Multicomponent reaction-based combinatorial chemistry for accelerating the discovery of therapeutic protein and nucleic acid delivery materials

Abstract

Biopharmaceutical drugs, such as therapeutic proteins and nucleic acids, are gaining prominence in modern medicine, presenting great opportunities for addressing major diseases that are challenging to treat with traditional small-molecule drugs. However, delivering these biomacromolecular drugs to specific intracellular targets remains a significant challenge. Due to the large size and hydrophilicity, biomacromolecules cannot easily permeate across cell membrane, thus requiring the development of carrier materials to enable their effective delivery into cells. Despite this need, the labor-intensive synthesis approach and inefficient structure optimization process significantly hinder the discovery of efficient carrier materials. Recently, multicomponent reaction (MCR)-based combinatorial chemistry has emerged as a powerful solution, enabling the rapid generation of a large materials library for screening and structure-activity optimization. In this perspective, we discuss the design principles of therapeutic protein and nucleic acid delivery materials, summarize recent advances, and propose future directions for using the MCR-based combinatorial chemistry to develop next-generation biomacromolecule delivery systems.

Introduction

Biomacromolecules, including peptides, proteins, and nucleic acids, have emerged as important therapeutic agents for a broad range of diseases because of their high specificity and potency, lower required dosages, and favorable safety profiles^1,2. As a matter of fact, biomacromolecular drugs accounted for 80% of the top ten best-selling drugs globally in 2023, including a peptide drug, a messenger RNA (mRNA) vaccine and six antibody-based drugs³. However, the inherent challenges of these biomacromolecules, such as poor stability, high molecular weight and hydrophilicity, impede their ability to cross biological barriers, making their delivery particularly challenging¹. This underscores the urgent need for the development of safe and efficient delivery systems for biomacromolecular therapeutics.

Effective biomacromolecular drug delivery systems must meet several key criteria⁴. First, they must ensure efficient encapsulation to protect the therapeutic cargo from degradation, maintaining high stability and bioactivity in biological environments. This requires precise adjustment of intermolecular interactions between the cargo and the delivery carriers. Second, the delivery systems must overcome cellular membrane barriers to enable direct intracellular delivery or facilitate endo-lysosomal escape, thus enhancing bioavailability of the drug within cells; Third, the delivery carriers need high biocompatibility and degradability to minimize adverse side effects. It is essential that the degradation products of these carriers are non-toxic and easily excreted from the body, improving the overall safety profile of delivery systems.

Recently, multicomponent reaction (MCR)-based combinatorial chemistry has emerged as a promising approach for accelerating the development of diverse biomacromolecule delivery materials. This approach offers a significant advantage by enabling the rapid exploration of vast chemical spaces, generating extensive libraries of carrier materials in an efficient and one-pot reaction^5,6. This significantly speeds up the discovery and optimization processes of delivery materials to address specific therapeutic challenges. In addition, combining high-throughput screening approach with MCR-based combinatorial chemistry allows researchers to systematically investigate structure-activity relationships (SARs) of delivery materials. This provides valuable insights into how specific structural features impact on delivery performance of carrier materials, thus guiding the rational design of high-performing biomacromolecular drug delivery systems.

In this perspective, we highlight the design principles, current advancements, and emerging trends in the application of the MCR-based combinatorial chemistry for discovering therapeutic protein and nucleic acid delivery materials. The integration of rational design principles with the versatility of MCR-based combinatorial chemistry approach holds great promise for addressing the current limitations of biomacromolecule delivery systems, paving the way for the development of next-generation delivery materials for various therapeutic scenarios.

Design principles of carrier materials

The design principles of carrier materials to deliver protein and nucleic acid therapeutics should focus on addressing the following problems: effective binding and encapsulation of biomacromolecules, efficient cellular internalization and escape from the endo-lysosome, as well as controlled drug release within cells. Proteins have a large molecular size, chemically diverse and heterogeneous surface, featuring various hydrophilic, hydrophobic, and charged domains (Fig. 1a)⁷, which present challenges for the formation of strong interactions between delivery carriers and proteins of interest. To enhance their binding affinity, multiple intermolecular interactions such as electrostatic, hydrophobic, and cation-π interactions, coordination bonding, and hydrogen bonding were usually employed either individually or in combination (Fig. 1b)⁸. These driving forces were confirmed to assist the formation of protein/carrier complexes, thus improving protein stability against protease degradation and controlling drug release inside cells. In contrast, nucleic acids were single- or double-stranded with negatively charged phosphate backbones, enabling them to be preferentially condensed and encapsulated by the cationic delivery materials via the electrostatic interaction, and the resulting nucleic acid/carrier complexes can be further stabilized by additional hydrophobic interactions within the delivery carriers⁹.

Fig. 1: The scheme of design of carrier materials for intracellular protein and nucleic acid delivery.

a Basic structures and properties of proteins and nucleic acids. b Effective encapsulation of proteins or nucleic acids with various delivery materials depending on multiple intermolecular interactions. c The main challenges of intracellular biomacromolecule delivery. Endocytosis pathway (i) and direct intracellular delivery by the membrane fusion (ii) are two major mechanisms of cellular internalization. For endocytosis pathway, the delivery system would undergo endo/lysosome escape to reach the cytoplasm. Then the cargo release was triggered in response to specific intracellular conditions by the stimuli-responsive chemistry (iii).

The ability to across cell membrane barrier is crucial for effective intracellular delivery of proteins and nucleic acids. As shown in Fig. 1c, cell membrane, primarily composed of amphiphilic phospholipid bilayers with the negatively charged surface, presents significant barriers to deliver large and hydrophilic biomacromolecules. Enhancing electrostatic interactions between delivery carriers and the negatively-charged cell membrane, along with fine-tuning amphiphilicity of carrier materials, can significantly improve efficiency of cellular internalization⁴. In addition, rationally designing carrier materials to modulate membrane affinity and influence fluidity can increase membrane fusion processes, thus facilitating direct transport of biomacromolecules into the cytosol¹⁰. Once the entry into cells by the endocytosis pathway, the delivery system must escape from the endo/lysosome to avoid premature degradation of biomacromolecular cargo. The endo/lysosome compartments are characterized by an acidic microenvironment (pH ~5) as compared to the neutral cytosol (pH ~7.4)¹¹. pH-sensitive delivery carriers, such as ionizable cationic materials, can exploit the protonation effect of ionizable amines at the pH below their pKa to trigger the disruption of the endo/lysosomal membrane¹². In addition, inducing a structure transition from the laminar phase to the hexagonal II phase can also disrupt the endo/lysosomal membrane, facilitating cargo release into the cytosol (Fig. 1c)¹³. This mechanism has been involved in the rational design of cone-shaped ionizable cationic lipid materials for nucleic acid delivery.

For biomacromolecules to exert the therapeutic effect, they usually need to be effectively released from the delivery carriers once inside cells, thus requiring a delicate balance in the intermolecular interactions between carrier materials and therapeutic cargo. If the interactions are too strong, biomacromolecules may remain bound to the carriers, hindering intracellular drug release and thus decreasing the bioavailability. Conversely, if the interactions are too weak, the biomacromolecules may prematurely dissociate from the carriers during systemic circulation, leading to the enzyme degradation or off-target toxicity. To address this dilemma, stimuli-responsive chemistry is often introduced to design the delivery carriers that can trigger the release of biomacromolecular cargo in response to specific intracellular conditions, such as changes in pH, redox potential, or the presence of specific enzymes (Fig. 1c)^{14,15,16,17,18,19}. These ensure that drug release occurs precisely within the targeted intracellular environment, maximizing therapeutic efficacy while minimizing side effects.

The chemical design of carrier materials plays a crucial role in the development of safe and effective biomacromolecule delivery systems. Traditional synthesis method of delivery materials often involves some complex, multi-step reactions, which can be labor-intensive and costly, making the optimization of carrier candidates challenging⁵. To overcome the limitation, the combinatorial chemistry approach has emerged as a more efficient strategy for the design of biomacromolecule delivery materials. Several mild and efficient reactions, such as Michael addition reaction, epoxide ring-opening reaction, and reductive amination reaction, are widely used for high-throughput functional screening of nucleic acid delivery systems²⁰. In recent years, MCRs have attracted increasing attentions as a powerful approach for the development of various biomacromolecule delivery carriers. MCRs allow the simultaneous combination of three or more reactants to generate a product in a one-step and efficient manner (Fig. 2a). The MCRs can be generally categorized into two classes of isocyanide-mediated and non-isocyanide-based reactions (Fig. 2b). Among these, there are two widely known isocyanide-mediated MCRs²¹. The first is the three-component Passerini reaction (Passerini-3CR), discovered in 1921, combines an aldehyde, an isocyanide and a carboxylic acid to form the α-acyloxycarboxamide. Similarly, the four-component Ugi reaction (Ugi-4CR), discovered in 1959, involves an aldehyde, an isocyanide, an amine and a carboxylic acid to produce the α-acylaminoamide. Some subsequent isocyanide-mediated MCRs were derived from the framework of the Passerini-3CR and Ugi-4CR. Non-isocyanide-based MCRs can be divided into two major types²². The first type includes one-step and efficient reactions, such as the Hantzsch reaction (1881), which combines β-ketoester, aldehyde and ammonia to afford the dihydropyridine, and the Biginelli reaction (1891), which involves aldehyde, urea and β-ketoester to obtain the 3,4-dihydropyrimidin-2(1H)-ones. The second type involves tandem reactions carried out without intermediate purification, effectively yielding the product in a one-pot process. In general, these MCRs can simultaneously introduce diverse functional moieties such as hydrophobic, positively charged, stimuli-responsive or hydrogen-bonding into the rational design of carrier materials (Fig. 2c). Compared to traditional chemical methods that require stepwise to design the introduction of various functional moieties into materials, the MCRs provide a more streamlined approach to generate versatile delivery materials. This facilitates high-throughput screening of carrier materials and their SARs investigation. Currently, various types of MCRs, such as Passerini-3CR²³, Ugi-4CR²⁴, Van leusen-3CR²⁵, and the amine-thiol-acrylate-based tandem reaction²⁶, have been reported for developing various biomacromolecule delivery systems (Fig. 2b). In the following sections, we categorize recent designs for therapeutic protein and nucleic acid delivery materials, showcasing the powerful potential of the MCRs in this field.

Fig. 2: The MCR-based combinational chemistry approach for the synthesis of carrier materials.

a Advantages of the MCR-based combinational chemistry approach. b Typical MCRs used for the design of biomacromolecule delivery materials. c Rational design of delivery materials with diverse functional moieties such as (i) hydrophobic, (ii) charged, (iii) hydrogen-bonding, and (iv) stimuli-responsive ones.

Protein delivery

Chemical conjugation and physical encapsulation are currently main strategies of carrier systems for therapeutic protein delivery. Post-translational modification by chemically conjugating functional moieties to proteins can improve their stability and pharmacokinetics in vivo⁸. For instances, polyethylene glycol (PEG) conjugation (PEGylation) has been widely used in clinical practice to prolong half-life of proteins by reducing protease degradation, and minimizing renal clearance, and decreasing immunogenicity. Using the Ugi-4CR, Yang et al. synthesized a series of PEG-protein conjugations, in which proteins retained their native structures during conjugation process, demonstrating the feasibility and immense potential of the Ugi-4CR in protein conjugation applications²⁷. In another study, Sornay et al. employed the Ugi/Passerini reactions to enable the chemical modification of the antibody trastuzumab, in which commercially available aldehydes and isocyanides were used to react with spatially close amine and carboxylate groups that provided by the aspartate/glutamate residues in the antibody. This work demonstrated a new approach for site-specific protein modification²⁸. For physical encapsulation of proteins, our lab has recently used the Ugi-4CR-based modular strategy to establish a combinatorial library of small-molecule gemini amphiphiles (GAs), identifying effective vehicles that facilitate cytosolic delivery of a wide range of proteins²⁹. The typical GA comprises two alkyl tails and two ionizable units connected by a spacer, and we synthesized 150 GAs from a relatively small number of chemical inputs, including 3 aldehydes, 2 isocyanides, 5 amines, and 5 carboxylic acids (Fig. 3a i). Specifically, the charge properties of GA can be tuned by incorporating various amines, while aldehydes and carboxylic acids were introduced to control overall hydrophobicity of GAs. The SARs analysis revealed that increasing chain lengths of alkyl tails in the GAs drastically improved protein delivery efficiency with decreasing in vitro cytotoxicity, and the top candidates were GAs with 18–20 carbons in alkyl tails. Of note, diisocyanides as bifunctional linker determined not only hydrophilic/hydrophobic balances but also the spacer length of GAs. When comparing the hydrophilic diisocyanide with hydrophobic one, we found that the hydrophilic spacer was critical for improving intracellular protein delivery. Furthermore, the length of hydrophilic spacer was fine-tuned using the diisocyanides containing one, two, or three ether bonds. However, the GA synthesized by the diisocyanide with three ether bonds (A1I2-3R2C18) significantly decreased protein delivery efficiency as compared to those with one (A1I2-1R2C18) and two ether bonds (A1I2R2C18) (Fig. 3a ii), which is likely due to the increased hydrophilicity negatively affecting carrier’s affinity for cell membranes, thus decreasing intracellular protein delivery efficiency.

Fig. 3: The MCR-based design of protein delivery materials.

a (i) Combinational synthesis of gemini amphiphiles (GAs) via the Ugi-4CR. (ii) Chemical structures of GAs with one ether bond spacer (A1I2-1R2C18), two ether bonds spacer (A1I2R2C18) or three ether bonds spacer (A1I2-3R2C18). (iii) In vitro genome editing on 293T-EGFP cells by RNP/A1I2R2C18 complexes. (iv) In vivo genome editing efficiency on tumor-bearing mice. (v) Illustration of the proposed lipid raft-dependent membrane fusion mechanism mediated by GAs, bypassing the classical endocytic pathway. Reproduced with permission from the American Chemical Society²⁹. b (i) The combinatorial approach for establishing dimeric amphiphiles (DAs) and trimeric amphiphiles (TAs) via the Ugi-4CR. Intracellular delivery of IgG-FITC (ii) and anti-NUP153 antibody (iii) by the A1I14R2C51. Reproduced with permission from the American Chemical Society³⁰. c Synthesis route of a small library of cationic polypeptoids with biodegradable disulfide units, and the scheme of polypeptoids mediated intracellular protein delivery process. Reproduced with permission from the Science China Press¹⁵.

Furthermore, the GAs with alkyl tails containing 18 carbons and hydrophilic spacers with one or two ether bonds (A1I2-1R2C18 and A1I2R2C18) were identified for highly efficient intracellular delivery of various proteins. These GAs successfully translocated a wide range of proteins with different molecular weights and pIs into the cytosol, exhibiting superior delivery efficiency than the commercial protein delivery reagent PULSin. These delivered proteins also retained their bioactivity within cells, for instance, GA-delivered saporin led to ribosome inactivation and cell death, while delivered β-galactosidase retained high catalytic activity inside cells. Notably, the top GA can effectively deliver the hard-to-deliver Cas9 ribonucleoprotein both in vitro and in vivo (Fig. 3a iii and iv). In vitro result demonstrated that genome editing efficiency of Cas9 ribonucleoprotein delivered by the A1I2R2C18 was superior to the commercial reagent CRISPRMAX Cas9. Moreover, in vivo genome editing efficiency reached approximately 20% in tumor-bearing nude mice, leading to disruption of the KRAS mutation and inhibition of tumor growth. A key factor contributing to the highly efficient protein delivery both in vitro and in vivo was dependent on their special entry mechanism (Fig. 3a v), which was mainly mediated via the lipid raft-dependent membrane fusion, bypassing the classical endocytic pathway that limits cytosolic delivery efficiency of many current carriers.

In a subsequent study, we explored the impact of amphiphile architecture, specifically dimeric and trimeric structures, on intracellular protein delivery capacity³⁰. The architecture of these amphiphiles can be adjusted by multifunctional carboxylic acids, while the other three components were monofunctional in the Ugi-4CR. To this end, dimeric amphiphiles (DAs) and trimeric amphiphiles (TAs) were synthesized by using bifunctional carboxylic acid and trifunctional carboxylic acid, respectively (Fig. 3b i). The top DA screened from a materials library based on BSA-FITC delivery efficiency had two alkyl tails with 16 carbons each (A1I16R2C41), while the top TA had three alkyl tails with 14 carbons each (A1I14R2C51). This finding indicates the optimal alkyl tail length differs depending on the amphiphile architecture, while shorter alkyl tails are more suitable for multimeric amphiphiles. Notably, the amphiphile architecture has influenced several performances of DA and TA. First, the complexes formed by individual DA and TA were similar in particle size (250–300 nm), while protein/TA complexes showed an average diameter of 648 nm, which is 1.4 times larger than protein/DA complexes (460 nm). Second, protein/TA complexes exhibited a higher protein loading efficiency (93.5%) compared to protein/DA complexes (77.2%). Third, the top TA demonstrated a stronger protein delivery capacity than both DA and the commercial reagent PULSin, effectively delivering a range of proteins into the cytosol, including negatively-charged R-phycoerythrin and ovalbumin, as well as positively-charged cytochrome C and lysozyme. This result suggests that TA have a stronger binding affinity with proteins, likely due to the presence of three tertiary amines in each TA molecule, resulting in higher charge distribution.

Furthermore, the entry mechanism of protein/TA complexes involved multiple pathways, including lipid raft-, pinocytosis-, caveolae-, and phagocytosis-mediated endocytosis pathways. A notable advantage of TA is its capacity to transport the hard-to-deliver antibody inside cells. The feature of macromolecular and near-neutral surface of antibodies generally leads to its weak affinity with delivery carriers, resulting in low intracellular delivery efficiency. Interestingly, TA efficiently delivered the model antibody FITC-labeled human immunoglobulin G (IgG-FITC) into the cytosol, outperforming the PULSin (Fig. 3b ii). Moreover, the intracellularly delivered anti-NUP153 antibodies retained their bioactivity, successfully targeting the cellular nuclear pore complex protein NUP153 (Fig. 3b iii). We propose that increasing the charge distribution of delivery carriers may be an effective strategy to enhance antibody delivery efficiency.

The Ugi-4CR also offers a solution for synthesizing polymeric carriers for intracellular protein delivery, through using two bifunctional and two monofunctional components in the one-pot reaction¹⁵. These cationic polymers, known as polypeptoids, were synthesized by coupling dicarboxylic acids, diisocyanides, isobutyraldehyde and amines (Fig. 3c). Using disulfide bond-contained dicarboxylic acids, stimuli-responsive backbones were introduced into cationic polypeptoids. The optimized polypeptoids (A1-D2) could form compact complexes with proteins, improving their cellular uptake through the caveolae-mediated endocytosis and macropinocytosis, while also facilitating endo/lysosomal escape. Subsequently, protein cargo could be efficiently released into the cytosol due to the intracellular glutathione (GSH)-triggered cleavage of disulfide bonds in polymer backbones.

Nucleic acid delivery

Different types of nucleic acids, such as plasmid DNA (pDNA), messenger RNA (mRNA), small interfering RNA (siRNA), and circular RNA (circRNA), vary significantly in molecular weights (e.g., 21–23 nt for siRNA, versus >100 nt for others), topological structures (single-stranded for mRNA and circRNA, double-stranded for pDNA and siRNA), as well as the intracellular target sites for active function. Due to the significant differences, there is limited understanding of the specific criteria for selecting delivery materials based on their chemical structure or composition that are most suitable for each type of nucleic acid. As a result, most nucleic acid delivery materials were developed mainly through a trial-and-error approach. Currently, lipid nanoparticles (LNPs) are recognized as the safest and most effective nucleic acid delivery systems, as evidenced by the FDA approval of the first siRNA drug for treating polyneuropathies and the success of three well-known mRNA vaccines, two Covid-19 vaccines (Comirnaty and Spikevax) and one respiratory syncytial virus vaccine (mRNA-1345)³¹. LNPs consist of four lipid components including ionizable lipids (ILs), phospholipids, cholesterol, and polyethylene glycol-conjugated lipids (PEG-lipids). Among these, the chemical design of ILs has attracted significant attention due to its critical impact on nucleic acid delivery efficacy. The typical structure of ILs includes three key moieties: ionizable heads, hydrophobic tails, and linkers connecting them³².

Recently, the MCRs also show great potential for discovering highly efficient ILs for therapeutic nucleic acid delivery. In 2019, Miao et al. reported the use of three-component Ugi reaction (Ugi-3CR) to screen a large library of ILs (Fig. 4a i)³³. They used the Ugi-3CR to simultaneously incorporate 12 amines, 9 alkyl ketones and 10 isocyanides to obtain 1080 ILs. These ILs were designed with hydrophobic chains ranging from 6 to 18 carbons and different degree of unsaturation and varied ionizable headgroups. Through high-throughput in vivo screening via intramuscular (i.m.) administration, they found the ILs with longer and unsaturated hydrophobic tails, a dihydroimidazole linker and cyclic amine head groups demonstrated superior delivery efficacy. The top IL A18-Iso5-2DC18 formulated LNPs exhibited the STING-mediated adjuvant effect, inducing antigen-presenting cell maturation and enhancing anti-tumor efficacy (Fig. 4a ii and iii). Using the same Ugi-3CR, Chen et al. reported the IL design for in situ muscle-targeted mRNA delivery. The finding demonstrated that the lead IL iso-A11B5C1-based LNPs exhibited high transfection efficiency in muscle tissues with minimized off-target effects in the liver and spleen³⁴.

Fig. 4: The MCR-based design of nucleic acid delivery materials.

a (i) The schematic to illustrate the Ugi-3CR-based combinatorial design of IL library. (ii) The structure for ionizable head screening by enzyme-linked immunospot (ELISpot) analysis of IFN-γ-spot-forming cells among splenocytes after ex vivo re-stimulation with SIINFEKL peptide in different LNPs. (iii) The structure of top IL A18-Iso5-2DC18. Reproduced with permission from Springer Nature³³. b (i) Synthetic method of the called “1 + 1 + 1 + 1” strategy in the Ugi-4CR. The lead IL A4I18R2C18-2 demonstrated efficient liver-targeted mRNA delivery via i.v. administration. (ii) Synthetic route of the called “n + 1 + 1 + 1” strategy in the Ugi-4CR. The lead IL A1I4R22C18-2 demonstrated efficient spleen-targeted mRNA delivery via i.v. administration. Reproduced with permission from the John Wiley and Sons²⁴. c (i) The schematic of CLP strategy for lung-tropism LNPs design through different pair manners between ILs and quaternary ammonium lipids. (ii) Structure-activity relationships of CLP-based LNPs shown lung-targeted mRNA delivery properties that correlated with the corresponding ILs. Reproduced with permission from the American Chemical Society³⁵. d (i) Combinatorial design of a GA molecular library based on the Ugi-4CR and the assembly process of pDNA/GA or mRNA/GA binary complexes, cGAMP/mRNA/GA ternary complexes, and LNP-like quinary complexes. (ii) Schematic illustration of GA-assisted DC vaccine manufacturing. (iii) Tumor volume curves and mice survival rates verified the success of gene-engineered DC vaccines. Reproduced with permission from the Elsevier⁴².

Very recently, we first reported a molecular library of ILs designed by the Ugi-4CR for the liver and spleen-selective mRNA delivery in vivo (Fig. 4b)²⁴. Initially, 60 ILs were synthesized using the called “1 + 1 + 1 + 1” strategy (each reactant is monofunctional used for the Ugi-4CR) by orthogonally combining four monofunctional components (4 × 3 × 5 × 1 = 60): 4 amines, 3 aldehydes, 5 carboxylic acids, and 1 isocyanide. These ILs were chemically introduced three functional moieties: (1) Tertiary amine-containing ionizable heads, including aliphatic and cyclic structures, supplied by amines; (2) Linker moieties with linear or cyclic aliphatic chains, supplied by aldehydes; (3) Saturated or unsaturated alkyl chains for hydrophobic tails, supplied by both isocyanide and carboxylic acids. These ILs could be formulated into mRNA-LNPs with the diameters between 50 and 150 nm, narrow size distributions (PDI < 0.3), and surface ζ potentials ranging from −10 to 5 mV. In vivo results indicated that most ILs primarily facilitated protein expression in the liver. Furthermore, the SARs analysis revealed that the ILs with aliphatic ionizable moieties outperformed those with cyclic one, and at least one unsaturated hydrophobic tail was crucial for improving in vivo mRNA delivery. The top liver-targeted IL, A4I18R2C18-2 (Fig. 4b i), consisted of an N, N-dimethylpropyl headgroup, a 1-ethylpentyl linker, and two octadecyl hydrophobic tails with two degrees of unsaturation. The cell types targeted in the liver was hepatocytes as verified by delivering Cre recombinase mRNA to Ai14 mice.

Next, another 60 ILs were synthesized using a bi- or trifunctional component instead in the Ugi-4CR, altering the reaction strategy from the called “1 + 1 + 1 + 1” to the “n + 1 + 1 + 1” (one of reactants is multifunctional used for the Ugi-4CR. n = 2 represents bifunctional, n = 3 represents trifunctional). This new IL library was also built by orthogonally combining 1 isocyanide, 4 aldehydes, 5 amines, and 3 carboxylic acids (1 × 4 × 5 × 3 = 60). These ILs were designed with three functional moieties: (1) Tertiary amine-containing bifunctional isocyanides as ionizable heads; (2) Linker moieties, either hydrophilic or hydrophobic, supplied by aldehydes and amines; (3) Hydrophobic tails containing saturated or unsaturated octadecyl chains, supplied by carboxylic acids. Interestingly, these ILs facilitated protein expression primarily in the spleen rather than the liver. The SAR analysis revealed that the linker moieties provided by aldehydes and amines played a significant role in spleen-tropism mRNA delivery, with uncharged linker moieties being particularly crucial (A1I4R22C18-2, Fig. 4b ii). We further expanded the ILs by introducing trifunctional head moieties and various types of hydrophobic tails, yet the spleen-specific mRNA delivery of ILs remained consistent. This suggests that the tissue tropism of ILs is largely determined by the skeleton of multifunctional ionizable head moieties.

We also reported that the ILs designed by the Ugi-4CR was paired with its derived quaternary ammonium lipids (QLs) to form a cationic lipid pair (CLP) for lung-tropism mRNA delivery (Fig. 4c)³⁵. In the new LNPs, the CLP was utilized to replace both ILs and phospholipids in the traditional LNPs. The tertiary amine of A4I18R2C18-2 was substituted with the methyl (Q₁), ethyl (Q₂), hydroxyethyl (Q₃), and hydroxyhexyl groups (Q₄) to produce four kinds of QLs. In vitro results indicated that the Q₁-substituted, CLP-based LNP (Q₁L/IL) demonstrated a stronger membrane-disruptive ability in the simulated endosomal environments and improved cellular uptake compared to other CLP-based LNPs, enabling the most efficient mRNA transfection. In vivo biodistribution and mRNA delivery following intravenous (i.v.) administration also demonstrated superior lung tropism of Q₁L/IL compared to other CLP-based LNPs. The primarily transfected lung cells were endothelial (39.5%) and epithelial cells (23.4%), with the minimal transfection observed in immune cells (4.5%), as evidenced by delivering Cre mRNA to Ai9 mice. Of note, Vlasova et al. utilized the Ugi-4CR to synthesize a library of cationic polymers for lung-targeted mRNA delivery³⁶. They used linear PEI derivatives as the backbone amine reactant, and combined with aldehydes, isocyanides, carboxylic acids to create functionalized cationic poly(ethylene imine) derivatives. A total of 155 cationic polymers were synthesized and formulated with mRNA into lipopolymer-lipid hybrid nanoparticles. The top-performing cationic polymer-based hybrid nanoparticles (U155@lipid) have a particle size of approximately 120 nm and positive ζ potentials around 10 mV, demonstrating efficient lung-targeted mRNA delivery properties, and effectively delivering mRNA to the lung endothelial and immune cells in Ai9 mice. As a result, U155@lipid successfully delivered mRNA therapeutics for enhancing treatment efficacy in two different lung disease models.

Very recently, Dong et al. reported the use of three-component Van Leusen reaction (Van Leusen-3CR) to establish a rational screening method for a spleen-targeted IL library²⁵. The Van Leusen-3CR involved in coupling of amines, aldehydes, and aryl-substituted tosylmethylisocyanides to produce imidazole-contained ILs under mild conditions, with no need for anhydrous or anaerobic environment. A library of 792 ILs was produced using 9 amines as headgroups, incorporating hydroxyl groups, saturated and aromatic rings with ionizable tertiary amines. In addition, 8 tosylmethylisocyanides with varying aromatic rings substitution were used, along with 11 aldehydes that featured varied alkyl chain lengths and the presence or absence of degradable ester bonds as hydrophobic tails. For high-throughput screening of ILs, the authors utilized a multidimensional orthogonal batch testing strategy. ILs containing the same ionizable heads or hydrophobic tails were grouped into batches for in vivo screening to optimize ionizable heads and hydrophobic tails, respectively, thus minimizing animal usage and reducing discovery costs. Most of imidazole-contained ILs demonstrated spleen-targeted mRNA delivery properties via i.v. administration. The top IL A3B7C2 was formulated into LNPs with particle sizes around 100 nm and negative ζ potentials around −3 mV, exhibiting better spleen targeting as compared to the spleen-targeted SORT-LNPs³⁷. Moreover, A3B7C2-based LNPs were verified to effectively deliver mRNA to spleen dendritic cells in Ai9 mice, highlighting its potential for spleen-targeted therapies.

In addition, Han et al. utilized a tandem multicomponent reaction (T-MCR) based on amine–thiol–acrylate conjugation between amines, 2-iminothiolane hydrochloride and alkyl acrylates to screen a series of amidine-incorporated degradable lipids (AID-lipid)²⁶. This T-MCR proceeded quickly at room temperature in the presence of triethylamine in ethanol with the minimal by-products. A library of 100 AID-lipids were obtained using 25 types of amines including alkylamines, anilines and hydrazides, reacted with 4 types of alkyl acrylates with different hydrophobic chain lengths. Through high-throughput in vivo screening after i.m. administration in mice, the top AID-lipid 12T-O14-based LNPs mediated efficient protein expression at the injection site without noticeable toxicity. 12T-O14 features a bulky benzene ring in hydrophobic tails, which likely enhance membrane disruption and thus endosomal escape by promoting AID-lipid to adopt a more conical shape. Moreover, the authors demonstrated that 12T-O14 could serve as a superior supplementary lipid to change organ selectivity of LNPs from the liver to the lung or the spleen through the formulation optimization. These facilitated CRISPR/Cas9-mediated gene editing in the lungs and mRNA vaccine delivery to the spleen. Isaac et al. utilized a facile, versatile and catalyst-free MCR platform to create a series of ILs with the new tetrahydropyrimidine (THP) backbones. By adjusting two independent primary amines, a small library of 26 THP TLs were synthesized using 5 amines as hydrophobic groups and 6 amines as ionizable groups. Through in vivo evaluation of mRNA delivery efficacy through i.m. administration in mice, the top IL THP1 demonstrated the highest transfection efficiency which was higher than the commercial IL MC3. The THP1-based LNPs showed minimal toxicity as demonstrated by hematological, histopathological and proinflammatory cytokine analysis³⁸.

The MCR-based combinatorial chemistry approach also shows great potential in discovering delivery vehicles applied for circRNA, siRNA, pDNA or gene-engineered cell therapy. Xu et al. utilized the Ugi-4CR to screen a series of ILs for i.v. and intratracheal delivery of interleukin-12 (IL-12) encoded circRNA³⁹. They constructed a 96 ILs library by combining 2 amines, 3 isocyanides, 4 aldehydes and 4 carboxylic acids. The top IL, H1L1A1B3, successfully encapsulated IL-12 circRNA into LNPs with high encapsulation efficiency (~90%) and good colloidal stability. Further studies in multiple lung cancer mouse models demonstrated potent efficacy of this circRNA delivery system for anti-tumor therapy. Molla et al. utilized a three-component thiolactone chemistry to screen a series of ILs for siRNA delivery^40,41. They used nucleophilic ring-opening of thiolactones followed by disulfide exchange to establish a one-pot, three-component synthesis method. Through the combination of amines as ionizable heads with pyridyl disulfide and thiolactone derivatives as hydrophobic tails, these synthesized ILs incorporated biodegradable disulfide bonds and double amide backbones, successfully delivering both siRNA and pDNA into cells.

Gene-engineered cell therapy holds great promise for preventing or treating major diseases due to its high specificity, potential for personalized treatment, and long-lasting therapeutic effects. For example, ex vivo DNA or mRNA-pulsed dendritic cell (DC) vaccines have demonstrated the ability to continuously produce antigens. However, these vaccines face several challenges, including low gene delivery efficiency and insufficient DC maturation and cytokine release, which limits antigen cross-presentation and T cell activation. To address this limitation and thus advance the development of DC vaccines for a broader range of clinical applications, our lab recently developed a GA materials library for enhancing DC transfection and vaccine production (Fig. 4d)⁴². Through materials screening and the SAR analysis, we identified the top GA material can form stable binary complexes with pDNA or mRNA, exhibiting strong binding affinity at a low mass ratio (less than 2:1 for GA to nucleic acid). These DNA/GA and mRNA/GA complexes were efficiently endocytosed by DCs, releasing the cargo into the cytosol after endo/lysosomal escape. As a result, they achieved significantly higher transfection efficiency compared to the commercial transfection reagent Lipofectamine 2000. The mRNA/GA binary complexes promoted DC maturation when delivering mRNA encoding model protein antigen. Further investigation revealed that GA significantly stimulated type I interferon expression through the intracellular STING (stimulator of interferon genes) pathway, enhancing both DC maturation and activation, which is critical factor for increasing the immunogenicity of DC vaccines. In addition, GA demonstrated the ability to co-deliver the STING agonist cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) by the formation of cGAMP/mRNA/GA ternary complexes, which further boosted DC activation and maturation. Adoptive transfer of DC vaccines pulsed with these ternary complexes led to robust T cell activation and accumulation in a B16F10-OVA allograft tumor model, ultimately resulting in significant tumor growth inhibition and prolonged survival in mice. In addition, due to structural similarity of GA to ILs, GA was used to construct LNP-like quinary complexes. When formulated with mRNA, DOPE, DMG-PEG, and cholesterol, the quinary complexes improved mRNA transfection in DCs by 6.9-fold compared to Lipofectamine 2000. Therefore, the GA molecular library offers a valuable platform for optimizing binary, ternary, and LNP-like quinary complexes, significantly enhancing mRNA or pDNA transfection and facilitating the production of potent gene-engineered DC vaccines.

Although high-throughput screening is a viable approach for discovering high-performing delivery materials, this method can be labor-intensive and costly due to the vast array of combinatorial chemical structures generated by the MCRs. The rise of artificial intelligence (AI) technology provides a powerful solution for analyzing extensive datasets and further optimizing the structure of delivery materials. Currently, AI technology has already been applied to the discovery of mRNA delivery systems by screening hundreds of thousands of virtual ILs. Xu et al. developed an AI-guided ionizable lipid engineering (AGILE) platform combined with the Ugi-3CR to accelerate ILs discovery for mRNA delivery (Fig. 5a i)⁴³. Using the AGILE platform, the author generated a virtual library of 60,000 ILs, initiating self-supervised model training and then refining the model accuracy through supervised fine-tuning with empirical data from a synthesized library containing 1200 ILs. Following the retrained algorithms, the model was then used for predictive analysis on a candidate library of 12,000 virtual ILs, followed by ranking for final candidate selection. The top 15 IL candidates were evaluated for i.m. administration of mRNA-LNPs, with the lead IL H9 (Fig. 5a ii and iii) showing superior mRNA delivery efficiency compared to the established benchmarks like the MC3 and ALC-0315 (Fig. 5a iv). This work represents a groundbreaking fusion of the MCR and AI technology, broadening insights into the development of mRNA delivery materials. In addition, Li et al. extended this AI technology to the Ugi-4CR for IL screening (Fig. 5b)⁴⁴. From an experimental library of 584 ILs, they used the best-performing model to probe a virtual library of 40,000 ILs, evaluating the top 16 predicted ILs. The lead IL 119-23 (Fig. 5b ii) was identified to outperform the commercial lipids such as MC3 and SM-102 by i.m. administration. Very recently, Jacob et al. demonstrated the application of neural networks and deep learning for structure optimization in the discovery of high-performing ILs⁴⁵. In this report, the Ugi-4CR was used to establish a virtual IL library containing 86,400 lipids, a collection not previously reported. Utilizing an optimized AI model, they predicted their mRNA delivery efficacy and identified the top ILs FO-32 and FO-35 with high-performing local mRNA delivery to the mouse muscle and nasal mucosa. These works highlight the cutting-edge potential of AI technologies in exploring next-generation ILs for in vivo mRNA delivery.

Fig. 5: AI-guided discovery of delivery materials through the MCR-based combinatorial chemistry.

a (i) Schematic of the workflow of AGILE platform. First, a virtual IL library was used to pre-train the model, then synthesis of an experimental IL library for the fine-tuning of the model. Finally, the fine-tuned model was used to predictive ILs on a candidate library. The virtual IL library was design through the Ugi-3CR. (ii) In vitro transfection of 15 IL candidates predicted by the AGILE. (iii) The structure of the lead IL H9. (iv) In vivo evaluation of the H9 and the commercial IL MC3 and ALC-0315 through i.m. administration. Reproduced with permission from Springer Nature⁴³. b (i) Schematic of IL library designed by the Ugi-4CR to generate data for training machine learning algorithms. (ii) The structure of the lead IL 119-23. (iii) Comparison of mRNA delivery efficacy of 119-23 with the commercial MC3 and SM102 through i.m. administration. Reproduced with permission from Springer Nature⁴⁴.

Future outlook

LNP-based carrier materials achieved significant success for nucleic acid delivery in recent clinical developments. In addition to well-known therapies like the siRNA drug (Onpattro) and mRNA vaccines (Comirnaty and Spikevax), other LNP-based RNA therapeutics, including cancer vaccines, protein replacement therapies or gene editing, have advanced beyond phase I clinical trials⁴⁶. Of note, although many protein delivery materials are developed, to the best of our knowledge, no available carrier material for intracellular protein drug delivery has yet been approved for clinical trials until now. Recently, the rational design of carrier materials using the MCR-based combinatorial chemistry provide the chance to further improve in vitro and in vivo delivery efficacy of therapeutic proteins or nucleic acids. However, in this field, the challenges still remain in the development of safe and efficient biomacromolecule delivery systems for various application requirements.

Explore new multicomponent chemistry approaches

Although current works described the MCRs for the design of lipid or polymer-based carrier materials for protein or nucleic acid delivery, most of them have mainly relied on isocyanide-mediated MCRs, such as the Ugi-3CR and the Ugi-4CR. To increase structural diversity in delivery materials and further broaden the application potential for various biomacromolecular therapeutics, other types of isocyanide-mediated and non-isocyanide-based MCRs should be explored. For examples, amine-formaldehyde-thiol-based catalyst-free three-component reaction is suitable for generating sulfur- and nitrogen-contained products⁴⁷. The Mannich multicomponent reaction is widely used for the synthesis of β-amino carbonyl products. The Biginelli multicomponent reaction can combine aldehydes, β-ketoesters, and urea to form dihydropyrimidinone products. Predictably, these chemical tools enable the facile and efficient synthesis of diverse delivery materials in short time, without requiring harsh conditions.

AI-guide structural design and optimization of delivery materials

Currently, the AI technology has been utilized to combine with materials science for accelerating the discovery of effective biomacromolecule delivery carriers. For example, AI technology was used to screen virtual IL libraries and LNPs for mRNA delivery, offering the potential to greatly increase efficiency of materials discovery process, with reducing animal usage. Despite the advances, significant challenges remain to be addressed. One major issue is the inaccuracy of predictions. Most existing machine learning model rely on these results of in vitro high-throughput screening tests, but these often showed a weak correlation with in vivo efficacy outcomes. This discrepancy eventually leads to inaccurate predictions in the subsequent animal study. Moreover, current predictive models largely focused on intramuscular delivery, which mainly involved local drug delivery efficacy. In contrast, systemic administration such as intravenous delivery is more complex due to the complicated physiological environments, which make AI difficult to predict organ tropism or in vivo targeted cell types of delivery systems. In addition, there is an urgent requirement for standardizing data collection to facilitate accurate AI analysis. The lack of criteria for quantifying delivery performance and efficacy of different carrier systems hampers the comparison of results from different labs, limiting the effectiveness of AI technology. Therefore, more efforts are needed to make for AI combined with delivery materials discovery, further improving their accuracy of predictions for various in vivo applications.

Enhance intracellular delivery bioavailability and safety

One of the formidable challenges for biomacromolecule delivery systems is their low delivery efficiency in cellular internalization and subsequent endo-lysosomal escape. For instance, even clinically successful nucleic acid delivery systems like LNPs only exhibit endosomal escape efficiency less than 2%⁴⁸. Carrier materials with enhanced endo-lysosomal escape capability are expected to further improve the bioavailability of biomacromolecular therapeutics. Alternative strategy that bypass classic endocytosis pathways, such as direct cytosolic delivery through the membrane fusion, was proved effective and thus suggested to develop in the near future. Adjusting degradability of carrier materials is also crucial for ensuring efficient intracellular cargo release, thus improving bioavailability while minimizing side effects. Future works should prioritize incorporation of degradable linkages like hydrolyzable or enzymatically cleavable bonds into delivery materials, which can break down into non-toxic products, thus decreasing immunogenicity and inflammatory response. In addition, biomimetic delivery materials, such as polypeptides and extracellular vesicles, show promise in reducing the risk of material accumulation and toxicity in vivo.

In summary, MCR-based combinatorial chemistry has showed significant advantages in accelerating the development of protein and nucleic acid drug delivery systems. Future research should explore new multicomponent combinatorial chemistry, and improve AI-based predictive approaches for functional screening and discovery of efficient delivery materials. With the aid of structure-function insights and AI technology, the next-generation of delivery materials that tailored for peptides, proteins, nucleic acids or other biomacromolecules will hold great potential for facilitating the clinical translation of intracellular biomacromolecular therapeutics.