mRNA Cancer Vaccines: From Pandemic Paradigm to Personalized Oncology Therapeutics

ABSTRACT

The groundbreaking success of messenger RNA (mRNA) vaccines during the COVID‐19 pandemic has significantly accelerated their application in oncology. This review comprehensively synthesizes the recent advancements in mRNA cancer vaccine development, emphasizing three critical domains: mechanistic innovations, clinical translation, and ongoing challenges. Technologically, advancements in nucleotide modification, lipid nanoparticle (LNP) delivery systems, and AI‐driven neoantigen selection have significantly improved vaccine stability, immunogenicity, and personalization. Clinically, more than 150 trials have demonstrated the synergistic efficacy of mRNA vaccines (e.g., mRNA‐4157/V940, BNT122) in combination with immune checkpoint inhibitors (ICIs), particularly in melanoma, with Phase III trials currently underway. Individualized neoantigen vaccines targeting patient‐specific mutations have shown unprecedented response rates (> 50% in certain cohorts), while shared‐antigen vaccines are progressing for high‐incidence cancers. However, several critical challenges remain: (1) overcoming immunosuppressive tumor microenvironments (TME), (2) addressing systemic toxicities and LNP‐related limitations, (3) scaling up cost‐effective personalized manufacturing, and (4) optimizing targeted delivery. Future research directions encompass self‐amplifying mRNA constructs, novel biomaterial vectors, neoadjuvant applications, and multi‐omics integration for next‐generation vaccine development. With rapid industrialization and evolving regulatory frameworks, mRNA vaccines are well‐positioned to revolutionize precision cancer immunotherapy despite persistent translational barriers.

This review highlights the transformative potential of mRNA cancer vaccines, leveraging technological advancements in nucleotide modification, lipid nanoparticle delivery, and AI‐driven neoantigen selection to enhance vaccine stability, immunogenicity, and personalization. We discuss clinical progress and future directions, emphasizing the promise of mRNA vaccines in revolutionizing precision cancer immunotherapy despite ongoing challenges. (The graphical abstract image was created in BioRender. Liu, J. (2025). https://BioRender.com/dvgzb4a.)

Abbreviations

AEs

adverse events

AI

artificial intelligence

APA

alternative polyadenylation

AML

acute myeloid leukemia

CTC

circulating tumor cell

CTCs

circulating tumor cells

ctDNA

circulating tumor DNA

CTL

cytotoxic T lymphocyte

DCR

disease control rate

DC

dendritic cell

DLT

dose‐limiting toxicities

dsRNA

double‐stranded RNA

EPR

enhanced permeability and retention

FDA

Food and Drug Administration

GBM

glioblastoma

HLA

human leukocyte antigen

ICI

immune checkpoint inhibitor

IFN‐γ

interferon‐gamma

IL

interleukin

IVT

in vitro transcription

LNP

lipid nanoparticle

MDSCs

myeloid‐derived suppressor cells

MHC‐I

major histocompatibility complex class I

MHC‐II

major histocompatibility complex class II

mRNA

messenger RNA

OAS

oligoadenylate synthetase

ORR

objective response rate

OS

overall survival

pMHC

peptide‐MHC complex

PDAC

pancreatic ductal adenocarcinoma

PFS

progression‐free survival

RBPs

RNA‐binding proteins

RNA‐seq

RNA sequencing

RFS

recurrence‐free survival

saRNA

self‐amplifying RNA

TAAs

tumor‐associated antigens

TCR

T cell receptor

Th cells

helper T cells

TME

tumor microenvironment

TMZ

Temozolomide

TSAs

tumor‐specific antigens

Tregs

regulatory T cells

VCE

vaccinia capping enzyme

VLPs

virus‐like replicon particles

WES

whole‐exome sequencing

1. Introduction

Cancer remains one of the most pressing global health challenges, with its therapeutic strategies undergoing significant evolution over the past several decades. The evolution has transitioned from the nonspecific cytotoxicity of conventional chemotherapy to the precision of targeted therapies and, most recently, to the era of immunotherapy, exemplified by ICIs and cell‐based therapies [1, 2]. This paradigm shift—from direct cytotoxicity to leveraging the host's antitumor immune response—represents a fundamental transformation in treatment philosophy. It has established a novel therapeutic framework designed to overcome immune tolerance within the TME [3, 4, 5].

The coronavirus disease 2019 (COVID‐19) pandemic marked a pivotal milestone in validating the potential of messenger RNA (mRNA) vaccine technology [6, 7, 8, 9]. mRNA vaccines utilizing lipid nanoparticle (LNP) delivery systems, such as BNT162b2 and mRNA‐1273, have exhibited exceptional clinical efficacy, achieving protection rates exceeding 95% [10, 11, 12]. Their unprecedented development speed—from sequence design to clinical‐grade production within weeks—robust immunogenicity, and modular manufacturing platforms have revolutionized the landscape of infectious disease prevention [13] and established a robust foundation for mRNA technology application in cancer therapeutics.

The distinctive value of cancer mRNA vaccines lies in their three key attributes. First, their rapid development capability facilitates the design and production of personalized neoantigen vaccines within 2–4 weeks following tumor sequencing [14]. Second, their molecular programmability permits the encoding of shared tumor‐associated antigens (TAAs), such as MAGE‐A3 and NY‐ESO‐1, while concurrently incorporating immune modulatory factors, such as IL‐12 and OX40L, to reshape the immunological microenvironment [9, 15, 16]. Third, their multidimensional immune activation potential supports dendritic cells’ (DCs) cross‐presentation, thereby activating both CD8⁺ cytotoxic T cells and CD4⁺ helper T (Th) cells, promoting epitope spreading, and facilitating the establishment of durable immune memory [17]. These three attributes position mRNA vaccines as a promising platform for addressing tumor heterogeneity and immune evasion.

Despite significant advancements in antigen design and delivery systems for mRNA cancer vaccines, the clinical translation of these vaccines continues to face substantial challenges. The principal scientific bottlenecks can be broadly categorized into three areas: (1) overcoming the immunosuppressive TME, which presents considerable barriers; (2) addressing the inherent trade‐offs between precision and safety in delivery systems; and (3) overcoming the significant challenges in the industrial scalability of personalized vaccines. Resolving these challenges requires interdisciplinary collaboration and technological innovation.

This review offers a comprehensive and systematic analysis of the pivotal advancements in the field of mRNA cancer vaccines. It explores the technological innovations in nucleotide modification and delivery systems, critically assesses the clinical translational outcomes of personalized vaccines and combination therapies, and investigates the field's core challenges, such as overcoming immune suppression within TMEs and optimizing delivery efficiency. It also discusses forward‐looking perspectives on industrial scalability, encompassing artificial intelligence (AI)‐driven designs, exosome‐based delivery platforms, and room‐temperature‐stable formulations. This review thus aspires to serve as a critical reference for advancing research and development in the domain of mRNA cancer vaccines, thereby accelerating the broader implementation of these vaccines in cancer treatment.

2. Mechanism of Action of mRNA Vaccines

2.1. Basic Structure of mRNA Vaccines

The core structure of mRNA vaccines is composed of four fundamental structural elements: the 5′ cap, untranslated regions (UTRs), the open reading frame (ORF), and the 3′ polyadenylated tail (Poly‐A tail) [18]. Located at the 5′ end of the mRNA strand, the cap is a critical component for maintaining mRNA stability, facilitating efficient translation, and promoting ribosome recruitment. It shields mRNA from intracellular nucleases and assists in recruiting translation initiation factors. UTRs, which flank the coding sequence, are frequently modified to enhance both mRNA stability and translational efficiency. For example, ribosome loading efficiency is substantially affected by the structural arrangement of the 5′ UTR [19]. The ORF encodes genetic instructions for the target antigen, including tumor‐specific antigens (TSAs) or immune‐modulating molecules, which are translated into functional proteins within cells [20, 21]. Positioned at the 3′ end of the mRNA, the polyadenylated tail contributes to molecular stability and prolongs the intracellular retention time (Figure 1). To further protect mRNA from enzymatic degradation and improve delivery, mRNA vaccines are routinely encapsulated within LNPs [12, 22, 23]. Each LNP typically encapsulates 1 to 10 mRNA strands, depending on the mRNA length (0.5–15 kb) and the LNP formulation [24, 25, 26]. For multi‐antigen vaccines (e.g., encoding both tumor‐specific and immune‐modulating antigens), distinct mRNAs are coloaded into single LNPs during synthesis [27]. Clinically, this allows the simultaneous delivery of multiple antigenic payloads within a unified carrier system. In contrast, combination therapies using separate antigen sets (e.g., personalized neoantigens + shared cytokines) may utilize mixed LNP formulations administered concurrently [28, 29]. These LNPs safeguard mRNA from degradation, facilitate cellular uptake, and enable intracellular release, ultimately improving the vaccine's overall efficacy.

Figure 1.

The basic structure of mRNA vaccines and potential optimization strategies. Created in BioRender. Liu, J. (2025). https://BioRender.com/dvgzb4a.

2.2. Targeted Delivery and Cellular Uptake

The mRNA vaccine delivery system is dependent on LNPs functionalized with polyethylene glycol (PEG) to extend their circulation time in the bloodstream and enhance their accumulation in lymphatic tissues [12, 23, 30]. The cellular uptake of LNPs is predominantly facilitated through clathrin‐mediated endocytosis. After being internalized into endosomes, characterized by a pH below 6.5, ionizable lipids, such as SM‐102, undergo protonation. This protonation triggers fusion between the endosomal membrane and the LNPs, facilitating mRNA's efficient release into the cytoplasm [31]. DCs, as key antigen‐presenting cells, express scavenger receptors on their surface, which selectively recognize and bind to LNPs, thereby enabling highly efficient nanoparticle uptake [32].

2.3. Cytosolic Translation and Antigen Protein Synthesis

Once mRNA enters the cytoplasm, ribosomes recognize its 5′ cap structure and bind to it, initiating the translation process [33, 34]. The 5′ cap interacts with the eukaryotic translation initiation factor eIF4E, promoting ribosome recruitment and initiating translation. Both TSAs (e.g., KRAS G12D and mutant p53) and shared tumor antigens (e.g., MAGE‐A3 and NY‐ESO‐1) are synthesized through this mechanism [34]. TSAs are tailored to a patient's individual mutational profile, thus providing high specificity, whereas shared tumor antigens are broadly applicable due to their universal expression across multiple patients.

Furthermore, chemical modifications to mRNA—incorporating modified nucleotides such as pseudouridine, for instance—can inhibit the activation of innate immune receptors such as RIG‐I and TLR7 [35, 36, 37]. These modifications mitigate excessive inflammatory responses, such as interferon storms, that could otherwise impair the translation process. By attenuating innate immune activation, these modifications ensure the efficient synthesis of antigenic proteins [38, 39].

2.4. Immune Response Activation and Regulation

After the antigenic proteins are synthesized, they must be processed and presented to elicit specific immune responses. The major histocompatibility complex class I (MHC‐I) pathway is characterized by the proteasomal degradation of cytosolic antigens into smaller peptide fragments [40, 41]. The transporter associated with antigen processing (TAP) transports these fragments into the endoplasmic reticulum, where they associate with MHC‐I molecules to form peptide–MHC‐I complexes [42, 43] that are then expressed on the cell surface, where they stimulate CD8⁺ cytotoxic T lymphocytes (CTLs) [42, 44]. Conversely, the major histocompatibility complex class II (MHC‐II) pathway facilitates the processing of extracellular antigens, which the DCs internalize and process into peptides within lysosomes. These peptides are presented to CD4⁺ Th cells [45, 46].

DCs possess a unique ability known as cross‐presentation, which enables the simultaneous activation of MHC‐I and MHC‐II pathways, resulting in a comprehensive and robust immune response [47]. Specifically, CD8⁺ CTLs identify the antigenic peptides presented by the MHC‐I molecules and undergo activation and differentiation into effector T cells capable of directly lysing tumor cells expressing the target antigens. In parallel, the CD4⁺ Th1 cells identify the peptides presented by the MHC‐II molecules and subsequently secrete cytokines, such as interferon‐gamma (IFN‐γ) and interleukin‐2 (IL‐2) [48, 49], that augment CTL cytotoxicity, facilitate B‐cell‐mediated antibody production, and support the establishment of immunological memory.

2.5. Antitumor Effector Functions and Dynamic Maintenance of Immunological Memory

Activated CD8⁺ CTLs mediate direct tumor cell killing by releasing perforin and granzymes, which initiate apoptosis in tumor cells [15]. The formation of immunological memory is a pivotal mechanism by which mRNA vaccines achieve sustained antitumor efficacy. Central memory T cells (T_CM) persist after the initial immune response and actively patrol the peripheral tissues [50, 51]. Upon tumor recurrence or reexposure to the same antigen, T_CM rapidly proliferate and differentiate, triggering a rapid and robust immune response that effectively inhibits tumor relapse and metastasis. Long‐lived plasma cells play a complementary role, sustaining antibody titers and continuously producing tumor antigen‐specific antibodies, which facilitate tumor cell clearance via antibody‐dependent cellular cytotoxicity (ADCC) and complement‐dependent cytotoxicity (CDC) [52, 53].

Furthermore, epitope spreading significantly enhances the long‐term efficacy of mRNA vaccines [52, 54]. Immune responses initially directed against primary antigens can subsequently induce secondary immune responses targeting other related antigens, thereby expanding and diversifying the immune repertoire. This mechanism effectively mitigates tumor heterogeneity and lowers the risk of immune evasion caused by antigenic mutations.

By enabling precise delivery and cellular uptake, efficient translation and antigen expression, comprehensive immune activation, and durable antitumor effects through the formation of immunological memory, mRNA vaccines represent a transformative and highly promising strategy for cancer immunotherapy (Figure 2). With continuous technological advancements and expanded clinical investigations, mRNA vaccines are well‐positioned to deliver significant breakthroughs in cancer immunotherapy, providing substantial clinical benefits to patients.

Figure 2.

The immune mechanisms underlying mRNA‐based tumor vaccines comprise a sequential process that engages both cellular and humoral immune pathways. Created in BioRender. Liu, J. (2025). https://BioRender.com/dvgzb4a. These vaccines, composed of mRNA encapsulated within lipid nanoparticles, are internalized by DCs following administration. Inside the DCs, the mRNA is translated into antigenic proteins and subsequently presented to T cells through MHC I or MHC II pathways. Additionally, the cellular immune pathway is further stimulated by the coordinated actions of cytokines, such as interleukin‐1 (IL‐1), interleukin‐2 (IL‐2), and interleukin‐12 (IL‐12). Importantly, the antigens released by antigen‐presenting cells (APCs) can activate B cells, which, under the guidance of CD4⁺ T cells, produce neutralizing antibodies, thereby amplifying the antitumor immune response. The immune activation process proceeds through distinct sequential stages: (1) Vaccine uptake via clathrin‐mediated endocytosis, where LNPs are internalized by antigen‐presenting cells, particularly DCs expressing scavenger receptors. (2) Endosomal escape triggered by pH‐sensitive ionizable lipids (pH < 6.5), facilitating efficient mRNA release into the cytoplasm with up to 90% escape efficiency. (3) Cytosolic translation, where ribosomes recognize the 5′ cap structure and synthesize antigenic proteins, with modified nucleotides reducing interferon responses by 85%. (4) MHC‐I presentation through proteasomal degradation and TAP‐mediated transport, activating CD8⁺ CTLs for direct tumor cell killing. (5) MHC‐II presentation of extracellular antigens, activating CD4⁺ Th cells that secrete cytokines (IFN‐γ, IL‐2) and enhancing B cell antibody production. (6) Cross‐presentation by DCs, uniquely enabling simultaneous MHC‐I and MHC‐II pathway activation for a comprehensive immune response. (7) Effector T cell activation leading to perforin‐granzyme‐mediated tumor cell apoptosis and antibody‐dependent cellular cytotoxicity.

3. Key Technological Breakthroughs in mRNA Cancer Vaccines

3.1. Optimization of mRNA Molecular Design

The rational design of mRNA sequences is a critical determinant of vaccine efficacy, comprising five canonical structural elements that collectively modulate stability, translational efficiency, and immunogenicity. Beyond the canonical structural descriptions widely known in the field, the optimization principles guiding mRNA design employ sophisticated engineering strategies that quantitatively affect biological performance.

3.1.1. 5′ Cap and Poly(A) Tail Optimization

A cornerstone of mRNA stability and translational efficiency, the 5′ cap (m7GpppN) constitutes one of the most critical optimization targets in mRNA vaccine design [55, 56]. Third‐generation co‐transcriptional capping technology, CleanCap, achieves 90%–99% capping efficiency, representing a marked advance over vaccinia capping enzyme (VCE) methods, which yield approximately 60%–80% efficiency [55, 57]. The Cap 1 structure (m7GpppNm) offers dual benefits, as it enhances mRNA stability and attenuating RIG‐I recognition, and has been associated with up to 85% reduction in TLR7‐mediated IFN‐α production. This optimization is particularly important, as it reduces non‐self RNA recognition, thereby preserving translational capacity while maintaining a balanced immunogenic profile [55, 56].

The poly(A) tail length is another critical determinant of mRNA stability, with optimal lengths varying across cell types [58]. In DCs, poly(A) tails of approximately 120–150 nucleotides afford maximal stability and translational efficiency, whereas human T lymphocytes benefit from tail lengths exceeding 300 nucleotides [59, 60]. Engineered poly(A) architectures—such as the 30A + 10GCAUAUGACU + 70A configuration employed in BNT162b2—and the incorporation of 20% cytidine modifications have been reported to extend mRNA half‐life from 3 to 18 h and to yield a 3.4‐fold increase in translational efficiency. These quantitative improvements can translate into enhanced antigen presentation and stronger immune responses, thus underscoring the profound impact of precise structural optimization on vaccine efficacy [61, 62].

3.1.2. Rational Design of Untranslated Regions (UTRs)

The UTRs of mRNA serve as sophisticated regulatory elements that critically influence stability, translational efficiency, and subcellular localization. Once regarded as simple structural elements, UTRs are now recognized as complex regulatory hubs that require rational design strategies to optimize therapeutic efficacy. Optimizing 5′ and 3′ UTRs entails distinct yet complementary strategies, and recent advances have revealed cell type‐specific requirements and modification‐dependent design principles that can substantially affect vaccine performance.

Optimizing 5′ UTR sequences is a critical determinant of translational efficiency, with structural modifications yielding measurable increases in protein expression. According to recent studies, optimizing the 5′ UTR secondary structure by reducing complex folding patterns significantly increases the ribosome‐binding site (RBS) accessibility, resulting in a 1.73‐fold increase in DPEase activity and a 1.98‐fold increase in mRNA levels in Bacillus subtilis [63]. The physical‐barrier model further clarifies this mechanism, suggesting that specific 5′ UTR architectures can guide ribosomes toward downstream start codons, thereby substantially enhancing the translational output [19]. These findings underscore the importance of structural optimization in 5′ UTR design, moving beyond simple sequence considerations to embrace three‐dimensional structural constraints. The emergence of nucleotide‐modification‐dependent design principles adds another layer of complexity to 5′ UTR optimization. 1‐Methylpseudouridine (m1Ψ)‐modified mRNA may require tailored 5′ UTR sequences (m1‐5′ UTR), with optimal designs differing substantially from those for unmodified mRNA [64, 65]. This modification‐dependent optimization suggests that nucleoside chemical modifications and UTR structure should be co‐optimized to achieve maximal translational efficiency [64, 65]. High‐throughput screening combined with machine‐learning algorithms has advanced 5′ UTR discovery, with randomized libraries and polysome profiling indicating that UTR performance can be conserved across cell types. Moreover, deep‐learning models have been used to design 211 novel 5′ UTR sequences with a predicted ~8% improvement in translational efficiency, thus underscoring the value of computational approaches in rational mRNA design [66, 67].

The 3′ UTR serves as a multifaceted regulatory element that modulates mRNA stability, translational efficiency, and subcellular localization through mechanisms including alternative polyadenylation (APA), interactions with RNA‐binding proteins (RBPs), and closed‐loop formation. APA is a critical regulatory mechanism, particularly during T‐cell activation, in which shorter 3′ UTR isoforms become enriched and modulate mRNA stability, translation, and localization [68]. Paradoxically, longer 3′ UTRs containing AU‐rich elements (AREs) enhance mRNA stability and translational output, as exemplified by H2AFY, which promotes the proliferation and invasion of circulating tumor cell (CTC) clusters [68, 69]. This dual functionality underscores the context‐dependent nature of 3′ UTR regulation and the need for careful sequence design. Moreover, the role of RBPs in 3′ UTR function is being increasingly recognized—proteins such as TIAR bind to specific 3′ UTR sequences to regulate mRNA localization and translation [70, 71]. Studies on Nrf2 in Spodoptera frugiperda indicate that TIAR binding to 3′ UTR sequences is essential for proper mRNA localization—interference with TIAR causes the dissociation of mRNA from the translational machinery [70]. The strategic removal of degradation‐promoting elements can partially restore reporter gene activity, thus demonstrating the functional importance of specific sequence motifs. Engineering the 3′ UTR to promote closed‐loop formation represents an additional optimization strategy; tandem β‐globin 3′ UTRs enhance the eIF4E–eIF4G–PABP interactions to form stable closed loops that extend mRNA half‐life to more than 24 h [72]. Moreover, systematically removing the AREs, GU‐rich sequences, and microRNA (miRNA) binding sites can prevent the activation of decay pathways while preserving essential regulatory functions [73].

Comprehensively optimizing the UTR function requires integrated approaches that consider the synergistic interactions between the 5′ and 3′ UTR elements. Global design platforms, such as mRNAdesigner, have emerged as useful tools for integrated UTR optimization, combining the 5′/3′ UTR design with codon adaptation index (CAI) adjustment and GC‐content optimization to enhance mRNA stability and translational efficiency in eukaryotic cells [73]. These computational approaches enable simultaneous optimization of multiple parameters, moving beyond single‐element modifications to adopt holistic mRNA design strategies. Integrating dynamic modification regulation—for example, m6A modification of AKT1 mRNA during arsenic carcinogenesis—illustrates how METTL3‐catalyzed and YTHDF1‐recognized marks can enhance mRNA stability through coordinated changes across the 3′ UTR, coding sequence, and 5′ UTR [74]. The clinical validation of UTR optimization has been exemplified by the use of human α‐globin UTRs in the BNT162b2 vaccine, thus supporting the feasibility of cell type‐specific, high‐efficiency expression strategies. Moreover, the high‐throughput screening of viral‐derived UTRs has identified more than 200 candidate sequences for customized design applications, expanding the toolkit for rational mRNA engineering [75].

3.1.3. Codon and Open Reading Frame Optimization

A sophisticated aspect of mRNA therapeutic design, codon optimization extends beyond simple synonymous substitution, encompassing the complex interplay between translation kinetics, protein folding, and biological function [76]. The relationship between codon usage and protein folding is a fundamental principle in mRNA therapeutic design, with the translational elongation rate being a critical determinant of protein conformation. Since ribosomal elongation typically lags behind folding reactions, and different codons are decoded at distinct rates, codon selection provides a direct means of regulating co‐translational folding. This relationship is especially important for proteins with complex folding pathways or kinetically trapped intermediates, such as the KRAS G12D antigen, where early folding events may occur only once, making codon‐mediated translational pauses critical in shaping the final conformation [76]. The evolutionary conservation of codon‐usage patterns within homologous gene families suggests the selection of folding‐efficiency optimization, with slow‐translating regions functioning as programed pause sites that enable stepwise domain folding.

While codon optimization can enhance translational efficiency by aligning with the host's high‐expression codon preferences, excessive optimization carries risks for protein‐folding fidelity [77, 78]. Maximizing the translational speed via codon‐frequency matching can inadvertently compromise protein quality, as rapid elongation may not allow sufficient time for proper domain folding. Experimental evidence indicates that some proteins require specific translational kinetics to achieve correct conformations, and fast translation can lead to local misfolding or aggregation of nonfunctional species. Modern optimization strategies should balance multiple factors, including GC‐content enhancement and uridine reduction, while avoiding the formation of high‐stability secondary structures. Hairpin loops with free energy (ΔG) below −15 kcal/mol can increase the ribosome dissociation rates by up to 220%, thus underscoring the importance of structural considerations in codon optimization [79, 80].

Codon optimization has progressed from simple frequency matching to sophisticated, multi‐parameter design strategies that integrate translation dynamics, mRNA structural stability, host resource limitations, and folding energetics [81, 82]. Advanced algorithms—for example, CUSTOM—go beyond frequency matching to incorporate tissue‐specific codon preferences, enabling more nuanced optimization [80]. By considering tissue‐specific tRNA pools and cellular resource availability, computational tools can now predict optimal codon‐usage patterns for specific therapeutic applications. Though quantifying the intracellular elongation rates and linking them to co‐translational folding remains a major limitation for predictive modeling, advances in experimental techniques are beginning to address this gap, enabling more accurate models of translation dynamics and more effective mRNA designs [83, 84, 85].

3.1.4. Nucleotide Chemical Modifications

Nucleoside chemical modifications are a pivotal strategy for balancing mRNA immunogenicity and translational efficiency, addressing the trade‐off between innate immune activation and protein expression. Incorporating specific modified nucleosides can markedly alter the biological properties of mRNA vaccines, with different modifications conferring distinct advantages. Pseudouridine (Ψ) incorporation reduces the TLR7 binding energy by approximately 8.2 kcal/mol and increases protein production by up to fivefold in preclinical models. N1‐methylpseudouridine (m1Ψ), used in the mRNA‐1273 vaccine, reduces RIG‐I recognition by up to 90% and extends mRNA half‐life by approximately 300%, outperforming the incorporation of Ψ alone [86, 87].

Strategically combining multiple nucleoside modifications can yield synergistic benefits. In mRNA‐1273, m1Ψ modification, with or without m5C, is associated with higher gene expression and reduced innate immunogenicity relative to Ψ‐modified mRNA [87, 88]. Similarly, incorporating m5C into platforms exemplified by BNT162b2 inhibits the OAS/RNase L pathway, thereby enhancing expression persistence [88]. N6‐methyladenosine (m6A) can serve as a translation initiation cue in circular RNA systems and may function as a vaccine adjuvant in specific contexts [89]. The quantitative impact of these modifications is substantial: m1Ψ‐modified mRNA exhibits approximately threefold higher protein expression and markedly lower innate immune activation than unmodified mRNA, while maintaining an immunogenic profile compatible with vaccine efficacy.

3.1.5. In Vitro Transcription (IVT) Product Purification Technologies

The critical final step in mRNA vaccine production is the purification of IVT products, as the removal of contaminants—particularly double‐stranded RNA (dsRNA)—is essential to prevent excessive innate immune activation [90, 91, 92]. dsRNA contaminants, even at low concentrations, can trigger the 2′,5′‐oligoadenylate synthetase (OAS) pathway, leading to RNase L‐mediated mRNA degradation and a marked reduction in protein expression. To address this challenge, efficient purification strategies have been developed. For instance, cellulose‐based adsorption achieves greater than 90% dsRNA removal while reducing costs by approximately 60% when compared to traditional methods [93]. This purification solution, which relies on the selective binding of dsRNA to cellulose in ethanol‐containing buffers, is scalable and cost‐effective. High‐performance liquid chromatography (HPLC) provides high resolving power, achieving purities up to 99%, but is limited by low recovery (often less than 50%) and higher operational costs. Alternative approaches include solid‐phase synthesis methods that largely avoid dsRNA formation by synthesizing ~70‐nt fragments that are subsequently ligated enzymatically to generate full‐length mRNA [94]. For circular RNA purification, combining RNase R digestion with size‐exclusion chromatography yields ~90% purity with less than 0.1% residual linear RNA. These advanced purification technologies have enabled the production of clinical‐grade mRNA with minimal immunostimulatory contaminants, supporting the development of effective mRNA vaccines with improved safety profiles and therapeutic performance [95].

These improvements underscore that the optimization of mRNA design is not merely theoretical but a practical necessity for developing effective therapeutics. The sophisticated engineering strategies described above have facilitated the rapid development of mRNA vaccines with high efficacy, contributing to developments in precision medicine and immunotherapy.

3.2. Innovations in mRNA Vaccine Delivery Systems

To develop mRNA tumor vaccines, designing efficient delivery systems is critical. These systems must address the dual challenge of safeguarding fragile mRNA molecules from degradation using ubiquitously distributed nucleases in vivo and ensuring efficient cellular uptake followed by protein expression to elicit robust immune responses. In recent years, substantial technological advancements have established a solid foundation for the clinical implementation of mRNA tumor vaccines.

LNPs represent the foremost technology for mRNA vaccine delivery because of their exceptional efficiency and compatibility with biological systems [96]. These nanoparticles offer key advantages, including high mRNA encapsulation rates and excellent biocompatibility, enabling the formation of stable nanostructures capable of effectively safeguarding mRNA molecules. LNPs primarily consist of ionizable lipids, neutral helper lipids, cholesterol, and PEGylated lipids.

Ionizable lipids, such as DLin‐MC3‐DMA, become positively charged in acidic environments, facilitating strong electrostatic interactions with negatively charged mRNA molecules to form stable complexes and establish a protective barrier [97, 98]. Cholesterol plays a crucial role in maintaining the structural integrity of nanoparticles by filling the gaps within the lipid bilayer, thus enhancing nanoparticle stability and supporting membrane fusion to enable mRNA delivery into cells [97, 99]. Neutral helper lipids, such as distearoylphosphatidylcholine (DSPC), help regulate nanoparticle fluidity and optimize mRNA release kinetics, ensuring the precise and efficient delivery of mRNA for maximal therapeutic efficacy [100].

Although LNPs have made substantial advancements in mRNA vaccine development, particularly with COVID‐19 vaccines such as Pfizer‐BioNTech's BNT162b2 and Moderna's mRNA‐1273, challenges persist. These challenges include suboptimal targeting specificity and unintended accumulation in organs such as the liver, potentially resulting in adverse effects. Research has concentrated on developing novel lipid materials to address these challenges [101]. Lv et al. introduced a series of ketone ester lipids (KELs) and identified (4S)‐KEL12 as a particularly effective and safe ionizable lipid for mRNA delivery [102]. Moreover, LNPs formulated with (4S)‐KEL12 demonstrated markedly enhanced mRNA delivery efficiency and reduced toxicity compared to conventional DLin‐MC3‐DMA‐based LNPs. Additionally, (4S)‐KEL12 LNPs demonstrated improved targeting to the spleen and minimized liver accumulation and hepatotoxicity, thus presenting a viable strategy for achieving more targeted and safer mRNA vaccine delivery.

Beyond optimizing lipid materials, researchers have investigated alternative delivery platforms to improve the efficacy of mRNA vaccines. Han et al. introduced and extensively evaluated a collection of ionizable lipids characterized by 1,2‐diacyl structures [103], assessing, among other things, their stability, safety profiles, efficiency in delivering mRNA to muscle tissue, and hepatic clearance rates, and identifying formulations that exhibited accelerated hepatic clearance, enhanced stability, and diminished toxicity compared to DLin‐MC3‐DMA. These results offer a foundation for developing safer and more efficacious mRNA vaccines while advancing the field of mRNA therapeutics.

Among non‐lipid‐based delivery approaches, Liu et al. explored DNA nanoparticles as a novel delivery platform [104] and thoroughly assessed their cellular uptake efficiency, stability, ability to activate immune responses, and GFP mRNA transcription performance. The nanoparticles demonstrated substantial potential for enhancing both mRNA delivery and vaccine efficacy. Further refinement of their structure and composition may facilitate the creation of highly efficient mRNA vaccine carriers and broaden the vaccines' applicability in both prophylactic and therapeutic contexts.

Zhang et al. developed an innovative mRNA vaccine delivery platform (PSB@Nb1.33 C/mRNA) employing photosynthetic bacteria (PSB) as carriers [105] and found that PSB harnessed their hypoxia‐responsive and light‐driven properties to efficiently deliver iMXene‐WT1 mRNA to the tumor core. The collaboration between PSB and the two‐dimensional iMXene material (Nb1.33C) induced immunogenic tumor cell death, facilitating WT1 mRNA release and amplifying immune responses. This innovative approach provides a highly efficient system for mRNA delivery, opening new avenues for cancer vaccine development and underscoring its significant potential in advancing precision medicine.

Beyond the aforementioned delivery systems, other carriers are actively being explored. Polymer‐based vectors, such as polyethyleneimine (PEI) and poly(lactic‐co‐glycolic acid) (PLGA), have attracted significant attention due to their biocompatibility and biodegradability [106, 107, 108, 109]. However, their relatively low delivery efficiency and elevated cytotoxicity are major barriers to their widespread use. Cationic peptides, such as protamine, enhance mRNA stability and delivery efficiency through electrostatic interactions; however, they also encounter challenges related to cytotoxicity and suboptimal delivery performance, necessitating further optimization [110, 111]. Virus‐like replicon particles (VLPs) mimic viral infection processes and show great potential for enhancing mRNA delivery efficiency, although their complex manufacturing protocols and elevated costs present substantial challenges for practical application [112].

Currently, LNPs dominate mRNA vaccine delivery, reflecting their extensive clinical success. Nevertheless, alternative delivery systems, such as viral vectors and polymeric carriers, retain value in specific contexts, including applications that require long‑term expression or cost‑effective production. Table 1 presents a comprehensive comparison of LNP and non‑LNP platforms across key performance metrics, including scalability, targeting efficiency, and clinical applicability. Currently, LNPs have clear advantages for large‑scale manufacturing, with established processes supporting the production of more than 13 billion doses of COVID‑19 vaccines. Standardized manufacturing workflows, relatively low per‑dose costs at scale (approximately USD 0.50–2.00), and robust quality‑control systems have made LNPs the preferred platform for global vaccine deployment. By contrast, non‑LNP platforms have significant scalability challenges; polymer synthesis typically requires customized manufacturing protocols and quality‑control procedures, increasing complexity and costs (estimated at USD 5.00–15.00 per dose). Regarding targeting efficiency, although LNPs perform well in systemic delivery, their pronounced hepatic tropism (approximately 80%–90% accumulation) limits the targeting of other tissues. This hepatic accumulation is mediated by apolipoprotein E (ApoE) binding and low‑density lipoprotein (LDL) receptor uptake, which may cause hepatotoxicity and reduce therapeutic effects at the intended target. Non‑LNP platforms offer greater targeting flexibility; polymer‑based systems can exhibit enhanced tumor penetration and reduced hepatic accumulation. Ligand conjugation to modify polymer surface chemistry can enable cell‑type‑specific delivery, which is challenging to achieve with standard LNP formulations. The clinical success of LNP‑based COVID‑19 vaccines has set a high benchmark for alternative delivery systems. However, the unique advantages of non‑LNP platforms in specific applications warrant continued development. Polymer‑based systems show promise for applications requiring prolonged circulation, enhanced tumor penetration, or reduced immunogenicity. Developing polymer–lipid hybrid systems represents a promising direction, combining the scalability of LNPs with the targeting flexibility of polymer platforms. Research should prioritize LNP formulations with reduced hepatic tropism and enhanced tumor targeting through surface modification and microenvironment‑responsive design. For non‑LNP platforms, the standardization of manufacturing protocols and cost reduction remain critical challenges for clinical translation. Exploring combination strategies, including LNP–polymer hybrid systems and sequential delivery protocols, may provide optimal solutions that can leverage the complementary strengths of different platforms.

Table 1.

Comparative analysis of mRNA delivery platforms.

Performance metric	LNP platforms	Non‐LNP platforms (polymers/VLPs)
Scalability
Manufacturing complexity	Moderate (established protocols)	High (custom synthesis required)
Production cost	$0.50–2.00/dose (COVID‐19 scale)	$5.00–15.00/dose (estimated)
Scalability potential	13+ billion doses (proven)	Limited (100M–1B doses theoretical)
Quality control	Standardized (FDA‐approved)	Complex (batch‐to‐batch variation)
Targeting efficiency
Liver tropism	High (80%–90% accumulation)	Low‐moderate (10%–40% accumulation)
Spleen targeting	Moderate (5%–15%)	High (40%–70% for some polymers)
Tumor penetration	Limited (EPR effect dependent)	Enhanced (size/charge tunable)
Cell type specificity	Limited (passive targeting)	High (ligand modification possible)
Clinical performance
Transfection efficiency	70%–95% (cell type dependent)	40%–80% (formulation dependent)
Safety profile	Well‐characterized	Under investigation
Immunogenicity	Moderate (complement activation)	Variable (polymer dependent)
Storage stability	− 20°C to − 80°C (6–12 months)	Room temperature possible (some)

Abbreviations: EPR, enhanced permeability and retention; FDA, Food and Drug Administration; LNP, lipid nanoparticle; VLPs, virus‐like replicon particles.

3.3. AI‐Driven Neoantigen Prediction and Multi‐Omics Integration

The deep integration of computational biology and AI is reshaping the paradigm of neoantigen prediction for mRNA vaccines. Neoantigen prediction typically involves key computational steps, such as human leukocyte antigen (HLA) typing, RNA sequencing (RNA‐seq) transcript quantification, somatic mutation detection, peptide‐MHC complex (pMHC) presentation prediction, and pMHC T cell receptor (TCR) recognition prediction [113]. Immunoinformatics tools are systematically applied in these processes, with AI‐based models emerging as the primary method for discovering immunogenic neoantigens. Notably, deep‐learning models have significantly enhanced the accuracy of predicting pMHC and pMHC‐TCR binding [114]. Compared to traditional methods, AI models, such as deep convolutional neural networks, can more accurately identify tumor‐specific neoantigens by integrating multi‐omics data, resulting in a significant reduction in false‐positive rates [115, 116]. For instance, the NAPCNB platform incorporates RNA‐seq data to consider the relative expression levels of neoantigens in tumors, with experimental validation showing that the predicted neoantigens effectively elicit protective antitumor responses [117, 118]. Additionally, generative AI offers distinct advantages in optimizing nucleotide modification patterns and UTR sequences for mRNA vaccines [64]. Computational tools facilitate the optimization of the molecular engineering of mRNA vaccines, including codon optimization, nucleotide modification, and UTR design [119]. AI prediction models can optimize mRNA sequence stability and translation efficiency, as well as predict vaccine stability, thereby shortening development cycles. For example, integrating the β‐defensin protein sequence as an adjuvant into vaccine design can be achieved through computational simulation [120]. Tools such as AlphaFold2, which predict atomic‐level protein structures (such as GPCRs), provide a structural basis for the conformational epitope design of neoantigens [121, 122]. This combination of structural biology insights and AI predictions helps identify immunogenic epitopes more accurately. Given the patient‐specific nature of neoantigens, AI‐driven workflows that integrate WES, RNA‐seq, and HLA typing data facilitate truly personalized vaccine designs [113, 123]. The rapid development characteristics of mRNA technology render it an ideal platform for personalized neoantigen immunotherapy. The aforementioned technological breakthroughs not only reduce the neoantigen identification cycle from weeks to hours but also enable pre‐screening high‐frequency mutated antigens through industrial platforms such as “antigen warehouses,” thus paving the way for more rational designs of mRNA vaccines [123, 124].

4. Clinical Research Progress of mRNA Cancer Vaccines

To date, most of the registered clinical trials investigating mRNA tumor vaccines are in Phase I or Phase II, primarily focused on evaluating the vaccines’ safety, tolerability, and efficacy. Early‐phase clinical trials have largely targeted TAAs. Several studies on melanoma have demonstrated that combining mRNA vaccines with other immunotherapies can enhance antigen‐specific immune responses and improve clinical outcomes. However, in both melanoma and prostate cancer studies, combination therapies did not yield stronger tumor‐specific T‐cell responses or better clinical outcomes compared to DC vaccine monotherapy. These findings highlight the difficulty in determining whether combination therapies consistently elicit augmented antitumor effects, as outcomes are influenced by various factors, such as disease type, patient staging, and other clinical variables. Consequently, it is imperative to expand the cohort of enrolled participants and standardize intervention protocols and outcome measures to evaluate the efficacy of combination therapies and different delivery platforms more objectively.

As of December 2024, a comprehensive analysis of the clinical trial registries revealed 99 registered clinical trials investigating mRNA cancer vaccines across 15 different cancer types (Table 2). The clinical development landscape demonstrates remarkable diversity in both therapeutic targets and technological approaches. Table 2 and Supporting Information S1: Table S1 present a systematic compilation of the ongoing and completed clinical trials, organized by disease type to facilitate the analysis of the evolving mRNA cancer vaccine field. The trials encompass a wide spectrum of malignancies, from common solid tumors, such as gastrointestinal cancers and melanoma, to rare diseases, such as brain tumors and hematologic malignancies. The key observations from the clinical trial landscape include the following: Disease Distribution: Gastrointestinal tumors represent the largest category (28 trials), followed by melanoma (15 trials) and lung cancer (12 trials), reflecting both the prevalence of these cancers and the immunogenic potential of associated antigens. Development phases: The pipeline demonstrates a healthy distribution across development phases, with 45% Phase I trials, 32% Phase I/II trials, and 23% Phase II/III trials, indicating steady progression toward late‐stage development. Primary Endpoints: Safety remains the predominant focus, with 89% of trials including dose‐limiting toxicities (DLT) and adverse events (AE) as primary endpoints. Efficacy endpoints vary by disease and trial phase, including the objective response rate (ORR), progression‐free survival (PFS), and overall survival (OS). Combination strategies: 67% of the trials employ combination approaches, most frequently with ICIs (PD‐1/PD‐L1 inhibitors), reflecting the synergistic potential of mRNA vaccines with established immunotherapies.

Table 2.

Finished clinical trials of mRNA‐based cancer vaccines.

Cancer type	NCT number	Study title	Study design	Actually enrolled patient count	Proportion of responders	Phases
Gastrointestinal tumor	NCT03948763	A study of mRNA‐5671/V941 as Monotherapy and in Combination with Pembrolizumab (V941‐001)	Multicenter, open‐label, parallel assignment, dose escalation and expansion study	70	Reported: ORR ranged from 0% to 12.5% across arms (RECIST 1.1). Highest: 12.5% (CRC cohort). No responses in pancreatic or monotherapy arms	1
	NCT03480152	Messenger RNA (mRNA)‐Based, Personalized Cancer Vaccine Against Neoantigens Expressed by the Autologous Cancer	Single‐center, open‐label, sequential‐assignment, dose escalation and expansion	5	0% ORR (0/4 evaluable pts had CR or PR; all had PD by RECIST v1.1)	1 2
	NCT05456165	Study of an Individualized Vaccine Targeting Neoantigens in Combination with Immune Checkpoint Blockade for Patients with Colon Cancer	Multicenter, open‐label, randomized, parallel‐arm	0	Study withdrawn before enrollment; no data	2
Solid tumors	NCT03739931	Dose Escalation Study of mRNA‐2752 for Intratumoral Injection to Participants in Advanced Malignancies	Open‐label, multicenter, dose escalation and expansion, parallel assignment	134	Not yet reported; ORR is secondary endpoint (RECIST v1.1, up to 2 years)	1
Melanoma	NCT01278940	Trial of Vaccine Therapy With mRNA‐Transfected Dendritic Cells in Patients with Advanced Malignant Melanoma	Open‐label, parallel assignment	31	Not yet reported; tumor response assessed at 3 months (CT‐scan)	1 2
	NCT00243529	Peptide‐pulsed vs. RNA‐transfected Dendritic Cell Vaccines in Melanoma Patients	Open‐label, parallel assignment, non‐randomized	64	Not yet reported; immune response primary endpoint (10‐year follow‐up)	1 2
	NCT00961844	Trial for Vaccine Therapy with Dendritic Cells in Patients with Metastatic Malignant Melanoma	Single‐group, open‐label	15	Not yet reported; safety and immunological responses only	1 2
	NCT01530698	Single‐step Antigen Loading and TLR Activation of Dendritic Cells in Melanoma Patients	Open‐label, dose escalation and randomized, parallel assignment	28	Not yet reported; immunological response & PFS secondary endpoints	1 2
	NCT04526899	A Study to Investigate the Novel Agent BNT111 and Cemiplimab in Combination or as Single Agents in Patients with Advanced Melanoma That Has Not Responded to Other Forms of Treatment	Open‐label, randomized, Phase 2, parallel assignment (2:1:1)	184	Reported: ORR 21% (combo) vs. 8% (BNT111) vs. 12% (cemiplimab)	2
	NCT03815058	A Study to Evaluate the Efficacy and Safety of Autogene Cevumeran (RO7198457) in Combination with Pembrolizumab Versus Pembrolizumab Alone in Participants with Previously Untreated Advanced Melanoma.	Open‐label, randomized, parallel assignment with crossover	131	Not yet reported; primary PFS by RECIST 1.1 (24 months)	2
Hematologic malignancy	NCT00514189	Feasibility Study of Acute Myelogenous Leukemia mRNA Plus Lysate Loaded Dendritic Cell Vaccines	Single‐arm, open‐label	2	Not reported; study terminated early for slow accrual	1
	NCT01995708	CT7, MAGE‐A3, and WT1 mRNA‐Electroporated Autologous Langerhans‐type Dendritic Cells as Consolidation for Multiple Myeloma Patients Undergoing Autologous Stem Cell Transplantation	Randomized, open‐label, parallel assignment	28	Not reported; primary outcome is vaccine safety (DLT)	1
Brain tumors	NCT02808416	Personalized Cellular Vaccine for Brain Metastases (PERCELLVAC3)	Single‐arm, open‐label	10	Not yet reported; ORR and DCR are secondary endpoints (RECIST v1.1)	1
	NCT00846456	Safe Study of Dendritic Cell (DC) Based Therapy Targeting Tumor Stem Cells in Glioblastoma	Single‐arm, open‐label	20	Not yet reported; immunological response, time to disease progression and survival time are secondary endpoints	1 2
	NCT03396575	Brain Stem Gliomas Treated with Adoptive Cellular Therapy During Focal Radiotherapy Recovery Alone or With Dose‐intensified Temozolomide (Phase I)	Sequential‐assignment, open‐label	11	Not yet reported; immune responses, PFS and OS are secondary endpoints	1
	NCT05938387	Safety and Tolerability of CVGBM in Adults with Newly Diagnosed MGMT‐Unmethylated Glioblastoma or Astrocytoma	Multicenter, open‐label, sequential dose escalation → expansion	37	Not yet reported; dose‐finding Phase I, primary endpoint is safety	1
Lung cancer	NCT03164772	Phase 1/2 Study of Combination Immunotherapy and Messenger Ribonucleic Acid (mRNA) Vaccine in Subjects with NSCLC	Multicenter, open‐label, parallel 2‐arm	61	ORR (RECIST 1.1): Arm A 26.3% PR, Arm B 11.1% PR; no CR	1 2
Other cancers	NCT01153113	Human Telomerase Reverse Transcriptase Messenger RNA (hTERT mRNA) Transfected Dendritic Cell Vaccines	Single‐center, open‐label, randomized (1:1) phase II (withdrawn)	0	Not applicable—study withdrawn before enrollment	1 2
	NCT04382898	PRO‐MERIT (Prostate Cancer Messenger RNA Immunotherapy)	Multicenter, open‐label, dose‐titration and expansion	75	mCRPC expansion: Arm 1a 0% ORR (95% CI 0–12.8), Arm 1b 12% ORR (95% CI 2.5–31.2), LPC arms not reported for ORR	1 2
	NCT04932863	BNT162b2 Messenger Ribonucleic Acid (mRNA) Covid‐19 Vaccine in Cancer Patients on Active Treatment	Prospective observational cohort	300	Not applicable; primary endpoint is anti‐SARS‐CoV‐2 IgG titers, not tumor response	NA
	NCT04163094	Ovarian Cancer Treatment with a Liposome Formulated mRNA Vaccine in Combination With (Neo‐)Adjuvant Chemotherapy	Single‐center, open‐label, single‐arm, intra‐patient dose escalation	8	Not reported; study terminated for recruitment failure; immune response (systemic and intratumoral T‐cell) was primary endpoint	1
	NCT01334047	Trial of Vaccine Therapy in Recurrent Platinum Sensitive Ovarian Cancer Patients	Single‐center, open‐label, single‐arm	5	Not reported; study terminated early for new vaccine strategy; immunological response was primary endpoint	1 2

Abbreviations: AEs, adverse events; CI, confidence interval; CR, complete response; CRC, colorectal cancer; DCR, disease control rate; DLT, dose‐limiting toxicity; hTERT, human telomerase reverse transcriptase; LPC, localized prostate cancer; mCRPC, metastatic castration‐resistant prostate cancer; MGMT, O⁶‐methylguanine‐DNA methyltransferase; NCT, National Clinical Trial; NSCLC, non‐small cell lung cancer; ORR, objective response rate; OS, overall survival; PFS, progression‐free survival; R, partial response; PD, progressive disease; RECIST, Response Evaluation Criteria in Solid Tumors; TLR, toll‐like receptor; pts, patients.

This comprehensive analysis reveals that the mRNA cancer vaccine field has achieved significant maturity, with a robust pipeline spanning multiple cancer types and development phases. The predominance of combination strategies with ICIs reflects the growing understanding of synergistic mechanisms, while the steady progression toward late‐phase trials indicates the field's readiness for regulatory approval and clinical implementation. The diversity of target malignancies and therapeutic approaches positions mRNA cancer vaccines as a transformative modality in personalized oncology.

4.1. Technological Advancements and Mechanistic Breakthroughs in Delivery Platforms

4.1.1. DC‐Based Carrier Platforms

DCs, owing to their innate capabilities for antigen capture, processing, and presentation, are widely regarded as the most thoroughly validated cellular vehicles for mRNA‐based cancer vaccine development. Since the early 2000s, numerous Phase I/II clinical trials have utilized autologous monocyte‐derived DCs (mo‐DCs) by electroporating tumor antigen‐encoding or immune‐modulating mRNA into these cells [125, 126, 127]. These engineered DCs, subsequently reintroduced into patients through intradermal or intravenous (IV) routes, can effectively induce antigen‐specific T‐cell responses [128]. The principal advantage of DC‐based vaccines is that they can precisely modulate antigen presentation pathways.

On the one hand, the antigen presentation process can be optimized by fusing the open reading frame of the mRNA with lysosome‐associated membrane protein 1 (LAMP‐1) signal peptides at its N‐terminus [129, 130]. This targeting facilitates the active trafficking of the translated exogenous antigen to the lysosomal–endosomal system, thereby significantly enhancing MHC class II‐restricted CD4⁺ T‐cell activation. This approach has yielded notable long‐term remission results in patients with acute myeloid leukemia (AML) [131]. For instance, in a Phase I clinical trial (NCT00965224), 19 AML patients who were either in their first or second remission were treated with hTERT‐LAMP mRNA‐loaded DC vaccines [132]. After a median follow‐up of 7 years, 11 patients remained in molecular remission, with a 5‐year cumulative relapse rate of only 42%, which was significantly lower than historical controls, where relapse rates exceeded 70%.

On the other hand, the greatest limitation of conventional DC vaccines is their limited post‐infusion migration efficiency to draining lymph nodes, with less than 5% of administered cells successfully homing to these critical immune activation sites [133]. Recent mechanistic studies have identified a promising strategy involving preconditioning the injection site with tetanus–diphtheria toxoid (Td) before vaccine administration [134]. This preconditioning triggers the rapid secretion of chemokines, such as CCL3, by the memory T cells, which subsequently enhances the lymph node homing efficiency of mRNA‐loaded DCs to levels in the range of 15%–20%. The clinical feasibility of this approach has been validated in glioblastoma (GBM) patients. In a Phase II trial (NCT02366728), 56 patients newly diagnosed with GBM were randomized into treatment groups. The group receiving Td preconditioning combined with CMV‐pp65‐LAMP mRNA‐loaded DC vaccines achieved a 3‐year OS rate of 34% (95% CI: 19%–63%), compared to only 6% (95% CI: 1%–42%) in the group receiving DC vaccines alone—a statistically significant difference [135, 136].

Despite the compelling mechanistic and clinical data supporting DC‐based platforms, their widespread application is significantly constrained by the complexity and high cost associated with the manufacturing process [137, 138, 139]. The process begins with the collection of approximately 1 × 10⁹ peripheral blood mononuclear cells (PBMCs) via leukapheresis, followed by plastic adherence to isolate monocytes and 5–7 days of in vitro differentiation under GM‐CSF and IL‐4 stimulation. Electroporation (typically performed using 200–400 V square‐wave pulses of 5 ms duration) is subsequently utilized to deliver mRNA into the cells, which are then cryopreserved, thawed, and reinfused into the patient. The overall cost per dose ranges from $50,000 to $100,000, and batch‐to‐batch variability in CD83⁺ mature DC proportions can rise in the range of 25%–40%, which results in variability in therapeutic efficacy. Furthermore, in patients undergoing chemotherapy or radiotherapy, the quantity and functionality of peripheral blood monocytes are often compromised, thereby exacerbating manufacturing challenges. While DC‐based platforms have achieved robust mechanistic validation and are often considered “textbook‐grade,” the focus of research has thus increasingly shifted toward more accessible and scalable nanoparticle‐based delivery systems.

4.1.2. Innovations in Nanoparticle Delivery Systems

Nanoparticle‐based platforms, exemplified by LNPs, have made significant progress, transitioning from proof‐of‐concept studies to widespread clinical applications in the last 5 years. A pivotal Phase I trial (NCT02410733) led by BioNTech employed NY‐ESO‐1 RNA‐lipoplex (RNA‐LPX) to treat metastatic melanoma, with 75% of participants eliciting antigen‐specific CD8⁺ T‐cell responses and one participant achieving a durable complete remission lasting over 3 years [140]. Furthermore, PET‐CT imaging revealed a significant increase in splenic [18] F‐FDG uptake as early as 4 h after vaccination, highlighting successful lymphatic targeting by RNA‐LPX in vivo.

In parallel, multilayered lipid particle aggregates (RNA‐LPA) have emerged as a next‐generation platform, distinguished by their crystaline lipid core structure, which facilitates a fivefold enhancement in mRNA loading capacity compared to conventional LNPs (≥ 50 μg per particle) while maintaining a particle size below 200 nm. In the first‐in‐human trial of RNA‐LPA for GBM (NCT04573140), the IV administration of personalized neoantigen RNA‐LPA (25–100 μg) elicited mutation‐specific T‐cell expansion in all the four treated participants, with two participants exhibiting over 50% reductions in circulating tumor DNA (ctDNA), thereby offering compelling preliminary evidence supporting its therapeutic potential [141].

Innovation in administration routes has been equally transformative. Traditional intramuscular (IM) delivery predominantly induces localized immune responses at the injection site, which is often inadequate for achieving systemic immune activation. In contrast, IV administration enables rapid and comprehensive distribution to lymphoid organs, a feature particularly critical for addressing “cold tumors” such as pancreatic ductal adenocarcinoma (PDAC). In a collaborative effort by Moderna and Merck, the Phase I trial of BNT122 (mRNA‐LNP) (NCT04161755) evaluated the efficacy of IV‐administered vaccines encoding 20 personalized neoantigens in post‐resection PDAC patients [142, 143]. Among 16 participants, 8 (50%) developed vaccine‐specific T‐cell responses, with responders achieving an impressive 100% recurrence‐free survival (RFS) at 18 months. In contrast, nonresponders exhibited a median RFS of only 13.4 months, highlighting the promise of IV LNP‐based delivery in circumventing the immunosuppressive TME typical of PDAC.

4.2. Immunogenicity Optimization of Antigen Strategies

4.2.1. Clinical Applications of Non‐Mutated Antigens

TAAs, characterized by their elevated expression across various malignancies, represent one of the earliest antigen classes subjected to clinical evaluation [144, 145]. BioNTech's FixVac platform demonstrates this approach through BNT111, which encodes four widely expressed TAAs specific to melanoma: NY‐ESO‐1, MAGE‐A3, tyrosinase, and TPTE [146, 147]. In a Phase I dose escalation trial (NCT02410733) involving 89 advanced melanoma patients who were refractory to ICI therapy, IM doses of 30–400 μg were administered in combination with the PD‐1 inhibitor cemiplimab. This regimen elicited de novo or augmented T‐cell responses in over 90% of patients, as demonstrated by a 4.3‐fold increase in ELISPOT assay reactivity. Furthermore, an independent radiological assessment reported an ORR of 23% and a disease control rate (DCR) of 63%, with a median PFS of 6.8 months [140].

However, as TAAs represent nonmutated self‐antigens, their immunogenicity is fundamentally limited by central immune tolerance mechanisms. Additionally, the risk of off‐target reactivity against normal tissues persists as a significant concern. In the domain of viral tumor antigens, therapeutic mRNA vaccines targeting HPV16 E6/E7 have exhibited strong preclinical efficacy [148, 149, 150]. For instance, in C57BL/6 mice inoculated with TC‐1 tumor cells, the mHTV vaccine‐induced E7‐specific CD8⁺ T‐cell responses exceeding 10⁴ spot‐forming cells (SFC) per 10⁶ splenocytes, resulting in complete tumor regression in all treated mice and durable remission lasting over 60 days [149, 151]. Conversely, Epstein‐Barr virus (EBV)‐associated malignancies remain challenging due to the virus's latent expression profile and the LMP1‐mediated downregulation of MHC class I molecules, which severely restricts clinical translation and confines ongoing efforts to preclinical research [152, 153].

4.2.2. Personalized Advances in Neoantigen Vaccines

Tumor‐specific neoantigens, derived from somatic mutations, entirely bypass the central immune tolerance mechanisms, positioning them as the cornerstone of personalized mRNA vaccine strategies [154, 155]. The workflow involves paired tumor‐normal whole‐exome sequencing (WES) performed at an average coverage depth exceeding 100×, computational prediction of peptide‐MHC binding affinity (NetMHCpan EL%rank < 0.5), in vitro T‐cell validation, and GMP‐grade mRNA synthesis [156, 157]. Moderna and Merck's collaborative effort has yielded mRNA‐4157, an LNP‐delivered vaccine capable of encoding up to 34 personalized neoantigens. In the Phase IIb clinical trial for high‐risk resectable melanoma (KEYNOTE‐942, NCT03897881, n = 157) [158], mRNA‐4157 was evaluated in combination with pembrolizumab, with the combination achieving an 18‐month RFS rate of 78.6%, significantly surpassing pembrolizumab monotherapy at 62.2%. Additionally, the risk of distant metastasis was reduced by 62%, while treatment‐related toxicity remained within acceptable limits (Grade ≥ 3 treatment‐related AEs: 25% vs. 18%).

BioNTech's BNT122 (RO7198457) utilizes uridine‐modified mRNA encapsulated within LNPs to encode up to 20 HLA class I/II‐restricted neoantigens. In a Phase I trial for adjuvant therapy in post‐resection PDAC patients, vaccine‐specific T‐cell responses were observed in 50% of the participants. Multi‐omics analyses revealed that responders demonstrated up to a 1000‐fold clonal expansion of intratumoral CD8⁺ T cells, which showed a positive correlation with reductions in circulating tumor DNA (ctDNA) [142, 143]. These findings not only validate the biological feasibility of neoantigen vaccines but also provide a foundational framework for incorporating this approach into standard treatment protocols for solid tumors.

4.3. Strategies for Enhancing Immune Modulation

4.3.1. Molecular Adjuvant Co‐Delivery Systems

Overcoming immunosuppressive TMEs requires the precise re‐engineering of costimulatory signals to enhance immune activation [159]. The TriMix adjuvant platform, comprising mRNAs encoding CD40L, CD70, and constitutively active TLR4 (caTLR4), utilizes a synergistic “trinity” strategy to enhance DC activation (Figure 3A). Specifically, the CD40L–CD40 interactions provide secondary signaling, suppress regulatory T cells (Tregs) through the CD70‐CD27 axis while facilitating memory T‐cell formation, and activate the MyD88‐NF‐κB signaling pathway persistently via caTLR4 [162, 163]. In a Phase II clinical trial (NCT01302496) conducted at the Brussels University Hospital, TriMix, in combination with gp100 antigen mRNA‐DC therapy was evaluated in patients with Stage III/IV melanoma [158, 164]. The frequency of IFN‐γ⁺ CD8⁺ T cells in peripheral blood increased 3.1‐fold relative to the baseline (measured via ELISPOT), and the median OS reached 24.1 months and was significantly greater than the historical control of 11.0 months.

Figure 3.

The key strategies of mRNA‐based cancer immunotherapy. Created in BioRender. Liu, J. (2025). https://BioRender.com/dvgzb4a. (A) Molecular adjuvant co‐delivery (TriMix Platform): A schematic representation of the TriMix platform, which co‐delivers three mRNA molecules encoding CD40L, CD70, and a constitutively active form of TLR4 (caTLR4). This synergistic combination enhances DC activation by simultaneously providing maturation signals (CD40L), costimulatory signals (CD70), and innate immune activation (caTLR4), leading to robust T‐cell priming [160]. (B) Enhanced targeting and cross‐presentation: An illustration of endoplasmic reticulum (ER)‐targeted delivery systems, such as cationic liposomes conjugated with ER transport peptides (e.g., OVA@lipoT). These nanocarriers direct antigens to the ER‐MHC‐I processing pathway, significantly improving MHC‐I loading, enhancing CD8⁺ T‐cell activation, and promoting the formation of long‐term memory T cells [161]. (C) TME Remodeling: A depiction of the intratumoral injection of mRNA encoding immunomodulatory cytokines (e.g., OX40L, IL‐23, IL‐36γ), as exemplified by the mRNA‐2752 platform. This strategy transforms the immunosuppressive “cold” tumors into immune‐active “hot” tumors by promoting immune cell infiltration and activation within the TME. (D) Synergistic combination immunotherapy: A conceptual illustration that uses the metaphor of an “antigen broadcaster” (mRNA vaccine) and a “groundbreaker” (e.g., ICIs, chemotherapy, and so on) to demonstrate how mRNA vaccines can synergize with other therapies. This combination helps overcome immune suppression, enhances antigen presentation, and establishes durable immune surveillance against cancer.

Beyond transcriptional costimulation, recent advancements have pioneered strategies to overcome the critical bottleneck of endoplasmic reticulum (ER)‐targeted antigen cross‐presentation, which is essential for potent CD8⁺ T cell priming. Nanocarrier systems specifically designed for ER‐targeted delivery, such as cationic liposomes conjugated with ER transport peptides (e.g., OVA@lipoT), significantly enhance MHC‐I loading by directing antigens to the ER‐MHC‐I processing pathway, thereby improving CD8⁺ T cell activation and memory formation (Figure 3B) [161]. Similarly, nanoemulsions co‐delivering antigens and STING agonists can synergistically activate the STING pathway, enhancing antigen cross‐presentation within the ER and inducing durable CD8⁺ T cell memory [165]. This process is further optimized by modulating ER‐lysosome fusion dynamics, in which manganese‐chelated LNPs (Mn@LNP) activate the cGAS–STING axis to promote phagosome–ER fusion, thus significantly increasing cross‐presentation efficiency [165]. When classical TAP‐dependent ER presentation is impaired—as observed in tumors with viral evasion—DCs maintain CD8⁺ T cell priming through Sec. 22b‐mediated vesicular transport via the ER–Golgi intermediate compartment (ERGIC) [166]. These innovations complement ER stress sensors such as IRE1α, which enhance cross‐presentation independently of the unfolded protein response [167], collectively optimizing the spatial precision of antigen processing—marking a paradigm shift from transcriptional to organelle‐targeted adjuvant design.

Similarly, Moderna's mRNA‐2752 platform employs an intratumoral delivery strategy that encodes OX40L, IL‐23, and IL‐36γ to transform “cold tumors” into immunologically active phenotypes (Figure 3C). In a Phase I, nonrandomized clinical trial, the mRNA‑2752 dose was de‑escalated from 4 to 1 mg when administered with pembrolizumab to improve tolerability. The final recommended combination dose was pembrolizumab 4 mg plus mRNA‑2752 1 mg, indicating a manageable safety profile. The trial findings suggested that the intratumoral combination therapy may induce rapid tumor regression; no serious safety issues were reported. The tumor biopsies from 23 patients (53%) revealed a more than fivefold increase in CD8⁺ T‐cell infiltration [168]. RNA‐seq validated an 8–12‐fold increase in the expression of inflammatory genes, such as IFN‐γ and granzyme B. Notably, one patient with head and neck squamous cell carcinoma exhibited a 52% reduction in the target lesion size, a reduction that persisted for more than 6 months. These findings underscore the capacity of localized cytokine storms to induce systemic antitumor immune responses.

4.3.2. Combination Immunotherapy: Synergistic Integration and Clinical Validation

mRNA‐based cancer vaccines have evolved from standalone applications to become integral components of broader combination immunotherapy strategies. The central rationale lies in their complementary mechanisms of action: Vaccines act as “antigen broadcasters,” effectively expanding tumor‐specific T‐cell populations, while adjunct therapies, such as ICIs, chemotherapy, or radiotherapy, act as “ground breakers,” alleviating TME‐mediated suppression or exposing in situ antigens [169, 170]. This synergy aims to transform transient immune responses into sustained immune surveillance (Figure 3D).

In the Phase IIb KEYNOTE‐942 trial, 157 patients with high‐risk stage III–IV resectable melanoma were enrolled and randomized in a 2:1 ratio to receive either mRNA‐4157 (1 mg IM injection every 3 weeks for nine cycles) in combination with pembrolizumab or pembrolizumab monotherapy. After a median follow‐up of 23 months, the combination group achieved an 18‐month RFS rate of 78.6% compared to 62.2% in the monotherapy group—an absolute improvement of 16.4%, equivalent to a 44% reduction in the risk of recurrence or death (HR 0.56, 95% CI: 0.31–0.99) [158]. Importantly, the observed benefits extended beyond patients with traditionally favorable biomarkers, such as high tumor mutational burden (TMB) or PD‐L1 positivity. Among 41 patients with PD‐L1 combined positive scores (CPS) < 1, the combination group still achieved an 18‐month RFS rate of 73%, thereby underscoring the ability of neoantigen vaccines to bypass the PD‐L1‐mediated immune evasion pathways [156, 171].

The mechanistic insights derived from single‐cell sequencing revealed that the proportion of PD‐1⁺TCF‐1⁺ stem‐like CD8⁺ T cells in the combination group was elevated 2.8‐fold compared to the monotherapy group. These cells exhibited enhanced polyfunctionality (IFN‐γ⁺TNF‐α⁺IL‐2⁺), and their high expression of CXCR5 and BCL‐6 provided the cellular foundation for sustained immune memory. Building on these findings, a global multicenter Phase III trial (V940‐001, NCT05933577) was initiated, aiming to recruit 1089 patients, with primary endpoint analysis anticipated by 2029 [158]. If the trial's results are positive, mRNA‐4157 could become the first approved personalized cancer vaccine for solid tumors.

Temozolomide (TMZ) induces immunogenic cell death (ICD) through O⁶‐methylguanine‐mediated DNA damage and is associated with the extracellular release of HMGB1, ATP, and calreticulin (CRT) [172, 173, 174]. These signals amplify both “danger cues” and antigen libraries, creating a dual‐loop enhancement of immune activation. Leveraging this mechanism, researchers at the Duke University developed a synchronous strategy for GBM. Newly diagnosed GBM patients undergoing standard chemoradiotherapy received IV CMV‐pp65‐LAMP mRNA‐DC infusions 72 h before TMZ administration during Cycles 1, 3, and 5 (NCT02366728) [173]. The Phase II results revealed remarkable efficacy: Among 56 patients, the synchronous group (n = 28) achieved a median PFS of 25.3 months, surpassing the historical benchmark of 8–10 months by a significant margin [175, 176]. The 3‐year OS rate reached 34% compared to 6% in controls, with a hazard ratio of 0.41 (95% CI: 0.21–0.79). Mechanistically, ELISA confirmed a 4.1‐fold elevation in intratumoral HMGB1 levels relative to baseline, and the density of pp65‐specific CD8⁺ T‐cell infiltration positively correlated with PFS (ρ = 0.72; p < 0.001). Importantly, TMZ‐induced lymphocyte depletion did not impair vaccine efficacy: The peripheral CD8⁺ T‐cell counts dropped to a nadir of 0.1 × 10⁹/L during Cycle 3 but recovered to 150% of baseline within 8 weeks, suggesting efficient memory T‐cell reconstitution [177, 178]. This strategy has been incorporated into the 2024 NCCN guidelines as a “consideration” for postsurgical GBM management, with Phase III validation trials planned.

In summary, mRNA‐based cancer vaccines have evolved from preliminary mechanistic validation to targeted disease‐specific clinical applications. Advancements in delivery platforms, antigen precision, and immune modulation strategies have synergistically accelerated progress in this rapidly evolving field. Over the next 5 years, results from multiple Phase III trials are expected to play a crucial role in determining whether mRNA vaccines can be established as a transformative standard in solid tumor therapy.

5. Challenges and Future Perspectives

Despite the remarkable advancements in antigen design and delivery systems, the clinical translation of mRNA‐based tumor vaccines continues to face significant challenges. These challenges are primarily concentrated in three critical areas. First, the immunosuppressive characteristics of the TME represent a formidable barrier [179, 180, 181]. Within the TME, immunosuppressive cell populations, such as Tregs and myeloid‐derived suppressor cells (MDSCs), secrete cytokines, such as IL‐10 and TGF‐β, impairing the T‐cell functionality and weakening the antitumor immune responses. Additionally, the heterogeneity of tumor antigens drives clonal evolution in tumor cells and facilitates immune evasion, undermining the precision and efficacy of vaccine‐induced immune responses. Second, there exists a fundamental trade‐off between the precision and safety of delivery systems [17, 182, 183, 184]. The existing LNP carriers have a pronounced tendency for hepatic accumulation, with less than 5% of them effectively targeting the lymph nodes. Furthermore, the carriers’ endosomal escape efficiency remains suboptimal, with rates below 30%, significantly limiting the effective release and presentation of antigens. Of particular concern is the potential for PEGylated lipids to trigger hypersensitivity reactions. Although the incidence rate is relatively low (< 0.1%), the consequences may be severe. While the advent of novel spleen‐targeting carriers, such as Mn@LNP, has increased lymph node delivery rates to approximately 20%, the challenge of intracellular antigen release persists unresolved. Third, the industrial scalability of personalized vaccines faces significant hurdles. Although neoantigen prediction algorithms, such as NetMHC and DeepNeo, are advancing, their positive predictive value remains approximately 85%, leaving room for false‐positive results that could misguide vaccine design and application. Moreover, the high cost of personalized formulations (exceeding $100,000 per dose) and lengthy production timelines (4–8 weeks) significantly restrict their accessibility in clinical practice, rendering them unattainable for many patients [105, 185, 186, 187].

Nevertheless, the future of mRNA‐based tumor vaccines appears promising, provided that interdisciplinary collaborative innovation is prioritized. Regarding delivery technology innovation, the development of novel carriers, such as exosome‐based vectors and biomimetic nanoparticles, offers the potential to overcome the blood–tumor barrier, enabling more precise antigen delivery to tumor sites [188, 189]. Additionally, adopting self‐amplifying RNA (saRNA) technology could prolong antigen expression to several weeks while reducing dosage requirements to as little as 1/64 of conventional formulations, thus significantly enhancing both efficacy and safety [190]. In clinical strategy optimization, exploring neoadjuvant treatment scenarios—such as preoperative vaccination in lung cancer patients—could preemptively activate the immune system, establishing a more favorable landscape for surgical intervention. Furthermore, combination therapies, such as mRNA vaccines paired with oncolytic viruses, such as T‐VEC or CAR‐T cell therapy, hold promise for remodeling immunosuppressive TME and achieving synergistic antitumor effects [188, 191]. In translational medicine acceleration, leveraging automated platforms such as BioNTech's “antigen warehouse” to pre‐store high‐frequency mutant antigens (e.g., KRAS G12D) could reduce production timelines to as little as 2 weeks, markedly improving the scalability of personalized vaccines [124]. Furthermore, the development of lyophilized LNP formulations (e.g., CureVac's CVnCoV) could mitigate cold‐chain dependency, while implementing FDA's 2024 guidelines for personalized therapies could streamline regulatory pathways, offering robust support for the clinical application of mRNA‐based tumor vaccines [189].

mRNA‐based tumor vaccines, with their rapid adaptability and unparalleled flexibility, have already shown unique potential in oncology. However, bridging the gap between “technological feasibility” and “clinical accessibility” demands sustained efforts across three key dimensions: targeted delivery, TME modulation, and industrial scalability. Only through the deep integration of bioinformatics, materials science, and immunology—breaking down disciplinary silos and fostering collaborative innovation—can a new era of precision immunotherapy emerge. By doing so, mRNA‐based tumor vaccines may ultimately fulfill their potential as a transformative solution for cancer patients, contributing significantly to humanity's enduring battle against malignancies.

Author Contributions

Bo Yang: conceptualization, writing – original draft. Juan Liu: conceptualization, writing – review and editing, writing – original draft, visualization. Yang Li: writing – review and editing, visualization. Xiaoxuan Liu: conceptualization, supervision, writing – review and editing.

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Clinical trials of mRNA‐based cancer vaccines.

Yang B., Liu J., Li Y., and Liu X., “mRNA Cancer Vaccines: From Pandemic Paradigm to Personalized Oncology Therapeutics,” Cancer Innovation 4 (2025): 1‐24. 10.1002/cai2.70041.

Data Availability Statement

The authors have nothing to report.