Lipid Nanoparticle-Mediated Hit-and-Run Approaches Yield Efficient and Safe In Situ Gene Editing in Human Skin

Abstract

Despite exciting advances in gene editing, the efficient delivery of genetic tools to extrahepatic tissues remains challenging. This holds particularly true for the skin, which poses a highly restrictive delivery barrier. In this study, we ran a head-to-head comparison between Cas9 mRNA or ribonucleoprotein (RNP)-loaded lipid nanoparticles (LNPs) to deliver gene editing tools into epidermal layers of human skin, aiming for in situ gene editing. We observed distinct LNP composition and cell-specific effects such as an extended presence of RNP in slow-cycling epithelial cells for up to 72 h. While obtaining similar gene editing rates using Cas9 RNP and mRNA with MC3-based LNPs (10-16%), mRNA-loaded LNPs proved to be more cytotoxic. Interestingly, ionizable lipids with a pK_a ∼ 7.1 yielded superior gene editing rates (55%-72%) in two-dimensional (2D) epithelial cells while no single guide RNA-dependent off-target effects were detectable. Unexpectedly, these high 2D editing efficacies did not translate to actual skin tissue where overall gene editing rates between 5%-12% were achieved after a single application and irrespective of the LNP composition. Finally, we successfully base-corrected a disease-causing mutation with an efficacy of ∼5% in autosomal recessive congenital ichthyosis patient cells, showcasing the potential of this strategy for the treatment of monogenic skin diseases. Taken together, this study demonstrates the feasibility of an in situ correction of disease-causing mutations in the skin that could provide effective treatment and potentially even a cure for rare, monogenic, and common skin diseases.

Keywords: ARCI; base editing; gene delivery; gene editing; genodermatoses; lipid nanoparticles; skin.

Figure 1

(A) Schematic depiction of LNPs and chemical structures of the helper lipids (DSPC, DOPC, DOPE, ES, DSPG) used for LNP preparation. (B) Cell uptake efficiency of LNPs containing different helper lipids 24 h after incubation with primary keratinocytes (KCs). (C, D) Frequency of indel% (normalized to wild-type (WT) cells) in the model gene HPRT after transfection of KCs with (C) RNP- and (D) mRNA-loaded LNPs at three different L/R (mol/mol) and N/P (mol/mol) ratios, respectively. * indicates statistically significant differences over RNAimax; *p < 0.05; **p < 0.01; ***p < 0.001. (E) Effect of RNP-loading at different pH on indel% (normalized to WT) indicative of gene editing efficacies in the model gene HPRT. Data are presented as the mean ± SD of at least three biologically independent replicates.

Figure 2

Cell viability of primary human KCs after exposure to (A) unloaded LNPs (μM refers to lipid concentration), (B) RNP-loaded LNPs (depicted as lipid-to-RNP (L/R) ratio), and (C) mRNA-loaded LNPs (depicted as nitrogen-to-phosphate (N/P) ratio) at different concentrations and ratios after 48 h. Data are presented as the mean of three biological replicates ± SD. (D) Representative live/dead cell assay images comparing the toxicity of RNP and mRNA-loaded LNP over 48h. Green is indicative of viable cells; red dots represent dead cells. (E) Preincubation with endocytosis pathway inhibitors indicates that LNP internalization in KCs is mainly dynamin-dependent. Scale bars = 50 μm. (F) Confocal microscopy images showing the time-dependent cell uptake kinetics of RNP-loaded LNPs over 24h.

Figure 3

Immunofluorescence staining against EEA1 (early endosome marker), RAB11A (recycling endosome marker), and LAMP1 (late endosome marker) in primary human KCs 24 h after treatment with mRNA-loaded, RNP-loaded, and unloaded LNP. Scale bar = 50 μm.

Figure 4

Impact of ionizable lipids and their pK_a on the frequency of indel% (normalized to wild-type (WT) cells) in (A) primary human KCs and (B) bronchial epithelial cells as well as nonepithelial cells including endothelial cells and dermal fibroblasts. (C) Cell viability of primary human KCs after a 48h treatment with the different mRNA-loaded LNP formulations. Data are presented as mean ± SD of at least three biologically independent replicates. * indicates statistical significance over RNAimax: *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 5

(A) Schematic representation of 3D skin models including histological and immunofluorescence staining verifying the comparability to human skin. (B) Frequency of indel% (normalized to wild-type (WT) models) of topically applied RNP- or mRNA-loaded DOPE LNP in 3D skin models after 48 h following pretreatment with 400 μm solid microneedles and (C) pretreatment with laser ablation. (D) IL-6 and IL-8 levels of untreated (control) and DOPE-LNP and LNP H treated skin models after pretreatment with laser ablation.

Figure 6

(A) Western blot analysis of Cas9 protein expression in primary human KCs and primary human bronchial epithelial cells 24, 48, and 72 h after treatment with Cas9 RNP loaded onto DOPE-LNPs. GAPDH served as housekeeper. (B) Using rhAMPseq, effective on-target editing of our model target HPRT with LNP H was confirmed while no off-target effects were observed in any of the eight predicted sgRNA-dependent off-target sites. (C) Visualization of the distribution of the most frequently identified alleles around the cleavage site for sgRNA AATTATGGGGATTACTAGGA in donor 1. Nucleotides are indicated by unique colors (A = green; C = red; G = yellow; T = purple). Substitutions are shown in bold font. Red rectangles highlight inserted sequences. Horizontal dashed lines indicate deleted sequences. The vertical dashed line indicates the predicted cleavage site.

Figure 7

(A) The sequence of one of the most common ARCI causing mutation TGM1 c.877–2 A>G. (B) Schematic depiction of the underlying mechanism of a cytosine base editor. (C) Base editing of TGM1 c.877–2 A>G ARCI patient cells using RNAimax and NG-BE4max editor with 15 μg/mL modified and unmodified mRNA. Editing of the A>G mutation on the coding strand is displayed as the mean normalized editing rates from three technical replicates. Chromatograms of sequences after treatment with (C) RNAimax show a double peak at the target site, indicative of base editing. The edited sites are marked with a green arrow.