In Vivo Engineered CAR-T Cell Therapy: Lessons Built from COVID-19 mRNA Vaccines

mRNA vaccines enable the precise encoding of target antigens. The spike protein (S protein) of SARS-CoV-2 is a trimeric structure with a highly dynamic receptor-binding domain (RBD). In the RBD-up conformation, the RBD is exposed and capable of binding to the ACE2 receptor on host cells, playing a critical role in viral entry [34]. The natural state of the S protein prior to host cell fusion, known as the prefusion conformation, is essential for initiating membrane fusion and, thus, represents a key step in viral infection. Consequently, interventions targeting this conformation can effectively block viral entry. Studies have shown that antibodies targeting the prefusion conformation of the S protein, such as RBD-specific antibodies, exhibit stronger neutralizing activity, as they prevent the virus from interacting with the ACE2 receptor. By introducing artificial modifications—such as beneficial proline substitutions and stabilizing mutations—the prefusion conformation can be stabilized, preventing its transition to the postfusion state [35,36]. This stabilization allows vaccines to more effectively mimic natural infection, thereby eliciting robust immune responses.

The optimization of mRNA sequences is a critical step in improving vaccine efficacy. Thoughtful design of mRNA sequences significantly enhances their stability, translational efficiency, and antigen expression. This section focuses on the pivotal strategies of codon optimization, minimizing secondary structure formation, and the optimization of 5′ and 3′ untranslated regions (UTRs) to improve mRNA performance.

Codons are the fundamental coding units of mRNA, and their usage frequency varies across host cells, a phenomenon known as codon bias. By replacing rare codons in foreign genes with codons preferred by the host cell, researchers can enhance ribosomal recognition and increase translation efficiency [37]. Interestingly, to address the rapid evolution of SARS-CoV-2, codon deoptimization (CDO) has been employed in the development of the live-attenuated vaccine candidate CDO-7N-1. This approach involves substituting frequently used codons in the viral genome with rarely used ones in the host cell, significantly reducing viral replication capacity and attenuating virulence while preserving broad immunogenicity against multiple variants [38].

mRNA, as a single-stranded nucleic acid, often forms secondary structures such as hairpins and stem loops through base pairing, which can hinder ribosome binding and movement along the mRNA strand. To mitigate this issue, researchers have employed computational algorithms to optimize mRNA sequences by predicting and minimizing regions prone to forming stable secondary structures. By strategically rearranging nucleotide sequences to create a more open configuration, ribosomal accessibility and translation efficiency are significantly improved, thereby boosting the production of antigenic proteins [39].

The 5′ and 3′ UTRs of mRNA play pivotal roles in regulating mRNA stability, translation efficiency, and half-life [40]. The 5′ UTR is particularly important for enhancing ribosome binding efficiency [41]. Recently, polysome profiling and deep learning were used to analyze a library of 280,000 randomized 5′ UTR sequences, resulting in a predictive model and the design of novel 5′ UTRs that precisely modulate ribosome loading levels to optimize protein expression [42]. On the other hand, the 3′ UTR is integral to controlling mRNA stability and translation efficiency [43]. Furthermore, the poly(A) tail has become a widely utilized and essential component of mRNA technology, although its relationship with translation and mRNA decay still requires further refinement and understanding [44]. Emerging studies have examined various aspects of the poly(A) tail, such as its length and dynamic behavior in specific biological processes, uncovering its diverse roles in natural mRNA translation and regulation [45,46]. The findings above provide valuable insights for designing synthetic mRNA to enhance its stability and functionality.

To reduce the immune response triggered by cellular recognition of mRNA as foreign RNA, modified nucleosides are commonly incorporated into mRNA [47]. These modifications enhance stability and expression levels while minimizing side effects. Pseudouridine (Ψ) is one of the most frequently used modifications, effectively reducing recognition by the innate immune system through Toll-like receptors (TLR7 and TLR8), thereby improving translation efficiency [31]. Furthermore, research has shown that N1-methylpseudouridine (m1Ψ) enhances these effects even further and has been widely employed in the BNT162b2 and mRNA-1273 vaccines [48,49]. In addition, some companies, such as CureVac, have opted for sequence engineering strategies to prevent mRNA immune activation rather than relying on nucleoside modifications [50], indicating that nucleoside modification and sequence optimization represent two complementary approaches for reducing immunogenicity and improving mRNA efficacy.

mRNA is inherently a highly degradable molecule, rendering it extremely unstable in physiological environments and necessitating effective delivery systems for its protection [60]. LNPs achieve this by encapsulating mRNA within their phospholipid bilayer structure and cholesterol components, providing stability and preventing degradation before reaching target cells. The efficient delivery capability of LNPs stems from their ionizable lipid content, which exhibits unique pH-responsive properties. At the neutral pH of the bloodstream (~7.4), ionizable lipids remain uncharged, ensuring LNP stability. Upon entering cells through endocytosis, LNPs are enclosed within endosomes, where the acidic environment (pH ~5.0–6.0) protonates these lipids, resulting in a positive charge. This charge alteration disrupts the negatively charged endosomal membrane, enabling LNPs to escape and release mRNA into the cytoplasm [61]. This pH-sensitive characteristic grants LNPs dual functionality: maintaining stability in systemic circulation and enabling precise cargo release within target cells, enhancing efficacy while minimizing side effects.

Compared to viral vectors, LNPs exhibit relatively low immunogenicity, mitigating the risk of triggering strong immune responses, which makes them safer choices for mRNA delivery [62]. Moreover, LNPs have the unique potential to act as adjuvants, modestly stimulating innate immune responses to increase vaccine efficacy [63,64]. Studies have shown that ionizable lipids in LNPs drive adjuvant activity by inducing the production of cytokines and chemokines such as IL-1β, IL-6, and GM-CSF. These molecules activate innate immune pathways and support robust adaptive immune responses [63,65,66]. Meanwhile, PEG-modified lipids prolong circulation time by reducing immune clearance; however, they may elicit mild immune reactions, particularly in individuals with pre-existing antibodies against PEG, potentially leading to hypersensitivity or reduced delivery efficiency [67,68]. Optimizing the concentration and structure of PEG-modified lipids is, therefore, critical to balancing stability and immunogenicity.

LNPs offer remarkable tunability, allowing precise targeting through modifications, such as specific ligands or component adjustments, and enabling the delivery of nucleic acids to specific tissues or cell types [69]. This adaptability makes LNPs a versatile platform for various therapeutic applications, including mRNA vaccines, gene therapies, and siRNA-based treatments.

In addition to tunability, LNPs exhibit industrial scalability, largely enabled by the ethanol-loading technique. This method temporarily alters lipid configurations in an ethanol environment to encapsulate nucleic acids, with LNPs reassembling into stable nanoparticles upon ethanol removal [70]. The technique is both efficient and reproducible, making it ideal for large-scale manufacturing and supporting the rapid production of LNP-based therapies, including mRNA vaccines and siRNA treatments such as Patisiran.

Building on the structure of pseudouridine, N1-methylpseudouridine (m1Ψ) offers enhanced immune evasion capabilities by reducing recognition by toll-like receptors. In a study on multiple sclerosis, m1Ψ-modified mRNA was used to encode disease-associated self-antigens. By delivering antigens non-inflammatorily and promoting Treg expansion, this m1Ψ-modified mRNA effectively suppressed disease progression in several mouse models of experimental autoimmune encephalomyelitis [82]. m1Ψ has been regarded as a key advancement in mRNA therapeutics, offering improved stability, reduced immunogenicity, and enhanced translational efficiency for diverse applications.

N⁶-Methyladenosine (m6A) is one of the most prevalent internal modifications of mRNA and plays a critical role in regulating mRNA stability, translational efficiency, and nuclear export. Mechanically, this modification interacts with certain proteins, including ‘writers’ such as the METTL3-METTL14 complex, which catalyzes its addition, ‘readers’ such as the YTHDF family, which bind to m6A-modified RNA to mediate downstream effects, and ‘erasers’ such as FTO and ALKBH5, which remove the methyl group, maintaining the dynamic regulation of RNA metabolism [83,84]. m6A, which is naturally abundant in endogenous mRNA, has emerged as a promising target for addressing multiple complex diseases, including cancer [85], and has also been explored for its ability to increase the efficacy of mRNA vaccines and therapeutic mRNA applications. In our previous work, we introduced m6A modifications to the therapeutic mRNA, leading to further acceleration of the anti-tumor response, convincingly demonstrating the potential of m6A-equipped mRNA to optimize therapeutic efficacy.

5-Methylcytosine (m5C) is a well-characterized modification in DNA, essential for gene regulation and stability. Although it is present at low levels in eukaryotic mRNA, recent studies have begun to reveal its biological significance [86]. Enzymes like NSUN2 catalyze m5C modifications, protecting RNA from cleavage under stress [86]. Notably, m5C at specific mitochondrial tRNA sites, such as C34 in tRNAMet, enhances oxidative phosphorylation, a process with critical implications in tumor metabolism, by improving mitochondrial translation efficiency [87]. We might actively leverage these functions in therapeutic mRNA design, where they could enhance stability and translation efficiency, thereby boosting the performance of mRNA-based treatments.

In addition, other modifications, such as 5-methyluridine (m5U) and N1-methyladenosine (m1A), have drawn significant attention. m5U has been shown to participate in ribosomal function by regulating tRNA transport and stability. It also enhances mRNA translation efficiency and improves its stability [88]. Studies suggest that alterations in m5U expression levels under certain pathological conditions may interact with disease processes, making it a potential therapeutic target or an optimization strategy for mRNA modifications [88]. m1A, on the other hand, is notable for its role in regulating RNA secondary structure and protein-RNA interactions. Specifically, the methylation characteristic of m1A alters the base-pairing capacity of adenosine, inducing a bend in the RNA strand at the modification site. This structural change can expose or obscure protein-binding sites on RNA, influencing its interactions with other molecules. In cancer-related research, m1A demethylation has been closely linked to metabolic reprogramming and tumor progression, further highlighting the importance of m1A regulation in disease processes [89].

2′-O-Methylation, despite its name, refers to the modification of the 2′ hydroxyl group of the ribose sugar rather than the nucleotide base itself. It is a common and significant RNA modification found in tRNA, rRNA, and mRNA, and plays a critical role in enhancing RNA stability and reducing immunogenicity. In tRNA, 2′-O-methylguanosine (Gm18) has been shown to effectively suppress the activation of TLR7 and TLR8, thereby attenuating innate immune responses [90]. If the 2′-O-methyl group is absent from Gm18, its immunosuppressive effect is substantially diminished. However, when combined with m5U to form 2′-O-methylthymidine (Tm or m5Um), the RNA demonstrates markedly enhanced immune evasion from TLR recognition [91]. This dual modification not only effectively suppresses immune activation but also improves RNA stability and translational efficiency.

2-Thiouridine (s2U) is a common tRNA modification that enhances tRNA stability and functionality through the sulfur modification of the uridine base. Studies have shown that a low level of s2U modification is associated with translational stalling at lysine (LysAAA) codons. In combination with 5-carboxymethylaminomethyl (mcm5) modification, s2U forms mcm5s2U, which is widely present at wobble positions in tRNAs for lysine, glutamate, and glutamine, and plays a critical role in translational regulation. Under stress conditions, such as artemisinin resistance in the malaria parasite Plasmodium falciparum, reduced levels of mcm5s2U modification alter codon usage preferences, prioritizing the translation of specific proteins, thereby enabling the parasite to adapt to drug pressure through targeted protein expression strategies [92]. Intriguingly, s2U and related modifications hold promise for optimizing translation efficiency and selectivity in therapeutic applications.

In summary, epigenetic modifications of mRNA play a crucial role in regulating various physiological and pathological processes. From tumor metabolism and drug resistance to the precise modulation of tRNA functionality, many fascinating discoveries have emerged from these fields of study. These findings, in turn, offer valuable insights and foundational knowledge for advancing therapeutic mRNA modifications. In other words, this domain remains a largely untapped frontier, presenting immense opportunities for further exploration and innovation. Nevertheless, we should also take into account that recent studies have suggested that certain modifications, such as m1Ψ, may induce +1 ribosomal frameshifting, potentially impacting translational fidelity, although no human-derived reports have indicated adverse effects [93]. Future research could focus on several key areas: optimizing combinations of multiple modifications, such as investigating the synergistic effects of m6A with other modifications like m1Ψ, to achieve an optimal balance between immunogenicity and antitumor activity; developing cancer-specific modification strategies tailored to different tumor types, enhancing therapeutic efficacy while minimizing side effects; and delving deeper into the mechanisms of these modifications, leveraging single-cell technologies and multi-omics analyses to uncover the systemic impacts of mRNA modifications on immune regulation, thereby providing a solid theoretical foundation for next-generation CAR-T therapies.

Passive targeting primarily relies on the interactions between specific microenvironments and the physicochemical properties of nanocarriers. For example, in tumor or inflamed tissues, newly formed blood vessels often have incomplete structures with enlarged endothelial gaps, allowing macromolecular substances like nanoparticles to pass through more easily and accumulate within the tumor microenvironment. Additionally, the lack of effective lymphatic drainage in tumor tissues hinders the clearance of nanoparticles, prolonging their retention time. This phenomenon, known as the enhanced permeability and retention effect, facilitates the accumulation of LNPs in tumor tissues while minimizing their impact on healthy tissues [94]. Beyond size and shape control, the composition of LNP lipids plays a critical role in enabling passive targeting. For example, the Spleen Selective ORgan Targeted strategy incorporates anionic lipids such as 18:1 PA into LNP formulations to mitigate their typical liver accumulation and achieve targeted delivery to extrahepatic tissues [76]. This modification operates via charge-based targeting by lowering the pKa of LNPs, enhancing their compatibility with the mildly acidic microenvironment of the spleen and altering their organ distribution. Similarly, by customizing lipid tail length (shorter tails are preferred), bond types (ester bonds outperform amide bonds), and head group structures (methyl groups show superior performance), LNP delivery efficiency to lymph nodes can be significantly improved [95]. Moreover, a partial substitution of cholesterol with hydroxycholesterol in LNP formulations alters the binding kinetics with Niemann–Pick C1 enzymes, reducing endosomal recycling and enhancing T cell delivery by 1.8 to 2 fold [96]. The critical role of lipid composition in shaping LNP performance has also driven advancements in screening and development methodologies. Li, B. and colleagues utilized combinatorial chemistry and machine learning to establish a high-throughput platform for ionizable lipid screening, enabling the prediction of lipid performance while identifying optimal lipid combinations [97]. These innovative techniques, which transitioned incremental improvements into transformative advancements, significantly accelerate the development of LNPs and expand the boundaries of the field.

Active targeting enhances the specificity of LNPs for target cells or tissues by modifying their surface with specific molecules such as antibodies, ligands, or small molecules. The antibody-conjugated LNP platform developed by Billingsley, M. M., is a notable example, which achieved precise T-cell targeting by incorporating antibodies against markers like CD3, CD5, and CD7 on the LNP surface. This approach significantly reduced off-target distribution to non-target tissues, such as the liver [98]. Similarly, a study on LNP-mediated mRNA delivery for treating hereditary retinal disorders employed M13 phage display technology to identify retinal-targeting peptides for LNP modification. This enabled specific recognition of photoreceptor cells within the highly barriered retinal structure, demonstrating excellent biocompatibility and markedly reduced off-target toxicity [99].

Another strategy involves designing LNPs to respond to external stimuli. LNPs made from sugar alcohol fatty acid monoesters exhibit lower or upper critical solution temperature behavior, allowing them to undergo structural changes at specific temperatures. This property is particularly advantageous in pathological environments with temperature variations, such as tumor sites with localized heating, enabling controlled drug release for more precise delivery [100]. In another study, Fe₃O₄ superparamagnetic nanoparticles were incorporated into LNPs, leveraging external magnetic fields to drive mechanical targeting and accumulation of LNPs in tumor regions. These LNPs were designed as dual-contrast nanoparticles with both optical absorption and magnetic properties. By integrating photoacoustic and magneto-ultrasound imaging technologies, these LNPs demonstrated high sensitivity for tumor imaging, offering significant potential in precision diagnostics and targeted therapies [101].

Enhancing the blood–brain barrier (BBB)penetration is a promising approach for improving the efficacy of certain brain tumor treatments and in vivo CAR-T strategies. Techniques such as focused ultrasound and microbubble-assisted BBB opening have been investigated for transiently increasing BBB permeability, allowing LNP-mediated mRNA delivery into brain tissues [102]. Peptide-functionalized LNPs, including those modified with T7, RVG29, AP2, and mApoE peptides, have further improved brain-specific mRNA transfection and may offer potential applications for future in vivo CAR-T therapies [103,104]. Modern LNP technologies integrate biological engineering, nanomaterial design, and external modulation strategies to enhance the specificity and efficacy of drug delivery while minimizing side effects. In the future, LNP targeting strategies are likely to become increasingly refined and personalized, and tailored to meet the therapeutic demands of various diseases.