Mitochondria-targeted gene delivery using fluorinated lipid nanoparticles to alleviate Leber’s hereditary optic neuropathy

Mitochondrial disorders are a broad group of hereditary and multisystem diseases that affect approximately 1 in 4300 individuals¹. Mitochondrial DNA (mtDNA) mutation is recognized as a causative factor in various mitochondrial disorders, common symptoms surrounding muscle weakness, heart dysfunction, neurological impairment, vision and hearing impairments, and stunted growth^2,3. However, due to the lack of effective clinical treatments for regulating mutated genes in mitochondrial diseases, the majority of these diseases lead to severe disability or even mortality⁴. This underscores the urgent need for methods to regulate and correct mitochondrial gene expression, which could effectively intervene in the onset and progression of mitochondrial disorders.

Building upon the success of nuclear gene therapy, an increasing number of researchers have sought to regulate the mitochondrial genome using modified tools, thereby driving the development of mitochondria-targeted gene delivery strategies. These systems involve introducing exogenous plasmid gene (pDNA) into the mitochondrial matrix, allowing target proteins to be expressed under the mitochondrial codon system and achieving functional complementation of mutant genes within the organelle^5,6,7. In addition to the various challenges faced by traditional gene therapies, a key to mitochondrial gene therapy is the development of a tool for overcoming the barrier posed by the mitochondrial double membrane, introducing nucleic acid drugs into the mitochondrial matrix⁸.

Current mitochondria-targeted delivery systems for transporting therapeutic agents to mitochondrial compartments primarily rely on the use of cationic hydrophobic groups (such as triphenylphosphonium, TPP) and mitochondrial targeting sequences (MTS)^9,10. These strategies have proven critical for the targeted delivery of small molecules into mitochondria¹¹. However, reports on mitochondrial gene delivery using nanoparticles modified with mitochondria-targeting signal peptides remain scarce^12,13,14. MTS-mediated transport primarily facilitates the import of proteins essential for mitochondrial function. This process involves the recognition and binding of the MTS to the translocase of the outer membrane (TOM) complex, followed by the unfolding of the protein to enable its translocation into the mitochondrial intermembrane space^15,16. Moreover, the mitochondria-targeting capability mediated by cationic hydrophobic groups is highly dependent on mitochondrial membrane potential (MMP). Although some experimental data suggest colocalization of certain gene cargos with mitochondria under normal conditions, this efficiency is significantly diminished under pathological conditions where MMP is compromised^6,17. The requirement for MMP-dependent mechanisms in mitochondria-targeted gene delivery systems presents a significant challenge when applying mitochondrial gene therapy to real mitochondrial disease conditions. The excess cationic charge density of mitochondria-targeted gene delivery systems is essential for approaching the outer mitochondrial membrane through electrostatic interactions, but it inevitably leads to adverse effects, including significant cytotoxicity¹⁸. How to develop technologies that transcend the conventional reliance on the MMP of mitochondria to create MMP-independent formulations remains a critical concern for achieving safe and effective mitochondria-targeted gene delivery.

Fluorinated polymers have been widely adopted to enhance the intracellular delivery efficiency of proteins, nucleic acids, and small molecules^19,20. In our research group, as well as by other researchers, fluorinated polymers have also been used to deliver functional pDNA, significantly improving cytoplasmic gene delivery efficiency^21,22. In recent years, fluorinated PEGylated lipid nanoparticles have emerged as a promising platform, offering notable advantages in improving the efficiency of mRNA delivery^23,24. Notably, the unique physicochemical properties of fluorinated ligands, including their low surface energy, hydrophobicity, and lipophobicity, play a crucial role²⁵. These properties not only enhance the affinity of fluorinated groups with the cell membrane, facilitating cellular uptake, but also promote endosomal escape^26,27. It can thus be hypothesized that fluorinated modifications may exploit their unique physicochemical properties to penetrate intracellular membrane structures, offering a membrane translocation mechanism that is independent of MMP. Additionally, studies have shown that replacing traditional hydrocarbon chains on amphiphilic compounds with fluoroalkyl groups can reduce cytotoxicity and hemolytic activity, improving the biocompatibility of polymers^28,29. Therefore, introducing fluorinated modifications into delivery vectors may represent a breakthrough in designing safe and efficient gene delivery systems in an MMP-independent manner. However, whether fluorination can penetrate the mitochondrial membrane remains unknown.

Here, we introduce the fluorinated group as a membrane-penetrating component into the hydrophobic tail of ionizable lipids for the construction of an MMP-independent fluorinated lipid nanoparticle (F-LNP). Particularly, we are interested in whether the fluorine content in F-LNPs affects their efficiency in mitochondria-targeted gene delivery. Therefore, we synthesize ionizable lipids with varying numbers of fluorinated groups in their hydrophobic tails and used these fluorinated ionizable lipids to construct 16 different formulations of F-LNPs (Fig. 1A–C). Using fluorescently labeled gene drugs (Cy5-plasmid gene, Cy5-pDNA) and a mitochondria-specific green fluorescent protein plasmid (mtGFP) as model genes (Supplementary Sequence 1), we systematically evaluate the mitochondrial targeting capability and gene transfection efficiency of each F-LNP formulation and explore some structural criteria required in fluorinated lipids for efficient mitochondria-targeted gene delivery. When the mass ratio of fluorine atoms to total lipids was 7.94%, mitochondrial gene transfection efficiency is optimized, reaching levels five times higher than those of LNPs without fluorinated modification. Additionally, the fluorinated groups facilitate the MMP-independent delivery of exogenous genes by F-LNPs directly into the mitochondrial matrix rather than to the outer mitochondrial membrane or intermembrane space. We find that this effect is likely due to the ability of fluorinated groups to interact with cardiolipin, a unique component of mitochondria, and rapidly traverse the membrane structure. To enhance the efficiency of mitochondrial selective delivery, we add MTS to the optimal fluorinated formulation to produce F-M-LNPs. In cells from patients with rare mitochondrial diseases, F-M-LNP has proven to be an exceptional gene delivery system targeting mitochondria. Additionally, F-M-LNP enhances the expression of targeted mitochondrial proteins, resulting in significant alleviation of symptoms in a mouse model expressing a mutant variant of mtDNA. Overall, our study provides proof of concept that incorporating fluorinated modifications into delivery vectors could offer a viable strategy to bypass the mitochondrial membrane barrier and modulate the mitochondrial genome, which has translational value for treating mitochondrial disease.

Synthesis of fluorinated ionizable lipids

We fluorinated the hydrophobic tails of ionizable lipids to different degrees (Fig. 1A), synthesizing 4 types of fluorinated lipids: fluorinated with 4 carbon atoms (4F, Supplementary Fig. 1), 6 carbon atoms (6F, Supplementary Fig. 2), 8 carbon atoms (8F, Supplementary Fig. 3), and non-fluorinated lipid (0F, Supplementary Fig. 4). In brief, hydrophobic tails with varying degrees of fluorination at the terminal carboxyl group were synthesized. Then, the hydrophobic tails with different degrees of fluorination were connected to hydrophilic head groups to form fluorinated lipids through esterification reactions. The chemical structures of these fluorinated lipids were confirmed using ¹H NMR spectra (Supplementary Fig. 5–8).

The cytotoxic effects of all four fluorinated lipids were investigated by MTT assay³⁰. In the cytotoxicity assay, the concentration range of fluorinated lipids was set at 10–60 µg/mL, as the maximum concentration established here was significantly higher than the concentration used in the subsequent transfection experiments. We found that all F-LNPs with different fluorinated lipids exhibited cellular viability higher than 90% across the tested concentration range, demonstrating that the synthesized fluorinated lipids had good safety in Neuro 2a (N2a) cells (Supplementary Fig. 9). Therefore, all four fluorinated lipids were used for subsequent studies.

Preparation of fluorinated lipid nanoparticles

LNP formulation (positive control group) for FDA-approved lipids included ionizable lipid ((4-hydroxybutyl)azanediyl)bis(hexane-6,1-diyl)bis(2-hexyldecanoate), ALC-0315), 1,2-Distearoyl-sn-glycero-3-phosphocholine (DSPC), cholesterol, 1,2-Distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (DSPE-PEG2000) at molar ratio of 50:10:38.5:1.5. Sixteen F-LNPs were prepared by adjusting the molar ratios of the ionizable lipid and fluorinated ionizable lipid, as well as varying the types of fluorinated lipids. The formulations of F-LNPs involved fluorinated lipid (0F, 4F, 6F and 8F), ALC-0315, DSPC, cholesterol and DSPE-PEG2000 with a:(50 – a):10:38.5:1.5 mol% (a = 12.5, 25, 37.5, 50) (Fig. 1B, C). Lipids were solubilized in ethanol, and the pDNA was added to the water phase. Various LNPs were prepared by microfluidic mixing of a lipid-containing organic phase and a pDNA-containing aqueous phase at a 1:3 volume ratio.

Fluorinated lipid nanoparticles enable gene delivery to mitochondria in an MMP-independent manner

During the development of fluorinated groups in ionizable lipids, we optimized the fluorination ratio in LNPs by adjusting the lipid tail fluorinated with different numbers of fluorine atoms and varying the molar ratios of fluorinated lipids and ALC-0315. The first step was to explore the cellular uptake and mitochondrial uptake of varying F-LNPs/Cy5-pDNA. From top to bottom, as the number of fluorine atoms in the fluorinated ionizable lipids increased, both cellular uptake and mitochondrial uptake of gene cargo first increased and then decreased (Fig. 1D, E). Subsequently, we prepared various F-LNPs/mtGFP and used flow cytometry to quantitatively analyze the mitochondrial transfection efficiency and the percentage of mtGFP-positive cells. The mtGFP plasmid was specifically designed to eliminate the possibility that pDNA released into the cytoplasm could lead to mtGFP expression followed by mitochondrial import via a signal peptide, thereby artificially increasing mitochondrial fluorescence. This was achieved through two design strategies: (1) the insertion of a stop codon compatible with cytoplasmic codons^6,31 and (2) the omission of any mitochondrial targeting sequence in the plasmid construct^32,33. Both mitochondrial transfection efficiency and the percentage of mtGFP-positive cells initially increased and then decreased with higher fluorination levels or molar ratios of fluorinated lipids, which were consistent with the cellular and mitochondrial uptake results (Fig. 1F, G). Based on different numbers of fluorine atoms and the molar ratios of fluorinated lipids and ALC-0315, we calculated the mass ratios of fluorine atoms to total lipids (F wt%) in different F-LNPs and found that the F-LNP with a fluorine atom content comprising 7.94% of the total lipid mass, which optimized its mitochondrial transfection efficiency. For subsequent studies, we determined the optimal formulation for mitochondria-targeted gene delivery to be 6F, ALC-0315, DSPC, cholesterol, and DSPE-PEG2000 at a lipid molar ratio of 25:25:10:38.5:1.5. These nanoparticles were abbreviated as 6F-LNPs, with the F wt% in 6F-LNPs being 7.94% (Fig. 1H).

Given that conventional mitochondrial targeting groups largely rely on MMP, which can be significantly diminished under pathological conditions, as reported in previous literature^34,35, we sought to determine whether fluorination could mediate MMP-independent mitochondrial targeting, thereby expanding its application under disease conditions. We prepared FITC-labeled 6F-LNPs using FITC-labeled DSPE-PEG2000. N2a cells were pre-incubated with or without the uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP, 50 µM) for 1 h. Then, free mitochondria were isolated from CCCP-pretreated and untreated cells and incubated with FITC-labeled 6F-LNPs. Meanwhile, MMP-dependent Rhodamine 123 (Rh123) dye was used as the control group. As shown in Fig. 1I, the proportion of FITC-positive mitochondria in the 6F-LNPs group reached 90.8 ± 0.67% and 90.8 ± 1.84%, with no significant difference observed between the CCCP-pretreated and untreated groups. In contrast, the FITC-positive mitochondria in the CCCP-treated Rh123 dye group were significantly lower than those in the untreated group. This indicated that the uncoupler CCCP did not affect 6F-LNP-mediated mitochondrial targeting but significantly reduced the mitochondrial targeting efficiency of MMP-dependent Rh123. Overall, 6F-LNPs with 7.94% fluorine atoms enabled gene delivery to mitochondria in an MMP-independent manner.

The mechanistic framework for how fluorinated lipid nanoparticles deliver the cargo DNA to the mitochondrial matrix

To further explore the mechanistic study of fluorinated lipid-mediated mitochondria-targeted gene delivery, we validated this process at multiple levels, including the binding affinity between fluorinated molecules and mitochondrial membrane, transmembrane process, and mitochondrial transport behavior of 6F-LNP (Fig. 2A).

We attempted to determine which specific components of the mitochondrial membrane have a stronger affinity for the fluorinated groups. Previous studies have speculated that fluorinated groups tend to interact with mitochondrial lipid components, which could be due to the strong binding capacity of fluoroamphiphiles with mitochondrial hydrophobic structures (such as membrane phospholipids)³⁶. Therefore, the interaction between fluorinated molecules and mitochondrial membrane lipids was carried out by non-targeted lipidomic analysis. Firstly, undecylfluorhexylamine, as a classical fluorinated group was grafted to the surface of carboxyl magnetic beads (300 nm) to form undecylfluorhexylamine-grafted magnetic beads by Michael’s addition reaction. The mitochondrial lipids extracted from isolated mitochondria were redissolved by methanol and methyl tert-butyl ether and incubated with undecylfluorhexylamine-grafted magnetic beads (named fluorinated group). After magnetic separation and removal of the supernatants, samples were obtained by elution of magnetic beads with organic solvent and analyzed by LC-MS/MS. The samples obtained from oleic acid-grafted beads (named non-fluorinated group) incubated with mitochondrial lysate were used as the control group. Experimental data indicated that the lipid fractions eluted from the fluorinated group exhibited an increase in cardiolipin (CL) and phosphatidylcholine (PC) by 38.0% and 20.8%, respectively, compared to the non-fluorinated group (Fig. 2B). Notably, CL is a lipid component unique to the mitochondrial membrane³⁷, supporting the mitochondrial specificity of F-LNPs. To further validate this result, we constructed two types of mitochondrial membrane mimicking liposomes composed of key lipids identified in the lipidomics screening (CL and PC). FITC-labeled LNPs (6F-LNPs or 0F-LNPs) were incubated with each type of liposome, and their binding affinities were evaluated using microscale thermophoresis (MST). The result showed that the dissociation equilibrium constant (Kd) between 6F-LNPs and CL was significantly lower than that of 0F-LNPs, suggesting a stronger binding affinity between 6F-LNPs and the mitochondrial-specific lipid component CL (Fig. 2C).

Molecular dynamics simulation was designed to analyze the process of 6F-LNPs traversing the inner mitochondrial membrane. As shown in the snapshots at different time points, 6F-LNPs first approached and embedded into the inner mitochondrial membrane. Over the subsequent 50-75 ns, the penetration of the nanoparticle induced a downward invagination of the membrane. By 75 ns, the nanoparticle began to detach from the mitochondrial membrane. During the 75-100 ns simulation period, 6F-LNPs progressively translocated across and fully disengaged from the membrane (Fig. 2D). To assess the stability of the 6F-LNPs and the mitochondrial membrane during the simulation, the root mean square deviation (RMSD) of each component in the system was calculated. RMSD quantifies the degree of structural deviation from the initial conformation, measured in angstroms (Å), with values below 2 Å indicating minimal structural changes and values exceeding 2 Å suggesting significant conformational alterations. The analysis revealed a progressive increase in the system’s RMSD over time (Supplementary Fig. 10A), indicating that the mitochondrial membrane underwent increasing perturbation due to nanoparticle interaction as the simulation progresses. The RMSD of 6F-LNPs showed a gradual increase over time. However, unlike the mitochondrial membrane, the RMSD of 6F-LNPs remained below 2 Å throughout the transmembrane process, suggesting that its overall three-dimensional structure remained stable. Combined with Fig. 2D, these findings indicated that while the mitochondrial membrane underwent significant structural perturbation during translocation, the structural integrity of 6F-LNPs was preserved, with only minor conformational changes. By analyzing the system’s compactness through the radius of gyration (Rg), the results indicated that the mitochondrial membrane became increasingly disordered over the course of the simulation (Supplementary Fig. 10B). The Rg value of 6F-LNPs initially decreased before returning to its original level. The decrease suggested that during membrane entry, the nanoparticle experienced resistance from the mitochondrial membrane, leading to a more compact structure. After translocation, the Rg value of 6F-LNPs gradually returned to its initial state, indicating the restoration of its original compactness (Supplementary Fig. 10C). Analysis of the solvent-accessible surface area (SASA) revealed a progressive increase as the nanoparticles penetrated, indicating that the mitochondrial membrane temporarily loosened upon the entry of 6F-LNPs, which facilitated nanoparticle permeation (Supplementary Fig. 10D). Notably, the SASA of 6F-LNPs initially decreased before increasing, suggesting that the nanoparticles underwent compression during transmembrane transport, leading to a transient reduction in SASA (Supplementary Fig. 10E). Further analysis of the number of hydrogen bond (HBNUM) between 6F-LNPs and the mitochondrial membrane showed a gradual increase followed by a decrease, reflecting the dynamic nature of their interactions (Supplementary Fig. 10F). Energy calculations during molecular dynamics simulations indicated a continuous decline in system energy from 0 ns, demonstrating a convergence process. This suggested that as diffusion progressed, the total system energy tended to decrease, signifying a more stable state and confirming that the diffusion process occurred spontaneously. In terms of energy dynamics, a sharp decline in system energy was observed between 30 and 75 ns, which coincided with the nanoparticle permeation event. This rapid energy shift suggested that the system stabilized more quickly during the penetration process (Supplementary Fig. 10G). As shown in Supplementary Fig. 10H, the total free energy remained negative, indicating the spontaneous nature of this permeation process. Breaking down the individual energy components, the van der Waals interaction energy in total gas phase free energy (GGAS) was negative (VDWAALS < 0), suggesting that hydrophobic interactions contributed to binding. Additionally, the total solvation free energy (GSOLV) was negative, indicating that solvation favored binding. Molecular dynamics simulations initially revealed the process of 6F-LNPs traversing the inner mitochondrial membrane, as well as the structural and energetic changes of both the mitochondrial membrane and 6F-LNPs during this process. Hydrophobic interactions and polar solvation were found to facilitate the interactions between fluorinated molecules and the mitochondrial membrane.

At the cellular level, we further evaluated whether 6F-LNPs could mediate mitochondria-targeted gene delivery. Experimental verification of mitochondrial transport behavior included assessments of 6F-LNP affinity for isolated mitochondria, mitochondrial colocalization analysis using confocal laser scanning microscopy (CLSM), and mitochondrial transport tracing using transmission electron microscopy (TEM). We investigated the affinity of different F-LNPs for extracted mitochondria. We co-incubated FITC-labeled F-LNPs with free mitochondria extracted from N2a cells. Prior to measuring the mean fluorescence intensity (MFI) of FITC-labeled F-LNPs, we pre-treated the isolated mitochondria with proteinase K (proK) to eliminate the outer mitochondrial membrane. We prepared various F-LNPs with different F wt%, including 0.00, 5.57, 7.94 and 10.06%. We found that the MFI in the free mitochondria increased with increasing F wt%. The F-LNPs with 7.94% demonstrated the highest affinity for free mitochondria (Fig. 2E). The result from the ex vivo mitochondrial experiment was consistent with those from cellular experiments (Fig. 1E). The variation in the affinity of F-LNPs with different F wt% for extracted mitochondria appeared to explain why the mitochondrial transfection efficiency of F-LNPs began to decline when the fluorine content exceeded 7.94% (Fig. 1F, G). We inferred that an increase in the fluorination ratio led to a greater binding affinity of F-LNPs with 10.06% fluorine atoms to itself compared to its affinity for membrane structures, thereby reducing its mitochondrial transfection efficiency. To verify this hypothesis, we examined the changes in nanoparticle size of F-LNPs at different time points. As the fluorine content increases, the particle sizes of F-LNPs showed minimal variation (Supplementary Fig. 11). The enhanced self-affinity of F-LNPs with 10.06% fluorine atoms resulted in slight changes in particle size and a more uniform size distribution. However, the affinity of F-LNPs with 10.06% fluorine atoms for mitochondria was significantly lower than that of F-LNPs with 7.94% fluorine atoms.

The proK protection assay was conducted to investigate whether F-LNPs can overcome the mitochondrial double membrane barrier and achieve mitochondria-targeted gene delivery³⁸. We treated N2a cells with 6F-LNP/Cy3-pDNA and then extracted mitochondria from these cells. CLSM images exhibited that in the absence of proK treatment, the outer membranes and matrix of free mitochondria were labeled green and red by TOMM20 and SDHA antibodies, and these overlapped with the blue fluorescence of 6F-LNP/Cy3-pDNA (Fig. 2F). This confirmed that 6F-LNP/Cy3-pDNA colocalized with mitochondria; however, it remained unclear where exactly the Cy3-pDNA in 6F-LNP/Cy3-pDNA was localized within the mitochondria-whether it was on the outer mitochondrial membrane, in the intermembrane space, or had entered the mitochondrial matrix. To clarify this, proK treatment was applied to the isolated mitochondria to remove the outer membrane, eliminating interference from the outer membrane and the intermembrane space, thus allowing a detection of genetic cargo that had entered the mitochondrial matrix. As shown in the CLSM images of proK (+) treatment, proK treatment indeed resulted in the disappearance of the outer mitochondrial membrane, as evidenced by the loss of fluorescence in the TOMM20 channel. The SDHA-labeled mitochondrial matrix protein fluorescence continued to merge with the Cy3-pDNA fluorescence, demonstrating that fluorinated lipids could assist nanoparticles in delivering genetic cargo to the mitochondrial matrix rather than to the outer mitochondrial membrane or intermembrane space. In addition, we found that the 0F-LNP/Cy3-pDNA did not colocalize with either the mitochondrial matrix protein (SDHA) or the outer membrane marker (TOMM20). This indicated that 0F-LNPs were incapable of delivering genetic cargo into the mitochondrial matrix.

To visualize the mitochondria-targeted delivery process mediated by 6F-LNPs, we used TEM to capture the spatial relationship between 6F-LNPs and cellular mitochondria. 6F-LNPs with gold colloid (10 nm) were prepared to trace the intracellular transport of 6F-LNPs (Supplementary Table 1). It was observed that 6F-LNP/gold colloids surrounded the mitochondria (Fig. 2G(i)), with some 6F-LNP/gold colloids embedded in the outer and inner mitochondrial membranes (Fig. 2G(ii)). More importantly, after 12 h, some 6F-LNP/gold colloids were already located within the mitochondrial matrix (Fig. 2G(iii)).

Therefore, the mechanistic study of how 6F-LNPs deliver cargo DNA to the mitochondrial matrix was conducted from three perspectives: the binding affinity between fluorinated molecules and mitochondrial membrane, transmembrane process and mitochondrial transfer behavior. These results demonstrated that fluorinated groups enhanced the affinity of 6F-LNPs for the mitochondrial membrane, particularly for the CL components unique to mitochondrial membranes. Under the combined effects of hydrophobic interactions and polar solvation, 6F-LNPs exhibited increased interactions with the mitochondrial membrane. The entry of 6F-LNPs disrupted the mitochondrial membrane structure, temporarily reducing its stability and compactness. During the permeation process, the mitochondrial membrane exhibited a downward invagination trend, facilitating the translocation of 6F-LNPs across the membrane. Notably, 6F-LNPs maintained good stability and structural integrity throughout the transmembrane process, which was beneficial for escorting cargo DNA into the mitochondria. Additionally, cargo DNA remained protected within 6F-LNPs, ensuring its targeted delivery to the mitochondrial matrix.

Mitochondrial targeting sequence-modification enhances mitochondria-selective gene delivery

As mentioned above, 6F-LNPs enabled gene delivery to mitochondria in an MMP-independent manner. However, CLSM results of 6F-LNP-mediated mitochondrial targeting revealed that although the Pearson’s correlation coefficient (R value) between Cy5-pDNA and mitochondria in the 6F-LNP group was 0.79, not every cell in the 6F-LNP group exhibited strong colocalization (Supplementary Fig. 12). To enhance mitochondrial selectivity, we incorporated MTS-modified DSPE-PEG into the 6F-LNP formulation. We anticipated that MTS would improve the mitochondrial selectivity of fluorinated LNPs while leveraging the properties of fluorinated lipids to overcome the limitations of MTS, such as its difficulty in assisting nanoparticles in penetrating the mitochondria and its MMP-dependent mitochondrial targeting.

Lipids including 6F, ALC-0315, DSPC, cholesterol, DSPE-PEG2000 and DSPE-PEG-Mal were solubilized in ethanol and pDNA was added in the water phase. MTS peptide (MLSLRQSIRFFKC) was covalently attached on the surface of LNPs by thiol-maleimide coupling between DSPE-PEG-Mal and the terminal thiol group of MTS (Fig. 3A). The molar ratio between DSPE-PEG2000 and DSPE-PEG-Mal was screened using cellular uptake and mitochondrial gene transfection efficiency. The formulations of 6F-M-LNPs involved 6F, ALC-0315, DSPC, cholesterol, DSPE-PEG2000 and 1,2-Distearoyl-sn-glycero-3-phosphoethanolamine-N-[maleimide(polyethylene glycol)-2000] (DSPE-PEG2000-Mal) with 25:25:10:38.5:(1.5 – b):b mol% (b = 0, 0.5, 0.75, 1, 1.5). These results in Fig. 3B, C confirmed that when the molar ratio of DSPE-PEG2000-MTS was set at 0.75% (b = 0.75%), the cellular uptake and mitochondrial gene transfection efficiency of 6F-0.75%MTS were higher than those of other groups. Following a series of optimizations in the F wt% and the molar ratio of MTS modification, the formulation of the nanoparticles (abbreviated as 6F-M-LNP) with the optimal mitochondrial transfection efficiency was determined to consist of 6F, ALC-0315, DSPC, cholesterol, DSPE-PEG2000, and DSPE-PEG2000-Mal with molar ratios of 25:25:10:38.5:0.75:0.75.

After formulating these LNPs, the physicochemical properties of 6F-M-LNP were characterized. The particle size of 6F-M-LNP was determined as 130.52 ± 0.77 nm with a low polydispersity index (0.259 ± 0.014). The zeta potential of 6F-M-LNP was −10.20 ± 0.45 mV. The encapsulation efficiencies (EE) of 6F-LNPs and 6F-M-LNPs were 92.99 ± 2.01% and 94.10 ± 3.53%, respectively (Fig. 3D). In addition, we detected the stability of 6F-LNPs and 6F-M-LNPs. We found that the effective diameters of these two LNPs were between 100 and 150 nm and that they ensure high stability of nanoparticle size in PBS buffer at pH 7.4 for up to 7 days. We also prepared a buffer solution that simulates the vitreous humor based on previously reported literature and examined the 7-day stability of nanoparticle sizes of 6F-LNP and 6F-M-LNP in this medium^39,40. This experimental result showed that both 6F-LNP and 6F-M-LNP exhibited no significant changes in particle size over 7 days, demonstrating that these nanoparticles retained their sizes under vitreous humor conditions (Supplementary Fig. 13A). Under the optimal formulation conditions, 6F-M-LNPs were able to form a uniform spherical shape, although with some variability in morphology (Fig. 3E). The pKa of 6F-M-LNP was determined to be 6.47 using the 2-(p-Toluidino)naphthalene-6-sulfonic acid (TNS) probe (Supplementary Fig. 13B), which facilitated lysosomal escape while allowing the nanoparticles to maintain a negative charge under neutral conditions. This property enhanced biological safety by minimizing cytotoxic effects. We then assessed the in vitro safety of 0F-M-LNP and 6F-M-LNP. For comparison, we prepared a control group of LNPs without fluorinated lipids⁴¹. Through MTT assays and the evaluations of mitochondrial function (ATP generation), we found that both 0F-M-LNP and 6F-M-LNP exhibited no significant cytotoxicity and did not impair mitochondrial function (Supplementary Fig. 14).

We further evaluated whether MTS modification could enhance mitochondria-selective gene delivery through CLSM. As shown in Fig. 3F, G, the colocalization coefficient between the Cy5-pDNA cargo delivered by 6F-M-LNP and the mitochondria was as high as 0.90, significantly higher than that of the 6F-LNP group without MTS modification. This demonstrated that MTS modification enhanced mitochondria-selective gene delivery. However, the MTS modification in 0F-LNPs appeared to have little effect on the colocalization coefficient. By isolating mitochondria after administration of each formulation and measuring the MFI of Cy5-pDNA within mitochondria, we observed the same results: MTS modification significantly enhanced the efficiency of 6F-M-LNPs in mitochondria-selective gene delivery but did not affect the efficiency of 0F-M-LNPs (Fig. 3H). Therefore, we hypothesized that fluorinated lipids played a more critical role in mitochondria-selective gene delivery.

The TOM complex is widely recognized as the primary gateway for MTS-mediated protein import into mitochondria^42,43. Therefore, several representative TOM complex proteins (TOMM20, TOMM22, TOMM40, and TOMM70) were further examined by western blot. 6F-M-LNPs exhibited higher adsorption of TOMM20 and TOMM22 compared to 6F-LNPs (Fig. 3I). However, the levels of TOMM40 and TOMM70 adsorbed onto 6F-M-LNPs were not different from those on 6F-LNPs. Furthermore, we evaluated the mitochondria-targeted gene transfection efficiency mediated by 6F-M-LNPs after siRNA-mediated knockdown of Tomm20 and Tomm22. The siRNA sequences were designed based on previously reported literature¹³, and their gene silencing efficiencies were verified. The knockdown efficiencies for Tomm20 and Tomm22 were 94.28 ± 1.19% and 93.84 ± 0.07%, respectively (Supplementary Fig. 15). Notably, treatment with siTomm20 or siTomm22 significantly reduced the mitochondrial gene transfection efficiency of 6F-M-LNPs, whereas treatment with a negative control siRNA (siNC) had no effect (Fig. 3J). These results suggested that the MTS moiety in 6F-M-LNPs facilitated selective binding to TOMM20 and TOMM22 on the outer mitochondrial membrane.

The mitochondrial gene transfection was evaluated in the other cell lines, including HEK293T and SH-SY5Y cells. The MFI of mtGFP in the 6F-M-LNPs/mtGFP group was obviously higher than that in the other groups, suggesting that 6F-M-LNP demonstrated effective mitochondria-targeted gene delivery across multiple cell lines (Supplementary Fig. 16A, B). In addition, we have investigated the mitochondrial gene transfection of 6F-M-LNP in various cell lines under mitochondrial impairment conditions. Since it is currently challenging to establish mtDNA mutation cell models across a wide range of cell lines, we followed previously reported studies and employed rotenone (Rot), a mitochondrial complex I inhibitor, to pre-treat various cell lines and thereby construct mitochondrial impairment models^44,45. We found that 6F-M-LNP retained its mitochondrial gene transfection capability even in different cell lines under mitochondrial impairment conditions (Supplementary Fig. 16C, D).

Underlying rationale for mitochondria-targeted gene delivery of 6F-M-LNP

To understand the mitochondria-targeted gene delivery performance mediated by 6F-M-LNP, we have investigated the key processes in mitochondrial gene delivery, including the cellular uptake mechanisms, intracellular distribution, and mitochondria-targeted transport.

For cellular uptake efficiency and mechanistic study, we carried out Cy5-labeled pDNA to detect the cellular internalization of 6F-M-LNP. The result of flow cytometry showed a stronger MFI in the 6F-M-LNP group compared to that in the other groups, suggesting that 6F-M-LNP exhibited more efficient cellular internalization compared to 0F-M-LNP and 6F-LNP (Supplementary Fig. 17A). In addition, various inhibitors were applied to study cell uptake mechanisms. Amiloride, genistein and chlorpromazine were respectively inhibitors of macropinocytosis, caveolae-mediated endocytosis, and clathrin-mediated endocytosis⁴⁶. The internalization of 6F-M-LNP/Cy5-pDNA was mediated through multiple uptake pathways, with caveolae-mediated endocytosis being the predominant mechanism (Supplementary Fig. 17B). It has been reported that caveolae-mediated endocytosis transported genetic cargo into the cytoplasm in the form of neutral caveosomes, with a small portion entering acidic lysosomes^47,48.

We subsequently examined the intracellular distribution of Cy5-pDNA using CLSM. The colocalization of Cy5-pDNA (red) with mitochondria (green) or lysosomes (green) was analyzed, respectively (Supplementary Fig. 18). After 3 h of incubation, both 0F-M-LNP/Cy5-pDNA and 6F-M-LNP/Cy5-pDNA exhibited the colocalization with lysosomes. At 8 h, the R value between 0F-M-LNP/Cy5-pDNA and lysosomes remained high at 0.61, whereas that of 6F-M-LNP/Cy5-pDNA decreased to 0.52, indicating that fluorinated lipids promoted lysosomal escape⁴⁹. With prolonged incubation (12 h), the R value for 6F-M-LNP/Cy5-pDNA and mitochondria progressively increased. In contrast, 0F-M-LNP/Cy5-pDNA showed lower mitochondrial colocalization, with a considerable portion remaining cytosolic even after 12 h. The result suggested that fluorination not only enhanced lysosomal escape but also substantially improved mitochondrial targeting efficiency compared with non-fluorinated 0F-M-LNPs. We performed qPCR to quantitatively evaluate the efficiency of 6F-M-LNPs in selectively targeting mitochondria. 6F-M-LNPs were loaded with model plasmid DNA (mtGFP), and 0F-M-LNP/mtGFP was used as the control group. Specifically, N2a cells were treated with different formulations, followed by mitochondrial isolation using a Mitochondria Isolation Kit. The isolated mitochondria were then treated with proK and DNase to remove any DNA adherent to the mitochondrial surface. Simultaneously, the remaining cellular components (excluding mitochondria), including the cytoplasm, were prepared for the qPCR assay. The level of mtGFP in both the mitochondrial and non-mitochondrial fractions was detected based on the results of qPCR assay. As shown in Supplementary Fig. 19, 6F-M-LNP treatment resulted in mitochondrial mtGFP levels that were approximately 9-fold and 12-fold higher than those observed in the LNP and 0F-M-LNP treatment groups, respectively.

To supplement target proteins within the mitochondria, 6F-M-LNP should effectively transport the genetic cargo to the mitochondria, particularly to the mitochondrial matrix. We began our investigation by exploring the specific location within the mitochondria, using the Multi-SIM (Multimodality Structured Illumination Microscopy) and proK protection assay⁵⁰. The images in Multi-SIM showed that 6F-M-LNPs were able to deliver Cy5-pDNA (green) into the internal cristae structure of mitochondria, where it colocalized with the inner mitochondrial membranes (red) (Fig. 3K). Moreover, the proK protection assays also confirmed that the fluorescently labeled gene cargo could be delivered by 6F-M-LNP to the mitochondrial matrix rather than adhering to the outer mitochondrial membrane or residing in the intermembrane space (Fig. 3L).

The mtGFP was carried out to evaluate 6F-M-LNP-mediated mitochondrial gene transfection. After different formulations treatment, we isolated mitochondria from N2a cells and detected the proportion of mtGFP-positive mitochondria in the total extracted mitochondria. As shown in Supplementary Fig. 20, the mtGFP-positive mitochondria were 16.87% in the 6F-M-LNP group, which was obviously higher than that in the PBS and 0F-M-LNP groups. These results indicated that pDNA could be efficiently delivered into mitochondria and expressed as the target protein.

In summary, the underlying rationale for 6F-M-LNP enabling mitochondria-targeted gene delivery was that 6F-M-LNP significantly facilitated the gene drugs in overcoming various intracellular delivery barriers, including the cellular membrane barrier, lysosomal degradation, and the mitochondrial double membrane barrier. The 6F-M-LNP maximized the delivery of nanoparticles to the cytoplasm through caveolae-mediated endocytosis and the lysosomal escape capability of fluorinated ionizable lipids. Furthermore, the modification with MTS significantly enhanced the interaction of 6F-M-LNP with the TOM complex on the mitochondrial membrane. Ultimately, under the influence of fluorinated lipids, 6F-M-LNP enabled mitochondrial membrane penetration, transporting the genetic cargo into the mitochondrial matrix (Fig. 3M).

Excellent performance in enhancing mitochondrial protein expression in cells derived from Leber’s hereditary optic neuropathy patients

Leber’s hereditary optic neuropathy (LHON) is a typical example of an ocular mitochondrial disease. Nearly all LHON patients are associated with mutations in one of three mtDNA-encoded subunits of mitochondrial complex I, a key component of the mitochondrial respiratory chain⁵¹. The lack of adequate clinical treatments for correcting mutated genes in LHON leads to severe, asynchronous bilateral vision loss, particularly in young adult males⁵². However, the mitochondrial double membrane barrier is considered a significant challenge to the delivery of gene therapeutics to mitochondria for the functional complementation of mutant mtDNA under pathological conditions. The optimal fluorinated lipid nanoparticle with MTS (6F-M-LNP), by virtue of its ability to gene delivery to mitochondria in an MMP-independent manner, is anticipated to facilitate the expression of target mitochondrial proteins as a potential treatment for LHON.

About 50% of patients with LHON carry the G11778A mutation, leading to the 340^th amino acid of the ND4 gene in complex I to change from arginine to histidine, making it a promising target for mtDNA delivery⁵³. We constructed a plasmid (human ND4, hND4) optimized for mitochondrial codons that encodes hND4, Flag, and luciferase for subsequent in vitro and in vivo studies (Supplementary Fig. 21A and Supplementary Sequence 2)⁶. The mitochondrial gene transfection of 6F-M-LNP/hND4 in N2a cells was confirmed using the detection of luciferase activity and the expression of Flag. 6F-M-LNP/hND4 treatment significantly enhanced luciferase activity (Supplementary Fig. 21B), suggesting that the 6F-M-LNP loaded with the hND4 functional gene still achieved superior mitochondrial gene transfection efficiency compared to other treatments. The immunofluorescence (IF) images showed that the Flag fluorescence intensity in the 6F-M-LNP/hND4 group was higher than that in other treatment groups. Remarkably, the Flag fluorescence overlapped with the COX IV-labeled mitochondrial fluorescence, confirming that ND4 was indeed expressed within the mitochondria under the influence of mitochondrial codons (Supplementary Fig. 21C).

To better align with the key characteristic that LHON is caused by mitochondrial gene mutations, patient cells with mutations in G11778A (LHON disease cells GM10742) were used to evaluate the in vitro therapeutic efficacy of 6F-M-LNP-mediated mitochondrial gene therapy^54,55. If the wild-type ND4 protein is successfully integrated into the functional respiratory complex, the respiratory dysfunction in LHON cells will be restored. The LHON disease cells GM10742 were treated with 6F-M-LNP/hND4 and other formulations for further evaluation (Fig. 4A). Cy5-labeled pDNA was used to prepare 6F-M-LNP/Cy5-pDNA, and the mitochondria-targeted gene delivery capability in GM10742 was detected using CLSM. The 6F-M-LNP enabled gene delivery to mitochondria in LHON disease cells GM10742 (Supplementary Fig. 22). The mitochondrial gene transfection efficiency was assessed by luciferase activity and hND4 expression. These results suggested that 6F-M-LNP/hND4 promoted the luciferase activity and hND4 expression compared with the other formulations, indicating that the 6F-M-LNPs maintained their mitochondria-targeted gene delivery capability in patient cells with G11778A mutations (Fig. 4B–D).

Key mitochondrial functions were investigated. The 6F-M-LNP/hND4 treatment repaired the damaged MMP level (Fig. 4E) and promoted ATP generation (Fig. 4F) in LHON disease cells GM10742. The 6F-M-LNP-mediated mitochondrial gene therapy significantly reduced the mitochondrial ROS level in LHON disease cells GM10742 (Fig. 4G). Oxygen consumption rate (OCR) assay is a widely used experimental technique to measure cellular respiration. The OCR curves were recorded under different treatments (Fig. 4H). Various parameters of mitochondrial function were analyzed according to Fig. 4H. As shown in Fig. 4I, 6F-M-LNP/hND4 increased the mitochondrial functions, including basal respiration (the energy demand of the cell under baseline conditions), ATP production, maximal respiration (the maximum capacity of the mitochondria to consume oxygen). Although all mitochondrial respiration-related indicators in the 6F-M-LNP/hND4 group were significantly higher than in other groups, they were similar to those in the 6F-LNP/hND4 group. This may be due to the enhanced membrane penetration capability of fluorinated lipids playing a more crucial role in mitochondria-targeted gene delivery in pathological states.

These results indicated that the optimized formulation of fluorinated lipid nanoparticles with MTS (6F-M-LNP) exhibited excellent performance in enhancing mitochondrial protein expression in cells derived from LHON patients, thereby restoring mitochondrial functions (Fig. 4J).

Inhibition of Leber’s hereditary optic neuropathy progression in a mouse model through the regulation of mitochondrial gene expression

Inspired by the superiority of in vitro mitochondrial gene transfection efficiency, we explored the intraocular delivery behavior of 6F-M-LNP and its ability to functionally complement damaged and mutant mtDNA in vivo. We examined whether 6F-M-LNP could facilitate the import of an exogenous functional gene in male mice, with a particular focus on achieving the expression of hND4 protein in retinal ganglion cells (RGCs) within the eye, as RGCs are the primary targeted cells affected in LHON patients⁵⁶. The hND4 was used as a functional pDNA for the treatment of LHON in mice. This was because the hND4 mtDNA levels in mitochondria reached 80% of its mouse homolog, and previous research has shown that hND4-based therapy can be used in the study of LHON in mice¹⁴. We detected the expression of hND4 protein at different time points in vivo. Within the 1-month observation period, 6F-M-LNP/hND4 led to sustained expression of the hND4Flag protein, with the protein expression peaking 3–7 days after administration. RGCs are a layer located near the inner side of the retina and are among the most crucial types of neurons in the visual pathway⁵⁷. IF staining also provided a direct visualization, showing a pronounced perinuclear distribution of hND4Flag protein, with hND4Flag protein (green) co-localizing with COX IV-labeled mitochondria (red) (Supplementary Fig. 23). Hence, 6F-M-LNP/hND4 treatment promoted the expression of the hND4 protein in the mitochondria of targeted cells in vivo, with the functional protein remaining consistently expressed for up to 14 days following a single administration.

The safety of various doses of 6F-M-LNP/pDNA (6F-M-L: 1 μg pDNA/eye; 6F-M-M: 2 μg pDNA/eye; 6F-M-H: 4 μg pDNA/eye) was assessed in vivo (Supplementary Fig. 24A). The mouse weight in different groups was distributed between 20–25 g, suggesting that various doses of 6F-M-LNP/pDNA did not cause significant systemic toxicity (Supplementary Fig. 24B). Additionally, there were no significant differences in organ index between the groups treated with different doses of 6F-M-LNP/pDNA and the PBS group (Supplementary Fig. 24C). After administration, liver and kidney functions remained within normal ranges across all groups (Supplementary Fig. 24D–H). We also examined the structure of the major organs, as well as the cornea and retina using H&E staining. None of the doses of 6F-M-LNP/pDNA caused noticeable damage to the major organs, cornea, or retina (Supplementary Fig. 24I). These results indicated that 6F-M-LNP demonstrated high in vivo safety and could be utilized for further studies in mitochondrial gene delivery.

Next, we detected whether the delivery of hND4 into mitochondria would repair visual damage in mice. We established a mutant ND4 mitochondria transgenesis (mtTg) LHON mouse model to validate the in vivo therapeutic effects of 6F-M-LNP/hND4. The mutant ND4 mCherry was encapsulated into the rAAV backbone and microinjected into fertilized eggs. Female founders (F0) exhibiting mCherry fluorescence on ophthalmoscopy were backcrossed with normal males for producing the first generation of offspring (F1): mutant ND4 mtTg LHON model male mice^14,58. Once mutant ND4 mtTg LHON model male mice reached 3 months of age, the phenotype of the transgenic mice named Mut appeared grossly normal, except for a larger body size compared to age-matched wild-type male mice named WT (Supplementary Fig. 25A). This increase in body size was attributed to systemic expression of the mutant ND4 protein, leading to mitochondrial dysfunction and subsequent metabolic abnormalities. In addition, RGC loss was also evident in male mice aged 3 months (Supplementary Fig. 25B), indicating that mutant ND4 protein caused loss of RGCs, the hallmarks of LHON patients. Electroretinogram (ERG) assay was carried out to analyze the changes in visual function. The ERG waveform indicated that mutant ND4 protein significantly reduced the amplitudes of both a- and b-waves compared to the WT mice, suggesting that mutant ND4 protein could lead to vision loss in mice (Supplementary Fig. 25C).

After the mutant ND4 mtTg LHON model male mice reached 3 months of age, these mice were randomly divided into three groups: the Mutant (Mut) group, the Idebenone (Ide) group, and the 6F-M-LNP/hND4 group. These mutant ND4 mtTg LHON model male mice received intravitreal injections of PBS or 6F-M-LNP/hND4 every 14 days for a total of two treatments. Ide is the positive drug approved by the European Medicines Agency for clinical treatment of LHON, which is set as a positive control group with daily oral gavage administration (60 mg/kg). Three-month-old wild-type C57BL/6J male mice were set as a control group, named WT group. Twenty-eight days after the first treatment, optomotor and ERG assays were carried out to analyze the changes in visual functions. Following these tests, the mice were sacrificed to assess the total ND4 protein expression in eye and ATP levels in the retina (Fig. 5A). The total ND4 protein refers to the combined expression of hND4 protein and mouse ND4 protein within the eyes of the mutant ND4 mtTg LHON model male mice, rather than hND4 protein alone. The total ND4 expression in mutant ND4 mtTg LHON model male mice was significantly lower than that in the WT group, while the 6F-M-LNP/hND4 treatment promoted the total content of ND4 protein (Fig. 5B, C). The 6F-M-LNP/hND4 enhanced the ATP levels compared with the Mut and Ide groups (Fig. 5D). In addition, the visual functions in mice were determined using optomotor and ERG tests. As shown in Fig. 5E and Supplementary Movie 1–4, the number of head movements in the 6F-M-LNP/hND4 group was significantly higher than in the Mut and positive control Ide groups, and there was no significant difference compared to the WT group. The ERG waveform and the quantitative data for the a- and b-waves indicated that 6F-M-LNP/hND4 significantly increased the values of both a- and b-wave amplitude compared to the Mut mice, suggesting that 6F-M-LNP/hND4 could improve vision loss caused by mitochondrial gene mutation (Fig. 5F–H). We hypothesized that the recovery of vision in the mice was due to the protective effect of 6F-M-LNP/hND4 treatment on RGCs, shielding them from apoptosis caused by the gene mutation. The images of H&E staining also supported this hypothesis, showing that the number of RGCs in the 6F-M-LNP/hND4 treatment group was significantly higher than in the other groups and comparable to that in the WT group (Fig. 5I, J). These results suggested that 6F-M-LNP/hND4 provided functional supplementation of the mutant gene and rescued visual dysfunction in LHON-like mouse models by delivering the hND4 plasmid into the mitochondria.

Mutations in the mtDNA result in mitochondrial dysfunction, ultimately causing various mitochondrial disorders without available cures. However, mitochondrial gene therapy remains a tricky challenge due to its double membrane barriers and severe pathological microenvironment, especially decreased MMP. Currently, mitochondria-targeted delivery strategies are mainly based on cationic hydrophobic groups or MTS-mediated transport, dependent on MMP. These inevitably lead to adverse events, including high cytotoxicity, reduced delivery efficiency in the presence of decreased MMP, or difficulty in the transportation of nanoparticles. Therefore, it is urgent to study how to overcome the mitochondrial double membrane barrier and discard the dependence on the high MMP to achieve safe and effective mitochondria-targeted gene delivery. Inspired by the literature using the hydrophobic and lipophobic properties of fluorinated modifications to penetrate cell membranes, we hypothesized that fluorination groups provided a membrane translocation mechanism independent of MMP to achieve mitochondrial membrane penetration. Here, we have shown that fluorinated lipid nanoparticles with MTS can be applied to efficiently mitochondrial gene delivery in an MMP-independent manner.

The phenomenon that fluorinated lipids help nanoparticles deliver genetic cargo into the mitochondrial matrix inspires us to find the underlying mechanism. Due to the hydrophobic characteristics and lack of positive charge of fluorinated groups, fluorinated lipid nanoparticles are more likely to interact with mitochondrial hydrophobic structures, such as lipid components, resulting in high mitochondrial affinity. Non-targeted lipidomic analysis further demonstrated that cardiolipin, a unique component of mitochondria, has a stronger affinity for the F-LNPs. After binding to mitochondria, F-LNPs can traverse the mitochondrial membrane structure as quickly as traversing the membrane structure owing to hydrophobic and lipophobic properties. However, due to the lack of effective real-time monitoring methods for organelles, the specific mitochondrial penetration process of fluorinated lipid nanoparticles still needs to be further studied. Moreover, the fluorine content in F-LNPs was unexpectedly found to influence their efficiency in mitochondria-targeted gene delivery, prompting us to investigate the structural criteria necessary for efficient gene delivery to mitochondria using fluorinated lipids. Both mitochondrial uptake and transfection efficiency were highest when the fluorination to total lipid mass ratio reached 7.94% but decreased when it exceeded 7.94%.

MTS-modification has been shown to enhance mitochondria-selective gene delivery of fluorinated lipid nanoparticles. Based on MTS, the nanoparticles were able to locate the mitochondrial surface and then transport into the mitochondrial matrix using penetration properties of fluorinated lipids. Furthermore, the results in mtDNA mutation-induced cell model and male mice model all verified that fluorinated lipid nanoparticles with MTS still had good mitochondrial transfection and therapeutic effects under the pathological condition. It was further suggested that fluorinated lipid nanoparticles enabled gene delivery to mitochondria in an MMP-independent manner.

In summary, our work established a proof of concept for safe and efficient mitochondrial gene delivery mediated by fluorinated lipid nanoparticles. Fluorinated modified vectors may offer a potentially universal tool for delivering various gene therapies into mitochondria, enabling mitochondrial gene manipulation, including functional supplementation of mutant gene, mitochondrial gene silencing, and even mitochondrial gene editing.

Materials

ALC-0315, DSPC, cholesterol, DSPE-PEG2000 and DSPE-PEG2000-Mal were obtained from A.V.T. Pharmaceutical Co. Ltd. (Shanghai, China). Label IT^® Tracker™ Intracellular Nucleic Acid Localization Kit, Cy^®3 and Cy^®5 were purchased from Mirus Bio LLC (USA). MTS peptide (MLSLRQSIRFFKC) was synthesized by Go Top Peptide Biotech Co., Ltd. (Hangzhou, China). Endo-free Plasmid Maxi Kit was prepared by Omega Bio-tek, Inc. (USA). The siTomm22 and siTomm22 were obtained from Ribo Biotechnology Co., Ltd (Guangzhou, China). DMEM, DMEM/F12, RPMI 1640 were obtained from Jiangsu KeyGEN BioTECH Co., Ltd. (Nanjing, China). Quant-iT PicoGreen DNA Assay Kit, Lyso-Tracker Red, Mitotracker Red or Green or Deep Red, MitoSOX Red, SDS-PAGE and Lipofectamine 3000 were obtained from Thermo Fisher Scientific Inc. (USA). Minute™ Mitochondria Isolation Kit was purchased from Invent Biotechnologies, Inc. (USA). Carboxyl magnetic beads (300 nm) and gold colloid (10 nm in size) were prepared by Nanoeast Biotech Co., Ltd. (Nanjing, China). Cell Mitochondria Isolation Kit, Tissue Mitochondria Isolation Kit, 4,6-diamino- 2-phenylindole (DAPI), Mitochondrial Membrane Potential Test Kit with JC-1, Enhanced ATP Assay Kit, Hoechst 33342, proteinase K (proK) and Rhodamine 123 (Rh123) were obtained from Beyotime Biotechnology Co., Ltd. (Shanghai, China). Nano-Glo Luciferase Assay System was obtained from Promega Biotech Co., Ltd. (Beijing, China). Seahorse XF RPMI 1640 pH 7.4, glucose, l-glutamine, sodium pyruvate and Agilent Seahorse XF Cell Mito Stress Test Kit were obtained from Agilent Technologies, Inc. (USA).

Plasmid construction

We constructed the plasmid DNA (pDNA) containing key gene sequence and epitope tag^6,20. The concrete gene sequences were provided in Supplementary gene sequence, including the plasmid mitochondrial green fluorescent protein (mtGFP) and plasmid mitochondrial luciferase containing human ND4 gene (hND4-3xFLAG-mtLuc). For in vitro tracking experiments, pDNA was stained by Label IT^® Tracker™ Intracellular Nucleic Acid Localization Kit, Cy^®3 or Cy^®5.

Cell culture

N2a (American Type Culture Collection (ATCC), CCL-131), HEK293T (ATCC, CRL-3216), and SH-SY5Y (ATCC, CRL-2266) cell lines were obtained from ATCC. GM10742 (LCL from B-Lymphocyte) cells were purchased from the Coriell Institute for Medical Research. N2a cell line was cultured in MEM medium. HEK293T cell line was cultured in DMEM medium. SH-SY5Y cell line was cultured in DMEM/F12 medium. GM10742 cells were cultured in RPMI 1640 medium.

Animals

C57BL/6J male mice (6-8 weeks) were purchased from East China Normal University Laboratory Animal Technology Co. Ltd. (Shanghai, China). These mutant ND4 mtTg LHON model male mice (3 months of age) were obtained from Aurora Bioscience Co. Ltd. (Suzhou, China). The mice were maintained under a 12-h light/dark cycle at a temperature of 25 °C with 50% relative humidity. All animal procedures were conducted in accordance with protocols approved by the regional ethics committee of China Pharmaceutical University (2023-08-013) and JOINN Laboratories Co., Ltd. (Suzhou, China, S-ACU24-0975).

Synthesis of 4F (compound 1)

The synthesis route of 4F (compound 1) was shown in Supplementary Fig. 1. Compound 1a was synthesized from 1-hexene (9.0 mmol), 1,4-diiodoperfluorobutane (9.0 mmol), vinyl benzoate (9.0 mmol), Na₂S₂O₄ (17.8 mmol) and NaHCO₃ (18.0 mmol) in CH₃CN/H₂O (8:7 v/v) for 50 min at 0 °C. Then, the reaction mixture was adjusted to pH 5-7 with HCl 3 N. After stirring for 20 min at room temperature (r.t.), H₂O was added, and the aqueous layer was extracted with CH₂Cl₂. The organic layer was dried and concentrated. Silica gel chromatography (C₆H₁₄/CH₂Cl₂ = 95:5) was used to purify the residue to yield compound 1a (1.42 g, 22%).

Compound 1a (2.0 mmol) in 50 mL of Et₂O was mixed with tributyltin hydride (10.0 mmol) and AIBN (0.20 mmol), then heated under reflux for 15 h under the light of a halogen lamp before concentration under vacuum. Flash chromatography over silica gel (n-C₆H₁₄/CH₂Cl₂ = 9:1) was used to purify the residue to yield compound 1b (740 mg, 80%).

The solution of compound 1b (1.5 mmol) in methanolic 1 M LiOH was stirred for 3 h at r.t. before concentration. Then, the solution was extracted with CH₂Cl₂, followed by drying and concentrating. Silica gel chromatography (n-C₆H₁₃/CH₂Cl₂ = 1:1) was used to purify the residue to yield compound 1c (403 mg, 75%).

The solution of compound 1c (1.1 mmol) in acetone and H₂O was added dropwise with Jones’ reagent at r.t. to obtain a red-brown solution. The mixture was added with i-PrOH for another 15 min before concentration. Then, the mixture was added with H₂O, extracted with Et₂O, followed by drying and concentrating. The yield of compound 1d was 91% (372.7 mg).

Compound 1d (1.0 mmol) in CH₂Cl₂ was stirred with 3-(dipropylamino)propane-1,2-diol (0.5 mmol), DMAP (0.1 mmol) and DIC (1.0 mmol) for 24 h at r.t. before filtration and concentration. Silica gel chromatography (PE/EtOAc = 3:1) was used to purify the residue to yield compound 1 (310 mg, about 71%).

Synthesis of 6F (compound 2)

The synthesis route of 6F (compound 2) was shown in Supplementary Fig. 2. Compound 2a was synthesized from 1-hexene (9.0 mmol), 1,6-diiodoperfluorohexane (9.0 mmol), vinyl benzoate (9.0 mmol), Na₂S₂O₄ (17.8 mmol) and NaHCO₃ (18.0 mmol) in CH₃CN/H₂O (8:7 v/v) for 50 min at 0 °C. The following treatment was similar to compound 1a, 1b, 1c, 1 d and 1 to yield compound 2a (1.51 g, 20%), compound 2b (842 mg, 83%), compound 2c (516 mg, 80%), compound 2 d (450 mg, 85%) and crude compound 2. The crude product was purified by silica gel chromatography (PE/EtOAc = 3:1) to yield compound 2 (359 mg, about 35%).

Synthesis of 8F (compound 3)

The synthesis route of 8F (compound 3) was shown in Supplementary Fig. 3. Compound 3a was synthesized from 1-bromobutylene (5.0 mmol), 1,8-diiodoperfluorooctane (5.0 mmol), vinyl benzoate (5.0 mmol), Na₂S₂O₄ (10.0 mmol) and NaHCO₃ (10.0 mmol) in CH₃CN/H₂O (8:7 v/v) for 50 min at 0 °C. The following treatment was similar to compound 1a, 1b, 1c, 1d and 1 to yield compound 3a (0.94 g, 20%), compound 3b (320 mg, 53%), compound 3c (190 mg, 72%), compound 3d (158.8 mg, 81%) and crude compound 3. The crude product was purified by silica gel chromatography (PE/EtOAc = 3:1) to yield compound 3 (65.6 mg, about 56%).

Synthesis of 0F (compound 4)

The synthesis route of 0F (compound 4) was shown in Supplementary Fig. 4. Compound 4a (0.3 mmol) in CH₂Cl₂ was stirred with 3-(dipropylamino)propane-1,2-diol (0.3 mmol), DMAP (0.3 mmol) and DIC (0.75 mmol) for 24 h at r.t. before filtration and concentration. Silica gel chromatography (PE/EtOAc = 5:1) was used to purify the residue to yield compound 4 (125.2 mg, about 70%).

Preparation of F-LNPs and F-M-LNPs

For F-LNP formulation, fluorinated lipids (0F, 4F, 6F, 8F), ALC-0315, DSPC, cholesterol and DSPE-PEG2000 were solubilized in ethanol. The mtGFP plasmid, hND4-3xFLAG-mtLuc pDNA, Cy3-pDNA or Cy5-pDNA was solubilized in 10 mM citrate buffer (pH 4.0) to form the aqueous phase. Organic phase and aqueous phase were mixed using a microfluidic chip device (LNP-B1, FluidicLab, Shanghai, China) at a volume ratio of 1:3. In contrast, the weight ratio between total lipids and pDNA was at 20:1. The resultant mixture was ultrafiltered to remove residual ethanol and concentrated using a Millipore 30 kDa ultrafiltration tube, and stored at 4 °C. For F-M-LNP formulation, DSPE-PEG2000-Mal was used to replace the part molar ratio of DSPE-PEG2000 (0%, 0.5%, 0.75%, 1% or 1.5%) in the lipid prescription. After preparation, the MTS peptide (MLSLRQSIRFFKC) was covalently attached to the surface of LNPs by thiol-maleimide coupling between DSPE-PEG-Mal and the terminal thiol group of MTS.

Physiochemical property characterizations

Plasmid DNA encapsulated in LNPs was measured by a Quant-iT PicoGreen assay. The morphology of LNPs was determined by TEM. TNS assays were used to calculate the pKa of LNP. In brief, 6 µM TNS probe and pDNA formulations (100 µM total lipids) were added into various buffer solutions with sodium chloride (150 mM), ammonium citrate (25 mM), sodium phosphate (20 mM) and ammonium acetate (20 mM), and then measured using a luminescence spectrophotometer.

Cell uptake

The N2a cells were incubated with amiloride (0.5 mM), genistein (0.21 mM), chlorpromazine (0.024 mM) or fresh DMEM for 1 h. Then, the 6F-M-LNPs/Cy5-pDNA was added into corresponding wells (1 μg Cy5-pDNA/well) apart from the control group. Cellular uptake was detected by flow cytometry.

Intracellular trafficking and subcellular localization

Lysosome escape assay: N2a cells were cultured as 4 × 10⁴ cells/well overnight and incubated with 0F-M-LNP/Cy5-pDNA or 6F-M-LNP/Cy5-pDNA (Cy5-pDNA = 2 μg/mL/well) for different time points. Then, these cells were labeled by Lysotracker Green (100 nM) and Hoechst 33342. Image was finally recorded via CLSM (LSM800, Carl Zeiss, Germany). Colocalization between endosome/lysosome and Cy5-pDNA was analyzed using the software ImageJ by Pearson’s correlation coefficient (R).

Mitochondrial colocalization assay: N2a cells or GM10742 cells were cultured as 4 × 10⁴ cells/well. Then, different cell lines were cultured with different LNPs/Cy5-pDNA (Cy5-pDNA = 2 μg/mL). Mitotracker Green (200 nM) was applied to label mitochondria in these cell lines. CLSM finally recorded the images, and colocalization between Cy5-pDNA and mitochondria was analyzed using the software ImageJ by Pearson’s correlation coefficient.

To verify whether the cargo is delivered into mitochondria, mitochondria were isolated, and proK protection assays were carried out according to previous reports³⁰. In brief, cells were cultured as 3 × 10⁶ cells/well to a 10 cm dish and treated with different groups (Cy3-pDNA = 2 μg/mL/well). After 24 h treatment, mitochondria in various treatment groups were extracted by the Minute™ Mitochondria Isolation Kit. The free mitochondria were resuspended with mitochondrial reserve solution and incubated with an equal volume of proK (30 µg/mL) at 4 °C for 30 min. Phenylmethylsulfonyl fluoride (PMSF, 100 mM) was added to inactivate proK for 10 min. Mitochondria were fixed with paraformaldehyde (PFA, 4%), permeabilized with Triton X-100 (0.2%), and incubated with primary antibodies (Mouse mAb to TOMM20, 1:400, ABclonal, #A27799; Rabbit pAb to SDHA, 1:100, Proteintech, #14865-1-AP). Mitochondria were further incubated with the appropriate secondary antibodies (Coralite488-conjugated goat Anti-mouse lgG, 1:200, Proteintech, #SA00013-1; Alexa Fluor 647 goat anti-rabbit IgG, 1:200, Beyotime Biotechnology, #A0468) at r.t. for 2 h. Samples were added to the coverslips before observation, and CLSM finally recorded the images.

To observe the delivery of cargo to specific locations within the mitochondria more clearly, the outer and inner mitochondrial membranes were labeled, respectively. In brief, N2a cells were cultured as 4 × 10⁴ cells/well overnight. Outer mitochondrial membranes in cells were labeled using Tomm20-mEmerald plasmid (1 μg). Cell lines were cultured with 6F-M-LNP/Cy5-pDNA (Cy5-pDNA = 2 μg/mL/well) for 24 h. Then, inner mitochondrial membranes were stained by Mitotracker Red (200 nM). Images were acquired on a Multi-SIM (Multimodality Structured Illumination Microscopy) imaging system (NanoInsights-Tech Co., Ltd.) equipped with a 63 × 1.40NA oil objective (ZEISS Objective Plan-Apochromat 63×/1.4 Oil M27), and a Photometrics Kinetix camera.

Mitochondrial uptake and the exploration of the mitochondrial uptake mechanism

To quantify the amount of pDNA internalized into the mitochondria, N2a cells were cultured as 3 × 10⁶ cells/well to a 10 cm dish and treated with different groups (Cy5-pDNA = 2 μg/mL/well) for 24 h. Mitochondria were isolated, and the Cy5 fluorescent intensities in different samples were measured via flow cytometry.

To study the binding affinity of F-LNPs with mitochondria, FITC-labeled LNPs (F wt%: 0.00, 5.57, 7.94 and 10.06%) were incubated with the isolated mitochondria. After 1 h, the isolated mitochondria were resuspended in mitochondrial storage buffer and incubated with an equal volume of proK (30 µg/mL) at 4 °C for 30 min. Then, the FITC fluorescence in mitochondrial matrix was detected by flow cytometry. To determine whether fluorination could mediate MMP-independent mitochondrial targeting, the cells were treated with the uncoupler, CCCP (50 µM), for 1 h before mitochondrial isolation. Isolated mitochondria were treated with the FITC-labeled 6F-LNPs and proK for assay. Mitochondrial bindings to Rh123 pretreated with or without CCCP were used as the control group.

The interaction between fluorinated molecules and mitochondrial membrane lipids was carried out by non-targeted lipidomic analysis. Firstly, undecylfluorhexylamine was grafted to the surface of carboxyl magnetic beads (300 nm) to form grafted magnetic beads of undecylfluorhexylamine (fluorinated) by EDC/NHS chemical reaction. Then, the mitochondrial lipids were extracted from isolated mitochondria using 300 µL of methanol and 1 mL methyl tert-butyl ether, followed by phase separation by adding 300 µL water. After centrifugation, the methyl tert-butyl ether layer was concentrated and dried as mitochondrial lipid extract for further analysis. Mitochondrial lipid extracts were resuspended by methanol and methyl tert-butyl ether. Then, these solutions were incubated with grafted magnetic beads of undecylfluorhexylamine (fluorinated) overnight at 4 °C. Following magnetic separation and decanting of the supernatant, the sample was obtained by elution of magnetic beads with organic solvent and analyzed by LC-MS/MS. The samples obtained from grafted beads of oleic acid (non-fluorinated) incubated with mitochondrial lysate were used as the control group.

The binding affinities between fluorinated lipids and mitochondrial lipid components were measured by microscale thermophoresis (MST) assay. Firstly, mitochondrial membrane mimicking liposomes were prepared with 80% phospholipids (POPC or TOCL) and 20% cholesterol. The liposomes were then diluted and incubated with FITC-labeled LNPs (6F-LNPs or 0F-LNPs) at a 1:1 volume ratio for 1 h. MST measurements were performed on a NanoTemper Monolith NT.115 system.

6F-LNP encapsulating gold colloid (10 nm) was prepared to trace the intracellular transport of 6F-LNPs. N2a cells were incubated with 6F-LNP/gold colloids at 37 °C for 12 h, and observed by TEM.

To investigate MTS-mediated interactions between 6F-M-LNPs and mitochondrial proteins, mitochondrial protein lysates were prepared and incubated with either 6F-M-LNPs or 6F-LNPs for 12 h. Key mitochondrial membrane proteins, including TOMM20, TOMM22, TOMM40, and TOMM70, were further validated through western blot analysis [Mouse mAb to TOMM20, 1:5000, ABclonal, #A27799; Rabbit mAb to TOMM22, 1:1000, ABclonal, #A9666; Rabbit mAb to TOMM40, 1:3000, ABclonal, #A24644; Rabbit mAb to TOMM70, 1:1000, ABclonal, #A21210].

In-situ mitochondrial gene transfection

For mtGFP transfection, various LNPs/mtGFP (mtGFP = 2 μg/mL/well) were cultured with cells in plates for 24 h. After that, a fresh complete medium was added for 48 h of culture. Then, the MFI in N2a cells was explored by flow cytometry. Additionally, mitochondrial impairment models were established by rotenone (Rot, a mitochondrial complex I inhibitor). To quantify the expression of mtGFP in mitochondria, N2a cells were treated with LNP/mtGFP for 24 h and cultured with a fresh complete medium for 48 h. Mitochondria were isolated and detected by flow cytometry.

For mtGFP transfection after siRNA treatment, N2a cells were first transfected with siRNA targeting Tomm20 or Tomm22 followed by transfection with LNPs/mtGFP. The MFI of mtGFP was determined by flow cytometry. The sequences of siRNA were: siTomm20 (sense: 5’-AUUCUCUGACUAAUGGUCG-3’, antisense: 5’- CGACCAUUAGUCAGAGAAU-3’) and siTomm22 (sense: 5’-AUCAUGAAGGAAGUGGUCC-3’, antisense: 5’-GGACCACUUCCUUCAUGAU-3’). The knockdown efficiency of siRNA targeting Tomm20 or Tomm22 was detected by quantitative polymerase chain reaction (qPCR) analysis (MA-6000, Molarray, China). The primer pairs were: Tomm20 (forward: 5’-GCCCTCTTCATCGGGTACTG-3’, reverse: 5’-ACCAAGCTGTATCTCTTCAAGGA-3’), Tomm22 (forward: 5’-CCCGAGGAATTACTCCCGAAA-3’, reverse: 5’-GGTCTCGTCTAGCTCGTCGT-3’) and Gapdh (forward: 5’-CCTCGTCCCGTAGACAAAATG-3’, reverse: 5’- TGAGGTCAATGAAGGGGTCGT-3’).

For hND4 transfection, N2a cells or GM10742 cells were incubated with various groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well) for 24 h. After that, a fresh, complete medium was added for 48 h of culture. According to the Nano-Glo Luciferase Assay System protocol, mitochondrial luciferase was assessed.

Expression of therapeutic protein in vitro

Western blot analysis: GM10742 cells were cultured as 3 × 10⁶ cells/well to a 10 cm dish and treated with different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well). Mitochondrial proteins were extracted by mitochondrial lysis buffer. The samples were incubated with primary antibodies [Mouse mAb to ND4, 1:1000, Abcam, #ab219822; Rabbit pAb to COX IV, 1:1000, ABclonal, #A6564] at 4 °C overnight. Membranes were incubated with the corresponding secondary antibody at r.t. for 2 h and then imaged with an ECL substrate by electrogenerated chemiluminescence (ECL, Tanon, China). COX IV was selected as the mitochondrial internal control.

Immunofluorescence (IF) staining analysis: N2a cells were plated 4 × 10⁴ cells/well and treated with different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well). The cells were cultured with various primary antibodies [Mouse mAb to Flag Tag, 1:500, Servicebio, #GB15938; Rabbit pAb to COX IV, 1:200, ABclonal, #A6564;] and the appropriate secondary antibodies [Alexa Fluor 647 goat anti-mouse IgG, 1:500, Beyotime Biotechnology, #A0473; Coralite488-conjugated goat Anti-rabbit lgG, 1:200, Proteintech, #SA00013-2]. DAPI was used to counterstain nuclear. CLSM finally recorded the images.

Mitochondrial membrane potential detection

N2a cells or GM10742 cells were cultured and treated with different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well). These cells were switched to a complete culture medium for 48 h before detection using the protocol of Mitochondrial Membrane Potential Test Kit (JC-1).

Mitochondrial ROS generation detection

MitoSOX Red probes were used to detect the mitochondrial ROS. GM10742 cells were cultured and treated with different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well). MitoSOX Red (5 μM) was used to stain mitochondrial ROS in cells.

Oxygen consumption rate (OCR) analysis

The OCR of cells was assessed using the Agilent Seahorse XF Cell Mito Stress Test Kits. GM10742 cells were cultured into six-well plates (5 × 10⁵ cells/well), and different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well) were added into cells. Then, the cells (8 × 10⁴ cells/well) were collected and incubated in a poly-D-lysine coated Seahorse XF Cell Culture Microplate by Seahorse XF RPMI 1640 (pH 7.4) medium containing glucose (10 mM), sodium pyruvate (1 mM) and l-glutamine (2 mM) before analysis. Final good concentrations of oligomycin, FCCP, and rotenone-antimycin A were 1.75, 5.0, 0.5, and 1.0 µM, respectively.

Therapeutic studies in vivo

Mutant ND4 mtTg LHON model male mice were assigned into three treatment groups randomly with six mice per group, including Mut group with PBS, Ide group, and 6F-M-LNP/pDNA group. C57BL/6J male mice were set as WT group. Then, the mice were anesthetized and injected with PBS or 6F-M-LNP/pDNA into both vitreous chambers once every 2 weeks for 1 month. The intravitreal injection volume was 3 μL each eye, and the administration dose of pDNA was 1 μg/eye. The Ide group was administered once daily at an oral dose of 60 mg/kg. After treatment, optomotor tests and electroretinography (ERG) measurements were performed. The mice were sacrificed to collect the eyeballs of each group for western blot analysis, retinal ATP content determination, and H&E staining of the eyeballs.

Safety assessment

C57BL/6J male mice were randomly assigned into four groups with five mice per group. Both eyes of mice received intravitreal injections of various doses of 6F-M-LNP/pDNA (6F-M-L: 1 μg pDNA/eye; 6F-M-M: 2 μg pDNA/eye; 6F-M-H: 4 μg pDNA/eye) once a week for 3 weeks. Blood was collected after another three weeks of observation, and the serum samples were prepared for blood biochemical analysis. Major organs and eyeballs were collected for H&E staining. The organ index was calculated by the relative organ weight (the weight ratio of organ/body).

Western blot analysis in vivo

To detect the amount of total ND4 protein in vivo, eyes were homogenized and standard protocols extracted the total protein. The anti-ND4 antibody [Rabbit pAb to ND4, 1:1000, ABclonal, #A17970] was chosen as the primary antibody, and the β-actin [Rabbit mAb to β-actin,1:1000, Beyotime Biotechnology; #AF5003] was used as the internal control.

Adenosine triphosphate (ATP) detection

N2a cells or GM10742 cells were cultured as 3 × 10⁵ cells/well and incubated with different groups (hND4-3xFLAG-mtLuc = 2 μg/mL/well) for specified durations. ATP contents were assessed by ATP Assay Kits.

To measure ATP content in retina, the mouse eyes were harvested rapidly and the retina were separated. 100 μL of lysis buffer was used for homogenization, and these samples were determined based on ATP Assay Kits.

The optomotor test

The mice were positioned on a central platform with a 10 cm diameter, encircled by a motorized roller capable of rotating either clockwise or counterclockwise, featuring vertical black and white stripes with a thickness of 1.0 cm. Clockwise or counterclockwise rotation of the roller was used to test the visual acuity of the various eyes. In brief, following a 5-min adaptation period to the experimental environment, the drum was rotated in a clockwise direction for 2 min, then counterclockwise for another 2 min, with a 30-s pause between rotations. A movie camera was recorded the movements, and then the number of head movements was recorded only when the head of the mouse moved in the direction of the roller, but the body did not.

The measurement of electroretinography (ERG)

Mice were placed in a dark room overnight to adapt to the dark. Ring electrodes were put on the corneas, and the needle electrodes were inserted into the skin and tail before being measured. Dark adaptation 10.0 ERG detections (flash intensity = 10 cd·s/m²) were performed by Espion Visual Electrophysiology System (Diagnosys LLC, USA).

MD simulations

MD simulations were used to study the interactions between 6F-LNPs and inner mitochondrial membranes. The 6F-LNP and inner mitochondrial membranes were respectively constructed by PACKMOL and PACKMOL-Memgen as initial configurations for MD simulations. The 6F-LNP was constructed with the ratio of ALC-0315, 6F lipid, DSPC, cholesterol, and DSPE-PEG2000 as 25:25:10:38.5:1.5, while the inner mitochondrial membrane was constructed with the ratio of TOCL/POPC as 20:80. Gromacs2019.6 was chosen as the dynamics simulation software. In the simulations, 6F-LNP was initially placed near the outside surface of inner mitochondrial membranes. The TIP3P water model was used and 6F-LNP was described with the GaFF2 force field, while the inner mitochondrial membrane was described with the Lipid 21 force field. The particle-mesh Ewald method was used for evaluation of long-range Coulomb interactions. The time step was set to 2 fs. The simulation systems were energy-minimized using the steepest-descent algorithm for 50000 steps. All systems were equilibrated for 30 ps in the NVT and NPT ensemble (pressure p = 1 bar, temperature T = 300 K), followed by a 100 ns MD simulation.

Statistics and reproducibility

Statistical analysis was performed using SPSS Version 25. Quantitative data in these experiments were presented as mean ± standard deviation (SD) from sample numbers (n). Statistical tests utilized for each experiment and the reproducibility of experiments were specified in the legends of figures. Independent two-tailed unpaired Student’s t test was analyzed comparisons between the two groups. Comparisons between more than two groups were analyzed by one-way analysis of variance (ANOVA) with a two-tailed Tukey’s multiple comparisons test when the data satisfied the homogeneity of variance or two-tailed Games-Howell post-hoc test when not met the homogeneity of variance.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

1. Gorman, G. S. et al. Prevalence of nuclear and mitochondrial DNA mutations related to adult mitochondrial disease. Ann. Neurol. 77, 753–759 (2015).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

2. Gorman, G. S. et al. Mitochondrial diseases. Nat. Rev. Dis. Prim. 2, 16080 (2016).

   Article  PubMed  Google Scholar 

3. Ng, Y. S. et al. Mitochondrial disease in adults: recent advances and future promise. Lancet Neurol. 20, 573–584 (2021).

   Article  CAS  PubMed  Google Scholar 

4. Pitceathly, R. D. S., Keshavan, N., Rahman, J. & Rahman, S. Moving towards clinical trials for mitochondrial diseases. J. Inherit. Metab. Dis. 44, 22–41 (2021).

   Article  PubMed  Google Scholar 

5. Bae, Y. et al. Functional nanosome for enhanced mitochondria-targeted gene delivery and expression. Mitochondrion 37, 27–40 (2017).

   Article  CAS  PubMed  Google Scholar 

6. Ishikawa, T., Somiya, K., Munechika, R., Harashima, H. & Yamada, Y. Mitochondrial transgene expression via an artificial mitochondrial DNA vector in cells from a patient with a mitochondrial disease. J. Control. Release 274, 109–117 (2018).

   Article  CAS  PubMed  Google Scholar 

7. Yu, H. et al. Gene delivery to mitochondria by targeting modified adenoassociated virus suppresses Leber’s hereditary optic neuropathy in a mouse model. Proc. Natl. Acad. Sci. USA 109, E1238–E1247 (2012).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

8. Chen, W. H., Luo, G. F. & Zhang, X. Z. Recent advances in subcellular targeted cancer therapy based on functional materials. Adv. Mater. 31, e1802725 (2019).

   Article  PubMed  Google Scholar 

9. Guo, X. et al. Mito-bomb: targeting mitochondria for cancer therapy. Adv. Mater. 33, e2007778 (2021).

   Article  ADS  PubMed  Google Scholar 

10. Yamada, Y. et al. Power of mitochondrial drug delivery systems to produce innovative nanomedicines. Adv. Drug Deliv. Rev. 154-155, 187–209 (2020).

    Article  CAS  PubMed  Google Scholar 

11. Zielonka, J. et al. Mitochondria-targeted triphenylphosphonium-based compounds: syntheses, mechanisms of action, and therapeutic and diagnostic applications. Chem. Rev. 117, 10043–10120 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

12. Wang, Y., Yang, J. S. & Jiang, H. L. Mitochondrial endogenous substance transport-inspired nanomaterials for mitochondria-targeted gene delivery. Adv. Drug Deliv. Rev. 211, 115355 (2024).

    Article  CAS  PubMed  Google Scholar 

13. Yoshinaga, N. et al. Design of an artificial peptide inspired by transmembrane mitochondrial protein for escorting exogenous DNA into the mitochondria to restore their functions by simultaneous multiple gene expression. Adv. Funct. Mater. 34, 2306070 (2024).

    Article  CAS  Google Scholar 

14. Yu, H. et al. Consequences of zygote injection and germline transfer of mutant human mitochondrial DNA in mice. Proc. Natl. Acad. Sci. USA 112, E5689–E5698 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

15. Pfanner, N., Warscheid, B. & Wiedemann, N. Mitochondrial proteins: from biogenesis to functional networks. Nat. Rev. Mol. Cell Biol. 20, 267–284 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Schmidt, O., Pfanner, N. & Meisinger, C. Mitochondrial protein import: from proteomics to functional mechanisms. Nat. Rev. Mol. Cell Biol. 11, 655–667 (2010).

    Article  CAS  PubMed  Google Scholar 

17. Yamada, Y. et al. Validation of a mitochondrial RNA therapeutic strategy using fibroblasts from a Leigh syndrome patient with a mutation in the mitochondrial ND3 gene. Sci. Rep. 10, 7511 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

18. Liew, S. S. et al. Smart design of nanomaterials for mitochondria-targeted nanotherapeutics. Angew. Chem. Int. Ed. Engl. 60, 2232–2256 (2021).

    Article  CAS  PubMed  Google Scholar 

19. Li, J. et al. Fluoroalkane modified cationic polymers for personalized mRNA cancer vaccines. Chem. Eng. J. 456, 140930 (2023).

    Article  ADS  CAS  PubMed  Google Scholar 

20. Xu, J. et al. A general strategy towards personalized nanovaccines based on fluoropolymers for post-surgical cancer immunotherapy. Nat. Nanotechnol. 15, 1043–1052 (2020).

    Article  ADS  CAS  PubMed  Google Scholar 

21. Wang, M., Liu, H., Li, L. & Cheng, Y. Y. A fluorinated dendrimer achieves excellent gene transfection efficacy at extremely low nitrogen to phosphorus ratios. Nat. Commun. 5, 3053 (2014).

    Article  ADS  PubMed  Google Scholar 

22. Wang, Y., Hu, L. F. & Jiang, H. L. Pathologically responsive mitochondrial gene therapy in an allotopic expression-independent manner cures Leber’s hereditary optic neuropathy. Adv. Mater. 33, e2103307 (2021).

    Article  ADS  PubMed  Google Scholar 

23. Zhang, H., Li, H. J. & Gu, Z. Fluorinated lipid nanoparticles for enhancing mRNA delivery efficiency. ACS Nano 18, 7825–7836 (2024).

    Article  CAS  PubMed  Google Scholar 

24. Chen, M., Bao, J. & Liu, S. Ultrasound-enhanced spleen-targeted mRNA delivery via fluorinated PEGylated lipid nanoparticles for immunotherapy. Angew. Chem. Int. Ed. Engl. 64, e202500878 (2025).

    Article  CAS  PubMed  Google Scholar 

25. Zhang, C. et al. Biological utility of fluorinated compounds: from materials design to molecular imaging, therapeutics and environmental remediation. Chem. Rev. 122, 167–208 (2022).

    Article  CAS  PubMed  Google Scholar 

26. Gong, J. H., Wang, Y. & Jiang, H. L. Biocompatible fluorinated poly(beta-amino ester)s for safe and efficient gene therapy. Int. J. Pharm. 535, 180–193 (2018).

    Article  CAS  PubMed  Google Scholar 

27. Han, H., Xing, J., Chen, W., Jia, J. & Li, Q. Fluorinated polyamidoamine dendrimer-mediated miR-23b delivery for the treatment of experimental rheumatoid arthritis in rats. Nat. Commun. 14, 944 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

28. Santaella, C., Vierling, P. & Riess, J. G. Perfluoroalklylated phospholipids as surfactants and co-surfactants for injectable fluorocarbon emulsions. Biomater. Artif. Cells Immobil. Biotechnol. 20, 835–837 (1992).

    CAS  Google Scholar 

29. Zhang, Z. et al. The fluorination effect of fluoroamphiphiles in cytosolic protein delivery. Nat. Commun. 9, 1377 (2018).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

30. Bi, Y. Y., Chen, Q., Yang, M. Y., Xing, L. & Jiang, H. L. Nanoparticles targeting mutant p53 overcome chemoresistance and tumor recurrence in non-small cell lung cancer. Nat. Commun. 15, 2759 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

31. Yamada, Y., Ishikawa, T. & Harashima, H. Validation of the use of an artificial mitochondrial reporter DNA vector containing a cytomegalovirus promoter for mitochondrial transgene expression. Biomaterials 136, 56–66 (2017).

    Article  CAS  PubMed  Google Scholar 

32. Yang, S. et al. Long-term outcomes of gene therapy for the treatment of Leber’s hereditary optic neuropathy. EBioMedicine 10, 258–268 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

33. Y, Z. et al. Prognostic factors for visual acuity in patients with Leber’s hereditary optic neuropathy after rAAV2-ND4 gene therapy. Clin. Exp. Ophthalmol. 47, 774–778 (2019).

    Article  Google Scholar 

34. Johnson, L. V., Walsh, M. L. & Chen, L. B. Localization of mitochondria in living cells with rhodamine 123. Proc. Natl. Acad. Sci. Usa. 77, 990–994 (1980).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

35. Pegoraro, C. et al. Polyproline-polyornithine diblock copolymers with inherent mitochondria tropism. Adv. Mater. 37, e2411595 (2025).

    Article  PubMed  Google Scholar 

36. Liu, Y., Zhang, J., Tu, Y. & Zhu, L. Potential-independent intracellular drug delivery and mitochondrial targeting. ACS Nano 16, 1409–1420 (2022).

    Article  CAS  PubMed  Google Scholar 

37. Oemer, G. et al. Molecular structural diversity of mitochondrial cardiolipins. Proc. Natl. Acad. Sci. USA 115, 4158–4163 (2018).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

38. Zhou, Y. X. et al. Delivery of low-density lipoprotein from endocytic carriers to mitochondria supports steroidogenesis. Nat. Cell Biol. 25, 937–949 (2023).

    Article  CAS  PubMed  Google Scholar 

39. Thakur, S. S., Shenoy, S. K., Suk, J. S., Hanes, J. S. & Rupenthal, I. D. Validation of hyaluronic acid-agar-based hydrogels as vitreous humor mimetics for in vitro drug and particle migration evaluations. Eur. J. Pharm. Biopharm. 148, 118–125 (2020).

    Article  CAS  PubMed  Google Scholar 

40. Qin, X. et al. Wearable electrodriven switch actively delivers macromolecular drugs to fundus in non-invasive and controllable manners. Nat. Commun. 16, 33 (2025).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

41. Jung, H. N., Lee, S. Y., Lee, S., Youn, H. & Im, H. J. Lipid nanoparticles for delivery of RNA therapeutics: current status and the role of in vivo imaging. Theranostics 12, 7509–7531 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Zhang, J. et al. Mitochondrial-targeted delivery of polyphenol-mediated antioxidases complexes against pyroptosis and inflammatory diseases. Adv. Mater. 35, e2208571 (2023).

    Article  PubMed  Google Scholar 

43. Jiang, L. et al. Mitochondrion-specific dendritic lipopeptide liposomes for targeted sub-cellular delivery. Nat. Commun. 12, 2390 (2021).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

44. Zhou, H. et al. Melatonin protects against rotenone-induced cell injury via inhibition of Omi and Bax-mediated autophagy in HeLa cells. J. Pineal Res. 52, 120–127 (2012).

    Article  CAS  PubMed  Google Scholar 

45. Aoyama, Y. et al. Involvement of endoplasmic reticulum stress in rotenone-induced Leber hereditary optic neuropathy model and the discovery of new therapeutic agents. J. Pharmacol. Sci. 147, 200–207 (2021).

    Article  CAS  PubMed  Google Scholar 

46. Qi, L. Y., Wang, Y. & Jiang, H. L. Enhanced nuclear gene delivery via integrating and streamlining intracellular pathway. J. Control. Release 341, 511–523 (2022).

    Article  CAS  PubMed  Google Scholar 

47. Islam, M. A. et al. Essential cues of engineered polymeric materials regulating gene transfer pathways. Prog. Mater. Sci. 128, 100961 (2022).

    Article  Google Scholar 

48. Yuan, X. & You, J. Virus-like nonvirus cationic liposome for efficient gene delivery via endoplasmic reticulum pathway. ACS Cent. Sci. 6, 174–188 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

49. Lv, J., Wang, H., Rong, G. & Cheng, Y. Y. Fluorination promotes the cytosolic delivery of genes, proteins, and peptides. Acc. Chem. Res. 55, 722–733 (2022).

    Article  CAS  PubMed  Google Scholar 

50. Guo, Y. et al. Visualizing intracellular organelle and cytoskeletal interactions at nanoscale resolution on millisecond timescales. Cell 175, 1430–1442 (2018).

    Article  CAS  PubMed  Google Scholar 

51. Lin, C. S. et al. Mouse mtDNA mutant model of Leber hereditary optic neuropathy. Proc. Natl. Acad. Sci. USA 109, 20065–20070 (2012).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

52. Lyseng-Williamson, K. A. Idebenone: a review in Leber’s hereditary optic neuropathy. Drugs 76, 805–813 (2016).

    Article  CAS  PubMed  Google Scholar 

53. Wallace, D. C. et al. Mitochondrial DNA mutation associated with Leber’s hereditary optic neuropathy. Science 242, 1427–1430 (1988).

    Article  ADS  CAS  PubMed  Google Scholar 

54. Chin, R. M., Panavas, T., Brown, J. M. & Johnson, K. K. Patient-derived lymphoblastoid cell lines harboring mitochondrial DNA mutations as tool for small molecule drug discovery. BMC Res. Notes 11, 205 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

55. Yi, Z. et al. Strand-selective base editing of human mitochondrial DNA using mitoBEs. Nat. Biotechnol. 42, 498–509 (2024).

    Article  CAS  PubMed  Google Scholar 

56. Zhuo, Y., Luo, H. & Zhang, K. Leber hereditary optic neuropathy and oxidative stress. Proc. Natl. Acad. Sci. USA 109, 19882–19883 (2012).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

57. Sanes, J. R. & Masland, R. H. The types of retinal ganglion cells: current status and implications for neuronal classification. Annu. Rev. Neurosci. 38, 221–246 (2015).

    Article  CAS  PubMed  Google Scholar 

58. Zollo, O., Tiranti, V. & Sondheimer, N. Transcriptional requirements of the distal heavy-strand promoter of mtDNA. Proc. Natl. Acad. Sci. USA 109, 6508–6512 (2012).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Author notes

1. These authors contributed equally: Yi Wang, Min Zhao, Hai-Xin Xie.

Authors and Affiliations

1. State Key Laboratory of Natural Medicines, Department of Pharmaceutics, China Pharmaceutical University, Nanjing, China

   Yi Wang, Min Zhao, Hai-Xin Xie, Hao-Yuan Yu, Jing-Song Yang, Lu-Xin Qie, Na-Hui Liu, Jia-Qi Chen, Zi-Juan Yi, Tian-Jiao Zhou, Lei Xing & Hu-Lin Jiang

2. Jilin Provincial Key Laboratory of Stress and Cardiovascular Disease, Department of Cardiology and Hypertension, Yanbian University Hospital, Yanji, China

   Xian Wu Cheng & Hu-Lin Jiang

3. Joint International Research Laboratory of Target Discovery and New Drug Innovation, MOE, China Pharmaceutical University, Nanjing, China

   Hu-Lin Jiang

4. Department of Precision Medicine, School of Medicine, Sungkyunkwan University, Suwon, Korea

   Hu-Lin Jiang

Contributions

H.L.J., Y.W., M.Z., H.X.X., T.J.Z., L.X. conceived the project, and designed all experiments; Y.W., M.Z., H.X.X., H.Y.Y., J.S.Y., L.X.Q., N.H.L., J.Q.C., Z.J.Y., T.J.Z. conducted the experiments and analyzed the data, H.L.J., X.W.C., Y.W. provided financial support for this work; and all authors edited the manuscript.

Corresponding author

Correspondence to Hu-Lin Jiang.

Competing interests

The authors declare no competing interests.

Peer review information

Nature Communications thanks anonymous reviewer(s) for their contribution to the peer review of this work. [A peer review file is available].

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions