contained in a lentivirus particle is limited with estimates between 20 and 250 molecules [28,29]. IN-fusion proteins may be present at lower levels because of potentially inefficient IN-fusion protein expression and packaging into new particles in producer cells. We have not determined the number of particles lacking the INfusion protein in our vector preparations. However, since gag-pol is transcribed in a fixed relationship to gag [9], the stoichiometry between (Gag-)Pol and Gag should be preserved when using the IN-fusion protein-containing packaging plasmids for vector production. Based on a stoichiometry value of 2000 copies of p24 capsid proteins per viral particle, a 10 ng dose should correspond roughly to 10-100 TUs per cell in our experimental settings [30,31]. Considering the strong effects on cellular viability encountered with low to moderate vector doses, we can conclude that the number of IN-molecules per vector particle does not represent a limiting factor, at least not in the two different study types presented here. This information could be useful for applications such as DSB-enhanced HR, which could benefit from the protein transduction technology we have described. Once inside the cell, the protein can be delivered into the nucleus or potentially to the cytosol, if the step of nuclear import is disabled for the modified vectors. According to our results, IN-I-PpoI LVVs do not cause extensive cytotoxicity in all cell lines when administered at low doses (Figures 1, 4, and S3), despite the fact that I-PpoI has several hundreds of recognition sites in the human genome. About 30% of the genomic I-PpoI sites were recently analyzed to become cleaved after I-PpoI administration [32].
Susceptibility to DSB formation and reduced DNArepairing capability through ionizing radiation or other forms of cellular stress are present in most cancers. In normal cell lines, the number of DSBs evoked by a certain amount of stress is largely constant and the outcome is mediated by the efficiency of the DSB repair, whereas, in cancer cells, the increased total number of DSBs may present a major challenge for maintaining cellular viability. Apparently nucleases into target cells without adding Vpr to the vector production system. For HR applications, endonucleases with unique cleavage sites in the human genome would be preferred catalysts for DSB generation. However, it remains to be determined whether the amount of packaged INfusion proteins is sufficient for proteins with lower activity than observed for I-PpoI, to exert their specific cellular functions. In summary, by incorporating the DNA-cleaving meganuclease I-PpoI into the 3rd generation LVVs, we showed that LVVs with cis-packaged LVVs could be used as versatile tools to transfer genome modifying proteins into target cells.  Figure S1B) and after LVV protein transduction in MRC-5 cells ( Figure S1C)