Highly efficient CRISPR-mediated large DNA docking and multiplexed prime editing using a single baculovirus

Abstract CRISPR-based precise gene-editing requires simultaneous delivery of multiple components into living cells, rapidly exceeding the cargo capacity of traditional viral vector systems. This challenge represents a major roadblock to genome engineering applications. Here we exploit the unmatched heterologous DNA cargo capacity of baculovirus to resolve this bottleneck in human cells. By encoding Cas9, sgRNA and Donor DNAs on a single, rapidly assembled baculoviral vector, we achieve with up to 30% efficacy whole-exon replacement in the intronic β-actin (ACTB) locus, including site-specific docking of very large DNA payloads. We use our approach to rescue wild-type podocin expression in steroid-resistant nephrotic syndrome (SRNS) patient derived podocytes. We demonstrate single baculovirus vectored delivery of single and multiplexed prime-editing toolkits, achieving up to 100% cleavage-free DNA search-and-replace interventions without detectable indels. Taken together, we provide a versatile delivery platform for single base to multi-gene level genome interventions, addressing the currently unmet need for a powerful delivery system accommodating current and future CRISPR technologies without the burden of limited cargo capacity.


INTRODUCTION
CRISPR/Cas represents a game-changing, Nobel prize winning gene editing tool (1). A programmable DNA nuclease (Cas9) is precisely guided to a specific DNA locus by means of a short single guide RNA (sgRNA) to elicit double strand DNA breaks (DSBs) subsequently repaired through non-homologous end-joining (NHEJ) introducing small insertions-deletions (indels), giving rise to frameshift mutations and functional gene knock-outs (KOs) (2). Unpredictable indels that likewise occur are however undesirable in the context of therapeutic gene editing. Precise gene editing in contrast is typically achieved through homology directed repair (HDR) by providing a DNA template flanked by homology arms of variable length (2,3), resulting in precise gene correction or knock-in (KIs) (2). HDR activity however is intrinsically low and mostly restricted to S/G2 cell cycle phases (4)(5)(6), reducing the efficiency of the desired gene editing outcome. Significant effort is being made to increase the efficacy of CRISPR-HDR by means of small-molecule NHEJ inhibitors (7), cell cycle stabilised Cas9 variants (8,9) and other strategies (10,11), however, gene editing efficiency in vivo has remained low.
More recently, homology-independent targeted integration (HITI), was shown to efficiently induce base pair precise KIs in both dividing and non-dividing cells (12-14 by exploiting NHEJ and Cas9 cleaved donors, with exciting potential for gene editing applications in vivo (12,15). Moreover, single base substitutions can also be achieved by using base editors (BEs) involving catalytically impaired Cas9 variants fused to cytosine or adenine deaminase (16,17) and, more recently, prime editors (PEs) using Cas9 nickase fused to reverse transcriptase, achieving genomic interventions with little to no indels and reducing the risks associated with DSBs (18). BE, PE and HITI share a reliance on multiple functional DNA and protein elements that must be simultaneously delivered into target cells. This limits their applicability, particularly for future bona fide therapeutic interventions that necessitate a systemic approach and where co-transfection of plasmids and proteins, and likewise coinfection of viral vectors, will be problematic or unfeasible. In summary, the large DNA cargo capacity required for implementing these next-generation genomic interventions in vivo stands at odds with the limited cargo capacity of available technology, including the currently dominating adenoassociated virus (AAVs) and lentivirus (LVs) vector systems (19). In response to the CRISPR delivery challenge, new viral delivery vectors, with higher DNA cargo capacity, transduction efficiency and safety features are urgently required (20). To address this unmet need, we developed an ad hoc method for rapid generation of customizable baculoviralvectors (BVs) for precise genome-engineering applications (MultiMate). BVs have a heterologous DNA cargo capacity far exceeding AAV and LV (19,21,22) and are widely used to transduce mammalian cells and living organisms (21)(22)(23)(24). However, early attempts at CRISPR delivery using BVs have not provided significant benefits over other delivery systems, so far resulting in only modest gene editing efficiencies (21,25).
Here, we deploy a single baculoviral vector (BV) encoding all the required components, achieving high efficiency HITI, single and multiplexed prime editing in a range of human cell lines. We exploit our approach to correct a genetic defect in SRNS patient derived podocytes. By achieving site-specific integration of very large DNA payloads and multiplexed prime editing mediated trinucleotide insertion at four different loci we unlock baculovirus as a vector of choice for next-generation genome engineering approaches.

LR recombination
LR recombination was carried out using LR Clonase II (Thermo Fisher #11791020) or LR Clonase II plus (Thermo Fisher #12538120). Although LR Clonase II plus is specifically designed for MultiSite Gateway recombination, LR Clonase II worked with comparable efficiency in our hands. LR reactions were carried out following manufacturer's instructions. One DEST and four ENTR vectors Nucleic Acids Research, 2022, Vol. 50, No. 13 7785 were diluted to 20 femtomoles/l each in TE buffer pH 8.0 (Thermo Fisher #12090015). 1 ul of each diluted vector was added to a 0.2 ml PCR tube with 2 ul LR Clonase II and 3 ul TE buffer, followed by a brief spin and incubation at 25 • C for 16 hours. The next day the reaction was terminated by addition of 1 ul proteinase K (provided with LR Clonase II enzymes) and incubation at 37 • C for 10 min. 2-3 ul were transformed into homemade electrocompetent Top10 or Pir + E. coli, followed by 2 h recovery at 37 • C and plating on LB/agar plates with the appropriate antibiotics. LR recombination products were predicted using APE (37) with custom recombination reactions. To quickly load Mul-tiMate LR reaction prototype in APE, the following code can be copied and used in Tools/Recombination Reaction Editon/New reaction from clipboard: MultiMate LR Reaction for APE: {ApE recombination reaction:}{MultiMate LR Reac- When imported as GenBank or Fasta files, the Mul-tiMate LR reaction in APE will automatically recognize pMMK ENTR1-4 vectors and one DEST donors when launched through Tools/Recombination tools. For manual prediction of MultiMate assembly products, a list of the attL/R sequences and their attB products is provided in Supplementary Table S3.

Plasmid propagation and stability
Upon LR recombination, plasmids assembled through MultiMate were propagated in Top10 or DH5␣ E. coli under standard culturing conditions (37 • C) in LB broth supplemented with the appropriate antibiotics. We assembled plasmids up to 33 kb in size and containing a number of identically repeated elements (e.g. promoters, terminators), but did not observe spontaneous recombination with any of our assemblies.
Recombination in E. coli systematically occurred for MultiMate HITI-2c vectors in absence of an AcrII4 module. Additionally, empty pMMK ENTR4 and pMKK ENTR AcrII4 modules occasionally recombined during bacterial propagation. This is due to the similarity and proximity of attL5/2 in pMKK ENTR4 and to the constitutive bacterial expression of AcrII4 in pMMK ENTR AcrII4. When the distance between attL5/L2 was increased by cloning an intervening cassette (e.g. pMMK4 CMV eGFP), no recombination events were observed for pMMK ENTR 4 modules. When the pMMK ENTR AcrII4 modules were assembled into MultiMate HITI-2c vectors, no additional recombination of the AcrII4 cassette was observed.

Cre-mediated recombination of DNA elements
One acceptor and one or multiple donor vectors were assembled using Cre-mediated recombination as previously described (33,(38)(39). One acceptor and one or more donors were mixed with a ratio of 1:1.1 in distilled H2O with 0.5 ul (7.5 U) of CRE recombinase (NEB # M0298M) and 1 ul Cre buffer (provided with CRE recombinase) to a final volume of 10 ul in distilled H 2 O. 500-1000 ng of total DNA were used for each reaction. Cre-reactions were incubated for 1 hour at 37 • C, followed by heat inactivation at 70 • C for 10 min. 2-3 ul were transformed into homemade electrocompetent Top10 E. coli, followed by 2 h recovery at 37 • C and plating on LB/agar plates with the appropriate antibiotics. Cre-recombination products were predicted using Cre-ACEMBLER Vers. 2.0 (38).

Cell culture methods
Human cells (HEK293T, HeLa, H4, RPE-1 hTERT and SH-SY5Y) were purchased from ATCC and propagated as adherent cultures in 60 or 100 cm dishes in a humidified incubator (37 • C, 5% CO 2 ). For passaging cells were washed with phosphate saline buffer (DPBS, Gibco # 14190144), detached using 0.25% Trypsin (Thermo Fisher #25200056) followed by a brief incubation at 37 • C, centrifuged at 300× RCF and resuspended in fresh media in a new plate at the desired concentration. Podocyte cell culture was carried out as previously described (40). Suspension cultures of Sf21 insect cells were grown in 125 ml or 250 ml polycarbonate Erlenmeyer flasks with vent cap (CORNING, #431143, #431144) at 27 • C in a shaking incubator. Sf21 were split every 2-3 days and maintained at concentrations between 0.5-2 × 10 6 cells/ml. Origin and media formulation recipe for each cell line is reported in Supplementary Table S4. For transfection in HEK293T, 2 × 10 5 cells/well were seeded in multi-24 wells. Transfections were carried out using Polyfect (QIAGEN #301105), following manufacturer's instructions. Briefly 500 ng of DNA were resuspended in 25 ul of Optimem (Gibco #31985062), supplemented with 5 ul Polyfect and incubated for 15 min at room temperature. Transfection mix was resuspended with 100 ul of complete media and added dropwise to each well. Cells were cultured for at least 48 hours before assessing the phenotype (e.g. fluorescence markers expression).
For puromycin selection of HITI-2c edited cells, puromycin dihydrochloride (Gibco A1113803) was used at a final concentration of 1 g/ml. Puromycin was added at 5-7 days post-transfection/transduction for 7 days. After selection, cells were maintained in absence of puromycin for at least 1 week prior downstream analysis. For puromycin/hygromycin selection of large DNA payload editing events, cells were first selected with puromycin as described above, followed by 10 days selection with 1 g/ml puromycin and 250 g/ml hygromycin (Hygromycin B, ThermoFisher #10687010). Upon completion of double selection, cells were maintained in absence of puromycin and hygromycin for at least 1 week prior downstream analysis.

Baculovirus vector amplification
Assembled MultiMate vectors were shuttled on baculovirus genomes (bacmids) propagated in E. coli using Tn7 transposition. 200-1000 ng of MultiMate vector were transformed in chemically competent DH10-MultiBacMam-VSV-G (21), DH10-EMBacY (33) or commercial DH10Bac (ThermoFisher # 10361012) as previously described (33). DH10-EmbacY were used to generate baculoviruses for multiprotein expression in insect cells and low MOT transduction of HEK293T, DH10-MultiBacMam-VSV-G were used for high MOT human cells transduction.DH10Bac were used for high MOT transduction of single and multiplexed prime editing constructs in human cells, by providing an additional module encoding for VSV-G in insect cells. DH10Bac were included in this study as they are widely wide used,and to demonstrate their compatibility with MultiMate. DH10-MultiBacMam-VSV-G are however preferrable for mammalian cells transduction, as the VSV-G module is already integrated in the baculoviral genome (21), alleviating the need for additional DNA assembly. Bacteria were streaked on LB/Agar plates with Gentamycin, Kanamycin, Tetracyclin, IPTG and Bluo-Gal and incubated for 24-hours for blue-white screening. White colonies were picked and grown overnight in 3 ml of LB supplemented with Gentamycin/Kanamycin to extract bacmid DNA through alkaline lysis/ethanol precipitation as previously described (32,33).
For transfection in insect cells, 0.8-1 × 10 6 Sf21 cells/well were seeded on multi-6 well plates in 3 ml of Sf-900 II media. 10 ul of purified bacmid were resuspended in 130 ul Sf-900 II media with 10 ul X-treme XP transfection reagent (Roche # 06366236001) and incubated at room temperature for 15 min. The entire transfection mix was added dropwise to a single well and cells were incubated at 27 • C in a static incubator. V 0 viral stocks were harvested collecting the supernatant of transfected cells 72-96 hours post transfection as previously described (19,32,47). 1-3 ml of V 0 viral stocks was added to 10 ml of fresh Sf21 cells at 0.8 × 10 6 cells/ml. Cells were cultured in 50 ml Falcon tubes while shaking at 27 • C and counted every day to monitor cell proliferation and size using Luna cell-counter (Logos-Bio). Successfully infected cells displayed arrested proliferation and increased average cell size (13-14 m control, 16-20 m infected). V 1 viral harvest were collected as previously described 2 days after proliferation arrest (DPA + 24) (19,32,47) by centrifugation at 4500 × rcf. 500 l/1 ml of V 1 viral stocks was added to 50 ml of fresh Sf21 cells at 0.8 × 10 6 cells/ml in 125 ml Erlenmeyer flasks and cells were cultured at 27 • C in a shaking incubator. V 2 viral harvests were collected by centrifugation at 4500 × rcf, and concentrated 20 times by high-speed centrifugation at 11000 × rcf, followed by resuspension in DPBS supplemented with 3% heat inactivated FBS and 1% glycerol for storage at −80 • C.

Baculovirus vector titration and transduction
For BV expressing fluorescent markers in human cells, titration was performed as previously described (41). HEK293T were used to determine viral titers. Briefly 1 × 10 5 cells/well were seeded in multi-48 wells plates in 200 ul of complete media. Concentrated virus was serially diluted in DPBS and 50 l were dispensed to each well. Spinoculation (30 at 600 × rcf at 27 • C) was used to enhance transduction as previously reported (42). Twenty four hours after transduction, cells were analysed using flow-cytometry to determine the percentage of transduced cells. TU/ml values from dilutions giving <20% transduction efficiencies were averaged to estimate the titer as previously described (41) and using Supplementary Equation 1. For experiments in which different viral titers were used, multiplicity of transduction (MOT) was calculated as TU/Cn. Transduction in various cell lines was performed. 2 × 10 5 cells per well were seeded in multi-48 wells chambers in 200 l DPBS, 50 ul of diluted virus at the appropriate multiplicity of transduction (MOT) were added and cells were spinoculated as described. After spinoculation, viral supernatant was removed, the cells were detached by trypsinization and replated to multi-24 wells. For podocytes transduction, BacMam enhancer (Thermo Fisher # B10107) was added at 1:1000 dilution for 24 h where indicated and removed by 3× DPBS washes.

Confocal and widefield imaging
Confocal images were acquired using a Leica Sp8 equipped with 405, 458, 476, 488, 496, 514, 561, 594, 633 nm laser lines and 37 • C stage. For time lapse confocal experiments on living cells the stage was supplemented with 5% CO 2 . For higher magnification, cells were plated on Lab-Tek borosilicate multi-8 wells (Thermo Fisher # 155411). For experiments in which transduced cells expressed more than three different fluorochromes, laser intensity and detection filters were adjusted to reduce spectral overlap using individual fluorescence controls transfections. Widefield and phase contrast images were acquired using a Leica DMI6000 equipped with excitation/emission filters optimised for DAPI, GFP, Rhodamine, Texas Red and Far red.

Flow cytometry analysis
For flow-cytometry analysis cells were trypsinised and resuspended in complete media supplemented with 3 M DRAQ7 (Abcam #ab109202) to counterstain dead cells. Cells were analysed on a Becton Dickinson LSR Fortessa X-20 (4 lasers 16 colours, HTS), fluorochromes were detected as follow: eGFP and EYFP (FITC-A), mCherry (PECF594-A), mTagBFP (BV421-A), DRAQ7 (AlexaFluor700-A). SSC-A and FSC-A were used to discriminate single cells and cell populations by size. FlowJo X was used to analyse FCS files. All data represented are percentages of live single cells (DRAQ7-).

PCR genotyping, Sanger sequencing and deconvolution
Genomic DNA was extracted with QIAamp DNA Mini Kit (QIAGEN # 51306) following manufacturer's instruction. A list of the predicted gene editing outcome sequences and genotyping oligos is provided in Supplementary Table S2. PCRs were performed using KAPA2G Fast Genotyping mix (SigmaAldrich # KK5621) following manufacturer's instruction. Amplicons were run on 0.8% agarose gels, purified using QIAquick Gel Extraction Kit (QIA-GEN # 28706) and eluted in distilled ddH 2 O. For Sanger sequencing 15 ul of purified PCR at 5-10 ng/l were mixed with 2 l of diluted (10 M) sequencing primer and sent to an external sequencing service (Eurofins). Electropherograms (.ab1) from parental and transduced cells were fed into ICE (43) from Synthego for sequence deconvolution and indels/knock-in estimation.

Western blot
Total protein extracts from HEK293T were obtained by lysing the cells with ice-cold RIPA Buffer (Thermo Fisher Nucleic Acids Research, 2022, Vol. 50, No. 13 7787 # 89901) supplemented with protease inhibitors (Thermo Fisher # 78429) for 30 on ice. Insoluble material was pelleted by centrifugation at 16 000 × rcf at 4 • C for 5 min. Total protein extracts from insect cells were obtained as previously described (32). Proteins concentrations were determined using Nanodrop. 10 g protein/sample were stained with Laemmli buffer, boiled at 95 • C for 5 min, separated using pre-cast NuPage 4-12% Bis-Tris SDS-Gels (Thermo Fisher # NP0321BOX) and transferred to PVDF membranes using iBlot. Membranes were blocked with 5% nonfat dry milk in T-TBS (50 mM Tris-Cl, pH 7.6, 150 mM NaCl, 0.5% Tween) for 1 h at room temperature. Membranes were incubated with primary antibodies diluted 1:1000 in T-TBS 5% milk overnight at 4 • C while rocking, followed by two T-TBS washes and incubation with HRPconjugated secondary antibody diluted 1:2000 in T-TBS 5% milk for 1 h at room temperature. Membranes were washed again with T-TBS and developed using Pierce ECL reagents (Thermo Fisher # 34579) following manufacturer's instructions. Finally, membranes were imaged using MyECL Imager. A list of the primary and secondary antibodies used is provided in Supplementary Table S5.

Immunofluorescence
For immunofluorescence, cells were seeded on glass coversleep and fixed with 4% paraformaldehyde ( Figure 4G) or methanol ( Figure 4H) for 15 min at room temperature or at −20 • C, respectively. Cells were washed twice with DPBS following incubation in blocking solution containing 3% BSA (Sigma) and 0.1% Triton X-100 (Sigma) for 1 h at room temperature. The cells were then left overnight at 4 • C in 0.5× blocking solution containing the primary antibody. The next day, the cells were washed three times with DPBS and then incubated with the secondary antibody for 1 h at room temperature in PBS. Coverslips were mounted on glass slides using Vectashield + DAPI (2bscientific # H-1200-10) and imaged using confocal microscopy. A list of the primary and secondary antibodies used is provided in Supplementary Table S5.

CCT/TriC complex purification
Recombinant MultiMate-CCT BV was produced as previously described (33) and used to infect Sf21 insect cells at a cell density of 1.0 × 10 6 /ml in Sf-900 II medium. Cells were harvested 72-96 h after proliferation arrest by centrifugation at 1000 × g for 15 min. Cell pellets were resuspended in lysis buffer (50 mM HEPES-NaOH, 200 mM KCl, 10 mM Imidazole, 20% Glycerol, pH 7.5, supplemented with EDTA-free protease inhibitor (Sigma-Aldrich) and Benzonase (Sigma-Aldrich)) and lysed by short sonication. The lysate was cleared by centrifugation at 18 000 rpm, 4 • C, in a F21-8 × 50y rotor (Thermo Fisher Scientific) for 60 min. The supernatant was loaded on a TALON column (Generon), equilibrated in TALON A buffer (50 mM HEPES-NaOH, 200 mM KCl, 10 mM Imidazole, 20% glycerol, pH 7.5) with a peristaltic pump. The column was washed with ten column volumes (CV) of TALON A buffer before eluting the bound protein complex with a step gradient of TALON B buffer (50 mM HEPES-NaOH, 200 mM KCl, 250 mM Imidazole, 20% glycerol, pH 7.5). The CCT protein complex was buffer exchanged in Heparin A buffer (50 mM HEPES-NaOH, 100 mM KCl, 10% glycerol, pH 7.5) while concentrating. It was then subjected to a Heparin column (GE Healthcare) and eluted with a 1 M KCl gradient in Heparin A buffer. Fully formed complexes and disassembled subunits were separated on a Superose 6 10/300 column (GE Healthcare) equilibrated in SEC buffer (20 mM HEPES-NaOH, 200 mM KCl, pH 7.5, 1 mM DTT, 10% glycerol). Peak fractions were pooled and concentrated and the purity of the CCT complex was analyzed by SDS-PAGE.

Electron microscopy
For electron microscopy, copper grids with carbon coating (300 mesh, Electron Microscopy Sciences) were glow discharged for 10 s, and 5 l of purified CCT was placed on the grids for 1 min. Afterwards the grid was washed for 15 s and floated onto a drop of filtered 3% uranyl acetate for 1 min. Excess solution on the grids was blotted off using filter paper between each step. Grids were visualized under a FEI Tecnai 20 transmission electron microscope (TEM), and digital micrographs were taken using a FEI Eagle 4K × 4K CCD camera. Particle picking and processing was performed using RELION 2 (44,45), 2D class averages were generated without applying symmetry or reference models.

Rapid assembly of highly modular baculovirus multigene vectors using MultiMate
We have previously developed methods to rapidly assemble functional DNA elements into multicomponent circuitry in a baculoviral vector (BV) (39,46,47). Here, we optimized and fine-tuned our DNA assembly approach (MultiMate) by combining proven MultiSite Gateway technologies with MultiBac (32,33)

Baculovirus-vectored homology independent targeted integration (HITI)
To date, baculovirus-vectored gene editing approaches were confined to CRISPR-HDR of small insert DNAs with low efficacy (21,(24)(25). HITI toolkits using a viral vector re-quired donor and Cas9/sgRNA to be split between two AAVs due to their limited cargo capacity (12), restricting successful gene editing to the fraction of co-infected cells. Moreover, manufacturing complete 'all-in-one' HITI vectors is not possible when viral packaging is performed in mammalian cells (typically HEK293T for AAV, LV), because simultaneous expression of Cas9 and sgRNA would inevitably excise the HITI donor, fatally compromising virus production. In marked contrast, BVs are manufactured in insect cells, where the mammalian promoters controlling Cas9 and sgRNA expression are poorly used and all-in-one HITI construct packaging into the BV is thus entirely feasible.
To minimize unpredictable indels and maximise correctly-edited alleles, we sought to analyse and compare HDR and HITI-2c (12) strategies by targeting the intronic ␤-actin (ACTB) locus, introducing a synthetic C-terminal exon fused to mCherry and a self-cleaving peptide (T2A) Nucleic Acids Research, 2022, Vol. 50, No. 13 7789 (51) followed by a puromycin selection cassette ( Figure  2A). We counteracted leaky bacterial Cas9 expression and potential leaky Cas9 expression in insect cells by outfitting our DNAs with the Cas9 inhibitor AcrII4 (34) under control of a dual prokaryotic and baculoviral promoter (J23119-polH), preventing vector self-cleavage during bacterial and viral amplifications stages. An additional CMV-eGFP module was added to track transduction efficiency. We next tested plasmid and BV-mediated delivery of MultiMate HDR and HITI-2c constructs in HEK239T cells. The HITI-2c strategy outperformed the HDR based approach (∼4-fold improvement), with a marked gain in editing efficiency when baculovirus transduction was used instead of plasmid transfection ( Figure  2B, C, Supplementary Figure S3a, b). The BV backbone was rapidly diluted as we expected (eGFP loss), while mCherry+ cells were stably maintained over time ( Figure  2B, C) with absolute gene editing efficiencies reaching ∼5% (HDR) and ∼20% (HITI-2c) (Figure 2b). Notably, when compared to plasmid transfection, baculovirus-vectored delivery increased absolute and relative gene editing efficiency up to ∼4and ∼2-fold respectively, regardless of the editing approach ( Figure 2B Figure S3g). For optimal efficacy, we prepared VSV-G pseudotyped BVs (21) and transduced HEK293T, HeLa, H4 and SH-SY5Y cells at different multiplicities of transduction (MOT) achieving higher transduction (up to 100%) and editing efficiencies (up to 30%) depending on the cell line ( Figure 2H, Supplementary Figure S3). To the best of our knowledge this is the first DNA assembly and delivery platform to enable efficient homology independent targeted integration in mammalian cells using a single all-in-one viral vector.

Safe-harbour integration of large DNA payloads
Precision docking of large multicomponent DNA circuitry in mammalian genomes remains an impeding challenge for currently available viral delivery systems which are constrained by their intrinsic packaging limitations (AAV: ∼4 kb; LV: ∼8 kb) (19,22). We assessed the aptitude of our system for precision DNA docking exploiting MultiMate-HITI-2c by repurposing the ACTB locus as a safe-harbour docking site for large DNA payloads, equipped with 5 and 3 fluorescent and selectable integration markers ( Figure  3A and Supplementary Methods). We used a new Cre insertion site to generate a series of HITI-2c payloads ranging from 4.7 kb to 18 kb with mTagBFP as a transduction efficiency reporter, resulting in all-in-one MultiMate plasmids of up to 30 kb ( Figure 3A). Transduction with EM-BacY BV (33) markedly outcompeted plasmid transfection and editing in HEK293T (Supplementary Figure S4a-d). Upon puromycin selection, cells remained >98% mCherry+ (Supplementary Figure S4e) confirming precise 5 -end integration. We observed silencing of the 3 -end fluorescent marker correlated with cargo size, which we could fully restore by hygromycin selection (Figure 3B, Supplementary Figure S4f). We confirmed correct integration by PCR genotyping and Sanger sequencing at pool level ( Figure 3C, D). To confirm integrity of the inserted DNA payload, we FACS sorted mTagBFP-/mCherry+/eGFP+ MultiMate-HITI-2c 18K-CGH transduced HEK293T in absence of any antibiotic treatment. 10 individual clones were cultured and expanded in absence of puromycin/hygromycin. At three months post-transduction, mCherry and eGFP expression were confirmed in all clones ( Supplementary Figure S4g). Seven out of nine clones were bona fide homozygous KIs (absence of wild-type ACTB amplicon) and all contained intact 5 and 3 junctions (Supplementary Figure  S4h). Large DNA payload integrity (18 kb) at the ACTB locus was confirmed by 6 spaced PCRs on the intervening DNA cassette. Nine out of 10 clones contained intact DNA payload, while only one clone lost a significant fragment possibly due to recombination (Supplementary Figure S4hi). We deployed MultiMate-HITI-2c 18K-CGH (30 kb) for baculovirus-vectored delivery with VSV-G pseudotyped BV achieving 100% transduction efficiency in both HEK293T and SH-SY5Y cells giving rise to 20% and 30% absolute genomic insertion efficiency, respectively ( Figure 3E). Of note, mCherry expression remained constant over time in the absence of any selective pressure while silencing of 3 eGFP expression ( Figure 3F, G) was again promptly restored by puromycin/hygromycin selection, remaining stable thereafter ( Figure 3H). Our results demonstrate that safe-harbour integration of extensive DNA payloads with base-pair precision can be achieved with high efficiency using all-in-one MultiMate-HITI-2c BVs, setting the stage for future large synthetic gene regulatory network engineering in human genomes.

Rescue of the disease causing R138Q podocin mutation in patient derived podocytes
Podocin is a key membrane scaffolding protein of the kidney podocyte essential for intact glomerular filtration. Mutations in the slit diaphragm protein podocin result in the most common form of monogenic steroid-resistant nephrotic syndrome (SRNS) (52-55. This disease manifests as early childhood onset of proteinuria, fast progression to end-stage renal disease (ESRD) and focal segmental glomerulosclerosis on kidney biopsy (FSGS) with no current treatment option. The most frequent podocin gene mutation in European children is R138Q, causing retention of the misfolded protein in the endoplasmic reticulum (ER) and degradation by the proteasome (40,56). Using temperature-sensitive transgene technology, we have developed human podocyte cell lines from both normal glomeruli (WT ciPods) and glomeruli from a kidney removed due to congenital nephrotic syndrome containing the R138Q mutation of podocin (PM ciPods) (40,57). To rescue NPHS2 expression in these cells we sought to deploy our large DNA payload integration strategy to dock a wild-type full-length NPHS2-Myc-Flag CDS downstream the ACTB locus under the control of a CMV constitutive promoter flanked by T2A::mCherry::P2A::Puro and CMV Hygro acting as 5 and 3 selectable markers, respectively ( Figure 4A). The resulting construct, (pMm HITI-2c NPHS, 22.6 kb) was tested by transfection in HEK2393T cells. Edited cells (mCherry+) were cultured in presence of Puromycin and Hygromycin for 1 week, showing correct expression and subcellular localization of NPHS2 by immunofluorescence and western blot ( Supplementary Figure S4j, k). Next, we transduced patient derived conditionally immortalized podocytes (PM ciPods) (57) with VSV-G pseudotyped BV pMm HITI-2c NPHS2. Although PM ciPods were readily transduced (>70% efficiency) ( Figure  4B), transgene expression levels (monitored through mTag-BFP expression) were weak and rapidly downregulated within 72 hours ( Figure 4B-E), leading to overall low gene editing efficiency ( Figure 4D). To counteract this premature silencing, we treated transduced cells with BacMam enhancer (BE), a histone deacetylase inhibitor widely reported to enhance BV-mediated transgene expression. Upon 24 h BE treatment, transduced PM ciPods retained 100% mTag-BFP expression up to 72 h post transduction ( Figure 4B, C, E), with gene editing efficiencies peaking at ∼40% (Figure 4D). Edited mCherry+ cells were detectable as early as 24 h post-transduction and stably maintained follow-Nucleic Acids Research, 2022, Vol. 50, No. 13  ing transient selection with Puromycin/Hygromycin (Figure 4F, Supplementary Figure S4l, m). Following transduction, NPSH2 expression in mCherry + cells was confirmed by immunofluorescence prior antibiotic selection using either ␣-Flag and ␣-NPHS2 antibodies. As expected (56), little to no expression of NPHS2 R138Q could be detected in untransduced PM ciPods (Figure 4G). To better characterize the subcellular localization of NPHS2 in engineered PM ciPods, we sought to compare them with WT ciPods transduced with lentiviral vectors overexpressing NPHS2 R138Q. While NPHS2 R138Q localized entirely at the endoplasmic reticulum, engineered PM Pods displayed NPHS2 localized at the ER and at the plasma membrane, demonstrating functional rescue of NPHS2 trafficking ( Figure 4H, top panel). Importantly, transgene expression remained active even after cells were allowed to differentiate by thermo-switching to 37 • C, demonstrating correct NPHS2 expression and subcellular localization under non proliferative conditions ( Figure 4H, bottom panel). Taken together, these results suggest that large cargo DNA payloads integration using all-in-one BV could be efficiently used to rescue recessive disease-causing 7792 Nucleic Acids Research, 2022, Vol. 50, No. 13  genes by integrating a functional copy at a safe harbour locus, in the future ideally under the control of tissue and cell specific promoters.

Highly efficient multiplexed prime editing
Base editors (BEs) (16,17) and prime editing (PEs) (18) are new additions to the CRISPR toolkit that could potentially correct up to 89% of the human disease-causing mutations in the absence of DNA cleavage (18). PE in particular, can be harnessed to precisely edit genomes with little to no indels production. PE exploits the nickase Cas9-H840A fused to reverse transcriptase from MMLV (PE2) to make a ssDNA copy of the engineered PegRNA at the edited site, allowing for the generation of all possible point mutations, insertions (up to 44 bp) and deletions (up to 80 bp) in the absence of DNA cleavage (18). Prime editing ef-ficiency can be additionally boosted by adding a nicking sgRNA (PE3) which nicks the non-edited strand. Both PE2 and PE3 rely on a Cas9-RT fusion which spans 6.3 kb of DNA (excluding promoter) and codes for a 240 kDa protein. While PE could be virally delivered only through multiple split-intein lentiviral vectors (18) due to cargo limitation, both PE2 and PE3 machinery are entirely within a single baculovirus cargo capacity. We first chose to insert a CTT trinucleotide in the HEK3 locus using prime editing (18). We therefore assembled MultiMate-PE2 HEK3 comprising PE2, HEK3 PegRNA cassette, VSV-G (for pseudotyping), aeBlue chromoprotein (36) (for visual readout of virus titer) and mTagBFP (for transduction tracking) ( Figure 5A, Supplementary Figure S5a). MultiMate-PE3 HEK3 was iteratively assembled by in-vitro CRE fusion ( Figure 5A). We next tested transfection-based delivery of MultiMate-PE2 HEK3 in HEK293T which, despite high  Figure 5C, D). While PE3 generally outperformed PE2, BV delivery resulted in almost 100% correct CTT insertion in HEK293T ( Figure 5C) after a sin-gle viral administration and in absence of any selective pressure. Given the higher susceptibility of HEK293T to BV transduction, we reasoned that this could attributed to a higher transgene expression in this cell line. To investigate this, we serially increased the transduction titer of MultiMate-PE2 HEK3 and observed a dose-dependent effect boosting correct editing events up to 45%, compared to the 25% of standard viral titers ( Figure 5C) without any detectable indels (Supplementary Figure S5f-i) indicating that, despite cell intrinsic factors affecting gene editing outcome, the PE2 expression levels and persistence in target cells are key to enhanced prime editing efficiencies. Since both PE2 and PE3 are well below the cargo capacity of BVs, we sought to explore multiplexed prime editing approaches to simultaneously perform trinucleotide insertions at four different loci. Since MultiMate correctly hosted up to seven CMV promoters ( Figure 1B), we chose to exploit individual hU6 promoters/sgRNA cassettes rather than polycistronic sgRNA expression modules (58). MultiMate-PE2 quadruplex was built to simultaneously target DNMT1, EMX1, RNF2 and RUNX1, while its PE3 counterpart was iteratively assembled by in-vitro CRE fusion with a MultiBac donor harbouring the respective nicking sgRNAs ( Figure  5E). Both MultiMate-PE2 and PE3 quadruplex BVs efficiently transduced a panel of immortalized human cell lines ( Figure 5E) with similar efficiencies to their single target counterpart ( Figure 5B) despite the increase cargo size. Despite overall high editing efficiencies, multiplexed PE2 did not always succeed in simultaneously correcting the four loci, presumably for a rate limiting expression of PE2, now occupied on four different targets ( Figure 5G, Supplementary Figure S5j). Efficient and simultaneous targeted PE3mediated trinucleotide insertions at the four loci could be readily detected in HEK293T (60-100%), RPE-1 (22-76%) and SH-SY5Y (15-64%) but unsatisfactory editing levels in HeLa ( Figure 5G, Supplementary Figure S5j), suggesting that transgene expression levels, DNA repair pathways and PegRNA efficiencies weight in more in multiplexed experiments than their single target counterpart. Perhaps due to Sanger sequencing deconvolution detection threshold (43), we were unable to detect low frequencies INDELs events at edited loci, although these are likely to be present as previously described (18). All together, these results provide the first evidence of a single viral delivery vector application for multiplexed prime editing approaches, a valuable tool for multiple gene correction or to establish polygenic disease cell-based models.

HDR and HITI, but not prime editing, promote undesired backbone integration regardless of the delivery system
Low backbone integration rate is an important safety feature when choosing a suitable delivery system for gene editing interventions. While baculoviruses have been shown to integrate at very low levels in mammalian cells (59), their long-term persistence when coupled with delivery of all-in-one HDR, HITI o prime editors toolkits has not been assessed yet. Indeed, gene editing approaches involving wild-type Cas9 and generation of knock-ins, regardless of the downstream DNA repair pathway used, nature of DNA donor (ssDNA, mini-circle, plasmid), or delivery system (AAVs, plasmid), have inevitably lead to frequent undesired integration of one or more DNA components (14,(60)(61)(62). While undesired integrations are seemingly unavoidable when using DSBs-producing Cas9, it is reasonable to assume that prime editing approaches, which rely on fusion of nickase-Cas9 with different enzymatic effectors (18), should promote little or no unspecific vector integration.
Since CRISPR toolkits are usually delivered through multiple vectors, fluorescent markers are rarely added on all the DNA or viral species involved, and ad hoc sequencing or genotyping experiments are usually designed to unveil the ratio of undesired backbone integration (14,(60)(61)(62). In contrast, our all-in-one delivery approach allows for real-time monitoring of transfection or transduction marker loss providing a readout of residual backbone integration events.
To assess backbone integration rates, we monitored fluorescent marker expression in HEK239T (eGFP or mTag-BFP) for up to 29 days following transfection or transduction with HDR, HITI-2c, PE2 quadruplex and PE3 quadruplex constructs ( Figure 6A-D). While eGFP expressed from a control plasmid was gradually lost over time, MultiMate-HDR and HITI-2c eGFP levels declined up to day 10, afterwards displaying stable eGFP expression in ≈1% and ≈9% of total cell population, respectively (Figure 6A). In contrast, mTagBFP expressed from MultiMate PE2 and PE3 quadruplex plasmid, was gradually lost over time ( Figure 6B). While the overall backbone integration rate for a standard eGFP control BV was ≈1%, MultiMate-HDR and HITI-2c BVs resulted in excess backbone integrations as in transfection experiments ( Figure 6C, Supplementary Figure S6a). By contrast MultiMate-PE2 and PE3 quadruplex BVs displayed integration rates similar to a control eGFP BV ( Figure 6D, Supplementary Figure S6a,b). When normalised for their initial transfection and transduction efficiencies, the backbones of HDR and HITI-2c constructs always integrated at higher rates compared to controls, notwithstanding the delivery system used (Figure 6E), while prime editing constructs residual marker expression levels were indistinguishable from the controls, for both transfection and transduction ( Figure 6E). Accordingly, PCR of gentamycin resistance (plasmid and BV specific) or gp64 (BV specific) on genomic DNA, confirmed backbone integration events above threshold for HDR and HITI-2c, but not for PE2 and PE3 quadruplex constructs ( Figure 6F), regardless the delivery system.
Cells transduced with MultiMate-HDR BVs resulted in excess eGFP integration mostly within the mCherry+ cells, while MultiMate-HITI-2c transduced cells displayed increased backbone integration rates in both mCherry + and mCherry-cells (Supplementary Figure S6a), most likely as result of donor excision through Cas9 activity. HDR and HITI-2c puromycin selected cells, revealed overall high levels of backbone integration, independently from the delivery method or the editing approach (Supplementary Figure  6Sc-e). In particular, within edited cells, both HDR and HITI-2c had similar levels of backbone integration upon transfection or transduction (25-30%), with BV-mediated delivery partially mitigating the backbone integration rates only for HDR (Supplementary Figure S6e).
Finally, we isolated eGFP-and eGFP+ clones from HEK293T transduced with BV MultiMate-HITI-2c and puromycin selected. Two out of three eGFP-clones where homozygous KIs (absence of WT ACTB band), while all the eGFP + clones were heterozygous. 5 integration was correctly detected in all clones, and BV integrations were only detectable in eGFP + clones (Supplementary Figure  S6f), confirming that eGFP late expression is a bona fide marker for backbone integration. In contrast, no BV integration events were detectable at clonal level in BV Mul-tiMate PE3 quadruplex transduced cells, again confirming that prime editors do not promote undesired backbone integration events.
Overall, we confirmed that basal baculovirus integration levels do not exceed 1%, rendering them well suited to de- liver different CRISPR technologies. Backbone integration levels are however dependent on the editing technology of choice, and both HDR and HITI-2c, which rely on DSBsproducing Cas9 variants and DNA donors, result in exceedingly high levels of backbone integration regardless the delivery system. By contrast prime editors promote little to no backbone integration by either transfection or baculovirusmediated transduction.

DISCUSSION
The rapid evolution of CRISPR approaches for precise genome-engineering precipitated an urgent need for nextgeneration viral delivery vector systems with superior DNA cargo capacity, efficiency and safety as compared to the state-of-the-art (20). The limited packaging capacity of currently dominating viral vectors (AAVs, LVs) constrains precise genome editing interventions (19,22), requiring cotransduction with multiple viruses (12,18) or implementation of smaller Cas variants (CasX, Cas ) (63,64) even for simple Cas/sgRNA approaches. While these technologies can cope, although suboptimal, with delivery into cultured cells, the viral titer required for successful co-transduction in vivo will likely dampen gene editing efficiency. Additionally, the size of the genome editing intervention in gene replacement experiments, cannot physically be extended beyond the cargo capacity of the vector of choice. We demonstrated here the efficacy of our baculovirusvectored approach to overcome this limitation and tackle the CRISPR delivery challenge. We optimized DNA as-sembly (MultiMate) to facilitate vector construction and improve BV manufacturing. We implemented homology independent targeted integration (HITI) for precise insertion achieving efficient precision C-terminal tagging in the ACTB locus, markedly outcompeting HDR efficiency. Using our approach, we demonstrated precision safe-harbour integration of large multicomponent DNA payloads with outstanding efficiencies and immediate potential for synthetic biology applications. We show that correct integration, however, is not sufficient to ensure stable expression of all the integrated components, likely due to transgene silencing events. Reassuringly, we demonstrate that usage of 5 and 3 junctions-linked selectable markers (fluorescent proteins or antibiotic resistance cassettes) can efficiently select for phenotypically and genotypically correctly edited cells. Synthetic biology applications in primary cell lines however, will require testing of alternative promoters (65), different safe-harbour integration sites (66) and incorporation of DNA insulator elements (67,68) to counteract gene silencing and ensure functionality of multicomponent synthetic circuits in absence of selection. There is no indication that we reached the cargo limit of baculovirus-vectored delivery. In fact, given the wide variation in size of naturally occurring baculoviruses (69) we expect that delivery of DNA cargos exceeding 100 kb will likely be feasible, enabling insertion of entire metabolic pathways and gene regulatory networks in safe-harbour sites or elsewhere in genomes. The large DNA payload integration strategy could in future be used to rescue recessive disease-causing genes by supplementing cells with wild-type CDS integrated at a safe harbour site. In this regard, we provided proof-ofconcept rescue of NPHS2 expression in SRNS patient derived podocytes (57). By installing a wild-type NPHS2 copy in the ACTB locus, we rescued NPHS2 trafficking phenotype and expression levels, theoretically demonstrating the feasibility of this approach for any recessive disease-causing gene. While we chose to install a CMV driven NPHS2 expression module, tissue and cell specific promoters will need to be tested and implemented in future to ensure transgene expression at physiologically relevant levels only within specific cell types. While in vivo applications remain theoretical, this approach could be applied to ex vivo modification of stem or primary cells, for instance enabling more affordable CAR-T manufacturing.
Importantly, we demonstrate the utility of our approach for seamless search-and-replace gene editing by implementing recently developed prime editing (PE) technology (18). Particularly PE, is considered safer when compared to HDR or HITI-2c, which rely on DNA cleavage. Using baculovirus-vectored delivery, we achieved highly efficient PE3-mediated trinucleotide insertion (CTT) in the HEK3 locus with up to 100% efficiency in the absence of detectable indels by using a single viral vector, in contrast to cotransduction with four lentivectors as previously reported (18). Additionally, we provide evidence of the feasibility of multiplexed prime editing approach to simultaneously insert trinucleotide insertions with high efficiencies (up to 60-97%) and, again, using a single viral vector with up to 8 hU6 PegRNA/sgRNA expression cassettes. We foresee that updated prime editing variants as PEmax (70) or more complex seamless drag-and-drop approaches (71) could be readily implemented for baculoviral-mediated delivery in future, further expanding the range of genome engineering applications.
We further demonstrate here that baculovirus vectors are inherently safe and do integrate in mammalian cells at rates close or below 1%. However, we have found that editing approaches which rely on DSBs Cas9 variants, inevitably lead to excess backbone integrations, regardless of the delivery system. In our hands, up to 25%-30% of edited cells displayed long term retention of transfection/transduction marker in HDR and HITI-2c experiments, suggesting plasmid and viral backbone integrations triggered by DSBs in addition to correct edits. Accordingly, various degrees of backbone integration have been reported with HDR and HITI approaches, notwithstanding the delivery system or the nature of the DNA donor (14,(60)(61)(62). In marked contrast, prime editing approaches, which rely on nickase Cas9 variants, could be safely delivered by either plasmid or baculoviruses, and displayed similar integration rates to nonediting constructs. While these two classes of gene editing enable two distinct sets of interventions (gene replacement for HDR and HITI and gene correction for PE), it appears evident that HDR and HITI are best suited to synthetic biology or ex vivo gene editing, while PE applications could be more impactful on the short term for in vivo applications. A new class of editors that combines the safety of prime editing with the size of the intervention range of HDR and HITI has yet to emerge, although hybrid approaches combining prime editing and integrase technologies such as twinPE and PASTE (71,72), could provide enhanced safety and reduced backbone integration, while still allowing for large DNA editing. Although requiring additional modules, both technologies could be easily accommodated on a single baculovirus, while delivery through other viral or non-viral delivery system will prove even more challenging.
Taken together, our results establish baculovirus as a vector of choice for precision engineering of large DNA cargoes and single/multiplexed prime editing in mammalian cells, and we anticipate baculovirus-enabled large-scale genome interventions, even combining safe-harbour integration with concomitant, if needed multiplexed, base or prime editing strategies, enabling complex synthetic biology approaches and, in future, ex vivo gene editing in clinically relevant cells.
Several roadblocks, which we recently reviewed (73), still stand in the way of bona fide baculovirus applications in vivo. Compared to more common viral vectors, only a handful of reports exist for baculovirus mediated gene delivery in vivo (22,(74)(75)(76), none of them including clinical trials in humans. Human serum complement, for instance, is known to inactivate gp-64 enveloped baculoviruses in vitro. This has been addressed by engineering modified envelopes to overcome this limitation. While VSV-G is able to partially shield from mouse serum complement in vivo (74,75), it provides little to no protection against rat and human serum in vitro (75). Pseudotyping with decay accelerating factor (DAF) however, has been shown to effectively shield virions from human serum complement (77,78). On this note, complement shielding has been achieved also by functionalising BVs with magnetic nanoparticles, enabling efficient CRISPR editing in vivo in mice (76). In addition, promoter silencing (79,80) and intracellular innate immunity driven by the cGAS/STING signalling pathway inactivate baculovirus in cultured mammalian cells (81,82), representing a challenge for both ex vitro and in vivo gene delivery and editing approaches. Histone deacetylase inhibitors (79,80) and small molecule STING antagonist (83) have been used with varying degree of success to counteract these challenges in cultured cells. While these molecules could theoretically pave the way for baculovirus mediated gene editing ex vivo in the future, more work will be needed to engineer baculoviral vectors suitable for efficient gene delivery in vivo. Contrary to other viral vectors however, the large cargo capacity of baculovirus will easily enable the incorporation of transduction helper modules, pending the elucidation of the molecular mechanisms underlying their inactivation in mammalian cells.

DATA AVAILABILITY
All plasmid sequences are provided in Supplementary  Table S1. MultiMate-CellCycle, MultiMate-Rainbow and MultiMate-HITI-2c ACTB reagents will be made available for distribution by Addgene. Raw flow-cytometry data have been deposited under the following DOI: https://doi.org/10. 6084/m9.figshare.20110364.v1. All other reagents are available from the authors upon reasonable request.