Poly(carboxybetaine) lipids enhance mRNA therapeutics efficacy and reduce their immunogenicity

Abstract

Messenger RNA (mRNA) therapeutics are a promising strategy to combat diverse diseases. Traditional lipid nanoparticle (LNP) formulations for mRNA delivery contain poly(ethylene) glycol (PEG), a polymer widely used in drug delivery carriers but that recently has been associated with efficacy and immunogenicity concerns. Here we report poly(carboxybetaine) (PCB) lipids as surrogates for PEG-lipids used in mRNA formulations. In vitro studies with immortalized and primary cells show that PCB-containing LNPs have higher mRNA transfection efficiency than PEG-containing LNPs across different formulations. Moreover, primary cell engineering and in vivo immunization studies in mice further demonstrate greater therapeutic efficacy of PCB-containing LNPs over their PEG counterparts. Mechanistic assays show that this improvement is attributed to enhanced endosomal escape of PCB-containing LNPs. These formulations exhibit a safe immunotoxicity profile and effectively mitigate the accelerated blood clearance effect that has been observed for PEG-containing LNPs, enabling repeated administrations without efficacy loss. Overall, these findings highlight PCB-containing LNPs as a potent and safe mRNA delivery platform for clinical applications.

Successful implementation of messenger RNA (mRNA) COVID-19 vaccines is greatly credited to the development of lipid nanoparticles (LNPs) that protect mRNA from degradation and ensure effective cellular uptake and mRNA endosomal escape^1,2. A typical LNP system has four components, including an ionizable cationic lipid, a phospholipid, cholesterol and a polymer-conjugated lipid. The structure and ratio of each component determine the critical characteristics of the LNPs, such as RNA encapsulation, particle stability and transfection efficiency³. Remarkably, all three FDA-authorized RNA-LNP drugs (Onpattro, Comirnaty and Spikevax) use a poly(ethylene glycol) (PEG)-conjugated lipid (PEG-lipid) in their nanoparticle formulations as the polymer-conjugated lipid^4–6. For years, PEG has been widely used in many approved drugs to shield the immunogenicity of biologics and prolong their blood circulation half-life⁷. In the LNP system, PEG provides a stealth coating to the particles to avoid aggregation during the self-assembly process and storage. When applied to the body, PEG increases the stability of LNP particles in vivo to avoid rapid clearance by reticuloendothelial clearance³. From the perspective of transfection efficiency, however, the steric barrier from PEG interferes with the cellular uptake and subsequent endosomal escape of mRNA, which greatly impairs the performance of mRNA-LNPs. To solve the dilemma, researchers have been looking for a ‘sweet spot’ to achieve a balance between particle stability and transfection efficiency by optimizing the PEG structure and composition⁸. However, recent studies have correlated the unwanted allergic effects of LNP vaccines with PEG-related immunogenicity^7,9. Moreover, the accelerated blood clearance (ABC) effect has been observed in LNP following repeated dosing in mouse studies, leading to the loss of its efficacy due to the induction of anti-PEG antibodies^10,11. Thus, exploring alternatives to PEG-lipids is underway to mitigate potential immunogenicity concerns^9,11–15.

In this work, poly(carboxybetaine) (PCB)-conjugated lipids (PCB-lipids) were synthesized to replace PEG-lipids in the LNP system for mRNA delivery (Fig. 1a). PCB has previously been used in various carriers for gene delivery, including the polyplexes for DNA delivery and lipoplexes for small interfering RNA (siRNA) delivery^16,17. Zwitterionic PCB polymer derived from naturally occurring glycine betaines is known for its superhydrophilicity that endows it with strong stealth properties. Zwitterionic PCB exhibits much stronger hydration properties than PEG due to its electrostatic interactions with water molecules, whereas PEG achieves hydration primarily through hydrogen bonding¹⁸. PCB has been applied in various biomedical applications, such as protein conjugates and nanoparticles, to replace PEG and address the immunogenicity of PEG that can hinder drug efficacy^16,19–21. Here we performed a comprehensive analysis of the structure and molar composition of PCB-lipids in the LNP system for mRNA delivery. Our results demonstrate the versatility of PCB-lipids across various LNP systems for efficient mRNA delivery, highlighting their potential as an alternative for PEG-lipids. Significantly, PCB-containing LNPs (PCB-LNPs) achieve higher transfection efficiency with minimal toxicity than LNP formulations with PEG (PEG-LNPs). A series of mechanistic studies were conducted to elucidate the mechanisms underlying the enhanced efficiency of PCB-LNPs, including cryogenic electron microscopy (cryo-EM), biological small-angle X-ray scattering (BioSAXS) and model membrane studies. The optimized PCB-LNPs were used to express functional proteins including Cas9 for gene editing and the model antigen ovalbumin to highlight its prospective applications in primary cell engineering and vaccines, respectively. In the setting of repeated systemic administrations of LNPs, PCB-LNPs maintain their transfection efficiency even after multiple doses. We showed that PCB-LNPs mitigated the ABC effect found in PEG-LNPs due to the induction of anti-PEG antibodies and rapid clearance of the injected particles.

Fig. 1 |. Synthesis and in vitro evaluations of PCB-containing LNPs.

a, Schematic of this work, where PEG-lipid was fully replaced by PCB-lipid in the LNP-containing ionizable cationic lipid, phospholipid, cholesterol and mRNA cargo. b, Reversible addition–fragmentation chain transfer polymerization (RAFT) was used to synthesize different variants of PCB-lipids with different average carboxybetaine repeat units (m = ~5, m = ~15 or m = ~27) and different acyl chain lengths (n = 12 or n = 16). c, Luciferase expression was measured to test the transfection efficiency of LNPs with different polymer lipids listed in b at different molar percentages. Transfections were performed in THP-1 and HeLa cell lines using MC3 and SM102 formulations, respectively. Transfection efficiency quantified as relative luminescence unit (RLU) is shown as the mean of three biological replicates in the heat map. d, Seven different cell lines were tested for the transfection efficiency of LNPs formulated with DMG-PEG 2K (PEG) or M2 containing FLuc-mRNA. Comparison between these two polymer lipids was conducted in three LNP systems including MC3, SM102 and ALC0315. Data from the CKKE12 system are shown in Supplementary Fig. 2. The relative luminescence signal from the transfection is shown in a facet grid plot, where the RLU has been normalized to the PEG group. In the box plot, the upper and lower box edges extend from the first to third quartile, the centre line represents the median and the whiskers are extended to the maxima and minima. Experiments were performed in three biological replicates. A two-tailed unpaired t-test was performed for statistical analysis (the exact P values are displayed in the facet grid). Panel a created with BioRender.com.

Synthesis and in vitro evaluation of the library of PCB-LNPs

PCB-lipids were synthesized through reversible addition–fragmentation chain transfer polymerization by a ‘graft-from’ method to generate a library of lipids with different molecular weights of polymer (2 kDa, 4 kDa and 7 kDa) and two acyl chains with different lengths (DMG: 1,2-dimyristoyl-rac-glycero-; DSG: 1,2-distearoyl-rac-glycero-; Fig. 1b). In our initial experiments, PCB-lipids were used to replace PEG-lipids in two commercial LNP formulations, respectively. SM102 formulations are used in Moderna’s mRNA-1273 COVID-19 vaccines, whereas MC3 formulations are used in Alnylam’s siRNA therapeutics, Patisiran³. The molar ratios of the ionizable lipid, 1,2-distearoyl-sn-glycero-3-p hosphocholine, cholesterol and polymer lipid were set as 50:10:40–x:x, where x = 1.5, 2.5 and 5, respectively. Encapsulation of mRNA into LNP was prepared by rapidly mixing lipid components in ethanol and mRNA in an acidic buffer following a pipette mixing protocol according to a previous study²². All formulations were characterized by dynamic light scattering for size measurement and by an encapsulation assay for RNA entrapment (Supplementary Tables 1 and 2). RNA encapsulation is mostly maintained above 90% and does not change with the type and amounts of polymer lipids. Then, the efficiency of the formulations was evaluated by transfecting the mRNA encoding firefly luciferase (FLuc) to two different cell lines: HeLa and THP-1, where the mRNA expression was analysed using a luciferase assay (Fig. 1c and Supplementary Fig. 1). Among all the formulations, M2 demonstrated the highest transfection efficiency and surpassed its PEG counterparts. We then compared the luciferase expression from M2 with that from PEG across seven cell lines (Fig. 1d). Three LNP systems studied in this experiment have been FDA approved, which contain ionizable lipids including SM102, MC3 and ALC0315. As a result, M2 consistently outperformed PEG in all formulations studied above. In addition, we tested the performance of M2 in CKKE12 systems, which have been frequently cited in the literature for their potent mRNA transfection in the liver (Supplementary Fig. 2)²³. Results highlighted the superior transfection efficiency for M2 over its PEG counterpart.

Although the luciferase assay produces sensitive results in a high-throughput manner, detecting the expression of enhanced green fluorescent protein (eGFP) allows us to analyse the transfection efficiency at the single-cell level, providing details on transfection percentage. Here we compare the expression of eGFP from SM102 LNP with M2 and PEG at a polymer molar ratio of 1.5 in a human monocyte cell line, THP-1 cells (Fig. 2a–c), and a human T cell line, Jurkat cells (Fig. 2d–f). Both SM102-M2 and SM102-PEG achieved a high transfection percentage in THP-1 cells, as shown in the flow cytometric analysis. However, the mean fluorescence intensity (MFI) of the eGFP signal from SM102-M2 is significantly higher than that from SM102-PEG, showing that more protein was translated from the mRNA delivered by SM102-M2. Similarly, SM102-M2 also has more efficient transfection in Jurkat cells. The MFI of the eGFP signal from SM102-M2 increases as the dosage of LNP increases, whereas SM102-PEG remains at a similar level of MFI at all dosages. The trend is generally consistent in other LNP systems formulated with MC3 and ALC0315 (Fig. 2). Additionally, we corroborate this finding of better efficiency by SM102-M2 in the 293T cells and by CKKE12-M2 in the THP-1 cells (Supplementary Fig. 3). From the flow cytometric analysis, we did not detect a significant viability loss in these formulations (Supplementary Fig. 4). These results highlight the efficiency benefit of replacing PEG with PCB in the LNP for mRNA delivery.

Fig. 2 |. PCB-LNPs show higher transfection efficiency than PEG-LNPs across different formulations.

a–f, In vitro eGFP expression of PCB-LNP and PEG-LNP in THP-1 (a–c) and Jurkat (d–f) cells. Comparison between these two polymer lipids was conducted in different LNP systems including MC3, SM102 and ALC0315. Transfection was performed in three biological replicates (n = 3) at a dosage of 25, 50 and 100 ng RNA per well. The expression of eGFP was analysed by a fluorescence microscope (Cytation 7) after 24-h transfection of the LNP to the indicated cell lines. Representative fluorescence images (a and d) and flow cytometric histograms (b and e) shown here were selected from one of the replicates at 100 ng RNA per well. The MFI of the eGFP was measured from transfection in THP-1 (c) and Jurkat (f) cells at the indicated dosages. Data are shown as mean ± s.e.m. For statistical analysis, a two-tailed unpaired t-test was performed (ns, not significant; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Enhanced efficiency of PCB-LNPs over PEG-LNPs

Efficient protein expression at a low dosage of LNPs is of significance in therapeutic applications. After showing that M2 formulations are better than their PEG counterparts using reporter genes, we chose SM102-M2 as an example to further evaluate its performance in different applications such as primary cell engineering and vaccines. We tested their translatability in primary cells. Primary cells are much harder to transfect than immortal cell lines. Therefore, an efficient mRNA carrier for primary cell engineering is consistently sought after²⁴. Among primary T cells, CAR-T is one of the most popular applications of ex vivo gene delivery. The success of the FDA-approved CAR-T therapy for B cell lymphoma has opened a new era for immune cell engineering^25,26. To test the feasibility of the ex vivo applications of SM102-M2 for CAR-T cell engineering, we first transfected Jurkat T cells with LNP encapsulating anti-CD19 CAR-mRNA and evaluated CAR expression using flow cytometry (Fig. 3a). A fluorescently labelled CD19 protein was used to detect the expression of CAR from SM102-M2 and SM102-PEG, respectively (Fig. 3b and Supplementary Fig. 6). As a result, SM102-M2 yielded more than 95% CAR⁺ Jurkat cells, which is twofold more than SM102-PEG. Additionally, SM102-M2 exhibited higher MFI values indicating the high expression of CAR, whereas SM102-PEG resulted in moderate CAR expression (Fig. 3c,d). No significant viability loss was detected in either group compared with media-only treated Jurkat cells (Supplementary Fig. 6). We also observed the inhibition of cell growth in an assay co-culturing CAR-T cells with Raji B cells, suggesting that the expression of CAR from this formulation is functional (Supplementary Fig. 7)²⁷. Following CAR-T cells, we then delivered mRNA encoding tdTomato fluorescent proteins to primary human T cells isolated from peripheral blood mononuclear cells (hPBMCs). Fluorescence images showed that SM102-M2-transfected cells have much stronger tdTomato expression than SM102-PEG after transfection with LNPs for 24 h (Fig. 3e and Supplementary Fig. 8). Flow cytometric analysis revealed significantly higher MFI from SM102-M2 at various dosages (Fig. 3f). Furthermore, we also observed that SM102-M2 showed more potent Cre recombination in primary splenic cells extracted from Ai14 mice, where more CD11b⁺ cells were edited and expressed tdTomato efficiently after transfection with Cre for 16 h ex vivo (Supplementary Figs. 9–11). Finally, the ex vivo engineering of primary neutrophils is another actively pursued application, which requires an efficient gene carrier for the generation of CAR-neutrophils²⁸. We compared eGFP-mRNA transfection by SM102-PEG and SM102-M2 in primary neutrophils extracted from the bone marrow of the mouse. Results show that SM102-M2 displayed a higher transfection efficiency with minimal cytotoxicity (Supplementary Fig. 12).

Fig. 3 |. PCB-LNPs outperform PEG-LNPs in various therapeutic applications.

a–l, Comparison of SM102-PEG and SM102-M2 in various applications, including the production of CAR-T (a–d), ex vivo primary cell transfections (e and f) and gene editing on primary T cells (g–i), as well as vaccines (j–l). a, Anti-hCD19 CAR-mRNA was delivered to the Jurkat T cells to create CAR-T cells. Transfection was performed at 100-ng RNA per 2 × 10⁴ cells for 24 h. b, Transfection efficiency was analysed using flow cytometry to detect surface expression of CAR19. c,d, Percentage (c) and MFI (d) of CAR⁺ T cells were compared with SM102-M2 and SM102-PEG. Measurements were performed in biological replicates (n = 3). e, Representative fluorescence images of human primary T cells transfected with LNP to express tdTomato proteins. f, MFI of the tdTomato-expressing cells was obtained from the flow cytometric analysis to compare the transfection efficiency in primary cells with SM102-M2 and SM102-PEG. Measurements were performed in biological replicates (n = 3). g, Cas9-mRNA and sgRNA were co-delivered by LNPs to human primary T cells for 48 h to induce TCR knockout. h, A TIDE assay was used to evaluate the editing rate. The red dashed box highlights the altered chromatogram peaks, indicating successful editing events. i, Editing rate of TCR knockout was compared with SM102-M2 and SM102-PEG in three biological replicates (n = 3). j, LNPs encapsulating OVA-mRNA were injected intramuscularly (5 μg per mouse) on day 0 and day 5. Splenocytes were extracted on day 12 for intracellular stimulation cytokine assay. k,l, Representative flow cytometric images (k) and statistical analysis (l) demonstrate IFN-γ% from CD3⁺CD8a⁺ T cells on OVA₂₅₇ peptide stimulation. Experiments were conducted in biological replicates (n = 5). Data are shown as mean ± s.e.m. Statistical analysis was performed using a two-tailed unpaired t-test (*P < 0.05, **P < 0.01, ****P < 0.0001). Panels a, g and j created with BioRender.com.

In addition to efficient expression, genetic manipulation is a key aspect of primary cell engineering for gain-of-function or loss-of-function applications. For CAR-T, disrupting the expression of endogenous T cell molecules, such as T cell receptor (TCR) and PD-1, is crucial for enhancing the efficacy of allogeneic CAR-T products²⁹. Genome engineering using the CRISPR-Cas9 system has emerged as a useful tool for inducing targeted gene alterations and consequently manipulating the protein expression. As a proof of concept, we co-delivered the Cas9-mRNA and single guide RNA (sgRNA) that targets the genomic locus encoding TCR α and β chains to human primary T cells and compared the knockout efficiencies between SM102-PEG and SM102-M2 (Fig. 3g). The efficiency of the TCR-targeted gene disruption was evaluated by a tracking of indels by decomposition (TIDE) assay on genomic DNA extracted from cells treated with LNPs for 48 h. As shown in Fig. 3h, Sanger sequencing data revealed multiple peaks surrounding the sgRNA target regions, verifying an alteration in the genomic reading frame downstream of these sites. For SM102-M2, the editing rate increased with higher dosages, whereas SM102-PEG remained at a low level (Fig. 3i). Cell viability was not changed even at a higher dosage of SM102-M2, indicating its potential for ex vivo applications at elevated concentrations.

The success of mRNA-based COVID-19 vaccines has inspired researchers to extend the application of this technology to other vaccines, such as those for cancer immunotherapy³⁰. However, these applications present greater challenges than COVID-19 vaccines, as they often require higher doses, placing significant demands on the transfection efficiency, which plays a critical role in vaccine efficacy. Here we chose OVA as a model antigen and vaccinated C57BL/J mice intramuscularly with 5 μg per mouse OVA-encoding RNA to compare the induction of antigen-specific immune responses by SM102-PEG and SM102-M2 (Supplementary Fig. 13). Briefly, mice were given a prime shot on day 0 and a booster shot on day 5. Then, splenocytes were collected on day 12 to analyse the IFN-γ expression using intracellular stimulation assays (Fig. 3j and Supplementary Fig. 14). The mRNA sequence contained an OVA₂₅₇ peptide and an OVA₃₂₃ peptide in tandem, linked by a GGGS flexible linker. The sequence was further flanked by a secretion peptide and MHC class I trafficking signalling peptide for enhanced antigen presentation (Fig. 3j)³¹. For flow cytometric analysis, CD3⁺CD8a⁺ T cells were gated to evaluate the response against the OVA₂₅₇ peptide, and CD3⁺CD4⁺CD44⁺ T cells were gated to evaluate the response against the OVA₃₂₃ peptide (Supplementary Fig. 14). As a result, we observed significant increases in CD3⁺CD8a⁺ T cells expressing IFN-γ in the spleen of SM102-M2-treated mice than the SM102-PEG-treated ones after the stimulation by OVA₂₅₇ peptide (Fig. 3k,l). For stimulation by OVA₃₂₃ peptide, the frequency of IFN-γ-secreting CD3⁺CD4⁺CD44⁺ T cells in SM102-M2 is comparable with that in SM102-PEG (Supplementary Fig. 15a). Negligible amounts of IL-5-expressing cells were detected in both formulations, indicating that our formulations induced a Th1-skewed but not a Th2-skewed response (Supplementary Fig. 15b).

Enhanced endosomal escape of PCB-LNPs

In the various models discussed above, we have demonstrated superior mRNA delivery capabilities of PCB-LNP. To understand the structural characteristics responsible for its high transfection efficiency, we next characterized SM102 LNPs formulated with PEG and PCB using BioSAXS and cryo-EM analyses. Several studies have used BioSAXS to reveal the internal structure of the LNP^15,32–35. Figure 4a shows the BioSAXS profile of SM102 LNP formulated with PEG and M2 at a molar percentage of 1.5. BioSAXS profiles of other PCB-lipids at different molar ratios are also shown (Supplementary Fig. 16). Both SM102-PEG and SM102-M2 show a structural peak at around q = 0.12 Å⁻¹, consistent with the data reported from other studies^32,34,35. This peak signifies the internal ordering of LNPs, which is usually described as a single lamellar layer with an electron-dense core formed by inverted micelles containing oligonucleotides, or the so-called current bun organization^6,15,33,36. These structural peaks were analysed using Lorentzian functions to obtain the peak position, peak width and peak area (Supplementary Fig. 17b). A slight shift in the peak position and width was observed as a function of the molar percentage of polymer lipids in all samples (Supplementary Methods). To gain more information on the LNP structures, we performed cryo-EM on SM102-M2 and SM102-PEG (Fig. 4b). SM102-M2 exhibits more prominent ‘bleb’ structures than more compact SM102-PEG. It has been reported that a higher level of bleb structures leads to enhanced mRNA transfection properties³⁷, which could partly explain the improved potency by SM102-M2, as shown above.

Fig. 4 |. Mechanistic assays show an enhanced endosomal escape efficiency by PCB-LNPs.

a, BioSAXS spectra of SM102-PEG (red) and SM102-M2 (blue) encapsulating a Fluc-mRNA. Radius of gyration (R_g) was calculated from a low-q range, and the peak position in the higher-q range of each formulation was also listed. b, Representative cryo-EM images of SM102-PEG and SM102-M2 in PBS (pH 7.4). Scale bar, 100 nm. c, Haemolysis assay to evaluate membrane disruption by SM102-PEG and SM102-M2 at pH 5.5 and pH 7.4, respectively. Measurements were done in biological replicates (n = 5). d, A FRET assay was performed by the co-incubation of a fluorescence-labelled endosome-mimicking liposome (EndoLipo) with LNPs. Data in c and d are shown as mean ± s.e.m. Measurements were done in biological replicates (n = 3). For statistical analysis, a two-tailed unpaired t-test was performed (***P < 0.001, ****P < 0.0001). e, pKa values of SM102-PEG and SM102-M2 were measured by a TNS assay. The dashed line indicates half of the maximal fluorescence signal and defines the apparent pKa of the formulation. Data are shown as mean ± standard deviation. Panels c and d created with BioRender.com.

Endosomal escape of mRNA is one of the most important barriers that dictates the transfection efficiency of LNPs. We speculated that the hydrophilicity of PCB and the structural characteristic observed above could lead to better endosomal escape efficiency of SM102-M2. We first compared the endosomolytic activity of SM102-PEG and SM102-M2 using a haemolysis assay, which is a surrogate model that measures the haemoglobin release after the incubation of LNPs with red blood cells (RBCs). The endosomolytic activity at physiological pH (7.4) indicates the cytotoxicity of nanoparticles, whereas lysis at 5.5 serves as a model for the capability of LNPs to escape endosomal compartments. As a result, SM102-M2 exhibited significantly higher haemolytic activity than SM102-PEG at various concentrations (Fig. 4c). Both formulations have undetectable RBC lysis at physiological pH. Besides, we also assessed their membrane disruptive activity using a fluorescence resonance energy transfer (FRET) assay. In this assay, the change in fluorescence signal at E_m = 590 nm was detected after the incubation of endosome-mimicking liposome (EM-lipo) and LNP samples, and the intensity indicates the relative extent of lipid fusion (Fig. 4d and Supplementary Fig. 18). As a result, SM102-M2 exhibits more than twofold stronger lipid fusion than SM102-PEG. The result echoes the structural observation of LNPs and implies a rationale that SM102-M2 has better transfection efficiency due to improved endosomal escape. The electrostatic interaction between LNPs and inner endosomal lipids is essential to drive the process of mRNA endosomal escape. However, neutral PEG-lipid in the LNP system does not provide electrostatic interaction as well as creates steric hindrance due to its nature of amphiphilicity. By contrast, the superhydrophilicity of PCB facilitates the interactions, thereby enhancing mRNA endosomal escape. In addition, the pKa value of SM102-M2 is slightly higher than SM102-PEG, which further implies that the M2 formulation is more sensitive to the change in pH in solution and, consequently, results in easier endosomal fusion (Fig. 4e and Supplementary Fig. 19). Furthermore, we found that M2 confers improved fusogenicity to the LNP without the induction of particle aggregation that could compromise the shelf life. At 4 °C, SM102-M2 exhibited a similar profile of size change as SM102-PEG (Supplementary Fig. 20). In addition, SM102-M2 displayed similar efficiency retention as SM102-PEG following the freezing process (Supplementary Fig. 21).

Safe immunogenicity profile of PCB-LNPs

Safety profile and immunogenicity are critical in the translation of a new LNP formulation. First, we evaluated the liver toxicity of the PEG-LNPs and PCB-LNPs by measuring two clinical chemistry markers, including alanine aminotransferase (ALT) and aspartate aminotransferase (AST)^22,38,39. We compared the AST and ALT levels in serum collected from mice administered systemically with SM102-PEG, SM102-M2, ALC0315-PEG and ALC0315-M2. As a result, most formulations show no significant increase in the ALT or AST levels compared with the phosphate-buffered saline (PBS)-treated group, except that ALC0315-PEG exhibited a slightly higher AST level (Fig. 5a). This result indicates a minimal level of liver toxicity from PCB-LNPs. Although any immunological findings from animal studies can be helpful for our understanding of how the immune system responds to LNPs, the FDA recommends that in vitro evaluation in human cells should be conducted to predict any potential toxicities in the clinic⁴⁰. Therefore, we incubated hPBMCs with RNA-LNPs for 24 h and analysed the cytokine levels in the supernatant, including IL-1β, IL-6, TNF-α, IL-2, IFN- γ, IFN-α, IL-4, IL-5, IL-10, GM-CSF and IL-8. Figure 5b shows the heat map of cytokine secretion from hPBMC after incubation with LNP with or without mRNA. As a result, we detected minimal rises in indicators of inflammation of all the samples in the absence of RNA in LNPs (Supplementary Fig. 22). In the presence of mRNA in the LNP, key proinflammatory cytokines, including IL-1β, IL-6, TNF-α and IFN-γ, were elevated in SM102-PEG at high doses, showing higher levels than those observed with SM102-M2 (ref. 41).

Fig. 5 |. PCB-LNPs exhibit a safe immunogenicity profile.

a, Liver toxicity in C57BL6/J mice was assessed after three weekly systemic injections of PEG-LNPs or PCB-LNPs at an RNA dosage of 10 μg per mouse. Data are shown as mean ± s.e.m (biological replicates, n = 3). A two-way analysis of variance test with Dunnett’s post hoc analysis was performed to compare each group with the control (treated with PBS). b, Multiplex assay was used to assess the release of cytokines from hPBMC after incubating 24 h with SM102-PEG and SM102-M2 with or without mRNA cargo at various dosages as indicated. Data are shown as the mean of three biological replicates in a heat map. c, Schematic illustrating the injection schedule of repeated administration studies of PEG-LNPs and PCB-LNPs. C57BL/6J mice at the age of 6–8 weeks were intravenously injected with 5 μg (RNA) per mouse. d–i, Comparisons were performed in the SM102 formulation (biological replicates, n = 8; d–f) and ALC0315 formulation (biological replicates, n = 3; g–i). d,g, Representative IVIS images of the extracted liver from each group are displayed. e,h, Total flux of the luminescence signal of the extracted liver was measured. f,i, Anti-polymer IgM was detected by incubating the serum with polymer-coated beads, followed by probing with fluorescently labelled secondary antibodies and analysis using flow cytometry. Serum was obtained on day 7 after the fourth injection. The level of anti-polymer IgM (IgM⁺) in the serum was quantified as a percentage of the fluorescence signal, with gating thresholds set using the serum from untreated mice. Data are shown as mean ± s.e.m. For statistical analysis, a two-tailed unpaired t-test was performed (*P < 0.05; ns, not significant). Panels b and c created with BioRender.com.

The induction of anti-PEG antibodies has been widely reported, which leads to efficacy loss after repeated administrations of PEG-LNPs^10,11,42. In this work, we compared the delivery efficiency of PEG-LNPs and PCB-LNPs in a multiple injection scenario. In brief, mice in group 1 received a single injection of FLuc-mRNA. On the other hand, mice in group 2 received four weekly injections of eGFP-mRNA followed by a fifth injection of FLuc-mRNA. FLuc expression in group 2 was measured via IVIS and compared with group 1 to determine any efficacy loss resulting from repeated administrations (Fig. 5c). By using eGFP-mRNA instead of FLuc-mRNA for the first four injections, we can prevent the induction of protein-specific immunogenicity that might inhibit protein expression in the final injection. Consequently, the reduced efficacy observed in the last injection is probably due to ABC effects caused by the immunogenicity of the polymer lipid or other lipid components. The efficacy of LNP delivery was evaluated by measuring the bioluminescence signal from the liver—the primary organ in which most mRNA expression occurs in systemic delivery (Supplementary Fig. 23). As a result, a decrease in the FLuc expression between group 1 and group 2 in SM102-PEG was observed (P = 0.0236), showing a reduction in the LNP transfection efficacy. By contrast, no significant drop was found between the two groups in SM102-M2, suggesting that the delivery efficacy remains after repeated administrations (P = 0.4810; Fig. 5d,e). Furthermore, serum from animals in group 2 was collected the day before the final injection of FLuc-mRNA to measure anti-polymer IgM levels. Consistent with the discovery in reported studies^10,11,42, repeated administrations of PEG-LNPs induce the generation of anti-PEG antibodies. However, mice in SM102-M2 were found with significantly lower polymer-specific antibodies (P = 0.0125). To verify the result in an alternative system, we also compared PCB and PEG in the ALC0315 formulation, which is used in Pfizer–BioNTech COVID-19 vaccines (Fig. 5g–i). After repeated administrations, no ABC effect was observed in ALC0315-M2, but ALC0315-PEG showed significant efficacy loss (P = 0.0397) as well as the induction of anti-PEG IgM (P = 0.0267). Collectively, results from both SM102 and ALC0315 formulations underscore a marked reduction in the antibody-mediated ABC effect by PCB-LNPs, strongly suggesting enhanced efficacy in repeated dosing scenarios.

Outlook

In this work, we demonstrate that PCB-LNPs have a higher transfection efficiency than PEG-LNPs across various formulations. Mechanistic studies reveal that this improvement is due to enhanced endosomal escape of PCB-LNPs. We attribute this effect to the differences in hydrophilicity between PCB and PEG. We hypothesize that the hydrophilic PCB in PCB-LNPs enables the proximity between the protonated ionizable cationic lipids and the anionic lipids on the endosomal membrane. It facilitates the initiation of electrostatic interactions of these polar groups that cause the endosomal fusion and drive the phase transition of lipids towards the hexagonal phase, resulting in membrane disruption and mRNA release^43,44. On the contrary, the amphiphilic PEG of PEG-lipid creates steric hindrance, impeding these interactions and, consequently, requiring higher energy to achieve endosomal disruption. Future studies on molecular simulations of these processes at the molecular level could provide deeper insights and guide the rational design of LNPs with enhanced endosomal escape efficiency. Another critical finding of our work is that PCB-LNPs have a safer immunotoxicity profile, enabling repeated intravenous administrations in mice without efficacy loss. Repeated administration of PEG-LNPs has been shown to induce anti-PEG antibodies, which, in turn, trigger the ABC effect that diminishes the efficacy of LNPs in subsequent administrations¹⁰. Zwitterionic polymer PCB has shown to be a low-immunogenic biomaterial due to its non-fouling properties and biocompatibility^18,19,21,45. This is attributed to the superhydrophilicity of the polymer, which introduces an additional energy barrier for epitope recognition, making it harder for polymer-specific antibodies to develop. Overall, we introduced PCB-lipids as a replacement for PEG counterparts in LNPs, enabling enhanced mRNA transfection and repeated administration for various clinical applications.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41563-025-02240-8.

Methods

Chemical synthesis of PCB-lipid

Chemical and reagents.

The synthesis of 3-acrylamido-N-(2-(tert-butoxy)-2-oxoethyl)-N,N-dimethylpropan-1-aminium was previously reported by our previous work⁴⁵. Also, tert-butyl N-(2,3-dihydroxypropyl)carbamate and azobisisobutyronitrile were purchased from Asta Tech and were used as received. Stearoyl chloride was purchased from TCI America and was used as received. All the other chemicals were purchased from Sigma-Aldrich and were used as received. Slide-A-Lyzer Dialysis Cassette (2K MWCO) were purchased from Thermo Fisher Scientific.

All the other chemicals were purchased from Sigma-Aldrich and were used as received.

Synthesis of PCB-lipid.

Details of the synthesis steps and structural validation are provided in Supplementary Information. In brief, PCB-lipids were synthesized using the graft-from method, where the PCB polymers grew from a chain transfer agent bearing either C14 or C18 lipid tails. The molecular weight was controlled by adjusting the monomer-to-lipidated chain transfer agent ratio (Supplementary Fig. 21 and Supplementary Table 3). Detailed ¹H nuclear magnetic resonance and ¹³C nuclear magnetic resonance characterizations of the intermediates and final polymer lipids are provided in Supplementary Figs. 25–38.

Preparation of LNPs

The lipids 1,2-distearoyl-sn-glycero-3-phosphocholine, 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 and methoxypolye thyleneglycoloxy(2000)-N,N-ditetradecylacetamide were purchased from Avanti Polar Lipids. Cholesterol was purchased from Sigma. For the ionizable lipids, DLin-MC3-DMA (MC3) was purchased from Organix. SM102 and ALC0315 were purchased from Cayman Chemical, and CKKE12 was purchased from Echelon Biosciences. Other components include 1,2-dioleoyl-sn-glycero-3-phospho-l-serine (DOPS; Avanti Polar Lipids) and cholesterol (Sigma).

Encapsulation of mRNA into LNPs were prepared by mixing lipid components in ethanol and mRNA in an acidic buffer at a ratio of 1:3 (v/v). Pipette mixing was used in formulation screening following the protocol described elsewhere²². LNPs used in the in vivo experiments were prepared in a microfluidic mixing device called NanoAssembr Ignite (Precision NanoSystems). mRNA encoding FLuc was purchased from Trilink Biotech and dissolved in 20-mM citrate buffer (pH 3) for MC3 and CKKE12 formulations, or 20-mM acetate buffer (pH 5) for SM102 and ALC0315 formulations. To remove residual organic solvents, the resulting LNP was washed with PBS in a dialysis cassette (Slide-A-Lyzer cassettes, molecular weight cut-off, 10 kDa; Thermo Fisher). For cryo-EM and BioSAXS studies, LNPs at high concentrations were used. In this case, a 10-kDa Amicon centrifugal filter was used to concentrate the LNPs in PBS to acquire a good resolution for cryo-EM images. If LNPs need to be stored frozen, a storage buffer containing 30% sucrose in Tris buffer (pH 7.4) was added to the final product. LNPs were characterized by measuring their size in NanoBrook Omni using dynamic light scattering. To quantify the amount of mRNA in the sample, an encapsulation assay was performed using a QuantiFluor assay (Promega).

Synthesis of mRNA by in vitro transcription

The sequence encoding eGFP was cloned from a commercial vector (Addgene, category no. 13031). The sequence encoding FMC was adapted from a clinical study⁴⁶ and synthesized by Twist Bioscience. The sequence of OVA peptide is shown in Fig. 3e and synthesized by Twist Bioscience. The Cas9-mRNA was purchased from Trilink Biotechnologies (category no. L-7206). The mRNA was prepared by in vitro transcription from a linearized plasmid pTNT4.0 (constructed in-house) containing a modified T7 promoter⁴⁷, UTR sequences used in BNT162b vaccines⁴⁸ and a poly A tail containing 120 repeating units. Briefly, linearization of the plasmids was done by an overnight enzymatic digestion reaction at XmaI, which is 3′ downstream of the poly A tail. Agarose gel electrophoresis was performed to confirm the successful linearization of plasmid templates. Next, the linearized template was used to generate an mRNA transcript using the IVTpro mRNA Synthesis System. (3′-OMe-m7G)(5′)ppp(5′)(2′-OMeA)pG was added to the reaction for 5′ capping, and N1-methyl-pseudo-UTP was added to replace the wild-type UTP. In addition, RNAse Inhibitor (NEB, no. M0314L) was added to a final concentration of 4 U μl⁻¹. The reaction was performed at 37 °C for 2 h in a thermocycler (Bio-Rad, C1000) and then treated with DNase I at 37 °C for 1 h to remove the template. Final products were purified by Monarch RNA Cleanup Columns (NEB). The concentration and purity of the DNA templates and mRNA product were analysed in a Take3 plate using a Cytation 7 microplate reader (BioTek). The integrity of the mRNA was confirmed by agarose gel electrophoresis or Bioanalyzer 5200 (Agilent) available at the Biotechnology Resource Center at Cornell University.

Cell lines and in vitro transfection

The following cell lines were cultured in Eagle’s minimum essential medium (ATCC no. 30-2003) supplemented with 10% foetal bovine serum (FBS): HeLa (ATCC, CCL-2) and HepG2 (ATCC, HB-8065). The following cell lines were cultured in Dulbecco’s modified Eagle’s medium (ATCC no. 30-2002) supplemented with 10% FBS: 293T (ATCC, CRL-3216), C2C12 (ATCC, CRL-1772), NIH/3T3 (ATCC, CRL-1658) and Raw264.7 (ATCC, TIB-71). The following cell lines were cultured in RPMI 1640 (ATCC no. 30-2001) supplemented with 10% FBS and 55 μM of 2-mercaptoethanol: DC2.4 (MilliporeSigma SCC142), Jurkat (ATCC, TIB-152) and THP-1 (ATCC, TIB-202). All the cell lines were maintained at 37 °C with 5% CO₂ in a humidified environment. Cells were seeded in a black 96-well plate (CellAh, Stellar Scientific) the night before transfection. Before transfection, the LNP was first premixed with the corresponding complete media so that the RNA dosage is 100 ng (or the indicated amount) per well. To transfect the cell with LNPs, the premixed LNP-media solutions were aliquoted to each replicate so that the final volume of each one is 100 μl per well. Transfections were performed with three biological replicates. After 24 h of incubation, the culture medium containing the LNP was carefully removed. For luciferase expression, 30 μl of passive lysis (Promega, category no. E1501) was added to each well for cell lysis. Luciferase expression was measured using a luciferase assay (Promega, category no. E1501) following the manufacturer’s protocol. For cells transfected with eGFP-encoded mRNA, the expression of GFP was visualized and imaged in Cytation 7. Also, cells were stained with live/dead dye and fixed with 4% paraformaldehyde solution before proceeding for flow cytometric analysis. For cells transfected with CAR-encoded mRNA, the expression of CAR was analysed by flow cytometry. Briefly, cells were first stained with live/dead dye and then stained with PE-human CD19 (AcroBiosystems) diluted (1:50) in a fluorescence-activated cell sorting buffer. The flow cytometric analysis was performed in Thermo Fisher Attune NxT at Cornell Biotechnology Center or in Agilent NovoCyte Quanteon 4025 as indicated in the experiment.

Transfection of primary human T cells

LNP transfections in primary T cells.

Human primary T cells were isolated from the hPBMC cyrostocks (STEMCELL) using a positive selection kit (Miltenyi Biotec, category no. 130-096-535) and a magnetic sort device (Miltenyi Biotec, category no. 130-042-401). Isolated T cells were verified by staining the surface marker CD3 (BioLegend, category no. 300405). Viable T cells were first seeded with 1 × 10⁶ cells ml⁻¹ in R10 media (RPMI 1640 + 10% FBS + 2-mercaptoethanol + 10 ng ml⁻¹ hIL2). Cells were pre-activated with CD3/CD28 T cell activator (STEMCELL, category no. 10971) for 24 h before LNP transfections. Cell clustering indicates successful activations. Then, activated T cells were seeded at 1 × 10⁶ cells ml⁻¹ and 0.25 ml per well in a 48-well plate. The indicated amounts of LNP were added and incubated for 24 h for tdTomato expression or 48 h for gene editing. Fluorescence images of tdTomato expression were taken in Cytation 7, and flow cytometric analysis was performed in NovoCyte Quanteon 4025. For CRISPR-Cas9 editing of TCR knockout, Cas9-mRNA produced in-house described above was co-delivered with sgRNA (Synthego) at a weight ratio of 1:1. The target sequence of sgRNA (AGAGTCTCTCAGCTGGTACA) was adapted from a previous study²⁹.

TIDE assay to analyse CRISPR-Cas9 editing on human primary T cells.

Genomic DNA was extracted from the samples following the manufacture’s protocol (NEB, category no. T3010S). Then, 100 ng of genomic DNA was subjected to polymerase chain reaction (PCR) to amplify the DNA fragments around the sgRNA target site. Q5 Hot-Start HF polymerase (NEB, category no. M0494S) was used and primers were adopted from a previous study²⁹. PCR products were purified and proceeded for Sanger sequencing (Genewiz, Azenta). The primers for PCR were adapted from a previous study²⁹: (TRAC-F)-TGAGTCCCAGTCCATCACG and (TRAC-R)-TTGGACTTTTCCCAGCTGACAGA. The primer for Sanger sequencing is CGAAGGCACCAAAGCTGC. The raw chromatogram file was analysed by an online tool at http://tide.nki.nl. DNA fragments from a FLuc(mRNA)-transfected sample were used as a reference sequence. The indel size range for analysis was set to 10 bp, and the decomposition window was adjusted to cover the maximum sequence with a high-quality signal. P-value threshold of decomposition was 0.001 by default.

BioSAXS

BioSAXS experiments were performed at the ID7A1 station at Cornell High Energy Synchrotron Source. The measurements were conducted at an X-ray energy of 11.2577 keV in a 0.25 × 0.25 mm² (h × v) aperture. The SAXS data were recorded by the EIGER X4M detector in the angular range between 0.006 Å⁻¹ and 0.5 Å⁻¹. The distance between the sample and the detector was set to 1,717 mm. For each measurement, 30–40 μl of the LNP sample was resuspended in 1× PBS at a lipid concentration of 2.8 mg ml⁻¹ and loaded into a 1.5-mm quartz capillary. During the measurement, samples were constantly oscillated to minimize damage due to overexposure to radiation. Buffer (1× PBS) was first measured for background subtraction in data analysis. Data analysis including baseline reduction and model fitting was performed using RAW software⁴⁹. Data presentation was done on GraphPad Prism v. 8.

Cryo-EM

mRNA-LNPs were prepared as described above and washed with Dulbecco’s PBS. mRNA-LNPs were then concentrated to 8 mg ml⁻¹ (lipid concentration) in a 100-kDa ultracentrifugation filter (MilliporeSigma). For cryo-EM, 2.5 μl of concentrated LNPs in Dulbecco’s PBS were applied to glow-discharged holey grids (Au-Flat 1.2/1.3) mounted in an EM GP2 automatic plunge freezer (Leica). Temperature and humidity were set to 6 °C and 85%, respectively, and the blot time was 6 s. Grids were loaded into a Gatan 914 cryo-stage and transferred into a JEOL JEM-2100F transmission electron microscope located at the State University of New York College for Environmental Science and Forestry Analytical & Technical Services facility. Micrographs were recorded at an electron optical magnification of ×40,000 on a Gatan OneView 4k complementary metal–oxide–semiconductor camera. Imaging was performed twice with similar results. Multiple images were acquired across different areas of each grid and the representative images are shown here.

Lipid fusion analysis by FRET assay

The fusogenicity of LNP was determined by a FRET assay. Rho-PE (Lissamine rhodamine B 1,2-dihexadecanoyl-sn-glycero-3-p hosphoethanolamine, triethylammonium salt) and NBD-PE (N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2-dihexadecanoyl-sn-glyce ro-3-phosphoethanolamine, triethylammonium salt) were purchased from Thermo Fisher Scientific. The FRET probes conjugated to DHPE, NBD-PE and Rho-PE were integrated into the same endosomal-like nanoparticle, resulting in an increase in rhodamine fluorescence due to FRET from NBD. The fusion of these FRET-probe-containing endosomal-like nanoparticles would lead to a decrease in the rhodamine signal as the distance between the two probes expanded. EM-lipos were prepared by mixing DOPS:DOPC:DOPE:NBD-PE:Rho-PE in a molar ratio of 25:25:48:1:1 in chloroform. This mixture was concentrated by rotary evaporation and high-vacuum drying to yield a thin lipid film. This dried film was later hydrated in PBS (pH 7.4), followed by 20 min each of vortexing and sonication. The total lipid concentration was kept constant at 1 mM. PBS (pH 5.5) was added to black 96-well plates (100 μl per well), followed by the addition of 2 μl of EM-lipos (1 mM) and 2 μl of LNPs to each well. Following a 5-min incubation at 37 °C, the fluorescence (F) of Rho-PE was measured using a BioTek Cytation 7 Cell Imaging Multimode Reader at an excitation wavelength of 470 nm and an emission wavelength of 590 nm. Negative control (F₀) consisted solely of EM-lipos in PBS with pH 5.5. As positive control (F_min), EM-lipos were incubated with Triton X-100 solutions (2 wt%). The percentage lipid fusion was then calculated as Lipid fusion(%) $\left(\%\right)=\frac{F_0-F}{F-F_{\text{min}}}\times 100\%$.

Lipid fusion analysis by haemolysis assay

A haemolysis assay was conducted by the co-incubation of mouse RBCs and LNPs at defined pH values to characterize the cytolytic property and endosomal escape efficiency of LNPs. Briefly, mouse RBCs were collected in heparin-coated tubes and washed five times with cold PBS (pH 7.4) by centrifugation at 10,000g for 5 min. RBCs were pelleted and resuspended in PBS (pH 7.4) or 20-mM citrate buffer saline (pH 5.5) to a final concentration of 4% (volume of blood/volume of buffer). Then, the RBC solutions were added to a V-bottom-shaped 96-well plate (100 μl per well) and mixed with 100 μl per well of LNPs at the indicated amounts of LNPs. The plate was incubated at 37 °C for an hour and centrifuged at 1,000g at 4 °C for 5 min to obtain the supernatants. The RBC solution was incubated with Triton X-100 (1 wt%) as a positive control and was incubated with buffer alone as a negative control. The haemoglobin substances in the supernatant were measured at 440 nm and calculated using the following formula:

$$\text{Haemolysis}\text{ratio}\left(\%\right)=\frac{\left({\text{OD}}_{\text{sample}}-{\text{OD}}_{\text{nagative}\text{control}}\right)}{\left({\text{OD}}_{\text{positive}\text{control}}-{\text{OD}}_{\text{nagative}\text{control}}\right)}\times 100\%.$$

TNS assay for measuring pKa of LNPs

A 6-(p-toluidino)-2-naphthalenesulfonic acid (TNS) assay for measuring the pKa value of the LNP has been reported previously⁵⁰. Briefly, buffers at different pH values (3, 4, 5, 6, 6.2, 6.4, 6.6, 6.8, 7, 7.4, 8 and 9) were prepared from a master buffer containing 20 mM of sodium phosphate, 25 mM of citrate, 20 mM of ammonium acetate and 150 mM of sodium chloride. In a black 96-well plate, 1 μl of LNP samples at a concentration of 40 μg ml⁻¹ were added to the 80 μl of TNS solution (6.5 μM) for all the pH values. The measurement was done in replicates. After mixing, the plate was then read using a Cytation 7 plate reader (E_x = 321 nm and E_m = 435 nm). Relative fluorescence was normalized by the equation $\frac{I_x-I_{\text{min}}}{\left(I_{\text{max}}-I_{\text{min}}\right)}$. A pKa curve was created by setting the relative fluorescence as the y axis and pH value as the x axis. The data were fit nonlinearly by a four-parameter logistic sigmoidal model. The apparent pKa of the LNPs was defined by the log of the inflection point of the curve.

Intracellular cytokine staining for immunization experiment

Male C57BL/6J mice were intramuscularly administered with LNPs encapsulating OVA-mRNA on day 0 and day 5. On day 12, splenocytes were collected from the vaccinated mice following a similar procedure described in the ‘Transfection of primary human T cells’ section. A total of 1 × 10⁶ cells per well were plated in a round-bottomed 96-well plate in RPMI 1640 media supplemented with 10% FBS, pen–strep antibiotic and 55 μM of 2-mercaptoethanol. Cells were stimulated with OVA257 and OVA323 peptides (Genscript) at 2 μg ml⁻¹ for 6 h in the presence of a protein transport inhibitor cocktail (eBioscience, category no. 00-4980-03). After stimulation, cells were first washed with 1× PBS and stained with a live/dead viability dye. Then, cells were washed by a fluorescence-activated cell sorting buffer and stained with an FcR blocking reagent followed by antibodies against cell surface markers. Antibodies used for this study include CD3e-FITC (Miltenyi Biotec, category no. 130-118-958), CD19-eF450 (Invitrogen, category no. 48-0193-80), CD4-PerCPCy5.5 (BioLegend, category no. 100434), CD8a-PECy7 (BioLegend, category no. 100722), CD44-AF700 (Invitrogen, category no. 56-0441-82) and CD62L-SB600 (Invitrogen, category no. 63-0629-42). After staining the surface markers, cells were fixed and permeabilized using BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD, category no. 554714). Cells were then washed in perm/wash solution and stained for intracellular cytokines, including IFNγ-PE (BioLegend, category no. 505808) and IL-5-PE (BioLegend, category no. 504304). After all the staining, cells were resuspended in a fluorescence-activated cell sorting buffer for flow cytometric analysis. Analysis was performed in the Agilent NovoCyte Quanteon 4025. Data were processed using NovoExpress v1.6.1. Background cytokine expression was captured by samples vaccinated with LNPs encapsulating an irrelevant mRNA.

Liver toxicity study

Male C57BL/6 mice at the age of 6–8 weeks were used in the animal experiments. Mice in a group of three were intravenously injected with mRNA-LNPs for each formulation at a dosage of 1 mg kg⁻¹. Mouse sera were collected at 6 h and 24 h after the last injection. Levels of liver enzyme ALT were determined using a Colorimetric Activity Assay kit from Sigma-Aldrich and the AST levels were determined using an Fluorescent Activity Assay kit from Abcam.

Detection of cytokine release from hPBMC using Luminex assays

Human PBMCs were purchased from STEMCELL and sourced from three different donors. PBMCs were carefully thawed and then resuspended in RPMI 1640 media supplemented with 10% FBS, 5-mM HEPES, 2-mM pen–strep and 2-mM glutamine, and rested for 2 h at 37 °C and 5% CO₂. Then, 1 × 10⁶ viable PBMCs per well in a 48-well plate were added with LNPs at the indicated dosages to a final media volume of 0.5 ml. Media alone were used as an untreated control. Supernatant was then collected and stored at −80 °C until analysis. Cytokine levels released from PBMCs were measured using a Human Magnetic Luminex Assay (Bio-Techne, category no. LXSAHM). Detection was performed in the Magpix system at the Human Nutritional Chemistry Service Laboratory at Cornell University. Analytics measured in this study include GM-CSF, IFN-α, IFN-γ, TNF-α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-8 and IL-10.

Repeated administration studies

RNA used in this experiment was purified using Monarch RNA Cleanup Kits followed by cellulose chromatography to remove residual double-stranded RNA (dsRNA). dsRNA is a well-known by-product of in vitro transcription that can activate TLR7/8 and is sometimes utilized as an adjuvant to enhance the immunogenicity of formulations^51,52. The procedure for this was adopted from a previous study⁵². A dsRNA-specific ELISA kit was used to detect the dsRNA content in the RNA samples (Vazyme, category no. DD3509). C57BL/6J mice at the age of 6–8 weeks were used in this study. RNA-LNPs were intravenously injected weekly to the animal at an RNA dosage of 5 μg per mouse. On the day before the last injection, mouse serum was collected by chin bleeding to detect the presence of anti-polymer IgM, and the details of the detection method are described in Supplementary Information. At 24 h after the injection of FLuc-mRNA, mice were injected intraperitoneally with 150 mg kg⁻¹ of d-luciferin and sacrificed after 10–12 min. Organs were rapidly extracted and imaged using an IVIS imaging system (PerkinElmer).

Animal studies

Mice were housed at Cornell University’s CARE facility. All the experiments adhered to Cornell’s Institutional Animal Care and Use Committee protocol (nos. 2020-0037 and 2023-0101).

Data analysis and schematics in figures

All the bar plots and heat maps were generated and all the statistical analyses were performed using Prism 8.0 (GraphPad). A two-tailed unpaired t-test was used when comparing two groups, whereas a one-way analysis of variance with Tukey’s correction was performed when comparing two groups or more than two groups. Difference was considered to be significant if P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001, unless otherwise indicated). The facet grid plots showing the luciferase expression data (Fig. 1d and Supplementary Fig. 2) were generated using R with the packages ggplot2, patchwork, gsgh4x and rstatix.

Reporting summary

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

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41563-025-02240-8.

Data availability

All data in this study are available within the Article and its Supplementary Information. The raw and analysed datasets generated during the study are available for research purposes from the corresponding author on request. Source data are provided with this paper.