A general genome editing strategy using CRISPR lipid nanoparticle spherical nucleic acids

Significance

Genome editing holds broad therapeutic potential, but clinical translation requires efficient delivery of genome editors to target cells and tissues. We modified lipid nanoparticles (LNPs) with a dense shell of DNA to create CRISPR LNP–spherical nucleic acids (LNP–SNAs), which exhibit enhanced cellular uptake, biocompatibility, and transfection efficiency. LNP–SNAs enable both insertion–deletion (indel) formation, which disrupts gene function, and homology-directed repair (HDR) for precise, user-defined genome modifications. Across multiple cell lines and genomic targets, LNP–SNAs achieve indel rates of 15 to 68% and HDR efficiencies of 13 to 30%, representing a 1.5 to 3-fold improvement over LNPs. These findings establish CRISPR LNP–SNAs as an effective platform for genome editing.

Abstract

Genome editing with CRISPR–Cas systems hold promise for treating a wide range of genetic disorders and cancers. However, efficient delivery of genome editors remains challenging due to the requirement for the simultaneous delivery or intracellular generation of Cas proteins, guide RNAs, and, in some applications, donor DNAs. Furthermore, the immunogenicity and toxicity of delivery vehicles can limit the safety and efficacy of genetic medicines. Here, we combine two nucleic acid delivery approaches to create CRISPR lipid nanoparticle–spherical nucleic acids (LNP–SNAs) that are both efficient and biocompatible. Compared to lipid nanoparticles (LNPs) lacking a surface-bound DNA shell, CRISPR LNP–SNAs exhibit two- to three-fold higher cellular uptake, reduced cytotoxicity, and improved gene transfection efficiency. Across multiple cell lines and genomic loci, CRISPR LNP–SNAs induce insertion–deletion mutations at average frequencies two- to three-fold higher than those observed with LNPs. When codelivered with donor templates, CRISPR LNP–SNAs enable homology-directed repair at an average efficiency of 21 ± 7%, a 2.5-fold improvement over LNPs (8 ± 4%). The ease of synthesis and biocompatibility of CRISPR LNP–SNAs highlight their potential as a versatile delivery platform for CRISPR–Cas and other gene therapies.

Genome editing using CRISPR–Cas systems has emerged as a transformative approach to treating genetic disorders, infectious diseases, and various forms of cancer (1). A typical CRISPR–Cas system utilizes a Cas nuclease and a guide RNA to induce double-strand breaks (DSBs) at specific genomic loci (2, 3). These DSBs are primarily repaired through nonhomologous end joining (NHEJ) or microhomology-mediated end joining (MMEJ), both of which result in error-prone insertions and deletions (indels), or through homology-directed repair (HDR), a more precise pathway that requires an exogenous DNA template (4, 5). Precision genome editing has also been demonstrated using base and prime editors, which utilize engineered Cas variants and guide RNAs without inducing DSBs (6, 7). However, realizing the full therapeutic potential of these technologies requires a safe and efficient delivery vehicle capable of codelivering multiple protein and nucleic acid cargos, which has remained challenging.

Strategies for the delivery of gene editing machinery typically involve electroporation, microinjection, or complexation (with nanoparticles or cationic lipids), but such methods are often limited due to high toxicity or poor delivery efficacies (8–13). The use of adenoviral or retroviral vectors is constrained by insertional mutagenesis, immunogenicity, and payload size limitations (14). Viral-like particles (VLPs) represent an emerging class of delivery systems, but their immunogenicity, delivery efficiency, and scalability require further investigation (15–20). Lipid nanoparticles (LNPs) are promising nonviral alternatives (21–27) and have been employed in clinical trials (28, 29), but their use for multiplexed cargo delivery is underexplored (30). Moreover, recent studies highlight the proinflammatory effects of ionizable and PEGylated lipids, undermining the safety and therapeutic efficacy of LNPs (31). These challenges emphasize the need to develop biocompatible and effective delivery vehicles for gene editors.

Spherical nucleic acids (SNAs) are nanostructures composed of an oligonucleotide shell conjugated at high density (so the oligonucleotides stand upright and adopt the shape of the particle template) to the surface of a nanoparticle core (32–34). SNAs have been shown to penetrate over 60 different cell types to date as well as tissues (35, 36), and they have been found to be safe and efficacious in human clinical trials (37, 38). Previously, we developed CRISPR–SNAs composed of a GeoCas9 ribonucleoprotein core to induce indel mutations (39). However, modifying other Cas variants with a DNA shell is challenging. To expand editing capabilities beyond indel formation, we hypothesized that SNAs with LNP cores (40) could be used as a biocompatible and generalizable platform for genome editor delivery. Here, we report the design and synthesis of CRISPR LNP–SNAs capable of accessing multiple gene editing pathways with high efficiency. As a proof-of-concept, we encapsulated and delivered spCas9 enzymes and guide RNAs to induce indels, and in a separate application, codelivered these components with repair templates to perform HDR edits. We found that CRISPR LNP–SNAs exhibited enhanced cellular uptake, reduced cytotoxicity, and higher transfection efficiency compared to CRISPR LNPs, and improvements in delivery efficiency resulted in enhanced editing outcomes.

Results and Discussion

The LNP core used here consists of four components (Fig. 1A): ionizable cationic lipids (Dlin-KC2-DMA) that bind to anionic nucleic acids and facilitate endosomal escape, phospholipids (DOPC) that assist in encapsulation, cholesterol that provides structural integrity, and lipid-anchored polyethylene glycol (PEG) that controls size and colloidal stability (40, 41). The genome editor, spCas9 protein, and its single-guide (sg)RNA are encoded by a plasmid, with U6 and CBh promotors regulating the expression of the sgRNA and protein, respectively (Fig. 1B and SI Appendix, Table S1) (5).

Fig. 1.

Design of CRISPR lipid nanoparticle–spherical nucleic acids (LNP–SNAs). (A) Lipid components that assemble into LNPs. (B) Plasmid construct encoding a single-guide (sg)RNA and a spCas9 protein tagged with 3 × FLAG epitopes and nuclear localization signals (NLSs). (C) Synthetic scheme for CRISPR LNP–SNAs. (D) Scheme for CRISPR LNP–SNA cellular delivery and genome engineering.

To synthesize CRISPR LNPs, the lipid components were first dissolved in ethanol and then mixed with the plasmid in an acidic buffer, followed by dialysis (Fig. 1C). A repair template was included in the lipid and plasmid mixture to create HDR-accessible constructs. After optimization, we selected the formulation (12.5% DOPC, 50% Dlin-KC2-DMA, 2.5% DPPE-PEG–maleimide, and 35% cholesterol) that resulted in a LNP size of 110 ± 19.4 nm and a plasmid encapsulation efficiency of 84% (SI Appendix, Table S2). CRISPR LNP–SNAs were subsequently synthesized by reacting 5′ thiol-modified oligonucleotides with maleimides on lipid-PEG, where a guanine-rich sequence, 5′–T₄(TGG)₁₀–3′, was used to enhance cellular uptake via G-quadruplex formation (42). To expand the range of conjugation chemistries accessible, we also used an alternative lipid-PEG with an azide moiety (DPPE-PEG–azide) to facilitate conjugation of 5′ DBCO-modified, guanine-rich DNA (SI Appendix, Fig. S1).

The DNA shell of the SNA facilitates the delivery of the genome editors through scavenger receptor A and caveolae-mediated endocytosis, and endosomal escape is aided by the ionizable lipids (Fig. 1D) (41, 43). Once in the cytosol, the plasmid is transported into the nucleus to initiate transcription (44). Following cytosolic expression, genome editors tagged with nuclear localization signals (NLSs) enter the nucleus and associate with sgRNA to perform gene editing steps.

The CRISPR LNP–SNAs prepared using thiol-maleimide chemistry were characterized using gel electrophoresis (Fig. 2A). The CRISPR LNP–SNA sample was less mobile in the gel than the free DNA used as the LNP–SNA shell, indicating the formation of larger structures based on the chemical conjugation of the CRISPR LNPs with DNA. Surfactant treatment disrupts the lipid core of CRISPR LNP–SNAs, resulting in their dissociation into lipid–DNA conjugates, which have lower molecular weights. Dynamic light scattering (DLS) measurements confirmed a progressive increase in hydrodynamic diameter: LNPs < CRISPR LNPs < CRISPR LNP–SNAs (Fig. 2B). These size increases correlate to what would be expected given the sequential addition of plasmids and DNA shells. In addition, the zeta potential went as LNPs > CRISPR LNPs > CRISPR LNP–SNAs, with CRISPR LNP–SNAs being the most negatively charged (Fig. 2C), consistent with the conclusion, the higher the nucleic acid content, the more negatively charged the constructs (45). Similar trends in particle size and zeta potential were observed for the analogous system prepared using DPPE-PEG–azide and DBCO-modified DNA, indicating that this conjugation chemistry can also be used to synthesize CRISPR LNP–SNAs (SI Appendix, Figs. S2 and S3).

Fig. 2.

Characterization of CRISPR LNP–SNAs synthesized via the thiol-maleimide reaction. (A) Agarose (0.5%) gel electrophoresis showing the migration of Cy5-labeled 5′-thiol-DNA, CRISPR LNP–SNAs, and CRISPR LNP–SNAs ruptured using a surfactant (1% Triton X-100). (B) Hydrodynamic diameters of LNPs, CRISPR LNPs, and CRISPR LNP–SNAs measured using DLS. (C) Zeta potentials of LNPs, CRISPR LNPs, and CRISPR LNP–SNAs. The mean ± SD from three biological replicates is shown.

Particle morphology was characterized using transmission electron microscopy (TEM). Cryo-TEM images of maleimide-modified CRISPR LNPs revealed monolayer and bilayer structures; upon negative staining, lamellar rings were observed (SI Appendix, Fig. S4). CRISPR LNP–SNA structures were predominantly observed as bilayers, with cores that appeared more electron-dense than those of the CRISPR LNPs, likely due to the presence of the DNA shell (SI Appendix, Fig. S5). Similarly, lamellar structures were seen in samples of azide-modified CRISPR LNPs and CRISPR LNP–SNAs synthesized via cycloaddition (SI Appendix, Fig. S6).

We assessed the cellular uptake of rhodamine-labeled, maleimide-modified CRISPR LNPs and CRISPR LNP–SNAs using flow cytometry. Over a 4-h incubation period, epidermal keratinocytes (HaCaT) treated with CRISPR LNP–SNAs exhibited fluorescence intensities 2-to-3 times greater than those observed when cells were treated with CRISPR LNPs (Fig. 3A). Confocal microscopy images further confirmed the internalization of CRISPR LNP–SNAs with rhodamine-labeled lipids and Cy5-labeled, thiol-modified DNA (Fig. 3B). These data indicate that when the CRISPR LNP system is formulated as an SNA architecture via surface DNA modification, rapid cellular entry (compared to the CRISPR LNP system) occurs.

Fig. 3.

Cellular uptake, biocompatibility, and functional delivery of LNP–SNAs (prepared using thiol-maleimide chemistry). (A) Flow cytometry analysis of rhodamine-labeled CRISPR LNPs and CRISPR LNP–SNAs incubated with HaCaT cells over a 4-h period. (B) Confocal imaging of rhodamine- and Cy5-labeled CRISPR LNP–SNAs in HaCaT cells. (C) Cell viability assays for CRISPR LNPs and CRISPR LNP–SNAs across four different cell lines at different particle concentrations (0 to 40 nM) after a 24-h incubation. (D) Flow cytometry analysis of EGFP fluorescence following the incubation of HaCaT cells with an EGFP plasmid (pEGFP), EGFP LNPs, and EGFP LNP–SNAs (with encapsulated pEGFP). (E) Quantification of protein expression levels, as determined by EGFP fluorescence intensity. The mean ± SD from three biological replicates is shown.

Next, the cytotoxicity of maleimide-modified CRISPR LNPs and CRISPR LNP–SNAs was evaluated in multiple cell lines, including HaCaT, human bone marrow stem cells (hBMSCs), macrophages (RAW 264.7), and human embryonic kidney cells (HEK293T) (Fig. 3C). Following 24 h of treatment, CRISPR LNPs at concentrations as low as 10 nM resulted in a ~20 to 40% reduction in cell viability. In contrast, treatment with 40 nM CRISPR LNP–SNAs maintained cell viability above 80%, indicating that the SNA formulation mitigated the cytotoxicity typically associated with LNPs (46, 47). Prolonged treatment (96 h) with CRISPR LNPs further decreased viability by ~45 to 80%, whereas CRISPR LNP–SNAs maintained viability above 55% across all particle concentrations tested (SI Appendix, Fig. S7). These results suggest that, rather than designing alternative lipid formulations (48), modifying existing LNP systems with a DNA shell to create an SNA architecture offers a simple and effective route to enhance biocompatibility.

We quantified gene expression by measuring protein fluorescence following the delivery of maleimide-modified LNPs and LNP–SNAs, each encapsulating an enhanced green fluorescent protein (EGFP) plasmid (SI Appendix, Fig. S8, referred to as EGFP LNP and EGFP LNP–SNA, respectively). Flow cytometry analysis showed that treatment of HaCaT cells with the plasmid alone (without a delivery vehicle) did not result in a detectable increase in protein expression, as expected (Fig. 3D). Notably, protein expression with EGFP LNP–SNAs was found to be approximately three-fold higher than that observed with EGFP LNPs, mirroring the observed trend in cellular uptake (Fig. 3E). These findings indicate that plasmids delivered using LNP–SNAs can efficiently undergo intracellular transcription and translation.

NHEJ/MMEJ and HDR are the primary pathways for repairing DSBs that result in indel formation or templated repair, respectively (Fig. 4A). To assess the indel frequencies induced by CRISPR LNP–SNAs, three loci were selected: DNAse I hypersensitive sites, associated with transcriptional activity, the human GRIN2B gene, and its murine ortholog Grin2b, implicated in neurodevelopmental disorders (49, 50). We treated the hBMSCs, HaCaT, HEK293T, and RAW 264.7 cells with 10 nM CRISPR LNP–SNAs (without repair templates) for 3 h. 3 d posttreatment, we extracted the genomic DNA and PCR-amplified the target regions (SI Appendix, Table S3). Subsequently, we quantified indel efficiencies by T7 endonuclease I (T7EI) assays (39) or sequencing (51). At the GRIN2B locus, sequencing identified indels near the Cas9 cleavage site in hBMSC and HaCaT cells, indicating error-prone repair (SI Appendix, Fig. S9); T7EI assay confirmed similar results (Fig. 4B and SI Appendix, Fig. S10). We observed no significant difference in indel rates between particles prepared via azide-DBCO or thiol-maleimide chemistry, indicating that terminal lipid functional groups have minimal impact on editing outcomes. GRIN2B-targeting LNP–SNAs generated approximately 1.5- to 2-fold higher average indel rates than LNPs (Fig. 4 B and C). We observed similar trends at DNAse I and grin2b loci (Fig. 4D and SI Appendix, Table S4), consistent with the enhanced uptake of LNP–SNAs. As expected, plasmid-only controls showed no detectable editing (SI Appendix, Fig. S11).

Fig. 4.

Genome engineering with CRISPR LNP–SNAs. (A) Scheme illustrating the gene editing strategies, including indel mutations and HDR-mediated gene modification with a repair template. (B) Indel frequencies at the GRIN2B target in hBMSCs and HaCaT cells. Cells were treated with CRISPR LNPs or LNP–SNAs conjugated via either DBCO–azide (Azide) or thiol-maleimide (Mal) chemistry, and indel rates were assessed by T7EI or sequencing (Seq). The mean ± SD across three to four biological replicates is shown. Statistical significance was determined by two-tailed Welch’s t test. (C) Indel frequencies at the GRIN2B target in HEK293T cells. Cells were treated with CRISPR LNPs or LNP–SNAs with DBCO–azide formulation, and indel rates were assessed by T7EI. The mean ± SD across three biological replicates is shown, and statistical significance was determined by two-tailed Welch’s t test. (D) Indel frequencies at the DNAse I and Grin2b targets using CRISPR LNPs or LNP–SNAs formulated via thiol-maleimide chemistry (in hBMSC and HaCaT cells) or DBCO–azide chemistry (in HEK293T and RAW 264.7 cells). Indel rates were assessed by T7EI. The mean ± SD across three to four biological replicates is shown, and statistical significance was determined by two-tailed Welch’s t test. (E) HDR efficiencies at the GRIN2B and DNAse I targets using lipofectamine, CRISPR LNPs, or LNP–SNAs formulated via DBCO–azide chemistry. HDR efficiencies were assessed by HindIII digestion. The mean ± SD across three biological replicates is shown, and statistical significance was determined by two-tailed Welch’s t test. (F) Surveyor assay in HEK293T cells targeting the EGFP site. (G) Confocal microscopy images of untreated and CRISPR LNP–SNA-treated cells, with arrows indicating cells whose fluorescence decreased following SNA treatment. (H) Flow cytometry histogram of HEK293T/EGFP cells treated with CRISPR LNPs or CRISPR LNP–SNAs.

We assessed HDR using a donor DNA template encoding a 12-bp HindIII recognition sequence (SI Appendix, Table S1). In three human cell lines, CRISPR LNP–SNAs yielded ~2 to 3-fold higher HDR efficiencies compared to LNPs or lipofectamine (Fig. 4E and SI Appendix, Table S5 and Fig. S12http://www.pnas.org/lookup/doi/10.1073/pnas.2426094122#supplementary-materials). We did not observe HDR at the Grin2b locus in RAW 264.7 cells, likely due to their reduced transfection efficiency. Nevertheless, these findings demonstrate that SNA architecture enhances codelivery of repair templates and genome editors, potentially leading to higher frequencies of DSBs. The increased intracellular concentration of both editors and templates near cleavage sites likely contributed to elevated gene knock-in efficiencies.

Finally, the gene silencing capabilities of CRISPR LNP–SNAs targeting the EGFP gene were evaluated in HEK293T cells (Fig. 4F). The indel efficiency was measured to be 36%, as confirmed by the loss of fluorescence in confocal microscopy images (Fig. 4G). Flow cytometry experiments revealed that treatment with CRISPR LNPs resulted in a fluorescence reduction of approximately 13% of the cell population, whereas CRISPR LNP–SNAs induced reduction in 32% of cells, indicating an enhanced gene knockout efficiency by the SNA constructs (Fig. 4H and SI Appendix, Fig. S13).

This work is important for the following reasons. First, it shows that transforming an LNP delivery system into an LNP–SNA enhances cellular uptake and editing efficiency. While LNPs are primarily internalized via clathrin-mediated endocytosis and micropinocytosis (52), SNAs engage class A scavenger receptors and are internalized through caveolae-mediated endocytosis with rapid kinetics (43). The distinct uptake pathways of LNP–SNAs likely underlie their efficient cellular internalization and improved gene editing efficiency. Second, it shows that multiple nucleic acid cargos can be codelivered to the same cells using this approach. Third, transforming LNPs into SNAs capable of dual nucleic acid delivery enables access to the HDR pathway, which has been a significant challenge for those interested in increasing the precision of gene editing. The versatility and ease of synthesis of LNP–SNAs should make them broadly applicable across the gene editing community. Finally, our findings highlight delivery efficacy as a key factor in enhancing gene editing outcomes and underscore the critical role of the chemical structure (not just the components) of the delivery vehicle in accessing two distinct gene editing pathways.

Conventional LNPs primarily accumulate in the liver, which poses a major challenge in delivering genome editors to other organs of interest (53). Intravenous administration of LNP–SNAs has been shown to result in preferential accumulation in the spleen (40), and we expect a similar biodistribution profile for CRISPR LNP–SNAs. Modifying the nucleic acid shell with alternative targeting moieties, such as aptamers, could further redirect delivery to additional organs. This type of programmable targeting could expand the range of accessible tissues and cell types for gene therapy. Importantly, SNAs have demonstrated favorable safety profiles in human clinical trials (37, 38), suggesting that the enhanced biocompatibility observed for CRISPR LNP–SNAs in vitro is likely translatable in vivo.

Future development of CRISPR LNP–SNAs could involve testing different amounts of DNA donors to optimize HDR efficiency and the delivery of high-fidelity Cas9 variants (54) or base and prime editors (6, 7). In addition, mRNA or ribonucleoprotein complex delivery could mitigate prolonged protein expression, thereby enhancing editing precision. Given that, in principle, any genome editor can be encapsulated by LNP–SNAs, these structures point toward a versatile and promising platform for the next generation of gene-editing therapies.

Materials and Methods

CRISPR Plasmid Construction.

Ten units of BbsI endonuclease (New England Biolabs), 1 µg of pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene #42230) (3), and 1 × NEBuffer r2.1 (New England Biolabs) were added to a 50-µL reaction and incubated overnight at 37 °C for plasmid digestion. The reactions were purified under 1% agarose gel electrophoresis, and the digested plasmids were extracted using an Omega Bio-Tek E.Z.N.A Gel Extraction Kit. Complementary gene fragments encoding sgRNAs for respective genomic loci (SI Appendix, Table S1) were purchased from Integrated DNA Technologies and annealed in 1 × NEBuffer 2 (New England Biolabs). The annealed fragments were then cloned into the hSpCas9 vector using T7 ligase (New England Biolabs). The ligation products were transformed into E. coli DH5α competent cells (Thermo Fisher) under 100 μg/mL ampicillin and cultured overnight from single colonies. Finally, plasmids were extracted using a QIAprep Spin Miniprep Kit (Qiagen), and the sequences were confirmed by Sanger sequencing (forward: 5′–GAGGGCCTATTTCCCATGATTCC–3′ and reverse: 5′–AAAAGCACCGACTCGGTGCCAC–3′).

Synthesis of CRISPR LNPs.

Lipid concentrations were screened to determine the best formulation with the optimal overall encapsulation efficiency and size distribution (SI Appendix, Table S2). The optimized components include 12.5% 18:1 (Δ9-Cis) PC phospholipid (DOPC, Avanti Polar Lipids), 50% Dlin-KC2-DMA (MedChemExpress), 2.5% DPPE-PEG(2000) Maleimide (Nanocs), and 35% cholesterol (Sigma), which were dissolved in 18 µL of ethanol. For cellular uptake experiments, an additional 1% 18:1 Liss Rhod PE (Avanti Polar Lipids) was added and mixed in ethanol for labeling and tracking. CRISPR plasmids (10 µg) were dissolved in 54 µL of 10 mM citrate buffer (pH 4.0) and mixed with lipids in the ethanol phase. The mixtures were dialyzed against 1 × PBS for 2 h, in 4 °C using Pierce 96-well Microdialysis Plates (Thermo Scientific). The concentration and size of the CRISPR LNPs were determined using a NanoSight (Malvern Panalytical).

To encapsulate both the CRISPR plasmid and HDR template, an additional 2 µg of the single-stranded, Alt-R™ HDR donor templates (Integrated DNA Technologies, SI Appendix, Table S1) was mixed with 10 µg of the CRISPR plasmid in 54 µL of 10 mM citrate buffer (pH 4.0). The synthesis was then completed using the procedures outlined above. For EGFP cellular uptake and expression experiments, 10 µg pCMV-GFP (Addgene #11153) (55) was used and encapsulated in lieu of the CRISPR plasmid. The synthesis was then completed using the procedures outlined above. To synthesize azide-terminated LNPs, DPPE-PEG(2000) Azide (Avanti Polar Lipids) was used in lieu of DPPE-PEG(2000) Maleimide. The final lipid ratio and concentrations remained constant, and the synthesis was then completed using the procedures outlined above.

Determination of Plasmid Encapsulation Efficiency.

Quant-iT PicoGreen dsDNA Reagent (Invitrogen) was diluted in 1 × TE buffer or 1 × TE buffer supplemented with 0.1% Triton X-100 and mixed with known concentrations of CRISPR plasmids to generate two standard curves. LNPs were mixed with the same amount of PicoGreen reagent and then diluted 500-fold in either 1 × TE buffer or 1 × TE buffer with 0.1% Triton X-100. The florescence of the samples and standards were measured using a plate reader, and the concentrations of nucleic acids were determined in the linear region of the standard curve. The concentration of free plasmids, [Free], was determined from the 1 × TE standard curve, and the concentration of total plasmids, [Total], was determined by the particles lysed in 0.1% Triton X-100. From this, the encapsulation efficiency was calculated from the following formula:

$$\begin{array}{c}Encapsulation efficiency(\%)=([Total]-[Free])/[Total].\hfill \end{array}$$

Synthesis of CRISPR LNP–SNAs.

5′ Thiol DNA (sequence: 5′–SH-TTTT(TGG)₁₀–3′, incorporated with 5′-Thiol-Modifier C6, Glen Research) and 5′ DBCO DNA (sequence: 5′–DBCO-TTTT(TGG)₁₀–3′, incorporated with 5′-DBCO-TEG Modifier, Glen Research) were synthesized on the solid-phase. 5′ Thiol DNA was deprotected under 1:1 mixture of 30% ammonia solution and 40% methylamine (aq) for 25 min at 55 °C. 5′ DBCO DNA was deprotected under 30% ammonia solution for 3 h at 55 °C. Subsequently, the solution was evaporated to dryness, and deprotected DNA were resuspended in water and purified using reversed-phase high-performance liquid chromatography. For cellular uptake experiments, Cyanine 5 phosphoramidite (Glen Research) was incorporated near the 5′ end of the thiol DNA (sequence: 5′–SH-Cy5-TTTT(TGG)₁₀–3′). The purified DNA was lyophilized, redissolved in nuclease-free water, and reduced under dithiothreitol for 1 h. Dithiothreitol was then removed via NAP-10 columns (Cytiva), and the eluent was vacuum-concentrated to dryness in a centrivap (Labconco). For DNA functionalization of CRISPR LNPs, 30 nmol DNA was resolubilized by mixing with 500 pM respective LNPs (i.e., 5′ Thiol DNA with maleimido-terminated LNPs and 5′ DBCO DNA with azido-terminated LNPs) in 1 × PBS buffer immediately after dialysis. The reaction was incubated overnight at room temperature with shaking at 300 rpm. The size of CRISPR LNP–SNAs was determined using DLS (Malvern Panalytical).

Gel Electrophoresis.

Samples of 5′-SH-DNA, CRISPR LNP–SNA (synthesized via the thiol-Michael addition), and CRISPR LNP–SNA lysed with 1% Triton X-100 (ruptured SNA, synthesized via the thiol-Michael addition) were loaded onto 0.5% agarose gel on ice. Gel electrophoresis was run under 1 × TAE buffer, 70 V, and 2 h.

Electron Microscopy.

CRISPR LNP and LNP–SNA samples were drop-cast onto glow-discharged formvar-coated carbon grids (Ted Pella) and incubated for 2 min. Excess samples were blotted away and stained with 1% uranyl acetate. For cryotransmission electron microscopy, samples were deposited onto lacey carbon grids (Ted Pella) and plunge-frozen into liquid ethane using a FEI Vitrobot Mark IV. All samples were imaged using a JEOL 1400 TEM operating at 120 kV.

Cellular Viability Assay.

Cells (i.e., HaCaT, hBMSC, RAW 264.7, and HEK293T) were seeded in 96-well cell culture plates at a density of 1 × 10⁴ per well in DMEM media with 10% fetal bovine serum (for HaCaT, RAW 264.7, and HEK293T) or mesenchymal stem cell growth medium 2 (for hBMSC) overnight. Then, the cell culture was replaced with fresh media containing different concentrations of CRISPR LNPs and CRISPR LNP–SNAs (synthesized via the thiol-Michael addition) and incubated for another 24 or 96 h. The cells were washed with 1 × PBS buffer, replaced with 10% CCK8 solution (Cell Counting Kit-8), and incubated for 30 min. The absorbance value at 460 nm was measured using a BioTek Synergy plate reader.

Cellular Uptake and Flow Cytometry.

HaCaT cells were seeded in 48-well plates at a density of 3 × 10⁴ per well and cultured overnight in DMEM containing 10% fetal bovine serum and 1% penicillin and streptomycin. The cell culture media were replaced with Opti-MEM and then incubated with 10 nM rhodamine-labeled CRISPR LNP or CRISPR LNP–SNA for various lengths of time (i.e., 0.5, 1, 2, and 4 h). At the end of each treatment, the cells were washed with PBS, trypsinized (Gibco), centrifuged at 300×g for 5 min, and fixed with fixation buffer (BioLegend). Flow cytometry was conducted using a BD FACSymphony A3 Cell Analyzer. The fluorescence of rhodamine was measured at least 1 × 10⁴ single-cell events per sample. The measurements were conducted in biological triplicates.

Confocal Microscopy.

A confocal laser scanning microscope (Zeiss LSM 810) was used to determine the intracellular delivery of CRISPR LNP–SNA (synthesized via the thiol-Michael addition). HaCaT cells were seeded at a density of 1 × 10⁴ per well in borosilicate eight-chambered cover glass slides (Nalge Nunc International). After 8 h, the cells were incubated with 10 nM CRISPR LNP–SNA for 4 h. The cells were then washed with 1 × PBS buffer to remove excess CRISPR LNP–SNAs, stained with 1 µg/mL Hoechst 33342 for 1 min, and fixed with 4% paraformaldehyde (Thermo Fisher) for 15 min. The cell nuclei, lipid, and 5′ thiol DNA were imaged via the Hoechst, rhodamine, and Cy5 channels, respectively.

Surveyor Nuclease Assay.

HaCaT, hBMSC, HEK293T, HEK293T/EGFP, and RAW 264.7 cells were seeded in 48-well plates at a density of 3 × 10⁴ per well and cultured overnight. The cells were then incubated with 10 nM CRISPR LNP or LNP–SNA in Opti-MEM for 3 h. Excess CRISPR LNPs or LNP–SNAs were washed away, and the cells were incubated in fresh media for an additional 3 d. Transfected cells were dissociated and centrifuged at 300×g for 10 min. The media were removed, and the cells were resuspended in 500 µL of QuickExtract DNA Extraction Solution (Biosearch Technologies). DNA was extracted by incubation at 65 °C for 6 min, and then 98 °C for 2 min. Then, 5 µL of the extraction solution was amplified with target region primers (SI Appendix, Table S3). To detect indel mutations, PCR amplified products were annealed in 1 × NEBuffer 2 (New England Biolabs) to allow for heteroduplex formation (95 °C for 10 min, 95 to 85 °C ramping at –2 °C/s, and 85 to 20 °C ramping at –0.2 °C/s). The annealed products were incubated with T7 endonuclease I (New England Biolabs) for 15 min at 37 °C and analyzed on 4 to 15% polyacrylamide or 2% agarose gels. Bands (cleavage and uncleaved) were quantified using densitometry analysis (5). Alternatively, PCR amplicons were sequenced by Sanger sequencing, and indel rates were assessed by TIDE (51).

HDR-Mediated Target Modification.

HaCaT, hBMSC, and HEK293T cells were seeded in 48-well plates at a density of 3 × 10⁴ per well and cultured overnight. The cells were then incubated with 10 nM CRISPR LNP or LNP–SNA in Opti-MEM for 3 h. Excess CRISPR LNP–SNAs were washed away, and the cells were incubated in fresh media for an additional 3 d. Transfected cells were dissociated and centrifuged at 300×g for 10 min. The media was removed, and the cells were resuspended in 500 µL of QuickExtract DNA Extraction Solution (Biosearch Technologies). DNA was extracted by incubation at 65 °C for 6 min, and then 98 °C for 2 min. Then, 5 µL of the extraction solution was amplified using target region primers (SI Appendix, Table S3). PCR amplified products were annealed in 1 × rCutSmart™ buffer (New England Biolabs) to allow for heteroduplex formation (95 °C for 10 min, 95 to 85 °C ramping at –2 °C/s, and 85 to 20 °C ramping at –0.2 °C/s). The annealed products were digested with HindIII-HF endonuclease (New England Biolabs) for 60 min at 37 °C and analyzed on 4 to 15% polyacrylamide or 2% agarose gels (5).

Statistical Analysis.

Statistical analyses were performed using GraphPad Prism 9, with specific tests detailed in each figure caption. Normality of the data was assessed using Q–Q plots, and Welch’s correction was applied when unequal variances were observed. Power analyses conducted with PASS 2021 (v21.0.8) indicate that with 3 to 4 replicates per group, we have 80% power at a two-sided α = 0.05 to detect standardized effect sizes of Δ = 3.1 and Δ = 2.4, respectively, based on a two-sample t test.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.