Rational design and applications of piperazine and cyclohexane ionizable lipids for PKU and SSADH deficiency

Abstract

A significant challenge of mRNA-based protein replacement therapies is the diminishing efficacy and escalating toxicity associated with repeated dosing of lipid nanoparticles (LNPs). Many existing lipid formulations were originally designed for vaccine and are not optimized for therapeutic applications. We developed two libraries of ionizable lipids—one based on piperazine and the other on a newly introduced cyclohexane structure—with variations in linker and tail groups to enhance molecular diversity. GC Biopharma’s cyclohexane- and piperazine-based LNPs (GCP LNPs) supported stable and high-level expression without inducing liver toxicity under repeated dosing regimens. These LNPs effectively corrected disease markers in mouse models of phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency. Specially, we observed that the rigid structure and chemical stability of cyclohexane-based lipids contributed to sustained delivery performance. These findings offer a promising direction for the development of LNPs suitable for chronic mRNA-based therapies.

Graphical abstract

This manuscript has focused on the development and evaluation of piperazine- and cyclohexane-head-based ionizable lipids for mRNA delivery, highlighting significant implications in mRNA-based therapeutics, phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency.

Introduction

With the US Food and Drug Administration (FDA) approval of Onpattro in 2018 for the treatment of hereditary transthyretin-mediated amyloidosis (small interfering RNA [siRNA]-based therapeutic) and Comirnaty and Spikevax for the prevention of COVID-19 (mRNA-based vaccines), lipid nanoparticles (LNPs) have been validated as a clinically effective platform technology for RNA delivery.^1,2,3 Notably, a considerable number of mRNA-LNP vaccines for various diseases are currently undergoing clinical evaluations.^1,2,3 However, their use in chronic therapeutic applications remains underdeveloped.^4,5 LNPs primarily designed for vaccination are optimized for single or limited doses, and their adaptation to chronic treatment regimens is constrained by issues of cumulative immunogenicity and hepatotoxicity.^6,7,8,9

The clinical use of mRNA therapeutics on the other hand remains restricted to a few well-defined metabolic and genetic diseases so far (propionic acidemia [PA], methylmalonic acidemia [MMA], glycogen storage disease type 1a [GSD1a], ornithine transcarbamylase deficiency [OTCD], and cystic fibrosis [CF]).^{8,10,11,12,13} A major barrier to wider application is that current LNP platforms are not fully designed for chronic and repetitive delivery.^4,5,6

To overcome these limitations, a deeper understanding of LNP structure—particularly the role of ionizable lipids—is essential. LNPs are composed of four lipid components: ionizable lipid, phospholipid, cholesterol, and PEG-lipid, among which ionizable lipids play a critical role in controlling delivery efficiency and potency.¹⁴ Upon protonation, the cationic charge of the head group facilitates binding with anionic mRNA and promotes endosomal escape, releasing therapeutic mRNA into cytoplasm.¹⁴ The linker connects the head group with an array of functional groups and can incorporate ester groups to increase biodegradability and minimize accumulation, which plays a vital role in minimizing possible side effects and modulating tolerability.^15,16 Additionally, the tails are hydrophobic carbon chains that affect the overall physicochemical properties of the ionizable lipid.^17,18 Given their fundamental importance, establishing a novel library of ionizable lipids with improved efficacy and tolerability has become essential for the development of mRNA-based therapeutics. Here, we generated extensive libraries of ionizable lipids based on either a piperazine or a cyclohexane scaffold.^19,20 Through large-scale screening, we identified that four leads, GC Biopharma cyclohexane- and piperazine-based (GCP) lipids, outperformed industry-standard lipids (MC3, SM-102, or ALC-0315) in sustaining protein expression and tolerability upon repeat dosing.

Furthermore, we aim to address this challenge by developing mRNA-LNPs customized for two autosomal-recessive metabolic diseases with high unmet medical needs: phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency.^21,22 Both conditions require lifelong, costly, and high-risk enzyme replacement therapy (ERT), and no ideal treatment is currently available.^21,22 In both cases, sustained restoration of a single protein could normalize downstream biochemical imbalances, presenting a compelling opportunity for mRNA-based intervention if the delivery platform is optimized for repeat dosing.^21,22

Here, LNPs prepared with the GCP lipids were further evaluated for their therapeutic utility in rodent models of phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency.^21,22 GCP LNPs restored disease biomarkers to normal levels, maintained high therapeutic protein expression, and exhibited a favorable safety profile. These results represent significant advancement in the rational design of LNPs for chronic protein replacement therapy and suggest their adaptability across diverse monogenic targets (Figure 1).

Figure 1.

The strategy of structure-guided design of novel piperazine- and cyclohexane-based ionizable lipids

This manuscript has focused on the development and evaluation of 83 of piperazine- and cyclohexane-head-based ionizable lipids for mRNA delivery, highlighting significant implications in mRNA-based therapeutics, phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency.

Results

Construction and screening of a piperazine-based ionizable lipid library and structure-activity relationship analysis

Guided by our platform design strategy, we synthesized and evaluated two ionizable lipid libraries based on piperazine and cyclohexane headgroups to investigate how structural features influence delivery performance. We first constructed 43 piperazine-based ionizable lipids by combinatorially pairing seven chemically diverse head groups with 33 biodegradable tails, varying in carbon length and ester positioning (Figure 2A; Table S1). The piperazine core was selected based on its prior use in both small molecules and LNPs, offering potential advantages in safety and structural stability.^23,24,25,26

Figure 2.

Screening of a piperazine-based ionizable lipid library and structure-activity relationship analysis

(A) Lipids were synthesized using the piperazine head as the core. The tail library contained different carbon tails with additional esters and branched and unsaturated segments. All tails in the library were combined with piperazine head A (PA) for screening. The tails that exhibited the best performance were then combined with the piperazine head library. PD22-1 and PD22-2, which differ only in the carbon chain length of their ionizable head groups, are shown separately due to their structural similarity. All lipids were synthesized with >98% purity. (B) LNPs were formulated by mixing lipids and Fluc mRNA with a microfluidic device. Two compositions were fixed for amino lipid- and lipidoid-based LNP formulations: composition A (50:10:38.5:1.5) and composition B (26.5:20:51.5:2) (ionizable lipid: helper lipid:cholesterol:PEG-lipid). (C) In vitro luciferase expression levels were evaluated in HEK293T cells 24 h after treatment with 25 ng of fLuc mRNA/LNPs per well. (D) The best performing lipid nanoparticles were delivered to BALB/c mice through i.v. injection at a dose of 0.5 mg/kg luciferase expression levels at 6 h postinjection were detected in the liver region with the mice in a supine position and quantified using IVIS imaging. (E and F) Representative subsets of LNPs were compared to assess the effects of total carbon number (x+y) (E) and carbon number of the terminal tail (F). Data are represented as the mean ± standard deviation.

The lipids were synthesized through the Michael addition reaction between the amine heads and the corresponding acrylates (Figure S1A). Notably, the inclusion of a hydroquinone catalyst in the reactions provided higher yield.²⁷ All synthesized lipids were of more than 98% purity, and their structures were confirmed by ¹H nuclear magnetic resonance (NMR) spectroscopy (Figures S1B and S1C).

We then investigated whether these lipids could be formulated into monodisperse LNPs with high encapsulation efficiency. Since the physicochemical properties of LNPs are determined by the ratio of the four components (ionizable lipid, phospholipid, cholesterol, and PEG-lipid), we tested two different compositions that had been commonly used in the literatures to form stable LNPs (Figure 2B).^23,28,29 Composition A was derived from the formulation of MC3 LNP in Onpattro (ionizable lipid:cholesterol:phospholipid:PEG-lipid = 50:38.5:10:1.5), the only FDA-approved LNP for systemic delivery.²⁹ Composition B was adapted from lipidoid-based LNPs that incorporate multi-tail structures (five or more), following previously applied formulation ratios.^23,28 As the mixing procedure used has also been shown to impact LNP properties,³⁰ we mixed the components via microfluidics to ensure robust LNP formulations. For quality control, we measured the hydrodynamic diameter, polydispersity index (PDI), and encapsulation efficiency of all 83 LNPs individually via dynamic light scattering and a RiboGreen assay. Although 86 formulations were initially designed based on 43 ionizable lipids and two compositions each, three formulations failed to produce stable LNPs during preparation and were excluded from further analysis, resulting in a final set of 83 LNPs. All LNPs were within the diameter range of 200 nm (PDI <0. 2), and LNPs with composition B were similar in size to or smaller than those with composition A (Figure S2). An increase in encapsulation efficiency was generally observed for LNPs with composition B compared to composition A, suggesting that a lower molar ratio of piperazine-derived ionizable lipids is required in LNPs than in aminolipids for binding with mRNA and avoiding steric hindrance due to multiple hydrophobic tails.

After characterizing the pool of LNPs, we performed initial in vitro studies, evaluating each composition according to the expression level of the luciferase protein in HEK293T cells (Figure 2C). To better understand the mRNA delivery efficiency and biodistribution of ionizable lipids following systemic administration, the lipid composition that exhibited superior performance in in vitro luciferase assays was subsequently selected for in vivo studies, which specifically evaluated luciferase protein expression in the liver. (Figure 2D).

To elucidate structure-activity relationships (SARs), we focused on the influence of tail structure on delivery outcomes (Figures 2E and 2F). To minimize structural variability, SAR analysis was restricted to lipids carrying a PA head group. We further narrow the scope to candidates featuring two ester bonds in tail, a motif previously associated with modulated degradation rates and intracellular trafficking.^31,32,33,34

Tail length was quantified by summing the carbon number at the x and y positions (Figure 2D). Lipids with fewer than nine total tail carbons exhibited poor delivery efficiency and unstable LNP formulation. Optimal efficacy was observed with total carbon tail lengths between 9 and 20, whereas a decline in expression was shown when the tail extended beyond 21 carbons (Figure 2E). Within the optimal range, the y position—representing the terminal segment of the tail—had a greater effect than the x position (Figure 2F). Lipids with six or more terminal carbons consistently yielded higher protein expression, suggesting that increased hydrophobicity at the distal tail may facilitate mRNA encapsulation and promote endosomal escape.^18,32 The head group also played a decisive role in delivery performance. Furthermore, hydroxyl-containing head tail length was quantified by summing the carbon number at the x and y positions. Lipids with fewer than nine total tail carbons exhibited poor delivery efficiency and unstable LNP formulation. Optimal luciferase expression was observed within a moderate range of 9–20 carbons, whereas efficacy declined when the tail extended beyond 21 carbons (Figure 2E). Within the optimal range, the y position—representing the terminal segment of the tail—had a greater effect than the x position (Figure 2F). Lipids with six or more terminal carbons consistently yielded higher protein expression, suggesting that increased hydrophobicity at the distal tail may facilitate mRNA encapsulation and promote endosomal escape.^18,32 The head group also played a decisive role in delivery performance. Furthermore, hydroxyl-containing head groups improved expression, presumably through hydrogen bonding with mRNA, while the inclusion of tertiary amines contributed positively—possibly enhancing endosomal release.^35,36 In contrast, bulky substituents imposed steric hindrance, reducing efficacy.³⁷ Considering the physicochemical characterization and storage stability evaluation results in combination with the greatest in vivo mRNA delivery efficiency, two lead compounds (PA12 and PC27) were identified for subsequent studies.

Construction and screening of a cyclohexane-based ionizable lipid library and SAR analysis

Although piperazine-based ionizable lipids demonstrated robust mRNA delivery and high protein expression, previous studies have raised concerns regarding their tolerability upon repeated dosing, potentially due to limited STING activation or inefficient metabolic clearance.^9,38 To address these limitations, we introduced a cyclohexane core. The chair conformation of cyclohexane offers a predictable three-dimensional geometry, allowing bulky hydrophobic tails to adopt equatorial positions.²⁰ This spatial regularity is hypothesized to reduce steric hindrance during lipid packing, thereby enhancing the reproducibility of LNP self-assembly and improving nanoparticle structural integrity.^20,39,40

Based on these design principles, we constructed a second library comprising 40 cyclohexane-based ionizable lipids with diverse combinations of head groups, linkers, and tail structures (Figures 3A and S3). In vitro screening in Hep3B and C2C12 cells (Figure 3B) revealed that lipids containing branched or unsaturated ester-linked tails consistently yielded higher luciferase expression than those with saturated tails. Among the head group variants, lipids incorporating the CA head group—featuring an amide bond—exhibited superior mRNA delivery efficiency compared to ester-based counterparts (CB, CC, and CD). This enhanced performance is likely attributable to the amide’s ability to form stable hydrogen bonds with mRNA, resist hydrolytic degradation, and maintain a favorable hydrophilic-lipophilic balance that supports LNP stability and circulation.^35,41,42

Figure 3.

Screening of a cyclohexane-based ionizable lipid library and structure-activity relationship analysis

(A) Lipids were synthesized with different cyclohexane-based heads as the core. The tail library was designed by incorporating branched and unsaturated functional components at different positions. (B) In vitro luciferase expression levels in Hep3B and C2C12 cells were evaluated 18 h after treatment with 50 ng of fLuc-mRNA-encapsulating LNPs per well; five lipids were selected as lead lipids on the basis of their in vitro transfection efficacy. (C) For formulation optimization, four compositions were tested for physicochemical properties including the z-average and encapsulation efficiency. The results shown are the averages of the five different lead lipids for each composition. (D) Relationship between in vitro luciferase expression levels in 293T cells evaluated 24 h after treatment with 50 ng of fLuc-mRNA-encapsulating LNPs per well and in vivo IVIS luciferase expression in BALB/c mice evaluated at 6 h after intramuscular injection with 0.25 mg/kg fLuc-mRNA LNPs (n = 2/group). Data are represented as the mean ± standard deviation.

The top-performing lipids (CA04, CA05, CA09, CA10, and CD10) were selected for further optimization of lipid composition ratios (Figure 3C). When formulated using alternative compositions C or D, these cyclohexane-based lipids exhibited improved physicochemical properties, including reduced particle size and increased encapsulation efficiency, compared to the conventional composition B.

Subsequent evaluation of mRNA delivery efficiency was performed both in vitro and in vivo by measuring luciferase expression (Figure 3D). Lipids with CA head groups consistently outperformed those with CD head groups. Notably, post hoc analysis revealed a strong correlation between in vitro and in vivo transfection efficiencies, underscoring the utility of in vitro screening as a predictive tool for systemic delivery performance.

Based on combined results from physicochemical characterization and in vitro and vivo efficiency, CA09 and CA10 were identified as lead candidates for further development.

Efficacy and tolerability assessments of lead LNPs in rodents

Evaluating protein expression at a single time point is insufficient to fully characterize the therapeutic potential of mRNA medicines.^1,43 To address this, we conducted a comprehensive in vivo evaluation of the pharmacodynamic behavior and tolerability of four selected lead LNPs—PA12, PC27, CA09, and CA10—under repeated dosing conditions (Figure 4A).

Figure 4.

Efficacy and tolerability assessments of lead LNPs in rodents

(A) Schematic illustration of selected GCP lipids. (B) Relationship between in vivo luciferase expression and hepatotoxicity in BALB/c mice. BALB/c mice were analyzed 24 h after treatment with poly(A)-encapsulating LNPs (5 mg/kg) to evaluate acute hepatotoxicity. (C) Whole-body luminescence was monitored for 48 h with IVIS after BALB/c mice received an i.v. injection of 0.5 mg/kg fLuc-mRNA-encapsulating LNPs. (D) Protein expression was evaluated in BALB/c mice after repeated dosing. Serum hEPO concentrations were measured 6 h after each treatment with hEPO-mRNA-encapsulating LNPs (0.5 mg/kg, i.v., every 3 days for 3 doses). Hepatotoxicity was assessed 24 h after the final dose by measuring AST and ALT levels. (E) Serum hEPO concentrations, as well as AST and ALT levels, were analyzed in SD rats. hEPO levels were measured 6 h after a single dose of hEPO-mRNA-encapsulating LNPs (1.5 mg/kg) with ELISA. AST and ALT levels were measured 24 h after each injection (1.5 mg/kg, every 3 days for 4 doses) to evaluate hepatotoxicity of the lead LNPs. Data are represented as the mean ± standard deviation.

The lead compounds from two lipid libraries were further optimized for their formulation stability.³² Specifically, an increase in particle size and aggregation formation were found for LNPs with the standard formulation composition (ionizable lipid:helper lipid:cholesterol:PEG-lipid = 50:10:38.5:1.5), upon freeze-thaw. Since maintaining colloidal stability during freezing is a critical criterion in formulation development, various compositions were screened to identify an optimal formulation with minimal size change after freeze-thaw. By adjusting the molar ratios, we identified a new composition (C11; molar ratio of 25:10:50:2) that preserved particle stability for at least 6 months at −80°C (Figures S4 and S5). This improvement is likely attributable to the unique packing behavior of hydrophobic multi-tail lipidoids.^40,44 The lead compounds from two lipid libraries were further optimized for their formulation stability.^4,32 Specifically, an increase in particle size and aggregation formation were found for LNPs with the standard formulation composition (ionizable lipid:helper lipid:cholesterol:PEG-lipid = 50:10:38.5:1.5), upon freeze-thaw. Maintaining the colloidal stability during freezing process is one of the critical criteria in development, so various formulation compositions were screened for the optimal composition with minimal size change upon freeze-thaw. By adjusting the molar ratios, we identified a new composition (C11; molar ratio of 25:10:50:2) that preserved particle integrity for at least 6 months at −80°C (Figures S4 and S5). This improvement is likely attributable to the unique packing behavior of hydrophobic multi-tail lipidoids.^40,44

Using the optimized composition, we evaluated both protein expression kinetics and acute toxicity following single and repeated intravenous (i.v.) administration. After a single dose of luciferase mRNA, all LNPs exhibited peak expression at 3 h postinjection, followed by a gradual decline, consistent with the liver-centric distribution observed in benchmark LNPs such as MC3, SM-102, and cKK-E12 (Figure 4B).^4,43,45 cKK-E12 was included as a benchmark due to its well-characterized liver-specific delivery efficiency and unique structural features, such as a piperazine ring and multi-tail configuration, allowing direct comparison with the newly developed piperazine- and cyclohexane-based lipids. Based on the protein expression and hepatotoxicity analysis, CA09 and CA10 outperformed not only PA12 and PC27 but also the benchmark LNPs.^17,25,46 This observation suggests that subtle structural differences among lead lipids may influence repeated-dose hepatotoxicity, underscoring the importance of structure-guided design in the development of LNPs for chronic mRNA therapy.^18,46

To corroborate the utility of GCP LNPs for protein therapeutics, pharmacodynamic effects and tolerability were assessed after multiple administrations of human erythropoietin (hEPO)-encoding mRNA (Figure 3C). While the typical dosing interval for mRNA/LNP products for protein replacement applications is every 2 or 3 weeks,^43,47 an aggressive dosing interval (every 3 days) was chosen to demonstrate the feasibility of chronic applications under stressed conditions, which would manifest the accumulated toxicity, and due to the relatively short half-life of the hEPO protein in vivo. As with the luciferase expression results, the cyclohexane-based LNPs (CA09 and CA10) performed better than the piperazine-based LNPs (PA12 and PC27) according to the hEPO expression levels. The hEPO protein expression level was maintained at subsequent doses, and all tested groups exhibited comparable hepatotoxicity (Figure 4D). CA09, which showed the most favorable efficacy and safety profile in mice among the tested formulations, induced serum IL-6 and MCP-1 levels that were similar to PBS controls or lower than those induced by benchmark LNPs, indicating minimal immune activation (Figure S6). Importantly, a single i.v. dose of hEPO mRNA resulted in consistent protein expression in both mice and rat models, and repeated dosing did not elevate serum ALT or AST levels beyond those observed in PBS-treated controls.

Evaluation of GCP LNPs in phenylketonuria

To assess the therapeutic applicability of our top-performing cyclohexane-based ionizable lipids (CA09 and CA10), we evaluated their efficacy in a mouse model of phenylketonuria (PKU), a genetic disorder caused by mutations in the phenylalanine hydroxylase (PAH) gene.²¹ In PKU, the absence of functional PAH in hepatocytes prevents the conversion of phenylalanine (Phe) into tyrosine, leading to toxic Phe accumulation in the blood.²¹ We aimed to restore functional PAH expression in the liver by delivering mRNA encoding the homotetrameric PAH enzyme via GCP LNPs (Figure 5A).

Figure 5.

Enzyme replacement therapy for PAH deficiency via mRNA/LNP delivery

(A) Schematic representation of the enzyme replacement therapy after PAH mRNA/LNP delivery. (B and C) In vitro PAH protein expression (B) and PAH activity (C) were evaluated in hPAH mRNA-MC3 LNP-treated HEK293T cells. (D) Phe levels were measured in serum from Pah^enu2 mice following a single i.v. dose of hPAH-mRNA-MC3 (n = 2–3/group, mean ± SD). (E–G) All experiments were performed using unmodified hPAH-mRNA synthesized with wild-type UTP. (E) In vitro PAH expression in hPAH-mRNA-GCP LNP-treated HEK293T cells. The data are presented as the expression relative to the expression of hPAH-mRNA in MC3-treated HEK293T cells (n = 2/group, mean ± SD). (F) In vivo PAH protein expression was evaluated by western blotting in normal (WT) mice at 6 and 24 h after a single i.v. injection of either vehicle or 1 mg/kg hPAH-mRNA/LNP (n = 3/group, mean ± SD). (G) Serum Phe levels were determined at multiple time points (before and 6, 24, and 48 h after dosing) in Pah^enu2 mice (n = 3–5/group, mean ± SD) following a single i.v. dose of hPAH-mRNA-LNPs. The dotted line indicates the target concentration of serum Phe. Data are represented as the mean ± standard deviation. The data were analyzed by the Kruskal-Wallis test with Dunn’s multiple comparisons test.

Base modifications in mRNA are known to suppress the activation of TLR3, TLR7, and TLR8, leading to reduced immune responses and lower inflammatory cytokine secretion, which ultimately enhances translation efficiency.⁴⁸ However, certain base modifications may instead impair translation efficiency.^49,50,51 In particular, N1m introduced at specific positions can interfere with ribosomal function, slowing down the translation process. Furthermore, differences in translation efficiency between 5-methoxyuridine (5moU) and N1-methylpseudouridine (N1m) have been reported, varying based on coding sequence (CDS) composition.^48,49,50,51

To identify the most effective mRNA construct, we encapsulated three variants of IVT-mRNA (unmodified UTP, 5-methoxyuridine, and N1-methylpseudouridine) into MC3 LNPs and compared their protein expression and enzymatic activity both in vitro and in vivo.^23,52 Among them, unmodified UTP mRNA showed the highest protein expression, enzymatic function, and Phe clearance (Figures 5A–5D) and was selected for subsequent studies.

Using the unmodified PAH mRNA, we next compared protein expression between MC3, PC27, and GCP LNPs. All GCP LNPs significantly outperformed MC3, with CA10 achieving the highest PAH protein levels (Figure 5E). To validate in vivo efficacy, we measured hepatic PAH protein levels at 6 and 24 h postinjection. Both CA09 and CA10 resulted in stronger expression than MC3 (Figure 5F).

Finally, we assessed therapeutic effect in the Pah^enu2 mouse model of PKU by quantifying blood Phe levels after a single i.v. administration of GCP LNPs carrying hPAH mRNA.⁵³ MC3, the first FDA-approved LNP system, was selected as a baseline comparator in both the PKU and SSADH models due to its well-established hepatic delivery profile.¹ In the PKU model, SM-102 was chosen for its proven hepatic mRNA delivery efficiency and clinical relevance.^43,54 Furthermore, recent studies have demonstrated its applicability in hPAH mRNA delivery for PKU,⁵⁵ supporting its inclusion as a comparator alongside MC3.

Within 6 h postinjection, GCP LNPs reduced serum Phe concentrations by more than 95%, and this reduction was sustained at 56% and 72% at 24 h for CA09 and CA10, respectively (Figure 5G). These findings indicate that GCP LNPs successfully deliver functional PAH protein in vivo and induce rapid and meaningful metabolic correction in a disease-relevant model.

Functional restoration in succinic semialdehyde dehydrogenase deficiency model using GCP LNP

To explore the broader therapeutic potential of GCP LNPs in protein replacement therapy, we tested their efficacy in a model of succinic semialdehyde dehydrogenase (SSADH) deficiency, a rare neurometabolic disorder.²² SSADH deficiency, like PKU, results from an inherited enzyme deficiency—in this case due to mutations in the ALDH5A1 gene, leading to impaired gamma aminobutyric acid (GABA) catabolism and pathological accumulation of GABA and its metabolite γ-hydroxybutyric acid (GHB) in the central nervous system (Figure 6A).²²

Figure 6.

Enzyme replacement therapy for SSADH deficiency via mRNA/LNP delivery

(A and B) Schematic representation of the strategy for SSADH-mRNA/LNP delivery for enzyme replacement therapy. (C and D) The abundance of liver SSADH protein and downstream biomarkers (GHB and GABA) in SSADH-KO mice at 1, 7, and 14 days after a single injection of 0.1 mg/kg SSADH-mRNA/ALC-0315 LNP were analyzed by immunoblotting (C) and LC‒MS/MS (D), respectively. (E) Immunization schedule for analysis enzyme replacement therapy for SSADH deficiency via mRNA/LNP delivery. (F) SSADH protein abundance and activity in SSADH-KO mice at 1 day after a single injection of 2 mg/kgSSADH-mRNA/LNPs were evaluated via western blotting and an activity assay, respectively. The relative quantity was normalized to that of GAPDH. The data are presented as a percentage relative to the SSADH activity of wild-type mice as 100%. Kruskal-Wallis. (G) Biomarkers (GHB and GABA) in the liver and serum of SSADH-KO mice were measured via LC-MS/MS one day after a single injection of 0.25 mg/kg SSADH-mRNA/LNPs. WT, wild-type mice, KO, SSADH-KO mice, NT, no treatment. Data are represented as the mean ± standard deviation.

We designed GCP LNPs to deliver ALDH5A1 mRNA to hepatocytes and restore cytosolic SSADH enzyme expression, thereby promoting the degradation of GABA and GHB (Figure 6B). MC3 and ALC-0315 were selected as comparators based on their hepatic delivery profiles and clinical relevance. ALC-0315, a clinically validated LNP used in mRNA vaccines, has demonstrated strong liver-targeted expression and tolerability in preclinical studies including repeated dosing scenarios.^56,57 MC3, widely used in liver-targeted delivery, was included as a secondary comparator to provide a balanced assessment of GCP LNP performance.^1,57

In a preliminary study, benchmark LNPs delivering SSADH mRNA were administered i.v., and GABA and GHB levels were monitored for up to 14 days. A single dose of LNP-encoded SSADH protein led to sustained reductions in hepatic and serum levels of both metabolites (Figures 6C and 6D).⁵³ We then tested the same mRNA sequence using CA09 LNP in SSADH knockout (KO) mice (Figure 6E). At 24 h postinjection, hepatic SSADH protein levels with CA09 LNP were 1.3- and 1.7-fold higher than those achieved by ALC-0315 and MC3. Functionally, CA09 LNPs restored SSADH enzymatic activity to 57% and 82% of wild-type levels in liver tissue (Figure 6F). At 24 h postinjection, hepatic SSADH protein levels with GCP LNPs were 1.3- and 1.7-fold higher than those achieved by ALC-0315.² Functionally, GCP LNPs restored SSADH enzymatic activity to 57% and 82% of wild-type levels in liver tissue (Figure 6F). Correspondingly, GABA and GHB levels were substantially reduced: GHB concentrations decreased by 52% in liver and 70% in serum, while GABA was lowered by 70% and 72%, respectively (Figure 6G). These results demonstrate that GCP LNPs not only enable robust expression of mitochondrial SSADH protein but also achieve effective metabolic correction, further supporting their utility as a versatile mRNA delivery platform for systemic protein replacement therapies.

Discussion

Messenger RNA therapeutics offer promising strategies for treating chronic protein deficiency disorders.^4,5 However, their clinical translation relies on the development of LNPs capable of delivering mRNA efficiently and tolerably across repeated administrations.^8,10 Most currently available LNPs are derived from vaccine-based platforms optimized for transient expression, which may not meet the demands of long-term protein replacement therapies.^4,45 In this study, we introduce a structure-guided ionizable lipid platform designed to support prolonged mRNA expression with minimal immunogenicity. To systematically explore structure-activity relationships (SAR), we constructed two proprietary lipid libraries, built around either a flexible piperazine or a rigid cyclohexane core.^20,39

In piperazine-based lipids, hydroxyl-functionalized headgroups paired with medium-length ester-linked tails—such as in PA12 and PC27—significantly enhanced intracellular protein expression.^23,24,25 This improvement is likely driven by a combination of favorable polarity and enhanced membrane fusion properties, where hydroxyl groups facilitate hydrogen bonding with mRNA to improve encapsulation stability, while ester-linked tails contribute to endosomal escape by modulating lipid packing density.⁵⁷ In contrast, cyclohexane-based lipids with amide-linked headgroups—such as CA09 and CA10—achieved superior in vivo delivery and sustained expression under repeated dosing. These advantages may stem from the rigid cyclohexane core, which supports consistent nanoparticle morphology,^20,39 and from the amide bond, which confers hydrolytic stability and prolonged interaction with mRNA.^35,41,42 Together, these findings illustrate how flexible and rigid core architectures enable distinct but complementary delivery advantages, guiding the rational selection of lead candidates optimized for both efficacy and tolerability.

In addition, the LNP composition study highlighted the importance of storage stability study, where additional composition screening was required to minimize aggregation upon freeze-thaw cycle. This suggests the chemical structure of ionizable lipids affects the storage stability of LNPs, so the assessment of LNP composition for novel ionizable lipids to ensure storage stability would be recommended. The series of preclinical studies in rodent models have demonstrated the mRNA delivery efficiency and tolerability of GCP LNPs, which outperformed clinically validated benchmark LNPs.

The therapeutic potential of these GCP LNPs was also evaluated in murine models of phenylketonuria (PKU) and succinic semialdehyde dehydrogenase (SSADH) deficiency. GCP LNPs restored relevant metabolic markers to near-physiological levels, and these findings confirm functional delivery in vivo and underscore the translational value of our design strategy.^3,9,58

Despite these advances, several limitations should be acknowledged. Translation from rodent models to humans is inherently uncertain due to interspecies differences in pharmacokinetics and immune responses.^32,59 Bridging studies in non-human primates will be necessary to assess biodistribution, long-term expression, and long-term toxicity. Furthermore, although PKU and SSADH represent monogenic liver disorders, the generalizability of these LNPs to other therapeutic contexts—such as lysosomal storage diseases or hematologic conditions—remains to be validated.^11,12,13

With the expansion of mRNA therapeutics into diverse indications, the demand for programmable and scalable LNP platforms continues to grow.¹³ Our GCP lipid platform offers a proof of concept for how SAR-informed lipid engineering and iterative formulation strategies can produce delivery systems compatible with chronic, systemic administration across a broad therapeutic landscape.

In summary, we present a rational design strategy for next-generation LNPs, leveraging SAR insights from piperazine and cyclohexane derivatives to construct a delivery platform with improved durability, tolerability, and translational potential for mRNA-based therapies.

Materials and methods

Lipid synthesis

All lipids were synthesized and characterized using protocols described in the supplemental information. All materials were produced under contract by Aragen, Inc. (Hyderabad, India).

mRNA synthesis

mRNAs were synthesized in house by GC Biopharma Corp. (Yong-in, Republic of Korea). Briefly, N1-methylpseudouridine-substituted mRNAs encoding the firefly luciferase (fLuc, L-8102, Trilink), human erythropoietin (hEPO, L-8109, Trilink), or deleted human phenylalanine hydroxylase (hPAH, first 117 aa regulatory domain deleted) protein were synthesized with CleanCap AG 3′OMe (Sandiego, CA, Trilink) in vitro via T7-RNA-polymerase-mediated transcription from a linearized DNA template containing the 5′ and 3′ untranslated regions and a poly(A) tail. The codon-optimized coding sequence of SSADH was cloned and inserted into a GC Biopharma’s mRNA production plasmid (optimized 3′ and 5′ UTRs with a poly(A) tail), transcribed in vitro in the presence of N1-methyl-pseudouridine-modified nucleoside, co-transcriptionally capped with CleanCap AG 3′OMe (TriLink, San Diego, CA, USA), and then purified using Oligo dT affinity chromatography. The purified intact mRNA was subjected to ultrafiltration and diafiltration (UF/DF) and quality control. The purity of the mRNA was confirmed by IP/RP-UPLC analysis and exceeded 90% in all samples. Additionally, double-stranded RNA (dsRNA) was monitored as a critical impurity via dot immunoblot analysis, and its level was controlled to remain below 0.1% w/w. All the mRNAs were stored at −80°C until use.

Formulation of mRNA-encapsulating LNPs

mRNA-loaded LNPs were constructed via microfluidic mixing. Briefly, the lipid components (novel ionizable lipids, DOPE, cholesterol, and DMG-PEG2000) were dissolved in ethanol and rapidly mixed with mRNA in citrate buffer (50 mM, pH 4.0) using Ignite (Precision Nanosystems, Vancouver, BC, Canada) at a total flow rate of 12 mL/min and a ratio of 3:1 aqueous to organic phases. The resulting formulation was neutralized through buffer exchange with 1× DPBS using an ultrafiltration device (Amicon Ultra Centrifugal Filter, MWCO 30 kDa; Millipore, Billerica, MA, USA). The solution was subsequently sterilized using a Millex 0.2 μm PES syringe filter. Sucrose was used as a cryoprotectant. The final LNPs were stored in liquid form at 4°C or frozen at −20°C or −80°C. The hydrodynamic size, polydispersity index (PDI), and zeta potential of the LNPs were measured with a Malvern Zetasizer Nano ZS90 instrument (Malvern, UK). The mRNA encapsulation efficiency of the LNPs was determined via a Quant-it RiboGreen RNA assay according to the manufacturer’s protocol (Invitrogen, Carlsbad, California, USA).

Cell culture and animal studies

The HEK293T cell line was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and maintained in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37°C in a humidified incubator with 5% CO₂.

All experiments involving animals were approved by the Institutional Animal Care and Use Committee (IACUC) of GC Biopharma (approval number:GC-23-044-3A) in strict accordance with ethical guidelines and also performed at GC Biopharma under standard laboratory conditions (temperature, 20°C–24°C; relative humidity, 50%–60%; 12-h light/12-h dark cycle). The housing, handling, and experimental procedures/protocols for the mice were conducted with the relevant guidelines and regulations. Male BALB/c mice, C57BL/6 mice, and Sprague Dawley rats were purchased from Orient Bio (Seongnam, South Korea). BTBR-Pah^enu2 mice are a model of heritable PKU generated by ethylnitrosourea (ENU)-induced mutagenesis, leading to a missense mutation in exon 7, and mice were obtained from the Jackson laboratory (Strain#002232). The SSADH-KO model was generated in C57BL/6 mice by KO of the ALDH5A1 gene using CRISPR technology by Macrogen (Seoul, South Korea).

In vitro and in vivo screening of LNPs

HEK293T cells were seeded in 96-well plates at a density of 15,000 cells per well overnight and then treated for 24 h with fLuc mRNA-LNPs at a dose of 25 ng mRNA/well. Luciferase expression was evaluated via the Bright-Glo Luciferase Assay System (E2610, Promega, Madison, WI, USA) or the Luciferase Reporter 1000 Assay System (E4550, Promega, Madison, WI, USA) according to the manufacturer’s protocol. For in vivo screening, BALB/c mice were i.v. injected with fLuc mRNA-LNPs at an mRNA dose of 0.25 or 0.5 mg/kg. At 4 or 6 h postinjection, the mice were intraperitoneally (i.p.) injected with 150 mg/kg D-luciferin potassium salt (IVISbrite RediJect), and bioluminescence imaging was performed using an IVIS Lumina S5 Imaging System (PerkinElmer, Waltham, MA, USA).

Systemic mRNA/LNP delivery

hEPO mRNA/LNPs were administered via i.v. injection, and blood was collected via retroorbital bleeding at predetermined time points at room temperature (RT) in a Vacutainer SST (BD, Franklin Lakes, NJ, USA). The tubes were centrifuged at 3,000 g for 10 min, and the serum was aliquoted and stored at −80°C until analysis. hEPO concentrations were quantified using a Human Erythropoietin IVD ELISA Kit (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instructions.

PAH mRNA/LNPs were i.v. injected into BTBR-Pah^enu2 mice, and blood samples were collected 7 days before and 6, 24, and 48 h after injection to assess the serum Phe level. To determine hPAH protein expression, C57BL/6 WT normal mice were euthanized 6 and 24 h after a single i.v. injection of 1 mg/kg PAH mRNA/LNPs.

SSADH mRNA/LNPs were administered to 13-day-old SSADH-KO or WT mice via i.v. injection. The recovery of liver SSADH protein and biomarker (GHB, GABA) levels was analyzed by immunoblotting and liquid chromatography-mass spectrometry (LC-MS/MS) at 1, 7, or 14 days after a single injection in SSADH-KO mice. The SSADH protein level and activity in SSADH-KO mice were evaluated via western blot/ELISA and an activity assay, respectively (described in detail below), at 1 day after a single injection of 2 mg/kg SSADH mRNA/LNPs.

AST/ALT measurements

Blood was collected from mice and rats via retroorbital bleeding at predetermined time points for clinical chemistry analysis by separating serum by centrifugation (3,000 g × 10 min, 4°C), and AST and ALT levels were analyzed using an AU680 Clinical Chemistry Analyzer (Beckman Coulter, CA, USA) at Chaon (Gyeonggi-do, South Korea).

Western blotting

HEK293T cells were lysed in M-PER buffer containing a protease inhibitor cocktail after mRNA/LNP treatment. Liver tissue collected at predetermined time points after mRNA/LNP administration was homogenized via bead beating in T-PER buffer and then centrifuged at 12,900 rpm for 15 min; the supernatants were collected and subjected to western blot analysis. The cell or liver lysates were separated via SDS-PAGE and transferred to PVDF membranes. Human PAH protein or SSADH protein was detected using mouse anti-PAH or anti-ALDH5A1/SSADH monoclonal antibodies and HRP-conjugated mouse immunoglobulin G (IgG) kappa binding protein as the primary and secondary antibodies, respectively. As the endogenous control, β-actin (PKU) and GAPDH (SSADH) were detected with a rabbit anti-β-actin and GAPDH primary antibody and HRP-conjugated goat anti-rabbit IgG secondary antibody. The membranes were scanned with an iBright Imaging System to obtain images, and the band intensity was analyzed using ImageJ software for the PKU study. The SSADH protein intensity was detected using the ChemiDoc Imaging System (Bio-Rad Laboratories, Hercules, CA, USA) and quantified using Image Lab software (Bio-Rad Laboratories, Hercules, CA, USA).

In vitro PAH enzyme activity assays

HEK293T cell pellets were collected and lysed with PBS containing 50 mM sucrose and a protease inhibitor at 24 h after 250 ng/mL hPAH mRNA-LNP transfection. For the enzymatic reaction, the cell lysates were incubated with 2.5 mM L-phenylalanine, 0.5 mg/mL bovine liver catalase, and 50 mM HEPES buffer for 5 min at RT. Then, 0.2 μM ammonium iron (II) sulfate and 50 mM HEPES buffer were added, and the mixture was incubated for 1 min at RT. Tetrahydro-L-biopterin (BH4) (0.2 μM) dissolved in oxygen-free water and 4 mM dithiothreitol were added, and the mixture was incubated for 30 min at RT with gentle mixing using a rotating mixer. Finally, the enzymatic reaction was quenched by the addition of 10% formic acid. The supernatants were collected and stored at −80°C until further analysis. For HPLC analysis, the transparent supernatant was transferred to an LC glass vial after centrifugation at 12,000 rpm for 10 min and subjected to LC for chromatographic separation. The data were analyzed with Waters Empower software.

Assay to measure PKU metabolite concentrations

The mice were administered PBS or PAH mRNA-LNPs, and blood (serum) samples were obtained from the retro-orbital venous of mice before and after injection. Serum phenylalanine was analyzed using a Waters Alliance e2695 HPLC system (Milford, MA, USA). In brief, for each sample, the serum was diluted in water, and the protein was precipitated by combination with perchloric acid containing each of the internal standards. The precipitate was pelleted, and the supernatants were transferred to a new tube and subjected to LC for chromatographic separation. The data were analyzed with Waters Empower software.

SSADH activity assay

The liver homogenate was diluted to 30 μg/mL in assay buffer (100 mM Tris-HCl, 0.1 mM EDTA, 20 mM β-mercaptoethanol, 50 mM KCl, 0.1% Triton X-100). Ten microliters of the diluted samples was added to a black-bottom 96-well plate, followed by the addition of 290 μL of reaction buffer (90 mM succinic semialdehyde [SSA], 150 mM dithiothreitol [DTT], 150 mM glutathione [GSH], and 3 mM nicotinamide adenine dinucleotide [NAD+] in assay buffer). The reaction was carried out at 37°C, and the increase in NADH due to SSADH enzyme activity was measured kinetically over 1 h. A standard curve was prepared using 25 nmol of NADH (in 300 μL of assay buffer) serially diluted at 1:2 to produce 10 concentrations. The fluorescence was measured at an excitation wavelength of 355 nm and an emission wavelength of 470 nm using a SpectraMax ID5 (Molecular Devices, San Jose, CA, USA). The enzyme activities were determined via interpolation from a standard curve with a 4-parameter logistic regression model.

LC-MS/MS to quantify GHB and GABA levels

GABA and the internal standard 4-aminobutyric acid-2,2,3,3,4,4-d6 were purchased from Sigma-Aldrich, whereas GHB and the internal standard gamma-hydroxybutyric acid-d6 were obtained from LIPOMED AG (Arlesheim, Switzerland). A total of 10 μL of mouse serum was mixed with 10 μL of methanol, followed by the addition of 100 μL of the internal standard mixture (GABA-d6 or GHB-d6 in MeOH). Mouse liver tissue homogenates were prepared by homogenizing the median lobe of the liver in 1 mL of distilled water. Five microliters of sample was mixed with 45 μL of PBS, and 200 μL of the internal standard mixture was added. The samples were mixed at RT for 3 min and then centrifuged at 12,000 × g for 5 min at 4°C. The supernatant was collected and analyzed using LC-MS/MS. Calibration standards for GABA and GHB were prepared by spiking standard solutions into mouse serum or PBS as a surrogate matrix. The concentration ranges for serum were 2–200 μg/mL for GHB and 0.5–10 μg/mL for GABA, whereas the concentration ranges for the liver tissue were 0.5–10 μg/mL for both GHB and GABA. The LC-MS/MS system used was a Triple Quad 5500+ (AB SCIEX, Framingham, MA, USA) coupled with an Agilent Infinity 1290. For the GHB analysis, the MS conditions were optimized in negative electrospray ionization (−ESI) mode, and for GABA analysis, positive electrospray ionization (+ESI) mode was used. The MS parameters were set as follows: for GHB, declustering potential (DP) −20 V, collision energy (CE) −10 V, and collision cell exit potential (CXP) −10 V; for GABA, DP 65 V, CE 20 V, and CXP 10 V. The column used for LC analysis was a Waters Atlantis dC18 column (3 μm, 2.1 mm × 100 mm). The solvents used were (A) 0.1% formic acid in water and (B) 90% methanol in 0.1% formic acid in water. Gradient elution was employed for separation (5%–95% B over 5 min).

Statistical analysis

The experimental values are expressed as the means ± SD. The significance of differences between groups was evaluated via the Kruskal-Wallis test with Dunn’s multiple comparisons test. GraphPad Prism ver. 10.2.3 was used for the analysis. Differences between groups were considered significant at p < 0.05.

Data availability

The authors confirm that the data supporting the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

This work was supported by GC Biopharma Corporation. We would like to express gratitude to the Intellectual Property Team at GC Biopharma (Byung-Sun Kim, Chang-Soo Choi, and Ga-Yung Park) for their dedicated efforts in patent application and registration.

Author contributions

J.S., H.Y., J.P., H.P., and S.-E.C. conceived the project and wrote the initial manuscript. J.L., J.K., E.P., S.B., S.S., E.O., H.K., J.J., N.L., B.J., Y.Z., D.D.K., and S.D. performed the experiments. S.K. and J.B. analyzed the lipid synthesis data. J.S., Y.K.S., Y.D., and J.U.J. discussed the results. J.U.J. supervised the project.

Declaration of interests

J.S., H.Y., J.K., E.P., S.K., J.B., and J.J. are coinventors on a patent application covering the piperazine-based lipids described in the paper. Y.Z., D.K., S.D., and Y.D. are coinventors on a patent application covering the cyclohexane-based lipids described in the paper. GC Biopharma had a scientific research agreement with Y.D. Additionally, J.S., H.Y., J.K., E.P., S.B., S.K., J.B., H.P., S.C., S.S., E.O., H.K., H.K., J.J., N.L., B.J., and J.J. are or were at the time of the work employees of GC Biopharma. The other authors declare that they have no competing interests.