Enhanced genome editing efficiency of CRISPR PLUS: Cas9 chimeric fusion proteins

Efforts to improve CRISPR-Cas9 genome editing systems for lower off-target effects are mostly at the cost of its robust on-target efficiency. To enhance both accuracy and efficiency, we created chimeric SpyCas9 proteins fused with the 5′-to-3′ exonuclease Recombination J (RecJ) or with GFP and demonstrated that transfection of the pre-assembled ribonucleoprotein of the two chimeric proteins into human or plant cells resulted in greater targeted mutagenesis efficiency up to 600% without noticeable increase in off-target effects. Improved activity of the two fusion proteins should enable editing of the previously hard-to-edit genes and thus readily obtaining the cells with designer traits.

www.nature.com/scientificreports/ generated the longer deletions of DNA in Zebrafish relative to SpyCas9 alone 14 . In addition, a SpyCas9 chimeric fusion protein with the three prime repair exonuclease 2 (TREX2), a human 3′ to 5′ exonuclease involved in DNA repair, replication, and recombination, was also reported to increase mutagenic efficiency 15,16 . TEXT (Tethering EXonuclease T5 with FnCas12a)-a fusion strategy significantly increased the knockout efficiency of FnCas12a at multiple genomic loci in different human cell lines 17 . In case of the knock-in activity, when the human CtIP endonuclease was fused to SpyCas9, the chimeric protein resulted in more than two-fold increase in transgene integration efficiency relative to SpyCas9 backbone protein alone 18 . A dominant-negative mutant of 53BP1, DN1S, was fused to Cas9, and the Cas9-DN1S fusion proteins significantly blocked NHEJ events specifically at Cas9 cut sites and improved HDR frequency 19 . Rad52-Cas9 fusion strategies yielded approximately threefold increase in HDR during the surrogate reporter assays in human HEK293T cells 20 . Furthermore, the chimeric fusion of adenine or cytidine deaminases to nuclease-inactivated version of SpyCas9 resulted in base editing without DSB, concomitantly reducing unintended off-target effects 21,22 . Lastly, an engineered reverse transcriptase fused to a catalytically-impaired SpyCas9 and a prime editing-guide RNA (pegRNA) led to specific change of the base using the pegRNA as a repair emplate 23 .
Because the chimeric fusion proteins with the SpyCas9 were shown to widen the spectrum of capability of the site-specific nuclease SpyCas9.We attempted to test the accessary proteins that may weaken the binding energy of the chimeric proteins to the substrate DNA and thus increase a turnover rate to eventually enhance the genome editing efficiency. To this end, we first created fusion proteins in which the carboxy-terminus of SpyCas9 was linked with a DNA-modifying protein-either a 5′ to 3′ DNA exonuclease (RecE, RecJ, T5, or lambda) 24 , mung bean nuclease, or terminal deoxynucleotidyl transferase (TdT). T5 and RecJ exonucleases were reported to enhance FnCas12a, T5-FnCas12a increased the knockout efficiency of FnCas12a at multiple genomic loci in different human cell lines, while RecJ-FnCas12a decreased the knockout efficiency of FnCas12a 17 .
The green fluorescent protein (GFP) was also tested as a control. Of the seven tested, SpyCas9 fusion with RecJ (SpyCas9-RecJ, C9R) and surprisingly with GFP (SpyCas9-GFP, C9G) resulted in a marked increase in both mutagenesis and knock-in efficiency, while the off-target effects are not significantly increased relative to the SpyCas9 (C9) structure alone. We refer the two functionally enhanced fusion proteins of C9R and C9G to as CRISPR PLUS, and explore how efficiently they edit the DNA at different conditions such as in test tube, cultured cell line, primary human cells, and plant protoplasts.

Results and discussion
Increased genome-editing efficiency of Cas9 chimeric fusion with RecJ or GFP. To prepare Cas9 chimeric fusion proteins, we translationally fused the carboxy terminus of SpyCas9 to a modifier protein: 5′-to-3′ DNA exonucleases (including RecE, RecJ, T5, and lambda), mung bean nuclease, TdT, or GFP ( Fig. 1A) (S1 Text). The SpyCas9 chimeric fusion proteins, were expressed in bacterial cells and purified. The SDS-PAGE pattern shows that all chimeras were isolated as single proteins (S1 Fig), indicating successful fusion of the accessory proteins and purity to perform functional analysis. The SpyCas9-T5 protein was barely detectable (< 0.2 mg/L) due to its low expression level.
To investigate the activity of the seven fusion proteins, we performed an in vitro DNA cleavage assay with the therapeutically relevant loci examined by other colleagues for cross comparison 25 . The 1.5 kb PCR-amplified C-C chemokine receptor type 5 (CCR5) DNA was cut in vitro by C9 and the CRISPR PLUS proteins, 750 bp DNA fragments were produced (Fig. 1B). A 0.25 pmol of each SpyCas9 fusion protein and a 0.3 pmol of sgRNA CCR5 were mixed and incubated at room temperature for 15 min. A total of 250 ng of DNA was digested with preassembled RNP at 37 °C for 10 min. The enzymatic products were then quantified from the agarose gel using ImageJ and the corresponding enzymatic activities were calculated. C9 exhibited 14.1% cleavage activity, whereas C9R and C9G showed 40.6% and 26.3% activity, respectively (Fig. 1C). By contrast, fusion with RecE resulted in difficulty purifying the proteins possibly due to instability of C9-RecE chimeric fusion structure, to 1.3%, and fusion with lambda, mung bean nuclease, and hTdT did not significantly affect the C9 activity. Together, these results demonstrate that the fused fragments affect the overall enzymatic activities of the Cas9. Henceforth, we further analyzed only C9R and C9G due to their desired effects on activity.

Improved indel efficiency of CRISPR PLUS in the cells.
To check whether C9R induces DNA insertion and deletion (Indel) mutations more rapidly than does C9, we first confirmed that the RecJ moiety of C9R is functional in vitro and fluorescence activity of GFP domain of C9G (S2 Fig.). We then introduced C9, C9R, and C9G ribonucleoproteins (RNPs) targeting the CCR5 gene into HEK293T cells and stopped the incubation at 8, 16, or 24 h after transfection; cells were then harvested and subjected to a T7 endonuclease 1 assay, the DNA was resolved on a 2% gel, and the results of the three replicates were digitized using ImageJ. C9R and C9G reached 20% efficiency as early as 8 h, a level that was not achieved until at least 16 h for the C9 control (Fig. 1D). Relative to C9, CRISPR PLUS displayed approximately twofold increase in efficiency at 16 h.
Because the RecJ moiety of C9R as a 5′ to 3′ ssDNA exonuclease resects DNA from 5′ to 3′ direction, the deletion created by C9R at DSB could be longer that that by C9. To determine whether the increased indel efficiencies of the CRISPR PLUS proteins are associated with deletion size, we first analyzed the correlation between deletion size and the number of NGS-based deep sequencing reads (> 40,000 per locus) at the CCR5 target  (Fig. 1F), but significant increases in longer DNA deletions of 46-60 and 76-120 bp. Conversely, insertions tended to be shorter than those using C9 (Fig. 1G). Next, we evaluated the off-target effects of CRISPR PLUS. Five potential off-target sequences of the CCR5, HPRT1, and EMX1 genes were identified by Cas-OFFinder 7 (S3 Fig). For sequences with > 50,000 NGS reads, the error rates of CRISPR PLUS did not exceed that of conventional NGS sequencing error rate (0.01-0.1%) (S3 Fig), suggesting that the high on-target editing efficiency of CRISPR PLUS is accompanied by low off-target effects.
Taken together, we demonstrated that CRISPR PLUS consistently displayed greater efficiency in indel mutagenesis in three different genes of HEK293T cells, whereas the off-target effects are not increased as compared with C9 treatment. Now we turn to testing knock-in efficiency of the CRISPR PLUS proteins.  Supplementary Table 1. The CCR5 DNA was digested with the purified enzymes and resolved on an agarose gel. Successful digestion cuts the 1.5 kb substrate DNA into two 0.75 kb DNAs. (C) DNA band intensities in the gel image. Statistical comparison was made between C9 and each of the C9R or C9G proteins. Significance was indicated on top of each bar. One-way ANOVA using Tukey's multiple-comparison test. The fold change relative to C9 is indicated above each bar. (D) DNA insertion and deletion (Indel) efficiency of C9, C9R, and C9G for the CCR5 gene in HEK293T cells at 8, 16, and 24 h after RNP transfection as measured by a T7E1 assay. Averages from the three replicate experiments are plotted. Two-way ANOVA using Tukey's multiple-comparison test. (E) Indel efficiency of C9, C9R, and C9G for the CCR5, HPRT1, and EMX1 genes (left to right) in HEK293T cells. At 24 h (CCR5 and HPRT1) and 8 h (EMX1) post-transfection, cells were harvested and subjected to targeted deep sequencing. The experiments were performed in triplicate. NT group represent the NGS data from non-treated control. One-way ANOVA using Tukey's multiple-comparison test was performed using C9 data as control. (F) Distribution of the DNA deletion sizes after treatment with C9, C9R, and C9G. The number of deletion events around the protospacer site in the CCR5 gene was counted according to the 15 bp deletion intervals from targeted deep sequencing data (n = 3) data. (G) Distribution of DNA insertion sizes after treatment with C9, C9R, and C9G in the CCR5 gene. The number of insertion events was counted according to the 1 bp insertion intervals from targeted deep sequencing data (n = 3) data. 6-10 and 11-61 bp insertion intervals were also shown to include events with smaller numbers. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05, and n.s. indicates not significant. www.nature.com/scientificreports/ Enhanced homology-directed DNA repair (HDR) of CRISPR PLUS. In addition to indel mutagenesis, knock-in replacement of the sequence is the major area of CRISPR-Cas9 applications. We therefore investigated whether CRISPR PLUS improves HDR activity for the three target genes in HEK293T cells. The cell cycle was first synchronized using nocodazole 26 before transfection with both the preassembled RNP complex (protein:sgRNA = 30:60 pmol) and a 106-mer single strand oligodeoxynucleotide (ssODN) repair template (20 pmol) consisting of a 57-nt 5′ homology arm, 5′-CAT ATG -3′ NdeI restriction enzyme site, and 43-nt 3′ homology arm ( Fig. 2A). Each target site was amplified and analyzed using targeted amplicon sequencing, and the HDR efficiencies were evaluated by NdeI digestion of the amplified DNA (S4 Fig). C9 exhibited 20%, 23%, and 5% efficiency for the CCR5, HPRT1, and EMX1 sites, respectively; these values were 36%, 48%, and 12% for C9R and 36%, 39%, and 11% for C9G (S4 Fig). Therefore, relative to C9, CRISPR PLUS increased the knock-in replacement efficiency by 6-16%. When the efficiency was re-examined using targeted deep sequencing analysis, the efficiency of C9 for the CCR5, HPRT1, and EMX1 sites was found to be 26%, 27%, and 6%, respectively; that of C9R was 48%, 41%, and 13%; and that of C9G was 48%, 33%, and 12%, confirming that CRISPR PLUS increased the knock-in efficiency by 6-22% (Fig. 2B). It is likely that increased knock-in efficiency might be associated with the increased editing activity of CRISPR PLUS enzymes.
Increased multiplex genome editing efficiency in induced pluripotent stem cells (iPSCs) and plant protoplasts. Finally, to determine whether the enhanced genome editing efficiency might be appli- Comparison of homology-directed DNA repair (HDR) activity among C9, C9R, and C9G. HDR efficiency was evaluated by targeted deep sequencing of the CCR5, HPRT1, and EMX1 loci (left to right) after genome editing with the specific 106-mer template shown in panel a. (****P < 0.0001 and ***P < 0.001, one-way ANOVA using Tukey's multiple-comparison test). For each gene, fold change was calculated relative to C9 efficiency. (C) Four different RNPs, each preassembled with a different sgRNA (targeting B2M, CIITA, CTLA4, or PDCD-1) were simultaneously transfected into induced pluripotent stem cells (iPSCs) and incubated for 24 h, and the targeted deep-sequencing analysis was performed to determine the indel efficiencies (****P < 0.0001, ***P < 0.001, and **P < 0.01, two-way ANOVA using Tukey's multiple-comparison test). www.nature.com/scientificreports/ cable to induced pluripotent stem cells (iPSCs), a cell type that is recalcitrant to editing, we performed multiplex genome editing of the beta 2 microglobulin (B2M), class II major histocompatibility complex transactivator (CIITA), cytotoxic T-lymphocyte-associated antigen 4 (CTLA4), and programmed cell death protein 1 (PDCD1).
When it comes to regenerative medicine using universal iPSCs, deletion of B2M and CIITA in iPSCs suppresses expression of their human leucocyte antigen (HLA) Class I and Class II genes, respectively, such that the genome edited iPSCs can avoid host T cell-mediated clearance 27 . CTLA4 and PDCD1 are inhibitory receptors that take part in T cell exhaustion, therefore being the target of deletion when designing engineered T cells for cancer immunotherapy 28 . We first designed two sgRNAs for each of the four genes, tested in HEK293T cells, and selected the more efficient sgRNA from each pair, where sgRNA B2M -2, sgRNA CIITA -1, sgRNA PDCD1 -1, and sgRNA CTLA4 -1 were chosen for the iPSC experiments (S5 Fig).
After in vitro assembly of each of the four sgRNAs with effector proteins, namely, C9, C9R, and C9G, three sets of four RNPs were transfected into HEK293T before moving to iPSCs. At three days post-electroporation, T7 endonuclease I assay (T7E1) was performed. Three-day treatment resulted in indel efficiency 38-81% regardless of C9, C9R, and C9G (S6 Fig). In iPSCs, compared with C9, CRISPR PLUS increased indel efficiency 7.5-fold and 8.8-fold for B2M (Fig. 2C), 3.2-and 6.8-fold for CIITA, 3.9-and 11.5-fold for CTLA4, and 4.5-and 15.8-fold for PDCD-1, respectively, suggesting possibility to obtain multiplex homozygous mutations in the target genes within shorter time than using C9. Plant protoplasts are an example of difficult-to-edit cells partly because the nucleofection of the RNP complex damages the cells. Therefore, obtaining desired level of editing efficiency with a tolerable amount of RNP is a demanding technique. To determine whether C9R activity is reproducible in plants, C9 and C9R RNPs programmed to target α-1,3-fucosyltransferase 1 were transfected into protoplasts of the model plant Nicotiana benthamiana by the DNA-free genome editing method (as described in previous studies 29 ), and their editing efficiency were analyzed by NGS deep sequencing. Compared to C9, C9R had increased editing efficiency by 2.6-fold, (S7 Fig), suggesting that C9R activity is conserved regardless of the plant or animal system. Taken together, we have shown that the two CRISPR PLUS proteins display markedly increased efficiency in both indel mutagenesis and knock-in efficiency at the various conditions tested such as in vitro DNA cleavage, in plant protoplasts, human cells, and in iPSC cells. However, currently, we do not have a complete explanation about the mechanisms of action of the enhanced genome editing efficiency observed from CRISPR PLUS especially in the C9G protein. In fact, it was not too much surprising that the C9R showed greater efficiency relative to C9 alone, because RecJ moiety could contribute to leaving permanent mutation footprint at the site of DSB. However, it was indeed unexpected results obtained from C9G experiments.
To rule out any possibility of experimental errors, we examined the identity of the C9G protein and performed the editing experiments repeatedly with separately prepared batches of C9G, but obtained basically the same results. First, we double checked if our C9G protein surely has GFP moiety by examining the photochemical spectra of both C9 and C9G. Specific light absorption and emission of C9G peaked around 400 and 500 nm, respectively as expected, whereas C9 did not noticeably absorb or emit at these wavelengths (S2 Fig. B). Emission of green light from C9G protein confirms that we used the right proteins. In addition, it can be postulated that the GFP moiety might contribute to stabilizing the C9G protein, and the relatively stable C9G acts longer time than C9 to result in overall increase in genome editing efficiency. However, our data shows that the activity was found enhanced at as early as 16 h, and not significantly increased then after (Fig. 1D). On the other hand, it is also of note that the nucleotide deletion size of both C9R and C9G similarly decreased at smaller size (1-30 nt) but increased at larger section (46-60 nt) (Fig. 1F). Thus, the mode of action of C9G could be similar to C9R, but it needs further research to draw definitive conclusion about the enhanced genome editing activity of the CRISPR PLUS proteins.
One of the remaining hypotheses to explore through future research is that chimerically fused GFP might have transformed the Cas9 moiety to conformationally favorable ways, and this resulting C9G protein has increased the turnover rate. Conceptually opposite to the inhibitory activity of anti-CRISPR proteins including AcrIIA2 and AcrIIA4 30 , the fused GFP moiety might interact with Cas9 part so that the C9G proteins better interrogate PAM and target site or release from the target DNA after creation of DSB. Previously, it was shown that, depending on the linker structure, GFP moiety in the chimeric fusion with an acid phosphatase (Pho-C) could differentially affect the enzyme activity 31 . Lastly, understanding the crystal structure of C9 in complex with RecJ or GFP should give us better answer concerning how the chimeric proteins performs better in creation of DNA DSBs.

Conclusion
In conclusion, we have shown that CRISPR PLUS greatly enhanced genome editing efficiency while keeping the off-target effect low relative to C9 control. Three genes in human HEK293T cells, four in iPSCs, and one in plant cells were tested, and consistently showed that CRISPR PLUS exceeded the editing efficiency of C9. Given that the off-target effects are proportional to duration and concentration of the genome editing enzymes in the cells of treatment 11 , CRISPR PLUS would further decrease the possibility of off-target effects because CRISPR PLUS achieved a genome editing rate within shorter time with relatively lower concentration than those of C9. Greater efficiency is also useful when practicing multiplex genome editing because the possibility to have homozygous alleles simultaneously at the multiple loci equals to the product of the editing rate at an individual locus. Our results should represent breakthrough techniques for multiplex genome editing of the precious and hard-to-edit cells for therapeutics and expedite availability of the regenerative medicines closer.

Methods
Cloning of SpyCas9 fusion proteins. The SpyCas9 coding sequence (Addgene, #138566) was ligated into a pET28a vector (Merck Biosciences, Darmstadt, Germany) that had been linearized by NcoI and XhoI double digestion. The pET28a-SpyCas9 vector, which was named pNGP020, was again linearized by XhoI which was 9 bp downstream of the full-length SpyCas9 sequence. A BPNLS signal peptide and DNA-modifying proteins were introduced to the XhoI site in pNGP020 by digestion of the XhoI site; pNGP020-BPNLS was named pNGP021. Full-length RecJ exonuclease amplified from E. coli DH5α was cloned into the pNGP020 vector by Gibson assembly (NEB, Ipswich, MA), which was then named pNGP022. With the same manner, the GFP gene was amplified from modified GFP4 32 , and the SpyCas9-GFP-BPNLS was named pNGP056. RecE exonuclease was amplified from E. coli DH5α, and the SpyCas9-RecE-BPNLS was named pNGP024. T5 phage exonuclease was synthesized by GenScript (Piscataway, NJ, USA), and the SpyCas9-T5-BPNLS was named pNGP041. Lambda phage exonuclease was synthesized by IDT (Coralville, IA, USA), and the SpyCas9-Lambda-BPNLS was named pNGP097. Mung bean exonuclease was amplified from cDNA of mung bean roots purchased from a local grocery store, and the SpyCas9-mungbean exonuclease-BPNLS was named pNGP030. Full-length human TdT (NM_004088) was obtained from the GenScript cDNA library (OHu11174D), and the SpyCas9-hTdT-BPNLS was named pNGP072. The DNA sequences mentioned here are provided in Supplementary File 1. In vitro sgRNA transcription. To prepare sgRNAs for in tube cleavage assay, a double-stranded DNA template was generated by annealing two single-stranded oligonucleotides with complementary sequences (forward:  5′-AAT TTA ATA CGA CTC ACT ATAGGXXXXXXXXXXXXXXXXXXXXGTT TTA GAG CTA GAA ATA GCA  AGT TAA AAT AAG GCT AGT CCG TTA TCA ACT TGA AAA AGT GGC ACC GAG TCG GTG CTT TT-3′, reverse:  5′-AAA AGC ACC GAC TCG GTG CCA CTT TTT CAA GTT GAT AAC GGA CTA GCC TTA TTT TAA CTT GCT ATT TCT AGC TCT AAA ACXXXXXXXXXXXXXXXXXXXXCCT ATA GTG AGT CGT ATT AAATT-3′) at 95 °C for 5 min and 55 °C for 10 min. 'X's can be replaced with the spacer sequences for each sgRNA. The DNA was purified using QIAquick (Qiagen) columns, and the purified DNAs were then used as templates for a T7 in vitro transcription (IVT) reaction according to manufacturer's direction (MEGAshortscript T7, Invitrogen, Carlsbad, CA, USA). In vitro transcribed sgRNAs were DNase treated, precipitated using the ammonium acetate/ethanol method, and then resuspended in distilled water for use with the SpyCas9 and its chimeric fusion proteins.
In tube cleavage assay of the SpyCas9 fusion proteins. The CCR5 locus from HEK293T cells was amplified by PCR (Supplementary Table 5) with Q5 High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, USA). For RNP complex formation, 0.25 pmol of each SpyCas9 fusion protein and 0.3 pmol of sgRNA CCR5 were mixed and incubated at room temperature for 15 min. A total of 250 ng of DNA was digested with preassembled RNP at 37 °C for 10 min. Digested DNA was placed on ice, and the products were subjected to 2% agarose gel electrophoresis. Assessment of the digested DNA intensity was performed by ImageJ 33 . The digested fraction was calculated with the following formula: Digested fraction (%) = b/(a + b) × 100, where a represents the band intensity of the DNA substrate and b represents the digested DNA.
Cell culture and synchronization. HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM; HyClone) supplemented with 10% fetal bovine serum and cultured at 37 °C in a humidified 5% CO 2 incubator. HEK293T cells were seeded at 3 × 10 6 cell density in a 10-cm culture dish. To help improve the efficiency of the knock-in (described below), the cell cycles were synchronized by treatment with nocodazole (200 ng/ml) for 17 h before electroporation. Protoplasts (2 × 10 5 ) were transfected with SpyCas9 and C9R proteins (10 µg) premixed with in vitro-transcribed sgRNA (20 µg) targeting α-1,3-fucosyltransferase 1 (FucT 13-1) (Supplementary Table 4). Prior to transfection, SpyCas9 protein was mixed with sgRNA in 1 × NEB buffer 3 and incubated for 10 min at room temperature. A mixture of protoplasts was resuspended in 200 µl MMG solution and gently mixed with 10-20 µl of RNP complex and 210-220 µl of freshly prepared PEG (0.2 M mannitol, 40% w/v PEG-4000, 100 mM CaCl 2 ) solution and then incubated at 25 °C for 15 min. After a 15 min incubation at room temperature, transformation was stopped by adding 840-880 μl of W5 solution. Protoplasts were then collected by centrifugation for 2 min at 100g at room temperature, washed once with 1 ml of wash buffer, and then collected by centrifuging for another 2 min at 100g. The density was adjusted to 1 × 10 5 protoplasts/ml, and they were cultured in modified Protoplast