mRNA therapeutics: Transforming medicine through innovation in design, delivery, and disease treatment

Abstract

mRNA therapeutics have revolutionized medicine, offering a versatile platform to address previously untreatable diseases. Recent advancements in biotechnology have enabled the efficient production of functional proteins, antibodies, and peptides via mRNA, providing rapid and adaptable solutions for vaccine development and therapeutic interventions. The success of mRNA vaccines, exemplified during the COVID-19 pandemic, underscores their potential for combating infectious diseases with unparalleled speed and scalability. This review explores the latest developments in mRNA technology, including innovations in design, delivery, and disease treatment; modulation of immune response; and the role of AI. Particular emphasis is placed on optimizing mRNA constructs to maximize therapeutic efficacy, new delivery vehicles for mRNA, and the modulations of immune response evoked by mRNA vaccines. The potential applications of mRNA therapeutics in genetic disorders, infectious diseases, and cancer are highlighted, alongside a discussion of existing challenges such as delivery efficiency and production scalability. By integrating molecular biology, RNA technology, and nanotechnology, mRNA therapeutics hold the promise of transforming precision medicine. These advancements offer hope for patients with complex or intractable conditions, paving the way for a new era in targeted therapies and personalized healthcare.

Graphical abstract

The design and delivery of mRNA into living systems (especially humans) are being optimized from time to time to ensure that safety, stability, and immunogenicity are optimal depending on the various contexts of disease treatment. Advanced AI and machine learning methods and tools are, therefore, needed to achieve better results.

Introduction

In recent years, mRNA has regained prominence due to breakthroughs in biotechnology and the emergence of novel therapeutic strategies. The success of mRNA vaccines, particularly during the COVID-19 pandemic, has underscored the versatility and efficacy of mRNA platforms in addressing infectious diseases.¹ This achievement was largely driven by advancements in base sequence modifications and the development of lipid nanoparticles (LNPs) as a safe and effective mRNA delivery system.²

Despite its transformative potential, mRNA therapeutics face significant challenges. A primary concern is the inherent instability of mRNA, which is susceptible to rapid degradation by ribonucleases. While modifications such as pseudouridine (Ψ) incorporation and optimized capping techniques have reduced immunogenicity, enhanced stability, and improved translation efficiency, they also complicate synthesis and escalate production costs.

Delivery systems, particularly LNPs, have revolutionized mRNA therapeutics but still face limitations in achieving tissue-specific targeting and minimizing off-target effects. Additionally, the immunogenicity of both mRNA and its delivery vehicles necessitates precise modulation to balance therapeutic efficacy with safety, as excessive immune activation can lead to adverse outcomes. These challenges highlight the need for ongoing innovation in RNA chemistry, delivery technologies, and scalable manufacturing processes to realize the full potential of mRNA-based therapies.

This review, therefore, aims to summarize the latest advances in mRNA research, focusing on areas such as mRNA sequence modifications, mRNA types and delivery systems, the immune response generated by mRNA vaccines, as well as the future directions. The advancements in mRNA technology as shown in Figure 1 suggest that there is more room for developments.

Figure 1.

Timeline of key discoveries and advances in mRNA-based therapeutics

The development of mRNA therapeutics in the light of design, delivery, and disease treatment can be grouped into three distinct phases. Phase 1 (1961–1990) includes the identification of mRNA,³ the use of protamine for RNA delivery,⁴in vitro translation of isolated mRNA,⁵ discovery of the mRNA cap structure,⁶ liposome-encapsulated mRNA delivery,⁷ commercialization of cap analogs and T7 RNA polymerases, and cationic lipid-mediated mRNA delivery.⁸ Phase 2 (1990–2020) involves the use of mRNAs for cancer immunotherapy,⁹ establishment of an mRNA-based company and discovery that the 3′ UTR regulates mRNA localization,¹⁰ induction of antitumor T cell responses by mRNA,¹¹ first clinical trial with mRNA using ex vivo-transfected DCs,¹² development of mRNA-based immunotherapy for human cancer,¹³ preclinical study with intranodally injected DC-targeted mRNA,¹⁴ protective mRNA vaccines against influenza¹⁵ and respiratory syncytial virus (RSV),¹⁶ application of CRISPR-Cas9 mRNA for gene editing,¹⁷ and advancement of personalized mRNA cancer vaccines into clinical trials.¹⁸ Phase 3 (2020-present) includes clinical trials of mRNA-based vaccines for cancer and infectious disease, mRNA-1273,¹⁹ and BNT162b emergency use for SARS-CoV-2 pandemic.²⁰

Advances in mRNA sequence design

Since mRNA molecules are needed to be at best performances in vitro or in vivo, their design involves computational and experimental strategies that ensure stability, translation efficiency, and immune compatibility. Designing an optimal mRNA sequence ensures effective therapeutic outcomes by balancing expression levels and minimizing unwanted immune activation.²¹ mRNA modification involves chemical modifications to the mRNA molecule to enhance its functionality and reduce immunogenicity. Common modifications include nucleotide modifications, cap structure enhancements, and base pair modifications.²² Research developments have led to the construction of linear, circular, self-amplifying, and trans-amplifying mRNAs, which differ in their structure and by extension, expression, or translation levels.²³ mRNA synthesis refers to the processes involved in generating functional mRNA, which include template preparation, in vitro transcription (IVT), and purification.²⁴ The advances in the technologies of each of these concepts would be discussed in detail in this section.

mRNA sequence design

mRNA comprises five critical functional regions: the 5′ cap, the 3′ poly(A) tail, the open reading frame (ORF), and the 5′ and 3′ UTRs.²⁵ Each of these elements plays a pivotal role in regulating the translation efficiency and stability of the mRNA molecule. The 5′ cap, a modified guanine nucleotide, is essential for the initiation of translation and protects the mRNA from exonucleases.²² The 3′ poly(A) tail, consisting of a stretch of adenine nucleotides, also enhances stability and facilitates the export of mRNA from the nucleus to the cytoplasm.²⁶ The ORF contains the codons that are translated into the protein sequence, while the flanking 5′ and 3′ UTRs contain regulatory sequences that modulate translation and mRNA degradation.²⁶

5′ cap

Recent advances, including stereochemically optimized phosphorothioate caps and α-β/β-γ phosphate modifications, enhance performance in RNA vaccines. Beyond m7G, technologies like anti-reverse cap analogs (ARCAs), CleanCap, and Cap-1/Cap-2 optimize capping efficiency, reduce immunogenicity, and mimic natural mRNA. Emerging strategies, such as trinucleotide analogs, NAD/FAD caps, and ligand-conjugated caps, expand applications in immunotherapy and targeted delivery.

Innovation in 5′-cap design has progressed from a simple chemical add-on to the decisive determinant of synthetic mRNA potency. The canonical 7-methyl-guanosine (m⁷G) cap was originally viewed as a binary requirement—present or absent—for nuclear export and ribosome recruitment, yet within the cytosol, it is dismantled within minutes by the Dcp1/2 decapping complex and the DcpS scavenger hydrolase.²⁷ The earliest corrective strategy, ARCAs, merely inverted the orientation of the triphosphate to prevent incorporation in the wrong direction; even this modest tweak doubled translational yield in dendritic cells.²⁸ Since then, every accessible atom of the cap has been interrogated, generating a spectrum of analogs whose functional differences now outweigh their structural similarities.^29,30

Bridge modifications have provided the most dramatic gains in stability. Substitution of an oxygen in the α-β phosphoanhydride with a methylenebisphosphonate lengthens the half-life 4- to 5-fold in cytosolic extracts, but the same modification slightly diminishes translation, hinting that the eIF4E clamp is sensitive to backbone geometry.^31,32 Phosphorothioates, in which sulfur replaces a non-bridging oxygen, perform better: translational output increases 2-fold, while stability rises 6- to 8-fold, and the stereopure Sp isomer evades recognition by IFIT1, thereby dampening innate immune sensing.^28,33 Tetraphosphate extensions (m⁷Gppppm⁷G) elevate eIF4E affinity 3-fold and protein yield 2.5-fold, but the extra charge accelerates DcpS hydrolysis unless further locked by stereopure phosphorothioate linkages. Dithiodiphosphate caps—essentially a disulfide-locked triphosphate—surpass all others in durability (10-fold longer half-life) with only a modest drop in initiation efficiency, making them attractive for applications where prolonged antigen expression is desired.^34,35

Head group alterations have revealed that tighter eIF4E binding does not always correlate with better translation. 7-benzyl-guanine caps (BN⁷mGpppG) reduce equilibrium affinity yet accelerate eIF4E release, yielding 2- to 3-fold higher protein in vivo.³⁶ Co-transcriptional trinucleotide caps (CleanCap AG) achieve ∼94% Cap-1 occupancy, and the 2′-O-methyl on the +1 adenosine both stabilizes the helix and improves 43S pre-initiation complex positioning, explaining the robust 2- to 3-fold boost in protein expression observed across multiple pre-clinical models.³⁷ CleanCap integrates capping with transcription, streamlining synthesis, while minimizing immunogenicity with modifications being done to the structure to continue improving its efficacy—an advantage that has already led to its adoption in licensed SARS-CoV-2 vaccines.^38,39

Immunogenicity, long treated as a downstream formulation issue, has become a primary design filter. Unmethylated Cap-0 RNA is sensed by RIG-I and IFIT1, triggering type I interferon (IFN) and protein kinase R (PKR)-mediated translational shutdown. Addition of a single 2′-O-methyl (Cap-1) reduces RIG-I activation by >80% and abrogates IFIT1 binding entirely.^40,41 CleanCap AG and enzymatic vaccinia capping enzyme (VCE) both install Cap-1 in >90% of transcripts, yet VCE remains labor intensive and yields batch-to-batch variability in Cap-1 content (78%–92%), whereas CleanCap lots routinely achieve ≥94% Cap-1 with undetectable Cap-0.^42,43 Residual immunogenicity of CleanCap mRNA is, therefore, limited to double-stranded RNA (dsRNA) contaminants rather than cap chemistry itself.

Head-to-head comparative data (Table 1) were generated under identical reporter mRNA, LNP formulation, and dosing in BALB/c mice. Composite caps that combine N7-benzyl substitution, stereopure phosphorothioate bridges and Cap-1 methylation emerge as the current Pareto optimum, doubling protein expression, while reducing systemic cytokine release by an order of magnitude.³⁶

Table 1.

Core modifications and functional properties of 5′cap structures

Core modification	Notable structure(s)	eIF4E affinity (Kd-fold vs. m7GpppG)	Half-life in cytosolic extract (t½-fold)	In vitro translation boost (RLU-fold)	Reported immunogenicity
Triphosphate bridge	α-β methylenebisphosphonate (CH2)	0.9	4–5×	0.7	Low
Phosphorothioate (α or β)	m7G-PS-ppG	1.1	6–8×	2.0	Low (IFIT1 evasion)
Tetraphosphate extension	m7Gppppm7G	3.2	2×	2.5	Moderate
7-enzylguanine	BN7mGpppG	2.1	3×	3.0	Very low
Dithiodiphosphate	m7G-S-S-ppG	1.3	10×	1.8	Low
Trinucleotide (CleanCap AG)	m7GpppAm2′-O-Ψ	1.4	4×	2.1	Ultra-low

Clinical translation remains anchored to enzymatic capping (VCE +2′-O-methyl-transferase) or co-transcriptional CleanCap AG.^44,45 Next-wave candidates will demand lower effective doses or weeks-long expression windows, forcing the field to adopt composite caps now confined to pre-clinical studies. Key uncertainties include scalable GMP synthesis of multi-modified caps at ≥95% purity without post-transcriptional enzymatic steps and head-to-head toxicology in non-human primates comparing CleanCap AG, BN⁷mGpppG-S-S, and stereopure phosphorothioate tetraphosphate caps (see Table 1). Conditional caps—photocleavable or pH-labile linkers—and ligand-conjugated targeted delivery caps already hint at a future in which the 5′ end is a programmable switch for potency, durability, and cell-type specificity.^30,46

5′ UTR

UTRs at the 3′ and 5′ ends of mRNAs, although not directly involved in encoding proteins, play pivotal roles in controlling mRNA translation and protein synthesis. The 5′ UTR is mainly responsible for initiating translation, while the 3′ UTR predominantly regulates mRNA stability and half-life.⁴⁷ Collectively, these regions are essential for the overall efficiency and regulation of gene expression.⁴⁸

The 5′ UTR is crucial for translation initiation, containing elements that facilitate ribosome binding and recognition of the start codon. The cap structure at the 5′ end recruits eukaryotic initiation factor 4E (eIF4E), which is instrumental in ribosome binding. Additionally, the 5′ UTR may contain internal ribosome entry sites (IRESs), enabling cap-independent translation and expanding the conditions under which translation can proceed.⁴⁸ Ferizi et al.⁴⁹ evaluated 5′ UTRs from five naturally long-lived mRNAs, identifying the human CYBA UTR as a top performer in enhancing and stabilizing protein expression in A549 and NIH3T3 cells. This foundational study underscored the potential of leveraging natural UTRs for mRNA design, providing a baseline for further modifications.

Asrani et al.⁵⁰ demonstrated the importance of UTR optimization by constructing a library of 10 UTR variants, identifying complement factor 3 (C3) and cytochrome P450 2E1 (CYP2E1)-derived UTRs as enhancers of therapeutic mRNA translation efficiency, independent of 3′ UTR modifications. Similarly, refining the Kozak sequence within the human beta-globin 5′ UTR significantly improved protein expression, emphasizing its pivotal role in translation initiation.⁵¹ Furthermore, studies showed that engineering short 5′ UTRs with minor Kozak sequence modifications can achieve or surpass the performance of natural UTRs, underscoring the potential of minimalistic, rational designs.⁵²

However, recent innovations in 5′ UTR design have converged on three complementary approaches—deep-learning-guided de novo generation, massively parallel functional mapping, and chemistry-aware optimization—collectively pushing mRNA translation control far beyond the classical paradigm of strong Kozak sequences paired with natural UTRs.

Generative models now autonomously learn the regulatory grammar of 5′ UTRs without reliance on human-encoded rules. For instance, UTRGAN, a conditional generative adversarial network trained on over 10,000 endogenous human 5′ UTRs, produces synthetic sequences that enhance mean ribosome load by more than 2-fold and reporter expression by over 5-fold while preserving known regulatory motifs such as IRES elements, upstream ORFs (uORFs), G-quadruplexes, and Kozak sequences.⁵³ Another model, UTR-Insight, combines a pre-trained RNA language model (based on the ESM-2 architecture) with a CNN-Transformer hybrid to predict the mean ribosome load (MRL) for variable-length sequences. When integrated into a genetic algorithm, it generated the sequence UTR_r2_29, which surpassed both the canonical hHBA UTR and the best endogenous high-MRL controls, elevating GFP fluorescence by 82.9%.⁵⁴ Additionally, Smart5UTR addresses the impact of chemical modifications, recognizing that N1-methyl-pseudouridine (m1Ψ) alters the optimal 5′ UTR landscape. By training a deep generative model specifically on m1Ψ-modified mRNA datasets, the platform produced UTRs that enhanced anti-Delta and anti-Omicron antibody titers in mice beyond those achieved with high-expression endogenous UTRs.⁵⁵

Furthermore, systematic variant scanning is uncovering how single-nucleotide and structural changes influence protein output in a cell-type-specific manner. A recent study employed a massively parallel reporter assay (MPRA) in glutamatergic cortical neurons (P21) to quantify the translational effects of over 1,000 patient-derived 5′ UTR variants linked to autism. By improving barcode design and incorporating Cre-inducible cell-of-origin tagging, the study identified nine neuron-specific functional alleles in LRFN5, LRRC4, and ZNF644 that were undetected in conventional cell line assays.⁵⁶ These findings highlight the necessity of in vivo, cell-type-matched screening for uncovering context-dependent regulatory elements.

Rational pairing of 5′ and 3′ UTRs, along with chemical modification-aware design, has become a standard practice. A combinatorial screen in HEK293T cells identified the synthetic 5′ UTR “5UTR05,” which, when combined with the IGHG2 3′ UTR and the mitochondrial mtRNR1 3′ UTR, increased translation efficiency by over 130% compared to the mRNA-1273 benchmark.⁵⁷ The same study confirmed that 5UTR05 maintains its advantage regardless of m1Ψ incorporation, demonstrating that top-performing AI-designed 5′ UTRs remain robust across different mRNA manufacturing chemistries.⁵⁷

Together, these advances establish a streamlined workflow for 5′ UTR optimization: deep-learning generators propose candidate sequences, MPRA or cell-type-specific assays filter for context-dependent activity, and combinatorial 3′ UTR pairing with modification-aware validation finalizes the design. The result is an unprecedented level of translational control and cell-type specificity, leading to higher therapeutic protein yields.

3′ UTR

The 3′ UTR significantly influences mRNA stability and translation duration. This region contains elements that affect mRNA stability and half-life, such as pyrimidine-rich sequences and regulatory motifs that interact with RNA-binding proteins (RBPs). For instance, the 3′ UTR of α-globin contains a discontinuous pyrimidine-rich sequence that enhances mRNA stability, while β-globin sequences contribute to prolonged protein expression.⁵⁸

Studies show that appending two consecutive β-globin 3′ UTRs can boost protein production and mRNA stability, with effects varying by cell type. This strategy enhances protein expression in mature dendritic cells but is less effective in immature ones, while in human pluripotent stem cells, β-globin 5′ UTRs outperform dual β-globin 3′ UTRs.^59,60 Schrom et al.⁶¹ reported higher protein expression using optimized coding sequences (CDSs) paired with minimal 5′ UTRs, human alpha-globin 5′ UTRs, and CYBA 5′ UTRs, underscoring the interplay between 5′ and 3′ UTRs.

Recent breakthroughs in 3′ UTR design have transcended traditional approaches like the β-globin tandem repeat, instead leveraging three key strategies: AI-guided de novo synthesis, cell-type-specific combinatorial screening, and programmable RBP modulation. Together, these methods enable nucleotide-level control of mRNA half-life and translational efficiency across diverse manufacturing platforms and target tissues.

Modern deep-learning models now decode the cis-regulatory logic linking 3′ UTR sequence to mRNA stability. Yang et al.⁶² trained a transformer-based encoder-decoder on >90,000 endogenous 3′ UTRs, demonstrating that the model’s latent space cleanly segregates stabilizing from destabilizing motifs. This enabled the design of synthetic 3′ UTRs that outperform the dual β-globin benchmark by 2.1-fold in HEK293T cells and 3.7-fold in primary human T cells. Similarly, Morrow and colleagues trained a gradient-boosted model on >120,000 native 3′ UTRs to predict mRNA half-life (R² = 0.78) and then used a genetic algorithm to design 60 synthetic 3′ UTRs; eight surpassed the dual β-globin benchmark, with SynUTR-17 extending reporter mRNA half-life 2.4-fold and boosting liver protein output 1.9-fold in mice, all without triggering innate immune sensors.⁶³

High-throughput assays now uncover context-dependent regulatory rules invisible in bulk analyses. Orlandini von Niessen et al.⁶⁴ screened 25,000 fragmented 3′ UTRs in dendritic cells, identifying a 114-nucleotide (nt) mitochondrial 12S rRNA fragment that doubles luciferase expression—but only in mature dendritic cells (DCs), highlighting cell-state specificity. MPRAs further reveal that AU-rich elements (AREs) exert opposing effects based on their distance from the poly(A) tail, a positional rule now quantified by the machine learning (ML) tool ARE Score.⁶⁵

Rather than deleting destabilizing motifs, newer approaches engineer high-affinity RBP docking sites. Ma et al.⁶⁶ inserted a 27-nt optimized ARE (AUUUA repeat) upstream of the stop codon, recruiting HuR to extend mRNA half-life 2- to 3-fold and boost protein output up to 5-fold without activating PKR. The effect is dosage dependent: expanding the ARE tract from two to five repeats progressively enhances stability, offering a tunable “dial” for expression duration.

These designs remain effective in m1Ψ-modified mRNA, confirming compatibility with clinical formulations. Emerging innovations include “split 3′ UTR” architectures, where a stabilizing core is linked to a destabilizing ARE cassette via a ribozyme, enabling small-molecule-controlled mRNA clearance.⁶⁷ The integration of AI design, cell-type-specific screening, and programmable RBP recruitment establishes a robust framework for engineering 3′ UTRs with tailored stability and expression kinetics. This precision engineering promises to enhance the therapeutic index of mRNA vaccines and medicines.

Coding sequence

The CDS, also known as the ORF, is a fundamental component in mRNA therapeutics, as it encodes the target antigen or protein. The CDS typically represents the majority of the mRNA transcript, making its optimization essential for improving mRNA stability and translation efficiency.^68,69 A primary strategy for optimizing the CDS is codon optimization, which involves replacing rare codons with synonymous, more-frequently used codons in the host organism. This approach not only boosts protein production but also enhances the fidelity of translation.^70,71

The significance of codon optimization is evident in its influence on the translation rate, which is determined by the availability of tRNA species. By favoring common tRNAs, codon optimization enhances both the efficiency and accuracy of protein synthesis. Narula et al.⁷¹ demonstrated that codon optimization leads to increased protein yields and more consistent expression levels across various cell types.

Beyond codon optimization, incorporating modified nucleosides such as the Ψ within the CDS is a widely used strategy to minimize innate immune responses. This innovation, pioneered by the Nobel prize-winning work of Kariko and Weissman has led to the enhanced stability and reduced immunogenicity of mRNA molecules.^72,73 However, the impact of these modifications on translation is complex and context dependent. In particular, a study found that N6-methyladenosine (m6A) modifications can inhibit translation by hindering tRNA association.⁷⁴

Contrarily, another study observed that m6A modifications can enhance translation by reducing mRNA secondary structure, thereby facilitating ribosome movement.⁷⁵ Similarly, Eyler group reported that Ψ modifications can alter tRNA binding, resulting in decreased translation efficiency and increased amino acid substitutions.⁷⁶ This, therefore, calls for the use of computational methods and experimental validation in the use of these base modifications. Base nucleotide modifications are further discussed in mRNA modifications section.

Poly(A) tail

Poly(A) tail engineering in modulating translation efficiency and protein expression has advanced, with studies demonstrating that longer tails (e.g., 100–120 nucleotides) significantly enhance protein output—up to 35-fold compared to shorter tails—by stabilizing the closed-loop interaction between the 5′ cap, eIF4G/E, and poly(A)-binding proteins.⁷⁷ However, the relationship between tail length and stability is nuanced: while traditional thresholds suggested a minimum of 20 nucleotides, some mRNAs (e.g., β-actin) remain functional with shorter tails, and excessively long tails (e.g., 425–525 nucleotides) offer no additional benefit in certain cell types like primary T cells.^78,79 Innovations such as recombinant poly(A) polymerase and template-designed DNA now enable precise tailoring of poly(A) lengths, optimizing mRNA therapeutics for enhanced translational control.⁸⁰

The current conception of the poly(A) tail as a static stabilizer has been eroded by two converging lines of evidence. First is the demonstration that an interrupted, fragmented tail—formalized in the BioNTech 3×(A30C5) cassette and now standard in BNT162b2—outperforms a classical 110-nt homopolymer without increasing length or CpG content.^81,82 Length alone no longer explains pharmacology. Classic linear tails of ∼100 adenosines—the default encoded in BNT162b2 and in the original pEVL plasmid system—support robust translation for only 6–8 h in murine liver, whereas tails extended to 200–250 nt by post-transcriptional elongation (C3P3-G2) or by branched architectures remain translationally active for at least 14 days, cutting the effective dose required for 80% liver editing from 2 mg kg⁻¹ to 0.7 mg kg⁻¹.⁸³ Importantly, the branched topology itself, rather than increased adenosine content, is causal; reversing branch polarity abolishes the benefit, and chemically stabilized 30-mer branches on a 30-mer stem outperform a single 150-mer linear tail.

Chemical composition now rivals length as a design variable. Ninetails nanopore profiling across four independent datasets reveals that cytidine, guanosine, and uridine are routinely incorporated during both enzymatic Poly(A) Polymerase (PAP) and T7 polymerase tailing, but with enzyme-specific signatures: T7 predominantly misincorporates uridine, whereas PAP yields cytidine-rich tails when rNTP concentrations are equimolar.⁸⁴ Misincorporation is not noise; tails containing 3%–5% non-adenosines are enriched among transcripts that accumulate in macrophages after Moderna mRNA-1273 vaccination, and TENT5A-mediated readenylation preferentially appends adenosine to uridine-containing substrates, generating heterogeneous “mixed tails” that persist 24–48 h longer than pure adenosine counterparts.^85,86 Conversely, engineered uridine-free tails produced with the G47A + 884G T7 mutant reduce innate immune activation 2-fold in human dendritic cells.⁸⁶

Manufacturing route therefore dictates tail identity and, by extension, potency. Standard IVT templates encoding 110–120 A residues suffer from template instability and recombination; segmented A40-A40-A40 tails reduce plasmid recombination without compromising translational output,⁸⁷ while the pEVL linear plasmid system now enables encoded tails up to 300 A residues with ∼90% homogeneity.⁷⁸ Alternatively, enzymatic post-transcriptional tailing, exemplified by the C3P3-G2 system, in which a tethered, cytoplasm-relocalized PAPα mutant extends nascent 40 A tails to 250 nts, circumvents plasmid size constraints and offers a single-ORF solution compatible with lentiviral or LNP delivery.⁸⁸

Immunogenicity and safety remain tightly coupled to tail design. Branched PS/2MOE-terminated tails do not elevate serum tumor necrosis factor alpha (TNF-α), aspartate aminotransferase (AST), or alanine aminotransferase (ALT) above poly(C) controls, whereas cytidine-rich tails can trigger RIG-I when present at >10% abundance.⁸³ These data align with recent observations that 2′-O-methylated caps and LNA-modified 5′ ends synergize with long, chemically protected tails to reduce PKR activation and dsRNA formation.⁸⁹

Together, the emerging rule set for therapeutic mRNA tails is as follows: (1) 150–250 nts total length delivered via branched or post-transcriptionally elongated architectures; (2) ≤5% non-adenosines, biased toward uridine only when accelerated decay is desired; (3) PS/2MOE terminal protection for exonuclease resistance without immunogenicity; and (4) manufacture by either G47A + 884G T7 IVT followed by enzymatic “tune-up” or single-ORF cytoplasmic tailing systems to decouple tail length from plasmid stability. These principles are now being integrated into next-generation multi-antigen cancer vaccines and in vivo CRISPR editors, where programmable tail degradation kinetics is engineered to match the therapeutic window of each encoded payload.⁹⁰

Computational and AI-driven design of mRNA construct

mRNA construct design has advanced from codon-centric optimization—exemplified by the codon adaptation index (CAI) adopted by IDT, Twist, and GENEWIZ⁹¹—to multi-objective frameworks that simultaneously maximize CAI, minimum free energy (MFE), DegScore-predicted stability, and average unpaired probability (AUP) through algorithms such as LinearDesign (DFA-lattice parsing), RiboTree (Monte-Carlo tree search), LinearFold/LinearPartition for rapid folding, RiboGraphViz for structural visualization, and the tissue-specific CUSTOM tool.²¹ High-throughput UTR libraries coupled with deep-learning (CNN) and genetic algorithms trained on >260,000 5′ UTR variants now predict ribosome loading and translational yield with experimental precision, while DegScore’s ridge-regression model quantifies degradation rates from in-line probing data.⁹²

Over the past 3 years, construct design and optimization have moved beyond isolated in silico screens toward fully integrated pipelines in which multi-objective ML models co-evolve with Design of Experiments (DoE),⁹³ Bayesian optimization,⁹⁴ and first-principle simulations⁹⁵ to deliver mRNA constructs whose potency, stability, and biodistribution profiles are predicted before a single wet-lab validation step is undertaken.

Sequence-level optimization

Classical dynamic programming (DP) algorithms, long constrained to synonymous codon replacement, have historically balanced CAI with MFE predictions to maximize expression while preserving structural stability. The LinearDesign algorithm recently reduced computational complexity from O(n³) to O(W²×n) without sacrificing thermodynamic accuracy, thereby enabling genome-scale optimization of clinically relevant transcripts.^96,97 Nevertheless, DP-based methods remain inherently limited by their inability to capture higher-order epistatic interactions among synonymous sites.

Deep-learning architectures now transcend these limitations. RNop, a Vision Transformer model that encodes CDS as two-dimensional positional frequency maps, simultaneously optimizes four differentiable loss terms: GPLoss (GC-content penalty), CAILoss, tAILoss (tRNA adaptation index), and MFELoss.⁹⁸ Benchmarked against DP baselines on both in silico metrics and a 2,000-gene in vivo expression compendium, RNop achieved a 1.7-fold median improvement in luciferase output in murine liver (p < 0.001). In parallel, CodonBERT—a 340-million-parameter language model pretrained on 2.3 million microbial CDSs—demonstrated state-of-the-art performance in predicting low-expression genes (area under the receiver operating characteristic curve = 0.93) and generated codon-optimized sequences without altering the encoded proteome.⁹⁹ Importantly, to circumvent off-target amino acid changes that can arise from probabilistic sampling, BiLSTM-CRF hybrids first optimize CAI under a synonymous-only constraint and subsequently revert non-synonymous mutations via constrained Viterbi decoding.¹⁰⁰

Nanoparticle design and delivery optimization

The rational design of ionizable lipids remains the principal determinant of LNP potency. A LightGBM classifier trained on a curated library of 1,950 ionizable lipids using extended connectivity fingerprints (ECFP) attained 82% accuracy and 0.76 precision against the clinical MC3 benchmark (Onpattro).¹⁰¹ Building on this previous report, Wang et al.¹⁰² introduced an AI-driven pipeline that couples LightGBM models with SHapley Additive exPlanations (SHAP) to virtually screen about 20 million ionizable lipids for optimal apparent pKa (6–7) and mRNA delivery efficiency, of which six lead lipid candidates outperformed the conventional MC3. This finding exemplifies the power of interpretable ML to identify key ECFP substructures (e.g., cyclohexyl tails and ester linkers) that govern potency.

Furthermore, an I-optimal DoE matrix was employed to fabricate 24 LNP-mRNA prototypes, systematically tuning raw-material attributes and microfluidic processing variables to improve critical quality attributes—namely, hydrodynamic diameter, surface charge, and encapsulation yield. An ensemble learning framework, internally cross-validated and reaching >97% classification accuracy, was then used to deconvolute the impact of each factor and pinpoint optimal operating windows.⁹⁴ Complementing this, a hybrid artificial-neural-network-driven DoE strategy further refined the bioprocess by jointly optimizing lipid identity, lipid-to-cholesterol molar ratio, nitrogen-to-phosphate (N/P) charge ratio, and total flow rate, underscoring the accelerating role of data-centric methodologies in next-generation vaccine manufacture.¹⁰³

The AGILE platform integrates self-supervised graph neural networks with high-throughput Ugi chemistry to screen 12,000 ionizable lipids in silico, yielding H9 (7.8-fold more potent than ALC-0315 in muscle) and R6 (5-fold better in macrophages); SHAP analysis and DoE refinement uncover cell-specific rules—electronic descriptors and asymmetric tail lengths—that accelerate bespoke LNP discovery from months to days while preserving safety and scalability.¹⁰⁴

Across six independent datasets (Moderna, Acuitas, Protiva), regression models for delivery efficiency exhibited dataset-dependent R² values ranging from 0.50 to 0.78, underscoring the influence of batch effects and inter-laboratory variability.¹⁰² Classification models (LightGBM, XGBoost) displayed markedly lower variance (R² = 0.78–0.85) and were consequently adopted as the consensus modeling paradigm in cross-institutional benchmarking exercises.¹⁰³

The convergence of large-scale structural datasets, foundation language models, and active-learning-driven DoE positions AI/ML as the central engine for next-generation sequence design, optimization, and LNP discovery for better mRNA functionality. Immediate priorities include federated learning frameworks to pool privacy-preserving, multi-institutional data and causal-inference models capable of dissecting structure-activity relationships beyond mere correlation. Such advances will be indispensable as mRNA technologies expand.

mRNA modifications

mRNA modifications have revolutionized the field of RNA therapeutics by addressing the inherent challenges of immunogenicity and instability associated with unmodified in vitro-transcribed mRNA. Among these modifications, Ψ and its derivative m1Ψ stand out as pivotal innovations, particularly in the realm of mRNA vaccines.¹⁰⁵ The recognition of this innovation led to the award of Nobel Prize to the two scientists who worked on the base modifications.¹⁰⁶ Various other base modifications, such as N1-methyladenosine, 5-methylcytidine, N6,2′-O-dimethyladenosine, inosine, N4-acetylcytidine, 2′-O-methylated nucleotides, and internal N7-methylguanosine, have been identified.¹⁰⁷ Increasing research efforts are elucidating their biosynthesis pathways, patterns of distribution, regulatory mechanisms, and functional roles, thereby expanding our understanding of these modifications in cellular processes.

The immunogenicity of unmodified mRNA is largely due to its recognition by innate immune receptors, such as Toll-like receptors (TLR7 and TLR8), and cytoplasmic sensors like retinoic acid-inducible gene I and PKR.¹⁰⁸ The discovery in 2005 that substituting uridine with Ψ could inhibit TLR activation marked a watershed moment for mRNA therapeutics. This modification not only evades TLR7 and TLR8 detection but also mitigates the activation of PKR, 2′,5′-oligoadenylate synthetase, and other immune pathways, thereby reducing undesirable immune responses.^105,108

Ψ-modified mRNA has also demonstrated enhanced stability and translation efficiency. In vivo studies revealed that Ψ incorporation extends the half-life of mRNA and increases protein output, as evidenced by elevated erythropoietin levels in animal models.¹⁰⁸ Mechanistically, Ψ reduces PKR activation, which prevents phosphorylation of eukaryotic initiation factor 2α (eIF2α), a critical regulator of translation.¹⁰⁹

The development of m1Ψ further refined these advantages. As a naturally occurring derivative of Ψ, m1Ψ exhibits superior performance in dampening immune responses and boosting translation efficiency.¹¹⁰ This modification underpinned the success of the first Food and Drug Administration-approved mRNA vaccines against COVID-19, developed by Pfizer-BioNTech and Moderna. Both vaccines employed IVT mRNA with complete uridine substitution by m1Ψ, which enhanced antigen expression and minimized adverse reactions.¹¹¹

The incorporation of m1Ψ into synthetic mRNA has demonstrated superior performance compared to current advanced mRNA modifications. This modification offers several advantages, including reduced innate cellular immunogenicity, enhanced cell viability, and increased levels of gene expression.^112,113 The observed high protein expression associated with m1Ψ-modified mRNA is closely linked to its enhanced secondary structure¹¹⁴ and reduced phosphorylation of eIF2α, accompanied by lower PKR activation. This attenuation of PKR activity prevents the inhibition of protein translation, ensuring efficient expression.^109,112

While other uridine analogs, such as 5-methyluridine, 5-methoxyuridine, and 2-thiouridine, can also decrease mRNA immunogenicity, their use is limited due to lower translation efficiency compared to Ψ.^115,116,117 The widespread adoption of Ψ and m1Ψ modifications underscores their transformative impact on mRNA vaccines, fostering rapid advancements in combating infectious diseases and cancers. As research progresses, these modifications may extend to other therapeutic applications, heralding a new era in RNA-based medicine.

mRNA delivery vehicles: new platforms and challenges

Nanoparticle-based delivery systems have emerged as a promising platform that addresses mRNA instability, RNAse activity, and efficient delivery. Nanoparticles protect mRNA from enzymatic degradation by encapsulating it within their core, enhancing cellular uptake through facilitated endocytosis and promoting intracellular trafficking.^118,119 Several types of nanoparticles have been investigated for mRNA delivery, including LNPs, which are composed of lipid bilayers and widely used due to their stability, efficient encapsulation, and controlled release properties. Polymeric nanoparticles, made from polymers such as polyethyleneimine and poly(lactic-co-glycolic acid), also offer protective encapsulation during transit.¹²⁰ Additionally, peptide nanoparticles have been explored for their potential in mRNA delivery.

The mechanisms underlying mRNA loading into nanoparticles often involve electrostatic interactions, hydrogen bonds, or coordination interactions, facilitated by techniques like thin-film hydration, nanoprecipitation, or microfluidic mixing.¹²¹ Importantly, these delivery carriers must be safe, exhibiting low toxicity, and non-immunogenic, meaning they should have minimal potential to provoke an immune response on their own.¹²² In this section, we provide an overview of the latest advances in the design and development of mRNA delivery systems, focusing on their interactions with mRNA and their transfection efficiency.

Lipid nanoparticles

LNPs are the most widely used delivery vehicles for mRNA, especially in the context of vaccines. They are composed of ionizable lipids, cholesterol, phospholipids, and polyethylene glycol (PEG)-lipids. LNPs encapsulate the mRNA, protecting it from degradation and facilitating cellular uptake via endocytosis. Once inside the cell, the acidic environment of the endosome triggers the release of mRNA into the cytoplasm.¹²⁰ This section focuses on new technologies in ionizable lipids and PEG.

Ionizable lipids

Ionizable LNPs (iLNPs) have become a critical focus in mRNA delivery due to their unique properties, particularly their electric neutrality under physiological conditions, as highlighted by Han’s study.¹²³ This neutrality minimizes rapid elimination from the bloodstream and reduces immune activation, making iLNPs especially suited for therapeutic applications. The core advantage of ionizable lipids lies in their single ionizable head groups, which can include various structures like simple tertiary amines, branched polyamines, or cyclic structures such as piperazine, diketopiperazine, or benzene cores.¹²³ These head groups can undergo protonation in acidic environments, facilitating the escape of mRNA from endosomes into the cytoplasm—a critical step for efficient translation.

Recent studies have further advanced the understanding and application of iLNPs. Ionizable lipids can be classified into several categories, including tertiary amine lipids, quaternary amine lipids, imidazole core lipids, piperazine core lipids, and diketopiperazine core lipids.¹²⁴ The design of new ionizable lipids has shown promise in enhancing mRNA delivery efficiency by improving endosomal escape and subsequent translation.^125,126

Previous study has reported the successful loading of CFTR mRNA into an MC3 delivery system and restored chloride channel function in patient-derived bronchial epithelial cells, showcasing the therapeutic potential of iLNPs.¹²⁷ The clinical relevance of LNPs composed of MC3, DSPC, cholesterol, DMG-PEG2000, and mRNA was further demonstrated by a study showing that transfection efficacy varied significantly across 30 cell lines, largely due to differences in the timing of endosomal escape.¹²⁸ Also, a recent study has highlighted the targeting capabilities of MC3 LNPs, which were covalently conjugated with αPV1 antibodies to direct mRNA to the lungs by binding to plasma vesicle-associated protein,¹²⁹ championing the era of organ and tissue targeting via LNPs.

In terms of safety and tolerability, ionizable lipids that not only exhibited high tolerability but also reduced innate immune stimulation when administered intramuscularly, emphasizing the importance of route-specific considerations in mRNA delivery, have been synthesized.¹³⁰ This was extended by synthesizing piperazine-centered compounds, which were successfully used as vectors for chimeric antigen receptor (CAR) mRNA delivery in primary human T cells.¹³¹ Further innovations include the synthesis of cationic lipid-modified aminoglycosides designed specifically for liver-targeted delivery of Luc mRNA.¹³² Yang et al.¹³³ developed an iLNP based on the iBL0713 lipid for delivering erythropoietin (EPO) mRNA, demonstrating comparable efficacy to the clinically established Dlin-MC3-DMA-based formulations in the liver.

The ongoing refinement of iLNPs has also led to significant advancements in targeting specific tissues. For instance, a lead LNP candidate, A4, which, along with other candidates like B5 and C12-200, effectively facilitated in vivo mRNA delivery to endothelial cells, immune cells, and placental trophoblasts were identified recently.¹²⁴ Another study built on this work by screening 128 novel LNPs and discovering that stereopure LNPs, such as C12-200-S, achieved significantly higher mRNA delivery in vivo compared to their racemic counterparts, with better tolerability.¹³⁴

Despite these advancements, the design and optimization of novel LNPs continues to be a research focus with another study identifying an optimized ionizable lipid, OC2-K3-E10, after iterative design, which proved effective for intramuscular mRNA delivery.¹³⁵

Recent interest has also centered on mRNA delivery to the placenta. In addition to previous findings, the use of C12-200 in combination with 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine as the phospholipid resulted in high transfection efficiency in vitro and in vivo, targeting the placenta without evident toxicity .¹³⁶ Also, an epidermal growth factor receptor (EGFR) antibody-conjugated LNP platform that achieved significantly higher mRNA delivery in murine placentas compared to non-targeted LNPs, highlighting the potential of targeted mRNA delivery to specific tissues, including the placenta, has been reported.¹³⁷ These studies underscore the rapid advancements in the design and application of iLNPs, paving the way for more efficient, targeted, and safer deliveries.

Polyethylene glycol incorporation

The incorporation of PEG into LNPs has emerged as a crucial strategy to enhance the pharmacokinetics and biodistribution of mRNA. PEG-lipids extend the circulation time of LNPs and reduce immune recognition, which is essential for improving the efficacy of mRNA delivery systems. Recent innovations in PEGylation techniques have focused on optimizing these effects, leading to better biodistribution and reduced side effects associated with LNPs.

The impact of varying PEG content on the size and performance of LNPs has been explored by adjusting the PEG content from 5% to 0.5%; LNPs ranging from 50 to 150 nm in size were generated.¹³⁸ The study found that LNPs with 0.5% PEG content resulted in the highest expression levels of luciferase following subretinal injections in mice, indicating that lower PEG content may enhance transfection efficiency.¹³⁸ This finding suggests a delicate balance between PEG content and LNP size that can be optimized for specific therapeutic applications.

Tanaka et al.¹³⁹ further advanced this understanding by investigating how modifications to the phospholipid and PEG-conjugated lipid components of LNPs could influence transfection efficiency. They reported a 221-fold increase in luciferase activity by optimizing the lipid composition, which underscores the significant role of lipid formulation in enhancing mRNA delivery and expression.

Furthermore, surface modifications of LNPs with different PEG variants could influence cellular tropism and transfection outcomes. Recent studies showed LNP variants including positively charged amine-modified PEG-lipids (LNPa), negatively charged carboxyl-modified PEG-lipids (LNPz), and neutral unmodified PEG-lipids (LNPx).¹⁴⁰ Subretinal injections of these LNP variants revealed that LNPa induced a similar signal in the retinal pigmented epithelium as conventional LNPs. However, LNPz and LNPx showed unexpected transfection patterns, with 27% and 16% transfection in photoreceptors, respectively, and striking localization throughout the photoreceptor cells.¹⁴⁰ These results suggest that altering the surface charge of PEG-lipids can significantly influence the tropism and efficiency of mRNA delivery, potentially leading to improved therapeutic outcomes.

The role of PEG-lipids in LNP-mediated mRNA delivery has also been explored in the context of respiratory therapeutics. Ongun et al.¹⁴¹ investigated the efficacy of PEG-lipid content in LNPs designed for local delivery of mRNA to the respiratory tract, particularly for inhalable mRNA therapeutics. They hypothesized that PEG-lipid content would be critical for maintaining colloidal stability during aerosolization and for effective mucosal delivery. Their findings demonstrated that while increasing PEG-lipid content improved the colloidal stability of LNPs during aerosolization, it had a negative impact on transfection efficiency in vitro.¹⁴¹ This result highlights the trade-offs involved in optimizing PEG-lipid content for different routes of administration, suggesting that the ideal formulation may vary depending on the specific delivery method and therapeutic target.

These studies collectively illustrate the importance of PEG in LNP design and emphasize the need for precise optimization of PEG-lipid content to balance stability, immune evasion, and transfection efficiency. As such, the fine-tuning of PEG-lipid parameters will be crucial for maximizing their clinical potential across a wide range of applications.

Polymer nanoparticles

Polymeric nanoparticles are another versatile platform for mRNA delivery. These include poly(lactic-co-glycolic acid) nanoparticles, dendrimers, and polyethylenimine (PEI)-based systems. PEI is one of the most potent non-viral vector for gene delivery. However, PEI is highly toxic and non-biodegradable, limiting its application. Therefore, the modification of PEI remains inevitable.

PEI modification

PEI has been widely utilized as a vector for mRNA delivery due to its strong cationic nature, which facilitates complexation with negatively charged nucleic acids. However, the inherent cytotoxicity of PEI has driven the development of modified polymers to enhance biocompatibility and transfection efficiency, particularly targeting specific tissues like the lungs and liver. The potential of PEI1800-LinA5-PEG0.3, a modified PEI polymer, to target the pulmonary microvascular endothelium has been explored. This approach demonstrated successful mRNA delivery to lung endothelium and pulmonary immune cells, highlighting the importance of polymer modifications in tissue-specific targeting¹⁴²

Poly(β-amino esters)

Poly(β-amino esters) (PBAEs) have also gained attention due to their biocompatibility and biodegradability, making them suitable candidates for mRNA delivery. Kaczmarek’s research group synthesized PBAEs to target lung endothelium and immune cells, achieving efficient mRNA delivery in these tissues.¹⁴³ The versatility of PBAEs is further demonstrated by another study that utilized oligopeptide end-modified PBAEs (OMPBAEs) with enhanced endosomal escape and cytoplasmic penetration properties to transfect mRNA into liver tissues.¹⁴⁴ These OMPBAEs facilitated precise targeting and effective delivery of mRNA, underscoring their potential for liver-specific therapeutic applications.

Hyperbranched PBAEs

These have been specifically developed for inhalation delivery of mRNA to the lung epithelium. Studies have shown that hyperbranched PBAEs (hPBAEs) could achieve sufficient protein production in the lungs while maintaining safety and biocompatibility, making them promising candidates for respiratory mRNA therapies.¹⁴⁵ Additionally, a novel polycaprolactone-based PBAE was designed to deliver mRNA intravenously to the spleen, further expanding the application of PBAEs for targeted organ delivery.¹⁴⁶

Overcoming the challenge of transfecting cell lines with high glutathione levels, tetrasulfide-incorporated large-pore dendritic mesoporous organosilica nanoparticles modified by PEI were designed. These modified nanoparticles demonstrated excellent in vitro and in vivo delivery efficacy, providing a solution for efficient mRNA transfection in challenging cellular environments.¹⁴⁷ A study by Ren et al.¹⁴⁸ advanced the development of mRNA delivery systems by designing a self-assembling polymeric micelle based on the modification of PEI with vitamin E succinate. This modification not only reduced the toxicity associated with unmodified PEI but also enhanced mRNA transfection efficiency across multiple cell lines, demonstrating a promising approach for safer and more effective mRNA delivery. Recently, modification of polyethyleneimine with fluoroalkane enhanced intracellular delivery and the efficacy of mRNA cancer vaccine.¹⁴⁹

These studies underscore the critical role of polymer modifications in enhancing the specificity, efficiency, and safety of mRNA delivery systems. By tailoring the properties of polymers like PEI and PBAEs, researchers are paving the way for more targeted and effective mRNA-based therapies, with applications spanning various tissues and organs.

Peptide nanoparticles

Peptide-based nanoparticles offer another promising avenue for mRNA delivery. These nanoparticles are formed by self-assembling peptides that can encapsulate mRNA and facilitate its delivery.

Cell-penetrating peptides

Cell-penetrating peptides (CPPs) have been utilized to enhance mRNA delivery. CPPs can traverse cell membranes and deliver mRNA directly into the cytoplasm. Recent studies have demonstrated that CPP-modified nanoparticles can significantly improve mRNA uptake and translation. Peptides that target photoreceptors have been identified. LNPs were decorated with these peptides and delivered mRNA to Müller glia, retinal pigment epithelium, and photoreceptors with robust protein expression observed in these locations in both rodents and non-human primates, overcoming the ocular barrier.¹⁵⁰ Further information about the development of polypeptide-based particles for mRNA delivery has been reviewed by Zhao et al.¹⁵¹

Stimuli-responsive peptides

The development of stimuli-responsive peptides (SRPs) that release mRNA in response to specific triggers, such as pH or temperature changes, has shown promise in improving the efficiency of mRNA delivery. A vivid instance is the pulmonary delivery of mRNA via PEGylated pH-responsive peptide nano-self-assembly. Recently, in vitro mRNA binding and release, cellular uptake, transfection, and cytotoxicity were studied, and finally, a proper PEGylated peptide with enhanced pulmonary mRNA delivery efficiency and improved safety in mice was identified.¹⁵²

Virus-like particles

Virus-like particles (VLPs) have emerged as promising platforms for mRNA delivery due to their ability to mimic native viruses while lacking the infectious genome, ensuring safety. These nanoscale, multiprotein structures are highly immunogenic and capable of encapsulating or surface displaying nucleic acids such as mRNA, making them highly suitable for vaccine and therapeutic applications.¹⁵³

Recent studies highlight significant strides in VLP design for mRNA delivery. VLPs derived from bacteriophage MS2 and Qβ are widely used due to their modularity and ease of genetic and chemical manipulation.¹⁵⁴ Qβ VLPs were engineered to enhance encapsulation of mRNA by fusing capsid proteins with RNA-binding domains.^155,156 This approach facilitated the efficient packaging and protection of mRNA, improving delivery efficiency in cellular systems. Furthermore, innovations in modifying VLP sizes and assembly dynamics have been achieved, enabling optimization for specific delivery contexts such as targeted tissues or cellular compartments.^157,158

In vivo applications are also advancing. Studies demonstrate that VLPs carrying mRNA can efficiently transfect immune cells, eliciting potent antigen-specific immune responses. The use of enveloped VLPs (eVLPs) has expanded possibilities, offering improved interaction with lipid membranes and enhanced cellular uptake.¹⁵⁹ However, challenges remain, including instability of eVLPs and inconsistent mRNA loading, which are being addressed through computational design and bioprocessing optimizations.

Preclinical findings suggest that VLP-based mRNA delivery platforms outperform some LNP systems in specific scenarios by offering tunable immunogenicity and reduced off-target effects.^160,161 Additionally, incorporating pathogen-associated molecular patterns within VLPs has shown to activate innate immune pathways, enhancing the therapeutic efficacy of mRNA payloads without requiring external adjuvants.¹⁶²

Despite these advances, challenges such as large-scale manufacturing, particle stability during storage, and regulatory hurdles need to be overcome. New bioprocessing methods, including automation and scalable purification techniques, are being developed to address these limitations.¹⁶³

Overall, VLPs represent a versatile and efficient platform for mRNA delivery, with continuous innovations poised to enhance their application in both therapeutic and vaccine development.

Therapeutic application of mRNA

mRNA offers a versatile platform for treating a range of diseases, including infectious diseases, cancer, and genetic disorders. Beyond vaccines, mRNA is used for protein replacement therapies to address enzyme deficiencies and genetic mutations. In oncology, mRNA-based immunotherapies stimulate targeted anti-tumor immune responses. Vaccines against infectious diseases and cancer, gene editing, and replacement would be reviewed in this section. Figure 2 gives a sequential production pipeline for the therapeutic application of mRNA.

Figure 2.

mRNA drug/vaccine sequence production pipeline

The desired peptide or protein sequence is encoded into a plasmid DNA template, which is then transcribed in vitro into mRNA using bacteriophage polymerases. The resulting mRNA is purified via high-performance liquid chromatography or nanoprecipitation to eliminate impurities. The purified mRNA is then encapsulated in delivery vehicles, which interact with the molecule through three primary mechanisms: (1) electrostatic attraction to the ribonucleotide phosphate groups, (2) hydrogen bonding with nucleobases, or (3) coordination with phosphate ions. Common delivery systems include cationic compounds (e.g., lipids, polymers, peptides, virus-like particles). The efficacy, pharmacokinetics, and safety of mRNA therapeutics are assessed in preclinical models, including vaccinated mice and primates, before scaling up production for clinical trials.

Vaccines against infectious diseases

COVID-19 vaccines

The COVID-19 vaccines developed by Pfizer-BioNTech and Moderna have not only proven this concept but also achieved efficacy rates above 90%, a testament to the power of this technology.^19,110 This success has sparked renewed interest and optimism in the potential of mRNA vaccines to tackle other infectious diseases.¹⁶⁴ However, concerns of safety as well as efficacy have continued to evolve in recent times. Polack et al.¹¹⁰ had reported that the BNT162b2 COVID-19 vaccine was quite efficacious, conferring 95% protection against the infection in persons aged 16 years or older with a safety profile that was characterized by short-term mild pain at the site of injection, headache, and fatigue. Another study led 30,420 volunteers who are given either the mRNA-173 vaccine or placebo, having divided these patients into two groups. Moderate transient local and systemic reactivity was observed in the mRNA-1273 vaccine group, and serious adverse effects were rare. The efficacy of the vaccine at preventing the COVID-19 illness was about 94.1%.¹⁹

In contrast, Fraiman et al.¹⁶⁵ reported serious adverse events after mRNA COVID-19 vaccination among adults in randomized trials. Using the Pfizer and Moderna vaccines, they found out that the excess risk was 10.1 and 15.1 per 10,000 vaccinated individuals for Pfizer and Moderna, respectively, over placebo baselines. When combined, the mRNA vaccines showed an excess risk of 12.5 per 10,000 vaccinated individuals. The Pfizer trial showed a 36% higher risk of serious adverse events in the vaccine group, while the Moderna trial showed a 6% higher risk. Overall, mRNA vaccine recipients had a 16% higher risk of serious adverse events.¹⁶⁵ Despite the safety concerns and the reports that have shown that mRNA vaccines are safe,¹⁹ the efficacy of the vaccines has not been in doubt as many reports have shown that the efficacy ranges between 86% and 100%.^166,167,168

Other infectious diseases

HIV

As early as 2018, Leal et al.¹⁶⁹ reported a phase 1 clinical trial of intranodal administration of mRNA-based therapeutic vaccine against HIV-1 infection. It was found that the vaccine was safe and well-tolerated, with 31 grade 1/2 and one grade 3 adverse events, mostly unrelated to vaccination. Patients receiving the highest dose exhibited increased T cell responses, and the proportion of responders increased from 31% to 80% post-vaccination. This study proves that mRNA vaccine is not only safe but also able to induce strong T cell responses.

Influenza virus mRNA vaccine

Freyn et al.¹⁷⁰ developed a multi-targeting nucleoside-modified mRNA vaccine that could offer broad protection. The selected antigens include the conserved hemagglutinin (HA) stalk domain, matrix-2 ion channel, nucleoprotein, and broadly reactive neuraminidase. Intradermal administration of these LNPs conferred protection to mice against H1N1 virus challenge, even at a dose 500 times the median lethal dose.¹⁷⁰ To address the challenge of mutation during circulation, which causes antigenic mismatch, Ma et al.¹⁷¹ developed an mRNA vaccine encapsulated in LNPs (mRNA-LNPs) targeting the consensus full-length HA sequence (H1c). Both Th1- and Th2-biased immune responses were elicited, while Th1-biased response was particularly robust, leading to a full protection of H1N1 influenza viruses in mice. To further accentuate broad protection, a quadrivalent mRNA vaccine was developed.¹⁷² The study demonstrates that mRNA vaccination induces neutralizing and serum antibodies against each influenza virus strain included in the current quadrivalent vaccine. This vaccine targets four distinct influenza viruses, comprising two influenza A viruses and two influenza B viruses, as well as various antigenically diverse influenza virus strains. Both hemagglutination inhibition assay and virus neutralization assays confirm the antibody response.¹⁷²

Notably, the quadrivalent mRNA vaccines exhibit antibody titers comparable to those elicited by monovalent vaccines for each tested virus, regardless of dosage following an mRNA booster vaccine. Furthermore, mice vaccinated with mRNA encoding an H1 HA experience reduced weight loss and lower lung viral titers compared to mice not vaccinated with H1 HA mRNA.¹⁷² These studies indicate that broad protection against multiple influenza viruses has been achieved through mRNA vaccine in preclinical models.

Respiratory syncytial virus mRNA vaccine

One of the most notable developments is the approval of mRNA-1345 (mRESVIA) by Moderna, which has shown high efficacy in preventing respiratory syncytial virus (RSV)-associated lower respiratory tract disease (LRTD) in older adults. The vaccine, which encodes the membrane-anchored RSV-A preF protein, was found to be 83.7% effective against RSV-LRTD with at least two symptoms and 82.4% effective against RSV-LRTD with at least three symptoms.¹⁷³ The safety profile of mRNA-1345 was consistent with previous studies, with mild to moderate adverse reactions reported, such as fatigue, headache, myalgia, and arthralgia.¹⁷³

In addition to its approval for use in older adults, mRNA-1345 is also undergoing phase 1 and 2 trials as a pediatric and maternal vaccine (NCT06143046).¹⁷⁴ This expansion into different age groups highlights the potential of mRNA vaccines to provide broad protection against RSV across various populations.

Another promising candidate is the monovalent RSV mRNA vaccine being investigated in a phase 2 trial sponsored by Sanofi (NCT06686654 and NCT07071558).¹⁷⁴ Although the details of the vaccine’s target antigen are not available, its development underscores the ongoing efforts to leverage mRNA technology for RSV prevention.

The success of mRNA-1345 and other mRNA vaccines is partly due to the platform’s ability to rapidly produce vaccines that elicit strong immune responses. The preF conformation of the RSV-F protein, which is highly conserved across RSV subtypes, has been a key target for these vaccines.¹⁷³ This approach has been validated by the efficacy observed in clinical trials and the subsequent approval of mRNA-1345.

Rabies virus mRNA vaccine

Rabies, a zoonotic disease with significant global health implications, claims approximately 59,000 lives annually. Despite the availability of effective vaccines, challenges such as limited accessibility and the need for multi-dose vaccination regimens (typically four doses) impede effective disease control.¹⁷⁵ Addressing these barriers, CureVac AG has introduced CV7201, an innovative mRNA-based rabies vaccine. Encapsulated in the cationic protein protamine, CV7201 encodes the rabies virus glycoprotein. Clinical trials (NCT02241135) demonstrated its temperature stability and ability to achieve a World Health Organization-recommended antibody response in 70.3% of participants through intradermal administration.¹⁸

Based on CV7201, CureVac AG refined the LNP formulation, resulting in the development of CV7202. Like CV7201, CV7202 employs the same mRNA antigen. The optimized LNP composition includes an ionizable amino lipid, a PEG-modified lipid, phospholipid, and cholesterol. Notably, CV7202 (see Table 2) demonstrated favorable tolerability in a clinical trial (NCT03713086).¹⁷⁶

Table 2.

Clinical trials of mRNA vaccines against infectious diseases (except COVID-19)

Trial number	Disease indication	Sponsor/Funder	Name of mRNA product	Route of administration	Status	Phase
NCT06680128	Japanese encephalitis virus	SK Bioscience Co. Ltd	GBP560	I.M.	Recruiting	1/2
NCT06564194	RSV	Immorna Biotherapeutics, Inc.	JCXH-108	I.M.	Active, not recruiting	1
NCT06375512	Herpes zoster	Shenzhen Shenxin Biotechnology Co., Ltd	IN001	I.M.	Active, not recruiting	1
NCT06273553	Human papillomavirus-associated intraepithelial neoplasia	RinuaGene Biotechnology Co., Ltd.	RG002	I.M.	Not yet recruiting	1/2
NCT06237296	RSV and metapneumovirus	Sanofi Pasteur	VAV00027	I.M.	Completed	1
NCT05217641	HIV	National Institute of Allergy and Infectious Diseases	BG505 MD39.3	I.M.	Active, not recruiting	1
NCT05398796	Nipah virus	National Institute of Allergy and Infectious Diseases	mRNA-1215	I.M.	Completed	1
NCT05001373	HIV-1	International AIDS Vaccine Initiative	mRNA-1644	i.p.	Active, not recruiting	1
NCT05127434	RSV	Moderna TX, Inc.	mRNA-1345	I.M.	Active, not recruiting	2/3
NCT03713086	Rabies	CureVac	CV7202	I.M.	Completed	1
NCT05624606	Influenza immunization	Sanofi Pasteur	MRT5410	I.M.	Completed	1/2
NCT05553301	Influenza immunization	Sanofi Pasteur	MRT5407	I.M.	Completed	1/2
NCT04975893	Cytomegalovirus	Moderna TX, Inc.	mRNA-1647	I.M.	Enrolling by invitation	2
NCT05085366	Cytomegalovirus	Moderna TX, Inc.	mRNA-1647	I.M.	Active, not recruiting	2
NCT05164094	Epstein-Barr virus	Moderna TX, Inc.	mRNA-1189	I.M.	Active, not recruiting	1/2
NCT03392389.	Human metapneumovirus and human parainfluenza	Moderna TX, Inc.	mRNA-1653	I.M.	Completed	1
NCT05581641	Malaria	BioNTech SE	BNT165b1	I.M.	Completed	1
NCT04917861	Zika virus	Moderna TX, Inc.	mRNA-1893	I.M.	Completed	2
NCT05537038	Tuberculosis	BioNTech SE	BNT164a1	I.M.	Active, not recruiting	1
NCT05415462	Seasonal influenza	Moderna TX, Inc.	mRNA-1010	I.M.	Completed	3
NCT05333289	Seasonal influenza	Moderna TX, Inc.	mRNA-1030	I.M.	Completed	1/2
NCT03345043	Influenza A (H7N9)	Moderna TX, Inc.	VAL-339851	I.M.	Completed	1
NCT04144348.	hMPV/PIV3	Moderna TX, Inc	mRNA-1653	I.M.	Completed	Ib
NCT04062669	Rabies	GlaxoSmithKline	GSK3903133A	I.M.	Completed	1

hMPV; human metapneumovirus; I.M., intramuscular; i.p., intraperitoneal; PIV3: .parainfluenza virus type 3.

In 2020, Stokes and colleagues utilized a CNE (a carrier system) to encapsulate self-amplifying RNA encoding the alphavirus RNA-dependent RNA polymerase and the rabies glycoprotein G.¹⁷⁷ Additionally, a nucleoside-modified rabies mRNA-LNP vaccine was developed.¹⁷⁸ In mouse studies, a single vaccination with RABV-G mRNA, even at low doses, induced more robust humoral and T cell immune responses than three inoculations of a commercially inactivated vaccine. Importantly, these mice achieved full protection against lethal rabies challenge.¹⁷⁸

A single RABV-G mRNA vaccination in mice induces a durable humoral immune response lasting at least 25 weeks, extendable to over a year with a two-dose regimen, outperforming the three-dose inactivated vaccine. While an unmodified rabies mRNA vaccine showed high cross-neutralizing antibody titers after two doses, it did not replicate the long-term protection observed in Bai’s study, emphasizing the critical role of vaccine design in achieving sustained immunity.¹⁷⁹

Gene therapy

Gene therapy is a groundbreaking strategy aimed at treating or curing various diseases by directly altering the genetic material within an individual’s cells. This approach involves introducing, removing, or modifying DNA or RNA sequences to correct genetic defects and restore normal cellular function.¹⁸⁰ In the context of gene therapy, mRNA serves versatile roles, enabling both gene replacement and gene editing, thus expanding the scope of therapeutic possibilities.

Gene replacement therapy

mRNA can be engineered to encode a functional version of a gene that is defective or absent in a patient. Upon delivery to the patient’s cells, the mRNA facilitates the synthesis of the corresponding protein, effectively addressing the underlying genetic deficiency.¹⁸¹ This therapeutic strategy has been studied for genetic disorders such as hemophilia and cystic fibrosis (CF).

Advancements in mRNA technologies have broadened their applications, extending beyond vaccines to encompass gene therapy. Although still in its early stages, the progress in mRNA-based gene therapies highlights significant potential for treating genetic disorders.¹⁸² Below are some of the disease conditions where mRNA gene replacement therapy is being applied.

Methylmalonic acidemia

An et al.¹⁸³ explored the application of mRNA therapy to manage methylmalonic acidemia, showcasing the feasibility of this approach in metabolic disorders.¹⁸⁴ It has been reported that even well-managed patients with methylmalonic acidemia (MMA) remain at the risk for complications associated with the disease, including intermittent metabolic decompensation and disease-associated sequelae. An et al.¹⁸⁵ developed a potent human methylmalonyl-CoA mutase enzyme (hMUT) mRNA therapy and delivered the gene using LNP to protect it from degradation. A single dose of hMUT mRNA delivered via LNPs significantly lowered the plasma and tissue methylmalonic acid levels in a severe MMA mouse model, with effects lasting several days. The expressed hMUT protein had an estimated half-life of approximately 1.2 days.¹⁸⁵

Acute intermittent porphyria

mRNA’s role as an etiological treatment for acute intermittent porphyria (AIP) has been investigated.¹⁸⁶ AIP results from haploinsufficiency of porphobilinogen deaminase (PBGD), the third enzyme in the heme biosynthesis pathway.¹⁸⁷ Jiang and colleagues administered human PBGD (hPBGD) mRNA, encoded by the HMBS gene, intravenously to mice. They encapsulated this mRNA in LNPs, which induced dose-dependent protein expression in the liver cells of the treated mice.¹⁸⁶ This approach rapidly normalized the excretion of urine porphyrin precursors during ongoing attacks. The hPBGD mRNA also protected against several conditions, including mitochondrial dysfunction, hypertension, pain, and motor impairment.

Fabry disease

The effectiveness of systemically delivered mRNA encoding human alpha-galactosidase A (h-α-Gal A) for treating Fabry disease, a lysosomal storage disorder, in various animal models was examined.¹⁸⁸ They administered LNP-formulated mRNA to wild-type CD1 mice and observed that h-α-Gal A exhibited a prolonged half-life in the plasma, liver, kidney, and heart following a single treatment.¹⁸⁸

These studies made use of LNP formulation, which further strengthens the acceptability and broad use of LNP in mRNA delivery. Comparing the results of these studies show that systemic mRNA therapy holds promise as a treatment for genetic diseases.

Cystic fibrosis

In contrast to the studies above that investigated systemic administration of LNP-mRNA for the purpose of gene replacement, a clinical study that made use of inhaled mRNA therapy for the treatment of CF was reported.¹⁸⁹ CF transmembrane conductance regulator (CFTR) mRNA was delivered by aerosol in LNPs to CF adults.

Furthermore, this first-in-human study treatment was generally safe and well tolerated, quickly resolving fever; hypersensitivity reactions were noted in some subjects and lung function remained stable after treatment, but no benefit was observed.¹⁸⁹ The result obtained is quite different from the systemic delivery, which raises a concern about the effectiveness of inhalation delivery. However, similar to this result, a previous study had earlier reported the aerosolized delivery of optimized CFTR mRNA using a ReCode LNPs to primary human bronchial epithelial cells derived from patients with CF. They reported toleration and restoration of function in hBE cells derived from CF patients.¹⁹⁰ This, however, is an in vitro study only.

Gene editing therapy

mRNA has emerged as a pivotal vehicle for delivering gene editing tools, including CRISPR-Cas9, into cells. By encoding the Cas9 enzyme within an mRNA molecule, precise genome editing becomes feasible. The Cas9 enzyme, directed by a guide RNA, introduces targeted cuts in the genome to correct mutations or integrate new genetic sequences.¹⁹¹

CRISPR-Cas9 mRNA

CRISPR-Cas9 is the most widely utilized programmable nuclease due to its simplicity and adaptability compared to zinc finger nucleases (ZFNs)¹⁹² and transcription activator-like effector nucleases (TALENs).^193,194 It has achieved significant success in mRNA-mediated delivery, enabling precise targeted insertions and deletions, highlighting its potential in therapeutic gene editing and precision medicine.¹⁹⁵

CRISPR-Cas9 mRNA is particularly promising for T cell engineering, where simultaneous knockouts of programmed cell-death protein 1 (PD-1), T cell receptor (TCR), and human leukocyte antigen class I have enhanced antitumor activity in allogeneic CAR T cells in both in vitro and in vivo studies.¹⁹⁶ Additionally, novel lentiviral vectors with hybrid ΔU3-sgRNAs have further streamlined TRAC locus editing, producing TCR-negative CAR19 T cells with strong antileukemic effects in preclinical models.¹⁹⁷

TALEN/ZLN mRNA

mRNA-based delivery systems have advanced genome engineering, particularly for CAR T cell therapy and precision immunotherapy. Electroporated TALEN mRNA has demonstrated efficient genome editing, achieving over 50% CCR5 knockout with minimal off-target effects in primary T cells.¹⁹⁴ Similarly, combining TALEN mRNA and CRISPR-Cas9 guide RNAs led to an 81% TCR knockout rate in T cells.¹⁹⁸

In comparative studies, ZFN mRNA demonstrated higher specificity and efficient engraftment of CD34+ cells in mice while retaining multilineage differentiation, surpassing TALEN and CRISPR-Cas9 mRNA in preserving stem cell potential.¹⁹⁹ These findings underline mRNA’s versatility in delivering programmable nucleases for therapeutic genome editing.

mRNA-based cancer vaccines and immunomodulatory therapies

A cancer vaccine initiates and amplifies the antitumor immune response by antigen-presenting cells (APCs), especially DCs. mRNA cancer vaccine platforms have been developed and have achieved encouraging outcomes based on their unique efficacy in pushing the cancer immunity cycle and safety,²⁰⁰ while a large number of mRNA vaccines against various cancers have been and are still being evaluated. mRNA vaccines against renal cell carcinoma (RCC), glioblastoma, melanoma, and acute myeloid leukemia (AML) have shown positive and active response to this mRNA-based immunotherapy, which gives a strong indication to further explore intensively the mRNA cancer vaccine field.²⁰⁰

Renal cell carcinoma

In the treatment of RCC, DC-based mRNA vaccines have been developed, which evoked moderate efficacy in the context of advanced RCC treatment.²⁰¹ Also, another mRNA vaccine has been administered to patients with RCC via the intradermal route, and the immune response evoked by the vaccine seems to mediate long-term survival in RCC patients (see Table 3).²¹³ Recent advances in the treatment of RCC using mRNA cancer vaccines borders on the identification of neoantigens and immune subtypes for the development of potent personalized mRNA vaccines.^214,215,216

Table 3.

Clinical trials of mRNA vaccines targeted at various cancer

Disease indication	Target proteins	Name of mRNA product	Delivery/route of administration	Phase	Results	Reference
Melanoma	TPTE, NY-ESO-1, MAGE-A3, tyrosinase	Lipo-MERT	Lipoplex/i.v.	1	1/3 PR, 1/3 SD, and 1/3 RSM	Kranz et al.202
Melanoma	Neoantigens	IVAC MUTANOME	None/I.N.	1	5/13 progression and 8/13 PR	Sahin et al.203
Melanoma	P53, surviving and hTERT	–	DCs/I.D.	1	9/22 SD and 13/22 PD	Borch et al.204
CRPC	PSCA, PSMA, PSA, and STEAP1	CV9103	Protamine/I.D.	1/2	29.3 months median OS (44 patients)	Kübler et al.205
NSCLC	NY-ESO-1, 5T4, MAGE-C1, MAGE-C2, and survivin	CV9201	Protamine/I.D.	1/2	9/29 SD and 20/29 progression	Sebastian et al.206
NSCLC	NY-ESO-1, MAGE-C1, MAGE-C2, survivin, 5T4, and MUC1	CV9202	Protamine/I.D.	1/2	12/26 progression	Papachristofilou et al.207
Pancreatic cancer	Neoantigens	Autogene cevumeran	Lipoplex/i.v.	1	8 of 16 patients responded, higher RFS (>18 months)	Rojas et al.208
AML in CR with high relapse risk	WT1	CCRG-09–003	DCs/I.D.	2	6/30 CR1 and 11/30 CR2	Anguille et al.209
AML IN CR with high relapse risk	hTERT	AST-VAC1	DCs/I.D.	2	11/19 CR	Khoury et al.210
Glioblastoma	pp65	CMV-DC	DCs/I.D.	1	3/6 PR and 3/6 progression	Mitchell et al.211
Glioma	mRNA copy of tumor	DC-CAST-GBM	DCs/I.D.	1/2	5/7 progression	Vik-Mo et al.212

CR, complete remission; CRPC, castration-resistant prostate cancer; hTERT, human telomerase reverse transcriptase; I.D., intradermal; I.N., intranodal; i.v., intravenous; OS, overall survival; PR, progression free; PSA, prostate-specific antigen; PSCA, prostate stem cell antigen, PSMA, prostate-specific membrane antigen; RFS; recurrence-free survival; RMS, restricted mean survival; STEAP1, six-transmembrane epithelial antigen of the prostate 1; SD, stable disease; TR, tumor regression.

Glioblastoma

mRNA vaccination has emerged as a promising therapeutic strategy for glioblastoma treatment. DC-based mRNA vaccines, developed using mRNA transcripts from glioblastoma patients, have been shown to extend progression-free survival by 2.9 times compared to matched controls (NCT00961844).²¹² Similarly, pre-conditioning the vaccine administration site with a potent recall antigen, such as tetanus/diphtheria (Td) toxoid, has been demonstrated to significantly enhance the efficacy of tumor-antigen-specific DCs, thereby increasing bilateral DC migration and markedly improving survival outcomes in glioblastoma patients.²¹¹

Furthermore, a DC-based mRNA vaccine has been specifically engineered to enhance the homing of mRNA-pulsed DCs to lymphoid organs. Follow-up data from the initial blinded, randomized phase 2 clinical trial (NCT00639639) indicated that approximately one-third of the participants achieved sustained tumor remission 5 years post-diagnosis. In a parallel clinical trial (NCT00639639), a 5-year survival rate of 36% was observed from the time of diagnosis.¹ Recently, a personalized and customizable mRNA-based therapeutic strategy that effectively targets multiple tumor antigens, demonstrating robust anti-tumor efficacy in preclinical glioblastoma models, was introduced.²¹⁷ Their findings revealed a significant increase in tumor-infiltrating lymphocytes with enhanced effector functions both within the tumor and systemically following antigen-specific mRNA-directed immunotherapy. This intervention led to a favorable modulation of the tumor microenvironment, transforming it from immunologically cold to hot.

Melanoma

Over the years, seven DC-based and three non-DC-based mRNA vaccines have undergone clinical testing. Among these, one non-DC-based²¹⁸ and one DC-based mRNA vaccine²¹⁹ utilized complete mRNAs derived from tumor cells, while the others employed tumor-associated antigens (TAAs) encoded into mRNAs (See Table 3). However, none of the DC-based mRNA vaccines demonstrated a significant improvement in clinical outcomes for patients with metastatic melanoma. To address these challenges, the TriMix-mRNA, containing mRNAs encoding immunostimulatory molecules such as CD40L, CD70, and caTLR4, was introduced to improve DC-based vaccine efficacy.²²⁰ Additionally, BioNTech developed a personalized mRNA vaccine for metastatic melanoma, which led to the absence of detectable lesions on radiological imaging and maintained recurrence-free status for 23 months following intranodal vaccination (NCT02035956).²⁰³

Given the variability in these outcomes, further researches are being conducted to validate the potential of mRNA vaccines as an effective immunotherapy for melanoma. Carvalho reported the recruitment of resected melanoma patients for a phase 3 trial of personalized anti-cancer mRNA vaccine combined with pembrolizumab, with final results due in 2029.²²¹ In the earlier KEYNOTE-942 open-label phase 2 trial involving 157 participants, patients treated with mRNA-4157 in combination with the anti-PD-1 drug pembrolizumab exhibited a 44% reduction in the risk of post-surgical recurrence or death compared to those receiving pembrolizumab alone.²²²

In a related development, Husseini et al.²²³ advanced the application of iontophoresis technology for the transdermal and intracellular delivery of a minimal mRNA vaccine targeting melanoma. This approach elicited a robust immune response, characterized by the activation of skin-resident immune cells. The combination of iontophoresis and mRNA vaccine technology resulted in a potent stimulation of the immune system, as evidenced by significant tumor inhibition in melanoma-bearing mice. Additionally, there was an upregulation of mRNA expression levels for various cytokines, particularly IFN-γ, and an increased infiltration of cytotoxic CD8+ T cells within tumor tissue, which are critical for tumor clearance.²²³

Similarly, a recent study reported that the non-invasive transdermal administration of mRNA encoding multivalent neoantigens effectively inhibited melanoma growth, further supporting the efficacy of targeting multiple antigens in mRNA vaccine therapy.²²⁴

Acute myeloid leukemia

Two dendritic cell (DC)-based mRNA vaccines have been developed to mitigate the risk of relapse in patients with AML who have achieved complete remission (NCT00510133 and NCT00965224).²⁰⁹ In one study, electroporation of DCs with WT1 mRNA was shown to significantly improve relapse-free survival in patients who responded to vaccination, compared to non-responders.²⁰⁹ This highlights the potential of WT1 mRNA in enhancing the immune response and prolonging remission in AML patients.

Another approach involved the use of mRNA encoding human telomerase reverse transcriptase (hTERT), which was administered through intradermal (i.d.) vaccination. This study reported that 11 of 19 patients in complete remission maintained remission with a median follow-up of 52 months.²¹⁰ These findings suggest that hTERT mRNA vaccines may offer a viable strategy for maintaining long-term remission in AML patients. However, it is important to note that mRNA vaccines may be less effective in patients with progressive AML. The efficacy of these vaccines relies on a functional immune system to mediate the anti-tumor response, and AML is known to impair immune function.²⁰⁹ Therefore, the suitability of mRNA vaccination in AML patients must be carefully considered.

To address this challenge, Wang²²⁵ conducted a study to identify immune subtypes in AML patients, classifying them into two clusters: Cluster 1 and Cluster 2. The study concluded that patients within Cluster 1 are more likely to benefit from mRNA vaccination. Additionally, Wang identified novel antigens that could serve as potential targets for mRNA vaccine development, further refining the approach to AML immunotherapy. This stratification of patients based on immune subtypes could play a crucial role in optimizing the efficacy of mRNA vaccines in AML treatment.²²⁵

mRNA vaccine-induced immune response: mechanisms and implications

In order to allow host cells to express antigens to trigger an immune response, mRNA vehicles are able to deliver nucleic acid molecules encoding antigens of interest to target cells in the human host.¹⁷⁶ In this way, when an antigen-carrying pathogen invades or a tumor cell emerges, the host immune system can rapidly trigger antigen-specific cellular humoral immunological reactions to specifically prevent the disease.¹⁷⁶

mRNA vaccine-triggered immune responses can enter three types of host cells through intramuscular, intradermal, or subcutaneous injections of the mRNA vaccine. These include non-immune cells at the site of the injection (e.g., muscle cells and epidermal cells)¹⁷⁸ and immune cells (e.g., dendritic cells and macrophages) in the tissues at the site of injection,²²⁶ and after the injected mRNA has been transported to nearby lymph nodes or the spleen by the lymphatic system, immune cells in peripheral lymphoid organs will develop.²⁰⁰ After injection, mRNA vaccines can be captured by APCs, initiating both innate and adaptive immune responses, and fully activating humoral and cellular immunity.²²⁷

Interplay of T cell and B cell immune response

T follicular helper cells and germinal centers

Modified mRNA-LNP vaccines effectively stimulate robust T follicular helper (Tfh) cell responses, critical for the formation of germinal centers (GCs) that support long-term, high-affinity antibody production.^105,228 In individuals vaccinated with BNT162b2, both GC B cell and Tfh cell activities in draining lymph nodes remain strong for up to 6 months following the second 30 mg mRNA dose. This activity fosters the development of affinity-matured memory B cells and long-lived plasma cells residing in the bone marrow.^229,230,231 Additionally, COVID-19 mRNA vaccines induce antigen-specific circulating Tfh cells, contributing to the durability of immune responses.²³²

CD4+ and CD8+ T cell responses

COVID-19 mRNA vaccines prompt CD4+ T cell responses skewed toward Th1 polarization and CD8+ T cells producing IFN-γ, persisting for up to 6 months post-booster dose.^232,233,234 Sequence-optimized, chemically unmodified mRNA vaccines encoding rabies virus glycoprotein (RABV-G) elicit measurable RABV-G-specific CD4+ and CD8+ T cell responses, demonstrated by TNF, IFN-γ, CD107a, and interleukin (IL)-2 expression after peptide stimulation in mice and pigs. Notably, the CD4+ T cell response from these vaccines surpasses that of licensed alternatives like Rabipur, although in human trials, CD4+ T cell responses were transient, detected only after three doses.^18,179,235

While mRNA vaccines encoding RABV-G or influenza HA induce strong CD4+ T cell responses in mice and cynomolgus monkeys, CD8+ T cell responses are more pronounced in mice and often undetectable in monkeys or humans. For instance, modified mRNA vaccines encoding influenza HA elicited T cell responses in mice but failed to generate CD8+ T cell responses in rhesus macaques or detectable T cell responses in a phase I clinical trial.^{228,236,237,238} This discrepancy may stem from the HA’s weak T cell antigenicity.

B cell activation

The efficacy and quality of antibody responses by mature B cells following vaccination with LNP-formulated mRNA are influenced by nanoparticle size, which facilitates transport to draining lymph nodes. This allows direct targeting of B cells within these sites.²³⁹ Although B cells can internalize LNPs and synthesize protein antigens from encoded mRNA, their role in antigen production is less prominent than monocytes and dendritic cells. Instead, B cells primarily engage antigens presented or secreted by neighboring cells.²³⁶ CD4+ T cells play a crucial role in B cell differentiation and memory response establishment, further supporting robust B cell activation post-vaccination (see Figure 3).

Figure 3.

mRNA vaccine-mediated immune response. mRNA therapeutics involves three critical stages: synthesis, intracellular processing, and immune activation

First, in vitro-transcribed mRNA is packaged into delivery vehicles, such as lipid nanoparticles, and internalized by APCs. Upon endosomal escape, the mRNA reaches the cytoplasm, where ribosomes translate it into target antigens. These endogenous antigens are processed by the proteasome, loaded onto major histocompatibility complex (MHC) class I molecules, and presented to cytotoxic T cells (CD8+ T cells), triggering cellular immunity. Alternatively, the secreted antigens are taken up by APCs, degraded in endosomes, and displayed via MHC class II to helper T cells (CD4+ T cells). This pathway stimulates B cells to produce neutralizing antibodies, completing the humoral immune response. Together, these mechanisms enable mRNA drugs to elicit robust and adaptive immune protection.

Th1/Th2 immune response balance

The balance between Th1 and Th2 immune responses plays a critical role in the pathophysiology of various diseases, including cancer, autoimmune disorders, and viral infections. The Th1/Th2 paradigm is characterized by distinct cytokine profiles, with Th1 cells producing cytokines such as IFN-γ and TNF-α, which are involved in cell-mediated immunity, and Th2 cells producing cytokines like IL-4 and IL-5, which support humoral immunity. Alterations in this balance can significantly impact disease progression and patient outcomes.²⁴⁰

In cancer, particularly, the Th1/Th2 ratio is often disrupted, leading to a skewed immune response that can favor tumor growth. A decreased Th1/Th2 ratio has been documented in patients with various malignancies, including glioblastoma, metastatic melanoma, non-Hodgkin’s lymphoma, breast cancer, and head and neck cancer.^241,242,243 This shift toward a Th2-dominated response is associated with a suppressed cell-mediated immune response, which may allow tumor cells to evade immune surveillance. Conversely, a Th1-dominant phenotype in breast cancer is linked to better patient survival, particularly in cases with the highest ratios of Th1 cytokines to IL-5 levels. This suggests that a strong Th1 response may contribute to controlling more aggressive cancer phenotypes, such as estrogen receptor-negative and triple-negative breast cancer.²⁴⁴ The association of Th1 dominance with better outcomes is particularly strong in premenopausal women, hinting at the potential influence of immunosenescence and age-related changes in immune function in postmenopausal women.

Rhesus macaques vaccinated with Moderna’s mRNA vaccine predominantly elicited a Th1-skewed CD4+ T cell response, as identified through intracellular cytokine staining, with minimal or undetectable Th2 and CD8+ T cell responses.¹¹¹ In contrast, a study by Zhu et al.,²⁴⁵ reported that a cGAMP-adjuvanted multivalent influenza mRNA vaccine generated strong and balanced Th1 and Th2 cellular responses in mice, accompanied by robust antigen-specific antibody production and an increase in cytokine-secreting splenocytes. These findings underscore the notion that Th2 and CD8+ T cell responses are more pronounced in mice compared to other species such as monkeys and humans. Additionally, Tai et al.²⁴⁶ observed a Th1-biased response in humanized transgenic mice and macaques vaccinated with an mRNA-based T-cell-inducing antigen and COVID-19 vaccine. This further supports the previous findings by Corbett and Bahl, reinforcing the tendency of mRNA vaccines to induce Th1-biased responses across different species, with Th2 response more pronounced in mice.

Therapeutic strategies that manipulate the Th1/Th2 balance hold promise for improving cancer treatment outcomes.²⁴⁷ Reconstituting Th1 cells has shown potential in preventing CD8+ T cell decay and enhancing CD8+ T cell activity, as demonstrated in models of lymphocytic choriomeningitis virus (LCMV) infection.²⁴⁸ This approach may be valuable in cancer therapy by shifting tumor antigen-specific T cell responses toward a more immunostimulatory Th1 bias, which could enhance the efficacy of existing treatments. Additionally, polarizing the immune response has been proposed as a strategy for developing vaccines tailored to induce specific immune phenotypes, offering optimal protection against pathogens like the influenza virus.²⁴⁹

The Th1/Th2 balance also influences the severity of diseases beyond cancer. In autoimmune diseases like rheumatoid arthritis, a Th1-skewed immune response exacerbates inflammation,²⁵⁰ while in diseases such as AIDS, a shift toward Th2 immunity increases the risk of immunosuppression.²⁵¹ Modulating this balance with treatments like dexamethasone, which can rebalance Th1/Th2 responses and influence cytokine engagement with receptors, offers a potential therapeutic avenue in these contexts.²⁵²

IgG1/IgG2a immune response index

The IgG1 and IgG2a subclasses of antibodies are pivotal markers of immune response polarization, particularly in preclinical murine models.²⁵³ These immunoglobulin subclasses reflect Th2 and Th1 immune responses, respectively, offering insight into the immunological pathways activated by vaccines.²⁵⁴

IgG1 and IgG2a production is tightly regulated by cytokine milieu. Th2-polarized immune responses, characterized by IL-4, favor IgG1 production, indicating robust humoral immunity suited for neutralizing extracellular pathogens.^255,256 Conversely, Th1 responses, driven by IFN-γ, promote IgG2a production, reflecting enhanced cellular immunity vital for combating intracellular pathogens and tumors.²⁵⁷

mRNA vaccines have demonstrated the ability to induce both Th1- and Th2-biased responses, but the extent of polarization depends on antigen design, delivery system, and context.²⁵⁸ For instance, infectious disease vaccines generally aim for a balanced Th1/Th2 response to ensure effective pathogen clearance and immune memory. In contrast, cancer vaccines prioritize Th1 responses to potentiate cytotoxic T cell activation and anti-tumor immunity, often resulting in higher IgG2a levels.²⁵⁵

Infectious disease vaccines

Studies using mRNA vaccines encoding viral antigens, such as the HA of influenza virus, show a balanced induction of IgG1 and IgG2a. For example, Park et al.²⁵⁸ demonstrated that LNP-encapsulated mRNA vaccines against influenza elicited robust humoral and cellular immunity, including IgG1 and IgG2a responses, in murine models especially when primed. Similarly, Yavuz et al.²⁵⁹ reported balanced IgG subclass responses in preclinical studies of mRNA vaccines targeting viral pathogens, underscoring their capacity to engage both Th1 and Th2 pathways. This balance is crucial for neutralizing the virus while activating complement and Fc receptor-mediated pathways for viral clearance.

Cancer vaccines

Research on cancer mRNA vaccines, including those targeting neoantigens or overexpressed tumor antigens, demonstrates a preferential Th1 response. For instance, studies have shown that cancer mRNA vaccines encoding tumor-specific antigens, such as neoantigens, drive robust IFN-γ production, promoting IgG2a over IgG1. In particular, Chen et al.²⁶⁰ demonstrated that LNP-encapsulated mRNA vaccines targeting tumor antigens induced strong Th1-biased immunity in murine models. Additionally, Husseini et al.²²³ highlighted that a minimal mRNA vaccine encoding patient-specific tumor-associated antigenic epitope elicited high CD8+ T cell activation, correlating with a Th1-skewed IgG2a response.

Furthermore, the IgG1/IgG2a ratio serves as a quantitative marker of immune response polarization, as demonstrated by experimental studies that evaluate the balance of Th1 and Th2 immune pathways. For instance, research by Snapper and Paul showed that IFN-γ and IL-4 cytokines regulate IgG subclass switching in murine models, offering insights into how this ratio can reflect underlying immune mechanisms.²⁶¹ Further studies have corroborated the utility of this ratio in assessing mRNA vaccine efficacy in eliciting targeted immune responses, particularly in pre-clinical settings.^223,258,259 A low IgG1/IgG2a ratio indicates a Th1-skewed response, desirable for cancer immunotherapy and intracellular infections, while a high ratio signals Th2 dominance, suited for extracellular pathogens. In our group, optimization of this ratio in mRNA vaccine design is being tested via

• (1)Antigen engineering: Incorporating epitopes that selectively activate Th1 or Th2 responses.
• (2)Adjuvants and delivery systems: LNPs and other adjuvants can modulate cytokine profiles, influencing IgG subclass responses.²⁵⁹

IgG1 and IgG2a responses, along with their ratios, are critical parameters for evaluating mRNA vaccine efficacy. A deeper understanding of the molecular pathways governing subclass switching can guide the rational design of vaccines tailored to specific immunological needs.

Future research directions

The evolution of mRNA therapeutics will be driven not by isolated breakthroughs but by the synergistic integration of multiple advancing frontiers: scalable data analytics, deeper mechanistic understanding, expanded delivery capabilities, and enhanced clinical precision. Key to this progress is the transition from small-scale discovery to systematic, high-throughput exploration of mRNA structural elements (5′ cap, UTRs, poly-A tail) within vast molecular landscapes. Foundation models pre-trained on comprehensive datasets of ionizable lipids, polymers, and hybrid nanoparticles—followed by institution-specific fine-tuning—could overcome the batch-effect inconsistencies that currently hinder cross-platform reproducibility (e.g., discrepancies between Moderna, Acuitas, and Protiva datasets).

A critical challenge lies in diversifying delivery beyond hepatic dominance. Achieving organ-selective tropism (e.g., lung, spleen, CNS) will require the development of targeting ligands discovered through diffusion-based generative models trained on high-resolution imaging data. Concurrently, AI-driven formulation design must optimize LNPs for alternative administration routes (intranasal, intradermal, intrathecal), each presenting unique biophysical constraints (ionic strength, mucosal barriers, immune-microenvironment interactions). Further acceleration may come from dynamic “closed-loop” platforms, where real-time pharmacokinetic data from implanted biosensors inform Bayesian optimization algorithms, enabling rapid dose adjustments in vivo and drastically shortening iterative preclinical development.

Finally, mRNA therapeutics will shift from broad applications to patient-specific precision medicine. For oncology, neoantigen vaccines could be computationally designed against an individual’s tumor phylogeny, with AI predicting epitope immunogenicity while co-optimizing LNP chemistry to align with the patient’s immune-metabolic profile.²⁶² In protein replacement therapies, codon-stabilized, chemically modified mRNAs could be encapsulated in LNPs engineered to evade tissue-specific microRNA-mediated silencing, minimizing off-target immune activation.

Collectively, these innovations point toward a future where mRNA drugs are computationally designed, optimized, and personalized before being rapidly synthesized via automated microfluidic systems—replacing traditional trial-and-error workflows with a precision-driven, high-throughput paradigm.

Conclusion

In conclusion, although mRNA therapeutics have showcased immense promise across various clinical applications, their widespread success will ultimately depend on addressing several key challenges. Chief among these are improving delivery mechanisms to ensure targeted and efficient cellular uptake, enhancing the stability of mRNA molecules to prolong their therapeutic efficacy, reducing unintended immunogenicity that can lead to adverse reactions, and scaling up production to meet the global demand without compromising quality. However, ongoing innovations in these areas are rapidly driving progress. With continued advancements, mRNA technology is expected to evolve into a robust and adaptable platform, offering transformative solutions for the treatment and prevention of a broad spectrum of diseases, including those that have historically been difficult to target with conventional therapies. The future of mRNA therapeutics holds significant potential to revolutionize modern medicine.

Acknowledgments

This study was supported in part by the Shenzhen Medical Research Fund under Grant No. D2403003 (X.W.) and Shenzhen Basic Research Fund under Grant No. JCYJ20190807170801656 (J.L.).

Author contributions

X.W. and J.L. designed the study. M.A.I. wrote the manuscript. J.L., X.R., Y.Y., H.Z., C.S., and Y.J. revised manuscript. X.W. and J.L. supervised the study. All authors have read and approved the final manuscript.

Declaration of interests

The authors declare no competing interests.