The clustered regularly interspaced short palindromic repeats (CRISPR) system has become a powerful biotechnological tool that found a variety of applications in fundamental research and pharmaceutics (1–8). However, therapeutic translation of the CRISPR system is severely hampered by the availability of efficient delivery systems (9–11). Among the class II CRISPR system, Cas12a has distinct properties from Cas9. For example, Cas12a has ribonuclease (RNase) III activity that allows multiplexed gene editing with one single CRISPR RNA (crRNA) array (12). Cas12a requires one single crRNA for target recognition, which is shorter than the engineered single guide RNA (sgRNA) for Cas9 (13). Cas12a recognizes T-rich protospacer adjacent motif (PAM) region, which offers an alternative to the G-rich PAM targeting of Cas9 (14). In addition, Cas12a shows less off-targeting effects in human cells than Cas9 (15). From the perspective of formulation design, the smaller size of Cas12a makes it easier for vector construction and delivery (16).
Current development of delivery systems for CRISPR has been focused on the viral carriers, where the safety concerns with viral carriers call for the development of efficient nonviral delivery systems (17–19). There have been a lot of efforts in developing nonviral delivery systems for CRISPR-Cas9 (20–28); however, only a few reports demonstrated the delivery of the Cas12a/crRNA ribonucleoprotein (RNP), including the use of gold nanoparticle (29) and electroporation (30) for local therapy. Thus, a biocompatible and efficient delivery method that allows systemic delivery of Cas12a/crRNA RNP to the target tissues is desirable (31). Compared with the plasmid DNA-based delivery of CRISPR (32, 33), the delivery of CRISPR system in the format of RNP allows quick onset of the genome editing process and more stringent control of dosages and reduces the risk of off-target cleavage (34, 35).
Here, we demonstrate a layer-by-layer–based formulation for hepatocyte-targeted delivery of the Cas12a/crRNA RNP, aiming to reduce serum level of cholesterol. We used DNA as building blocks and prepared self-assembled DNA nanoclew (NC) as a core particle to load Cas12a/crRNA RNP (Fig. 1A). The Cas12a/crRNA RNPs were programmed to target enhanced green fluorescent protein (EGFP) in a model cell line or the Pcsk9 gene in mouse hepatocytes. Then, we coated the Cas12a/crRNA/NC with a cationic layer of polyethyleneimine (PEI) for condensing the negatively charged core and promoting endosome escape through “proton-sponge” effect. To allow systemic administration of the assembly, an anionic polymer layer was coated (Fig. 1B), which has a charge reversal behavior upon the exposure to acidic environment (anionic to cationic). Further conjugation of a hepatocyte-targeted ligand, galactose, to the charge reversal layer enables enhanced hepatocyte targeting after systemic administration. Overall, this DNA NC–based charge reversal formulation could circulate in the blood with high biocompatibility because of the negative charge at physiological pH. After binding to the asialoglycoprotein receptor (ASGP-R) on hepatocytes and getting internalized, the acidic endosomal environment can trigger charge conversion of assembly to facilitate endosome disruption, releasing the loaded Cas12a/crRNA cargo into the cytosolic compartment. After nuclear transportation of the RNP through the nuclear localization sequence (NLS) fused on Cas12a, the Cas12a/crRNA complex can cleave the target locus in the genome and introduce insertions/deletions (indels) for gene disruption.
Preparation of the assembly
To prepare the genome editing assembly, we chose Cas12a isolated from Lachnospiraceae as the cargo (13). Cas12a with a molecular weight of ~145 kDa was expressed in Escherichia coli and purified by the Ni–nitrilotriacetic acid (NTA)–based chromatography (fig. S1A). The crRNA was transcribed and purified in vitro (table S1), and it was later complexed with Cas12a to generate the Cas12a/crRNA RNP. By using EGFP as a model target, the double-stranded DNA cleavage activity of Cas12a/crRNA RNP was confirmed by an in vitro DNA cleavage assay, where a linearized EGFP-encoding plasmid DNA was used as the substrate (fig. S1B). EGFP disruption efficacy was then confirmed in a U2OS.EGFP reporter cell line (36). Using Cas9/sgRNA-EGFP from our previous study as control (37), comparable EGFP disruption efficacy could be observed for Cas12a/crRNA-EGFP (fig. S2).
To load the Cas12a/crRNA-EGFP, we prepared DNA NCs that complement the EGFP-binding region of crRNA-EGFP. The DNA NCs were generated by rolling circle amplification (RCA) (37, 38), and the RCA templates were designed with varying complementarity to the crRNA (table S1) to optimize the interaction between the carrier and the cargo. A sham DNA NC (NC-0) and DNA NCs with 6 base pairs (bp) (NC-6), 12 bp (NC-12), 17 bp (NC-17), and 27 bp (NC-27) of complementation to crRNA-EGFP were synthesized (fig. S1C). To prepare the assembly of Cas12a/crRNA-EGFP/NC, Cas12a was first complexed with crRNA-EGFP at a molar ratio of 1:1. NC-27 was used as the representative DNA NC for investigating the process of the assembly. Cas12a/crRNA-EGFP was loaded into the DNA NC-27 at a molar ratio of ~0.7:1 (Cas12a/crRNA RNP:repeating unit in the DNA NC, weight ratio at 4:1). The Cas12a/crRNA loading reduced the zeta potential of DNA NC from −19.8 ± 1.5 mV to −26.5 ± 2.5 mV and increased the mean hydrodynamic size from 131.4 ± 6.2 nm to 140.2 ± 4.9 nm.
Then, a layer of PEI was coated and the ratio of PEI to DNA NC was optimized (Fig. 2A). The weight ratio of PEI to NC-27 was varied between 0:1 and 3:1. When the ratio was increased from 0.5:1 to 1:1, an abrupt change of zeta potential was observed (−9.1 ± 2.3 mV to 17.2 ± 3.1 mV) and the size was reduced from 97.6 ± 4.3 nm to 66.9 ± 2.1 nm. The weight ratio of 1:1 for PEI to NC-27 was chosen for the next step coating of the charge reversal layer. The hepatocyte-targeted charge reversal layer was obtained by modifying PEI with galactose (Gal) and 2,3-dimethylmaleic anhydride (DM) following reported methods (39, 40). Preparation of charge reversal polymer Gal-PEI-DM was confirmed by Fourier transform infrared (FTIR) spectroscopy and 1H nuclear magnetic resonance (NMR) (fig. S3). The polymer Gal-PEI-DM is capable of targeting hepatocytes in vivo and changing its surface charge from negative to positive under an acidic condition. When coating Gal-PEI-DM onto Cas12a/crRNA-EGFP/NC-27, the weight ratio between Gal-PEI-DM and PEI was optimized between 0:1 and 3:1 (Fig. 2B). Increasing the ratio of Gal-PEI-DM to PEI converted the charge of the assembly from positive to negative. The size and charge of the assembly reached an equilibrium when the weight ratio between Gal-PEI-DM and PEI approached 1:1 (80.6 ± 5.1 nm and −25.1 ± 3.4 mV). Assembly of Cas12a/crRNA-EGFP/NC-27/PEI/Gal-PEI-DM was confirmed by confocal laser scanning microscopy (CLSM), where Cas12a was labeled with the infrared dye AF647, Gal-PEI-DM was conjugated with the green dye fluorescein isothiocyanate (FITC), and NC-27 was stained with the blue dye Hoechst 33342 (fig. S4A). Morphology of the assembly was further confirmed with transmission electron microscopy (Fig. 2C), which was consistent with its hydrodynamic size.
In vitro charge reversal performance
For tissue-targeted delivery of gene editing therapeutics by systemic administration, it is desirable that the carrier remains negatively charged in physiological condition (pH 7.4) but shows positive charge once it reaches intracellular environment so that it will be able to escape from the endosome (pH 4.5 to 6.5). To mimic the process of systemic circulation and intracellular trafficking of the assembly, we used phosphate-buffered saline (PBS) at three pH gradients (pH 7.4, 6.5, and 5.5) to test the charge reversal behavior of the assembly. The negatively charged β-carboxylic amide group on Gal-PEI-DM is stable in alkaline condition, which can be hydrolyzed under acidic environment to generate secondary amines and recover the PEI structure. As shown in Fig. 2D, the assembly Cas12a/crRNA-EGFP/NC-27/PEI/Gal-PEI-DM was stable under pH 7.4, where it remained negatively charged after 24 hours of incubation. Generally, galactosylated nanoparticles administered through intravenous injection could reach hepatocytes in 1 hour (41, 42). Therefore, the stability of Cas12a/crRNA-EGFP/NC-27/PEI/Gal-PEI-DM at pH 7.4 is sufficient for efficient hepatocyte targeting. In contrast, the assembly recovered its positive charge under pH 5.5 within 1 hour of incubation, while those incubated under pH 6.5 started to recover their positive charge after 4 hours of incubation. The rapid charge recovery under acidic pH is favorable for an efficient escape from the endosome once the assemblies are internalized by the target cells. Hydrolysis of the β-carboxylic amide group on Gal-PEI-DM was further confirmed by 1H NMR (fig. S3B).
To evaluate the biocompatibility of Cas12a/crRNA-EGFP/NC-27/PEI/Gal-PEI-DM under physiological pH and its membrane-lytic activity in acidic environment, we preincubated the assembly in PBS buffers of different pH for 8 hours and then tested their hemolysis capability in vitro. As shown in Fig. 3A, Cas12a/cRNA-EGFP/PEI/Gal-PEI-DM pretreated in pH 5.5 showed ~20% hemolysis activity, and those pretreated in pH 6.5 showed ~7% hemolysis, while the assemblies incubated in pH 7.4 PBS did not cause significant hemolysis, confirming the biocompatibility and acid-triggered membrane-lytic activity of the charge reversal nanoparticle.
GFP disruption in vitro
To evaluate the performance of the assembly Cas12a/crRNA/NC/PEI/Gal-PEI-DM in vitro, we used U2OS cell lines for the investigation. We examined the intracellular trafficking profile of the assembly with an EGFP-free U2OS cell line that allows the tracking of the endo/lysosomes with LysoTracker Green (Fig. 3B and fig. S4B). After 2 hours of incubation, the assemblies reached intracellular compartment but most of them were visualized in the endo/lysosome. After 4 hours of incubation, it could be observed that the assemblies started to escape from the endo/lysosome and reach the nuclei. After 6 hours of incubation, the assemblies efficiently escaped from the endo/lysosome and accumulated in the nuclei.
After confirming the efficient endosome escape of the assembly, we studied the gene editing (EGFP disruption) efficacy of the assembly. We used U2OS.EGFP cell for this study, which is an engineered reporter cell line that has an integrated copy of destabilized EGFP in the genome, making it a suitable reporter for the genomic disruption of EGFP (36). The Cas12a/crRNA-EGFP RNP was designed to target the open reading frame (ORF) of EGFP, where successful cleavage of EGFP gene could introduce indels to the ORF of EGFP and cause the disruption of EGFP. To optimize the interaction between Cas12a/crRNA RNP and the DNA NC, we prepared various versions of the assembly, where the DNA NCs were programmed with different complementarity to crRNA-EGFP (NC-27, NC-17, NC-12, NC-6, and NC-0). Dosages of the assemblies were also optimized in terms of the concentration of Cas12a (25 to 200 nM). Quantitative analysis of EGFP disruption by flow cytometry (Fig. 3C) showed a Cas12a concentration–dependent EGFP disruption profile, where higher dosages of Cas12a/crRNA-EGFP lead to increased EGFP disruption. Furthermore, the complementation with the DNA NCs significantly affected the delivery efficacy of Cas12a/crRNA. NC-17–cored assembly showed the highest EGFP disruption efficiency, followed by NC-12 and NC-6. Considering the fact that the annealing temperature of RNA/DNA hybrids is mainly affected by the length of complementation, suggesting that the phenomenon of medium complementation between DNA NC and Cas12a/crRNA leads to better Cas12a/crRNA delivery might be caused by a balance of Cas12a/crRNA loading into and release from the DNA NC. High complementation (NC-27) might hamper the release of Cas12a/crRNA from the DNA NC, while low complementation (NC-0) might lead to inefficient loading of Cas12a/crRNA into the assembly.
Overall, assemblies that contained the DNA NC core, irrespective of the complementarity, showed enhanced performance compared to the formulation without the DNA NC core (NC free). Under the optimal condition where NC-17 was used in the Cas12a/crRNA-EGFP/NC-17/PEI/Gal-PEI/DM assembly and Cas12a was administered at 200 nM, approximately 82% of EGFP disruption was observed in U2OS.EGFP. Representative fluorescent microscopy and flow cytometry images were shown in Fig. 3D and fig. S5.
Under the same optimized condition, we used a control crRNA that does not target any gene in U2OS.EGFP and confirmed that the EGFP disruption was due to the specific targeting of crRNA-EGFP (fig. S6). To evaluate biocompatibility of the assembly, we stained the cells with a TO-PRO-3 live/dead stain. No significant cytotoxicity against U2OS.EGFP cells could be observed, indicating the high biocompatibility of the charge reversal assembly and confirming that the efficient EGFP disruption was not caused by cytotoxicity. To assess the formation of indels at the targeted genomic locus of U2OS.EGFP cells, we applied the T7 endonuclease I (T7EI) assay that can detect the mismatches between mutated and wild-type (WT) genetic sequences. As shown in Fig. 3E, efficient cut of the amplicons was detected that paralleled the analysis by flow cytometry.
Pcsk9 disruption in vitro
PCSK9 (proprotein convertase subtilisin/kexin type 9) is a liver-secreted protein that binds low-density lipoprotein (LDL) receptor, a key receptor that mediates endocytosis of cholesterol (43). Naturally occurring loss-of-function mutation of PCSK9 in some families significantly reduced their risk of cardiovascular disease (~88%) (44, 45), which is a leading cause of death in the United States (46). Furthermore, recent evidence also showed that molecules inhibiting PCSK9 are promising candidates for lowering cholesterol in the blood circulation (47, 48). The PCSK9 inhibitors alirocumab and evolocumab have received U.S. Food and Drug Administration (FDA) approval for controlling cholesterol, where the traditional statin therapy may not work satisfactorily (49).
For efficient Pcsk9 gene disruption, we optimized the crRNA selection by designing two crRNAs targeting exon2 (crRNA-exon2) and exon3 (crRNA-exon3) of Pcsk9. DNA NCs that have 17 bp of complementation to the crRNAs-exon2 and crRNA-exon3 were synthesized, generating NC-exon2 and NC-exon3. Following the formulation optimized for EGFP disruption, Cas12a/crRNA-exon2/NC-exon2/PEI/Gal-PEI-DM and Cas12a/crRNA-exon3/NC-exon3/PEI/Gal-PEI-DM were incubated with the model mouse cell line 3T3-L1. The assembly prepared with the control crRNA that does not target any locus of the genome (Cas12a/crRNA-control/NC-control/PEI/Gal-PEI-DM) was used as the control. With the T7EI assay, approximately 75% indel formation was detected with the crRNA-exon3–containing assembly, while the crRNA-exon2–containing assembly induced ~44% indels (Fig. 4, A and B, and fig. S7). Thus, crRNA-exon3 was chosen for the following studies.
After confirming the efficacy of the assembly in inducing Pcsk9 disruption in vitro, we tested the performance of the nanoparticle in vivo. PCSK9 is mostly secreted from hepatocytes, making liver the organ of interest. In the design of the charge reversal layer, we have conjugated galactose as a hepatocyte-specific targeting ligand on the polymer. Galactose could bind to ASGP-R specifically overexpressed by hepatocytes (50), which enables rapid removal of galactose-terminated nanoparticles from the circulation (51). To demonstrate the liver-targeted accumulation of the delivered Cas12a/crRNA RNP, ex vivo imaging of fluorescently labeled nanoparticles was carried out. We labeled Cas12a with Cy5.5 (Cas12a-Cy5.5), free Cas12a-Cy5.5/crRNA-exon3 RNP and Cas12a-Cy5.5/crRNA-exon3/NC-exon3/PEI/Gal-PEI-DM were administered. One hour after the intravenous injection, the mice were sacrificed for analyzing the distribution of the nanoparticles by fluorescence-based in vivo imaging (IVIS). It can be observed that free Cas12a/crRNA has a much higher rate of renal clearance and more Cas12a/crRNA was delivered by the assembly to the liver after 1 hour of administration (Fig. 4C). Quantification of IVIS imaging by the region of interest analysis (Fig. 4D) showed that 2.6-fold more Cas12a/crRNA reached the liver when it was delivered by the assembly.
Pcsk9 disruption in vivo and therapeutic efficacy
Then, we investigated the in vivo Pcsk9 disruption efficacy of Cas12a/crRNA-exon3/NC-exon3/PEI/Gal-PEI-DM. PBS and Cas12a/crRNA-control/NC-control/PEI/Gal-PEI-DM were used as the controls. The assemblies were administered by intravenous injection into C57BL/6 mice every other day for three times. One week after the last drug administration, peripheral blood and livers were collected for analysis. T7EI assay identified around 53% disruption at the on-target site (Fig. 4E), while examination of the top three potential off-target sites (OT-1, OT-2, and OT-3) of crRNA-exon3 showed no detectable off-target cleavage (Fig. 4F). The serum levels of secreted PCSK9 protein were significantly reduced (~88%) when the mice were treated with the assemblies delivering crRNA-exon3, while the mice treated with the assemblies delivering crRNA-control did not show significant changes to serum PCSK9 levels (Fig. 4G). The augmented PCSK9 reduction in serum over the gene disruption efficacy has also been observed in other studies (52). It is possibly due to the increased expression of LDL receptor after Pcsk9 gene disruption, which facilitated the removal of PCSK9 from the serum (53). As a result of the reduced PCSK9 level, serum cholesterol level showed a ~45% reduction after the gene disruption (Fig. 4H). Then, we looked into the biocompatibility of the assemblies by monitoring liver damage–associated biomarkers. Examination of the levels of serum albumin and alanine aminotransferase (ALT) showed that the assemblies are highly biocompatible for delivering either crRNA-control or crRNA-exon3, indicating that neither the assembly nor the targeted gene disruption caused any observable hepatic toxicity (Fig. 5, A and B). Biocompatibility of the liver targeted assemblies was also evaluated by hematoxylin and eosin (H&E) staining of the liver sections. No observable differences between the PBS and assembly-treated groups could be detected (Fig. 5C and fig. S8), suggesting that the assemblies are biocompatible to the liver and other major organs. Deep sequencing of the on-target editing region showed an overall indel frequency of 48%, which is consistent with the T7EI assay. Mapping the deep sequencing results to WT sequence showed that most of the indels were deletions at the 3′ end of the protospacer sequence. Cas12a/crRNA-exon3 caused deletions around the target site, which confirmed the high specificity of the delivered RNP (Fig. 5D). Size of the deleted nucleotides varied widely and −19 deletion was detected to be the most frequent indel (Fig. 5E), and the representative deletions were shown in Fig. 5F.
Acknowledgments: Funding: This work was supported by grants from start-up packages of UCLA and UNC/NC State. A.K. was supported by the NIH (R01GM126571 and R01HL140951). Author contributions: W.S. and Z.G. designed the experiments. W.S., J.W., Q.H., and X.Z. performed experiments. W.S., J.W., Q.H., X.Z., A.K., and Z.G. analyzed the data and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.