Radioprotection of healthy tissue via nanoparticle-delivered mRNA encoding for a damage-suppressor protein found in tardigrades

Abstract

Patients undergoing radiation therapy experience debilitating side effects because of toxicity arising from radiation-induced DNA strand breaks in normal peritumoural cells. Here, inspired by the ability of tardigrades to resist extreme radiation through the expression of a damage-suppressor protein that binds to DNA and reduces strand breaks, we show that the local and transient expression of the protein can reduce radiation-induced DNA damage in oral and rectal epithelial tissues (which are commonly affected during radiotherapy for head-and-neck and prostate cancers, respectively). We used ionizable lipid nanoparticles supplemented with biodegradable cationic polymers to enhance the transfection efficiency and delivery of messenger RNA encoding the damage-suppressor protein into buccal and rectal tissues. In mice with orthotopic oral cancer, messenger RNA-based radioprotection of normal tissue preserved the efficacy of radiation therapy. The strategy may be broadly applicable to the protection of healthy tissue from DNA-damaging agents.

Over 60% of patients with cancer undergo radiation therapy at some point in their disease process, whether for curative intent or palliation¹. Most, if not all, patients develop injury to the surrounding healthy and normal tissue as a result of treatment^2,3. The primary mechanism of cell death is deoxyribonucleic acid (DNA) damage caused by single- and double-stranded DNA breaks⁴. The toxicities resulting from radiation-induced normal tissue injury are dependent on the location of treatment, with some of the most common short-term toxicities being oral mucositis and proctitis^5–7. This normal tissue injury may lead to severe morbidity and ultimately treatment breaks or discontinuation. Breaks in radiotherapy have been associated with inferior tumour control in most difficult-to-treat cancers^8–10. Thus, a major challenge in radiation therapy is to reduce the damage to normal tissue surrounding the target tumour while enhancing the cytotoxic effect within tumours. As an example of the burden and impact of adverse effects in clinical trials, we summarize events of acute oral mucositis, which were to occur in up to 100% of patients with head and neck cancer undergoing radiation therapy (Supplementary Table 1) and which were associated with costs of up to US$17,000 per patient^11–17.

Attempts to reduce these side effects include systemically administered radioprotectants, such as amifostine and GC4419; tissue spacing technologies, such as SpacerOAR; and radiation techniques, such as intensity-modulated radiation therapy^18–20. However, all of these methods to reduce side effects have limitations in protecting normal tissues, including selectivity, severe hypotension, user experience with optimal spacer placement and infection^19,21. New methods for radiation protection are warranted to reduce morbidity, as well as to improve treatment adherence and overall survival. Moreover, opportunities for radiation dose escalation to improve treatment efficacy may be possible as a function of the improved radiotolerance of dose-limiting normal structures.

Tardigrades are water-dwelling, eight-legged micro-animals that can be found everywhere from high altitudes atop mountains to the deep sea. Tardigrades have an extraordinary ability to tolerate immense doses of radiation that would be lethal to most other life forms on the planet (Supplementary Fig. 1). In a dehydrated state, tardigrades can also withstand a wide range of physical extremes such as extreme temperatures (up to 273 °C), pressure (7.5 GPa), immersion in organic solvents and direct exposure to open spaces^22,23. In the most stress-tolerant tardigrade species, Ramazzottius varieornatus, numerous ubiquitous and tardigrade-unique stress-related genes are overexpressed, enabling protection against extreme conditions. In particular, the tardigrade-unique damage-suppressor (Dsup) protein was found to co-localize with DNA, enabling tolerance to DNA-damaging radiation²². The Dsup protein associates with DNA and protects it from single- and double-stranded DNA breaks. In addition, this protein is highly basic, suggesting its potential association with DNA via electrostatic interactions. A Dsup expression vector was previously transfected into immortalized human cells, enabling expression of the Dsup protein. In stably transformed cells, radiation-induced DNA damage was reduced by approximately 40%. Human cells expressing a Dsup mutant lacking the DNA-associating domain showed inferior radiotolerance compared with those expressing full-length Dsup protein. Moreover, human cells expressing Dsup exhibit no reduction in cell proliferation²². Thus, the expression of the full-length Dsup protein is a promising avenue for imparting radiotolerance to human cells.

Although previous studies have elegantly established the pharmacological activity of the Dsup protein in vitro²², a clear path to its clinical translation is not yet available. We posited that messenger RNA-based local delivery of Dsup to high-risk areas could be used to minimize radiation-induced DNA damage (Fig. 1). We chose mRNA over plasmid DNA- and protein-based therapeutics for several reasons. First, the Dsup protein must be present intranuclearly to elicit its activity. Although several exciting platforms have been developed²⁴, intracellular delivery of proteins remains a considerable challenge. By contrast, mRNA-based delivery results in intracellular protein expression, facilitating subsequent nuclear co-localization. Second, following mRNA delivery, the protein is expressed for prolonged periods, which may be due to the multiple copies of protein produced for every copy of the delivered mRNA²⁵. Third, DNA-based therapies have progressed substantially in the past few decades. However, there remains the risk of genomic integration. mRNA-based therapies have been used in millions of individuals as a vaccination strategy during the severe acute respiratory syndrome coronavirus 2 pandemic^26,27. Hence, the safety profile of in vivo mRNA delivery using nanoparticles (NPs) has been well established. Finally, mRNA must be delivered into the cytoplasm and may be more readily converted to protein compared with DNA-based therapy, which needs to be delivered into the nucleus. Hence, in this project, we developed an mRNA-based therapy for the expression of Dsup protein in mouse models to assess radioprotection of the oral cavity and rectum.

Fig. 1 |.

Schematic illustrating the delivery of Dsup mRNA by NPs to protect against proctitis and oral mucositis caused by prostate and head and neck irradiation, respectively.

Results

Polymer–lipid nanoparticles for intracellular delivery of mRNA

We designed an NP system for mRNA delivery to the cytoplasm where it is translated into a therapeutic protein. Lipid nanoparticles (LNPs) are the most advanced formulation for mRNA delivery^28–30. Upon cellular uptake, LNPs enter the endolysosomes, where they destabilize the vesicular membrane through ionic interactions³¹. The level of endolysosomal escape determines the in vitro transfection efficiency^32,33. Interestingly, cationic polymers also mediate endolysosomal escape through pH modulation^34–36. We reasoned that the addition of polymer to the LNPs would endow them with two mechanisms of endolysosomal escape and improve in vitro transfection. We evaluated the in vitro transfection efficiency of >200 novel cationic branched poly(β-amino ester) (PBAE) polymers, which we recently described³⁷ (Fig. 2a). An mRNA encoding enhanced green fluorescent protein (eGFP) was complexed with polymers to form polymer-based NPs. Primary human oral epithelial cells (HOECs) were transfected with NPs for 24 h, and eGFP expression was measured using flow cytometry. Through this screen, we identified several polymers capable of transfecting >30% of cells in vitro (Fig. 2b). We incorporated one of our top polymers, polymer #1156, into the ionizable lipid-based NPs. The Fourier transform infrared (FTIR) spectroscopy, nuclear magnetic resonance (NMR) and gel permeation chromatography (GPC)-based characterizations of polymer 1156 are shown in Supplementary Fig. 2.

Fig. 2 |. Polymer–LNPs for intracellular delivery of mRNA.

a, Synthesis scheme for the PBAE polymers. b, HOECs were treated with polyplexes made by complexing PBAE polymers and eGFP mRNA (mRNA = 1 μg ml⁻¹). Transfection was measured 24 h later using a high-throughput flow cytometry screen. eGFP-expressing cells were used for gating. Data are shown as median of five biologically independent experiments. c,d, HOECs (c) and human colon epithelial Caco-2 cells (d) were transfected with LNPs or polymer–LNPs (represented as LNP-1156-xx or LNP-844-xx) loaded with eGFP mRNA. Transfection was measured 24 h later using flow cytometry. The numbers 1156 and 844 represent the polymer in the LNP. The last numbers in the names, namely ‘10’, ‘20’ and ‘40’, represent the weight ratios of 1:0.1, 1:0.2 and 1:0.4 between ionizable lipid:polymer, respectively. Data are represented as mean ± s.d. (n = 4) of four different biologically independent experiments. Statistical analyses were performed on the mean fluorescence intensity (MFI) dataset by one-way ANOVA, post hoc Tukey determined using GraphPad Prism 9.0. NS, not significant.

As we were interested in applications within the oral cavity and gastrointestinal tract, we used DLin-KC2-DMA (a lipid that does not contain a degradable ester bond) as the ionizable lipid. The formulation also contained cholesterol, 1,2-dioleoyl-sn-glycero-3-p hosphoethanolamine (DOPE) and 1,2-dimyristoyl-sn-glycero-3-p hosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (ammonium salt) (PEG-PE). Starting with a formulation identified by Kauffman and colleagues (referred to as Formulation A)³⁸, we tested four different molar ratios of the four components (Extended Data Fig. 1a). We maintained the molar content of the ionizable lipid and PEG-PE constant and changed those of DOPE and cholesterol. Particle size was analysed using dynamic light scattering. The formulations were comparable and had a particle size ~200 nm (Extended Data Fig. 1b). Membrane fluidity was characterized by measuring the generalized polarization value using a Laurdan assay^39,40. The formulations showed comparable values (Extended Data Fig. 1c). Next, we assessed the transfection efficiency of the four LNPs in HOECs in vitro. The formulation containing high content of DOPE and low content of cholesterol showed the highest transfection efficiency (referred to as Formulation D) (Extended Data Fig. 1d). We compared the in vivo transfection efficiencies of our top formulation (Formulation D) to a previously described formulation (Formulation A). Formulation D produced yielded higher transfection (Extended Data Fig. 1e) and was used in subsequent experiments to produce polymer–LNPs.

We incorporated our top-performing polymer, 1156, into Formulation D to produce polymer–LNPs. Polymer 1156 was mixed with the lipid mixture dissolved in ethanol. The polymer–lipid mixture was then complexed with mRNA dissolved in citrate buffer to yield the polymer–LNPs. Different weight ratios of ionizable lipid:polymer 1156 were tested (1:0.1, 1:0.2 and 1:0.4). Interestingly, the addition of polymer 1156 to the LNPs resulted in increased transfection. Transfection was maximal when the weight ratio of the ionizable lipid:polymer was 1:0.2 (Fig. 2c). This formulation is referred to as LNP-1156-20.

Next, we asked if polymer-based improvement in transfection was specific to HOECs and polymer 1156, or if it would be seen in other systems. We previously screened our polymer library using a colon epithelial cell line (Caco-2 cells)³⁷. From these studies, we identified polymer 844 as one of the top hits (the FTIR, GPC and NMR characterization of 844 is shown in Extended Data Fig. 1). We included polymer 844 in DLin-KC2-DMA-based LNPs (Formulation D from Extended Data Fig. 1) and tested their transfection efficiency in Caco-2 cells. Consistent with our previous results, the inclusion of the polymer in LNPs led to an increase in transfection efficiency. In these studies, the optimal weight ratio of ionizable lipid:polymer was found to be 1:0.2 (Fig. 2d). This formulation is referred to as LNP-844-20. Interestingly, LNP-844-20 did not outperform the LNPs in oral epithelial cells (Supplementary Fig. 3a). However, in colon epithelial cells, polymer 1156 improved the activity of LNPs, but not to the same degree as 844 (Supplementary Fig. 3b).

We evaluated the physicochemical properties of the LNPs and polymer–LNPs using dynamic light scattering, zeta potential analysis and transmission electron microscopy. LNPs and polymer–LNPs had a slight net negative zeta potential and a particle size of ~200 nm (Extended Data Fig. 2a,b). Electron microscopy imaging revealed that the morphology of the formulations was comparable (Extended Data Fig. 2c,d). The generalized polarization value was determined using a Laurdan assay. The addition of the polymer 844, but not 1156, produced a decrease in membrane fluidity at pH 6.5 and 4.5 (Extended Data Fig. 2e). This suggests that the inclusion of the polymer may be affecting the packing of the lipid membrane, at least under some pH conditions. Finally, mRNA encapsulation efficiency was measured using a RiboGreen assay. The encapsulation efficiency in LNPs was 85%. The encapsulation efficiencies in LNP-1156-20 and LNP-844-20 were 86% and 78%, respectively (Extended Data Fig. 2f). Particle size and encapsulation efficiencies upon storage at −20 °C in sucrose containing buffers did not change greatly over the course of 3 weeks (Supplementary Fig. 4). Longer-term stability studies (for example, 2 years) will be of much value for clinical translation.

We were interested in understanding the mechanism underlying the improved transfection of the polymer–LNPs. First, we compared the cell uptake of polymer–lipid and LNPs. HOECs were transfected with NPs loaded with eGFP mRNA conjugated to Cy5 dye. As expected, protein expression was higher in cells treated with polymer–LNPs than in cells treated with LNPs. Surprisingly, the cellular uptakes of the two NPs were nearly identical (Supplementary Fig. 5a). Next, we measured membrane disruption, a key mechanism for the endolysosomal escape of LNPs, by measuring red blood cell lysis in vitro. Indeed, polymer–LNPs resulted in greater red blood cell lysis (Supplementary Fig. 5b). These preliminary studies suggest that polymer–LNPs have the same cellular uptake as LNPs; however, they may elicit improved endolysosomal escape.

Local delivery of mRNA polymer–LNPs facilitates local protein expression in vivo

Next, we evaluated the local protein expression in vivo upon the injection of firefly luciferase (fluc) mRNA polymer–LNPs. LNP-1156-20 was used for buccal delivery and LNP-844-20 was used for rectal delivery. The mice underwent local injection into the buccal or rectal tissue and were then assessed at 3 h, 6 h, 24 h and 96 h. Luciferase expression was significantly increased for the fluc mRNA polymer–LNP at 3 h, 6 h and 24 h compared with fluc mRNA alone (Fig. 3a–c). Maximal protein expression was noted at 6 h in both buccal and rectal tissues. Minimal to no protein expression was observed after 96 h.

Fig. 3 |. In vivo expression after local delivery of NPs.

a, Representative images of mice with an overlaid luminescent signal at 3 h, 6 h, 24 h and 96 h post-injection. b, Total emission (photons per second) in right buccal tissue at 3 h, 6 h, 24 h and 96 h after injection of luciferase mRNA NPs. c, Total emission (photons per second) in rectal tissue at 3 h, 6 h, 24 h and 96 h after injection of luciferase mRNA NPs. Data are presented as mean ± s.d. (n = 4 per arm). P values were determined for each time point by unpaired t-test using GraphPad Prism 9.0.

To obtain a more detailed measurement of protein expression and NP distribution, we conducted a pharmacokinetic study. Mice were treated with fluc mRNA-loaded LNP-1156-20 containing a lipophilic dye, 1,1-dioctadecyl-3,3,3,3-tetramethylindotricarbocyanine iodide (DiR). Treatments were administered in the buccal tissue, rectal tissue and intravenously via retro-orbital injection. At 3 h, 6 h and 24 h post-dose, mice were euthanized for analysis of bioluminescence and fluorescence in major organs. In mice receiving intravenous mRNA NPs, maximal fluorescence and bioluminescence were found in the liver (Supplementary Fig. 6). By contrast, in mice receiving buccal mRNA NPs, maximal fluorescence and bioluminescence were found at the site of injection in the cheek. Importantly, the signal was focused on the cheek that was treated with the NPs and no signal was observed in the opposite cheek (Supplementary Fig. 6). Similar results were obtained with colonic injections with LNP-844-20 (Supplementary Fig. 7). To investigate the distribution of particles post-injection, we evaluated buccal tissue at 15 min and 6 h post-injection of fluc mRNA-loaded LNP-1156-20 containing the DiR dye. We found that NPs were distributed up to 1.8 mm away from the proposed injection site (Supplementary Fig. 8). These data suggested that injection of our NPs into buccal and rectal tissues resulted in only local expression of the protein, and this coincided with the predominantly local retention of the NPs.

Delivery of Dsup mRNA polymer–LNPs reduces radiation-induced DNA damage and promotes cell viability

Our data indicated that we could deliver different types of mRNA into the oral and rectal cells and tissues. Next, we evaluated the efficacy of mRNA encoding the Dsup protein. As antibodies recognizing the Dsup protein are not commercially available, we designed an mRNA encoding the Dsup protein fused with an eGFP protein at its C-terminus. HOECs were transfected with polymer–LNPs loaded with eGFP and Dsup-GFP mRNA and observed using a fluorescence microscope (Supplementary Fig. 9). In eGFP mRNA-treated cells, fluorescence was found to be diffusely present throughout the cell. By contrast, in cells expressing Dsup-GFP mRNA, fluorescence was concentrated in the nucleus. This was likely due to the nuclear localization signal present in the C-terminal region of the Dsup protein^41,42. We determined if the expression of Dsup protein remained intracellular or if it was secreted. HOECs were treated with LNP-1156-20 loaded with GFP mRNA and Dsup-GFP mRNA. Fluorescence in the cell supernatant and in the cell lysate was measured using a plate reader. Indeed, in both mRNA treatment groups, the signal was localized to the cell lysate, suggesting intracellular expression of the protein (Supplementary Fig. 10).

To further confirm that the Dsup-GFP mRNA yielded a fusion protein, we evaluated the protein using western blotting and silver staining. Specifically, HOECs were transfected with LNP-1156-20 loaded with GFP mRNA or Dsup-GFP mRNA. Twenty-four hours later, cells were lysed and the lysate was subjected to western blot analysis using an anti-GFP antibody and silver staining analysis. In both cases, the Dsup-GFP band appeared at a higher molecular weight compared with the GFP band (Supplementary Fig. 11a,b). The Dsup-GFP band was analysed using mass spectrometry. Fragmentation patterns of the protein are shown in Supplementary Fig. 11c. We identified 57 peptides that represented 73% coverage of the Dsup protein. This analysis is included in Supplementary Data 1 of the paper.

To evaluate the effect of Dsup expression on radiation-induced DNA damage in oral and colonic epithelial cells, we initially performed comet assays on human oral and colonic epithelial cells. Comet assays are routinely used to measure DNA damage (that is, strand breaks, DNA adducts, excision repair sites and cross-links) at the single-cell level⁴³. The amount of DNA damage was directly related to the percentage of DNA in the comet tail per cell. Dsup expression conferred a significant degree of protection from 10 Gy of radiation compared with control mRNA (GFP) and no treatment (Fig. 4a).

Fig. 4 |. Delivery of Dsup mRNA reduces radiation-induced DNA damage and promotes cell viability.

a, An alkaline comet assay of cells pretreated with Dsup or control mRNA NPs and exposed to radiation. HOECs and colorectal epithelial (Caco-2) cells that expressed Dsup had significantly less single-stranded DNA damage than the control (GFP) mRNA (P < 0.0001) (n = 300 cells for buccal tissue and n = 150 cells for rectal tissue). The no treatment control cells were not exposed to radiation. Results are from three biologically independent experiments. b, To validate a decrease in DNA damage resulting from radiation exposure, HOEC and Caco-2 cells pretreated with Dsup or control mRNA NPs that expressed Dsup had significantly less γ-H2AX foci per nucleus than the control mRNA (P < 0.0001) (n = 36 z-projection images per arm evaluated with ~100 cells per image). Images were analysed by CellProfiler 4.2.1. Fluorescence images were obtained on the red channel. Results are from three biologically independent experiments. c, Clonogenic assays were performed to evaluate cell viability. Human oral (HOEC) and colorectal (Caco-2) epithelial cells that expressed Dsup had significantly more colonies than the control (GFP) mRNA. Data are presented as mean ± s.d. (n = 3). Each data point represents one biologically independent experiment. P values were determined by unpaired t-test using GraphPad Prism 9.0.

To validate these results, we performed phosphorylated histone H2AX (γ-H2AX) staining of irradiated cells. γ-H2AX is used as a biomarker of double-stranded DNA damage⁴⁴. Cells were transfected as described above and exposed to radiation. Directly after irradiation, cells were subsequently placed on ice, fixed and stained for γ-H2AX foci/nuclei. The quantity of γ-H2AX foci per nucleus was compared between Dsup mRNA NP-treated cells and control mRNA NP-treated cells. We found that Dsup expression resulted in a significant amount of radioprotection compared with that of the control mRNA (Fig. 4b).

Finally, clonogenic cell survival assays were performed to determine the impact of Dsup mRNA on radioprotection for cell viability and proliferation. We found that cells expressing Dsup had a significantly greater number of colonies after exposure to multiple doses of radiation than the control (Fig. 4c). We compared the activity of Dsup-GFP mRNA NPs (1 μg ml⁻¹ mRNA) to WR-1065 (active metabolite of amifostine; in a concentration ranging 0.5–2 mM). Dsup-GFP mRNA NPs produced radioprotection comparable to WR-1065 at 2 mM and were more efficacious than WR-1065 at lower concentrations (Extended Data Fig. 3a–d).

In summary, we used three orthogonal assays to confirm that Dsup mRNA-treated cells had greater protection against radiation-induced DNA damage, and that these cells had greater survivability.

Dsup mRNA delivery protects local tissue from high-dose radiation therapy

To determine the impact of Dsup on radioprotection in vivo, we treated healthy C57BL/6 mice locally with Dsup and control mRNA polymer–LNP (Fig. 5a,b). Mice that received no treatment were used as the controls. All mice were exposed to a single radiation dose directed to the oral cavity or rectum 6 h after the mRNA treatment. The mice were euthanized an hour after radiation therapy, and the tissues were preserved and stained for immunohistochemical analysis of γ-H2AX, a clinical readout for DNA damage, and 4′,6-diamidino-2-phenylindole (DAPI), a nuclear stain. We found that animals receiving Dsup treatment had significantly lower γ-H2AX foci per nucleus than the control mRNA NP-treated and untreated animals (Fig. 5c–f). The protective effects were noted at the injection site where there was GFP expression compared with the peri-injection site where there was no GFP expression (Extended Data Fig. 4a). Furthermore, the local protection was more pronounced in mice irradiated 6 h and 24 h after Dsup treatment compared with irradiation at 96 h (Extended Data Fig. 4b).

Fig. 5 |. Local delivery of Dsup mRNA NP protects normal tissue from high-dose radiation and does not impact treatment of contralateral orthotopic syngeneic tumour growth.

a,b, Schematic for local mRNA NP delivery into the buccal mucosa (a) and rectal mucosa (b). The schematics were created with http://BioRender.com. c, Quantification of DNA damage in buccal tissue from radiation therapy (10 Gy in 1 fraction) showing Dsup mRNA NP reduces the number of γ-H2AX foci per nucleus compared with a control (GFP) mRNA NP (P = 0.0007) and no treatment (P < 0.0001) (n = 2–4 z-projection images per animal, 7 animals per arm). Images were analysed by CellProfiler 4.2.1. d, Quantification of DNA damage in rectal tissue from radiation therapy (10 Gy in 1 fraction) showing Dsup mRNA NP reduces the number of γ-H2AX foci per nucleus compared with a control (GFP) mRNA NP (P = 0.0007) and no treatment (P < 0.0001) (n = 2–4 z-projection images per animal, 6 animals per arm). Images were analysed by CellProfiler 4.2.1. e,f, Representative images from the quantification of γ-H2AX foci per nucleus in the buccal (e) and rectal (f) tissue, respectively. Fluorescence images showed merged images from red (AF568) and blue (DAPI) channels. g, Schematic for testing the impact of the local delivery of Dsup mRNA on contralateral orthotopic syngeneic tumour growth combined with radiation (n = 7 per arm). The schematic was created with http://BioRender.com. RT, radiation therapy. h, Tumour growth curves of mice treated with Dsup mRNA NP, control mRNA NP and no treatment before radiation demonstrating no change in tumour growth as a result of NP delivery. Data are presented as mean ± s.d. P values were determined by one-way ANOVA using GraphPad Prism 9.0.

To ensure that Dsup expression did not produce a systemic immune response, we performed cytokine and transaminase analyses of the plasma of mice treated with one and two (1 week apart) buccal injections of the Dsup mRNA NPs compared with untreated mice. There was no difference in cytokine values and transaminase values between the different treatment groups (Supplementary Fig. 12).

Local delivery of Dsup mRNA does not protect the contralateral orthotopic syngeneic tumour growth

Critical to the use of any radioprotectant is to confirm the lack of protection of the tumour or surrounding areas at risk for microscopic cancer cell invasion. To assess radioprotection in a relevant cancer model, we evaluated tumour growth in an orthotopic syngeneic oral cancer mouse model (MOC-2) where the Dsup mRNA polymer–LNP was injected in the contralateral buccal mucosa, and the tumour was exposed to radiation (Fig. 5g). There was no difference in tumour growth between mice treated with Dsup mRNA NP compared with the control mRNA NP and untreated controls (Fig. 5h). In addition, there was no difference in tumour growth between non-irradiated mice administered Dsup mRNA NP, control mRNA NP or no treatment (Supplementary Fig. 13).

In the experiments described above, the specificity of Dsup mRNA NPs is likely mediated by the localized injections and lack of spread of NPs to the contralateral cheek, as evidenced by our pharmacokinetic studies (Supplementary Fig. 5). To confirm this, we treated orthotopic MOC-2 tumour-bearing mice with Dsup-GFP mRNA NPs in the contralateral cheek. Healthy tissue (injected with the NPs) and tumour tissue were excised 6 h post-dose. The tissues were stained with an antibody against GFP and imaged using confocal microscopy. Indeed, a strong signal was observed in the healthy tissue and at the site of injection. However, only background fluorescence was observed in the tumour (Supplementary Fig. 14).

We were interested in understanding if our NPs would be specific to normal cells if the tumour was adjacent to the site of injection. To evaluate this, human oral cancer MOC-2 and normal oral epithelial cells were treated with Dsup-GFP mRNA NPs (LNP-1156-20). Transfection efficiency was monitored using flow cytometry. LNP-1156-20 transfected >90% of HOECs (normal cells) and only 15% of the tumour cells (Supplementary Figs. 15 and 16). This selectivity may be of benefit for potential clinical applications.

Discussion

Radiation therapy is a highly effective anticancer modality that has long been part of the clinical oncologist’s arsenal; however, it is associated with substantial acute and chronic toxicities as exemplified in the acute oral mucositis reviewed in Supplementary Table 1 (refs. 11–17). Oral mucositis is a primary cause of treatment breaks or complete discontinuation of therapy, which can result in inferior tumour control in the larynx, pharynx and oral cavity⁹. Proctitis and urethritis are commonly observed in patients with prostate cancer undergoing radiation therapy^2,45–47. Rectal toxicities, such as rectal bleeding, are associated with the radiation dose as well as the volume of the rectum exposed to radiation⁴⁸. Furthermore, as the field of oncology has successfully improved the treatment of certain localized cancers, including prostate, breast and human papillomavirus positive oropharyngeal cancers, we leave patients with severe chronic side effects to treatment³. Opportunities to limit these toxicities will improve therapeutic efficacy and quality of life of patients and reduce treatment-related regret⁴⁹. In patients undergoing radiation therapy in an adjuvant setting, there is an increased risk of wound complications and chronic wound formation⁵⁰.

Radiation, through direct interaction with DNA and the production of reactive oxygen species, produces DNA strand breaks. This causes massive cell death and initiates a local inflammatory cascade⁵¹. This mechanism lies at the heart of radiation-induced mucositis and its toxicity. Hence, we posited that a treatment that reduces radiation-induced DNA strand breaks in normal tissues may reduce radiation-induced toxicities. Here we show that the local delivery of Dsup mRNA can provide substantial radiation protection in normal tissues. Our preclinical results suggested that Dsup expression protects against DNA damage and leads to an increase in cell viability and proliferation. In addition, the local delivery of Dsup mRNA does not confer a radioprotective benefit to the primary tumour of an oral cavity cancer mouse model. These studies show the utility of the transient expression of radioprotectant proteins to reduce radiation-induced damage without compromising the effectiveness of radiation treatment on the tumour.

Furthermore, we demonstrate that cells expressing Dsup exhibit a magnitude of radioprotection comparable to that provided by a clinically approved radioprotectant. The primary mechanism of action for Dsup has been identified as the binding to nucleosomes, thereby protecting DNA from hydroxyl radicals generated by ionizing radiation^41,52–55. This mechanism markedly differs from that of all clinically approved and trialed small-molecule radioprotectants, which function primarily as free radical scavengers and antioxidants.

To achieve this goal of local Dsup expression, we needed to identify potent vectors for mRNA delivery. There are numerous challenges in the delivery of mRNA, including protection of mRNA under physiological conditions, cell uptake, intracellular release, endolysosomal escape and minimization of toxicity²⁹. This has led to enormous global efforts directed towards the discovery of new materials (such as lipids and polymers) for RNA delivery^56–59. As has been recently noted, this has yielded a toolbox for mRNA delivery vectors that offers different properties, advantages and disadvantages⁶⁰. We focused our efforts on combining two delivery modalities, which have been widely studied for nucleic acid delivery: ionizable lipids and cationic polymers. We found that a combination of these two modalities yielded higher transfection rates. This observation prompted the question of whether existing and clinically characterized materials can be combined to maximize mRNA delivery. This strategy may be valuable, as it carries a relatively lower risk than the use of new materials, where clinical safety, performance and large-scale manufacturing are less established⁶¹. Hypothesis-driven experiments and unbiased high-throughput screening may be effective methods for answering this question.

Although our study establishes a preliminary proof of concept regarding the utility of Dsup-based therapy, several challenges and unanswered questions remain. In this study, we used xenoprotein for radioprotection. Despite showing no toxicity after two injections 1 week apart, it is possible that repeat dosing will neutralize the protection of Dsup owing to anti-Dsup antibodies. To overcome the potential immunogenicity, we could consider using mRNA with modified bases, humanizing the Dsup protein or using the ApoE protein, which was previously shown to have anti-apoptotic properties⁶². Furthermore, with an mRNA-based system, transient expression of ~1 day was achieved. This may provide only a short window for the administration of radiotherapy and may require repeated mRNA treatment for subsequent rounds of radiotherapy. Alternative platforms, such as self-amplifying mRNA, which provide protein expression for a longer duration, may be more suitable for this application. We used tissue injections for the local administration of mRNA therapy. The path to translation may include packaging of mRNA delivery systems into clinically suitable forms outside of direct injection, including microneedles or topical agents^63,64. In addition, testing the utility of this approach in other tumour models may be of great interest.

Our study has two major limitations. First, mouse models of oral mucositis are limited to inflammation of the oral tongue and not buccal tissue⁶⁵. As a result, we were limited to measuring markers of radioprotection, which, although clinically relevant, provide an indirect measurement of efficacy. Second, antibodies against the Dsup protein are not commercially available. Consequently, we were limited to using antibodies against GFP, which was fused to Dsup in our mRNA construct. The availability of antibodies against the Dsup protein may be beneficial to intensify our characterization efforts.

The use of Dsup mRNA delivery may be co-opted for several other clinical applications, including protection of normal tissue from DNA-damaging chemotherapies or progressive degeneration of specific tissues, cancer predisposition, chromosomal instability and hypersensitivity to DNA-damaging agents. In addition to cancer-related applications, the use of Dsup protein can be extended to total body exposure to space radiation or as prophylaxis against nuclear radiation exposure^66,67. Overall, Dsup mRNA delivery could potentially offer an entirely new modality for the prevention of radiation-induced toxicity.

Methods

Synthesis of hybrid PBAE polymer library

To screen for cationic polyplexes, a library of >200 branched hybrid PBAE polymers was synthesized using a previously described protocol³⁷. An amine and a diacrylate were mixed in an 8 ml glass vial containing a magnetic stir bar. The vial was placed on a magnetic stir plate at 90 °C for ~18 h to yield the linear PBAE polymer. The linear polymer was dissolved in dimethyl sulfoxide (DMSO). The two linear polymers were mixed with a branching agent and allowed to react on a shaker for 24 h at room temperature. Finally, an end-capping agent was added to the mixture, which was allowed to react for 1 h at room temperature. The polymer library was stored at −20 °C until further use. The lead polymers, 1156 and 844, were purified and incorporated into LNPs. A detailed description of their syntheses is provided below.

Synthesis of polymer 1156.

Two linear PBAE polymers (198 and 95) were produced by melt polymerization of an amine and diacrylate. To synthesize polymer 198, 0.4 g of 3-amino-1-propanol (1 equiv.) was mixed with 1.45 g 1,6-hexanediol diacrylate (1.2 equiv.) in a glass vial equipped with a magnetic stir bar. To synthesize polymer 95, 0.4 g of 3-amino-1,2-propanediol (1 equiv.) was mixed with 1.09 g 1,6-hexanediol diacrylate (1.1 equiv.) in a glass vial equipped with a magnetic stir bar. The vials were placed on a stir plate at 90 °C for 16 h. The polymers were dissolved in dimethylformamide at a concentration of 166.7 mg ml⁻¹. Next, 0.3 ml polymer 198 (50 mg) and 0.18 polymer 95 (30 mg) were mixed with 0.295 mg N-methylethylenediamine (branching agent), which was also dissolved in dimethylformamide (100 mg ml⁻¹). The reaction was allowed to proceed under shaking at room temperature for 24 h. The polymer was capped by reacting it with 25 mg 1-(3-aminopropyl)-4-methylpiperazine in dimethylformamide (0.193 ml) at room temperature for 1 h. The reaction mixture was added to diethyl ether to precipitate the polymer, which was separated by centrifugation at 1,500 RCF for 30 min at 4 °C in a swing-bucket centrifuge. The polymer was dried under argon stream/vacuum overnight and then stored at −20 °C until further use.

Synthesis of 844.

Polymer 844 contained two linear polymers, 198 and 13. Polymer 198 was synthesized as described above and dissolved in dimethylformamide at a concentration of 166.7 mg ml⁻¹. To synthesize polymer 13, we added 0.4 g 5-amino-1-pentanol (1 equiv.) and 0.93 g 1,4-butanediol diacrylate (1.2 equiv.) to a glass vial equipped with a stir bar. The vial was placed at 90 °C for 16 h to enable polymerization. The resultant polymer was dissolved in dimethylformamide at a concentration of 166.7 mg ml⁻¹. Next, 0.3 ml polymer 198 (50 mg) and 0.18 polymer 13 (30 mg) were mixed with 0.185 mg tris(2-aminoethyl) amine (branching agent), which was also dissolved in dimethylformamide. The reaction was allowed to proceed under shaking at room temperature for 24 h. The polymer was capped by reacting it with 25 mg 1-(3-aminopropyl)-4-methylpiperazine in dimethylformamide (0.193 ml) at room temperature for 1 h. The reaction mixture was added to diethyl ether to precipitate the polymer, which was separated by centrifugation at 1,500 RCF for 30 min at 4 °C in a swing-bucket centrifuge. The polymer was dried under argon stream/vacuum overnight and then stored at −20 °C until further use.

The molecular weights of the polymers were evaluated using GPC. Polymers were dissolved in dimethylformamide at a concentration of 5 mg ml⁻¹. Analysis was conducted on an Agilent Technologies 1200 Series GPC equipped with an Agilent 1260 Infinity diode array detector for ultraviolet analysis. The stationary phase was a cross-linked polystyrene–divinylbenzene resin (PLGel Mixed-B) column and dimethylformamide was used as the mobile phase. The molecular weight of both polymers was calculated by the method of moments based on the polystyrene-based calibration. Furthermore, we evaluated the polymers using FTIR spectroscopy on an Agilent Technologies Cary 630 FTIR. Finally, polymers were also analysed using NMR after dissolving them in DMSO-d₆.

To measure polymer degradation in aqueous buffers, polymer 1156 was dissolved in ethanol and diluted in citrate buffer (pH 3) and phosphate-buffered saline (PBS) (pH 7.4). The polymer solution was placed on a shaker incubator at 37 °C. At various times, aliquots were taken, frozen to −80 °C and lyophilized. Dimethylformamide was added to the lyophilized pellet to dissolve the polymer. Undissolved residues were removed by centrifugation, and the supernatant was analysed using GPC as described above.

High-throughput screening of hybrid PBAE polymer library

HOECs (Celprogen) were seeded in a 96-well plate (Celprogen) at a density of 12,000 cells per well for 24–48 h before transfection. For the transfection, a library of polyplexes was prepared as synthesized before³⁷. In brief, polymer–DMSO solutions were diluted in 25 mM sodium acetate buffer (pH 4.2) to a concentration of 2 mg ml⁻¹ polymer. mRNA encoding for eGFP (Trilink) was diluted in the acetate buffer to a concentration of 0.02 mg ml⁻¹. Equal volumes of the polymer and mRNA solutions were mixed by pipetting 10–20 times. The mixture was allowed to stand at room temperature for 10 min to yield the polyplexes. The polyplexes were diluted 10-fold in DMEM serum-free media and added to the cells (mRNA concentration = 1 μg ml⁻¹). After 4–5 h, the treatments were removed, and fresh media were added to the cells. Twenty-four hours later, cells were dislodged using a Tryple dissociation agent, and intracellular fluorescence was measured using flow cytometry (BD FACSAria Fusion SORP).

Synthesis of LNPs

The optimal lipid ratio for local mRNA delivery in the buccal tissue was unknown. We started with the lipid ratio described by Kauffman and colleagues³⁸. Specifically, the molar ratio of DLin-KC2-DMA:DOP E:PEG-PE:cholesterol 35:16:2.5:46 was used (denoted as Formulation A in Supplementary Fig. 3). Three other formulations (B, C and D) were tested, where the molar concentrations of DLin-KC2-DMA and PEG-PE were held constant, while changing the molar ratios of DOPE and cholesterol. All formulations were prepared in an identical manner as follows. The lipids were dissolved in ethanol at 10 mg ml⁻¹. The lipid solutions were mixed to obtain a molar ratio described in Supplementary Fig. 3a. mRNA was dissolved in 10 mM citrate buffer (pH 3). The lipid solution was added to the mRNA solution at a volume ratio of 1:1 and mixed by pipetting 20 times followed by vortexing for 30 s. In this mixture, the weight ratio of DLin-KC2-DMA:mRNA was 10:1. The mixture was placed on ice for 10 min to allow for complexation. The NPs were then used for physicochemical and/or biological characterization. Particle size and zeta potential were measured using a Zeta NanoZS machine (Malvern).

Laurdan assay was used to measure membrane fluidity using protocols described before^39,40. Specifically, various LNPs were synthesized as described above. Laurdan reagent (dissolved in dimethylformamide, 100 μM) was diluted to 1 μM in 10 mM citrate buffer (pH 4.5 and 5.5) and 150 mM PBS (pH 6.5 and 7.4). To this, LNPs were added to achieve a total lipid to a Laurdan molar ratio of ~50:1. The samples were protected from light and incubated at room temperature for 60 min. The fluorescence intensity was measured at excitation/emission of 340 nm/440 nm (I440) and 340 nm/490 nm (I490). Generalized polarization (GP) was calculated using the following formula:

$$\text{GP}=\frac{I440-I490}{I440+I490}$$

Synthesis of polymer–LNPs

We synthesized polymer–LNPs by mixing our lead polymers with the four-lipid combination. Specifically, DLin-KC2-DMA, DOPE, PEG-PE and cholesterol were dissolved in ethanol (10 mg ml⁻¹) and mixed to obtain a molar ratio of 35:46.5:2.5:16 (that is, Formulation D). Polymers 1156 and 844 were dissolved in ethanol at a concentration of 10 mg ml⁻¹. The lipid and polymer solutions were mixed to achieve different weight ratios of the ionizable lipid:polymer. mRNA solution was prepared in a 10 mM citrate buffer (pH 3). The polymer–lipid solution was mixed with the mRNA solution to achieve an ionizable lipid-to-mRNA ratio of 10:1 (w/w). The mixture was pipetted 20 times, vortexed for 30 s and then stored on ice for 10 min to yield the polymer–LNPs. LNPs were prepared using an identical method but without the addition of the polymer.

To measure encapsulation efficiency, LNPs and polymer–LNPs were prepared as described above. To measure unencapsulated mRNA (free RNA), the NP dispersions were incubated with RiboGreen reagent. To measure total mRNA (total RNA), the particles were disrupted by the addition of 1% Triton X in Tris-EDTA buffer for 15 min at 37 °C. The disrupted particle dispersion was incubated with RiboGreen reagent. Fluorescence measurement was performed at excitation/emission of 485 nm/530 nm. Encapsulation efficiency was calculated using the following equation:

$$\text{Encapsulation efficiency}=\frac{\text{total RNA}-\text{free RNA}}{\text{total}}\text{RNA}\times 100$$

In vitro transfection, cell uptake and erythrocyte lysis using polymer–LNPs

Caco-2 cells (ATCC) were seeded in Corning 96-well plates at a seeding density of 20,000 cells per well and allowed to adhere overnight. HOECs (Celprogen) were seeded in a 96-well plate (Celprogen) as described above. Polymer–LNPs or LNPs loaded with mRNA encoding eGFP or fluc were synthesized (mRNA concentration = 0.12 mg ml⁻¹). The NPs were diluted in DMEM media containing 10% fetal bovine serum to obtain an mRNA concentration of 0.125–1 μg ml⁻¹, and added to the cells. To measure fluorescence, 24 h later, the cells were detached and analysed using flow cytometry (BD FACSCelesta flow cytometer). A Zeiss LSM980 laser scanning confocal microscope was used to determine intracellular location of the fluorophore. To measure bioluminescence, 24 h after transfection, cells were treated with Steady-Glo reagent, as prescribed by the supplier.

To measure the cell uptake of the NPs, fluorescently labelled mRNA was used. eGFP encoding mRNA (Trilink) was labelled with Cy5 Label IT reagent according to the manufacturer’s protocol (Mirus Bio). HOECs were incubated with LNP and LNP-1156-20 loaded with Cy5-labelled eGFP mRNA at an mRNA concentration of 1 μg ml⁻¹. Six hours later, the treatments were removed and the cells were washed with ice-cold 1× PBS. Cells were trypsinized and analysed using flow cytometry as described above.

To estimate membrane disruption capability of the NPs, we used an erythrocyte lysis assay as described by Alabi and colleagues⁶⁸. Pig blood was collected in an EDTA tube and centrifuged at 1,500 × g for 10 min to isolate the erythrocytes. The cells were washed twice with 1× PBS. NPs were prepared in citrate buffer and diluted in 1× PBS (pH 5.5). The NPs were incubated with the cells on an incubator shaker at 50 RPM at 37 °C for 1 h. The cells were separated by centrifugation at 1,000 RPM for 5 min (the concentration of the ionizable lipid was 0.15 mg ml⁻¹). Lysis was estimated by measuring absorbance of the supernatant at 540 nm.

In vitro DNA damage and clonogenic cell survival assays

To assess protection from DNA single strand breaks, alkaline comet assays were conducted. The cells were transfected as above. Twenty-four hours after transfection, the cells were exposed to 10 Gy of radiation using ¹³⁷Cs (JL Shepherd & Associates). The plates were immediately placed on ice and then set up for the comet assay (Abcam, ab238544). In brief, HOEC and Caco-2 cells were detached from their plates and resuspended at a concentration of 1 × 10⁵ cells per ml in ice-cold PBS (without Mg²⁺ and Ca²⁺). The cells were then combined with comet agarose at a 1:10 ratio (v/v), mixed and transferred at 75 μl per well onto the top of the comet agarose base layer. The slides were hardened over 15 min and then transferred to lysis buffer and immersed for 60 min. The slides were then transferred to an alkaline solution for 30 min. The slides then underwent electrophoresis for 30 min (1 V cm⁻¹). The slides were washed with deionized water 3 times and then 70% ethanol for 5 min. The slides were allowed to air-dry and then 100 μl per well of diluted Vista Green DNA Dye for 15 min. The slides were reviewed using a confocal microscope using an fluorescein isothiocyanate filter.

To assess protection from DNA double strand breaks, γ-H2AX foci were quantified. The cells were transfected as above. Twenty-four hours after transfection, the cells were exposed to 1 Gy of radiation using ¹³⁷Cs. The 1 Gy dose enabled identification of discrete phosphorylated γ-H2AX foci in each nucleus. The cells were subsequently fixed using 2% PFA for 30 min. The cells were subsequently stained for γ-H2AX foci using an anti-Phospho-Histone H2A.X (Ser139) antibody (Cell Signaling Technology) at a 1:200 ratio. Cells were counterstained with DAPI by mounting solution (Vector Laboratories). CellProfiler 4.2.1 was used to quantify γ-H2AX foci per nucleus.

Cells were initially treated with Dsup mRNA and control mRNA NPs. The cells then underwent cell sorting to identify cells expressing GFP and then exposed to 4 Gy and 10 Gy of radiation. Cells were plated at 300 cells per well for HOEC and 500 cells per well for Caco-2. Two weeks after seeding, colonies were fixed, stained and counted if there were >50 cells present.

To compare the protective effect of Dsup and WR, cells were initially treated with Dsup mRNA and control mRNA NPs. After incubation for 24 h, control mRNA NP cells were treated with 500 μM, 1 mM and 2 mM WR 2 h before collecting. The collected cells then underwent cell sorting to identify cells expressing GFP and then exposed to 4 Gy of radiation. Cells were then plated to a 6-well plate at 300 cells per well for HOEC and 500 cells per well for Caco-2. Two weeks after seeding, colonies were fixed, stained and counted. In addition, a secondary cohort of cells were stained and counted for γ-H2AX foci, as above.

In vivo transfection using polymer–LNPs

All animal studies were approved by the Committee on Animal Care at the Massachusetts Institute of Technology (#0519-023-22) and the Institutional Animal Care and Use Committee at the University of Iowa (#1082416). All animals were exposed to a 12 h light/dark cycle, ambient temperature and humidity, and received food and water ad libitum through the studies.

The goal of these studies was to determine the local transfection efficiency of the NPs in the buccal and colonic tissue. Polymer–LNPs were synthesized as described above. In some cases, the NPs were dialysed against 1× PBS (pH 7.4) for 1 h at 4 °C. The NPs were stored on ice for a maximum of 1–2 h before injection.

C57BL/6 mice were anaesthetized using isoflurane gas. NPs (equivalent to 3 μg fluc mRNA in 50 μl) were injected into the mucosal side of the buccal tissue. A separate cohort of mice were injected with NPs into the mucosal side of the rectal mucosa proximal to the anus. Post-mRNA treatment mice were injected with an indicated volume of luciferin in 1× PBS based on body weight (10 μl of a 15 mg ml⁻¹ luciferin stock solution per gram of body weight). Bioluminescence was measured using an Ami HTX imager (Spectral Instruments Imaging). Bioluminescence was quantified using AMIView Imaging Software, version 1.5.0.

In vivo pharmacokinetics of polymer–LNPs

DLin-KC2-DMA, DOPE, PEG-PE and cholesterol were dissolved in ethanol (10 mg ml⁻¹) and then mixed to obtain a molar ratio of 35:46.5:2.5:16. Polymer 1156 (dissolved in ethanol at 10 mg ml⁻¹) was added to this solution to obtain a DLin-KC2-DMA:polymer 1156 weight ratio of 1:0.2. 1,1-Dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide (DiR) was dissolved in ethanol at 1 mg ml⁻¹ and added to the lipid–polymer mix such that the total lipid:DiR weight ratio was 100:1 (refs. 69,70). Fluc mRNA was diluted in 10 mM citrate buffer (pH 3). The lipid–polymer–dye solution and mRNA solutions were mixed at a 1:1 volume ratio by pipetting 20 times and then vortexing for 30 s. The NPs were placed on ice for 10 min to complete complexation. The NPs were dialysed against 1× PBS (pH 7.4) at 4 °C. The resultant NPs were diluted with an equal volume of 1× PBS and used for in vivo studies.

For buccal injections, 50 μl of the NPs (containing 3 μg mRNA and 0.9 μg DiR dye) was administered into the mucosal side of the cheek and via retro-orbital injection in mice. For rectal injections, 50 μl of the NPs (containing 3 μg mRNA and 0.9 μg DiR dye) was administered into the rectum and via retro-orbital injection in mice. At various times (3 h, 6 h and 24 h) post-injection, mice were treated with 0.2 ml potassium luciferin solution (150 mg ml⁻¹). Mice were euthanized 10 min later via carbon dioxide asphyxiation and cervical dislocation. Organs were collected and imaged using an IVIS imaging system. Fluorescence was monitored at excitation/emission wavelength of 750 nm/780 nm. In subsequent experiments evaluating local distribution of DiR dye-containing LNPs after buccal injection, the mice were euthanized at 15 min and 6 h post-injection. The buccal tissues were resected, frozen with liquid nitrogen and sectioned using a cryostat microtome. The sectioned tissues were stained with DAPI and then imaged on a Leica STELLARIS system.

Western blotting and mass spectrometry to evaluate Dsup expression

For validation of the Dsup protein expression, western blot and mass spectrometry analysis were performed on HOECs transfected with Dsup mRNA NPs. At 24 h after exposure to Dsup mRNA NPs, the cells were collected and subjected to the Chromotek GFP pull-down assay (GFP-Trap agarose kit, gtak-20). Cells were lysed and exposed to agarose beads coated with anti-GFP antibodies. The beads were washed multiple times, and an acidic elution was performed before neutralization. The eluate was then subjected to western blot analysed for anti-GFP antibody and mass spectrometry. For western blot analysis, SDS–PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis) was used to separate proteins, and the separated proteins were transferred onto a PVDF membrane. The membrane was incubated with a rabbit anti-GFP antibody (Cell Signaling Technology, 2956S) at a 1:400 dilution followed by an anti-rabbit antibody (Cell Signaling Technology) conjugated to horseradish peroxidase. The membrane was treated with a chemiluminescent substrate to visualize the enzyme’s activity, allowing for the detection of the protein bands using digital imaging techniques.

For mass spectrometry analysis, the eluate from the GFP pull-down assays was run on SDS–PAGE. The gel was then fixed in EtOH/acetic acid, washed and stained with silver stain (ThermoFisher Scientific) and developed. The gel was imaged using ChemiStation. The specific protein bands were cut from the gel, which was then washed with a 50% acetonitrile (ACN)/50% NH₄HCO₃ solution, using a higher concentration for darker gels, followed by dehydration with ACN. The modification phase involved treating the gel with 1,4-dithiothreitol, then with Iodoacetamide, each under specific conditions. This was followed by a cleaning phase using NH₄HCO₃ and ACN solutions in cycles, concluding with ACN dehydration. Trypsin digestion was performed by applying a diluted trypsin solution to the gel, with initial incubation on ice and then overnight at 37 °C. In the final extraction phase, the samples were treated with ACN in formic acid, undergoing sonication, shaking and centrifugation, with the process repeated to concentrate the supernatant. The concentrated supernatant was then run on a Thermo Orbitrap LUMOS. Data analysis was performed by SEQUEST and interrogated on Reactome.org.

Localization of Dsup protein expression

HOECs were treated with LNP-1156-20 for 24 h. At this time, the cell supernatant was collected, frozen to −80 °C and lyophilized. The cells were lysed using RIPA buffer. The supernatant was reconstituted in a small volume of 1× PBS. The fluorescence of the cell lysate and reconstituted cell lysate were analysed using a fluorescence plate reader at excitation/emission of 450 nm/490 nm.

In vivo DNA damage assays

The right buccal tissue was injected with 50 μl Dsup mRNA NPs, GFP mRNA NPs or no treatment. Using the 220 kVp X-rays generated from the small animal radiation research platform (SARRP, Xstrahl), the right buccal tissue was treated with 10 Gy in 1 fraction with image guidance using the Muriplan software. One hour after irradiation, the mice were euthanized, and the right buccal tissue was excised. The tissue was subsequently formalin fixed and paraffin embedded for staining. Cells were stained for γ-H2AX foci using an anti-Phospho-Histone H2A.X (Ser139) antibody (Cell Signaling Technologies) at a 1:200 ratio. Cells were counterstained with DAPI by mounting solution (Vector Laboratories). CellProfiler 4.2.1 was used to quantify γ-H2AX foci per nucleus.

Immunogenicity analyses

Safety analyses of Dsup mRNA NP administration in buccal tissue were performed by evaluating plasma cytokine levels and hepatocellular injury after one and two doses of Dsup mRNA NPs compared with no treatment. Dsup mRNA NPs (50 μl) were injected into the right buccal tissue. Twenty-four hours after one dose, terminal cardiac punctures were performed in a cohort of mice. A second cohort of mice were subjected to a repeat buccal injection of Dsup mRNA NPs, and terminal cardiac punctures were performed in this cohort of mice. Plasma was collected from the blood by centrifuging blood at 2,000 × g for 10 min. For cytokine analysis, the Proteome Profiler Array Mouse Cytokine Array Panel A was used. The manufacturer’s protocol was used to run the panel array. The membranes were exposed to an X-ray film for 2 min. Data analysis was conducted using Python pandas, numpy, scipy and stats modules. Data were filtered to only include cytokines in which two values above the background rate were present for a given cytokine, indicating a potentially true positive signal. The corrected mean was calculated by subtracting the background value from the mean intensity. The total brightness parameter was calculated by multiplying the corrected mean by the area in which the signal was detected. For evaluation of hepatocellular injury, 300 μl of plasma was analysed on an Idexx Catalyst Dx for AST and ALT values.

Orthotopic syngeneic oral cavity cancer model

Thirty thousand MOC-2 cells (Kerafast) in 30 μl were injected into the right buccal tissue of female C57BL/6J mice. Upon tumour size reaching ~50 mm³, the left buccal tissue was injected with 50 μl Dsup mRNA NPs, GFP mRNA NPs or no treatment. In the cohort of mice treated with radiation using the SARRP, the tumour was contoured, and treatment planned on Muriplan software delivering a single fraction of 10 Gy. Callipers were used to measure tumour size over the course of 16 days, and mice were then euthanized.

To evaluate the specificity of Dsup-GFP expression, we treated orthotopic MOC-2 tumour-bearing mice with Dsup-GFP mRNA NPs in the contralateral cheek. Six hours after treatment, mice were euthanized using an excess dose of carbon dioxide, and the healthy tissue (that is, site of injection) and tumour tissue were collected. Collected tissues and tumours were embedded with OCT and sectioned right away. Slides were checked under a Zeiss LSM980 laser scanning confocal microscope to determine the GFP signal in tissues and tumours.

Reporting summary

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

Extended Data

Extended Data Fig. 1 |. Identification of lipid composition for lipid nanoparticles.

a, Lipid nanoparticles were made using an ionizable lipid (DLin-KC2-DMA), cholesterol, PEG lipid (PEG-PE) and helper lipid (DOPE). Four formulations were produced and the molar ratios of the lipid components in those formulations is listed. b, Particle diameter and polydispersity index was measured using dynamic light scattering. Data is mean ± SD, n = 3. Individual data shown as circles. c, General polarization value was measured using Laurdan assay. Particle diameter and polydispersity index was measured using dynamic light scattering. Data is mean ± SD, n = 3. d, Transfection efficiency of the four formulations in oral epithelial cells was measured in vitro. Data is mean ± SD, n = 4. Individual data shown as circles. P value is measured by Student’s t-test using GraphPad Prism 9.0. (e) C57BL/6 mice were treated with lipid nanoparticles and transfection efficiency was measured using IVIS imaging in vivo. Data is mean ± SEM, n = 4. Individual data shown as circles.

Extended Data Fig. 2 |. Physicochemical characterization of polymer lipid nanoparticles.

a, Particle size and polydispersity index of polymer-lipid nanoparticles measured using dynamic light scattering. Data is represented as mean ± SD, n = 3. Circles show individual data. b, Zeta potential of polymer-lipid nanoparticles measured using dynamic light scattering. Data is represented as mean ± SD, n = 3. Circles show individual data. Transmission electron microscopy images of c, lipid nanoparticles and d, polymer-lipid nanoparticles containing polymer 1156. e, Generalized polarization values for lipid nanoparticles and polymer-lipid nanoparticles in buffers of different pHs. Data is represented as mean + SD, n = 3. P value was calculated by one-way ANOVA and post-hoc Bonferroni using GraphPad Prism 9.0. f, Encapsulation efficiency of mRNA in lipid nanoparticles and polymer-lipid nanoparticles. Data is represented as mean ± SD, n = 3.

Extended Data Fig. 3 |. Comparison of radioprotective efficacy of Dsup-GFP mRNA nanoparticles to WR1065.

a, Human oral epithelial cells and b, human colorectal epithelial cells were treated with Dsup-GFP mRNA nanoparticles or WR-1065 for 4 hours. Cells were then irradiated with 1 Gy of radiation and subsequently fixed and stained for phosphorylated γH2AX. The γH2AX were counted per nucleus. Horizontal lines show mean values. P values were calculated by one-way ANOVA and post-hoc Bonferroni using GraphPad Prism 9.0. c, Human oral epithelial cells and d, human colorectal epithelial cells were treated with Dsup-GFP mRNA nanoparticles or varying concentrations of WR-1065 for 4 hours. Cells were then irradiated with 4 Gy radiation and plated. Number of colonies were counted. Data is represented as mean + SD, n = 3. P values were calculated by one-way ANOVA and post-hoc Bonferroni using GraphPad Prism 9.0.

Extended Data Fig. 4 |. Comparison of radioprotective efficacy of Dsup-GFP mRNA nanoparticles in buccal tissue.

a, Quantification of DNA damage in buccal tissue from radiation therapy (10 Gy in 1 fraction) showing γ-H2AX foci/nucleus is significantly greater in buccal tissue where there is no Dsup-GFP expression (n = 3-4 samples per animal, 5 animals/arm). b, Quantification of DNA damage in buccal tissue from radiation therapy (10 Gy in 1 fraction) showing the number of γ-H2AX foci/nucleus increases as a function of time between Dsup treatment and irradiation (n = 2-3 samples per animal, 5 animals/arm).

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41551-025-01360-5.

Data availability

The main data supporting the results in this study are available within the paper and its Supplementary Information. The raw and analysed datasets generated during the study are available for research purposes from the corresponding authors on reasonable request. Source data are provided with this paper.