Multicomponent thiolactone-based ionizable lipid screening platform for efficient and tunable mRNA delivery to the lungs

Sequential Thiolactone Amine Acrylate Reaction (STAAR lipids)

Similarly to other ionizable lipid structures, STAAR lipids consist of a polar head, a linker, and hydrophobic tails. Given that one of the aims of this study was precisely to examine how the various fragments of these lipids can influence their in vivo performance, a combinatorial chemistry approach was followed to screen a selection of lipids that included different moieties in their structure. For simplicity, each lipid was named as AiBjCk to account for the reagents used for their synthesis. In this nomenclature, Ai represents an amine (polar head); Bj indicates the linker, which consists of a thiolactone incorporating a hydrophobic tail, and Ck represents an acrylate with another hydrophobic tail.

The synthetic process of STAAR lipids is illustrated in Fig. 1a. Briefly, this two-step synthesis starts with the coupling of homocysteine thiolactone with a carboxylic acid to yield the corresponding linker intermediate (Bj). The second step involves a tricomponent one-pot sequential reaction which starts with a ring-opening addition of the intermediate (Bj) with an amine (Ai) followed by a Michael addition with an acrylate (Ck) to yield the lipid AiBjCk²⁵. Remarkably, this multicomponent reaction can be carried out at room temperature, without need for catalysts or inert atmosphere, and within a 2-h reaction time. Moreover, the high purity (>80%) obtained makes it an optimal method for the rapid multidimensional screening of ionizable lipids, eliminating the need for further purification steps. This method simplifies the synthesis of new ionizable lipids in 1.5 mL tubes, reducing the reliance on expert organic synthesis knowledge, which allowed the rapid screening of nearly 100 novel lipids.

a Representative scheme describing STAAR lipids synthetic steps. First, the synthesis of Bj thiolactone derivative and then the one-pot tricomponent reaction. The amine opens the thiolactone ring of Bj (aminolysis) and subsequently the thiol group reacts with the acrylate to afford the final lipid with high purity in short times and at room temperature²⁵. b Scheme of the process followed for the formulation of lipids into LNPs for mRNA delivery in vivo.

The general structure of the resulting STAAR lipids includes two amide bonds (one originating from the reaction of the amine Ai and the thiolactone Bj and the other one connecting this thiolactone to its hydrophobic tail), a thioether bond (from Bj), and an ester group (from Ck). Consequently, any changes to be made in the hydrophobic tails can be achieved by modifying the structure of the thiolactone derivatives (Bj) and acrylates (Ck).

Two-phase screening strategy for ionizable lipids

To assess how different moieties affect the final structure and activity of the lipid, a two-phase screening strategy was devised in which one component of the lipid (Ai, Bj, or Ck) was initially kept constant while the other two were varied. This method enabled us to systematically evaluate the impact of each component on the lipid’s performance. By fixing one fragment and varying the other two, the intention was to efficiently identify the most promising candidates, thereby streamlining the screening process and optimizing our approach. Thus, the screening process was divided into two optimization phases, with each phase focusing on the optimization of a different lipid moiety (Fig. 2a, b). This methodological approach not only reduced the number of required experimental animals but also enhanced the efficiency of the screening process.

a Amines (Ai), Bj derivatives and acrylates (Ck) used for lipid screening. b Scheme summarizing the strategy followed for the two screening phases. c Total flux (entire mouse body) results of screening phase 1, where B was optimized by combining different acrylates and Bj derivatives while keeping polar head (A1) fixed. Mice were intramuscularly injected (i.m.) with mRNA-Luc-loaded LNPs at an mRNA dose of 0.05 mg/kg. Luminescence imaging acquisition was performed at 4 h post-treatment and total flux was quantified. In vivo mLuc expression displayed in a heat map (n = 3 biologically independent samples). Data are presented as mean values. d Total flux results table of Screening phase 2 where A and C were optimized by combining different amines and acrylates while keeping B2 fixed. Mice were i.m injected with mRNA-Luc-loaded LNPs at an mRNA dose of 0.05 mg/kg. Total flux luminescence imaging acquisition of the whole mouse body was performed at 4 h post-treatment and total flux was quantified. In vivo mLuc expression displayed in a heat map (n = 3 biologically independent samples). Data are presented as mean values. Below each table, representative luminescence images of each lipid in mice acquired using the IVIS Lumina XRMS Imaging System are shown.

For this optimization phase, the effectiveness of each IL candidate was evaluated by formulating each lipid individually into LNPs using a pipette rapid mixing procedure²⁶. Thus, each LNP was composed of an ionizable lipid (AiBjCk), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), cholesterol, and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG-2000) (50:10:38.5:1.5 molar ratios, respectively) and firefly luciferase mRNA (with a 6:1 molar N/P ratio Fig. 1b),we selected the same molar lipid ratios used for FDA-approved siRNA drug (Onpattro) and FDA-approved mRNA vaccine (mRNA-1273)²⁷.

Also, due to the weak correlation between in vitro and in vivo LNP performance²⁸, a direct screening in vivo using mice as animal model was followed, for which a firefly luciferase mRNA dose of 0.05 mg/kg was administered intramuscularly (i.m.) and the luminescence results were measured at 4 h post-inoculation.

Screening phase 1: optimization of linker’s hydrophobic chain (Bj)

The first phase of the structure optimization, aimed to select the optimal Bj component by synthesizing 21 lipids in which the polar head was fixed (A1 was selected) while combining three types of thiolactones with different acrylates (Fig. 2c). This amine was chosen as its structure closely resembles that of the MC3 polar head, featuring a spacer of three methylene units between a tertiary amine and a carbonyl group²⁹. The hydrophobic chains linked to the thiolactone derivatives (Bj) were selected to have distinctly different structures: a linear chain (B1) with 18 carbon atoms, an unsaturated chain (B2), and a branched chain (B3).

The analysis of the protein expression (Fig. 2c) achieved for the LNPs for each individual lipid revealed that A1B2Ck lipids (with k ranging from 1 to 7) resulted in increased protein expression for in vivo mRNA delivery, regardless of the acrylate used (Ck). This observation aligns with existing literature, which suggests that branched structures often confer enhanced performance in lipid-based delivery systems³⁰. Moreover, unsaturated chains yielded higher expression than saturated ones. Notably, hydrophobic moieties C3 and C7 were particularly effective.

Screening phase 2: optimization of polar head (Ai) and second hydrophobic chain (Ck)

Based on the findings from the screening phase 1, B2 was selected as the foundation for the next screening step. This second phase aimed to optimize the other two moieties of the molecule: the polar head (Ai) and the second hydrophobic chain (derived from Ck). Once again, the use of the STAAR synthesis approach for this optimization process allowed for a swift exploration of these structural moieties thanks to the high purities achieved in the sequential reactions used in their synthesis. More specifically, 70 AiB2Ck novel lipids were synthesized, with i and k ranging from 2 to 11 and 1 to 7, respectively (Fig. 2d).

Moreover, LNPs commonly consist of mRNA, help a comparative assay between crude and purified ionizable lipid to assess its impact on protein expression (Fig. S3). The results demonstrated no significant differences in performance between the two forms, thus validating the reliability of our approach. This result provides reassurance that the structure-function relationships derived from our screening data using crude lipids are reliable and directly translatable to purified compounds.

Regarding the Ai moiety, diverse polar heads, including cyclic, linear, aromatic, and analogs with a variable number of ionizable N groups, underwent testing. Upon in vivo evaluation, cyclic and aromatic amines (A5, A6, A7, A8, A9) failed to demonstrate significant expression, the only exception being that of A11. In contrast, among the linear amines, A2 and A4 displayed the best results, with lipid A4B2C3 notably yielding the highest expression levels. Remarkably, the comparison of amines A1 and A4, which differ by just one single methylene unit, resulted in an almost 5-fold increase in expression for lipid A4B2C3 over A1B2C3, emphasizing the role of the number of methylenes between the tertiary amine and the primary amine of Ai in the lipid performance. These observations showcase the sensitivity of lipid performance to minor structural changes in the spacer length of the polar head. Remarkably, the comparison of amines A1 and A4, which differ by just one single methylene unit, resulted in an almost 5-fold increase in expression for lipid A4B2C3 over A1B2C3, emphasizing the role of the number of methylenes between the tertiary amine and the primary amine of Ai in the lipid performance³¹. These observations showcase the sensitivity of lipid performance to minor structural changes in the spacer length of the polar head performance³¹. These observations showcase the sensitivity of lipid performance to minor structural changes in the spacer length of the polar head.

With regards to the hydrophobic tails (stemming from Ck), we confirmed the previously observed trend for Bj, with the 2-butyl-1-octanol derived branched acrylate (C3) yielding the lipids with highest expression levels. Additionally, linear 18-carbon monounsaturated (C1) also showed promising results in vivo. We also noticed that when shorter acrylate chains were used, the in vivo expression decreased up to 60 times, irrespective of it being branched or linear, this including isooctyl (C7), ethylhexyl (C2), and linear octyl (C6). Moreover, intermediate length chains (C5 and C4, 12 and 14 carbons in length, respectively), led to higher expression compared to shorter chains, yet the expression remained three times lower than C3 and C1. This suggests that longer chains, featuring unsaturations or branches in the hydrophobic moiety Ck, might be necessary to achieve the desired electronic and steric balance in the STAAR lipids.

Altogether, this screening phase highlighted the profound impact of the C moiety, where branched and unsaturated Ck chains, especially C3, demonstrated substantial improvements in protein expression.

In relation to physicochemical parameters of LNPs, observed sizes ranged from 150 to 230 nm, with the majority falling beneath 190 nm. Polydispersity indexes mostly lay below 0.25, while ζ potentials ranged between 5 mV and −25 mV, with all lipids but four exhibiting a negative ζ potential. mRNA encapsulation rates varied from 20% to 100% (Table S2). Notably, lipids with higher activity achieved encapsulation rates between 60% and 80% (with a direct pipetting method), hence suggesting this might represent the optimal encapsulation range for STAAR lipids LNPs in the screening phase. Moreover, we noted that even though employing an amine featuring two ionizable groups (A5) as lipid head results in high encapsulation rate, this does not reflect in its performance, as the in vivo activity was found to be 2 to 30 times lower compared to lipids containing amine A4, which contains only one ionizable group.

As expected, in terms of biodistribution, all intramuscular injections primarily displayed fluorescence signals concentrated in the muscle area of administration. This is consistent with previously reported findings, where mRNA-containing LNPs administered intramuscularly tend to express at the injection site and to some extent, in the liver³² (Fig. 2c, d).

Fine-tuning of the ionizable lipid structure

Following the discovery of lipid A4B2C3 as the top-performing candidate in the initial two-phase screening approach, the research focused on exploring the impact of further modifications to the lipid structure, specifically targeting the hydrophobic branched chains.

After identifying A4B2C3 as the most effective ionizable lipid in the two-phase screening, further structural studies were conducted to assess the impact of hydrophobic chain lengths and functional groups and their roles in influencing lipid performance. The purpose of this stage was to deepen the understanding of how variations in the chemical structure can influence protein expression, thereby refining the structure-activity relationship for improved delivery efficiency.

Assessment of the hydrophobic tails of lipid A4B2C3

Following the same synthetic approach as in the screening phase, we synthesized and purified eight additional ionizable lipids, hence introducing three distinct branched chain lengths via the Bj and Ck moieties (Fig. 3a). Having optimized the lipid structure in the initial phases, for these refining phases we relied on microfluidic techniques to confirm the previous conclusions while keeping the formulation process precisely controlled. Nanoparticles were obtained with sizes ranging from 80 to 130 nm, with the majority being under 100 nm, and they exhibited high encapsulation efficiencies (90–99%, Table S4). Thus, a purified batch of the optimized lipid A4B2C3 along with these new 8 lipids were formulated into LNPs along with three separate controls. Two out of three controls used MC3 as ionizable lipid in the same molar and N/P ratios as the screened lipids, one of them included DOPE as helper lipid (same as the formulated STAAR lipids), namely MC3 DOPE, whereas the other one replicated the MC3 LNP benchmark by using DSPC. We also added SM-102 used for FDA-approved mRNA vaccine as control. Remarkably, when comparing the in vivo results to those of the MC3 benchmark LNP²⁹ (Table S3), eight out of the nine tested lipids showed a total flux that was at least double that of the MC3 benchmark.

a Building blocks for hydrophobic tail optimization, combining A4, branched Bj derivatives and branched acrylates. b In vivo luminescence total flux results of a structures normalized respect to MC3 benchmark LNP, with the dashed line in the y-axis representing the results of control MC3 benchmark LNP. MC3(DOPE) LNP was also used as control employing DOPE as helper lipid. Mice were i.m injected with mRNA-Luc-loaded LNPs at an mRNA dose of 0.05 mg/kg. Luminescence imaging acquisition was performed at 4 h post-treatment and total flux was quantified. In vivo mLuc expression (n = 3 biologically independent samples). Data are presented as mean values ± standard deviation (SD). c Schematic lipids set structure displaying different functional groups varying in alpha (α) and beta (β) positions. d Results of lipids in c, in vivo mLuc expression (n = 3 biologically independent samples). Data are presented as mean values ± SD. Mice were i.m injected with mRNA-Luc-loaded LNPs at an mRNA dose of 0.05 mg/kg. Luminescence imaging acquisition was performed at 4 h post-treatment and total flux was quantified. e ζ potentials and pKa of the formulated LNPs (n = 3 biologically independent samples). Data are presented as mean values. In b and d one-way ANOVA with Tukey’s correction was used.

Among these, the result for A4B2C8 (known as CP-LC-0729) was particularly striking (Fig. S4), as it demonstrated an expression value 6.4 times higher than the MC3 benchmark control. Furthermore, it outperformed SM-102 control (Fig. 3b), which firmly establishes this lipid as the best candidate for further studies and optimization. The exception amongst these 8 lipids was observed with lipid A4B4C3, characterized by the shortest hydrophobic chain combination, and for which the expression dropped to 0.034 times relative to the MC3 benchmark. The results suggest that a tail with a minimum carbon number may be necessary for achieving high protein expression, aligning with the trend observed during the two-phase screening, where shorter chains such as C2, C6, and C7 consistently resulted in lower protein levels²¹.

Influence of functional groups in the structure-activity relationship of STAAR lipids

After exploring the impact of altering the hydrophobic chains and identifying lipid CP-LC-0729, which contains two branched chains (B2, C8), as the best candidate for high-performing in vivo activity, our focus then shifted to study the functional groups within the structure. As previously reported in literature, certain bonds, such as esters, play a pivotal role in endosomal escape and degradation pathways^9,16,31,33. Taking advantage of the chemical versatility of STAAR lipids, which allows for modular modification of functional groups, we investigated how the nature of specific bonds can influence the performance of the STAAR lipids.

We focused specifically on the role of the amide bonds, which are common to all previously synthesized STAAR lipids and are located at positions referred to as α and β^34,35, by synthesizing purified structural variations of CP-LC-0729 as depicted in Fig. 3c. These products were subsequently formulated with the same N/P and lipid ratios used in the initial screenings, employing microfluidic mixing.

As a first step, we inverted the C-N carbonyl order (N-CO to CO-N) of the β amide bond (CP-LC-0729-03) resulting in a threefold reduction of in vivo protein expression compared to CP-LC-0729 (Fig. 3c, d). This change may be attributed to the shift in the relative position of the β amide hydrogen bond, which can potentially affect the intermolecular force landscape and packing of the molecule³⁶.

Then, to further explore the role of the α amide bond, lipid CP-LC-0729-02 was synthesized, where the α amide is methylated, resulting in a 3.4-fold decrease of in vivo expression. In the next step, the amide bonds were partially or completely substituted with ester bonds in CP-LC-0729-01, CP-LC-0729-04 and CP-LC-0729-05. In this case, we noticed a significant decrease in effectiveness—by two orders of magnitude—when replacing both amides (α and β positions) with ester groups in CP-LC-0729-05 (Fig. 3d). This change resulted in the lowest performance among the proposed variations, and contrasts with the fact that other commercial ionizable lipids such as ALC-0315, SM-102, or MC3 only feature ester groups. As a middle-ground case, when combining an amide and an ester group, the in vivo performance was dependent on relative position of the groups, yielding intermediate outcomes. Interestingly, when lipids contain only one amide, their expression is four and a half times higher when located in the α position (CP-LC-0729-04) compared to the β position (CP-LC-0729-01). Thus, these results indicated that preserving the α amide bond (CP-LC-0729-04) led to better expression results than replacing it with an ester group (CP-LC-0729-05). Remarkably, the protein expression value of the methylated α amide lipid CP-LC-0729-02 was on par with CP-LC-0729-03 and CP-LC-0729-04. Interestingly, when lipids contain only one amide, their expression is four and a half times higher when located in the α position (CP-LC-0729-04) compared to the β position (CP-LC-0729-01). Additionally, in the case of the methylated α amide (CP-LC-0729-02), which no longer acts as a hydrogen bond donor, the expression decreased threefold compared to the non-methylated counterpart, suggesting that hydrogen bonds might play a key role. We hypothesize that the observed difference may arise from the α amide proximity to the polar head, potentially enhancing the pKa and its availability to interact with mRNA via hydrogen bonding or intermolecular interactions between ionizable lipids. The optimal combination, as evidenced by our findings, involves amides located at both the α and β sites, reflecting in the chemical structure of lipid CP-LC-0729.

Regarding the LNP physicochemical characterization, we noticed that altering the α amide bond led to lower LNP pKa. Specifically, pKa values were the lowest for CP-LC-0729-05 (pKa = 5.48) and CP-LC-0729-01 (pKa = 6.12). A similar trend was observed when measuring ζ potential, as CP-LC-0729-05 (ζ = −18.0 mV) and CP-LC-0729-01 (ζ = −10.3 mV) displayed the lowest values of the series (Fig. 3e).

Additionally, to further evaluate the behavior of lipid CP-LC-0729 in different formulations, an in vivo study using LNPs with different molar ratios of the four lipid components was conducted. The results, detailed in Table S7, indicate that CP-LC-0729 performs well even when the molar ratios diverge from the ones used in more conventional formulations, highlighting its versatility. We also characterized CP-LC-0729 LNP using the same formulation used in screening process through cryo-TEM to evaluate the structure and morphology. CP-LC-0729 LNP displayed a spherical shape with uniform size and solid-core morphology, with some bleb structures (Fig. 4a). Previous studies have documented the presence of bleb structures in LNPs. These structures are reported to improve the stability of the mRNA cargo and transfection efficiency in both in vivo and in vitro³⁷.

a Representative cryo-TEM images of CP-LC-0729 LNP. Scale bar, 100 nm. b Serum alanine aminotransferase levels (ALT), serum aspartate aminotransferase levels (AST), serum alkaline phosphatase levels (ALP), urea nitrogen concentration (UREA). Mice were intravenously injected (i.v) with mRNA-Luc-loaded LNPs at an mRNA dose of 2.5 mg/kg (n = 5 biologically independent samples). Serum was collected for ALT, AST, ALP and UREA analysis at 24 h and 48 h post-treatment. Data are presented as mean values ± SD. Two-way ANOVA with Tukey’s correction was used. c Body weight change for 15 days (measured made at days 0, 1, 2, 7, 9, 13 and 15). Mice were i.v injected with mRNA-Luc-loaded LNPs at an mRNA dose of 2.5 mg/kg (n = 5 biologically independent samples). Data are presented as mean values ± SD. d Uptake inhibition of CP-LC-0729 and MC3(DOPE) LNP by different endocytosis inhibitors (n = 3 biologically independent samples). The inhibitors used were Amiloride (inhibitor of macropinocytosis), Chlorpromazine (inhibitor of clathrin-mediated endocytosis), Genistein (inhibitor of caveolae mediated endocytosis), Methyl-β-cyclodextrin (inhibitor of lipid raft mediated endocytosis). Data are presented as mean values ± SD. e Hemolysis assay of CP-LC-0729 LNP and MC3(DOPE) LNP at pH 7.4 or 6.0. RBCs were incubated with CP-LC-0729 LNP and MC3(DOPE) LNP at an mRNA concentration of 2.5 μg/mL at 37 °C for 1 h. Positive and negative controls were carried out with 0.1% Triton-X and 1× PBS, respectively. Data are presented as mean ± SD (n = 3 biologically independent samples).

Additional biological assays of the top-performing ionizable lipid CP-LC-0729 LNP

In vivo safety profile of lipid CP-LC-0729

After selecting CP-LC-0729 as the best performing ionizable lipid candidate through the in vivo screening and subsequent optimization phases, different biological assays were carried out to evaluate the safety profile. Thus, in vivo toxicity assays on CP-LC-0729 LNPs were conducted to assess any potential organ damage by measuring liver enzymes (ALT, AST, ALP) and kidney function (UREA) at 24 and 48 h after systemic injection at a high dose (2.5 mg/kg). When compared to the PBS buffer used as a control, CP-LC-0729 LNPs showed no significant impact on the levels of these enzymes (Fig. 4b). The change in mice body weight was monitored for 15 days post-inoculation, revealing no apparent differences compared to the negative control (PBS buffer) (Fig. 4c). In order to further assess the intramuscular toxicity of CP-LC-0729, we examined the muscular area surrounding the inoculation site for signs of toxicity by comparing TRIS buffer, CP-LC-0729, and SM-102; no apparent signs of toxicity were observed (Fig. S9). While a more comprehensive assessment, including testing across different animal models, is necessary to establish a complete safety profile of this lipid, our preliminary results indicate excellent tolerability following systemic administration, with no signs of toxicity observed under the conditions studied.

Evaluation of cell internalization of CP-LC-0729 LNPs

To gain deeper insights into the mechanisms of cell internalization, the endocytosis pathways of CP-LC-0729 LNP were examined in vitro in hepatic HepG2 cells. Four endocytic pathways were assessed by incubating the LNPs with specific inhibitors for each route before in vitro transfection²¹. The results show that the uptake of CP-LC-0729 LNPs strongly relies on lipid rafts, as indicated by inhibition with methyl-β-cyclodextrin. Other internalization pathways, such as macropinocytosis and clathrin-mediated endocytosis, showed a smaller contribution. Additionally, genistein did not inhibit uptake, indicating that caveolae-mediated endocytosis is not significant (Fig. 4d). We also conducted hemolysis assays to assess endosomal escape of CP-LC-0729 LNPs through their membrane-disrupting activity under both acidic (pH 6) and physiological (pH 7.4) conditions^21,30. These tests showed that CP-LC-0729 LNPs increased hemolysis in the acidic medium up to 90%, ten times higher than the MC3 (DOPE) LNP used as control. In contrast, at physiological pH, the hemolytic activity remained comparable to the control level (PBS) indicating no hemotoxicity (Fig. 4e). This suggests a potential enhanced endosomal escape mechanism of CP-LC-0729 LNPs at acidic pH, while maintaining hemocompatibility at physiological conditions, a result that is in accordance with the fact that lipid CP-LC-0729 shows a higher in vivo performance than MC3 while displaying no signs of toxicity³⁸.

Extrahepatic organ targeting of CP-LC-0729 using permanently cationic lipids

Recognizing the critical importance of targeting specific organs for effective therapeutic outcomes, we focused on achieving extrahepatic mRNA delivery to broaden the scope and versatility of ionizable lipid CP-LC-0729 and the STAAR screening platform. As reported by Siegwart and coworkers^24,39,40, the addition of a fifth permanently charged lipid results in selective targeting of the lungs or spleen, attributed to changes in the protein corona and LNP surface charge¹³. Herein we explored a streamlined approach to modify the STAAR ionizable lipid CP-LC-0729 into its permanently cationic form (+) CP-LC-0729, using a single synthetic step, and later carried out a comparative study with what is reported in the literature following the fifth-lipid strategy. Hence, a novel cationic lipid was synthesized, denoted as (+) CP-LC-0729, as it is derived from the methylation of the tertiary amine of the top-performing ionizable lipid CP-LC-0729 (Fig. 5a). Subsequently, the performance of ionizable lipids MC3 and CP-LC-0729 was compared, in LNPs incorporating cationic lipids (+) CP-LC-0729 or the standard commercial alternative DOTAP as fifth lipids, using representative formulations with firefly luciferase mRNA (Fig. 5b)²⁴. As shown in Table S5, the cationic lipid molar percentages were compared using LNPs that incorporated different proportions of this fifth lipid (30%, 40% and 50% (mol/mol)). The experiments were conducted using microfluidics mixing for LNPs, followed by intravenous administration. LNPs were termed as [cationic lipid (percentage of cationic lipid in molar rate) ionizable lipid] LNP, for example, LNP using DOTAP as cationic lipid at 50% combined with MC3 as ionizable lipid was designated [DOTAP (50%) MC3] LNP.

a Synthetic scheme used to prepare a permanent cationic lipid, namely (+) CP-LC-0729, from an ionizable lipid CP-LC-0729. b Schematic representation of the formulated LNPs incorporating a STAAR permanently cationic lipid as fifth component for mRNA delivery in vivo through i.v. administration. c Top to bottom: Ex vivo luminescence of major organs from mice, pie charts showing a profile of the tissue-specificity of mLuc expression in the lungs, heart, liver, intestine, spleen and kidneys, bar representation of total luminescence found in lungs and table with the formulation details of LNPs (n = 3 biologically independent samples). Mice were i.v. injected with mRNA-Luc-loaded LNPs at an mRNA dose of 0.2 mg/kg. Images were acquired at 4 h post-treatment. For the bar graph, data are presented as mean values ± SD. One-way ANOVA with Tukey’s correction was used.

Regarding the physiochemical properties, LNPs with (+) CP-LC-0729 showed similar values to CP-LC-0729 LNP, including particle size (60–85 nm) and 99% mRNA encapsulation efficiency. However, we noticed positive surface ζ potentials in (+) CP-LC-0729 LNPs, which varied depending on the cationic lipid ratio used (Table S6).

After inoculation, ex vivo assays revealed distinct variations in protein expression in the lungs, depending on the cationic lipid ratio. The [(+) CP-LC-0729 (50%) CP-LC-0729] lipid nanoparticles (LNPs) demonstrated remarkable selectivity for the lungs, achieving 88% selectivity. Even the [(+) CP-LC-0729 (40%) CP-LC-0729] and [(+) CP-LC-0729 (30%) CP-LC-0729] LNP exhibited strong selectivity, 88% and 82%, respectively (Figs. S5 and S6), far exceeding the 38% selectivity observed for the benchmark LNP, [DOTAP (50%) MC3].

Moreover, when [(+) CP-LC-0729] was formulated with MC3, the resulting expression levels were twofold lower than when it was combined with CP-LC-0729 (Fig. 5c). Even more remarkably is the fact [(+) CP-LC-0729 (50%) CP-LC-0729] LNP exhibited 32-fold higher luciferase expression compared to the benchmark [DOTAP (50%) MC3] LNP (Fig. 5c). These findings demonstrate that both the selectivity and protein expression of CP-LC-0729 combinations significantly outperform the benchmark [DOTAP (50%) MC3] formulation. Furthermore, the optimal results observed when combining (+) CP-LC-0729 with CP-LC-0729 may suggest that structural similarity between the ionizable and cationic forms may improve lipid packing and enhance overall LNP performance. In order to broaden the scope of our approach, we synthesized and conducted in vivo assessments of two additional cationic lipids, (+) A4B2C1 and (+) A2B2C4, using the same one-pot strategy and ionizable lipid CP-LC-0729. Interestingly, the resulting LNPs also exhibited tropism to the lungs, with selectivity values of 66% and 59%, respectively. Notably, the total flux observed in the lungs was significantly higher for (+) A4B2C1 compared to (+) A2B2C4, which was consistent with the enhanced mRNA delivery performance observed for A4B2C1 in the phase 2 screening (Fig. S8). Overall, [(+) CP-LC-0729 (50%) CP-LC-0729] LNPs remained the most effective formulation for lung delivery (Fig. S7). These results show the potential of CP-LC-0729, while also expanding its application to therapies requiring targeted extrahepatic delivery.

Additionally, [(+) CP-LC-0729 (50%) CP-LC-0729] LNP was formulated with a small percentage of the fluorescent probe DiR (dialkylcarbocyanine, 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindotricarbocyanine iodide) to track fluorescence and determine LNP biodistribution. While most LNPs were directed to the liver, luciferase expression was observed predominantly in the lungs (Fig. S8).