Lipid-modified Aminoglycosides for mRNA Delivery to the Liver

Abstract

Cationic lipid nanoparticles (LNPs) are widely used as carriers for delivery of nucleic acids. Most synthetic routes towards cationic lipids have derived from simple amine cores. Greater chemical diversity could be obtained through starting with natural products containing basic nitrogen atoms, which offers routes to more complex molecules. Natural building blocks have not been extensively explored, such as aminoglycosides which are both structurally and functionally interesting for developing new carriers for nucleic acid delivery. Herein, we explore cationic lipid-modified aminoglycosides (CLAs) as a family of vehicles for messenger RNA (mRNA) delivery. CLAs were synthesized from natural existing aminoglycosides coupling with alkyl epoxides and acrylates. The top hit (GT-EP10) was able to deliver Luc mRNA to C57BL/6 mice at a dose of 0.05 mg/kg to achieve a 10⁷ average luminescence intensity in the liver. The Lox-Stop-Lox TdTomato mouse model was used to further demonstrate that this efficient mRNA delivery system could be potentially used for gene editing. Successful delivery of human erythropoietin (EPO) mRNA shows that CLA-based LNPs have promising opportunities for delivery of therapeutic nucleic acids in the future.

Graphical Abstract

Main Text

Gene therapy is a promising approach for various applications including treatment of genetic disorders, development of vaccines, and cancer immunotherapy.^[1] Despite great progress, delivery of genetic medicines remains an essential problem. A major obstacle in gene therapy is to develop effective and safe carriers to deliver nucleic acids to targeted organs. Various viral and non-viral carriers have been developed for delivery of nucleic acids^{[1b, 2]} used in research and clinical trials. The main drawbacks of viral carriers are their cost and safety issues.^[3] Therefore, significant attention has been focused on developing non-viral carriers including cationic lipids, cationic polymers, and inorganic carriers.^{[1b, 2a, 4]} Among them, cationic lipid nanoparticles (LNPs) are used most widely, including use in the first FDA-approved siRNA drug (Onpattro^®).^[5] Although progress for LNP delivery of siRNA has been very successful, their application for delivery of mRNA still needs to be further explored. Compared to siRNA, the unique structure of mRNA (single stranded, more than 1000 nt) requires safer and more stable vectors. To date, many successful LNPs contain amine-rich cores which are mostly built on synthetic building blocks such as alkyl amines and polyamines.^{[4b, 6]} Natural products containing basic nitrogen atoms have been rarely explored, such as aminoglycosides, which are both structurally and functionally interesting for developing new carriers for nucleic acids delivery.

Aminoglycosides are a class of natural products comprising amino sugars utilized as antibiotics in the clinic, which are a part of the modern arsenal for killing gram-negative pathogens.^[7] The general mechanism of action (MOA) of aminoglycosides are to interfere the protein synthesis through binding the 30S ribosomal (rRNA) subunit of bacterial causing the inhibition of translocation and misreading of mRNA.^{[7a, 8]} Evidence also shows that aminoglycosides are able to penetrate the cell membrane of pathogens by disrupting the lipopolysaccharide moieties^[7a] which might be beneficial for endosomal escape. Due to the similar physicochemical properties between rRNA, siRNA, and mRNA, the ability of binding to rRNA of aminoglycosides and penetrating cell membrane could be also applied in delivery of siRNA and mRNA. We further speculated that the biological activity of aminoglycosides could be beneficial for mRNA delivery. Previously, Lehn and co-workers^[9] reported a new class of cholesterol-modified kanamycins (KanaChol) for pDNA transfection both in vitro and in vivo. Anderson and co-workers^[10] reported a unique family of cationic lipid-modified aminoglycosides (CLAs) for in vitro and in vivo siRNA delivery. The most effective CLA (a hygromycin derivative, HG-C11) could knockdown FVIII expression in mice with the ED₅₀ of 0.04 mg/kg. Because the chemical structures of pDNA and siRNA are very different compared to the chemical structure of mRNA, we decided to investigate potential utility of CLAs for delivery of mRNA. Herein, we further explored lipid-modified aminoglycosides as a class of delivery materials and specifically tested them for delivery of mRNA both in vitro and in vivo.

We combined the advantages of aminoglycoside core structures that provide mRNA binding/release with aliphatic carbon chains providing the hydrophobicity to form stable nanoparticles. Among the synthetic CLAs, GT-EP10 was found to most efficaciously deliver luciferase mRNA in vitro and in vivo. We further employed the activatable Lox-stop-Lox tdTomato mouse model to demonstrate that this CLAs delivery system could be potentially used for gene editing. Based on the results, CLA nanoparticles represent a potent system for mRNA delivery to the liver and have great opportunities for future utility.

Four commercially available aminoglycosides were chosen as starting materials: hygromycin (HG), gentamicin (GT), amikacin (AM), and geneticin (GN). The hydrophobic side carbon chains of CLAs were installed by reacting with stoichiometric amounts of alkyl epoxides or acrylates using the similar conditions described in a previous report on CLAs.^[10] We focused on alkyl chain lengths of 8 to 14 carbons (Figure 1B and Figure S1, Supporting Information). The synthetic products (CLAs) were formulated with 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), cholesterol (Chol), 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG) and luciferase (Luc) mRNA to form CLA LNPs for initial screening. Recently, we optimized an ionizable amino dendrimer LNP formulation^{[2f, 11]} and determined that LNPs optimized for mRNA delivery should contain less ionizable cationic lipid and more zwitterionic phospholipid compared to standard siRNA formulations.^[6a] This formulation was able to effectively deliver fumarylacetoacetate hydrolase (FAH) mRNA to treat a model of hepatorenal tyrosinemia type I.^[6a] Our understanding of mRNA formulation optimization was guided by a Design of Experiments (DOE) methodology that we also applied herein to CLA LNPs.

Figure 1.

(A) The systematic optimization process to identify the optimal LNP. Five components of nanoparticles: CLA/DOPE/Cholesterol/DMG-PEG mixing with mRNA. (B) Structure of aminoglycosides and synthetic route toward GT-EP10.

Building on this knowledge, we prepared initial CLA formulations (A0) with molar ratios of 15/15/30/3 (CLA/DOPE/Chol/DMG-PEG) at weight ratio of CLA/mRNA (20/1, wt/wt). As shown in the initial in vitro screening results (Figure S2, Supporting Information), the most effective carriers for mRNA delivery were located in the region of lipid-modified gentamicin (9 tails) and amikacin (8 tails). Comparing epoxide- and acrylate-derived CLAs, more effective CLAs for in vitro delivery were epoxide-modified aminoglycosides, except for AM-Ac10 (decyl acrylate-amikacin). We chose six CLAs from the top hits of in vitro results and further tested the delivery efficacy in C57BL/6 mice (Figure S3, Supporting Information). The initial in vivo screening results clearly showed that GT-EP10 LNPs could deliver Luciferase mRNA with protein expression resulting effectively and specifically in the liver. GT-EP12 LNPs could also deliver mRNA to the liver effectively, but with much less specificity. With increasing the side chain to 14 carbons, the in vivo delivery efficacy was dramatically lost (GT-EP14). All amikacin derivatives showed either less efficacy or less specifically. These results identified GT-EP10 as a lead CLA for efficacious mRNA delivery. Interestingly, the identification of gentamicin-based CLAs (such as GT-EP10) was very different to lead CLAs previous reported for siRNA delivery which identified hygromycin as the most active core.^[10] The major length and physical property differences between double-stranded siRNA (18-22 bp) and single-stranded mRNA (1,000-6,000 nt) leads to differing carrier optimization for different roles.^{[6a, 6c]} After identifying the top hit (GT-EP10), we performed the systematic orthogonal DOE optimization method^{[6a, 12]} to further optimize the nanoparticle formulation to achieve higher efficacy and liver specificity.

The relative molar ratios of the four different lipids used to form LNPs were experimentally tuned in the following ranges: GT-EP10 (5 to 20), DOPE (5 to 20), Chol (20 to 35) and DMG-PEG2000 (1 to 4). Based on the principle of orthogonal experimental design^{[6a, 12]}, we can decrease the original 256 combinations down to only 16 combinations which is a huge relief of labor (Table S1, Supporting Information). The in vitro delivery efficacy was evaluated by luminescence intensity after transfection of Luc mRNA into IGROV1 ovarian cancer cells (Figure 2). Formulation A0 represented the original formulation (CLA/DOPE/Chol/DMG-PEG = 15/15/30/3, mol/mol). Compared with the original formulation, several formulations have shown great improvements in delivery efficacy with low cell toxicity. A11 and A12 had increased more than 2-fold compared with A0 formulation. Interestingly, the DMG-PEG percentage of these three formulations are quite different (A11: 3.85%; A12: 1.64%; A0: 4.76%). We plotted the trends of CLA, DOPE, Chol and DMG-PEG relative molar ratio on delivery efficacy in Figure S4 (Supporting Information). The trends showed that the increasing of DOPE relative molar ratio and decreasing of DMG-PEG and Chol relative molar ratio could benefit the delivery efficacy. The tread line of molar percentage of components versus in vitro delivery efficacy (Figure S5, Supporting Information) showed that only one of them has moderate correlations: DMG-PEG (%) versus in vitro delivery efficacy (negative).

Figure 2.

In vitro delivery efficacy of luciferase mRNA CLA LNPs in IGROV1 ovarian cancer cells (25 ng mRNA per well, 96-well plate, n = 3, mean ± SD). Luminescence intensity and cell viability were quantified 24h after adding LNPs. Pie charts show the relative molar percentages of CLA/DOPE/Chol/DMG-PEG in formulations, red: CLA (GT-EP10); orange: DOPE; cyan: Chol; purple: DMG-PEG.

Physical properties of all formulations were measured including RNA encapsulation capability, size, polydispersity, and surface charge (Figure S6, Supporting Information). All nanoparticle diameters were between 100 nm and 200 nm with low polydispersity. A7 and A12 were the largest two nanoparticles which contain the lowest percentage of DMG-PEG (1.64%), while A6 was the smallest nanoparticle, which contains the highest percentage of DMG-PEG (9.09%). This result is in line with expectations that higher PEG-DMG percentages would result in smaller nanoparticle diameters due to larger surface area.^[13] All nanoparticles could encapsulate mRNA very efficiently (>95%), which indicates that the RNA encapsulation capability is not the only factor that affects delivery efficacy of CLA lipid nanoparticles. The surface charge of nanoparticles was close to zero (−2 to 0 mV). The structure of A15 nanoparticle was also revealed by TEM (Figure 4E).

Figure 4.

(A, B) A15 nanoparticle stability monitored by DLS (n = 3) for 10 days. (C, D) Dose-dependent luciferase mRNA expression. C57BL/6 mice (0.05 mg/kg, 0.1 mg/kg, 0.25 mg/kg through I.V. injection, n = 3, mean ± SD). Ex vivo imaging of organs were recorded 6 h after injection. (E) The morphology of GT-EP10 A15 LNP shown in transmission electron microscopy (TEM).

We next tested all formulations in vivo to examine delivery efficacy. Luciferase (Luc) mRNA was administered in CLA LNPs to C57BL/6 mice through systemic IV administration at a dose of 0.1 mg/kg Luc mRNA.^{[2f, 14]} The luminescence intensity following mRNA translation to protein was quantified ex vivo in various organs (Figure 3). Several formulations were found with higher delivery potency compared to the parent A0 formulation. Among these, the A15 CLA LNP exhibited the highest luciferase expression in the liver at the screening dosage of 0.1 mg/kg. The A15 formulation significantly outperformed DLin-MC3-DMA LNPs containing mRNA. The trends of relative molar ratio of CLA, DOPE, Chol, and DMG-PEG on delivery efficacy in vivo were plotted in Figure S4 (Supporting Information), which were very different compared to those in vitro. Interestingly, the in vitro delivery efficacy of A15 formulation is quite low compared to others, suggesting that static cell culture cannot recapitulate the complexity of in vivo delivery barriers.^[15]

Figure 3.

(A) In vivo delivery efficacy of luciferase mRNA in C57BL/6 mice (0.1 mg/kg mRNA per mouse, through I.V. injection, n = 3, mean ± SD, statistical analysis was performed using a Student’s t-test in GraphPad Prism. *: p-value < 0.1). Luminescence were recorded 6 h after injection. Pie charts show the relative molar percentages in the formulations. (B) Ex vivo imaging of organs were recorded 6 h after injection. Formulation of MC3 (relative molar ratio): DLin-MC3-DMA/DSPC/Chol/DMG-PEG = 50/10/38.5/1.5, MC3/mRNA = 10/1, wt/wt). Pie charts show the relative molar percentages of CLA/DOPE/Chol/DMG-PEG in formulations, red: CLA (GT-EP10); orange: DOPE; cyan: Chol; purple: DMG-PEG.

We also plotted the trend line of percentage of each component versus in vivo delivery efficacy (Figure S5, Supporting Information). Two factors had moderate correlations: the percentages of CLA has slightly positive correlation to in vivo delivery efficacy and the percentages of Chol has slightly negative correlation to in vivo delivery efficacy. The A15 formulation contains relatively high molar percentage of DMG-PEG (6.25%) compared with others. The most effective formulation in vitro was A12, which only contains 1.64% of DMG-PEG. Reports^{[4b, 6d]} have shown that the PEG lipid has significant impact on the stability of nanoparticles and delivery efficacy in vivo. For instance, Dong and co-workers^[12b] reported that although nanoparticles with low or zero percentage of DMG-PEG could deliver the mRNA efficiently in vitro, the nanoparticles were unstable. Overall, we can conclude that both the relative ratio and molar percentages of each component play significant roles to achieve an optimized formulation.

Next, we examined the stability of A15 CLA LNPs by monitoring size and PDI in PBS over time (Figure 4A and 4B). As shown in the figure, the size and PDI of A15 CLA nanoparticle did not significantly change over 10 days storage at 4 °C.

Using DOE, we were able to identify potent formulations for in vitro (A12) and in vivo (A15) mRNA delivery. Dose response studies confirmed that A15 CLA LNP could deliver Luc mRNA effectively to the liver at a dose low to 0.05 mg/kg, producing > 10⁷ average luminescence intensity Figure 4C and 4D). Fluorescently labeled Cy5.5-mRNA was used to examine biodistribution. The results demonstrated that most mRNA was transported to liver following intravenous administration which is consistent with other reports for LNPs in this class (Figure S7, Supporting Information).^{[6a, 10]} To further examine tolerability of A15 CLA LNPs, we measured markers of liver function following intravenous administration. Injection of the optimized A15 LNPs did not result in any obvious toxicity, changes in ALT or AST, or liver injury in histological sections (Figure S8 and S9, Supporting Information) at the two tested doses.

To further explore the application of optimized CLA nanoparticles for mRNA delivery, Cre recombinase mRNA was used to perform gene editing in an engineered tdTomato mouse model. The engineered mice could express tdTomato in all cells after deletion of the Stop cassette in genome.^[16] The optimized A15 CLA nanoparticles encapsulated with Cre mRNA were injected intravenously at a dose of 0.25 mg/kg to tdTomato mice. Two days after injection, tdTomato was highly expressed (Figure 5B and 5C). This result was also confirmed by confocal fluorescence imaging of liver sections (Figure 5D). To examine the ability of CLA LNPs to deliver therapeutic genes, human erythropoietin (EPO) mRNA was encapsulated and delivered to wild type C57BL/6 mice as a proof of concept.^[17] The results (Figure 5E and 5F) showed that GT-EP10 nanoparticles can generate >3 times of EPO in blood than the benchmark MC3 LNPs after 6 h. These results establish GT-EP10 nanoparticles as promising carriers of mRNA to provide therapeutic proteins to the liver.

Figure 5.

(A) Schematic representation shows that delivery of Cre mRNA deletes the stop cassette and activates tdTomato gene to produce protein. (B, C) Cre mRNA delivery efficacy (GT-EP10, A15) tested with Td-tomato mice (Ai9) (0.25 mg/kg through I.V. injection, n = 3, mean ± SD). Ex-vivo imaging of organs were recorded 2 days after injection. (D) Confocal fluorescence microscopy of tissue sections showed tdTomato-positive cells in liver. Scale bars = 100 μm. (E, F) In vivo human erythropoietin expression following I.V. injection of EPO mRNA after 6 h and 24 h. Formulation of MC3 (relative molar ratio): Dlin-MC3-DMA/DSPC/Chol/DMG-PEG = 50/10/38.5/1.5, MC3/mRNA = 10/1, wt/wt). (n = 3, mean ± SD, statistical analysis was performed using a Student’s t-test in GraphPad Prism. **: p-value < 0.01, ns: not significant).

In summary, we report a family of highly effective mRNA delivery vehicles based on cationic lipid-modified aminoglycosides (CLAs). After formulation optimization though the orthogonal experimental design, the top hit (GT-EP10) was able to deliver Luc mRNA at a dose of 0.05 mg/kg to achieve a 10⁷ average luminescence intensity in liver. The TdTomato mouse model was used to demonstrate that this efficient mRNA delivery system could be potentially used in the gene editing area. Based on the results of delivery of human EPO mRNA, CLAs nanoparticles have great opportunities for delivery of therapeutic nucleic acids in the future.