Safer non-viral DNA delivery using lipid nanoparticles loaded with endogenous anti-inflammatory lipids

Abstract

The value of lipid nanoparticles (LNPs) for delivery of messenger RNA (mRNA) was demonstrated by the coronavirus disease 2019 (COVID-19) mRNA vaccines, but the ability to use LNPs to deliver plasmid DNA (pDNA) would provide additional advantages, such as longer-term expression and availability of promoter sequences. However, pDNA-LNPs face substantial challenges, such as toxicity and low delivery efficiency. Here we show that pDNA-LNPs induce acute inflammation in naive mice that is primarily driven by the cGAS-STING pathway. Inspired by DNA viruses that inhibit this pathway for replication, we loaded endogenous lipids that inhibit STING into pDNA-LNPs. Loading nitro-oleic acid (NOA) into pDNA-LNPs (NOA-pDNA-LNPs) ameliorated serious inflammatory responses in vivo, enabling safer, prolonged transgene expression-11.5 times greater than that of mRNA-LNPs at day 32. Additionally, we performed a small LNP formulation screen to iteratively optimize transgene expression and increase expression 50-fold in vitro. pDNA-LNPs loaded with NOA and other bioactive molecules should advance genetic medicine by enabling longer-term and promoter-controlled transgene expression.

Fig. 1 |. Unlike nucleoside-modified mRNA-LNP, pDNA-LNP delivery causes acute inflammatory responses.

a, Survival curve graph of naive C57BL/6 (‘Black-6’) mice i.v. injected with 1 mg kg⁻¹ of either mRNA-LNP or pDNA-LNP shows 100% mortality in mice treated with pDNA-LNPs compared with 0% in mice treated with mRNA-LNPs. b, Four hours after 1 mg kg⁻¹ i.v. injection of LNPs, movement of mice was tracked for 1 h, and total distance walked was assessed by AI software (DeepLabCuts), a validated metric of an infusion reaction. Mice treated with pDNA-LNPs had significantly lower total distance walked compared with mRNA-LNP controls, indicating severe lethargy. c, Weight change over time in mice given a much lower i.v. dose (5 μg) of mRNA-LNPs or pDNA-LNPs. d, Multiplex analysis of pro-inflammatory plasma cytokines 4 h after a 5 μg i.v. dose of pDNA-LNP indicates acute systemic inflammation compared with PBS and mRNA-LNP controls. e,f, Specifically, IFN-β (e) and IL-6 (f) levels were approximately 1,400-fold and approximately 1,000-fold higher, respectively, for mice injected with pDNA-LNP compared with mRNA-LNP. g, Five micrograms of pDNA formulated with three FDA-approved LNP formulations (D-Lin-MC3-DMA, SM-102 and ALC-0315 ionizable lipids) was i.v. injected in mice, and plasma was collected 4 h later for cytokine quantification, which indicated that this inflammatory response occurs across various LNP formulations. h, Cytokine levels in mouse plasma collected 4 h or 24 h after i.v. injection of 5 μg of pDNA-LNPs, highlighting acute-but-transient inflammation, with the majority of pro-inflammatory cytokine levels back to baseline at the 24 h timepoint. i–k, In vitro studies in a macrophage-derived cell line, RAW264.7. i, Effect of 1,000 ng ml⁻¹ mRNA-LNP or pDNA-LNP on cell viability over time. j, IFN-β levels in cell supernatant 4 h after exposure to 1,000 ng ml⁻¹ of empty-LNP, mRNA-LNP and pDNA-LNP show DNA-cargo-specific inflammatory cytokine production. k, IFN-β levels in cell supernatant increase exponentially as a function of pDNA-LNP dose. Statistics: n = 5 per group for a and c (biological replicates); n = 4 per group for k (biological replicates); and n = 3 per group for the rest (biological replicates). Data shown represent mean ± s.e.m. b,c,i, Unpaired, two-tailed t-tests were performed. For all other graphs, comparisons were made using one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01 and ****P < 0.0001. NS, not significant.

Fig. 2 |. STING activation drives pDNA-LNP-induced inflammation.

a, The proposed mechanism that drives pDNA-LNP inflammation. Any cytosolic DNA (endogenous or exogenous) is detected, independent of DNA sequence, by cGAS, leading to downstream activation of STING, which induces an acute inflammatory response. b, STING-KO mice i.v. injected with 5 μg of pDNA-LNPs have reduced levels of pro-inflammatory cytokines in plasma 4 h after dose compared with wild-type mice. c,d, pDNA-LNPs injected in STING-KO mice have complete reduction in IFN-β (c), leading to improved survival rates at the 1 mg kg⁻¹ dose (d). e, Representative images of p-STING 4 h after treatment indicate STING activation for the pDNA group (1,000 ng ml⁻¹ dose) in RAW264.7 cells. f, Quantification of p-STING MFI from original images. Statistics: n = 3 per group for b and c (biological replicates); n = 5 per group for d (biological replicates); and n = 6 per group for f (biological replicates). Data shown represent mean ± s.e.m. c, Unpaired, two-tailed t-test was performed. f, Comparisons were made using one-way ANOVA with Tukey’s post hoc test. All studies were done using C57BL/6 mice. *P < 0.05, **P < 0.01 and ***P < 0.001.

Fig. 3 |. Loading endogenous anti-inflammatory lipids that inhibit STING into standard pDNA-LNP formulations.

a, Schematic showing how cell stress caused by virus infections leads to nitration of endogenous unsaturated fatty acids to form NFAs that have anti-inflammatory properties, including potent inhibition of STING (modified from Fig. 1 of the article by Hansen et al.). b, Development of platform technology by adding AILs as the fifth component in the lipid mixture before LNP formation using microfluidics. c, Size distribution of pDNA-LNPs loaded with various AILs determined by DLS shows no differences in LNP size or polydispersity index (PDI). d, All tested AILs load more than 80% (AIL-to-lipid ratio of 0.2, mole-to-mole) and have no impact on pDNA encapsulation. Detailed methodology can be found in Methods. e, pDNA-LNPs loaded with various AILs reduce IFN-β levels in cell supernatant 4 h after 1,000 ng ml⁻¹ dose. AILs that are nitrated before LNP formation (NLA and NOA) are more effective than lipids that are not nitrated during LNP formation (DHA and EPA). Statistics: n = 4 per group for e (biological replicates) and n = 3 per group for c and d (technical replicates, similar findings with biological replicates). Data shown represent mean ± s.e.m., and comparisons were made using one-way ANOVA with Tukey’s post hoc test. **P < 0.01.

Fig. 4 |. NOA-pDNA-LNPs show superior safety profiles compared with standard pDNA-LNPs in vitro and in vivo.

a, Representative confocal images of p-STING show that NOA-pDNA-LNPs do not activate STING compared with standard pDNA-LNPs in RAW264.7 cells. b, Quantification of p-STING MFI shows significant decrease for the NOA-pDNA-LNP group relative to standard pDNA-LNP. c, Confocal imaging of a downstream marker of STING activation, phosphorylated TBK1 (p-TBK1), is also not activated for cells treated with NOA-pDNA-LNPs compared with standard pDNA-LNPs. d, Quantification of p-TBK1 MFI. e, RAW264.7 cell viability measured over time indicates better tolerability of NOA-pDNA-LNPs compared with standard pDNA-LNPs. f–h, IFN-β levels in cell supernatant are significantly lower for NOA-pDNA-LNPs compared with standard pDNA-LNP 4 h after 1,000 ng ml⁻¹ dose, irrespective of LNP formulation (D-Lin-MC3-DMA (f), SM-102 (g) and ALC-0315 (h)—all FDA-approved LNP formulations). i, Quantification of pro-inflammatory plasma cytokines 4 h after a 5 μg dose of pDNA-LNPs or NOA-pDNA-LNPs. j,k, Specifically, IFN-β (j) and IL-6 (k) levels are approximately four-fold and approximately eight-fold lower, respectively, for mice injected i.v. with NOA-pDNA-LNP compared with standard pDNA-LNP control. l, Survival curve in C57BL/6 (‘Black-6’) mice, comparing an i.v. dose of 1 mg kg⁻¹ of pDNA-LNP and NOA-pDNA-LNPs, shows that the addition of NOA in pDNA-LNPs completely prevents mortality. Statistics: n = 5 per group for b and d (technical replicates, similar findings with biological replicates); n = 3 per group for e–k (biological replicates); and n = 5 per group for l (biological replicates). Data shown represent mean ± s.e.m. e, Unpaired, two-tailed t-tests were performed. b,d,f–h,j,k, Comparisons were made using one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Fig. 5 |. NOA-pDNA-LNPs show prolonged transgene expression in vivo.

a, Representative IVIS images of BALB/c mice that were i.v. injected (retro-orbitally) with 25 μg of either pDNA-LNPs or NOA-pDNA-LNPs (encoding luciferase). p, photons. b, Quantified total flux (photons per second) from IVIS images shows that the addition of NOA does not hinder pDNA transgene expression capacity and shows prolonged expression (at least 1 month). c, At day 32, total transgene expression levels were significantly higher in mice treated with 25 μg NOA-pDNA-LNPs compared with PBS and mRNA-LNP (5 μg) controls. Statistics: n = 4 per group for PBS control group and n = 5 per group for pDNA-LNP and NOA-pDNA-LNP groups for a–c (biological replicates). Note that two mice that received 25 μg of pDNA-LNPs died within 2 d. Data shown represent mean ± s.e.m. c, Comparisons were made using one-way ANOVA with Tukey’s post hoc test. **P < 0.01.

Fig. 6 |. A small LNP formulation screen significantly boosts transgene expression of NOA-pDNA-LNPs in vitro, enabling efficient transfection in difficult-to-transfect cells.

a,b, DoE screening to optimize NOA-pDNA-LNP formulation in RAW264.7 cells. Using JMP software, a full factorial screen was designed by varying ionizable lipid mol% (30–50), total lipid-to-pDNA w/w ratio (10:1 to 40:1) and the type of helper lipid used (DSPC, DOPE, DOTAP and 18:0 PG). All LNPs contained a NOA-to-total lipid ratio of 0.2 (mole-to-mole). Twenty-four hours after 500 ng ml⁻¹ treatment in RAW264.7 cells, luciferase protein expression was measured, highlighting the influence of LNP formulation parameters for transgene expression (a). NOA-pDNA-LNP optimized with DOTAP as the helper lipid led to a 50-fold increase in transgene expression when compared with standard NOA-pDNA-LNP (b). c,d, Two-dimensional monoculture of difficult-to-transfect cell line human iPSC-derived type II alveolar epithelial cells (iAT2s) was treated with 1,000 ng of eGFP-encoding pDNA, using Lipofectamine, standard NOA-pDNA-LNPs or optimized NOA-pDNA-LNPs, and imaged after 48 h. tdTomato (red) signal indicates iAT2 positivity and eGFP (green) signal indicates successfully transfected cells (c). One hundred twenty hours after transfection, tdTomato⁺ cells that were also eGFP⁺ were quantified using flow cytometry, indicating similar transfection levels (trending higher) for optimized NOA-pDNA-LNPs to the gold standard, Lipofectamine 2000 (d). Statistics: n = 2 per LNP formulation for a (biological replicates) and n = 3 per group for b–d (biological replicates, representative images shown for c). Data shown represent mean ± s.e.m.