High-density brush-shaped polymer lipids reduce anti-PEG antibody binding for repeated administration of mRNA therapeutics

Abstract

Messenger RNA lipid-nanoparticle-based therapies represent an emerging class of medicines for a variety of applications. However, anti-poly(ethylene glycol) (anti-PEG) antibodies generated by widely used PEGylated medicines and lipid nanoparticles hinder therapeutic efficacy upon repeated dosing. Here we report the chemical design, synthesis and optimization of high-density brush-shaped polymer lipids that reduce anti-PEG antibody binding to improve protein production consistency in repeated dosing. Brush-shaped polymer lipid parameters, including side chain length, degree of polymerization, anchor alkyl length and surface regimes on lipid nanoparticles modulate anti-PEG antibody binding affinity and control their blood c ir cu la tion p ha rm ac ok in etics. Compared to widely used 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene g l y co l - 2000, lipid nanoparticles containing brush-shaped polymer lipids generate superior therapeutic outcomes in protein replacement therapy and genome editing models, reformulating structure–activity guidelines for the design of PEG lipid substitutes. Overall, these findings contribute to the general effort in the development of lipid nanoparticles with low immunogenicity to overcome current roadblocks to nucleic acid medicines.

Messenger RNA (mRNA) therapy holds immense potential to treat various liver conditions caused by genetic mutations, chronic infections, drug and alcohol abuse and serious injury that lead to more than two million deaths worldwide per year^1,2. Genetic metabolic diseases are particularly well-suited for mRNA-based protein replacement therapy and genome editing since the liver functions as an essential metabolic organ by regulating chemical and protein levels in the blood^3,4. Various viral and synthetic delivery systems have been developed for the delivery of nucleic acids and gene editors^5,6. Viral vectors can accumulate in the liver and transfect cells with high efficacy^7–12. However, the wide existence of neutralizing antibodies in humans prevents repeated dose efficacy for viral vectors^13–16. Since life-long liver diseases require continuous medication, attention has shifted to non-viral strategies that could potentially overcome immune recognition and antibody production.

Lipid nanoparticles (LNPs) represent the most promising synthetic delivery system class due to their clinical use in small interfering RNA (siRNA) liver therapy and multiple mRNA COVID-19 vaccines^17–21. Synthetic LNPs are expected to improve safety and redosing feasibility considerations over viral vectors. However, recent reports suggest that the generation of anti-PEG antibodies (APAs) after receiving PEGylated therapeutics, including COVID-19 vaccines, limits efficacy when re-administered²². Anti-PEG IgG was boosted 13.1-fold after Spikevax vaccination when gold standard 1,2-dimyristoyl-rac-glycero-3-m ethoxypolyethylene glycol-2000 (DMG-PEG2000) is used as a PEG lipid²². Meanwhile, specific APA-related phagocytosis was enhanced, leading to accelerated blood clearance of re-administered PEGylated drugs that weakens therapy. Thus, LNP neutralization by existing APAs in the serum raises concerns for LNP-mediated protein replacement therapy and genome editing that require multiple administrations.

To address this issue, we designed, chemically synthesized and optimized a systematic series of polymer lipids as substitutes for currently used DMG-PEG2000, generating LNPs with reduced APA reactogenicity and improved repeated dose efficacy (Fig. 1a). We thus introduce here a new chemical space for LNP optimization, revealing that the polymer lipid component is a promising LNP optimization element. Atom transfer radical polymerization (ATRP) was employed to generate a polymer lipid library with controlled architecture, chemical functionality and molecular weight using lipid-functionalized initiators and various monomers containing linear, brush, hydrophilic, hydrophobic, positively charged, neutral and zwitterionic features. From the above polymer lipid library, brush-shaped poly(ethylene glycol) methyl ether methacrylate (PEGMA) lipids (BPLs) exhibited superior luciferase (Luc) mRNA transfection in vivo and were selected to further examine three critical chemical design parameters: side chain length (x), degree of polymerization (DP, y) and alkyl length (n) of the initiator. Structure–activity relationships emerged whereby all three factors affected mRNA delivery efficacy. All BPL LNPs exhibited reduced APA binding affinity compared to standard DMG-PEG2000 LNPs owing to the advantage of the comb structure and mushroom polymer regimes that hindered protein adsorption and antibody binding. All these features resulted in superior BPL LNPs with slower clearance and higher mRNA delivery efficacy upon repeated dosing compared to DMG-PEG2000 LNPs. Moreover, BPL LNPs overcome the APA inhibitory effect induced by DMG-PEG2000 LNPs, rescue immunogenicity-stalled protein replacement therapy with extended survival in fumarylacetoacetate hydrolase knock-out (FAH^−/−) mice and achieve efficient in vivo CRISPR–Cas9 genome editing. Overall, an expanded chemical scope of LNP stabilizing polymer lipids revealed valuable insights into polymer physicochemical properties and architecture that improved repeated dose maintenance of high human protein expression and therapeutic outcomes.

Fig. 1 |. Synthetic polymer-lipid-incorporated LNPs mediate efficient mRNA delivery.

a, Schematic illustration of how high-density BPLs reduce APA binding to enable mRNA LNP repeated dose efficacy. b, Synthetic route to access polymer lipids by ATRP using lipid–Br species as initiators. c, In vivo Luc mRNA delivery by polymer lipid LNPs (0.1 mg kg⁻¹ Luc mRNA, i.v., 6 h). d, Quantitative data of Luc activity in each organ (from c) after delivery of Luc mRNA mediated by polymer lipid LNPs. bpy, 2,2′-bipyridyl.

Polymer chemical structure modulates mRNA delivery efficacy

To examine whether synthetic polymers with alternative chemical structures and physical properties could function as substitutes for the currently used DMG-PEG2000 in LNPs, we designed a chemically diverse series of polymer lipids containing two lipid tails, an ester linker and polymer chains derived from various radically polymerizable monomers (Fig. 1b and Supplementary Information). Luc mRNA was used to quantify the delivery efficacy of LNPs, where Luc protein is produced after successful cellular uptake, endosomal escape and mRNA translation. Nearly all polymer-lipid-stabilized LNPs were able to deliver mRNA in vitro, but to varying degrees (Supplementary Fig. 1). However, when transitioning to in vivo delivery, a much lower hit rate (mean luminescence of over 1 × 10⁵ photons s⁻¹ cm⁻² sr⁻¹) of 40% was observed (Fig. 1c,d). PEG lipids protect LNPs from clearance by evading mononuclear phagocyte system recognition during systemic circulation. Thus, LNPs without DMG-PEG2000 showed no signals in the main organs. Interestingly, polyHEA and polyHEMA (that is, presence or absence of a methyl substituent in the backbone) and polyHPMA (varied numbers of methylene units in the side chain), which have slight differences in hydrophilicity and are considered as effective PEG substitutes in many reports, also failed to transfect tissues in vivo. This may be due to the additional extracellular and intracellular barriers faced by LNPs versus PEGylated drug conjugate products. Encouragingly, LNPs prepared with polyDEAEMA, polyAEMA, polyNIPAM and polyDMAM containing positive charges transfected the liver and/or spleen, and LNPs prepared with the zwitterionic polymer polySBMA also transfected the liver and spleen with high efficacy. Among all the polymer lipids examined in the first library, polyEG₉MA, a brush-shaped polymer with ethylene glycol (EG) side chains, exhibited the highest in vivo transfection efficacy and was selected for further studies in an expanded second library.

Polymer architectures and mRNA delivery efficacy

Advanced synthetic techniques, including ATRP, can be used to precisely control individual parameters hypothesized to modulate LNP stability and efficacy. Specifically, the brush-shaped architecture of lipid-functionalized polyEGMAs provided versatile handles to carefully examine three factors that affect the polymer chemical structure: the side chain length (x), the DP of the polymer backbone (y) and the alkyl length of the initiator (n), as demonstrated in Fig. 1a.

We first investigated structure–activity relationships by careful control of polymer architectures using EGMA as a base structure. DMG-PEG_xMA species (x = 1, 2, 3, 5, 9, 19) were obtained that shared an identical lipid initiator DMG and similar DP, but varying side chain length (x; Fig. 2a). BPL LNPs were formulated with variable synthetic BPLs and constant amounts of 4A3-SC7, DOPE, cholesterol and mRNA during the formulation. An incremental increase in the side chain length resulted in increased hydrophilicity that assisted in the formation and stability of the LNPs (Fig. 2b and Supplementary Table 2). LNP size sharply decreased when the side chain length reached 5 and stabilized at approximately 120 nm, which was similar to DMG-PEG2000 and suitable for in vivo studies. Cryogenic transmission electron microscopy (cryo-TEM) images (Supplementary Fig. 2) indicated similar structures between BPL LNPs and DMG-PEG2000 LNPs. BPL LNPs were stable, enabled high mRNA encapsulation efficiency (>90%) and possessed low cytotoxicity (Supplementary Figs. 3a and 4a,b). BPL LNPs with the longest side chain length (for example, the DMG-PEG₁₉MA₃₆ BPL) produced the highest luminescence signal and human erythropoietin (hEPO) level, demonstrating that increased EG side chain length benefitted mRNA delivery efficacy (Fig. 2c–e and Supplementary Fig. 4c).

Fig. 2 |. Synthetic chemistry controls polymer architectures and mRNA delivery efficacy.

a–e, EG side chain length (x)-dependent mRNA delivery efficacy. a, BPL chemical structure and schematic illustration with different EG side chain lengths. b, Size and polydispersity index (PDI) of BPL LNPs measured using dynamic light scattering. Data are presented as mean ± s.d. (n = 3 biologically independent samples). c, In vivo Luc mRNA delivery (0.1 mg kg⁻¹). d, Quantitative data of Luc activity in livers after Luc mRNA delivery. Data are presented as mean ± s.d. (n = 3 biologically independent samples). e, Serum hEPO level quantified by enzyme-linked immunosorbent assay (ELISA) after hEPO mRNA delivery (0.3 mg kg⁻¹). Data are presented as mean ± s.d. (n = 3 biologically independent samples). f–j, DP (y)-dependent mRNA delivery efficacy. f, BPL chemical structure and schematic illustration with different DP values. g, Size and PDI of BPL LNPs. Data are presented as mean ± s.d. (n = 3 biologically independent samples). h, In vivo Luc mRNA delivery (0.1 mg kg⁻¹). i, Quantitative data of mean Luc activity in livers after Luc mRNA delivery. Data are presented as mean ± s.d. (n = 3 biologically independent samples). j, Serum hEPO level quantified by ELISA after hEPO mRNA delivery (0.3 mg kg⁻¹). Data are presented as mean ± s.d. (n = 3 biologically independent samples). k–o, Alkyl length (n)-dependent mRNA delivery efficacy. k, BPL chemical structure and schematic illustration with different alkyl lengths. l, Size and PDI of BPL LNPs. Data are presented as mean ± s.d. (n = 3 biologically independent samples). m, In vivo Luc mRNA delivery (0.1 mg kg⁻¹). n, Quantitative data of mean Luc activity in livers after Luc mRNA delivery. Data are presented as mean ± s.d. (n = 3 biologically independent samples). o, Serum hEPO level quantified by ELISA after hEPO mRNA delivery (0.3 mg kg⁻¹). Data are presented as mean ± s.d. (n = 3 biologically independent samples). DOG, 1,3-dioctanoyl glycero-; DLG, 1,3-dilauroyl glycero-; DMG, 1,3-dimyristoyl glycero-; DPG, 1,3-dipalmitoyl glycero-; DSG, 1,3-distearoyl glycero-.

Since polymer length can affect the hydrophilicity and flexibility of extended chains, we next carefully examined the effect of backbone DP on mRNA delivery using DMG-PEG₁₉MA_y (y = 5, 10, 21, 36, 60, 77; Fig. 2f). Since polymer lipids extend from the surface of LNPs in aqueous solvents, BPLs with longer DPs led to a slight increase in the LNP sizes (Fig. 2g, Supplementary Fig. 3b and Supplementary Table 2). Luc and hEPO transfection efficacy increased with increasing DP, peaking at a DP of 36 and declining with a further increase in DP (Fig. 2h–j and Supplementary Fig. 4d–f). A similar result was also found in the DMG-PEG₉MA_y polymer series with different DPs (Supplementary Fig. 5), further confirming that an intermediate polymer length (y = 36) benefits mRNA transfection. This finding may be related to both the hydrophilicity and the shielding effect of PEGMA polymers. When the DP is low (for example, DMG-PEG₁₉MA₅), the polymer chain is too short for stealth effects. Conversely, when the DP is high (for example, DMG-PEG₁₉MA₇₇), the long polymer may inhibit cellular uptake and hinder the release of mRNA, leading to a decrease in delivery efficacy²³.

Polymer lipids are incorporated into LNPs through hydrophobic interactions between hydrophobic anchors of the polymer lipid and LNP lipid components²⁴. Tuning of the lipid anchor chemistry alters the hydrophobicity and shedding rate of polymer lipids from LNPs^17,25. Therefore, lipid anchors with varying alkyl length (n = 4, 8, 10, 12, 14) were appended as ATRP initiators to synthesize BPLs with a fixed side chain length (x = 9 or 19) and backbone DP (y ≈ 36; Fig. 2k). BPLs with short anchors, for example, 1,3-dioctanoyl glycero (DOG)-terminated polymers (n = 4), exhibited larger LNP size and poorer stability because the weaker hydrophobic interactions led to a loose LNP structure (Fig. 2l, Supplementary Fig. 3c and Supplementary Table 2). In addition, DOG-terminated polymers more easily desorb from LNPs, thereby allowing faster clearance by the mononuclear phagocyte system and weaker transfection (Fig. 2m–o and Supplementary Fig. 6). Meanwhile, less effective mRNA delivery was also observed for LNPs containing 1,3-distearoyl glycero (DSG)-terminated polymers (n = 14). The strong hydrophobic interaction hinders desorption from LNPs to expose the LNP membrane for the binding of serum proteins, leading to low mRNA delivery efficacy by DSG polymers²⁴. Therefore, intermediate alkyl chain lengths (n = 10, 12) possess appropriate hydrophobic interactions to shed from LNPs and achieve high mRNA transfection efficacy in vivo (Supplementary Fig. 7).

Brush and mushroom regimes alter delivery outcomes

Having established the optimal alkyl length anchor (n), we further focused on the polymer regimes enabled on the LNP surfaces with respect to side chain length (x) and polymer backbone DP (y), as these chemical differences govern the resulting polymer architectures and density on the surface (Supplementary Information). As this is a brush-shaped polymer, the Flory radius (R_F) values of the backbone and side chain are calculated individually. The R_F and distance between polymer molecules (D) values of different EG side chain lengths are presented in Fig. 3a and Supplementary Table 3, indicating that the side chains are in the brush regime. For the polymer backbone, R_F values are presented in Fig. 3a and Supplementary Table 4. The molar ratio of polymer lipids was fixed at 4.76%, resulting in a D value around 29 Å (ref. 26). Thus, a mushroom to brush transition occurred from a DP of 36 to a DP of 60. A representative schematic illustration of the mushroom and brush regimes of brush-shaped polymers on LNP surfaces is presented in Fig. 3b. Mushroom regimes cover more surface area than brush regimes at the same density of polymer lipids. The high surface coverage may reduce the exposure of the hydrophobic domain of LNPs and weaken the protein adsorption driven by hydrophobic interactions. Moreover, the reduced surface space of mushroom regimes inhibits protein insertion, repels protein adsorption and maximizes the steric effect of the polymer layer²⁷. Therefore, polymer lipids in mushroom regimes are beneficial for prolonging the circulation by impeding protein-adsorption-mediated blood clearance, such as ApoE, opsonins and APAs.

Fig. 3 |. Polymerization parameters and polymer regimes control the pharmacokinetic profile and APA binding.

a, R_F values of BPLs and their regimes on the surface of LNPs. b, Representative schematic illustrations of mushroom and brush regimes. c–h, Plasma cholesteryl methyl ether (CME) levels quantified by GC-MS in the mice injected with BPL LNPs with different EG side chain lengths (c and d), DPs (e and f) and alkyl lengths (g and h). PK, pharmacokinetic. The area under curve values were found at 48 h. Data are presented as mean ± s.d. and statistical significance was analysed by one-way analysis of variance (ANOVA) multiple comparisons with DMG-PEG2000: ***P < 0.001; **P < 0.01; *P < 0.05 (n = 3 biologically independent samples). i, Schematic illustration of biolayer interferometry assay for the measurement of binding affinity between LNPs and APA. Left: a typical binding kinetic experiment. Right: association and dissociation between the LNPs and the APA immobilized biosensor. j–o, Equilibrium dissociation constant (K_D) values and representative association and dissociation phase curves of APA and BPLs with different EG side chain lengths (j and k), DPs (l and m) and alkyl lengths (n and o), indicating a chemical-structure-dependent association rate behaviour with eligible dissociation. A lower K_D value indicates stronger binding. In the biolayer interferometry experiment, the immobilized APA concentration was 100 nM. Data are presented as mean ± s.d. (n = 3 biologically independent samples).

BPL LNPs increase blood circulation and reduce APA binding

PEG lipids stabilize LNPs and function to prevent biofouling, reduce systemic clearance and prolong the blood circulation time in vivo. All BPL LNPs exhibited less protein adsorption than DMG-PEG2000 LNPs owing to the comb structure (Supplementary Fig. 8), indicating the superior antifouling ability for longer blood circulation. To examine the differences in pharmacokinetic profiles of LNPs composed of DMG-PEG2000 and different BPLs, we used an assay designed to avoid artificial contributions of reporter dyes or metals²⁸. We prepared LNPs with cholesterol fully substituted by cholesteryl methyl ether, a metabolic inactive sterol species, and measured plasma cholesteryl methyl ether level by gas chromatography/mass spectrometry (GC-MS) to profile the pharmacokinetics of LNPs²⁹. Notably, all BPL LNPs showed longer circulation than linear DMG-PEG2000 LNPs, indicating the advantage of the comb structure in reducing biofouling. Longer side chain length BPLs (DMG-PEG_xMA, x = 9, 19) exhibited longer circulation times than shorter counterparts (x = 5; Fig. 3c,d), in accordance with the finding that linear PEG with a longer PEG length showed a prolonged half-life^30,31. However, increased polymer backbone DP (y) led to decreased circulation time (Fig. 3e,f), which may be related to the mushroom to brush regime transition as y increases. Evidently, the interspace between neighbouring polymers on the LNP surface is more spacious for blood protein adsorption and phagocytic cell uptake in the brush regimes, leading to the rapid clearance of BPL LNPs with longer polymer backbone DPs (DMG-PEG₁₉MA_y, y = 60, 77). In the mushroom regimes, transition from an interdigitated mushroom to a mushroom occurred with the increased polymer length. DMG-PEG₁₉MA₅ polymer was more condensed to a random coil model, while DMG-PEG₁₉MA₃₆ lost some conformational freedom and tended to behave as if it was between the mushroom and brush regimes. Thus, more room for protein adsorption to DMG-PEG₁₉MA₃₆ LNPs exists than to DMG-PEG₁₉MA₅ LNPs, leading to faster clearance. The difference in alkyl length (n) also plays a role in the protein-mediated clearance, where LNPs composed of BPLs with longer alkyl tails (n = 12, 14) exhibited an increased circulation time compared with that of shorter counterparts (n = 8, 10) (Fig. 3g,h).

PEGylation is widely used in clinical medicines to improve the bioavailability of PEGylated therapeutics³² and is limited by the generation of APAs^33–37. To measure the APA binding affinity, a biolayer interferometry assay was employed (Fig. 3i). Standard, well-established DMG-PEG2000 containing LNPs exhibited a strong binding affinity with the lowest equilibrium dissociation constant (K_D) among all the LNPs tested, indicating that linear PEG is more likely to bind antibodies than brush-shaped PEG polymers. For BPLs, only the side chain length (x) and polymer backbone DP (y) affected K_D. Weaker APA binding correlated with shorter side chain length, as evidenced by DMG-PEG₃MA₃₆ with the highest K_D and DMG-PEG₁₉MA₃₆ with the lowest K_D (Fig. 3j,k). Mechanistically, APA binds to approximately three linear EG monomer subunits of a PEG chain. Thus, increased EG side chain length increased the possibility of APA binding and thus resulted in a stronger binding affinity with decreased K_D. On the other hand, K_D decreased with increased backbone DP (y), indicating that the binding affinity was enhanced with longer polymer length due to the increased EG amount (Fig. 3l,m). A sharp decrease in K_D was observed when y increased from 36 to 60, which is exactly the predicted range of the DP for the mushroom to brush conformation transition. The brush regime allows the antibody to squeeze between the chains and bind to EG segments in the side chain, leading to the drastic decrease in K_D. No significant difference was found in different alkyl length (n) groups (Fig. 3n,o), indicating that these effects are dominated by the polymer chemistry and not by the shedding of PEG lipid from LNP. Overall, BPLs reduced APA binding affinity compared to linear PEG even at higher EG amounts. In brush-shaped polymers, the increased EG side chain length (x) and polymer backbone DP (y) both increased the number of EG units that can bind APAs, leading to a stronger binding affinity. In addition, polymer regimes also affect the antibody binding affinity associated with the surface coverage of LNPs.

LNP performance in repeated dose regimens

Having validated that polymer lipid architectures and regimes lead to a reduction in APA binding affinity, we next examined whether this would contribute to consistent in vivo mRNA translation upon repeated dosing. The design to study the correlation of polymer lipid chemistry and surface configurations with repeated dose efficacy is illustrated in Fig. 4a. Mice were intravenously (i.v.) administered with LNPs on day 1, followed by administration of the same LNPs again on day 30. Due to the strong binding affinity of DMG-PEG2000 with APAs, the second dose of DMG-PEG2000 LNPs suffered rapid clearance and a sharp decrease (87.3%) of in vivo Luc mRNA transfection efficacy in the liver (Fig. 4b and Supplementary Fig. 9a,b). By contrast, BPL LNPs exhibited longer circulation with reduced anti-PEG IgG and IgM levels following the second dose (Supplementary Fig. 10) and demonstrated a similar mRNA delivery efficacy after the first and second doses. Structure–activity relationships of BPLs in enabling repeated dose transfection consistency were systematically studied. An increase of side chain length (x) resulted in less effective repeated dose efficacy, as evidenced by a smaller Luc liver delivery reduction of 9.0% for DMG-PEG₅MA₃₉ (x = 5) LNPs versus 35.4% for DMG-PEG₁₉MA₃₆ (x = 19) LNPs (Fig. 4b and Supplementary Fig. 9a,b). Similarly, a longer polymer backbone DP (y) resulted in a greater reduction of delivery efficacy following the second dose. Efficacy increased for DMG-PEG₁₉MA₅ LNPs and decreased for BPL LNPs with a longer backbone DP (DMG-PEG₁₉MA_y, y = 36, 60, 77) in the repeated dose scenario (Fig. 4c and Supplementary Fig. 9c,d). Altering alkyl tail length (n) did not lead to a distinguishable difference in repeated dose delivery efficacy, which was attributed to similar polymer backbone DP and side chain length (Fig. 4d and Supplementary Fig. 9e,f). Overall, BPL LNPs demonstrated improved maintenance in repeated dose efficacy than DMG-PEG2000 LNPs due to the weakened APA binding affinity.

Fig. 4 |. BPL LNPs outperform DMG-PEG2000 LNPs in repeated dose regimen studies.

a–d, Mice were first i.v. administered with standard DMG-PEG2000 LNPs or BPL LNPs carrying Luc mRNA (0.1 mg kg⁻¹). After 30 days, mice received an injection of the same LNP as the first injection (a). Reduction of luminescent signals by the second dose compared to that of the first dose with different EG side chain lengths (b), DPs (c) and alkyl lengths (d) is shown. Data are presented as mean ± s.d. and statistical significance was analysed by the one-tailed unpaired t-test: ***P < 0.001; **P < 0.01; *P < 0.05 (n = 3 biologically independent samples). e–i, Mice were first i.v. administered with BPL LNPs carrying Luc mRNA (0.1 or 0.5 mg kg⁻¹). After 30 days, mice received standard DMG-PEG2000 LNP injections (e). f,g, Relative serum anti-PEG IgG (f) and IgM (g) levels were measured seven days after the second dose (0.5 mg kg⁻¹). Data are presented as mean ± s.d. and statistical significance was analysed by one-way ANOVA multiple comparisons with DMG-PEG2000: ***P < 0.001; **P < 0.01; *P < 0.05 (n = 3 biologically independent samples). h,i, Reduction of luminescent signals by the second dose compared to that of the first dose with different side chain lengths (h) and DPs (i). Data are presented as mean ± s.d. and statistical significance was analysed by the one-tailed unpaired t-test: ***P < 0.001; **P < 0.01; *P < 0.05 (n = 3 biologically independent samples).

We hypothesized that LNPs with weaker APA binding affinity would be able to maintain efficacy after multiple doses. To study this, we performed another repeated dose regimen with different LNPs injected on day 1 followed by standard DMG-PEG2000 LNPs administration on day 30 (Fig. 4e). Serum anti-PEG IgG and IgM were measured after seven days to investigate the immune response. Over onefold higher anti-PEG IgG and IgM levels were detected in mice dosed with DMG-PEG2000 LNPs than all BPL LNPs following the first dose (Fig. 4f,g). The significantly lower antibody generation in mice treated with BPL LNPs indicated a weaker immune response. When mice received a second dose consisting of DMG-PEG2000 LNPs, the immune response was greatly boosted. Consequently, Luc activity was higher in all BPL-LNP-treated mice versus DMG-PEG2000-LNP-treated mice. Liver Luc transfection efficacy correlated with side chain length and DP (Fig. 4h,i and Supplementary Fig. 11a–d), in accordance with the structure-related difference in APA binding affinity.

To further support the role of APAs in generating an immune response that limits repeated dose efficacy, we examined LNP delivery in NOD.Cg-Prkdcscid/J severe combined immunodeficient mice that lack functional T cells and B cells. With this backdrop, severe combined immunodeficient mice treated with DMG-PEG2000 LNPs and BPL LNPs should not lose efficacy upon repeated dosing. In agreement with this hypothesis, the lack of an immune response indeed resulted in a comparable delivery efficacy of all LNPs in both first dose and repeated dosing regimens (Supplementary Fig. 12). In addition, no significant difference was observed in the mice administered with BPL LNPs as the first dose and DMG-PEG2000 LNPs as the second dose, further confirming that repeated dose loss of efficacy is caused by the APA immune response (Supplementary Information).

BPL LNPs overcome the APA inhibitory effect

Having established the advantage of BPLs in reduced APA binding and the fact of the wide existence of APAs produced by PEGylated drugs, we next examined if the inhibitory effect raised by APA production following DMG-PEG2000 LNP injection could be overcome (Fig. 5a). Mice were first administered with DMG-PEG2000 LNPs on day 1. On day 30, mice received DMG-PEG2000 LNPs or BPL LNPs. The second LNP dose exhibited reduced activity due to priming by DMG-PEG2000 LNPs. All BPL LNPs exhibited prolonged circulation time versus DMG-PEG2000 LNPs with a higher plasma cholesteryl methyl ether level and lower APA levels (Fig. 5b–d), indicating their excellent ability to overcome the inhibitory effect of APAs. In particular, certain BPL LNPs displayed superior mRNA delivery efficacy following the second dose compared to DMG-PEG2000 LNPs; for example, DMG-PEG₁₉MA₃₆ and 1,3-dipalmitoyl glycero (DPG)-PEG₁₉MA₃₆ BPL LNPs maximized Luc mRNA delivery efficacy and thus best overcame the neutralization of APAs (Fig. 5e,f). Turning to hEPO mRNA delivery, DMG-PEG₁₉MA₃₆ and DPG-PEG₁₉MA₃₆ BPL LNPs also produced higher hEPO levels (70% increase) versus DMG-PEG2000 LNPs (Fig. 5g). Overall, BPL LNPs were capable of overcoming the APA inhibitory effect stimulated by the systemic administration of DMG-PEG2000 LNPs, demonstrating potential in protein replacement therapy even against the backdrop of a widespread existence of APAs.

Fig. 5 |. BPL LNPs overcome the APA inhibitory effect induced by DMG-PEG2000 LNPs.

a, Schematic illustration of the rechallenge study. Mice were first i.v. administered with standard DMG-PEG2000 LNPs to create pre-existing APAs. After 30 days, mice received BPL LNP injections at the same dose. b, Plasma cholesteryl methyl ether level at 6 h post injection (0.1 mg kg⁻¹). Data are presented as mean ± s.d. and statistical significance was analysed by the two-tailed paired t-test: **P < 0.01; *P < 0.05 (n = 3 biologically independent samples). c,d, Relative serum anti-PEG IgG (c) and IgM (d) levels were measured seven days after the second dose (0.5 mg kg⁻¹). Data are presented as mean ± s.d. and statistical significance was analysed by one-way ANOVA multiple comparisons with DMG-PEG2000: ***P < 0.001 (n = 3 biologically independent samples). e–g, In vivo evaluation of Luc mRNA (0.1 mg kg⁻¹; e and f) and hEPO mRNA (0.3 mg kg⁻¹; g) delivery by BPL LNPs. Data are presented as mean ± s.d. and statistical significance was analysed by the two-tailed unpaired t-test: **P < 0.01; *P < 0.05 (n = 3 biologically independent samples).

BPL LNPs extend survival in protein replacement therapy

Next, we used a genetically engineered mouse model of hepatorenal tyrosinemia type 1 to compare mRNA-based protein replacement therapy of BPL LNPs to benchmark DMG-PEG2000 LNPs. Fumarylacetoacetate hydrolase (FAH)-encoding mRNA was delivered i.v. to FAH knock-out (FAH^−/−) mice (Fig. 6a) in a repeated dosing therapeutic setting. We first verified that BPL LNPs and DMG-PEG2000 LNPs could deliver FAH mRNA in vitro and in vivo, which resulted in clear protein expression in both cells and liver tissues (Supplementary Figs. 13 and 14). To mimic the clinical scenario that billions of people worldwide may have developed APAs, we induced the generation of APAs in FAH^−/− mice by i.v. administering DMG-PEG2000 LNPs 30 days before the start of protein replacement therapy. Groups of mice then received LNP treatment at a dose of 0.3 mg kg⁻¹ every three days after nitisinone (NTBC)-containing water was removed. Animals receiving phosphate buffered saline (PBS) treatment lost more than 20% of their body weight within 21 days (Fig. 6b and Supplementary Fig. 15). By contrast, protein replacement therapy slowed body weight loss. At the study end-point of 33 days, 100% of mice receiving repeated injections of DMG-PEG₁₉MA₃₆ BPL LNPs were alive (Fig. 6c). By contrast, only 37.5% of mice receiving repeated injections of DMG-PEG2000 LNPs were alive. FAH protein expression in DMG-PEG₁₉MA₃₆-LNP-treated and DPG-PEG₁₉MA₃₆-LNP-treated animals was substantially higher than that in DMG-PEG2000-LNP-treated animals (Fig. 6d,e and Supplementary Figs. 16–18). Moreover, bilirubin levels were most significantly reduced in the DMG-PEG₁₉MA₃₆ BPL LNP group (Supplementary Fig. 19). Overall, DMG-PEG₁₉MA₃₆ and DPG-PEG₁₉MA₃₆ BPL LNPs demonstrated a promising protein replacement therapeutic benefit and extended survival in diseased mice.

Fig. 6 |. BPL LNPs achieve superior therapeutic outcomes in protein replacement and genome editing.

a–e, BPL LNPs increase survival in protein replacement therapy. a, Schematic illustration of therapeutic regimen in FAH^−/− mice. Mice with pre-existing APAs (off NTBC water) were treated with PBS, DMG-PEG2000 LNPs, DMG-PEG₁₉MA₃₆ BPL LNPs or DPG-PEG₁₉MA₃₆ BPL LNPs every three days (FAH mRNA, 0.3 mg kg⁻¹). b, Body weight was monitored during the treatment. Data are presented as mean ± s.d. (n = 8 at day 0 biologically independent samples; mice with body weight loss over 20% were euthanized (red dashed line) and n decreased accordingly). c, Survival analysis of protein replacement therapy. Statistical significance was analysed by log-rank (Mantel–Cox) test. d,e, Western blot of liver tissues at end-point were measured to evaluate therapeutic efficacy. FAH expression is plotted as fold change relative to PBS. Data are presented as mean ± s.d. and statistical significance was analysed by one-way ANOVA multiple comparisons: ***P < 0.001; **P < 0.01 (n = 5 biologically independent samples). f–k, BPL LNPs achieve superior CRISPR/Cas9 editing over DMG-PEG2000 LNPs. f, Schematic illustration of genome editing study in C57BL/6 mice with pre-existing APAs. Cas9 mRNA and sgPCSK9 were co-delivered at a total dose of 1.5 mg kg⁻¹ (2:1, wt/wt). g,h, Indels detected in livers after ten days using Sanger sequencing and next-generation sequencing (NGS). Editing was quantified using TIDE (g) and CRISPResso2 (h) methods. i–k, Western blot of liver tissue (i) and ELISA of liver (j) and serum (k) PCSK9 levels were applied to evaluate the reduction of PCSK9 protein. Data are presented as mean ± s.d. and statistical significance was analysed by one-way ANOVA multiple comparisons: ***P < 0.001; **P < 0.01; *P < 0.05 (n = 3 biologically independent samples).

BPL LNP genome editing and protein inhibition

We next investigated the delivery capability of BPL LNPs for in vivo genome editing. We selected PCSK9 as a target gene, which functions in blood cholesterol level regulation and is associated with hypercholesterolaemia and atherosclerotic cardiovascular disease. Cas9 mRNA and single guide RNA targeting PCSK9 (sgPCSK9) were co-delivered to C57BL/6 mice with pre-existing APAs (Fig. 6f). DMG-PEG₁₉MA₃₆ and DPG-PEG₁₉MA₃₆ BPL LNPs induced 47.7% and 49.5%, respectively, insertions and deletions (indels) at the PCSK9 locus in livers 10 days after the i.v. injection, while DMG-PEG2000 LNPs induced only 13.1% indels at the same dose (Fig. 6g and Supplementary Fig. 20). Next-generation sequencing demonstrated that indels were threefold higher in the treatment with BPL LNPs than that with DMG-PEG2000 (Fig. 6h). The high genome editing efficacy mediated by BPL LNPs led to ~100% reduction in PCSK9 protein levels in the liver and serum, whereas DMG-PEG2000 LNPs mediated only a 60% reduction in PCSK9 protein in the liver and serum (Fig. 6i–k and Supplementary Fig. 21). In addition, apparent reductions in both serum cholesterol and triglyceride levels were also observed in animals treated with BPL LNPs (Supplementary Fig. 22). No obvious liver or kidney damage was found after treatment (Supplementary Figs. 23 and 24). These results provide an exciting insight that modulating polymer lipids with reduced APA binding can increase in vivo genome editing efficacy, which can benefit therapy.

Outlook

The finding that polymer chemical structure and architecture control nano–bio interfaces is anticipated to stimulate further polymer development to creatively open new avenues for engineering^38–40 LNP surfaces for various purposes. Further expansion of polymer chemistries with antifouling abilities, extrahepatic delivery ability and LNP formulation engineering will deepen the mechanistic understanding of LNP-related immune responses and disease-specific cellular effects. We envision that the versatile polymer lipids reported herein will support a robust and durable repeated dosing regimen and may find meaningful applications for a variety of therapies.

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/s41563-024-02116-3.

Methods

Materials

DMG-PEG2000 was purchased from NOF America Corporation. The 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) was purchased from Avanti Polar Lipids. The 1,3-dioctanoyl glycerol (DOG-OH), 1,3-dilauroyl glycerol (DLG-OH), 1,3-dimyristoyl glycerol (DMG-OH), 1,3-dipalmitoyl glycerol (DPG-OH) and 1,3-distearoyl glycerol (DSG-OH) were obtained from Cayman Chemicals. Cholesterol, CuBr, 2,2′-bipyridyl (bpy), N,N,N′,N″,N″-pentamethyldiethylenetriamine (PMDETA), 1,1,4,7,10,10-hexame thyltriethylenetetramine (TPMA), anti-PEG antibody clone 6.3 and all monomers were purchased from Millipore-Sigma. The 4A3-SC7 was synthesized following a previous report⁴⁰. The Quant-iT RiboGreen RNA assay kit was purchased from Life Technologies Corporation. The ONE-Glo + Tox Luc assay kit was purchased from Promega Corporation. The d-luciferin firefly, sodium salt monohydrate was purchased from Gold Biotechnology. The Anti-Mouse IgG Fc Capture (AMC) biosensor was purchased from Sartorius. The sgPCSK9 was purchased from and synthesized by Agilent Technologies, using solid phase synthesis and phosphoramidite chemistry, incorporating end modifications (3 × 2′-O-methyl-3′-phosphorothioate on the 5′ end and 3 × 2′-O-methyl-3′-phosphonoacetat on the 3′ end).

Synthesis of polymer lipids

Lipid-functionalized ATRP initiators were synthesized by an esterification reaction between lipid–OH and 2-bromoisobutyryl bromide. Typically, 1,3-dimyristoyl glycerol (DMG-OH, 1,000.0 mg, 1.95 mmol) and pyridine (231.4 mg, 2.93 mmol) were dissolved in 7.0 ml chloroform. The 2-bromoisobutyryl bromide (672.5 mg, 2.93 mmol) was dissolved in 5.0 ml chloroform and added dropwise into the above solution on ice. The mixture was allowed to react at room temperature for three days. The mixture was washed twice with 4% NaHCO₃ and dried with Na₂SO₄. Solvent was evaporated and the final product of DMG-Br was obtained through vacuum drying.

Polymer lipids were synthesized by ATRP using lipid–Br as the initiator, and bpy, PMDETA or TPMA as the ligand of CuBr, which depends on the monomers. For example, to synthesize DMG-PEG₉MA₄, DMG-Br (50.0 mg, 0.076 mmol), poly(ethylene glycol) methyl ether methacrylate with a number average molecular weight (Mn) of 500 (EG₉MA, 151.4 mg, 1.211 mmol) and PMDETA (31.5 mg, 0.182 mmol) were first dissolved in 2 ml isopropanol and degassed with nitrogen for 30 min. CuBr (13.0 mg, 0.091 mmol) was then added to the reaction and it was further degassed for 15 min. The mixture was reacted at 50 °C under the protection of nitrogen for 24 h. After that, CuBr was removed, and the mixture was dialysed against deionized water for two days to remove unreacted monomer and organic solvent. DMG-PEG₉MA₄ polymer lipid was obtained after freeze-drying.

The ¹H NMR spectra of initiators and polymer lipids are displayed in Supplementary Figs 25–53.

mRNA synthesis

Firefly Luc mRNA, hEPO mRNA, FAH mRNA and Cas9 mRNA were synthesized by in vitro transcription. In vitro transcription reactions were performed following standard protocols but with N1-methylpseudouridine-5′-triphosphate replacing the typical uridine triphosphate. Finally, the mRNA was capped (Cap-1) using the ScriptCap system (CellScript). The quality of in vitro transcription mRNA was evaluated by electrophoresis using an Agilent 4150 TapeStation system.

LNP formation and characterization

LNPs were formulated using the ethanol dilution method. The 4A3-SC7, DOPE, cholesterol and polymer lipids were dissolved in ethanol or a mixture of ethanol and dimethyl sulfoxide (DMSO) at a fixed molar percentage (Supplementary Table 1). The mRNA was diluted in citrate buffer (10 mM, pH 4.0). The two phases were rapidly mixed by pipette, vortex or T-mixing at a volume ratio of organic solvent to aqueous volume of 1:3. The weight ratio of 4A3-SC7 to mRNA was fixed at 17:1. The mixture was stabilized at room temperature for 15 min before dialysis against ×1 PBS for further use.

Standard DMG-PEG2000 LNPs were formulated with 4A3-SC7, DOPE, cholesterol and DMG-PEG2000 instead of polymer lipid at the same molar percentage using the same protocol as described above.

Dynamic light scattering (Malvern, v.7.13) was applied to measure the size and polydispersity index of LNPs. LNPs were diluted tenfold with ×1 PBS for dynamic light scattering studies. The stability of LNPs at 4 °C was investigated by recording the size and polydispersity index for seven days.

The encapsulation efficiency was measured using a RiboGreen assay.

Cryo-TEM was used to observe the morphologies of LNPs. To prepare samples for cryo-TEM imaging, holey carbon coated 300 mesh copper grids (Electron Microscopy Sciences) were glow discharged at 15 mA for 30 s with a PELCO easiGLOW glow discharge system (Ted Pella). Some 4.0 μl of sample was applied to the grids and the grids were then blotted for 2.5 s and plunged frozen in liquid ethane using Vitrobot Mark IV, under 4 °C and 100% humidity. The grids were kept in liquid nitrogen until imaging. Images were collected using a Glacios Cryo-TEM system (Thermo Fisher) at 200 kV with a Falcon 4 direct electron detector. Images were taken at −3.0 μm defocus to improve contrast. Two samples were prepared for each LNP. For each sample, at least 40 images were taken with similar results.

Protein adsorption assay

Plasma proteins adsorbed to LNPs were isolated using reported methods²⁵. Some 200 μl of LNPs (mRNA, 0.025 mg ml⁻¹) were mixed with 200 μl of C57BL/6 mouse plasma (K₂EDTA, Innovative Research) and incubated for 15 min at 37 °C. The mixture was loaded onto a sucrose solution (0.7 M) of equal volume and centrifuged at 17,000g and 4 °C for 1 h. The supernatant was removed, and the pellet was washed with ×1 PBS three times and centrifuged at 17,000g followed by removing the supernatant. The obtained pellet was resuspended in 2 wt% SDS buffer. Excess lipids were removed using a ReadyPrep 2-D Cleanup kit (Bio-Rad). The obtained pellets were dissolved in deionized water and the protein concentrations were measured by a Micro BCA protein assay kit (Thermo Scientific).

APA binding assay

Biolayer interferometry (Octet RED384, Sartorius) was employed to analyse the APA binding affinity with LNPs. To study antibody binding to LNPs, we loaded the pre-equilibrated AMC biosensors with 100 nM APA, clone 6.3 (in ×1 kinetics buffer from Sartorius). Then, binding curves with LNPs of different concentrations from 20 μM to 312.5 nM were measured. The time for each step was fixed as follows: 60 s for baseline phase, 120 s for loading phase, 300 s for association phase, 300 s for dissociation phase and 20 s for regeneration phase. The kinetic curves (association and dissociation steps) were fitted to a 1:1 kinetics model to calculate K_D, the association rate constant k_a (on) and the disassociation rate constant k_d (off) by Octet System Data Analysis Software (FortéBio).

In vitro transfection

In vitro Luc transfection was performed in HepG2 cells. Cells were purchased from the American Type Culture Collection (ATCC; catalogue no. HB-8065) and cultured in Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin (Gibco). Cells were seeded in a 96-well plate at a density of 10,000 cells per well. After incubation for one day, the medium was discarded and 100 μl of fresh medium containing LNPs was added to each well (50 ng mRNA per well). LNPs were incubated with cells for another 24 h. The cytotoxicity and Luc transfection efficiency were measured following the instructions of the ONE-Glo + Tox Luc assay kit.

In vitro FAH mRNA transfection was performed in A549 cells. Cells were purchased from ATCC (catalogue no. CCL-185) and cultured in DMEM with 10% FBS and 1% penicillin–streptomycin. Cells were seeded in a 12-well plate at a density of 100,000 cells per well and incubated for one day. Cells were then treated with LNPs (300 ng mRNA per well) for 24 h. Cells were washed with PBS and treated with T-PER tissue protein extraction reagent containing protein inhibitor (Thermo Fisher) to extract proteins. FAH protein levels were evaluated by western blot.

Animal experiments

All the animal experiments were approved by the Institution Animal Care and Use Committees of The University of Texas Southwestern Medical Center (protocol number 2016–101430; approval date, 7 June 2022) and were consistent with local, state and federal regulations as applicable. Mice were housed in a barrier facility with a 12 h light/dark cycle and maintained on standard chow (2916 Teklad Global). The temperature range for the housing room was 22 °C and the humidity range was 35–60% (average is around 50%).

Pharmacokinetic study

The pharmacokinetic profile of LNPs was evaluated using GC-MS. Cholesteryl methyl ether (Supplementary Fig. 54) was used to formulate LNPs instead of cholesterol. All LNPs were formulated at the same ratio as that described above. Male C57BL/6J mice were i.v. dosed with LNPs at an mRNA dose of 0.1 mg kg⁻¹. Blood samples were collected using BD Microtainer Capillary Blood Collector (lithium heparin with plasma separator tube) at predetermined time points, that is, 15 min, 30 min, 1 h, 3 h, 6 h, 12 h, 24 h, 48 h, 72 h, 96 h and 120 h. Plasma was obtained after centrifuging at 6,000g for 15 min, and the content of cholesteryl methyl ether was analysed via GC-MS.

In vivo Luc mRNA delivery in C57BL/6 mice

C57BL/6J mice were i.v. injected with LNPs at a Luc mRNA dose of 0.1 mg kg⁻¹. At 6 h post injection, mice were anaesthetized with isoflurane, and 100 μl of d-luciferin (30.0 mg ml⁻¹ in PBS) was intraperitoneally (i.p.) injected. After 5 min, the main organs were harvested, and the Luc activity was measured using an Ami imaging system (Spectral Instruments Imaging AMI-HTX).

In vivo hEPO mRNA delivery in C57BL/6 mice

C57BL/6J mice were i.v. injected with LNPs at an hEPO mRNA dose of 0.3 mg kg⁻¹. At 6 h post injection, mice were anaesthetized with isoflurane, and blood samples were collected to obtain serum. Serum hEPO level was evaluated by ELISA following a standard protocol.

Repeated dosing regimen

The repeated dosing study was performed by i.v. injecting LNPs containing Luc mRNA (0.1 mg kg⁻¹) or hEPO mRNA (0.3 mg kg⁻¹) on day 1 and day 30. Briefly, C57BL/6J mice were dosed with LNPs at a Luc mRNA dose of 0.1 mg kg⁻¹ via the tail vein. After 30 days, mice were dosed with the same LNPs at the same dose. At 6 h post injection, the Luc activity of the main organs was evaluated.

C57BL/6J mice were i.v. injected with different LNPs (Luc mRNA dose, 0.1 mg kg⁻¹) on day 1. After 30 days, standard DMG-PEG2000 LNPs (Luc mRNA dose, 0.1 mg kg⁻¹) were i.v. injected. At 6 h post injection, the Luc activity of the main organs was evaluated.

C57BL/6J mice were injected with standard DMG-PEG2000 LNPs at a Luc mRNA dose of 0.1 mg kg⁻¹ via the tail vein on day 1. After 30 days, a second dose of different BPL LNPs (Luc mRNA dose, 0.1 mg kg⁻¹) was injected via the tail vein. At 6 h post injection, the Luc activity of the main organs was evaluated, and plasma samples were collected to measure plasma LNP levels.

APA level measurement

The APA levels were measured in all three repeated dosing regimens. C57BL/6J mice were i.v. injected with LNPs (mRNA dose, 0.5 mg kg⁻¹) on day 1 and day 30. Seven days after the second dose, serum was separated. Serum anti-PEG IgG and IgM levels were measured by ELISA. MaxiSorp 96-well plates (Thermo Fisher) were coated with 100 μl of PEG-Catalase (Nanocs Bio, 10 μg ml⁻¹ in carbonate/bicarbonate buffer) overnight at 4 °C. After removing the coating agent, plates were blocked with 300 μl of 1% bovine serum albumin (BSA) for 2 h and washed with 0.145 M NaCl solution, three times. Then 100 μl of serum samples (1:1,000) were added and incubated for 90 min. After washing three times, 100 μl of anti-mouse IgG, horseradish peroxidase (HRP)-linked antibody (cell signalling technology, 1:1,000) or Peroxidase AffiniPure donkey anti-mouse IgM, μ chain specific (Jackson Immuno Research, 1:5,000) was added and incubated for 1 h. After another four washes, plates were incubated with 100 μl of 3,3′,5,5′-tetramethylbenzidine (TMB) substrate for 20 min and 50 μl of stop solution (R&D Systems). The absorbance was read at 450 nm.

In vivo Luc mRNA delivery in NOD.Cg-Prkdcscid/J mice

NOD.Cg-Prkdcscid/J mice were purchased from Jackson Laboratory. Mice were i.v. injected with BPL LNPs or DMG-PEG2000 LNPs (mRNA dose, 0.1 mg kg⁻¹) on day 1. After 30 days, standard DMG-PEG2000 LNPs (Luc mRNA dose, 0.1 mg kg⁻¹) were i.v. injected and the Luc activity of the main organs was evaluated at 6 h post injection.

In vivo FAH mRNA delivery in FAH^−/− mice

FAH^−/− mice were obtained from the lab of Professor Hao Zhu at the University of Texas Southwestern Medical Center and bred to maintain homozygous expression with NTBC water (Yecuris Corporation). Mice were i.v. injected with BPL LNPs or DMG-2000 LNPs (FAH mRNA, 0.3 mg kg⁻¹). After 6 h, livers were isolated and protein was extracted using T-PER tissue protein extraction reagent containing protein inhibitor for a western blot.

In vivo protein replacement therapy in FAH^−/− mice

FAH^−/− mice were first injected with DMG-PEG2000 LNPs (mRNA, 0.3 mg kg⁻¹) to generate APAs. After 30 days, NTBC water was removed and mice were treated with DMG-PEG2000 LNPs, DMG-PEG₁₉MA₃₆ BPL LNPs or DPG-PEG₁₉MA₃₆ BPL LNPs (FAH mRNA, 0.3 mg kg⁻¹) every three days via i.v. injection until day 33. The body weight of each mouse was monitored. Mice with a body weight loss over 20% were euthanized to comply with institutional guidelines on quality-of-life care. At each end-point, serum was separated for liver function evaluation, and liver was harvested for a western blot.

Western blot for protein replacement therapy

Protein concentration after extraction was measured using a bicinchoninic acid (BCA) assay kit. Some 30 μg of total protein was loaded. Separated proteins were transferred into a polyvinylidene membrane and blocked by 5% BSA for 1 h. Primary antibodies were incubated overnight at 4 °C (1:1,000 FAH recombinant rabbit monoclonal antibody, 1:1,000 primary actinin antibody). After washing four times, the membrane was incubated with anti-rabbit IgG, HRP-linked antibody (1:2,000). The membrane was imaged using a ChemiDoc imaging system (Bio-Rad).

In vivo genome editing (Cas9 mRNA/sgPCKS9) in C57BL/6 mice

C57BL/6J mice were first injected with DMG-PEG2000 LNPs (mRNA, 1.5 mg kg⁻¹) to generate APAs. After 30 days, mice were i.v. injected with DMG-PEG2000 LNPs or BPL LNPs for the co-delivery of Cas9 mRNA and sgPCKS9 at a total dose of 1.5 mg kg⁻¹ (2:1, wt/wt). Ten days after injection, serum was obtained for the evaluation of liver functions and PCSK9 protein, cholesterol and triglyceride levels. Liver was harvested for a histology study by haematoxylin and eosin (H&E) staining; PCSK9 protein analysis by western blot and ELISA assay; and genomic DNA collection.

Western blot for genome editing

Protein concentration after extraction was measured using a BCA assay kit. Some 40 μg of total protein was loaded. Separated proteins were transferred into a polyvinylidene membrane and blocked by 5% BSA for 1 h. Primary antibodies were incubated overnight at 4 °C (1:1,000 primary PCSK9 antibody (Abcam), 1:1,000 primary actinin antibody). After washing four times, the membrane was incubated with anti-rabbit IgG, HRP-linked antibody (1:2,000). The membrane was imaged using a ChemiDoc imaging system (Bio-Rad).

Statistics and reproducibility

Statistical analyses were conducted using GraphPad Prism v.9.5.1 software. The error bars in the results represent the mean ± s.d. The sample size was not predetermined by a specific statistical method, and dosing groups were filled by randomly selecting from the same pool of animals for both in vitro and in vivo experiments. No data were excluded from the analyses. Importantly, all investigators were blinded to group allocation during the data collection and analysis processes.

Reporting summary

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

Supplementary Material

Additional information

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41563-024-02116-3.

Data availability

All the data are available from the corresponding author upon request. Source data are provided with this paper.