Modification of mRNA can result in an increase in protein expression in vitro

To assess the impact of mRNA modification and LNP formulation on HA protein expression, mRNA encoding Sing16 HA in either a UNR or MNR format was formulated as OF-02, cKK-E10, or SM-102 LNPs (Supplementary Table 1) and transfected in primary human myoblast (HSKM) and primary human dendritic cells (hDCs) at different doses. Protein expression was measured by immunofluorescence imaging (IF) in HSKM cells (Fig. 1a), and by flow cytometry in hDCs (Fig. 1d) at 24 h post-transfection. Representative flow cytometry plots for measuring HA protein expression were shown in Fig. 1c. In both cell types and across multiple doses, the m1ψ MNR mRNA conferred significantly higher target protein expression compared to UNR mRNA for cKK-E10 and OF-02 LNPs. However, for the SM-102 formulation cell type-specific differences in protein expression between UNR and MNR mRNAs were observed. While MNR mRNA trended toward higher HA protein expression in dendritic cells with SM-102, UNR mRNA showed a higher protein expression in HSKM and there was no difference in cell viability (Fig. 1b, Supplementary Fig. 6). These results suggest that the choice of ionizable lipid and target cell type can result in differences in protein expression.

a Graph represents mean fluorescence intensity (MFI) of hemagglutinin (HA) protein expression of 1 μg or 10 μg of Sing16 HA mRNA-LNPs transfected per million cells, determined using IF 24 h post-transfection in primary human myoblast (HSKM) cells. b Graph represents the percentage of viable hDCs transfected with 5 μg of Sing16 mRNA-LNPs determined using flow cytometry 24 h following transfection. c Representative flow cytometry plots for determining HA protein expression in hDCs are shown. d Graph represents the percentage of Sing16 HA positive cells following transfection of hDCs with 5 μg of Sing16 HA mRNA-LNP, determined using flow cytometry 24 h post-transfection. P-value determined by unpaired T-test analysis. N = 3 experimental replicates and 6 replicates in each experimental repeat for figures a, b and d. e Graph represents mean MFI of HA protein from HSKM cells transfected with 1 μg/million cells of mRNA OF-02 LNP encoding four different strains: A/Wisconsin, A/Tasmania, B/Washington and B/Austria detected with strain-specific primary antibody. n = 3 experimental replicates with n = 8 replicates for each experimental repeat. P-values were determined by unpaired T-test analysis. N = 3 independent donor samples. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

In addition to Sing16 HA, we tested the expression of HAs from 4 other influenza strains- Wisconsin, Tasmania, Washington and Austria in the presence or absence of the m1ψ modification and formulated as OF-02 LNPs (Fig. 1e). As previously observed for Sing16 HA mRNA, MNR Wisconsin HA strain demonstrated a higher expression than UNR while Tasmania HA showed comparable expression between UNR and MNR. Washington HA and Austria HA results were considered inconclusive due to higher dsRNA levels. Overall, our data suggests that incorporation of chemically modified nucleosides such as N1-methylpseudouridine can increase protein production in mRNA vaccines although the extent of increase can depend on the impurities, sequence context and delivery.

Both unmodified and modified mRNA-LNPs can cause global translational repression in vitro

To assess the impact of mRNA modification on global cellular translation, we performed a puromycin incorporation assay which measures the total amount of active translation in the cell by quantifying the amount of puromycin incorporated into growing polypeptide chains (Fig. 2a). HSKM cells were transfected with either UNR or MNR Sing16 HA mRNA delivered as OF-02 or cKK-E10 LNPs and cultured for 20 h. Cells were then treated with puromycin to label all actively translating polypeptides, after which proteins were separated by SDS-PAGE and puromycin incorporation measured by blotting with an anti-puromycin antibody (Fig. 2b). Untreated cells showed a strong incorporation of puromycin that was abolished by treatment with the translation inhibitor cycloheximide. Transfection with both UNR and MNR mRNA at the lowest doses tested showed translational inhibition by ~58% with both LNPs (Fig. 2c), indicating transfection with mRNA in general impacts cellular translation. At all dose levels MNR showed higher global translation levels than UNR, with MNR LNPs demonstrating 40% and 46% higher translation levels with cKK-E10 and OF-02 LNPs, respectively. The increase in the differential observed over the dose response curve further supports that UNR mRNA has a stronger effect on global translation repression. Further, OF-02 LNPs exhibited greater translational repression than cKK-E10 at all dose levels.

Puromycin incorporation assay to measure global cellular translation by western blot. a Mechanism and workflow for puromycin incorporation assay by western blot. b Western blot image using anti-puromycin (green) and anti-actin (red) antibodies. Cells were transfected with unmodified and modified Sing16 HA mRNA-LNPs over a dose response curve from 1.25–10 µg/million cells. c Quantification % translational inhibition calculated from puromycin signal across the entire lane. Maximum inhibition from Cycloheximide control indicating 100% translational inhibition. Two-way Anova with Šídák’s multiple comparisons test was performed. p > 0.05, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001.

Both unmodified and modified mRNA-LNPs induced antiviral responses in vitro

To determine the effect of UNR and MNR on the global transcriptome, Sing16 HA mRNA delivered as OF-02, cKK-E10 or SM-102 LNPs were used to transfect HSKM cells, and the transcriptome was analyzed at 1, 4, and 24 h post-transfection. Across all LNPs, the transcriptome exhibited a high degree of induced differential response relative to the buffer controls (Fig. 3a, top). The top 50 upregulated genes globally across conditions showed a strong antiviral response signature, inclusive of genes such as the OAS family, MX1, and the IFIT family (Fig. 3a, bottom). To validate that antiviral associated genes were a key axis of response and elucidate other axes of differentiation amongst lipids, gene set variation analysis (GSVA) was performed using MSigDB module C5 (Gene Ontology [GO]) ( ~ 8000 genesets) to calculate enrichment scores across genesets per sample. From this, a principal component analysis (PCA) was calculated using the top variable genesets (n = 1527), where PC1 and PC2 explained 77.2% and 7.7% of the total variance, respectively (Fig. 3b). This PCA based on GSVA-derived enrichment scores shows clear axes of separation of the buffer and SM-102 LNP from cKK-E10 and OF-02 LNPs across PC1 and PC2 (Fig. 3b) confirming previous findings that uptake and protein expression can be dependent on cell type and that SM-102 LNPs may not deliver mRNA efficiently to muscle fiber in a murine model²¹. To assess the signatures driving the clustering, top genesets correlating with the principal components were selected based on the magnitude of the rotation component of the PCA per geneset. PC1-associated signatures showed a clear positive enrichment of processes associated with protein translation, such as the GO protein binding, modification, and metabolic process genesets (Fig. 3b, bottom), while PC2 showed an enrichment for immune-related processes, including the GO cytokine receptor binding, response to virus, cell death, and IL-1 response genesets (Fig. 3b, left). While OF-02 LNP showed a stronger antiviral signature with UNR at the 4 h time-point, with SM-102 this was not observed until 24 h post-transfection, and cKK-E10 showed little differentiation of UNR vs. MNR at any time-point across top PC1/PC2 associated signatures. The difference in gene expression profiles for OF-02 and SM-102 could be a result of differences in cell uptake kinetics and composition variance, ultimately leading to different amounts of mRNA being delivered to the cell. These results show that in addition to the modification of mRNA, the delivery vehicle has an impact on the global transcriptional profile of the cells.

a Waterfall plot showing each gene’s maximum induction across any condition shown by log2 fold change relative to buffer (top), with the top 50 upregulated genes by absolute max log2 fold change highlighted (bottom). b Principal component analysis (PCA) was constructed using enrichment scores of top variable genesets (n = 1517) calculated by gene set variation analysis (GSVA). Samples are shown as dots on the main plot. Waterfall plots are shown adjacent to Principal components 1 (PC1 [bottom]) and 2 (PC2 [left]) of the aggregated GVSA scores for top genesets (named at right).

Modification of mRNA had a moderate impact on functional antibody titers

To evaluate whether the increase in protein expression observed in MNR formulations could result in an increase in functional antibody titers, UNR and m1ψ MNR Sing16 HA-LNPs were evaluated in BALB/c mice with 0.08, 0.4 or 1 μg mRNA-LNP administered intramuscularly in a two-dose regimen, 3 weeks apart. LNPs prepared with ionizable lipids OF-02, cKK-E10 and SM-102 were evaluated for functional antibody responses as measured by hemagglutination inhibition (HAI) titers against the homologous A/Singapore/ INFIMH160019/2016 H3N2 strain (Fig. 4). There were no significant differences in HAI titers between UNR and MNR at all dose levels of OF-02 and SM-102 LNPs. However, significantly higher HAI titers (p 3). While our data showed tendency of higher reactogenicity with unmodified mRNA, consistent with published nonclinical and clinical studies on mRNA-LNP vaccines²², it is important to acknowledge that dsRNA contamination can be a confounding factor.

BALB/c mice (n = 8 per group) were immunized twice intramuscularly (IM), 3 weeks apart with 1, 0.4, and 0.08 μg of different Sing16 HA mRNA-LNP formulations. Serum titers were measured by hemagglutinin inhibition assay (HAI) against homologous A/Singapore/INFIMH160019/2016 H3N2 Influenza virus strain and geometric mean titers were graphed for a Sing16 HA-OF-02, b Sing16 HA-cKK-E10 or c Sing16 HA-SM-102 LNPs.

To further evaluate immunogenicity of the mRNA formulations, a dose range of 15, 45, and 135 μg of UNR and MNR Sing16 HA-cKK-E10 and Sing16 HA-SM-102 were tested in NHPs with two doses administered intramuscularly at 4-week intervals. Robust HAI titers were induced in all treatment groups with titers peaking between Days 42-56 (Fig. 5). Peak antibody responses were recorded at 135 μg for SM-102 LNP (log₁₀ HAI_t / HAI₀ = 2.51) and at 45 μg for cKK-E10 LNP (log₁₀ HAI_t / HAI₀) = 2.26) (Supplementary Fig. 1) and the impact of m1ψ modification was somewhat variable between the two LNPs. For cKK-E10, significantly higher antibody titers were recorded on Days 28 (p p p p 1). Beyond Day 56, antibody titers in the 45 μg and 135 μg cKK-E10 MNR groups were maintained, indicating potential persistence of functional antibodies in NHPs after modified mRNA vaccine-based immunization. For the SM-102 treatment group, titers elicited by MNR formulations were significantly better than UNR in the 135 μg group at Days 14 and 42 (p p = 0.33 and 0.08; logFC = 0.25 and 0.45, respectively in Supplementary Fig. 1). Unlike what was observed for MNR cKK-E10, the titers induced by the 135 μg dose of MNR SM-102 formulations were somewhat lower on Day 90 (p = 0.08; logFC = −0.45) possibly indicating a weaker persistence of antibody titers for MNR SM-102 (Supplementary Fig. 1, Fig. 2). Overall, the benefit of m1ψ modification was less apparent with the SM-102 formulation, where the MNR formulation induced significantly higher neutralizing antibody titers than the UNR only at the 135 μg dose on Days 14 and 42. As a marker of systemic inflammation, we measured C reactive protein (CRP) serum levels in vaccinated NHPs. Supplementary Fig. 4 shows that induction of CRP was transient, peaking at Day 2 and reduced to baseline by Day 7 for all treatments. Importantly, modification of mRNA was not sufficient to reduce induction of CRP following vaccination and empty LNP also increased CRP levels in NHPs.

Cynomolgus macaques (n = 4–6 per group) were immunized at Day 0 and Day 28 by intramuscular route with 135, 45, or 15 μg of different mRNA-LNP formulations encoding for the H3-HA protein from the A/Singapore/INFIMH-16-0019/2016 strain of influenza. Serum titers taken over the course of 1 year were measured by hemagglutinin inhibition assay (HAI) against homologous A/Singapore/INFIMH160019/2016 H3N2 Influenza virus strain and geometric mean titers were graphed for the following mRNA-LNPs: a Sing16 HA-cKK-E10 15 μg, b Sing16 HA-cKK-E10 45 μg, c Sing16 HA-cKK-E10 135 μg, d Sing16 HA-SM-102 15 μg, e Sing16 HA-SM-102 45 μg, f Sing16 HA-SM-102 135 μg.

The impact of mRNA modification on induction of inflammatory cytokines is species-dependent

To better understand the impact of mRNA modification on innate immune responses, serum cytokines and chemokines were assessed in mice (Fig. 6a) and NHPs (Fig. 6b). Mice were immunized with UNR and MNR versions of Sing16 HA delivered with either OF-02, cKK-E10 or SM-102 LNPs while NHPs were immunized with UNR and MNR versions of Sing16 HA delivered with either cKK-E10 or SM-102 LNPs.

a BALB/c mice (n = 8 per group) were immunized twice intramuscularly, 3 weeks apart with 1 μg of either UNR or MNR Sing16 HA-OF-02, Sing16 HA-cKK-E10 or Sing16 HA-SM-102 lipid nanoparticles (LNPs). Mice were bled pre-study and 24 h post-priming. Cytokine analysis was performed using Milliplex mouse cytokine/chemokine magnetic bead panel as described in methods. Cynomolgus macaques (n = 4-6 per group) were immunized at Day 0 and Day 28 by IM route with 135 μg of either UNR or MNR Sing16 HA-cKK-E10 or Sing16 HA-SM-102 LNPs. Blood was taken at 6-days pre-study (baseline), and 2-days post each administration. Samples were analyzed for b serum cytokine levels by Luminex and c whole blood transcriptomic analysis; volcano plots were made for select cytokines/chemokines of interest.

In mice, induction and upregulation of cytokines such as IFNγ, IL-1a, IP10, VEGF, MIP1b and IL-5 was similarly detected in both the UNR and MNR LNP vaccinated mice regardless of the lipid used in the formulation (Fig. 6a). However, OF-02 showed higher upregulation of cytokines, including IL-6, GCSF, IL-2 and IP10 than cKK-E10 and SM-102 LNPs. Overall, RNA modification did not substantially reduce these serum cytokine levels compared to unmodified mRNA. Similar observations were reported by others²³ wherein m1ψ modification of mRNA not only did not impact the serum cytokine profile in mice but also did not impact in vivo protein expression, immunogenicity or induction of liver enzymes alanine transaminase and aspartate transaminase²³. On the contrary, in NHPs, higher cytokine induction was demonstrated in UNR LNP vaccinated animals when compared to the MNR LNP cohort. Cytokines such as IFNγ, IFNα, and MCP1 were higher with UNR LNP compared to MNR LNP immunized animals, while other cytokines, such as TNFα (cKK-E10), IL-1RA and GCSF (SM-102) were still significantly induced in MNR LNP-immunized NHPs (Fig. 6b). An enhanced cytokine induction response after the second vaccine administration was demonstrated in the cKK-E10 cohort, a phenomenon which has been previously observed^24,25. For instance, at the 135 µg dose, UNR cKK-E10 induced a 2-fold increase in IFNγ at Day 2 which was further enhanced following the second administration to an 8-fold increase at Day 30, whereas UNR SM-102 induced a 2-fold increase of IFNγ at both Day 2 and Day 30 following the second administration. Similarly, UNR cKK-E10 induced a 16-fold rise in IFNα at Day 2 which was enhanced by the second administration to 32-fold at Day 30, as compared to UNR SM-102 which induced a 32-fold and 16-fold rise in IFNα at Day 2 and Day 30, respectively.

Transcriptional signatures of mRNA vaccination in NHPs

The sensitivity of existing cytokine assays and their specificity to NHPs could have an impact on our ability to accurately detect subtle changes in sera cytokines. To address these concerns as well as investigate the impact of mRNA modification on the transcriptional signatures of mRNA vaccination, we next performed bulk mRNA sequencing on whole blood drawn from NHPs treated with cKK-E10 or SM-102 LNP formulations made with UNR and MNR formats of Sing16 HA at pre-study and 2 days post primary immunization. Confirmation of the serum cytokine and chemokine findings was done by interrogating differential expression of cytokine and chemokine genes taken at Day 2 compared to pre-immunization. In concurrence with the sera cytokine findings, a significant dose- and LNP-dependent increase in cytokine genes in UNR formulations was observed, which was mostly abrogated by m1ψ-modification (Fig. 6c). This includes the gene encoding for IL1RA (IL1RN) which was observed in NHP sera after immunization (Fig. 6b). A strong dose-dependent upregulation of genes encoding for chemokines such as CCL8 and CXCL10 was also noted, as previously reported to be induced by mRNA-based vaccines²⁴. Comparing the serum cytokines (Fig. 6b) with the cytokine gene expression profile (Fig. 6c), we also observed a trend that the inflammatory cytokine profile of SM-102 appeared less impacted by m1ψ-modification as compared to cKK-E10, highlighting the potential impact of LNP selection as a driver of proinflammatory cytokine responses. Of note, CXCL10 protein was not detectable in the serum, supporting the inclusion of multiple technologies of varying sensitivities and specificities for biomarker discovery work.

As mRNA vaccines are likely to impact many aspects of the immune response beyond cytokine induction, we next assessed the global impact of UNR and MNR formulations on differential gene expression. Administration of mRNA vaccines, regardless of formulation, resulted in a substantial induction of differentially expressed genes in NHPs (Fig. 7). PCA analysis details a dramatic separation between pre- and post-vaccination for unmodified mRNA-LNP formulations, unlike those containing modified mRNA where greater overlap was observed between transcriptional profiles for formulations of both ionizable lipids (Fig. 7a). In general, both LNPs induced similar transcriptome profiles, with UNR being a stronger inducer of differentially expressed genes compared to modified mRNA (Fig. 7b). Although there was a diminished impact on differential gene expression in animals treated with both MNR formulations compared to UNR, it is of note that the magnitude of this impact was LNP-dependent, with more significantly upregulated genes observed in animals immunized with modified mRNA formulated in SM-102 compared to cKK-E10 (509 vs 75 genes, respectively; Fig. 7b). Substantiating the influence of LNP selection on the transcriptional modulatory potential of m1ψ-modification, we conducted a concordance analysis between the gene expression profiles induced by UNR and MNR formulated in SM-102 and cKK-E10 and found the profiles to largely be concordant for SM-102 (r = 0.8659) unlike the relationship observed with cKK-E10 (r = 0.2917; Fig. 7c).

a Principal-component analysis (PCA) of transcriptomes generated prior to and after vaccination (Day 0 and Day 2). The variation in the global gene expression profiles of the top 500 most variable genes is shown. Principal components 1 (PC1) and 2 (PC2) are shown, which represent the greatest biologically relevant variation in gene expression. b Volcano plots of differentially expressed genes (DEGs) comparing Day 2 and Day 0 for each lipid nanoparticle and dose evaluated. DEGs are defined as genes with p c Concordance analysis of logFC between unmodified and modified mRNA in ckk-E10 and SM102 LNPs. Concordant and discordant genes are magenta and cyan, respectively.

Consistent with the observations in HSKM cells treated with UNR and MNR Sing16 HA (Fig. 3b), amongst the top canonical pathways activated in NHPs following mRNA vaccination were genes associated with pattern recognition receptors, interferon signaling and the role of PKR in interferon induction (Fig. 8). Induction of genes associated with PKR signaling and interferon signaling was greatly reduced by mRNA modification, with a larger impact observed when mRNA was formulated in cKK-E10 LNP compared to SM-102. When examining the potential influence of LNP composition on the impact of mRNA modification, we observed a greater reduction in expression of genes associated with the PKR signaling cascades (e.g., OAS1, DDX58, STAT1) and interferon signaling (e.g., MXI, IFI6, IFIT3) in MNR formulated in cKK-E10 than SM-102 (Fig. 8b). Interestingly, genes associated with the interferon response continued to be induced by MNR formulations, irrespective of whether the mRNA was formulated with cKK-E10 or SM-102 LNPs (Fig. 8b). Further experiments were carried out with empty LNP (cKK-E10) to investigate if transcriptional responses seen could be due to the LNP delivery vehicle alone. Our data showed that certain pathways, such as those associated with phagosome formation, are induced following administration of empty LNPs, indicating that the LNP delivery system alone can activate specific innate immune pathways (Supplementary Fig. 5).

a The heatmap depicts overrepresentation of canonical pathways enriched in whole blood of non-human primates following a 135 μg dose of mRNA-LNP (cKK-E10 and SM-102) vaccines. The color of each cell refers to the -log₁₀ (p-value) and indicates the significance of the pathway enrichment (see color key). The color bar below the heatmap denotes the mRNA modification class (UNR: blue; MNR: red). b The heatmap depicts the top canonical pathways enriched following mRNA vaccination. The color of each cell refers to the log2 fold change of the differentially expressed genes (DEGs) composed within the pathway (see color key). Genes with a log2 fold change of more than 1 and a p-value of

To further compare in vitro findings to those observed in NHP, we next assessed genes associated with protein translation and degradation. Figure 9 shows a heatmap of a representative subset of these genes for NHPs vaccinated with 135 μg of Sing16 HA-cKK-E10 or Sing16 HA-SM-102 in UNR or MNR mRNA formats. In concurrence with in vitro findings (Fig. 3b), there was strong upregulation of 2’, 5’-oligoadenylate synthetase (OAS) family genes, including the OAS activated RNASEL, in NHP immunized with UNR formulations, with less pronounced upregulation observed in the MNR-treated group, indicating differences in the IFN-induced antiviral activity between the two mRNA formats. The expression of STAT2 48 h post-immunization revealed the ongoing transcription of IFN-induced genes, suggesting a common intracellular antiviral response to both UNR and MNR mRNA formats. Several genes known to contribute to mRNA degradation were heightened in the UNR group, including XRN1, XRN2, and certain LSM1-7 subunit genes associated with mechanisms leading to cytoplasmic RNA decay²⁶. Additionally, two genes that code for pivotal translation initiation inhibitors EIF2AK2 (a dsRNA-dependent kinase impeding ternary complex turnover)²⁷ and EIF4EBP1 (which competitively hinders the cap binding protein by blocking its eIF4G binding site)²⁸ were upregulated. A more pronounced downregulation was observed in UNR-immunized NHP for various ribosomal proteins of both large and small subunits along with numerous translation initiation factors, including eIF4 components, the gamma subunit of the eIF2 ternary complex, and the alternative ternary complex eIF2A. Our RNA sequencing data indicated that mRNA modification-dependent transcriptional signatures were present in NHP blood, with UNR mRNA-LNPs inducing stronger activation of interferon-stimulated genes and other innate immune pathways compared to MNR mRNA-LNPs. Attenuation of innate immune activation with m1ψ modification was consistent with reduced translational repression. These findings provide strong, physiologically relevant evidence that the differential effects of mRNA modification on translation and immune sensing observed in HSKM cells are recapitulated in vivo in a complex immune milieu.

Heatmap providing a set of genes representative of the translation and mRNA degradation pathways in the whole blood collected from non-human primates 2 days post-immunization. Heatmap color scale represents log2 fold change of Day 2 over pre-immunization (baseline) levels. Hierarchical clustering was performed using the Euclidean distance measurement.