Reversible one-step acylation facilitates mitochondrial delivery of functional RNA

Abstract

Mitochondria-targeted RNA therapeutics hold promise for treating mitochondrial disorders and cancer, yet effective mitochondrial gene modulation remains challenging due to the strong negative charge and intrinsic instability of RNA, which hinder transport across the mitochondrial double membrane. Herein, we demonstrate, for the first time, mitochondrial RNA delivery enabled by reversible 2′-hydroxyl (2′-OH) acylation chemistry. Installation of a triphenylphosphonium (TPP)-bearing acyl group at the RNA 2′-OH in a single step creates a plug-and-play platform that simultaneously enhances RNA stability and directs selective mitochondrial accumulation, while allowing spontaneous recovery of native RNA structure and activity. In cell-based assays, this strategy enables efficient and selective silencing of mitochondria-encoded genes with minimal off-target effects. Importantly, in vivo delivery of siND1 selectively suppresses MTND1 expression without affecting MTCO1 and significantly inhibits tumor growth in a xenograft model. Conclusively, this work establishes reversible RNA acylation as a simple and versatile chemical framework for mitochondrial RNA delivery and therapy.

---

Introduction

The mitochondrial genome (mtDNA) encodes essential components of the oxidative phosphorylation machinery and sustains autonomous transcription and translation within mitochondria. Consequently, mitochondrial gene expression is central to cellular energy metabolism and redox homeostasis.^1,2 Pathogenic mtDNA mutations underlie numerous severe diseases, and mitochondrial dysfunction is increasingly recognized as a hallmark of cancer.^3–6 These observations underscore a pressing need for strategies that enable precise and selective modulation of mitochondrial gene function.

Although RNA-based technologies such as RNAi, CRISPR, and mRNA therapeutics have transformed nuclear gene regulation,^7–9 progress in mitochondrial gene engineering has been markedly slower.^10–14 This limitation arises primarily from the impermeable double-membrane architecture of mitochondria and the intrinsic instability and polyanionic nature of RNA, which together severely restrict mitochondrial entry and persistence of exogenous RNAs. To address this challenge, mitochondrial delivery has been explored using ligands such as mitochondrial targeting sequence (MTS) peptides and lipophilic cations, most notably triphenylphosphonium (TPP).^15,16 These strategies rely on permanent attachment of mitochondria-targeting motifs to RNA,^17–21 which, despite enabling broadly applicable targeting, introduces irreversible structural perturbations that can interfere with native RNA folding, target recognition, and biological activity.^22,23 As a result, approaches that enable efficient mitochondrial RNA delivery while fully restoring the RNA to its unmodified, native state remain exceedingly limited.

In parallel, RNA instability remains a major challenge. While material-based carriers and permanent chemical modifications can improve RNA stability,^24–26 they often compromise biological compatibility and are poorly suited to mitochondrial systems with distinct transcriptional and translational machinery. Although vector-mediated delivery of unmodified RNA offers advantages,^27–29 the accumulation of delivery vehicles could cause a burden on mitochondrial homeostasis. Therefore, simple reversible RNA modification has emerged as an attractive chemical strategy.^30–35 In particular, pioneered by the Kool group, transient 2′-hydroxyl acylation (“RNA cloaking”) provides a simple and general means to enhance RNA stability while preserving the capacity for spontaneous recovery of the native structure and function. Subsequent work by Zhu and coworkers introduced stimuli-responsive 2′-OH RNA “uncloaking” strategies that enable controlled intracellular release. Despite these advances, RNA cloaking has thus far been limited to global stabilization and intracellular activation, and it has not yet been demonstrated as a strategy for targeted subcellular RNA delivery, particularly to mitochondria.

Herein, we report the first example of mitochondrial RNA delivery enabled by reversible RNA cloaking. We describe a modular chemical strategy that integrates transient 2′-OH acylation with a mitochondria-directing TPP motif (Fig. 1). This single-step modification simultaneously stabilizes RNA and enables its mitochondrial targeting while preserving its ability to regain the native structure and function upon intracellular traceless release. This plug-and-play platform eliminates the need for permanent peptide or linker conjugation and establishes reversible RNA acylation as a general chemical framework for mitochondrial RNA delivery and gene regulation.

Fig. 1 Scheme illustrating the key steps of mitochondria-targeted RNA delivery based on a TPP-conjugated acylimidazole reagent: (1) chemical structures of the TPP–acylimidazole derivatives; (2) activation (“uncloaking”) and traceless release of the targeted RNA; (3) mitochondria-specific delivery and subsequent gene-silencing function.

Results and discussion

Design, RNA modification, and reversible stabilization of IM@RNA

Four TPP-conjugated acylimidazole reagents (Fig. 1) were rationally designed and synthesized to enable reversible covalent modification of RNA at the 2′-OH position. These reagents differ primarily in the length of the linker connecting the TPP moiety to the acylimidazole reactive center: (i) TPIM, featuring a long-chain aliphatic linker; (ii) TEIM, incorporating a flexible diethylene glycol linker; (iii) TSIM, containing a short PEG linker with a redox-responsive disulfide bond; and (iv) T2IM, combining a longer PEG linker with a disulfide unit. The detailed synthetic routes and structural characterization of the corresponding acid precursors are provided in the SI. All acylimidazole reagents were freshly generated in situ from their corresponding carboxylic acids via 1,1′-carbonyldiimidazole (CDI) activation and subsequently reacted with RNA under mild conditions to afford IM-modified RNA (IM@RNA).³³ This one-step cloaking process enables efficient esterification of accessible 2′-OH groups along the RNA backbone. Importantly, the reversibility of the modification is inherently engineered to linker chemistry: TPIM and TEIM rely on ester bond hydrolysis triggered by nucleophilic or mildly alkaline conditions, whereas TSIM and T2IM incorporate disulfide bonds that could undergo reductive cleavage in thiol-rich environments, such as those containing glutathione (GSH), ultimately leading to a thermodynamically stable five-membered ring formation and subsequently to traceless release of native RNA (Fig. 1).³⁴

We selected siRNA (∼20 bp) as a model substrate to optimize the cloaking conditions. Based on earlier reports, five acylimidazole reagents—including a previously described but non-targeted control reagent, CDI-activated N,N-dimethylglycine (DGIM, Fig. 2A)—were incubated with siRNA in a DMSO/water mixture (1 : 4, v/v) at room temperature for 4 h with gentle agitation.³³ The amounts of all these reagents were in large excess relative to that of the RNA. The modified RNAs were purified by ethanol precipitation to afford the acylated products, termed IM@RNA. Concentration-dependent acylation was first evaluated by agarose gel electrophoretic mobility shift assays. As shown in Fig. S1, all reagents induced clear mobility shifts of siRNA, confirming successful acylation. Notably, compared with DGIM, TPP-conjugated ones produced more pronounced broadened band shifts at lower concentrations, suggesting that the positively charged TPP moiety enhances the acylation efficiency, likely through electrostatic interactions with RNA. Therefore, the TPP ones were applied at a concentration of 100 mM, approximately one-tenth that of DGIM, for the generation of IM@RNA (Fig. 2B and S2). The MALDI-TOF MS data further demonstrate that different acylation reagents resulted in varying degrees of RNA modification (Fig. 2C and S3). Specifically, TPIM displayed the highest level of acylation, with an average modification of approximately 8 added acyl groups per strand (for strands >18 nt). In contrast, TSIM and T2IM with disulfide bonds showed a noticeably lower degree of acylation. Furthermore, the data indicate that the specific length and conformation of the RNA (single-stranded versus double-stranded) also have a slight influence on the overall acylation levels.

Fig. 2 (A) DGIM formation. (B) Agarose gel electrophoresis of siRNA (500 ng) after incubation with different acylation reagents in ultrapure water at 37 °C for 4 h to form IM@RNA. The resulting complexes were purified using ice-cold ethanol. (C) MALDI-TOF mass spectrum of cloaked siCOX1 with T2IM reagent. Peaks are assigned according to the number of TPP-acylated adducts, indicated in red. (D) Uncloaking assay of IM@RNA under basic or reductive conditions. Samples were incubated for 24 h in 200 mM Tris buffer (pH 8.8) or with 20 mM GSH, followed by agarose gel electrophoresis. (E) Evaluation of RNase A resistance for FAM-labelled acylated siRNA. PAGE analysis (FITC channel) of unmodified FAM-labelled siRNA and TPIM@siNC^FAM in the absence (−) or presence (+) of RNase A treatment. Lane marked (*) corresponds to free siRNA.

For IM@RNA to function as a practical delivery and protection platform, the covalent modification must be reversible under intracellularly relevant conditions. Because IM@RNA is linked through ester bonds, deacylation can be triggered either by hydrolysis or by reductive cleavage of the linker, depending on reagent design. To evaluate reversibility, IM@RNA samples were incubated under mildly alkaline conditions (200 mM Tris buffer, pH 8.8) or in the presence of GSH (20 mM), followed by purification and agarose gel electrophoretic analysis. As shown in Fig. 2D and S4, all the complexes underwent efficient deprotection in an alkaline or a reductive environment. These results confirmed that both hydrolysis- and redox-triggered uncloaking pathways could enable efficient recovery of unmodified RNA, supporting the intracellular applicability of this platform.

We next investigated whether esterification of the 2′-OH group could confer protection against enzymatic degradation. Heat-deannealed TPIM@siNC^FAM (fluorescent FAM moiety modified siNC) were incubated with RNase A (100 µg mL⁻¹) for 30 min and analyzed by agarose gel electrophoresis (Fig. 2E). The PAGE results clearly demonstrate the heterogeneous protective effect of our modification against RNase A. Specifically, upon treatment with RNase A, the less-modified or uncloaked RNA populations are degraded, as shown by the attenuation of the lower bands. However, the populations with higher levels of modification (corresponding to the distinct upper bands) remain well-protected and retained. Therefore, this combination of reversibility and partial stability establishes IM@RNA as a robust chemical platform for subsequent mitochondrial delivery and functional studies.

Cellular uptake and mitochondrial colocalization of IM@RNA

We next evaluated the cellular uptake and intracellular localization of IM-modified RNA (IM@RNA). Based on preliminary results indicating that effective cellular delivery occurred after 4 h, all subsequent experiments were performed using extended incubation times (Fig. S5). Accordingly, HeLa cells were incubated with IM@siCOX1^TMR (100 nM, defined based on the siRNA concentration) for 48 h, followed by mitochondrial staining with MitoTracker Green and confocal microscopy analysis. As shown in Fig. 3A, the control DGIM@siCOX1^TMR exhibited minimal overlap with mitochondrial signals, indicating inefficient mitochondrial localization. In contrast, as indicated by Pearson's correlation coefficients exceeding 0.75, TPP-containing IM@RNA constructs demonstrated pronounced colocalization with mitochondria (Fig. 3A and B), confirming their effective transmembrane delivery into mitochondria. Similarly effective delivery was observed when siND1^TMR was administered to HepG2 cells, further confirming the generality of the mitochondrial targeting capability (Fig. S6).

Fig. 3 Cellular uptake and subcellular localization of TPP-acylated siCOX1 (TAMAR dye-labelled) complexes (IM@siCOX1^TMR). (A) Confocal microscopy images showing cellular uptake and mitochondrial colocalization of IM@siCOX1^TMR (red channel). Cells were treated with 100 nM IM@siCOX1^TMR for 48 h, followed by staining with MitoTracker Green (mitochondria, green channel). The red boxes indicate regions enlarged in the panels to the right. (B) Quantification of mitochondrial colocalization. The Pearson's correlation coefficient (PCC) between TMR and MitoTracker Green signals was calculated from confocal images (n = 3 independent experiments). (C) Flow cytometry analysis of cellular uptake. Cells treated with IM@siCOX1^TMR for 24 h were analyzed using the Cy3 channel. Histograms display the relative fluorescence intensity (RFU) of 10 000 cells per sample. Statistical significance was determined using one-way ANOVA with GraphPad Prism (***P < 0.001). (D) Co-localization analysis of IM@siCOX1^TMR complexes (yellow channel) with mitochondria and lysosomes. Right plot: quantitative fluorescence intensity profile analysis. ImageJ was used to extract and plot the fluorescence intensity along the line indicated in each channel.

Furthermore, quantitative analysis corroborated these observations: flow cytometry revealed significantly higher fluorescence intensities for all TPP-linked IM@RNA samples compared with the DGIM control after 48 h of incubation (Fig. 3C). These results confirm that incorporation of the TPP motif substantially enhances both cellular uptake and mitochondrial accumulation, consistent with its lipophilic and cationic properties that favor mitochondrial membrane permeation. Notably, linker architecture also influenced delivery efficiency. IM@RNA constructs bearing PEG-based linkers showed higher uptake efficiencies than TPIM, likely due to increased conformational flexibility and improved colloidal behavior, which may reduce nonspecific aggregation and lysosomal sequestration. To directly assess organelle specificity, cells were co-incubated with TEIM, T2IM, or the negative control DGIM and subjected to dual staining of lysosomes and mitochondria (Fig. 3D and S7). Compared with DGIM, both TEIM and T2IM produced stronger intracellular signals and preferential co-localization with mitochondria, while showing minimal overlap with LysoTracker, indicating that the delivered siRNA largely avoided lysosomal capture. Collectively, these results confirmed that TPP-mediated reversible RNA acylation enables efficient cellular uptake and sustained mitochondrial localization and provides insight that rational linker engineering is an effective strategy to improve delivery efficiency and organelle specificity.

Mitochondrial gene silencing efficiency of IM@RNA

Building on the validated mitochondrial delivery capability of the bifunctional IM@RNA platform, we next evaluated its ability to mediate functional gene silencing of mitochondria-encoded targets.^21,28 siRNAs against MTCO1 and MTND1, which encode key subunits of mitochondrial respiratory chain Complex IV and Complex I, respectively, were formulated into IM@RNA constructs. Cells were treated with IM@RNA formulations at a final concentration of 100 nM (based on the amount of siRNA, with each strand present at 100 nM), followed by western blot (WB) analysis to assess target protein expression. Vinculin protein was used as the reference loading control across all samples. Relative to the non-targeted control (DGIM@RNA), all TPP-conjugated IM@RNA formulations induced pronounced downregulation of their respective target proteins (Fig. 4). Among them, TEIM@RNA and T2IM@RNA consistently exhibited superior silencing efficacy. For MTCO1, TEIM@siCOX1 achieved a knockdown efficiency of 60% ± 10%, while T2IM@siCOX1 resulted in more robust suppression, reaching 78% ± 15% (Fig. 4A). Similarly, efficient silencing of MTND1 was observed across all TPP-linked constructs, with knockdown efficiencies exceeding 70%, and T2IM@siND1 displaying the strongest effect (Fig. 4B). These trends closely mirror the mitochondrial targeting efficiencies observed in the imaging and flow cytometry analysis, underscoring the importance of effective mitochondrial localization for functional gene silencing.

Fig. 4 Targeted silencing of mitochondrial genes by TPP-acylated siRNA complexes. (A) and (B) suppression of MTCO1 and MTND1 by different acylated complexes, respectively. Representative western blot and corresponding quantification of MTCO1\MTND1 protein levels in HeLa cells treated with 100 nM of the indicated IM@siND1 complexes for 48 h. (C) and (D) Time course of MTCO1 and MTND1 silencing, respectively. Representative western blot and quantification of MTCO1/MTND1 protein levels in HeLa cells treated with 100 nM T2IM@siCOX1 for the indicated durations. Blank: without any treatment. (E) Dose-dependent silencing of MTCO1. Representative western blot of MTCO1 protein levels in HeLa cells treated for 48 h with TEIM@siCOX1 or T2IM@siCOX1 at the indicated concentrations. (F) Dose-dependent silencing of MTND1. Representative western blot of MTND1 and MTCO1 protein levels in HeLa cells treated for 48 h with T2IM@siCOX1 at the indicated concentrations. Vinculin served as a loading control; treatment with siRNA only (without any transfection reagent) served as a negative control. (G) The formation of TPAM@RNA with a carbamate link. (Left) Suppression of MTND1 using T2IM, TPAM or Lipo2000 reagents. (Right) (H) Suppression of MIF protein using T2IM, TSIM or Lipo2000 reagents.

We next investigated the temporal and concentration dependence of mitochondrial gene knockdown using TEIM@RNA and T2IM@RNA. Progressive reduction of target protein levels was observed with increasing incubation time, with pronounced knockdown occurring between 6 and 48 h (Fig. 4C, D and S8). Dose–response studies were subsequently conducted by treating cells with IM@RNA at concentrations of 100, 50, 25, or 10 nM for 48 h. Both TEIM@RNA and T2IM@RNA exhibited clear concentration-dependent silencing behavior, with maximal knockdown at 100 nM and markedly reduced the efficacy below 25 nM (Fig. 4E). The specificity of mitochondrial gene silencing was further evaluated using T2IM@siND1. WB analysis revealed that knockdown was restricted to the intended target, with no detectable changes in the expression of vinculin or MTCO1 across all tested concentrations (Fig. 4F).

Moreover, to explore whether acylation can suppress siRNA activity and whether activity can be restored, we constructed control conjugates using triphenylphosphonium amine (TPAM, Fig. 4G), which forms relatively stable carbonate linkages with RNA, as reported by Kool and co-workers,³⁶ which is expected to be comparably irreversible under intracellular conditions. As shown in Fig. 4G, we compared three conditions: (i) siRNA modified with non-cleavable TPP-NH2 (TPAM@siND1), (ii) siRNA delivered via Lipofectamine 2000 (Lipo2000@siND1), and (iii) our cleavable TPP-modified siRNA (T2IM@siND1). We observed that only the cleavable conjugate (T2IM@siND1) reduced MTND1 expression, whereas the non-cleavable TPP-modified siRNA (TPAM@siND1) showed minimal silencing activity. This suggests that permanent acylation masks siRNA function and that restoration of activity requires cleavage of the linkage.

Furthermore, to evaluate the role of mitochondrial targeting, we performed parallel experiments using siRNA against a cytosolic protein, MIF. Mitochondria-targeted constructs (T2IM@siMIF and TSIM@siMIF) showed significantly reduced knockdown efficiency compared to Lipo2000-delivered siMIF (Lipo2000@siMIF), which efficiently decreased cytosolic MIF levels (Fig. 4H). This indicates that mitochondrial targeting restricts siRNA availability in the cytosol and is therefore critical for achieving selective knockdown of mitochondrial genes. Taken together, these results confirm that this platform enables time-/dose-dependent, highly specific silencing of mitochondria-encoded genes without off-targets using reversible linkage chemistry.

Mitochondrial functional perturbation induced by IM@siCOX1

Disruption of the mitochondrial respiratory chain is expected to induce pronounced functional and phenotypic alterations at the cellular level.^37–39 MTCO1 encodes a core subunit of respiratory Complex IV, whereas MTND1 encodes an essential subunit of Complex I (NADH: ubiquinone oxidoreductase) that is critical for its catalytic activity and proper assembly. Accordingly, silencing either gene is anticipated to directly impair oxidative phosphorylation. To assess the functional consequences of mitochondrial gene knockdown, we evaluated mitochondrial morphology, membrane potential (MMP), and reactive oxygen species (ROS) production.

HeLa cells were incubated with DGIM@siCOX1, TEIM@siCOX1, or T2IM@siCOX1 for 4 days, followed by staining with MitoTracker Green and confocal microscopy analysis. While the overall cellular morphology remained comparable across all treatment groups, pronounced alterations in mitochondrial architecture were observed in the TPP-linked groups (Fig. 5A). In the DGIM@siCOX1 control group, mitochondria retained an elongated, filamentous network. In contrast, mitochondria in TEIM@siCOX1-treated cells—and more prominently in T2IM@siCOX1-treated cells—displayed marked fragmentation and a loss of filamentous structure (Fig. 5A, enlarged views), a characteristic hallmark of mitochondrial dysfunction following MTCO1 depletion.^40,41 Consistent with these observations, migration assays revealed that TPP-linked groups exhibited reduced cell proliferation (Fig. S9) and migratory capacity (Fig. S10), reflecting compromised mitochondrial function and decreased cellular viability.

Fig. 5 Assessment of mitochondrial function following mtDNA silencing by TPP-acylated siRNA. (A) Morphological changes in mitochondria. Representative confocal microscopy images showing mitochondrial morphology in HeLa cells after 96 h of treatment with different IM@siCOX1 complexes (100 nM). Mitochondria were stained with MitoTracker Green (E_x/E_m: 490/516 nm). Scale bar: 20 µm. (B) Detection of mitochondrial membrane potential (ΔΨ_m). JC-1 assay was performed on HeLa cells after treatment with different IM@siCOX1 complexes (100 nM, 48 h). CCCP-treated cells served as a positive control for depolarization. Fluorescence images show the distribution of JC-1 aggregates (red) and monomers (green), with the corresponding scatter plot showing the quantitative ratio of JC-1 aggregates to monomers (R/G) analyzed from confocal images using ImageJ. Data are presented as mean ± SD (n = 3). Statistical significance was determined by one-way ANOVA (***P < 0.001, ns: not significant). (C) Measurement of intracellular reactive oxygen species (ROS). The DCFH-DA probe was used to measure ROS levels in HeLa cells treated with different IM@siND1 complexes (100 nM, 48 h). Flow cytometry histograms (FITC channel) show the distribution of DCF fluorescence intensity. The bar graph represents the relative mean fluorescence intensity (MFI) normalized to the control. Data are mean ± SD (n = 3); statistical analysis as in (B).

We next examined changes in mitochondrial membrane potential using the JC-1 probe. Cells treated with TEIM@siCOX1 or T2IM@siCOX1 (100 nM, 48 h) showed a substantial decrease in the red-to-green fluorescence intensity ratio compared with the blank and DGIM@siCOX1 controls, indicating pronounced MMP depolarization (Fig. 5B and S11A). Carbonyl cyanide m-chlorophenyl hydrazone (CCCP) served as a positive control. Notably, T2IM@siCOX1 induced the most severe reduction in MMP, approaching the level observed in CCCP-treated cells, thereby confirming effective disruption of mitochondrial respiratory function.

Mitochondrial dysfunction was further corroborated by analysis of intracellular ROS levels. Following 48 h incubation, cells stained with DCFH-DA exhibited markedly elevated ROS levels upon treatment with TEIM@siND1 and T2IM@siND1, as determined by confocal microscopy and flow cytometry (Fig. 5C and S11B). ROS fluorescence intensities exceeded those observed for the positive control Rosup, whereas the non-targeted DGIM@siND1 group showed only minimal ROS elevation, indicating insufficient mitochondrial delivery. Taken together, these results show that TPP-mediated siRNA delivery operates effectively in cultured cells, producing robust mitochondrial gene silencing that translates into profound structural and functional mitochondrial changes.

In vivo tumor suppression by mitochondria-targeted IM@siND1

To extend our in vitro findings demonstrating that mitochondrial targeting of IM@RNA and modulation of mitochondrial genes (e.g., MTCO1 and MTND1) produced pronounced intracellular effects, we next evaluated the therapeutic potential of this strategy in vivo. Given the central role of mitochondrial dysfunction in tumor metabolism and proliferation, HepG2 cells were selected as a clinically relevant hepatocellular carcinoma model, and MTND1—a key subunit of mitochondrial respiratory Complex I—was chosen as the therapeutic target.^42,43

A subcutaneous HepG2 xenograft model was established in nude mice to assess whether mitochondria-targeted IM@RNA-mediated MTND1 silencing could suppress tumor growth in vivo. Once tumors reached approximately 150 mm³, mice were randomly assigned to three treatment groups: PBS, DGIM@siND1, or T2IM@siND1. Each group received daily intratumoral injections (1 µg of double-stranded siRNA per dose) for eight consecutive days (Fig. 6A). As shown in Fig. 6B, treatment with T2IM@siND1 led to pronounced tumor suppression after four injections, significantly slowing tumor growth and inducing progressive tumor regression. In contrast, tumors in the PBS and DGIM@siND1 groups continued to grow steadily. Quantitative analysis at the experimental endpoint revealed that T2IM@siND1 reduced the tumor mass by approximately 60% relative to controls, whereas no statistically significant difference was observed between the PBS and DGIM groups (Fig. 6C and D). Throughout the treatment period, all mice maintained stable body weights, indicating minimal systemic toxicity (Fig. S12).

Fig. 6 In vivo antitumor efficacy of mitochondria-targeting siRNA delivery in a mouse xenograft model. (A) Schematic of the experimental design for tumor-bearing mouse treatment. Mice received daily intratumoral injections of either PBS, DGIM@RNA, or T2IM@RNA (1 µg RNA per mouse per day, n = 6). (B) Tumor growth curve. Tumor volume was measured every day using a digital caliper and calculated using the formula: V = L × W² × 0.52, where L is the tumor length and W is the tumor width. Data are presented as mean ± SD (n = 6 per group). (C) Representative photographs of excised tumors from each treatment group at the endpoint. (D) Quantification of tumor weight. Statistical significance was determined by one-way ANOVA (**P < 0.01) (n = 6). (E) Immunofluorescence staining of MTND1 and MTCO1 in tumor tissues. (F) Silencing of mitochondrial target proteins in tumor tissues. Representative western blot analysis and quantification (below) of MTCO1 and MTND1 protein expression in tumor lysates. Vinculin served as a loading control. Data are presented as mean ± SD (n = 3). (G) TUNEL staining and quantification of apoptotic cells within tumor tissues.

Mechanistic studies further confirmed effective and selective mitochondrial gene silencing in vivo. Western blotting and immunofluorescence analyses of tumor tissues showed that T2IM@siND1 selectively downregulated MTND1 protein expression by 70–80% without affecting other mitochondrial proteins such as MTCO1 (Fig. 6E, F, and S13). Functionally, MTND1 depletion is expected to disrupt electron transport chain activity, elevate ROS levels and trigger apoptosis. Consistent with this mechanism, TUNEL assays revealed extensive apoptotic signaling in tumors from the T2IM@siND1 group, whereas PBS- and DGIM-treated tumors exhibited minimal apoptosis (Fig. 6G). Together, these results demonstrate that mitochondria targeted delivery of siND1 via T2IM enables potent and selective mitochondrial gene silencing in vivo, resulting in robust tumor suppression through apoptosis induction while maintaining favorable biocompatibility and minimal off-target effects.

The simplicity, reversibility, and plug-and-play nature of this chemical platform distinguish it from existing mitochondrial delivery strategies that rely on permanent and structurally invasive modifications. Beyond its therapeutic potential, this approach provides a powerful tool for probing mitochondrial gene expression, protein biogenesis, and mitochondrial contributions to cellular physiology. Collectively, our results establish reversible 2′-OH RNA acylation as a versatile and translatable chemical framework for mitochondria-targeted RNA therapeutics and mitochondrial biology research.

Conclusion

In summary, we have developed a chemically simple, modular, and user-friendly strategy for efficient RNA delivery to mitochondria and precise modulation of mitochondrial gene function in living cells and animals. By developing a TPP-conjugated, reversible 2′-OH acylation platform, we achieved transient RNA stabilization and mitochondrial targeting without permanently altering the RNA scaffold. Notably, this work represents the first demonstration of mitochondrial RNA delivery enabled by reversible DNA cloaking, highlighting the power of a single-step chemical modification to overcome longstanding delivery barriers. Using this platform, we achieved robust and selective silencing of mitochondria-encoded genes in cultured cancer cells, leading to pronounced mitochondrial dysfunction, elevated ROS production, and impaired cellular viability. Importantly, the same strategy translated effectively in vivo, where mitochondria-targeted delivery of siND1 significantly suppressed tumor growth in a xenograft model with minimal systemic toxicity and off-target effects.

Author contributions

Y. Zhuang: conceptualization, investigation, formal analysis, methodology, writing of manuscript; X. Liu: organic synthesis, investigation, formal analysis; Y. Shen: investigation, formal analysis; Q. Wang, Z. Lin, Q. Zeng and X. Chen: investigation, review; L. Jiang: supervision, funding acquisition; Shao Q. Yao: supervision, funding acquisition; J. Ge: conceptualization, resources, supervision, funding acquisition, writing of manuscript.

Conflicts of interest

The authors declare that they have no competing interests.

Data availability

The authors confirm that the data supporting the findings of this study are included within the article and its supplementary information (SI). Supplementary information: detailed experimental procedures, supplementary figures, uncropped gel and blot images, synthetic routes and characterization data (¹H and ¹³C NMR spectra) of acylated compounds, as well as additional data supporting the findings of this study. See DOI: https://doi.org/10.1039/d6sc01497f.

Acknowledgements

We thank the National Natural Science Foundation of China (22377109, 22177104, and 22505224), the Natural Science Foundation of Zhejiang Province (LR26B060003), the Fundamental Research Funds for the Provincial Universities of Zhejiang (RF-C2025005), National Medical Research Council (NMRC, 23-0740-A0001) and the Ministry of Education, Singapore (MOE, T2EP10222-0002), for financial support.

Notes and references

1. G. Burger, M. W. Gray and B. F. Lang, Mitochondrial genomes: anything goes, Trends Genet., 2003, 19, 709–716.
2. X. J. Chen and R. A. Butow, The organization and inheritance of the mitochondrial genome, Nat. Rev. Genet., 2005, 6, 815–825.
3. O. M. Russell, G. S. Gorman, R. N. Lightowlers and D. M. Turnbull, Mitochondrial Diseases: Hope for the Future, Cell, 2020, 181, 168–188.
4. H. Wen, H. Deng, B. Li, J. Chen, J. Zhu, X. Zhang, S. Yoshida and Y. Zhou, Mitochondrial diseases: from molecular mechanisms to therapeutic advances, Signal Transduct. Target. Ther., 2025, 10, 9.
5. P. K. Kopinski, L. N. Singh and S. Zhang, et al., Mitochondrial DNA variation and cancer, Nat. Rev. Cancer, 2021, 21, 431–445.
6. D. R. Welch, C. Foster and I. Rigoutsos, Roles of mitochondrial genetics in cancer metastasis, Trends Cancer, 2022, 8, 1002–1018.
7. S. Y. Wu, G. Lopez-Berestein, G. A. Calin and A. K. Sood, RNAi therapies: drugging the undruggable, Sci. Transl. Med., 2014, 6, 240ps7.
8. J. Y. Wang and J. A. Doudna, CRISPR technology: A decade of genome editing is only the beginning, Science, 2023, 379, eadd8643.
9. H. Parhiz, E. N. Atochina-Vasserman and D. Weissman, mRNA-based therapeutics: looking beyond COVID-19 vaccines, Lancet, 2024, 403, 1192–1204.
10. P. Silva-Pinheiro and M. Minczuk, The potential of mitochondrial genome engineering, Nat. Rev. Genet., 2022, 23, 199–214.
11. A. N. Patananan, T. H. Wu, P. Y. Chiou and M. A. Teitell, Modifying the Mitochondrial Genome, Cell Metab., 2016, 23, 785–796.
12. J. D. Barrera-Paez and C. T. Moraes, Mitochondrial genome engineering coming-of-age, Trends Genet., 2022, 38, 869–880.
13. N. Li, B. Wu, Y. Xiao, Y. Hao, F. Wu, Y. Wei, Y. Xu and X. Han, Tools and delivery technologies for mitochondrial gene editing, Cell Biomater, 2025, 100254.
14. L. Jiang, B. Zhou, H. Qian, H. Wang, Y. Wang, W. Fan, G. Zheng and J. Ge, Cell-type-specific CRISPRization of mitochondrial DNA using bifunctional biodegradable silica nanoparticles, Chem. Commun., 2023, 59, 9251–9254.
15. S. S. Liew, X. Qin, J. Zhou, L. Li, W. Huang and S. Q. Yao, Smart Design of Nanomaterials for Mitochondria-Targeted Nanotherapeutics, Angew. Chem., Int. Ed., 2021, 60, 2232–2256.
16. J. Zielonka, J. Joseph, A. Sikora, M. Hardy, O. Ouari, J. Vasquez-Vivar, G. Cheng, M. Lopez and B. Kalyanaraman, Mitochondria-Targeted Triphenylphosphonium-Based Compounds: Syntheses, Mechanisms of Action, and Therapeutic and Diagnostic Applications, Chem. Rev., 2017, 117, 10043–10120.
17. L. D. Cruz-Zaragoza, S. Dennerlein, A. Linden, R. Yousefi, E. Lavdovskaia, A. Aich, R. R. Falk, R. Gomkale, T. Schöndorf, M. T. Bohnsack, R. Richter-Dennerlein, H. Urlaub and P. Rehling, An in Vitro System to Silence Mitochondrial Gene Expression, Cell, 2021, 184, 5824–5837.
18. L. D. Cruz-Zaragoza, D. Dahal, M. Koschel, A. Boshnakovska, A. Zheenbekova, M. Yilmaz, M. Morgenstern, J. N. Dohrke, J. Bender, A. Valpadashi, K. A. Henningfeld, S. Oeljeklaus, L. S. Kremer, M. Breuer, O. Urbach, S. Dennerlein, M. Lidschreiber, S. Jakobs, B. Warscheid and P. Rehling, Silencing mitochondrial gene expression in living cells, Science, 2025, 389, eadr3498.
19. F. Li, Y. Liu, Y. Dong, Y. Chu, N. Song and D. Yang, Dynamic Assembly of DNA Nanostructures in Living Cells for Mitochondrial Interference, J. Am. Chem. Soc., 2022, 144, 4667–4677.
20. Y. Li, J. Li, Y. Chang, J. Zhang, Z. Wang, F. Wang, Y. Lin and L. Sui, Mitochondria-targeted drug delivery system based on tetrahedral framework nucleic acids for bone regeneration under oxidative stress, Chem. Eng. J., 2024, 496, 153723.
21. K. Gao, M. Cheng, X. Zuo, J. Lin, K. Hoogewijs, M. P. Murphy, X. D. Fu and X. Zhang, Cell Res., 2021, 31, 219–228.
22. Y. Shi, X. Zhen, Y. Zhang, Y. Li, S. Koo, Q. Saiding, N. Kong, G. Liu, W. Chen and W. Tao, Chemically Modified Platforms for Better RNA Therapeutics, Chem. Rev., 2024, 124, 929–1033.
23. Y. Wang, D. Weissman and Y. Dong, RNA chemistry and therapeutics, Nat. Rev. Drug Discovery, 2025, 24, 828–851.
24. H. Zhou, D. S. Chen, C. J. Hu, X. Hong, J. Shi and Y. Xiao, Stimuli-Responsive Nanotechnology for RNA Delivery, Adv. Sci., 2023, 10, e2303597.
25. J. Witten, Y. Hu, R. Langer and D. G. Anderson, Recent advances in nanoparticulate RNA delivery systems, Proc. Natl. Acad. Sci. U.S.A., 2024, 121, e2307798120.
26. Q. Tang and A. Khvorova, RNAi-based drug design: considerations and future directions, Nat. Rev. Drug Discovery, 2024, 23, 341–364.
27. A. Shaikh, P. K. Neeli, G. Singuru, S. Panangipalli, R. Banerjee, S. R. Maddi, R. Thennati, S. R. Bathula and S. Kotamraju, A functional and self-assembling octyl-phosphonium-tagged esculetin as an effective siRNA delivery agent, Chem. Commun., 2021, 57, 12329–12332.
28. W. Lang, W. Tan, B. Zhou, Y. Zhuang, B. Zhang, L. Jiang, S. Q. Yao and J. Ge, Mitochondria-Targeted Gene Silencing Facilitated by Mito-CPDs, Chem. Eur. J., 2023, 29, e202204021.
29. L. M. A. Ali, D. D. Su, R. Mucci, K. Martin, M. Barboiu, J. Lacour, N. Bettache and S. Ulrich, Mitochondria-Targeted Delivery of siRNA by Amphiphilic Nanovectors Self-Assembled from a Cationic [4]Helicene-Squalene Ester, ChemBioChem, 2025, 26, e202500537.
30. L. Xiao, L. Fang and E. T. Kool, 2'-OH as a universal handle for studying intracellular RNAs, Cell Chem. Biol., 2024, 31, 110–124.
31. S. R. Wang, L. Y. Wu, H. Y. Huang, W. Xiong, J. Liu, L. Wei, P. Yin, T. Tian and X. Zhou, Conditional control of RNA-guided nucleic acid cleavage and gene editing, Nat. Commun., 2020, 11, 91.
32. A. Kohler, A. Barrientos, F. Fontanesi, M. K. Ott, A. M. Kietrys and E. T. Kool, RNA cloaking by reversible acylation, Angew. Chem., Int. Ed., 2018, 57, 3059–3063.
33. L. Fang, L. Xiao, Y. W. Jun, Y. Onishi and E. T. Kool, Reversible 2'-OH acylation enhances RNA stability, Nat. Chem., 2023, 15, 1296–1305.
34. J. Guo, S. Chen, Y. Onishi, Q. Shi, Y. Song, H. Mei, L. Chen, E. T. Kool and R. Y. Zhu, RNA control via redox-responsive acylation, Angew. Chem., Int. Ed., 2024, 63, e202402178.
35. H. Lei, W. Xiong, M. Li, Q. Qi, X. Liu, S. Wang, T. Tian and X. Zhou, Enhanced control of RNA modification and CRISPR-Cas activity through redox-triggered disulfide cleavage, Bioorg. Med. Chem., 2024, 112, 117878.
36. R. Shioi, L. Xiao and E. T. Kool, Aqueous Activation of RNA 2’-OH for Conjugation with Amines and Thiols, Bioconjugate Chem., 2024, 35, 40–50.
37. Y. Gao, R. Sun, M. Zhao, J. Ding, A. Wang, S. Ye, Y. Zhang, Q. Mao, W. Xie, G. Ma and H. Shi, On-Site-Specific Immobilization of Mitochondrial-Targeted NIR Fluorescent Probe for Prolonged Tumor Imaging, Anal. Chem., 2020, 92, 6977–6983.
38. Y. Zong, H. Li, P. Liao, L. Chen, Y. Pan, Y. Zheng, C. Zhang, D. Liu, M. Zheng and J. Gao, Mitochondrial dysfunction: mechanisms and advances in therapy, Signal Transduc. t Target Ther, 2024, 9, 124.
39. A. Kohler, A. Barrientos, F. Fontanesi and M. Ott, The functional significance of mitochondrial respiratory chain super complexes, EMBO Rep., 2023, 24, e57092.
40. R. Zhai, B. Fang, Y. Lai, B. Peng, H. Bai, X. Liu, L. Li and W. Huang, Small-molecule fluorogenic probes for mitochondrial nanoscale imaging, Chem. Soc. Rev., 2023, 52, 942–972.
41. Y. Ding, J. Li, J. Zhang, P. Li, H. Bai, B. Fang, H. Fang, K. Huang, G. Wang, C. J. Nowell, N. H. Voelcker, B. Peng, L. Li and W. Huang, Mitochondrial segmentation and function prediction in live-cell images with deep learning, Nat. Commun., 2025, 16, 743.
42. M. Komza and J. E. Chipuk, Mitochondrial metabolism: A moving target in hepatocellular carcinoma therapy, J. Cell. Physiol., 2025, 240, e31441.
43. L. Kamalian, A. E. Chadwick, M. Bayliss, N. S. French, M. Monshouwer, J. Snoeys and B. K. Park, The utility of HepG2 cells to identify direct mitochondrial dysfunction in the absence of cell death, Toxicol. in Vitro, 2015, 29, 732–734.

---

Footnote

† These authors contributed equally to this work.

---