Design of promiscuous sgRNAs
Candidate target sequences containing an NGG protospacer adjacent motif in the PCSK9 and Albumin genes in the mouse (mm10) genome were extracted using Cas-Designer [37]. The extracted sequences were aligned to the human genome (hg19) and only the aligned sequences with no mismatches were selected. The selected candidates were analyzed with Cas-OFFinder [18] and those with a diverse set of related sequences that contained different numbers of mismatches (ranging from 0 to 5 per site) and that were the most broadly distributed throughout the human and mouse genomes were chosen as targets.
Construction of plasmids for sgRNA and Cas9 expression
The Streptococcus pyogenes Cas9 sequence [38] and the designed promiscuous sgRNA sequences that target Albumin and PCSK9 were cloned into the AAV plasmid backbone used in a previous study [39] to create Cas9 (pAAV-Cas9) and sgRNA (pAAV-Albumin and pAAV-PCSK9) expression vectors. Cas9 expression is under the control of the CMV promoter and sgRNA expression is under the control of the U6 promoter. Guide sequences targeting FANCF, VEGFA, and HBB genes [14] were cloned into pRG2 vector (Addgene #104174).
GUIDE-seq
Human HEK293T cells and mouse NIH-3T3 cells were maintained in Dulbecco’s modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) (ATCC, HEK293T, CRL-11268; NIH-3T3, CRL-1658) and 1% penicillin-streptomycin at 37°C in the presence of 5% CO2. HEK293T and NIH-3T3 cells were subcultured every 72 h to maintain 80% confluency. For GUIDE-seq, 2x105 HEK293T cells were transfected with plasmids expressing sgRNA (500 ng, pAAV-Albumin or pAAV-PCSK9) and Cas9 (500 ng, p3s-Cas9HC; Addgene plasmid #43945) and 5 pmol dsODN using Lipofectamine 2000. 2 × 105 NIH-3T3 cells were transfected with plasmids expressing sgRNA (250 ng, pAAV-Albumin or pAAV-PCSK) and Cas9 (500 ng, p3s-Cas9HC; Addgene plasmid #43945) and 100 pmol dsODN using an Amaxa P3 electroporation kit (V4XP-3032; program EN-158). Transfected cells were transferred to a 24-well plate containing DMEM (1 mL/well) that had been pre-incubated at 37°C. After 72 h, genomic DNA was isolated using a QIAamp DNA Mini Kit (Qiagen).
1000 ng of purified DNA was fragmented using a Covaris system (Duty Factor: 10%, PIP: 50, Cycles per burst: 200, Time: 50 s, Temperature: 20 °C) and purified using Ampure XP beads (A63881). Sequencing libraries were generated from the DNA using an NEBNext® Ultra™ II DNA Library Prep Kit for Illumina (E7546L) per the manufacturer’s protocol. Next, the regions of the library containing dsODN sequences were amplified using dsODN-specific primers and sequenced using Miseq (Illumina, TruSeq HT Kit). The remaining procedures were as described previously [2]. For data analysis, GUIDE-seq (1.0.2; https://pypi.org/project/guide-seq/) was used, which is compatible with Python 3.
Construction of plasmids for sgRNA transcription and in vitro transcription reactions
To improve the yield and accuracy of sgRNA transcription, we modified a previously described method [40]. Briefly, sgRNA templates were generated by annealing two complementary oligonucleotides followed by PCR amplification. BamHI, BsaI, and KpnI restriction sites were attached to the ends of sgRNA templates with a second PCR. Tailed sgRNA templates were inserted into the pUC19 plasmid digested with BamHI and KpnI. sgRNA-encoding plasmids were linearized with BsaI, which resulted in proper sgRNA end sequences. Linearized plasmids were incubated with 7.5U/μl T7 RNA polymerase (NEB, M0251L) in reaction buffer (NEB, B9012S) containing 14 mM MgCl2 (NEB, B0510A), 10mM DTT (Sigma, 43816), 0.02U/μl yeast inorganic pyrophosphatase (NEB, M2403L), 1U/μl murine RNase inhibitor (NEB, M0314L), 4mM ATP (NEB, N0451AA), 4mM GTP (NEB, N0452AA), 4mM UTP (NEB, N0453AA), and 4mM CTP (NEB, N0454AA) for 8 h at 37 °C. Yeast inorganic phosphatase was included to enhance sgRNA synthesis. After the reaction, the mixture was mixed and incubated with DNase I to remove the DNA template; transcribed sgRNAs were then purified using a PCR purification kit (Favorgen, #FAGCK001-1).
Digenome-seq
Genomic DNA from HEK293T and NIH-3T3 cells was purified with a DNeasy Blood & Tissue Kit (Qiagen). Both types of genomic DNA (10 μg) were incubated with Cas9 protein (10 μg) and sgRNAs targeting Albumin and PCSK9 (10 μg each) in a 1-mL reaction volume containing NEB3 buffer [100 mM NaCl, 50 mM Tris-HCl, 10 mM MgCl2, 100 μg/mL bovine serum albumin (BSA), at pH 7.9] for 8 h at 37°C. Digested genomic DNA was then treated with RNase A (50 μg/mL, Qiagen) for 10 min to degrade sgRNAs and purified with a DNeasy Blood & Tissue Kit (Qiagen) again.
Genomic DNA (1 μg) was fragmented to the 300-bp range using a Covaris system (Life Technologies) and blunt-ended using End Repair Mix (Thermo Fischer). Fragmented DNA was ligated with adapters to produce libraries, which were then subjected to WGS using a HiSeq X Ten Sequencer (Illumina) at Macrogen. WGS was performed at a sequencing depth of 30–40×. DNA cleavage sites were identified using the Digenome 1.0 program [41].
In silico prediction of off-target sites
Genome-wide candidate off-target sites with fewer than seven nucleotide mismatches with the chosen sgRNAs were obtained using Cas-OFFinder (hg19). CROP scores (heuristic scores that indicate if the candidate off-target sites would be edited) were computed using the CROP prediction model and optimized parameters (https://github.com/vaprilyanto/crop) based on a previous paper [31]. CFD scores (percent activity values provided in a matrix of penalties based on mismatches of each possible type at each position within the guide RNA sequence) were calculated using “crisprScore” R package [32]. For both calculations, GX19 (GACATGCATATGTATGTGTG for Albumin and GAGGTGGGAAACTGAGGCTT for PCSK9) sgRNA sequences and X20 target sequences were used.
Extru-seq
In preparation for Extru-seq, the transcribed sgRNAs were refolded in 1X NEBuffer 3.1 reaction buffer (100 mM NaCl, 50 mM Tris-HCl, 10 mM MgCl2, 100 μg/mL BSA, at pH 7.9). sgRNAs were heated to 98 °C for 2 min, after which the temperature was lowered at a rate of 0.1 °C/s until 20 °C was reached. To reduce reaction inhibition from a high concentration of glycerol, Cas9 buffer (10 mM Tris-HCl, 0.15 M NaCl, 50% glycerol, at pH 7.4) was exchanged with elution buffer (100 mM NaCl, 50 mM Tris-HCl, 10 mM MgCl2, at pH 8.0). Buffer exchange was conducted through a 10K Amicon® Ultra-15 Centrifugal Filter (Millipore).
HEK293T and NIH-3T3 cells were harvested with 0.25% trypsin-EDTA and human bone marrow MSCs (BM-MSCs) were harvested with 0.05% trypsin-EDTA. Harvested cells were resuspended in Dulbecco’s phosphate-buffered saline (PBS). Buffer-exchanged Cas9 (800 mg) and refolded sgRNA (530 μg) were preincubated for 10 min at room temperature to form RNP complexes. (For multiplex Extru-seq, buffer-exchanged Cas9 (800 mg) and five different refolded sgRNAs (106 μg each) were used). 1 × 107 cells were mixed with 5000 nM RNP complexes in 1 mL 1X NEBuffer 3.1 reaction buffer (100 mM NaCl, 50 mM Tris-HCl, 10 mM MgCl2, 100 μg/mL BSA, at pH 7.9). To perform Extru-seq in the presence of SCR7, SCR7 pyrazine (Sigma, SML1546) was added (1 μM). After gentle pipetting, suspended cells were extruded 11 times through an 8-μm pore-sized polycarbonate membrane filter (Whatman) using a mini-extruder (Avanti Polar Lipids). The extruded sample was then incubated at 37°C for 16 h. Genomic DNA was purified from the extruded sample using a FavorPrep Blood Genomic DNA Extraction Mini Kit (Favorgen, #FAGCK001-2) after RNase A (2 mg/mL) was added to remove sgRNA and RNA. WGS was carried out at a sequencing depth of 30–40×. DNA cleavage sites were identified using the Digenome-seq standalone program (http://www.rgenome.net/digenome-js/standalone). Analysis filtering options were as follows: minimum depth, 10, minimum score, 0.05, and minimum ratio, 0.01; other options were the default. [As the developer of a new tool, we checked all the sites identified by Extru-seq with the Integrative Genomics Viewer (IGV). Some of the loci look as if they are false-positive candidates (that is, non-cleavage sites according to IGV (Additional File 4: Table S3)). These false positives were also observed in Digenome-seq. The relevant bam files are available at the NCBI Bioproject (https://www.ncbi.nlm.nih.gov/bioproject/) under accession number PRJNA796642.]
Assignment of off-target results of Digenome-seq and Extru-seq to CAS-OFFinder results
Unlike GUIDE-seq and CAS-OFFinder, the standalone Digenome-seq program does not have a sgRNA:off-target alignment function that provides information about the number of mismatches and type of bulge (DNA or RNA) between the guide and off-target site. [The web version of the Digenome-seq analysis tool (http://www.rgenome.net/digenome-js/#!) has an optional alignment function with an alignment score that does not provide any information about the number of mismatches or type of bulge.] Instead, we used CAS-OFFinder to identify off-target sites with up to seven mismatches and two bulges relative to the target sequence. The positions of the off-target candidates identified by Digenome-seq and Extru-seq were then compared with those identified by CAS-OFFinder so that the information about mismatches and bulge type from CAS-OFFinder could be assigned to the loci identified by Digenome-seq and Extru-seq.
Validation of candidate off-target sites using a human cell line
Human HEK293T cells were maintained in DMEM supplemented with 10% FBS (ATCC, CRL-11268) and 1% penicillin-streptomycin at 37°C in the presence of 5% CO2. To determine indel frequencies at candidate off-target sites, 2×105 HEK293T cells were transfected with plasmids expressing sgRNA (500 ng, pAAV-Albumin, pAAV-PCSK9, pRG2-HBB, pRG2-FANCF, or pRG2-VEGFA) and Cas9 (500 ng, pAAV-Cas9 or p3s-Cas9HC; Addgene plasmid #43945) using Lipofectamine 2000 (vendor, amount). The cells were incubated at 37°C for 3 days, after which genomic DNA was prepared using a FavorPrep Blood Genomic DNA Extraction Mini Kit (Favorgen, #FAGCK001-2). The deep sequencing data are available at the NCBI Bioproject (https://www.ncbi.nlm.nih.gov/bioproject/) under accession number PRJNA796642. We used the following criteria used by EDITAS Medicine [23] to determine whether the target was validated or false (Additional File 5: Table S4). First, the indel of the sample must be higher than 0.1% for the sample to be validated. Second, the treated/control ratio must be higher than 2.
AAV production
AAV8 carrying the desired cloned sequences (pAAV-PCSK9, pAAV-Albumin, and pAAV-Cas9) were produced by VigeneBioscience at large scale [1013 genome copies (GC)/mL]. The resulting AAVs were aliquoted and stored at −70 °C until use.
Animal studies
All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Yonsei University College of Medicine (IACUC number 2019-0215). C57BL/6 mice were maintained under a 12:12 h light-dark cycle.
AAV injection
Two forms of AAV8, respectively carrying pAAV-Cas9 and one of the two pAAV-sgRNAs (pAAV-PCSK9 or pAAV-Albumin), were delivered into C57BL/6 mice by systemic (intravenous) and subretinal injection. Both types of injections were performed at a 1:1 GC (pAAV-Cas9:pAAV-sgRNA) ratio. Each dose consisted of 2.5 × 1011 GC/animal for intravenous injections and 1.5 × 1010 GC/eye for subretinal injections.
For systemic injections, 7- to 9-week-old male mice received a 200-μl tail vein injection with a dose of 2.5 × 1011 GC of AAV8 diluted in PBS.
For subretinal injections, 7- to 9-week-old male mice were selected. Under general anesthesia, one pupil per mouse was dilated with an eye drop containing tropicamide and phenylephrine. The body temperature of the mice was maintained at 37°C with a heating pad during the experiment. A small incision was made with a 1/2 30G needle 1 mm from the limbus of the cornea. A Hamilton syringe with a 33G blunt needle, loaded with 2 μl of solution containing the AAV8 mixture, was inserted through the incision until the point at which resistance was felt (subretinal space). To prevent unnecessary tissue damage, we carefully and gently injected the volume, waited for 20–30 s to allow it to spread evenly, and then slowly removed the syringe. Antibiotic ointment was then applied to the surface of the eyeball. Four mice were used for each injection method and each sgRNA.
DNA preparation from harvested organs and tissues
Organs and tissues were harvested 2 weeks and 3 months after the injection. Animals were euthanized by cardiac puncture under isoflurane anesthesia at the experimental endpoint. The organs—including the eye, liver, spleen, lung, kidney, muscle, brain, and testis—were dissected, snap-frozen in liquid nitrogen, and stored at −70°C until further analyses.
In the case of the subretinal injections, the neural retina and retinal pigment epithelium (RPE) were separated and prepared. The cornea, iris, lens, and vitreous were removed from the enucleated eyeball. The remaining eye tissues were incubated in hyaluronidase solution at 37 °C, 5% CO2 for 45 min, and then incubated in cold PBS for 30 min to inactivate the hyaluronidase activity. Next, the eye tissue was transferred to fresh PBS and the neural retina was gently separated from the retina/RPE/choroid/sclera complex. The remaining RPE/choroid/sclera complexes were incubated in trypsin solution at 37 °C, 5% CO2 for 45 min, and gently shaken until the RPE sheets were fully detached. All separated RPE sheets and RPE cells were collected. The genomic DNA was extracted using a DNeasy Blood & Tissue Kit (Qiagen, Cat No. 69506) according to the manufacturer’s instructions.
Targeted deep sequencing
Genomic DNA from mouse RPE cells was amplified with a REPLI-g Single Cell Kit (Qiagen) according to the manufacturer’s protocol.
Target sites and potential off-target sites were analyzed by targeted deep sequencing. Deep sequencing libraries were generated by PCR. TruSeq HT Dual Index primers were used to label each sample. Pooled libraries were subjected to paired-end sequencing using MiSeq (Illumina).
Statistical analysis
Scores/sequence read counts were min-max normalized. In each population, the maximum value was normalized to 1 and the minimum value was normalized to 0. Wilcoxon Rank-Sum Test was performed on the samples in each intersection of Venn diagram to test the equality of score medians from two different groups. Results from the two-sided unpaired Mann-Whitney test calculated by Prism (version 9.4.1) are shown.
Authors’ contributions
J.K.L., S.B., J-S.K., Y.K., J.L., J.K.H., and J.E. supervised the research. J.K., M.K., A.J., G.H.H., M.J., and G.C. performed the experiments. W.H., U.K., and H.K. analyzed the data. The authors read and approved the final manuscript.