Lipid-based mRNA chimeric antigen receptor T cells for enhanced in vitro antitumor activity

Summary

Chimeric antigen receptor (CAR) T cell therapy provides substantial clinical benefits in hematologic malignancies. However, cell engineering using conventional viral vectors is costly, labor intensive, and prone to safety risks, while electroporation-induced cellular damage limits manufacturing scalability. This study reports a safe, high-efficiency, nonviral platform for generating mRNA-engineered CAR T cells. Herein, eight ionizable lipids are synthesized and formulated into lipid nanoparticles. Systematic screening and structural optimization identified lipid 6 as the lead candidate. Transfection efficiency is validated by delivering enhanced green fluorescent protein (eGFP) mRNA into DC2.4, peripheral blood mononuclear, dendritic, and natural killer cells. Murine models confirm robust in vivo expression and a favorable safety profile. Lipid 6-mediated delivery of MUC1 mRNA generates MUC1-CAR T cells characterized by high surface receptor expression and potent cytotoxicity against MUC1-positive tumor cells. These findings establish lipid 6 as a feasible strategy for primary human T cell engineering, facilitating functional CAR expression with minimal toxicity.

Graphical abstract

Highlights

• •A library of eight ionizable lipid scaffolds pinpoints lipid 6 for multi-immune mRNA delivery
• •DOE-optimized LNPs surpass benchmark lipids in vivo and stabilize T cell homeostasis
• •MUC1 mRNA-loaded LNPs generate low-exhaustion CAR T with durable tumor lysis
• •Non-integrating mRNA bypasses electroporation and preserves T signaling

Biological sciences; Cell biology; Immunology

Introduction

Chimeric antigen receptor (CAR) T cell therapy represents a promising curative strategy for cancer owing to its high specificity and potent efficacy.¹ Since the 2017 Food and Drug Administration (FDA) approval of CD19-targeted CAR T cell therapy for relapsed/refractory acute lymphoblastic leukemia (ALL), several drugs have entered clinical practice.^2,3 Standard manufacturing relies on ex vivo T cell isolation, CAR-encoding gene modification, expansion, and autologous reinfusion.⁴ Currently approved therapies and most clinical trials employ viral vectors, which enable stable CAR expression but present substantial drawbacks.^5,6 Beyond insertional mutagenesis and potential immunogenicity, viral systems suffer from constrained cargo capacity, labor-intensive production, and prohibitive costs, all of which limit broader clinical accessibility.^7,8,9,10,11 While cytokine release syndrome and neurotoxicity remain major on-target toxicities of CAR T cell therapy itself, the safety and practical limitations of viral delivery have collectively spurred intense interest in nonviral platforms to improve the safety, scalability, and affordability of engineered T cell therapies.¹¹

mRNA-induced CAR expression is a viable alternative, as its transient, non-integrating function confers a self-limiting safety profile by confining CAR expression to a defined window.¹² Importantly, mRNA-based CAR T cells have demonstrated efficacy against ALL, melanoma, and Hodgkin lymphoma comparable to virally transduced cells.^{13,14,15,16,17,18,19,20,21,22} Nevertheless, CAR protein levels decline within days, which restricts single-dose persistent antitumor activity but potentially mitigates long-term off-target toxicities.^13,21,23 Notably, naked mRNA’s inherent instability and membrane impermeability necessitate a delivery carrier, and electroporation (EP) is currently the clinical standard for T cell transfection.^13,22,24 However, EP requires specialized equipment, compromises membrane integrity, reduces cell viability, alters cellular physiology, yields inconsistent transfection outcomes, and thus demands further safety evaluation of its long-term effects on post-EP gene expression.^25,26,27,28

Nanoparticles (NPs) composed of lipids or polymers provide a compelling alternative to EP, as they enable stable nucleic acid encapsulation, high delivery efficiency, and low cytotoxicity without specialized equipment.^25,29,30 Notably, polymeric NPs have been shown to mediate mRNA delivery with markedly reduced cytotoxicity compared with EP.^31,32 Among these nanoplatforms, ionizable lipid NPs (LNPs) represent a clinically validated strategy, as demonstrated by the FDA-approved therapy Onpattro.^33,34 Mechanistically, LNPs enable efficient endosomal escape of mRNA through protonation in acidic endosomes.^{35,36,37,38,39} In addition, they outperform commercial reagents in immune cells and offer tunable physicochemical properties for optimized delivery.^{11,34,35,39,40,41} These attributes position LNPs as a promising platform for ex vivo human T cell engineering.

Herein, we systematically evaluated ionizable LNPs for mRNA delivery to T cells. We formulated a panel of eight LNPs through microfluidic mixing, each combining a distinct ionizable lipid with cholesterol, phospholipid, and polyethylene glycol (PEG)-lipid at fixed ratios. The LNPs were screened using DC2.4 and primary human T cells, and lipid 6 was identified as the lead candidate, achieving high transfection efficiency with minimal cytotoxicity. Subsequent optimization of the nitrogen-to-phosphate ratio, transfection parameters, and mRNA dose yielded >90% of enhanced green fluorescent protein (eGFP)-positive cells in primary T cells, natural killer (NK) cells, and dendritic cells (DCs), demonstrating broad utility across immune cell types. Finally, encapsulation of CAR mRNA into optimized lipid 6 generated functional CAR T cells characterized by high CAR expression, low cytotoxicity, and potent, specific tumor cell killing in vitro. This study establishes ionizable LNPs as an effective and translatable platform for mRNA-based ex vivo CAR T cell manufacturing.

Results

Optimized characterization of the ionizable LNP library

To develop an mRNA delivery platform for T cells, we focused on ionizable LNPs, owing to their high transfection efficiency and low cytotoxicity in vitro and in vivo. Building on recent reports of nucleic acid delivery to immune cells using LNPs,^21,33,37,42 we designed a library of eight novel ionizable lipids as structural analogs of previously reported candidates effective in immune cell transfection (Figures 1 and S1 and Table S1).^11,33,38,39 Each lipid was formulated into an LNP via microfluidic mixing with a fixed excipient composition: cholesterol (for membrane integrity and stability), the phospholipid 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) (to facilitate endosomal escape), and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000) (to confer colloidal stability and reduce nonspecific uptake).^39,40,43 Excipient molar ratios were selected based on established mRNA-LNP optimization protocols, which often employ DOPE as the helper lipid, a reduced ionizable lipid fraction, and increased cholesterol and PEG-lipid concentrations to fine-tune the properties of NPs.⁴⁴ These ratios were held constant to isolate the effect of the ionizable lipid head group during initial screening. For mRNA delivery, a molar ratio of 35 ionizable lipid:46.5 cholesterol:16 DOPE:2.5 DMG-PEG was used as a standard formulation for comparison.

Figure 1.

Design and synthetic routes of amine core scaffolds and representative ionizable lipid synthesis

(A) Synthesis of amine cores, illustrated using amine core #1 as an example.

(B) Structures of the amine cores employed in the generation of the ionizable lipid library.

(C) Reaction scheme for synthesizing ionizable lipids via an SN2 reaction, where excess alkyl chains react with polyamine cores, with lipid 1 serving as a representative example.

We characterized the resulting LNP library using dynamic light scattering (DLS) and RiboGreen. All formulations exhibited monodisperse size distributions, with Z-average diameters of 79.6–97.2 nm and polydispersity indices (PDIs) of 0.107–0.180 (Table S1). The encapsulated mRNA concentration and encapsulation efficiency (EE), determined by RiboGreen, were consistent across the library (encapsulated mRNA concentration: 76.47–90.64 ng/μL, EE: >95.77). These data confirm the generation of eight distinct, mRNA-encapsulating LNP formulations suitable for downstream evaluation in T cells.

Screening of ionizable LNPs for mRNA delivery

We assessed LNP-mediated mRNA delivery using eGFP for in vitro screening and firefly luciferase (FLuc) for in vivo validation. Both mRNAs contained N1-methylpseudouridine and 5-methylcytidine modifications, which enhance translational efficiency and reduce immunogenicity—a property that may influence delivery outcomes and warrants further investigation.^{45,46,47,48,49}

Initial in vitro transfection efficiency was evaluated in the murine DC line DC2.4 and activated human peripheral blood mononuclear cells (PBMCs). Cells were transfected with eGFP mRNA-loaded LNPs (200 ng per 5 × 10⁵ cells). After 24 h, flow cytometry revealed that four of the eight LNP candidates transfected >90% of DC2.4 cells, with the poorest performer achieving >60% positivity. Lipids 6 and 8 showed the highest eGFP positivity rates and mean fluorescence intensity (MFI), surpassing the performance of the benchmark formulation ALC-0315 (Figure 2A).

Figure 2.

Screening of ionizable lipid nanoparticles in vivo and in vitro

(A and B) Analysis of enhanced green fluorescent protein (eGFP) expression in DC2.4 cells and peripheral blood mononuclear cells (PBMCs) after treatment with the LNP library and Lipofectamine for 24 h at a dose of 200 ng per 5 × 10⁵ cells, and identification of top-performing LNPs. Assessment of eGFP expression in DC2.4 cells and PBMCs using flow cytometry indicated increased eGFP expression, evaluated as positive rate and mean fluorescence intensity (MFI).

(C) Schematic of the administration protocol.

(D and E) In vivo bioluminescence imaging following (D) intravenous (i.v.) and (E) intramuscular (i.m.) administration of LNPs encapsulating firefly luciferase (FLuc) mRNA at a dose of 0.5 mg/kg in mice. SM-102 and ALC-0315 are clinically validated benchmark ionizable lipids used for comparison.

(F and G) Calculation of tumor total flux or individual data (p/s) using in vivo imaging software. (F) i.v. and (G) i.m. routes of administration. Data are presented as mean ± SD, n = 3 (independent biological replicates). The statistical analysis was measured by one-way ANOVA. ∗∗p < 0.01 and ∗∗∗∗p ≤ 0.0001.

In this study, we prioritized primary human T cells—the relevant clinical substrate for CART cell manufacturing. Activated PBMCs from healthy donors were transfected under identical conditions. Consistent with the DC2.4 results, lipid 6 yielded the highest eGFP positivity and MFI (Figure 2B), demonstrating its robust mRNA delivery capability in primary human immune cells.

Based on its superior in vitro performance, we selected lipid 6 for in vivo evaluation via FLuc mRNA delivery. A route-of-administration study comparing intravenous (i.v.), subcutaneous, intramuscular (i.m.), intranasal, and ocular delivery indicated that i.v. injection produced the strongest and most systemic luciferase signal, whereas i.m. administration led to sustained local expression (Figures S2A and S2B). We directly compared lipid 6 with the benchmark lipids SM-102 and ALC-0315 via i.v. and i.m. injections (Figure 2C). Bioluminescence imaging showed that lipid 6 elicited significantly higher FLuc expression at the 6-h time point for both routes (Figures 2D and 2E). Although i.m. delivery generated lower peak expression than i.v. delivery, it produced a more prolonged, localized signal, which remained detectable at the injection site at 72 h (Figures 2F and 2G). Together, integrated in vitro and in vivo screening identified lipid 6 as the lead formulation for efficient mRNA delivery to immune cells.

Optimization of LNPs for mRNA delivery

Following the identification of lipid 6 as our lead candidate, we systematically optimized its formulation parameters for primary human T cell transfection. We evaluated the effects of mRNA:ionizable lipid weight ratio (nitrogen to phosphate [N/P]), four-component stoichiometry, APOE4 dosing, mRNA loading, and storage stability using eGFP mRNA-loaded LNPs at a standard dose of 200 ng per 5 × 10⁵ cells.

We first employed a design of experiments (DOE) approach to screen 13 formulations by varying the stoichiometry of the 4 LNP components: ionizable lipid, DOPE, cholesterol, and DMG-PEG2000 (Figure 3A). Both ionizable lipid and DMG-PEG2000 concentrations exerted the most substantial influence on transfection outcomes, affecting the percentage of eGFP-positive cells as well as the MFI. Reduced levels of either component diminished the proportion of eGFP-positive cells. Ionizable lipid percentages of 35%–50% yielded higher MFI values, although formulations with 50% lipid occasionally precipitated after storage. DMG-PEG2000 concentrations of 1.5%–2.5% produced the highest MFI; however, because high PEG-lipid concentrations can potentially increase immunogenicity, we minimized its amount (Figures 3B and 3C and Table S2). Based on these initial results, we refined the ionizable lipid and DMG-PEG2000 ranges and conducted a second DOE screen (13 formulations). After 24 h, all formulations achieved >90% eGFP positivity with no significant differences between them (Figure 3D and Table S2). Formulations B11 and B13 produced the highest MFI. Owing to its lower DMG-PEG2000 concentration, we selected B13 as the optimal stoichiometry for subsequent studies (Figure 3E). We titrated the N/P ratio to identify the optimal condition for primary human T cell transfection. The percentage of eGFP-positive cells increased with the N/P ratio and plateaued at N/P = 5. In contrast, the MFI peaked at N/P = 10 before declining at higher ratios (Figures 3F and 3G). By balancing both high positivity and intense expression, we selected an N/P ratio of 10 for all subsequent experiments.

Figure 3.

Systematic screening and optimization of LNP formulations for mRNA delivery to primary human T cells

(A) Schematic of LNP four-component assembly, the nitrogen-to-phosphate (N/P) mix ratio optimization, and the modes of T cell uptake (direct release vs. endocytosis/endosomal escape).

(B–E) Lipid library screening in T cells. Transfection efficiency (eGFP%) and MFI across a screening library of formulations (A1-A13 and B1-B13).

(F and G) Optimization of the N/P ratio. Dependence of eGFP% and MFI on N/P molar ratio.

(H and I) Effect of exogenous APOE4 supplementation on transfection. Dependence of eGFP% and MFI on APOE4 dose (mg/500,000 cells).

(J) Schematic for T cell transfection with different dosage optimization.

(K and L) mRNA dose-response in T cells. Dependence of eGFP% and MFI on mRNA dosage (ng/500,000 cells). All data points represent the mean ± standard deviation (SD) of results from T cells isolated from three independent healthy donors. NTC, non-transfected control.

Based on reports that APOE4 supplementation enhances LNP-mediated T cell transfection,^50,51 we performed a dose-response assessment of APOE4 during transfection of primary human T cells. The percentage of eGFP-positive cells increased with APOE4 concentration and plateaued at 1 μg/mL. This was chosen as the optimal dose (Figures 3H and 3I). We then titrated the amount of encapsulated eGFP mRNA delivered to human T cells (Figure 3J). Both the positivity rate and MFI reached a maximum at 150 ng per 5 × 10⁵ cells, indicating saturating transfection efficiency at this mRNA dose (Figures 3K and 3L).

To determine the optimal storage parameters for lipid 6, we monitored cell viability and transfection efficiency (eGFP%) under various thermal stresses. At 4°C, the LNPs exhibited remarkable stability, maintaining >80% cell viability and ∼100% eGFP expression up to 110 days, regardless of the presence of cryoprotectants (Figures S3A–S3C).

However, elevated temperatures significantly compromised LNP performance. At 25°C and 37°C, a progressive decline in cell viability was observed starting from day 8. By day 16, eGFP% (Figure S3B) and protein expression levels (MFI) (Figure S3C) in the 25°C and 37°C groups dropped precipitously, with the 37°C group showing near-total loss of activity. Interestingly, the addition of trehalose appeared to offer a marginal protective effect on cell viability on day 16 under thermal stress compared with the control and sucrose groups, although it could not rescue the functional delivery of mRNA at extended time points. These results indicate that although the LNP scaffold is relatively robust at refrigerated temperatures, the mRNA-LNP complex is highly sensitive to ambient and physiological temperatures for long-term storage.

LNPs deliver eGFP mRNA to T cells, DCs, NK cells, and tumor-infiltrating lymphocytes

After establishing the efficient mRNA delivery by lipid 6 to primary human T cells, we evaluated its broader applicability across other primary immune cell types relevant to immunotherapy. Using our optimized protocol, we transfected primary human DCs, NK cells, and tumor-infiltrating lymphocytes (TILs) with eGFP mRNA (Figure 4A). Lipid 6 achieved >90% eGFP-positive cells in all three populations, with TILs exceeding 97% positivity (Figures 4B–4E). Furthermore, both activated (>95% positivity) and unactivated (>80% positivity) primary T cells were transfected with high efficiency. These data confirm that the lipid 6 platform facilitates robust gene delivery across diverse primary human immune cells, supporting its potential for multiplexed cell-engineering applications.

Figure 4.

Broad-spectrum delivery efficiency of lipid 6-based LNPs across primary human immune cell subsets

(A) Schematic workflow of LNP-mediated eGFP mRNA transfection into various primary immune cells.

(B–D) Delivery efficiency (eGFP%) in (B) PBMC-derived dendritic cells (DCs), (C) NK cells, and (D) tumor-infiltrating lymphocytes (TILs) isolated from human hepatocellular carcinoma tissues.

(E) Comparison of eGFP expression between activated and naive T cells. Data are presented as mean ± SD from three independent biological donors (n = 3). Specific cell populations were identified by lineage markers (CD11c⁺/CD14⁻ for DCs; CD3⁻/CD16⁺/CD56⁺ for NK; CD3⁺ for T cells/TILs).

Detailed flow cytometry gating strategies are provided in Figure S4.

Ex vivo engineering of human primary CAR T cells

Following the demonstration of efficient eGFP mRNA delivery, we applied the optimized lipid 6 platform to deliver an mRNA encoding MUC1-targeting CAR to primary human T cells. This approach leverages the transient nature of mRNA-mediated expression, which may mitigate the toxicity risks associated with permanently modified CAR T cells.¹³ Herein, we encapsulated MUC1-CAR mRNA in lipid 6 to generate CAR T cells and assess their functional activity.

In this study, lipid 6 was used to deliver MUC1-CAR mRNA to primary human T cells at a concentration of 150 ng/μL. Cell viability and proliferation assays showed that LNP-mediated mRNA delivery did not significantly impair T cell proliferation or survival (Figures 5A and 5B), indicating that the delivery system is physiologically compatible. Flow cytometry confirmed that ∼90% of transfected T cells expressed the CAR protein (Figure 5C), and the CD4⁺/CD8⁺ subset ratios remained statistically unchanged (Figures 5D and 5E), demonstrating that lipid 6 does not preferentially affect any T cell subset. To assess whether the delivery process induced functional exhaustion, the surface expression levels of the exhaustion markers PD-1, LAG-3, and TIM-3 were measured. No significant differences were observed between lipid 6-treated and control groups (Figures 5F–5H), suggesting that the platform does not cause excessive T cell exhaustion or functional decline. Collectively, lipid 6 mediates highly efficient CAR mRNA delivery while preserving T cell fitness and subset integrity.

Figure 5.

Safety profile and phenotypic characterization of LNP-engineered CAR T cells

(A and B) Cell count (A) and viability (B) of activated primary human T cells at 0 and 24 h post-transfection with lipid 6 LNPs (ns, non-significant).

(C) CAR expression efficiency 24 h post-transfection.

(D) Ratio of CD4⁺ and CD8⁺ T cell subsets within the CAR⁺ population; note the adjacent legend for subpopulation identification.

(E) Comparative CAR expression in CD4⁺ vs. CD8⁺ subsets.

(F–H) Representative flow cytometry histograms and quantitative analysis of exhaustion markers LAG-3, TIM-3, and PD-1 in control vs. lipid 6-treated cells. T cells were activated with TransAct for 48 h prior to LNP treatment. Data represent mean ± SD from three independent donors (n = 3).

We evaluated the functional cytotoxicity of MUC1-CAR mRNA delivered by lipid 6 using real-time cell analysis (RTCA) (Figure 6A). T cells transfected with MUC1-CAR mRNA using lipid 6 were cocultured with the human tumor lines MDA-MB-468 (breast cancer) and SKOV3 (ovarian cancer) at various effector-to-target (E:T) ratios (2:1, 1:2, 1:4, and 1:8). After 24 h, RTCA curves demonstrated that the lipid 6 group had significantly enhanced tumor cell-killing capacity in the MDA-MB-468 cell line at all ratios relative to the controls (Figure 6B). Correspondingly, supernatant cytokine analysis revealed markedly increased IL-2, TNF-α, and IFN-γ secretion, indicating robust Th1-type activation (Figure 6C). Flow cytometry demonstrated a progressive upregulation of the exhaustion markers PD-1 and TIM-3 on T cells over time, whereas the proportion of CD8⁺ cells increased markedly (Figures 6D–6G). Parallel experiments with SKOV3 cells reproduced these findings: enhanced cytotoxicity at all E:T ratios (Figure 6I); elevated IL-2, TNF-α, and IFN-γ secretion (Figure 6J); and incremental PD-1/TIM-3 expression together with increased CD8⁺ T cell frequency (Figures 6K–6N). Subsequently, we longitudinally assessed CAR T functionality after lipid 6-mediated MUC1-CAR mRNA transfection (Figure 7A). Flow cytometry showed peak CAR expression at 24 h, followed by a progressive decline to <5% by day 4 (Figure 7B), reflecting the transient nature of mRNA-driven CAR expression. Concurrent CCK-8 assays demonstrated that lipid 6 delivery did not impair T cell proliferation or viability (Figure 7C). Cytotoxicity assays performed at 1, 2, and 3 days post-transfection using MDA-MB-468 target cells at various E:T ratios revealed that the lipid 6 group had significantly higher tumor cell-killing capacity than the controls at each time point (Figures 7D–7F). In summary, lipid 6 enables efficient, transient CAR expression that drives potent and sustained antitumor cytotoxicity while maintaining T cell proliferative capacity and a favorable viability profile, underscoring its utility for generating functional mRNA-based CAR T cells.

Figure 6.

In vitro antitumor efficacy and effector function of mRNA CAR T cells

(A) Enlarged schematic of the real-time cytotoxicity assessment (RTCA) workflow.

(B) Cytolysis of MDA-MB-468 cells at 24 h across various E:T ratios.

(C and D) Cytokine secretion (IL-2, TNF-α, IFN-γ, and CD4/CD8 subset distribution).

(E–G) Flow cytometry analysis of exhaustion markers.

(H) Schematic illustration of the mechanism for mRNA-CAR delivery into T cells via lipid nanoparticles.

(I) Cytolysis of SKOV3 cells at 24 h across various E:T ratios.

(J) Supernatant cytokine secretion (IL-2, TNF-α, and IFN-γ) following coculture with SKOV3 cells.

(K–N) Flow cytometry analysis of CD4/CD8 T cell subset distribution and exhaustion markers (LAG-3, TIM-3 and PD-1) on T cells over time in SKOV3 cocultures.

Data represent mean ± SD from three independent biological donors. Data are presented as mean ± SD, n = 3 (independent biological replicates). The statistical analysis was measured by paired two-tailed t tests. ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p ≤ 0.0001, and ns indicates no significance.

Figure 7.

Kinetic profile of transient CAR expression and long-term cytotoxicity

(A and B) Time course of CAR expression (%) and MFI from day 1 to day 4 post-LNP transfection.

(C) Growth and viability kinetics of LNP-engineered T cells.

(D–F) Continuous RTCA kinetic traces day 1 (D1)/day 2 (D2)/day 3 (D3) showing the percentage of cytolysis for 24 h against different tumor targets. The correlation between transient CAR expression and killing duration is highlighted. Data are presented as mean ± SD, n = 3 (independent biological replicates). The statistical analysis was measured by paired two-tailed t tests. ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p ≤ 0.0001, and ns indicates no significance.

Transcriptomic profiling of LNP-transfected T cells

To explore the global transcriptional changes underlying the high transfection efficiency and preserved T cell function mediated by lipid 6, we performed transcriptomic sequencing (RNA sequencing [RNA-seq]) on primary human T cells treated with lipid 6-MUC1 mRNA or control. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted to identify altered biological processes and signaling pathways.

GO enrichment analysis revealed significant upregulation of gene sets associated with transmembrane transport, membrane dynamics, cell junction organization, and intracellular trafficking, consistent with enhanced endosomal escape and intracellular mRNA delivery (Figures 8A and 8B). KEGG pathway analysis further demonstrated robust activation of cytokine-cytokine receptor interaction, T cell receptor signaling, chemokine signaling, and Th1/Th2 cell differentiation pathways, confirming that lipid 6-mediated MUC1 mRNA delivery triggered physiological T cell activation without exhaustion or aberrant inflammatory signaling (Figures 8C and 8D). These transcriptomic profiles support that lipid 6 enables efficient cytosolic delivery of mRNA while maintaining normal immune cell function and viability.

Figure 8.

Transcriptomic profiling of primary human T cells after lipid 6-mediated MUC1 mRNA transfection

(A and B) GO enrichment analysis of differentially expressed genes showing enriched biological processes, cellular components, and molecular functions.

(C and D) KEGG pathway analysis highlighting upregulation of immune activation, cytokine signaling, and T cell function-related pathways. n = 3 biological replicates.

Discussion

Our study establishes lipid 6 as a rationally designed ionizable lipid-based mRNA delivery system that enables highly efficient, nonviral engineering of primary human immune cells. By systematically optimizing lipid structure, formulation, and transfection conditions, we identified lipid 6 as the lead component that mediates robust mRNA expression in T cells, NK cells, DCs, and TILs while preserving cell viability, phenotype, and effector function. This platform overcomes critical limitations of viral vectors and EP, including insertional mutagenesis, high production costs, and cellular damage, thereby providing a scalable and safety-enhanced approach for ex vivo immune cell modification.

The superior performance of lipid 6 is likely driven by structural features that support efficient intracellular delivery and functional immune activation. Transcriptomic profiling of LNP-transfected T cells revealed significant enrichment of gene sets related to membrane trafficking, endosomal organization, and cellular transport processes, consistent with enhanced intracellular mRNA delivery and release. In line with these observations, the rationally optimized structure of lipid 6 appears to facilitate membrane interactions and fusogenicity, which may promote cytoplasmic mRNA delivery without inducing excessive cytotoxicity. Notably, unlike liver-tropic LNPs that depend on APOE-mediated uptake, lipid 6 shows robust activity in immune cells, potentially through alternative endocytic pathways. This profile, together with the upregulation of cytokine signaling and T cell activation pathways observed in our transcriptomic analysis, supports that lipid 6 enables effective mRNA delivery and downstream immune function in non-hepatic cells.

Despite these advances, several key aspects warrant further investigation. First, the precise intracellular trafficking, uptake mechanisms, endosomal escape processes, and exact role of APOE4 in lipid 6-mediated transfection remain incompletely elucidated. Second, comprehensive long-term safety, biodistribution, and immunogenicity profiles have yet to be fully characterized. Third, while lipid 6 enables efficient generation of functionally potent CAR T cells in vitro, its in vivo antitumor efficacy remains unvalidated in tumor-bearing models. Future studies will employ clinically relevant tumor models to assess antitumor activity, survival benefits, and in vivo persistence, thereby validating the translational potential of this platform. Finally, although transient expression improves safety, it may limit therapeutic persistence in solid tumor microenvironments; future optimization will aim to balance expression duration, antitumor efficacy, and safety using modified mRNA designs and formulation tuning.

In conclusion, this study establishes lipid 6 as a robust, nonviral LNP platform for ex vivo engineering of human immune cells with exceptional transfection efficiency and minimal toxicity. Through the systematic optimization of formulation parameters, the platform achieved high-efficiency mRNA delivery across multiple immune cell types, including primary human T cells, DCs, NK cells, and TILs. Notably, the platform enabled rapid and efficient generation of functional CAR T cells in vitro through CAR mRNA delivery. These cells maintained high viability while exhibiting potent antigen-specific tumor cell-killing ability and cytokine release capacity. Moreover, CAR expression was transient and controllable, potentially mitigating the long-term toxicity risks associated with traditional viral vectors or permanently integrated CAR T cells. While these results support further development toward safer and more scalable CAR T manufacturing, future in vivo efficacy studies will be necessary to confirm clinical utility.

Limitations of the study

Although this work constructed a novel ionizable lipid library and screened out lipid 6 to prepare functional MUC1 CAR T cells with potent in vitro antitumor capacity, several key limitations remain. First of all, all cellular assessments were restricted to in vitro tests; accordingly, long-term in vivo antitumor efficacy, biodistribution, and biosafety of LNP-engineered CAR T cells were not verified in tumor-bearing animal models. Furthermore, even though APOE4 facilitates LNP uptake, the detailed intracellular trafficking and endosomal escape mechanisms of lipid 6 in primary human T cells are still unclear. In addition, only MUC1-targeted CAR mRNA was examined, so the general applicability of this LNP platform for other tumor antigens lacks systematic validation. Finally, this study only relied on transient mRNA-based CAR expression without introducing the TAF3-JL transposon system to prolong CAR persistence, and scalable GMP-level LNP production processes were not established, which together impede further clinical translation.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Qijun Qian (qianqj@shu.edu.cn).

Materials availability

All eight custom-synthesized ionizable lipids (lipid 1–lipid 8) described in Figure S1 and Table S1 are available upon request from the lead contact.

Data and code availability

• •The authors declare that the data supporting the findings of this study are contained within the paper and will be shared by the lead contact upon request.
• •The RNA-seq transcriptomic datasets generated in this study have been fully deposited and publicly released in the NCBI Gene Expression Omnibus (GEO) database under accession number GSE338028 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE338028). All raw sequencing FASTQ files and processed expression matrices are openly accessible.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This research was supported by the Cell and Gene Therapy Project under the Shanghai Science and Technology Innovation Action Plan, Shanghai Municipal Science and Technology Commission (23J21900500). This research was also funded by Shanghai University and Shanghai Cell Therapy Group Co., Ltd. The authors would like to thank all researchers at the Center for Cell Medicine of the Shanghai Cell Therapy Group Corporation and Shanghai University.

Author contributions

H.G., Y.F., and Q.Q. designed the study. H.G., Y.F., Z.T., T.L., C.G., and P.Z. performed the experiments and collected the data. H.G. and P.Z. analyzed and interpreted the data. Statistical analysis was performed by Y.F. and H.G. Y.F. wrote the manuscript. H.G., Y.F., and Q.Q. revised the manuscript. Q.Q. and P.Z. funded the experiments. All authors contributed to the article and approved the submission version.

Declaration of interests

Several co-authors (Z.T., T.L., and Q.Q.) are listed as inventors of authorized Chinese invention patent CN 116456967 B, covering the ionizable lipid scaffold, LNP formulation, and primary T cell transfection platform described in this manuscript. The patent assignees include Zhejiang Jiliang Technology Co., Ltd., Shanghai Jiliang Pharmaceutical Engineering Co., Ltd., and Shanghai Cell Therapy Group Co., Ltd. No stock, equity, consulting fees, or commercial royalties were received by any authors during this study. All experimental design, data collection, statistical analysis, and manuscript writing were completed independently without commercial interference.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
MonoRab Rabbit Anti-Camelid VHH Cocktail [PE]	GenScript	Cat# A02018
Anti-human CD3 PE-Cy5/CD4 PE/CD8 FITC Cocktail	BioLegend	Cat# 319001
PE/Cy5 anti-human CD45RO	BioLegend	Cat# 304208
FITC anti-human CD197	BioLegend	Cat# 353216
PE anti-human CD62L	BioLegend	Cat# 304806
PE anti-human CD366 (Tim-3) Antibody	BioLegend	Cat# 345006
Brilliant Violet 421 anti-human CD279 (PD-1)	BioLegend	Cat# 367422
Alexa Fluor 647 anti-human CD223 (LAG-3) Antibody	BioLegend	Cat# 369304
PE/Cyanine7 anti-human CD107a (LAMP-1) Antibody	BioLegend	Cat# 328618
Anti CD3	Biolegend	Cat# 300426
Anti CD8	Biolegend	Cat# 344730
APC-Cy7 Mouse Anti-Human CD8	BD	Cat# 557834
Anti CD4	BD	Cat# 564419
Anti CD11c	BD	Cat# 559877
Anti CD14	Biolegend	Cat# 367116
Anti CD16	Invitrogen	Cat# 12–0168-42
Anti CD56	Biolegend	Cat# 362504
Chemicals, peptides, and recombinant proteins
Ionizable lipids (Lipid 1–8)	This paper	N/A
SM-102	SINOPEG	06040008800
ALC-0315	SINOPEG	06040008600
Cholesterol	Accela ChemBio	SY012864
DOPE (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine)	Aventis Pharmaceutical Technology	S03005
DMG-PEG2000	Aventis Pharmaceutical Technology	O02005
DSPC	Aventis Pharmaceutical Technology	S01005
Collagenase IV	Sigma	Cat# C5138
DNase I	Takara	Cat# 2270A
Dimethyl sulfoxide (DMSO)	Origen	Cat# CP-70
Fetal Bovine Serum (FBS)	Gibco	Cat# 10091–148
Dulbecco’s Phosphate Buffered Saline (DPBS)	Hyclone	Cat# SH30028.02
N1-Me-ψUTP, rATP, rGTP, rCTP	Hongene Biotechnology	Cat# R5-027, R1331, R2331, R3331,
Ficoll	Cytiva	Cat# 17144003
Dulbecco’s modified Eagle’s medium	Corning	Cat #10–013-CVR
AIM-V Medium	Gibco	Cat# 0870112DK
RPMI 1640	Gibco	Cat# A1049101
T7 RNA polymerase	Roche	Cat# 8140669103
NheI	NEB	Cat# R3131V
Vaccinia virus capping enzyme	Novoprotein	Cat# GMP-M062-M010
S-adenosylmethionine	Novoprotein	Cat# GMP-S062-M005
Oligo(dT)25 columns	ThermoFisher Scientific	Cat# A47384
Tangential flow filtration 100 kDa hollow fiber membrane	Repligen	Cat# D02-E100-05-N
Recombinant human IL-2	Guangheng Pharmaceutical	Cat# S200220004
Recombinant human IL-4	Novoprotein	Cat# CX03
Recombinant human GM-CSF	Novoprotein	Cat# C003
2′-O-methyltransferase	Novoprotein	Cat# GMP-M072-M01
Critical commercial assays
Quant-iT RiboGreen RNA Assay Kit	Thermo Fisher	Cat# Q33140
Human Th1/Th2 Cytokine Kit II (CBA)	BD	Cat# 551809
Pan T cell Isolation Kit	Miltenyi Biotec	Cat# 130–096-535
TransAct	Miltenyi Biotec	Cat# 130–111-160
NK Cell Induction Kit	Baso	Cat# T20153P3
EasySep Human T cell Isolation Kit	StemCell	Cat# 17951
Experimental models: Cell lines
DC2.4	Cell Bank of Chinese Academy of Sciences	GNM51
SKOV3	ATCC	HTB-77
MDA-MB-468	Cell Bank of Chinese Academy of Sciences	TCHu136
Experimental models: Organisms/strains
Mice (6–8 weeks)	Phenotek Biotechnology	N/A
Software and algorithms
FlowJo vX.0.7	BD	N/A
Kaluza v2.0	Beckman Coulter	N/A
GraphPad Prism 10.6.0	GraphPad Software	N/A
FCAP Array v3 software	BD Biosciences	N/A
RNA-seq transcriptomic dataset	NCBI GEO database	Accession No. GSE338028, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE338028
Illumina NovaSeq 6000 sequencing system	Illumina	N/A

Experimental model and study participant details

Cell lines

The SKOV3 cell line was purchased from ATCC, while the DC2.4 and MDA-MB-468 cell lines were obtained from the Cell Bank of Chinese Academy of Sciences. SKOV3 and MDA-MB-468 cells stably express MUC1. All cell lines were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% Fetal Bovine Serum () at 37 °C with 5% CO₂. Polymerase chain reaction–based short tandem repeat analysis was used to verify the cell lines, and the cell lines were regularly tested for mycoplasma contamination. All cell lines were authenticated via STR profiling upon receipt and tested negative for mycoplasma monthly during culture.

Primary human immune cells

Human PBMC samples

All experiments using healthy donor-derived peripheral blood mononuclear cells (PBMCs) were reviewed and formally approved by Milecell Biological Science & Technology Co., Ltd. (Approval No.: 2023-02). Any revisions to the research protocol, informed consent forms, or experimental procedures required supplementary submission and re-approval by this IRB prior to implementation. All PBMC-related research strictly complies with the ethical standards stated in the Declaration of Helsinki.

Whole blood samples from healthy adult volunteers were collected to isolate PBMCs via Ficoll density gradient centrifugation. Written informed consent was acquired from each healthy donor before blood sampling. Complete written informed consent was obtained from every healthy blood donor before blood collection. T cells were purified from PBMCs using Pan T cell Isolation Kit according to the manufacturer’s instructions by magnetically activated cell sorting. The T cells were cryopreserved in liquid nitrogen in a 90% FBS medium containing 10% dimethyl sulfoxide. The cells were then thawed, and TransAct was added to the T cells at a 1:1 ratio. The cells were subsequently cultured in an AIM-V medium containing 2% FBS and 100 U/mL of IL-2 at 37 °C and 5% CO2 for 2 days. NK cells were enriched and cultured from PBMCs using NK Cell Induction Kit.

To obtain human monocytes, PBMCs were cultured in a T75 flask for 2 h at 37 °C with 5% CO₂ at a concentration of 4–5 × 10⁶ cells/mL in an AIM-V medium containing 2% FBS and 100 U/mL of IL-2. After 2 h of incubation, monocytes adhered as a monolayer and nonadherent cells were removed by three washes with sterile Dulbecco’s phosphate-buffered saline. The monocytes were cultured for 5 days in the presence of 1000 U/mL each of recombinant human IL-4 and recombinant human GM-CSF on days 0 and 2. After 5 days, immature DCs were obtained and washed several times with cold and sterile Dulbecco’s phosphate-buffered saline (Dulbecco’s Phosphate Buffered Saline).

Patient-derived TIL samples

All experiments utilizing patient tumor-infiltrating lymphocytes (TILs) isolated from surgical tumor tissues were reviewed and granted ethical approval by the Institutional Review Board of Shanghai Mengchao Cancer Hospital (Approval No.: MCEC-H202101-P01). All protocol amendments and revised informed consent documents must be submitted to this ethics committee for supplementary review and approval before implementation, fully following the Declaration of Helsinki ethical principles.

Primary TILs were separated from surgically resected tumor specimens collected at Shanghai Mengchao Cancer Hospital. All adult tumor patients provided signed written informed consent prior to tumor tissue resection. Every participant was fully informed of the research objectives, in vitro cell processing procedures, foreseeable minor risks, and the unconditional right to withdraw their tissue samples from this study at any time without penalty or adverse medical impact.

Mouse models

All animal experiments were approved by Institutional Animal Care and Use Committee of Phenotek Biotechnology (Shanghai) Co.,Ltd (Approval No: IACUC-AUP-202109–02). All animal housing and experiments were conducted in strict accordance with the institutional guidelines for care and use of laboratory animals and fully comply with the ARRIVE 2.0 reporting guidelines. Six-to-eight-week-old mice were purchased from Phenotek Biotechnology and housed in individually ventilated IVC Sealsafe cages under controlled environmental conditions (20°C–26 °C, 12 h light/dark diurnal cycle). Investigators were not blinded during data collection and quantitative analysis; experimental and control animal cohorts received identical housing, handling and monitoring regimens to eliminate confounding variables. Experimental and control animals were treated equally. We randomly selected six mice and injected them with LNPs at a dosage of 0.5 mg/kg through i.v., i.m., nasal, ocular, and subcutaneous routes (three mice in each group). DPBS and FLuc mRNA were used as controls. After 6 h, we injected 200 μL of 10 mg/mL D-fluorescein potassium salt into each mouse through the tail vein. After 10 min, the mice were placed in the live imaging system to image total fluorescence intensity. After 24, 48, and 72 h, the mice were placed in the live imaging system to image total fluorescence intensity.

Method details

LNP formulation and characterization

LNPs were prepared using a microfluidic mixing technique. Briefly, an ethanolic lipid phase containing the ionizable lipid, DOPE, cholesterol, and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000) at a predetermined molar ratio were prepared in anhydrous ethanol. The mRNA was dissolved in 10 mM citrate buffer (pH 4.0) to form the aqueous phase. The two solutions were mixed using a microfluidic mixer at a volumetric flow rate ratio (ethanolic phase:aqueous phase) of 1:3 to form LNPs. Immediately after mixing, the resulting suspension was transferred into a dialysis membrane with a molecular weight cut-off of 14 kDa and dialyzed against a volume of DPBS 50 times that of the suspension at 4 °C overnight to remove residual ethanol and exchange the external buffer. The final LNP suspension was collected, filtered through a 0.22-μm sterile filter, and stored at 4 °C until use.

The physicochemical properties of the prepared LNPs were characterized. The average hydrodynamic diameter, polydispersity index, and zeta potential were measured at 25 °C using DLS. Each sample was diluted with 1× DPBS to an optimal concentration and measured in triplicate. Subsequently, encapsulated mRNA concentration were determined using Quant-iT RiboGreen RNA Assay Kit according to the manufacturer’s protocol.

In vitro transfection of immune cells

All primary human immune cells (T cells, NK cells, and monocytes for DC differentiation) were isolated from peripheral blood mononuclear cells (PBMCs) of three independent healthy donors (n = 3). For flow cytometry analysis, the following markers were used: CD11c and CD14 for monocyte-derived DCs; CD3, CD16 and CD56 for NK cells; CD3 for T cells.

Following isolation from PBMCs, T cells were cultured in AIM-V medium supplemented with 2% FBS and 100 U/mL of IL-2. Activation was performed using TransAct (Miltenyi Biotec; cat# 130–111-160) for 48 h prior to LNP transfection. LNPs were transfected into primary T cells at two differentiation statuses: nonactivated and activated T cells. Nonactivated T cells were isolated from fresh leukopaks using EasySep Human T cell Isolation Kit and thawed for subsequent LNP transfection without T cell activation. Activated T cells were prepared by stimulating thawed T cells with TransAct at a 1:1 ratio for 48 h in AIM-V medium supplemented with recombinant human IL-2 (100 IU/mL). The resulting cells were transfected immediately.

Cells were transfected with LNPs at different payload dose levels with 0.2 million DC2.4 cells or 0.5 million T cells, TILs, DCs, and NK cells per well in a 24-well plate. All experiments were performed in duplicate. LNP were mixed well and incubated with primary cells for 6 h inside a 37 °C CO₂ incubator and washed three times with culture medium. Cells were kept inside the 37 °C CO₂ incubator until ready for eGFP or CAR expression analysis or functional assessment.

Flow cytometry

CAR expression and surface markers of transfected T cells were flow cytometrically analyzed on Cytek Northern Lights or DxFLEX Flow Cytometer. The antibodies used were as follows: MonoRab Rabbit Anti-Camelid VHH Cocktail [PE]; Anti-human CD3 PE-Cy5/CD4 PE/CD8 FITC Cocktail, PE/Cy5 anti-human CD45RO, FITC anti-human CD197, PE anti-human CD62L, PE anti-human CD366 (Tim-3) Antibody, Brilliant Violet 421 anti-human CD279 (PD-1), Alexa Fluor 647 anti-human CD223 (LAG-3) Antibody, PE/Cyanine7 anti-human CD107a (LAMP-1) Antibody, Anti CD3, and Anti CD8; and APC-Cy 7 Mouse Anti-Human CD8, Anti CD4. Specifically, MonoRab Rabbit Anti-Camelid VHH Cocktail [PE] was utilized for CAR detection.

Cells were harvested and stained with fluorophore-conjugated antibodies against surface markers for 30 min at 4 °C in the dark. Briefly, intact lymphocytes were first gated based on forward scatterFSC and side scatter (SSC) profiles, followed by singlet gating to exclude doublets. T cells (CD3⁺), NK cells (CD3⁻ CD16⁺ CD56⁺), DCs (CD11c+ CD14⁻), and TILs (CD3⁺) were distinguished. CD3⁺ T cells were further subset into CD4⁺ and CD8⁺ populations. A comprehensive gating strategy is provided in Figure S4.

Cytotoxicity assay

The cytotoxic activity of T cells was evaluated using xCELLigence RTCA system. In total, 1 × 10⁴ SKOV3 or MDA-468 cells were seeded in each well of a 16-well microplate for 24 h. MOCK-T and MUC1 CAR-T cells were added at different E:T ratios. The same volume of culture medium was used as a negative control. The signal–time curve was plotted to show the cytotoxicity of T cells.

Cytokine secretion analysis

To verify the specificity of MUC1 CAR-T cells in antigen recognition, 5 × 10⁵ MOCK-T or MUC1 CAR-T cells were inoculated into 12-well plates coated with MUC1 antigen (5 μg/mL) and cultured for 24 h. Cell supernatants were analyzed using Human Th1/Th2 Cytokine Kit II. Cytokine detection was performed using the Navios flow cytometer; data were calculated using FCAP Array v3 software.

TIL isolation and culture

Tumor-infiltrating lymphocytes (TILs) were isolated from fresh surgical specimens of hepatocellular carcinoma patients (n = 3) under an approved ethical protocol (Approval No.: MCEC-H202101-P01). TILs were isolated via mechanical dissociation followed by enzymatic digestion, and identified as the CD3⁺, CD4+/CD8+ lymphocyte population within the tumor mass. Fresh tumor tissues were cut into small pieces (∼1 mm³) and enzymatically digested in PBS containing Collagenase IV, hyaluronidase, and DNase for 3–4 h at 37 °C. The digested suspension was filtered through a 70 μm cell strainer to obtain a single-cell suspension. TILs were isolated using discontinuous Ficoll density gradient centrifugation (75% + 100%). Purified TILs were defined as CD3⁺ TILs and cultured in RPMI 1640 medium supplemented with 10% FBS and 1000 U/mL recombinant human IL-2 for expansion and functional assays.

CAR mRNA construction and in vitro transcription

eGFP, Firefly luciferase (FLuc) and MUC1 mRNAs were synthesized in accordance with the established protocol. The main raw materials used were rNTPs (rATP, rGTP, rCTP and N1-Me-ψUTP) and T7 RNA polymerase. Prior to in vitro transcription, plasmid templates encoding eGFP and MUC1 were linearized using NheI. In vitro transcription was conducted at 37 °C for 4 h, using 200 μg of the linearized plasmid per 10 mL of reaction mixture. The plasmid template was digested using DNase I. The solution was buffer-exchanged by tangential flow filtration with a 100 kDa hollow fiber membrane to prepare for the capping reaction. The capping reaction was performed using Vaccinia virus capping enzyme and 2′-O-methyltransferase in the presence of S-adenosylmethionine. The resultant mRNAs were purified using affinity chromatography with Oligo(dT)25 columns on an ÄKTA pilot 600S chromatography system, and their quality was assessed using Fragment Analyzer 5200.

RNA-seq and transcriptomic analysis

Total RNA was isolated from activated primary human T cells at 24 h after Lipid 6–MUC1 mRNA transfection (treatment group) or from untreated activated T cells (control group), with n = 3 per group. RNA sequencing was performed using an Illumina platform. Raw reads were aligned to the human reference genome using HISAT2, and gene expression levels were quantified with featureCounts. Differential expression analysis was performed using DESeq2, with adjusted p < 0.05 and |log₂FC| > 1 set as the significance threshold. Gene Ontology (GO) biological process and Kyoto Encyclopedia of Genes and KEGG pathway enrichment analyses were performed using the clusterProfiler package in R.

Quantification and statistical analysis

Unless specified otherwise, all datasets originate from a minimum of three independent biological replicates, and all displayed images serve as representative experimental readouts. Quantitative outcomes are reported as mean ± standard deviation. Statistical evaluations were executed via two-tailed t tests or one-way analysis of variance (ANOVA) coupled with Tukey’s post-hoc comparison test, utilizing GraphPad Prism 10.6.0 software. Any p-value below 0.05 was defined as statistically meaningful. Comprehensive information regarding sample quantities and analytical workflows is documented within each corresponding figure legend.