Targeting intestinal inflammation using locked nucleic acids delivered via lipid nanoparticles

Abstract

Locked nucleic acids are a third-generation antisense oligonucleotides with high binding affinity. A major limitation is the high dosages they require to achieve efficacy which may induce unwanted adverse effects. Here, we report the use of Lipid-based nanoparticles to deliver locked nucleic acids for treating intestinal inflammation in mice. Eight formulations with novel ionizable lipids were screened for stability and toxicity. Particles were loaded with splice-switcher sequence, enabling a precise assessment of potency in vitro. Three lead candidates were tested in vivo, demonstrating a 30-fold dose reduction compared to the unformulated oligonucleotides. The most potent formulation, encapsulating a sequence against Tumor necrosis factor alpha, was evaluated in a mouse model of colitis. Treatment reduced disease severity and inflammatory cytokines, with good safety. These findings support the use of lipid nanoparticles for the precise delivery of locked nucleic acids and highlight their promise for future therapies.

Fig. 1. Structures of ionizable amino lipids.

The lipids were synthesized using different hydrophobic tails, with ethanolamine/hydroxylamine/hydrazine linkers and dimethylamine/piperazine head groups. Lipids 2, 6 & 10 consist of a linoleyl chain, and lipids 14 & 26 consist of a branched ester chain as hydrophobic tails. For Lipid 15, a linoleyl chain and a branched ester chain were included as hydrophobic tails, whereas a linoleyl chain and a non-branched ester chain were used as hydrophobic tails for Lipid 25. For Lipid 24, two different branched ester chains were used as hydrophobic tails.

Fig. 2. LNPs formulations and characterization.

In vitro screening of ionizable amino lipids in LNPs; a Schematic illustration of LNPs preparation using microfluidic mixers, created in BioRender. Hazan-halevy (2025) https://BioRender.com/r4vmcxs. b Representative CryoEM images of LNPs. The scale bars are 50 nm. c mean diameter ± SD (nm); d polydispersity index (PDI) ± SD. e ζ potential ± SD (mV), as measured by Zeta Sizer. f Percentage of encapsulation efficiency as measured by a RiboGreen assay. Bars in panels c-f show consist of n = 2 independent LNP preparations with at least n = 3 technical repeats for each. g Viability test measured by XTT to establish in vitro toxicity 96 h post-transfection (n = 3 independent biological repeats for cells that were treated with LNPs). The bars in (c–g) represent means with error bars that represent the standard deviation. One-way ANOVA was conducted, followed by two-sided Dunnett’s multiple comparison. P-values are indicated whenever significant.

Fig. 3. LNA-LNPs Splicing efficacy.

a schematic illustration of an LNA that causes exon 14 skipping in the Hif1α pre-mRNA, resulting in a completely exogenous isoform of the Hif1α mRNA and shifting the balance from a wild-type Hif1a mRNA to a “synthetic” isoform, created in BioRender. Hazan-halevy (2025) https://BioRender.com/4hcx0py. Splice-switching efficacy as a function of ascending concentration of aHα-loaded LNPs prepared with lipids 10 (b), lipid 15 (c), and lipid 2 (d), tested on three cell lines. Each graph represents a curve obtained from n = 2 independent biological repeats, and each point is composed of n = 2 technical repeats. Error bars represent the standard deviation of the mean. In points where the error bar is missing, the error bars are smaller than the symbols and are therefore not displayed.

Fig. 4. Assessing in vivo splice-switching efficiency with LNAs-LNPs in healthy mice.

a Experimental design, created in BioRender. Hazan-halevy (2025) https://BioRender.com/ka9d8uib lipid 10, c lipid 15, and d lipid 2, Hif1a SpSw efficiency was measured in the liver and spleen of mice. Each column represents the mean splice-switching % of a group of n = 3 mice, with error bars that represent the standard deviation.

Fig. 5. Distribution profile of LNAs- loaded LNPs.

a IVIS images of 50% CY5 lipid 15 LNPs in DSS-treated mice, b quantified as Avg Net Radiance, obtained from n = 3 independent biological repeats, c flow cytometry of LNPs’ association to different cell populations isolated from the colonic lamina propria obtained from n = 2 for UT and n = 3 for LNPs-treated independent biological repeats. For (b) and (c), data is represented as scattered dots, with lines representing mean and standard deviation. d Hif1a SpSw was measured in the colons of DSS-bearing mice versus healthy ones that received either unformulated aHα at two doses (1 mg/kg; n = 2, 3 mice in healthy and DSS-treated groups, respectively, and 30 mg/kg; n = 1, 2 mice in healthy and colitis-bearing groups, respectively), 1 mg/kg of aHα-loaded lipid 15 LNPs (n = 3 biological repeats for healthy, and n = 10 biological repeats for colitis-bearing), or PBS (n = 2 biological repeats). Hif1a SpSw was also measured in the e kidneys (n = 1 PBS-treated mice, n = 2 healthy ASO or LNPs-treated mice, n = 3 colitis-bearing free ASO-treated mice, n = 4 colitis-bearing LNPs-treated mice) and, f lungs (n = 1 PBS-treated mice, n = 2 healthy ASO or LNPs-treated mice, n = 3 colitis-bearing free ASO-treated mice, n = 4 colitis-bearing LNPs-treated mice) of DSS-bearing mice versus healthy ones that received either unformulated 30 mg/kg aHα, 1 mg/kg of aHα-loaded lipid 15 LNPs, or PBS. Data is represented as scattered dots with columns for the average, and error bars representing the standard deviation. For graphs in (d–f), Two-way ANOVA was conducted, followed by two-sided Sidak’s multiple comparison. P-values are indicated.

Fig. 6. Therapeutic efficacy of LNAs-loaded in LNPs.

a Experimental design, created in BioRender. Hazan-halevy (2025) https://BioRender.com/ri51v9d. b Percentage weight change in all treated groups. c Colon length was measured at the end of the experiment. Expression levels of pro-inflammatory cytokines; d TNFα, e IL-6, f IL-β. For graphs in (b–f). n = 5, 7, 10, 10, 10 mice for healthy, DSS-bearing, NC LNPs, Free Anti-TNF ASO, Anti-TNFα-loaded LNPs, respectively. Liver enzyme levels were measured as an assessment for toxicity (n = 7 mice); g ALP [IU/L], h ALT [IU/L], i AST [IU/L]. j Histological scoring of colon tissue sections (n = 10 mice). The center values represent the average, while the error bars represent the standard deviation of the mean. For the graph in (a), a Two-way ANOVA was conducted, followed by a two-sided Dunnett’s multiple comparison. For the rest of the graphs, a one-way ANOVA was conducted, followed by a two-sided Bonferroni’s multiple comparison. P-values are indicated.