mRNA-based CAR T cell engineering: Unmodified mRNA enables high CAR expression without innate immune activation in T cells

Abstract

Unmodified, uridine-containing mRNA is known to trigger antiviral immune responses, inflammatory signaling, and apoptosis in transfected cells. To avoid this and enable high expression, modified nucleosides such as N1-methylpseudouridine have become the gold standard for mRNA applications including T cell engineering, albeit at increased cost. Here, immune responses toward mRNA were evaluated across five primary human cell types. Remarkably, T cells, unlike other immune and non-immune cell types tested, exhibited no immune activation by unmodified mRNA. T cell viability and cytokine secretion remained unaffected, regardless of mRNA delivery method via lipid nanoparticles or electroporation. The absence of nucleotide modifications improved expression of chimeric antigen receptor (CAR) in activated T cells and CAR-T cell cytotoxic potency. By eliminating the need for mRNA-nucleoside modification in CAR-T cell engineering, our findings challenge existing paradigms and position mRNA as a non-inflammatory, minimally invasive and highly efficient tool for T cell engineering, while simplifying and reducing manufacturing cost.

Graphical abstract

Proinflammatory innate immune activation by unmodified mRNA is a limitation, which mandates mRNA-nucleoside modification across applications. However, Drzeniek and colleagues reveal that it is absent in cultured T lymphocytes. This eliminates the risk of inflammatory side effects and functional impairment in mRNA-engineered T cells, such as transient CAR-T cell therapy.

Introduction

The CAR-T cell landscape is being rapidly expanded by revolutionary clinical responses in autoimmune disease^1,2 and promising preclinical data in fibrosis and senescence.^3,4 However, virus-transduced CAR-T-cell therapies are often associated with high cost, burdensome logistics and lingering safety concerns, making their widespread application challenging.⁵ Consequently, non-malignant applications have sparked an acute interest in the use of in vitro transcribed (IVT-) mRNA technology for transient CAR-T cell manufacturing, which could improve safety and reduce manufacturing cost and complexity,^6,7 but has traditionally been limited by excessive immune activation upon recognition of IVT-mRNA by pattern recognition receptors (PRR).^8,9 Thus, mRNA nucleoside modifications which can avoid PRR recognition, such as N1-methyl-pseudouridine (m1Ψ) or 5-methoxy-uridine (5moU),^10,11 are often cited as a requirement, also in the context of mRNA-CAR-T.^12,13 m1Ψ has become a gold standard in the mRNA field and to the best of our knowledge, most studies on mRNA-based CAR-T cells where nucleoside chemistry is disclosed use m1Ψ-modified mRNA.^14,15 Other studies emphasize the role of dsRNA removal to ensure the efficacy of mRNA-CAR-T cells.¹⁶ Surprisingly little is known about the different sensitivity of different cell types toward mRNA-triggered immune activation and thus about cell type-specific requirements for mRNA modification.

In previous studies, we extensively characterized the immune response toward mRNA synthesized with different chemical modifications in bone marrow stromal cells. We found that sensing of uridine-containing synthetic mRNA leads to profound changes in cell phenotype and function, including pro-inflammatory paracrine signaling, and that uridine modifications are crucial to control these cell-intrinsic inflammatory responses and to ensure optimal mRNA expression. The efficacy of these improvements varied quantitatively between uridine modifications, with 5moU minimizing paracrine changes most effectively, and m1Ψ and 5moU yielding the highest peak expression.^10,11 Asking to what extent those insights could be transferred to other cell types, here we screened responses toward unmodified and nucleoside-modified (m1Ψ, 5moU) mRNA in 5 primary human cell types. Based on our interest in systemic autoimmune diseases, such as systemic lupus erythematosus, our focus was on putative targets of RNA based therapies, such as T cells, macrophages and cell types from the connective tissue, vasculature and kidney, to exemplify endothelial, epithelial and stromal cells, as well as a lymphoid and a myeloid immune cell type.

Remarkably, and in contrast to all other cell types tested, T cells exhibited no cell-intrinsic immune response to unmodified mRNA, regardless of the transfection method. Unmodified, uridine-containing mRNA did not negatively impact CAR-T cell functionality and even resulted in higher expression of CD19-CAR in primary human T cells and slightly higher CAR-T killing activity, compared to industry-standard m1Ψ-modified CAR-mRNA.

Results

Primary human T cells, monocyte-derived macrophages, and three non-hematopoietic primary cell types representative of different tissues (human umbilical vein endothelial cells, bone marrow stromal cells, and renal tubular epithelial cells) were screened for antiviral responses toward IVT-mRNA.

Non-hematopoietic cells showed very similar interferon response, caspase activation and cell death in response to unmodified mRNA, which was greatly reduced by uridine substitution with m1Ψ and further reduced by 5 moU (Figures 1A–1C). Macrophages showed the strongest immune response and apoptosis to unmodified mRNA (Figure 1D). This was also reflected in the relatively large impact of uridine modification on mRNA expression in macrophages.

Figure 1.

Cell-intrinsic responses toward mRNA and its nucleoside chemistries vary between cell types

(A) Umbilical vein endothelial cells, (B) bone marrow stromal cells, (C) renal tubular epithelial cells, (D) macrophages and (E) T cells (all primary human cells) were transfected with unmodified (U) or m1Ψ-modified or 5 moU-modified GFP-encoding mRNA using lipofectamine complexes (or LNP for T cells). UNTR = untransfected cells. Cellular responses were measured 24 h post-transfection (from left to right): The expression level (MFI = mean fluorescence intensity) and the fraction of mRNA-expressing cells measured by quantifying GFP fluorescence using flow cytometry with equal GFP-laser settings for all measurements of all cell types. Cell viability measured using flow cytometry. Increase in activity of caspases 3 and 7, expressed as fold change from untransfected cells, was measured using a luminescence-based assay. Concentration of type-I interferons α and β in supernatants. Data presented as mean ± SD, n ≥ 3. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns = not significant.

Unlike all other tested cell types, T cells showed no adverse response to unmodified mRNA in terms of interferon secretion, caspase activation, or cell death and all three mRNAs were expressed equally (Figure 1E).

The lack of a recognizable immune reaction toward unmodified mRNA in T cells did not depend on the transfection method: T cells displayed no increased cell death in response to unmodified mRNA for either electroporation (EP) or lipofection (LNP; Figures 2A and S1), and did not show any interferon-β induction or apoptosis in side by side comparison with macrophages, which did display both responses, although the same LNP-formulation was used to transfect both cell types (Figures 2B and 2C). Electroporated T cells exhibited some viability loss in response to EP-related stress (mock-EP), which did not increase when mRNA was added.

Figure 2.

Human T cells show no innate immune activation by unmodified IVT-mRNA irrespective of the transfection method

Unmodified mRNA was introduced into primary human T cells using lipofection (LNP) and electroporation (EP) to exclude dependency of immune activation on the delivery mechanism. (A) Representative set of flow cytometry plots of T cells 24 h following either transfection method is shown out of n = 12 from 4 biological donors (Figure S1). (B) Interferon β secretion from T cells was quantified using ELISA after transfection with unmodified mRNA (or control condition without any mRNA) through either EP or LNP. Monocyte-derived macrophages (MΦ) were used as a positive control. n = 3 donors. Median is shown. (C) Cell death following transfection with unmodified mRNA (or control condition without any mRNA) through either EP or LNP was quantified using DAPI staining and flow cytometry. Monocyte-derived macrophages (MΦ) were used as a positive control. n = 4 donors. Mean is shown. (D) Protein array measuring 105 cytokines in T cell supernatants 24 h post-transfection with unmodified mRNA. n = 1 array from culture supernatant of n = 1 biological donor. (E) Publicly available bulk RNA sequencing data¹⁷ was interrogated for transcripts of pattern recognition receptors relevant for intracellular recognition of mRNA (NOD2, RIGI, MDA5, TLR3, TLR7, and TLR8), DNA (AIM2 and CGAS) and related adapter proteins (MyD88, TRIF, and MAVS). TPM: transcripts per million. n = 4. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns = not significant.

To assess whether mRNA transfection influenced the production of cytokines beyond type-I interferons, a protein array profiling 105 cytokines was performed on T cell supernatants. T cells transfected with unmodified mRNA via either EP or LNP exhibited cytokine secretion patterns indistinguishable from the untreated and mock controls, including unaffected levels of IFN-γ, IL-2, and TNFα (Figure 2D).

Interrogation of RNA-sequencing data from two independent sources^17,18 revealed that among RNA-sensing PRRs, TLRs 7, and 8, which recognize unmodified uridine were expressed in T cells at lower levels than in other leukocytes, suggesting that T cells might not be equipped to recognize uridine-containing mRNA as a pathogen-associated pattern (Figures 2E and S2).

Finally, the effect of nucleoside modifications on CAR-T cell functionality was examined (Figure 3A). Unmodified CAR-mRNA resulted in higher CD19-CAR expression than m1Ψ-modified and 5 moU-modified mRNAs (Figures 3B, 3C, S3A, and S3B), while preserving equally high T cell viability (>90%; Figures 3D, 3E, and S3C) and unperturbed cytokine secretion patterns across several time points (Figures 3F and S4). Consistent with the absence of a type I IFN antiviral response toward uridine, IFNβ levels were not significantly elevated in the unmodified mRNA condition (Figure S4). Unmodified-mRNA-transfected CAR-T cells displayed cytotoxic activity against B cells (Figure 3G) which was CAR-specific (Figure S3D) and at least on par with CAR-T cells generated using modified mRNA across multiple time points and effector: target cell ratios (Figures 3H and S5). When exposed to a higher target cell burden (1:4), CAR-T cells generated with unmodified mRNA displayed more effective killing than m1Ψ-CAR-T cells (Figures 3G–3I).

Figure 3.

Unmodified mRNA enables high CAR expression without compromising T cell functions

(A) Schematic: mRNA with or without nucleoside modifications encoding anti-CD19 chimeric antigen receptor was electroporated into T cells. Resulting CAR-T cells were characterized with regards to their CAR expression, viability, paracrine activity and killing of target B cells. (B–D) n = 2 experiments with n = 2 separately synthesized CAR mRNA batches, total of n = 8 donors. Mock refers to cells subjected to the electroporation process with no mRNA present. Flow cytometric quantification of the fraction of CAR-positive T cells (B), CAR expression intensity (C) and T cell viability (D). (E) Activity of caspases 3 and 7 was measured using a luminescence-based assay, normalized to cell counts and expressed as fold-change relative to mock-electroporated T cells (dotted line), n = 16: n = 2 technical replicates from n = 8 cell donors. (F) Protein array of 105 cytokines in T cell supernatants 24 h post-transfection with unmodified or uridine-modified mRNA. n = 2 arrays, each from pooled culture supernatant of n = 4 biological donors transfected with n = 2 CAR mRNA batches. (G–I) Killing activity of CAR-T cells or mock-electroporated T cells co-cultured with GFP-labelled Nalm6 B-cells were assessed at 1:1, 1:2 and 1:4 effector:target ratios. The total GFP-positive area was measured every 4 h over a 28-h period using the ImageXpress PICO automated imaging system to quantify B cell elimination kinetics. n = 4 donors. (G) Representative image areas from the PICO system at effector:target = 1:4 ratio at 0 h, 16 h, and 28 h showing GFP-fluorescent target cells. Scale bars: 0.5 mm. (H) Killing kinetics at (CAR) effector:target ratios of 1:1, 1:2, and 1:4 and (1) significance levels from two-way ANOVA and Tuckey’s post-hoc testing for differences between mRNA groups at the effector:target = 1:4 ratio. Data in (B–E) and (H) presented as mean ± SEM. p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns = not significant.

Discussion

Pro-inflammatory responses toward IVT-mRNA can reduce cell viability, functional fitness, and mRNA expression,¹¹ but little is known about the cell type-dependence of those side effects. Despite multidimensional profiling, we found no indication of immune response or downstream effects to unmodified mRNA in primary T cells, in contrast to other human cell types of different lineages (Figure 1). Importantly, the T cell secretome including IFN-γ, TNFα, and IL-2, remained unaffected by mRNA treatment, in contrast to very prominent secretome shifts previously described in stromal cells.¹¹

Although T cells are difficult to lipofect¹⁹ and thus required mRNA-delivery in LNPs or via EP, the choice of mRNA-delivery method was not responsible for the absence of immune activation, as shown in side-by-side comparison with macrophages (Figure 2).

We were interested in the translational implications of our findings for CAR-T cell generation. Compared to virus-transduction of CAR-T cells, mRNA-based manufacturing could reduce cost and time,⁶ but the currently gold standard use of chemically modified nucleosides increases mRNA manufacturing cost by up to ∼35% and mRNA-triggered inflammation is considered a major limitation and potential risk of the technology. Cell-intrinsic sensing and responses toward exogenous nucleic acid are gaining relevance in T cell engineering as recent studies reveal production of type-I interferon in T cells upon activation of the STING pathway by exogenous DNA.^20,21 This results in reduced fitness and dose-dependent toxicity in DNA-based T cell engineering for example when DNA template for CRISPR-Cas9 knock-in is delivered.^21,22 In contrast, when mRNA was delivered to T cells, we found no evidence of a cell-intrinsic immune response or functional impairment. This positions RNA technology as an attractive alternative platform for T cell engineering, compared to DNA-based editing.

As a potential mechanistic link, we show that T cells express DNA sensors, such as CGAS and AIM2, while several RNA-sensing receptors, including TLRs 7 and 8 which recognize uridine-rich motifs on exogenous RNA,^9,23 are weakly expressed in T cells (Figure 2E). However, this study provides only indirect insight into RNA sensing in T cells, prompting exciting avenues for further mechanistic investigation. The RNAs used in this study contain double-stranded or uncapped RNA contaminants, which need to be removed or otherwise avoided in a clinical-grade product. However, in this study their presence did not lead to any detectable cell death or type-I interferon response in T cells (or in any of the other tested cell types when 5 moU-modified mRNA was used; Figure 1), further highlighting the apparently low sensitivity to synthetic mRNA in T cells on the one hand, and the importance of uridine modification in other cell types on the other hand. However, RNA sensing is highly complex and involves various recognizable motifs beside uridine, such as double-stranded or uncapped structures, as well as alternative functions of RNA sensing receptors in T cells aside from type-I interferon responses,²⁴ which should also be investigated with unmodified mRNA.

While in other contexts unmodified, uridine-containing mRNA is considered highly pro-inflammatory⁸ and can lead to functional impairment of therapeutic cells,^10,11 in CAR-T cells we did not find any evidence of functional impairment by unmodified mRNA (Figure 3). On the contrary, our data from two independently synthesized CAR-mRNA batches, two independent transfection experiments and a total of 8 biological donors, shows a higher expression of unmodified CAR-mRNA, which also translated to a slightly higher cytotoxic potency of CAR-T cells generated with unmodified nucleotide chemistry compared to m1Ψ and 5 moU (Figure 3). This observation is in line with a report that also finds minor increase in CAR expression by unmodified versus m1Ψ-modified mRNA in T cells but emphasizes a more prominent effect of dsRNA impurities,¹⁶ which need to be removed in a clinical-grade product.

Expression from unmodified mRNA was consistently high across different target specificities and sequences, whereas expression from the modified mRNAs was more variable or even lower. Reduced or variable translation of modified mRNA may occur due to less efficient ribosomal processing of chemically modified uridine derivatives or the formation of secondary structures that hinder translation compared to a natural transcript and may therefore also be influenced by the specific sequence of the mRNA.²⁵

Our report demonstrates that uridine-containing mRNA, which is highly pro-inflammatory in other cell types, neither elicits inflammation in expanded T cells nor impairs function or potency of CAR-T cells. This highlights mRNA technology as an effective and cost-efficient alternative for CAR-T cell manufacturing and removes concerns over undesired immune response in the ex vivo transfection setting. In the “in vivo CAR” setting,^26,27 where transfection of T cells occurs in situ by targeted LNPs, concerns over systemic inflammation still apply because off-target transfection of non-T cells cannot be completely avoided. Nonetheless, for in vivo CAR-T therapies, our study implies stricter targeting of T cells and de-targeting of macrophages as a potential strategy to limit inflammatory side effects.

Materials and methods

Cell culture

Blood was drawn from n ≥ 3 consenting adult donors (Charité ethics committee approval EA4/091/19). From the mononuclear fraction obtained by density gradient centrifugation, primary T cells and monocytes were isolated using magnetic cell sorting with beads against CD3 and CD14, respectively (Miltenyi Biotec, Germany).¹¹ T cells were cultured and activated in ImmunoCult-XF T cell expansion medium with recombinant IL-2 and ImmunoCult humanCD3/CD28/CD2 T cell-activator (Stemcell Technologies, Canada) for 48 h before transfection. Macrophages were differentiated within 7 days of culture in RPMI1640 containing 10% FCS, 1% penicillin/streptomycin and 50 ng/mL M-CSF (Miltenyi).^11,28

Bone marrow stromal cells (n = 2 donors) were supplied by the BIH Center for regenerative therapies (BCRT) core facility “Cell Harvesting”.²⁹ Written informed consent was given, and approval was obtained from the Charité IRB/local ethics committee (EA2/089/20). Human umbilical vein endothelial cells (n = 2 donors) were purchased from Promocell, Germany. Tubular epithelial cells (n = 2 donors) were a kind gift from Prof. Babel’s group at BCRT. Non-immune cells were cultured in low-glucose DMEM containing 10% FCS, 1% Glutamax, and 1% penicillin/streptomycin.

mRNA synthesis and transfection

DNA sequences encoding enhanced green fluorescent protein (GFP) or chimeric antigen receptor (CAR; Table S1) were cloned into the pRNA2-(A)128 vector, amplified by PCR and transcribed using the TranscriptAidT7 kit (Thermo Fischer, USA) into mRNA containing anti-reverse cap analog, uridine, m1Ψ or 5moU (all Jena Bioscience, Germany). mRNA was purified using lithium chloride precipitation, washed in 70% ethanol and stored in aqueous solution, as previously described.¹¹

mRNA was complexed with lipofectamine MessengerMax (LMM; ThermoFischer) at a ratio of 1 μg:2 μl¹⁰ and added to cells. Because LMM did not transfect T cells (Figure S6), mRNA lipid nanoparticles (LNPs) were formulated with Genvoy T cell lipid mix (Cytiva, USA) using a NanoAssemblr Spark (setting #3) and added to cells.

T cells or macrophages were electroporated using a 4D-Nucleofector (Lonza, Switzerland) in P3 buffer using the EH-115 or DP-148 programs, respectively. All transfections in this study were carried our using a dose of 2 μg mRNA per 1∗10⁶ cells.

Quantification of mRNA expression and cell death

GFP fluorescent imaging was performed on a Nikon ECLIPSE Ti (Nikon Instruments, Japan). GFP- and CAR-expression were quantified using MACSQuantVYB (Miltenyi) or CytoflexLX (Beckmann Coulter, USA) flow cytometers. CD19-CARs were stained using anti-Myc-Tag (CAR1³⁰ 1:50; 9B11; CellSignalingTechnology, USA) or anti-IgG-Fcγ (CAR2²²; 1:50; 109-605-098; Jackson ImmunoResearch, USA) antibodies. DAPI was used (1 μg/mL) for live/dead discrimination. Data was analyzed in FlowJov10; a representative gating strategy is shown in Figure S7. Caspase activity was quantified as previously described.¹⁰

Cytokine and interferon analysis

Concentration of interferons α and β and tumor necrosis factor α were measured by ELISA and T cell secretomes were measured using ProteomeProfilerXL Human Cytokine array (all R&D Systems, USA) 24 h post-transfection or at time points indicated in the respective figures, according to the manufacturer’s instructions.

Cytotoxicity assay

CAR-dependent cytotoxicity was assessed in a VITAL assay, as previously described.³⁰ 8 h post-transfection, CAR-T cells were incubated at decreasing ratios with Nalm6 target (CD19+ GFP+) and control (CD19- RFP+) cells for 8 h, stained with DAPI, and ratios of surviving target:control cells were analyzed by flow cytometry. Cytotoxicity was normalized against mock-transfected, CAR-negative T cells using the formula:

$$\text{Cytotoxicity}=1-\left(\left[\mathrm{T}:\mathrm{C}\right]\text{CAR}/\left[\mathrm{T}:\mathrm{C}\right]\text{mock}\right)$$ (Equation 1)

For killing kinetics, CAR-T cells or mock-electroporated T cells were co-cultured with CD19-expressing GFP-labeled Nalm6 B-cells (2.5∗10⁴/well) on a fibronectin-coated 96-well clear bottom imaging plate (PhenoPlate, PerkinElmer) at effector:target cell ratios of 1:1, 1:2, and 1:4. The total GFP-positive area was measured every 4 h over a 28-h period using the ImageXpress PICO automated imaging system (Molecular Devices) to assess target cell killing relative to signal at 0 h and normalized to wells without T cells (Nalm6 only).

Data analysis

Statistical analysis was carried out in Prism10 (GraphPad Software., USA). A one-way or two-way ANOVA with Tukey’s or Dunnet’s post-hoc test was used to test for significant differences between groups. Data shown as mean and standard error of the mean, unless specified. Heatmaps were generated in R and Python v3.11.0.

Data and code availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the supplemental information. All data generated in this study is available from the corresponding author on reasonable request. The pRNA2-(A)128 vector is available on Addgene (ID: 174006). The plasmid containing the CD19-CAR 2 sequence can be obtained from Addgene (ID: 183473).

Acknowledgments

This project had received funding from the Einstein Center for Regenerative Therapies (ECRT) Advanced Scientist Kickbox and the European Rare Diseases Research Alliance (ERDERA) under the European Union’s Horizon Europe research and innovation program, grant no. 101156595. Funding was provided by the Helmholtz Association and by the Federal Ministry of Education and Research, Germany, in the Program Health Research (BCRT grant no. 13GW0098 and 13GW0099). The authors would like to thank Prof. Katalin Karikó (Szeged University, Szeged, Hungary; University of Pennsylvania, Philadelphia, USA) for valuable discussion and Lisa Hemmerling and Yusuf Aydemir (both Charité, Berlin, Germany) for their help with experiments. The authors also thank Prof. Nina Babel (BCRT) for providing renal TEC cells and the BCRT Cell Harvesting Core Unit (BCRT-CH) of the Berlin Institute of Health, Charité—Universitätsmedizin Berlin for their excellent technical assistance and support. Figures were created using BioRender.com and GraphPad Prism.

Author contributions

Investigation, formal analysis, visualization, N.Kahwaji.; investigation, formal analysis, N.Kotzian.; investigation, visualization, J.M.P.; investigation, formal analysis, Y.P.; validation, methodology, J.K.; methodology, S.P.; software, A.L.H.; investigation, A.Klaas.; methodology, C.M.D.; methodology, writing – review & editing, M.L.; formal analysis, writing – review & editing, A.P.; writing – review & editing, A.Kleyer. writing – review & editing, D.N.S.; methodology, resources, D.L.W. writing – review & editing, C.P.; writing – review & editing, G.K.; methodology, resources, funding acquisition, writing – review & editing, M.S.H.; conceptualization, funding acquisition, writing – review & editing, H.D.V.; conceptualization, supervision, resources, writing – review & editing, M.G.; conceptualization, project administration, supervision, methodology, investigation, formal analysis, visualization, writing – original draft, funding acquisition, N.M.D.

Declaration of interests

N. Kahwaji and C.M.D. are currently employees of Pantherna Therapeutics GmbH, focused on RNA therapeutics. H.-D.V. is an employee and shareholder of Checkimmune GmbH. H.-D.V. and D.L.W. are co-founders of TCBalance Biopharmaceuticals GmbH focused on regulatory T cell therapy. The opinions expressed in this article are those of the authors and not necessarily those of Pantherna, TCBalance, or Checkimmune. Pantherna, TCBalance, or Checkimmune were neither financially involved in the creation nor in the publication of this article.