Cytidine acetylation yields a hypoinflammatory synthetic messenger RNA

Summary

Synthetic messenger RNA (mRNA) is an emerging therapeutic platform with important applications in oncology and infectious disease. Effective mRNA medicines must be translated by the ribosome but not trigger a strong nucleic acid-mediated immune response. To expand the medicinal chemistry toolbox for these agents, here we report the properties of the naturally occurring nucleobase N⁴-acetylcytidine (ac4C) in synthetic mRNAs. We find that ac4C is compatible with, but does not enhance, protein production in the context of synthetic mRNA reporters. However, replacement of cytidine with ac4C diminishes inflammatory gene expression in immune cells caused by synthetic mRNAs. Chemoproteomic capture indicates ac4C alters the protein interactome of synthetic mRNAs, reducing binding to cytidine-binding proteins and an immune sensor. Overall, our studies illustrate the unique ability of ac4C to modulate RNA-protein interactions and provide a foundation for using N⁴-cytidine acylation to fine-tune the properties of nucleic acid therapeutics.

eTOC Blurb

Messenger RNA (mRNA) therapeutic platforms benefit from nucleotide components that enable protein translation while evading immune activation. Here Nance et al. report the exploration of N4-acetylcytidine (ac4C) as a medicinal chemistry element in a synthetic mRNA.

Graphical Abstract

Introduction

Synthetic messenger RNA (mRNA) is an emerging therapeutic modality with a wide range of clinical applications in cancer, genetic disorders, and infectious disease (Pardi et al., 2018; Pastor et al., 2018). The power and portability of these platforms is perhaps most well-illustrated by their rapid deployment as vectors for the delivery of vaccines directed against SARS-CoV-2, the viral pathogen responsible for the COVID-19 pandemic (Corbett et al., 2020; Walsh et al., 2020). These synthetic mRNAs are comprised of four key molecular elements: i) a 5’ cap, which confers stability; ii) 5’ and 3’-untranslated regions (UTRs), important for translation initiation and stability; ii) a coding sequence (CDS), that determines the identity of the protein to be expressed; and iv) modified nucleobases, which can enhance an mRNA’s protein production and immune evasion (Figure 1a) (Weng et al., 2020). Depending on the intended application, activation of the innate immune system by a synthetic mRNA may be either desirable (e.g. for a vaccine) or detrimental (e.g. for a therapy requiring repeated dosing), emphasizing the importance of defining structure-function relationships in this area (Loomis et al., 2016).

Figure 1.

Design of a synthetic mRNA panel. (a) Components of synthetic mRNA. (b) Modified nucleobases, caps, and UTRs as graphical abbreviations used to describe mRNA structure. (c) eGFP reporter mRNAs applied to assess the effects of ac4C. This mRNA panel allows comparison of cytidine, ac4C, 5fC, and Ψ in the context of the following non-coding elements: 1-4: 5’-cap, Kozak 5’-UTR. 5-8: 5’-cap, EMCV IRES 5’-UTR. 9-10: uncapped, Kozak 5’-UTR. 11-12: uncapped, EMCV IRES 5’-UTR.

Substantial prior work has evaluated the impact of modified nucleosides on synthetic mRNA activity, revealing context-dependent effects. For example, early studies found universal incorporation of pseudouridine (Ψ) and related analogues can increase protein production and decrease immune stimulation within certain synthetic mRNAs (Karikó et al., 2005, 2008; Kormann et al., 2011). More recently, a systematic study of modified uridine analogues in over >100 mRNAs found modified uridines can cause higher, lower, or unchanged levels of protein expression compared to canonical uridine depending on the protein encoded and synonymous codon sequence used (Mauger et al., 2019). These examples illustrate a paradigm in which discovery research first defines the potential capabilities of modified nucleosides, providing a suite of chemical options that subsequent ‘medicinal chemistry-like’ efforts to optimize individual therapeutic mRNAs can draw upon.

Many of the nucleoside analogues that have been explored in synthetic RNAs are derived from naturally occurring modifications (Karikó et al., 2005; Pastor et al., 2018). Compared to uridine, cytidine modifications have been less well-explored in synthetic mRNAs, with the majority of studies focusing on 5-methylcytidine (mC) and its analogues (Kormann et al., 2011). Considering modifications to expand the chemical diversity of synthetic mRNA, we were intrigued by N⁴-acetylcytidine (ac4C; Figure 1b). This RNA modification is universally conserved amongst all domains of life, suggesting the evolutionary potential for either immune recognition or tolerance (Ikeuchi et al., 2008; Sas-Chen et al., 2020). N⁴-acetylcytidine triphosphate (ac4CTP) is compatible with conventional in vitro transcription protocols used to prepare synthetic mRNAs (Sinclair et al., 2017). Finally, a recent study demonstrated that ac4C-containing luciferase reporter mRNAs can be translated by human cells (Arango et al., 2018). However, the quantitative interplay of ac4C with other mRNA elements (e.g. 5’-cap and UTR), as well as the immune regulatory potential of this modification, remains to be determined for any discrete coding sequence.

Results

Design of a panel of ac4C reporter mRNAs

To explore ac4C as a functional element in synthetic mRNA, we designed a set of reporter constructs to determine how this modification quantitatively influences protein production (Figure 1c). We chose enhanced green fluorescent protein (eGFP) as an optimal reporter molecule, as immunoblotting could provide a facile readout of overall protein levels, while flow cytometry could allow complementary detection of folded eGFP at the single cell level. Next we considered the influence of the 5’-cap. Previous studies have found that modified nucleobases such as N⁶-methyladenosine (m6A) can reduce the necessity of 5’-end capping for translation and stability of synthetic mRNA reporters (Meyer et al., 2015). To extend these insights to ac4C, we envisioned testing constructs with and without an enzymatically introduced 5’-N⁷-methylguanosine (m⁷G-ppp-Nm) linkage. The 5’-UTR located between the cap and the ORF is another element which demonstrates sensitivity to RNA modifications (Thess et al., 2015). Interestingly, a literature survey found several well-cited papers which have explored the effects of modified nucleobases on synthetic mRNA function do not report or specify 5’-UTR sequence, suggesting a need for clearer structure-activity data (Karikó et al., 2008; Kormann et al., 2011; Mauger et al., 2019). Here we focused on two UTRs, a minimal 5’-GCCGCCACC-3’ Kozak sequence, and an internal ribosome entry site derived from encephalomyocarditis virus (EMCV IRES). We hypothesized that comparing the minimal Kozak UTR and the highly structured EMCV IRES may provide insight into whether ac4C can impact functional UTR structures. Of note, previous studies have found that universal substitution of ac4C for cytidine can increase RNA duplex stability (Sas-Chen et al., 2020). Finally, we sought to put ac4C’s effects in a broader context by comparing them to structurally and functionally related modifications. 5-formylcytidine (5fC), an oxidized analogue of 5mC, is known to form an exocyclic hydrogen bonding network (Wang et al., 2016). This network is analogous to the weak C-H-O interaction (Gu et al., 1999) made between the acetate on ac4C’s cytidine ring in its highly preferred anti-conformation (Parthasarathy et al., 1978; Sas-Chen et al., 2020). To provide additional context we also included pseudouridine, which constitutes the uridine analogue which has been most widely explored in synthetic mRNAs (Karikó et al., 2008; Mauger et al., 2019).

Chemoenzymatic synthesis of ac4C eGFP reporter mRNAs

Based on these design considerations, we synthesized eGFP mRNAs 1–12 (Figure 1c) using in vitro transcription (IVT). RNA modifications were universally incorporated by substitution of base-modified nucleotide triphosphates (NTPs) for the corresponding canonical NTP in IVT reactions (Sinclair et al., 2017). One critical consideration for our studies is that cytidine acetylation is a labile modification and can hydrolyze over time (Thomas et al., 2018). This led us to use freshly synthesized ac4CTP in all relevant IVT reactions. The effectiveness of this strategy was verified by digesting and dephosphorylating synthetic mRNAs to their constitutive mononucleosides followed by LC-MS, which observed identical incorporation of ac4C and the non-labile cytidine analogue 5fC into mRNA (Figure S1). To prepare them for biological evaluation, crude mRNA products were purified, subjected to enzymatic capping and poly(A)-tailing, and re-purified to furnish synthetic mRNAs in sufficient yields for downstream applications. Analysis by denaturing gel electrophoresis found that base-modified and unmodified synthetic mRNAs were intact and migrated as expected based on their sizes (Figure S1).

Defining the influence of ac4C on protein production from synthetic mRNAs

Next, the activity of these synthetic mRNAs was assessed by monitoring production of eGFP in HEK-293T cells (Figure 2). Pilot studies found complexation of synthetic mRNAs with TransIT lipid-polymer transfection reagent provided optimal transfection efficiency and no observable toxicity in this cell line. All samples were analyzed 24 h post-transfection, which is identical to the primary time point chosen in previous studies (Karikó et al., 2008; Mauger et al., 2019). First, we examined how ac4C impacted protein production from synthetic mRNAs containing different 5’-UTRs (1-8). Synthetic mRNAs containing ac4C and a minimal Kozak sequence were found to produce similar, albeit slightly reduced, levels of eGFP relative to those containing the cytidine nucleobase (Figure 2b-d). Northern blot confirmed uptake of canonical and ac4C mRNAs (Figure S2). Pseudouridine had effects akin to ac4C. In contrast, 5-fC-containing RNA failed to produce detectable protein in any setting. This suggests that even though conformationally-related, ac4C and 5-fC have extremely disparate effects when incorporated into the coding sequence of synthetic mRNA. Examining the IRES 5’-UTR mRNAs (5-8), we found for all constructs overall protein production was diminished relative to the Kozak sequence-containing analogues. Cytidine-containing mRNAs were once again found to most efficiently produce eGFP protein (Figure 2e-f); however, in the context of IRES mRNAs ac4C (6) was well-tolerated, while pseudouridine (8) was not. While this was initially surprising given the widespread use of pseudouridine and related analogues in synthetic mRNAs, a literature search found a prior study that had previously noted the poor compatibility of pseudouridine with IRES elements, corroborating our observation (Thess et al., 2015). Next, we explored the relative translation of “capless” variants of the eGFP mRNA (9-12). Here we focused our analysis only on ac4C and C, given the poor activity of 5-fC and prior studies of pseudouridine, which had been previously explored in a cap-independent synthetic mRNA reporter and shown minimal effects (Meyer et al., 2015). As expected, non-inclusion of the 5’-cap severely impeded eGFP expression mRNAs relative to their 5’-m⁷G-ppp-Nm containing analogues. However, in this weakly expressing context, ac4C outperformed the canonical cytidine nucleobase in both Kozak (Fig. 2g) and IRES-containing sequences (Figure S1). Overall, these studies illustrate the compatibility of ac4C with protein expression from synthetic mRNAs and offer new contexts where it may be explored for advantage relative to other nucleotides.

Figure 2.

Analysis of eGFP protein production from synthetic ac4C mRNAs. (a) Experimental schematic. (b) Western blot analysis of eGFP protein production by mRNAs 1-4 (1 μg/mL, 24 h). Representative data of three biological replicates. (c) Fluorescence microscopy analysis of eGFP protein production by mRNAs 1-2 (1 μg/mL, 24 h). Scale bar corresponds to 400 μm. (d) Flow cytometry analysis of eGFP protein production by mRNAs 1-4 (1 μg/mL, 24 h). Histograms are color coded according to the graphical abbreviations in Figure 1b. Representative data of three biological replicates. Replicate data may be found in Figure S1. (e) Western blot analysis of eGFP protein production by mRNAs 5-8. Data are representative of three biological replicates. (f) Flow cytometry analysis of eGFP protein production by mRNAs 5-8 (1 μg/mL, 24 h). Representative data of three biological replicates. Replicate data may be found in Figure S1 (g) Flow cytometry analysis of eGFP protein production by mRNAs 9-10 (1 μg/mL, 24 h). Representative data of three biological replicates. Replicate data may be found in Figure S1.

Defining the influence of ac4C on synthetic mRNA immune activation

Modified nucleobases often show altered interaction with the innate immune system. Immune evasion by synthetic mRNAs may be particularly important for clinical applications which necessitate repeated dosing, such as protein replacement therapy. However, whether ac4C incorporation affects the immune response to synthetic mRNA is unknown. Since HEK-293T cells lack many of the receptors involved in innate immunity we explored this phenomenon in human monocyte-derived dendritic cells (MoDCs) derived from patient donor specimens. We reasoned understanding ac4C’s effects in this model may prove especially valuable, given the key role DCs play in linking the innate and adaptive immune response. Accordingly, MoDC’s isolated from three patients were transfected with two capped Kozak eGFP mRNAs containing either cytidine or ac4C and assayed for inflammatory cytokine production (Fig. 3a). An additional set of cells was treated with TransIT transfection reagent (without mRNA) as a control. Assessment of cytokine secretion (TNF-ɑ) at 24, 48, and 72 h revealed that each set of patient-derived MoDCs displayed a unique absolute level of response to the synthetic mRNAs, but highly reproducible trends (Fig. 3b, Fig. S6). Cytidine-containing eGFP mRNA provoked the greatest response, stimulating TNF-ɑ secretion as much as 1000-fold relative to vehicle control. Substituting cytidine with ac4C drastically decreased this stimulation, in some cases reducing cytokine secretion to near background levels. Analogous results were observed across other cytokines (IL-1β, IL-10) and within different MoDCs sets (Figure S3); in all cases, the combination of cap and ac4C served to limit the inflammatory response that was observed when cells were treated with cytidine-containing synthetic mRNAs. Analysis of transfected MoDCs by flow cytometry revealed similar, albeit low, levels eGFP protein expression with both cytidine and ac4C mRNAs (Figure S3), suggesting the discrepant immune response was not due to differences in mRNA uptake.

Figure 3.

Analysis of immune activation by synthetic ac4C mRNAs. (a) Experimental schematic. (b) Analysis of TNF-α secretion after treatment of MoDCs with cytidine or ac4C-containing mRNAs (24 h, 1 μg/mL). Data are graphed as mean ± SEM, n = 3. (c) Nanostring analysis of inflammatory gene expression after treatment of MoDCs with cytidine or ac4C-containing mRNAs 1 and 2 (24 h, 1 μg/mL). Data are representative of biological duplicates. Output of nSolver analysis was Z-score transformed for visualization using Heatmapper. (d) Treatment of HEK-293T cells expressing the human dsRNA sensor TLR3 with eGFP mRNAs (36 h) does not induce an NF-κB reporter. Poly(I:C) = dsRNA control. Data are graphed as mean ± SEM, n = 3.

To further characterize this phenomenon, we compared the effects of cytidine and ac4C mRNAs on the expression of more than 200 inflammation-associated genes in MoDCs using a quantitative, digital gene expression assay (Nanostring). Consistent with our analysis of cytokine secretion, incorporation of ac4C in mRNA strongly diminished, but did not abolish, the ability of synthetic eGFP mRNA to induce inflammatory genes downstream of NF-κB and interferon signaling (Figure 3c, Figure S3, Table S1). The NF-κB pathway appeared more strongly influenced by N⁴-acetylation. This analysis also enabled the identification of genes such as HLA-DRB1 and HLA-DRA, responsible for immune presentation, whose expression was decreased by the canonical mRNA but maintained after incorporation of ac4C. Overall these studies demonstrate the potential of ac4C as a novel approach to chemically modulate the immune signature of synthetic mRNAs.

Previous studies have observed pseudouridine and 5-mC can limit activation of TLR3, an endosomal double-stranded RNA sensor protein. To test whether a similar mechanism was responsible for the effects of ac4C, we transfected TLR3-overexpressing HEK-293T cells with synthetic mRNAs and analyzed production of a secreted alkaline phosphatase (SEAP) reporter through the canonical NF-κB (p50-p65) pathway (Kaiser et al., 2004; Warren et al., 2010). Double-stranded poly(I:C) RNA capably triggered TLR3-dependent SEAP expression. However, neither cytidine nor ac4C mRNAs increased the activity of the TLR3-driven SEAP reporter, even at five-fold higher concentrations (Fig. 3d). This result suggests the reduced immunogenicity of ac4C does not only reflect altered interaction with TLR3 and raises the question as to whether this modification can disrupt the interaction of RNA with other immune receptors.

ac4C can disrupt the interaction of RNA with immunoregulatory proteins

Nucleic acid immunogenicity is attributed to the ability of these species to physically interact with pattern recognition receptors. In addition to TLRs, other key proteins shown to play a role in the immune response to synthetic mRNAs include PKR, OAS-L, and RIG-I (Kim et al., 2004; Li et al., 2011; Okamoto et al., 2009), some of which exhibit differential interaction with pseudouridine. To better understand the potential for cytidine acetylation to alter the immunoregulatory interactions of synthetic mRNAs, we employed a chemoproteomic profiling approach. Briefly, T4 RNA ligase was used to introduce a single biotinylated nucleotide at the 3’-end of either the cytidine or ac4C eGFP mRNA. Successful incorporation of biotin was validated by dot blot (Figure S4). Chemoproteomic probes were then immobilized on pre-blocked streptavidin resin and incubated with extracts from HEK-293T cells. Following capture, washing, and tryptic digest proteomes were analyzed by multidimensional protein identification technology (MudPIT) LC-MS/MS and quantified using distributed normalized spectral abundance factors (dNSAF), a label-free analysis method that normalizes peptide spectral counts for protein size. An approximately equal number of peptide spectra were identified in each experiment, suggesting the biotinylated RNA probes exhibit roughly similar stability in cell extracts (Table S2). To increase the stringency of our analysis, we narrowed our focus to 370 proteins which displayed i) more than 1 detectable peptide and ≥10 spectral counts in either cytidine or ac4C pulldown, ii) no detection in the background (bead) control, and iii) less than 10 spectral counts in a community database of proteomic background experiments (CRAP-ome). Of proteins meeting these criteria, 18 were selectively enriched (≥10-fold) by the ac4C mRNA, while only 5 were selectively repelled (≥10-fold) (Table S2). Amongst this latter group, only one immune receptor’s binding was lessened by ac4C incorporation: the cytosolic nucleic acid sensor PKR.

PKR is primarily considered to be activated by double-stranded RNA, with single-stranded RNA playing a less-explored role. However, our eGFP mRNAs did not activate TLR3, an immune receptor that recognizes RNA duplexes >90 nucleotides in length (Jelinek et al., 2011; Karikó et al., 2004). This led us to consider whether the differential protein interactions of our two eGFP mRNAs may reflect differences in their secondary structure. To assess this possibility, we probed the coding sequence of each RNA using selective 2’-hydroxyl acylation analyzed by primer-extension mutational profiling (SHAPE-Map). In this approach RNAs are treated with the electrophile 1-methyl-7-nitroisatoic anhydride (1M7), which causes an increased mutational rate at all four nucleotides when analyzed by next-generation sequencing. Higher error rates at specific positions are associated with greater levels of 1M7 modification and correlate with decreased secondary structure. One caveat to this approach is that the basal error rate at ‘C’ nucleotides in untreated samples was ~10-fold higher for the ac4C mRNA as compared to the cytidine mRNA (Fig. S8a). However, the overall level of mutations induced by 1M7 were highly similar between ac4C and C mRNAs, suggesting the method maintains utility even with this higher background (Fig. S8b). Applying this method to the eGFP mRNAs revealed a balanced effect of ac4C on secondary structure. Rather than exerting a unidirectional global change, we found ac4C increased the structural content of some segments, decreased others, and left many unchanged (Fig. 4c). Functional assays confirmed cytidine - but not ac4C -- mRNAs were able to induce PKR phosphorylation in HEK-293T cells (Fig. 4d). Of note, this cell line does not express high levels of TLR3 and similar effects were not caused by poly(I:C), both of which indicate that the differential induction we observe is not generically attributable to double-stranded RNA content. These studies suggest one way in which ac4C may decrease PKR activation is by altering the duplex character of synthetic mRNAs.

Figure 4.

Effects of ac4C on synthetic mRNA-protein interactions. (a) Structure of capped, 5’-Kozak, 3’-biotinylated eGFP mRNA capture probes containing C (top) or ac4C (bottom). (b) Differential enrichment of proteins by C and ac4C capture probes in LC-MS/MS experiments. Proteins showing a >10-fold enrichment by either C (left) or ac4C (right) are colored blue. Proteins identified exclusively in the ac4C sample are in the top right. (c) SHAPE-MaP analysis of ~900-nt eGFP coding sequence. Colors indicate normalized reactivity with 1M7 as assessed by 1M7-dependent misincorporation over a 30-bp sliding window for C and ac4C mRNAs. (d) Synthetic mRNAs containing cytidine, but not ac4C, cause functional activation of the immune sensor PKR (phospho-PKR, top) in HEK-293T cells (4 h). This increased phosphorylation is detected despite lower overall levels of PKR in cytidine mRNA-treated samples relative to control. (e) Differential enrichment of RNA-binding proteins (RBPs, gray), ribonucleotide cytidine-binding proteins (rCBPs, orange), and immune sensors (immune, red) in LC-MS/MS experiments. Proteins below the diagonal were preferentially enriched by C-containing mRNA. (f) ac4C has the potential to disrupt mRNA-MBNL1 interactions by occupying the Hoogsteen edge (red arc), where it may block hydrogen bonding and sterically clash with proximal the proximal amino acid Glu183.

Finally, we considered the influence of ac4C on single-stranded RNA binding proteins. Several proteins whose capture was disrupted by ac4C are known to functionally interact with single-stranded RNA, including PKR which can functionally bind both double and single-stranded sequences (Mayo and Cole, 2017; Toroney et al., 2012). Decreased interaction with the former may contribute to its resistance to activation by ac4C-containing mRNA. To get a global view of this phenomenon we performed a ‘filter-free’ analysis of our proteomic data, focusing on RNA cytidine-binding proteins (e.g. PCBPs, PTBPs, HNRNPK, termed ‘rCBPs,’ orange), immune receptors (red) and a generic collection of RNA binding proteins (e.g. HNRNPs RBMs, termed ‘RBPs,’ gray). Interestingly, 8/9 rCBPs were preferentially captured (≥2-fold) by the cytidine RNA, while only 7/54 generic RBMs displayed a similar preference (Fig. 4e). This suggests the potential of ac4C to broadly disrupt the interaction of single-stranded RNA-binding proteins with cytidine-containing sequences. To explore this notion in greater detail, we examined the structure of MBNL1, an RNA cytidine-binding protein that had been previously crystallized in the presence of a 5’-CGCUCU-3’ RNA ligand (Fig. 4f) (Teplova and Patel, 2008) and whose capture appeared to be disrupted by ac4C (Fig. 4e). The N⁴-amino group of cytidine in the native ligand is well-positioned to make two hydrogen bonds with MBNL1, only one of which could be maintained in the presence of ac4C. In addition, the strong thermodynamic preference of ac4C for a conformer in which the oxygen of the acetyl carbonyl lies proximal to cytidine’s C⁵ C-H group would be expected to selectively project steric bulk that may clash with proximal amino acids towards the nucleotide’s Hoogsteen edge (red arc, Fig. 4f) (Parthasarathy et al., 1978; Sas-Chen et al., 2020). The N⁴-amino group of cytidine has also been observed to mediate important interactions in crystal structures of HNRNPK (Backe et al., 2005) and ZAP1 (Meagher et al., 2019), both of which interact more strongly with cytidine-containing mRNA in our chemoproteomic experiments. Overall, our data suggest the unique ability of ac4C to alter RNA-protein interactions may underlie its non-inflammatory properties and other biological effects.

Discussion

Here we have reported the synthesis and characterization of a panel of synthetic mRNAs containing the modified nucleobase ac4C. Our data indicate that within the context of the eGFP coding sequence universal incorporation of ac4C is well-tolerated, functionally producing slightly lower levels of protein than cytidine. This is in stark contrast to the conformationally related nucleobase analogue 5fC, which does not appear to be functional as a cytidine surrogate. We also identified two contexts in which ac4C can be advantageous from the perspective of protein production relative to other nucleobases: IRES-containing mRNAs (ac4C preferable to Ψ), and capless mRNAs (ac4C preferable to C). Compared to its subtle effects on protein production, cytidine acetylation had a more profound impact on immune stimulation, with incorporation of ac4C dampening cytokine secretion by MoDCs up to 100-fold relative to cytidine-containing mRNAs. Follow-up studies suggested a role for cytidine acetylation in reducing activation of PKR. This may be due to the ability of ac4C to alter RNA-protein interactions, a property supported by our chemoproteomic analyses, and precedented by previous studies of pseudouridine. While uridine is the most well-characterized nucleotide determinant of RNA immune activation (Freund et al., 2019), prior studies have also suggested a role for cytidine (Durbin et al., 2016; Karikó et al., 2005; Sugiyama et al., 2005; Takata et al., 2017; Warren et al., 2010) and lend credence to the notion that modified nucleotides may influence PKR activation (Nallagatla and Bevilacqua, 2008). Interestingly, previous studies have suggested that replacement of less than 10% of all uridine residues with pseudouridine can significantly limit immune activation by synthetic mRNAs (Karikó et al., 2005), and in the future it may be interesting to ask whether substoichiometric ac4C can exert similar effects. Importantly, our results also do not rule out differential activation of TLRs, including TLR3, in MoDCs. The full scope and mechanism by which cytidine acetylation dampens immune signaling is unknown and will be an important subject of future research. Finally, an important limitation of our study is that it explores only a single coding sequence. Like other modified nucleobases, the utility of ac4C in any mRNA of interest will require empirical determination, and is likely to vary based on sequence, length, and codon composition (Mauger et al., 2019). This is conceptually analogous to medicinal chemistry, in which new functional groups provide fodder for ligand optimization, whose potency must be experimentally determined via synthesis and assay. Our studies suggest ac4C may be a worthwhile addition to the ‘medicinal chemistry’ toolbox used to optimize synthetic mRNAs. The modified nucleotide triphosphate that enables the construction of ac4C RNAs is now commercially available, which should facilitate such efforts. Looking forward, we envision methods enabling introduction of ac4C into discrete elements (e.g. UTR, CDS) or at specific positions of RNA may allow exploration of this modification in functional nucleic acids requiring more precise medicinal chemistry, such as those used in RNA silencing and gene editing, and also provide insights into the physiological function of cytidine acetylation. Such studies are currently underway and will be reported in due course.

STAR★METHODS

RESOURCE AVALIABILITY

Lead contact

Requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Jordan L. Meier (jordan.meier@nih.gov).

Materials availability

All unique reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Data and code availability

Mass spectrometry proteomics data has been deposited to the MassIVE (MSV000086841) and Proteome Xchange (PXD024145) repositories. Original mass spectrometry data may also be accessed after publication from the Stowers Original Data Repository at https://www.stowers.org/research/publications/LIBPB-1599. Review access can be obtained using the following location and password information:

Location: ftp://MSV000086841@massive.ucsd.edu

Password: MeierRNA2021

This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Culture

HEK293T (female) cells were obtained from ATCC. 293/hTLR3 and 293 null cells were obtained from Invivogen. All cells were cultured at 37 °C under a 5% CO₂ atmosphere. HEK293T cells were cultured in a growth medium of DMEM, 4.5 g/l glucose, 10% (v/v) fetal bovine serum (FBS), and 2 mM L-glutamine. 293/hTLR3 and 293 null cells were cultured in a growth medium of DMEM, 4.5 g/l glucose, 10% (v/v) 50 U/ml penicillin, 50 mg/ml streptomycin, 100 mg/ml Normocin^™, 2 mM L-glutamine, and 10 μg/mL Blasticidin. All cell lines were routinely tested for mycoplasma contamination.

Human peripheral blood cell samples used for generation of MoDCs were obtained from anonymous healthy donors (Transfusion Medicine Department, Clinical Center, National Institutes of Health, Bethesda, MD) with an approved human subject agreement, and were isolated from the leukopack by Ficoll-Hypaque density gradient centrifugation. Subsequently, monocytes were separated by MACS sorter using CD14 monocyte isolation kits (Miltenyi Biotech). The isolated monocytes were cultured at 7.0 × 10⁵ cells/mL in RPMI-1640 medium (Mediatech Inc) supplemented with 10% FBS (GemCell), 2 mM L-glutamine (Lonza), 25 mM HEPES (Quality Biological), 100 U/ml penicillin (Lonza), 100 μg/mL streptomycin (Lonza), and 50 μM of 2-mercaptoethanol (Sigma) containing 50 ng/mL hGM-CSF and 50 ng/ml hIL-4 at 37 °C in a humidified CO₂ (5%) incubator for 3 days. On day 3, 50% of the culture medium was replaced with prewarmed fresh complete RPMI-1640 medium containing the same concentration of hGM-CSF and hIL-4. On the 5th day of culture, monocyte derived dendritic cell (MoDC) suspensions were harvested as immature MoDCs and cultured at 7.0 × 10⁵ cells/mL for subsequent experiments.

METHOD DETAILS

Design of eGFP reporter constructs

Τ7 eGFP reporter constructs were designed with Τ7 promoter, Kozak (GCCGCCACCATG) or ΙRES (Bochkov and Palmenberg, 2006) 5’UTRs, eGFP CDS (Cormack et al., 1996), Xenopus β-globin 3’UTR (Wang et al., 2008), and NheI restriction site as synthetic DNA cloned into a pTwist high-copy cloning vector (Twist Bioscience, San Francisco, CA). Prior to transcription, plasmid DNA was linearized and gel purified to ensure proper transcription.

Preparation of eGFP reporter mRNAs

In vitro transcription of linearized eGFP constructs was performed with the HiScribe^™ T7 High Yield RNA Synthesis Kit (New England Biolabs, Ipswtich, MA), according to the manufacturer’s protocol using 0.2 units/μL T7 RNAP, 100 ng DNA template, and 100 mM NTPs. For ac4C- and 5fC-containing transcripts, CTP was replaced in the reaction mixture with ac4CTP or 5fCTP (50 mM) respectively. For pseudouridine-containing transcripts, UTP was replaced with ΨTP (50 mM). After transcription, reactions were treated with TurboDNase (ThermoFisher Scientific) to remove template DNA and precipitated using 2.5M LiCl at −20 °C for 1 hour. After washing pellets with 70% ethanol, mRNA was resuspended in nuclease-free water and RNA concentration was determined by measuring the optical density at 260 nm. Next, mRNAs were appended with poly(A) tails enzymatically using E. coli Poly(A) Polymerase (New England Biolabs, Ipswtich, MA) and the provided protocol. RNAs were again purified by LiCl precipitation before enzymatic capping using the Vaccinia Capping System and 2’OMe transferase (New England Biolabs) and provided protocols to give mRNAs a cap 1 structure. After a final LiCl precipitation, mRNAs were analyzed for integrity and proper tailing by loading 2 μg of mRNA on a 1% agarose gel alongside uncapped/untailed controls (Fig. S2).

Mass spectrometry analysis of nucleoside modified mRNAs

Mass spectrometry analysis of C, ac4C, and 5fC incorporation into IVT mRNAs was assessed following nuclease digestion. Briefly, in vitro transcribed mRNAs were treated with nuclease P1 (1U/10 μg RNA, #N8630, Sigma) in 50 μL of buffer containing 100 mM ammonium acetate [pH 5.5] for 16 h at 37 °C. Samples were heated at 60 °C for 5 minutes, returned to 37 °C for an additional hour and 5 μL of 1 M ammonium bicarbonate [pH 8.3] and Bacterial Alkaline Phosphatase (0.5U/10 μg RNA, #18011–015, ThermoFisher) were added. Samples were further incubated at 37 °C for 2 h, adjusted to 150 μL with nuclease-free water and filtered to remove enzymatic constituents (Amicon Ultra 3K, #UFC500396). Following lyophilization, samples were reconstituted in 10 μL RNase-free water and analyzed via LC-MS/MS using a reverse phase chromatography (Shimadzu LC-20AD) coupled to a triple-quadrupole mass spectrometer (Thermo TSQ-ultra) operated in positive electrospray ionization mode. Quantification was accomplished by monitoring nucleoside-to-base ion transitions and generating standard curves for each nucleoside using the stable isotope dilution internal standardization method. (Fig. S1)

Analysis of GFP expression in HEK-293T cells

Prior to mRNA transfection, eGFP mRNAs were quantified using Qubit^™ RNA BR Assay Kit and diluted to a concentration of 1 μg/uL in nuclease-free water. HEK-293T cells were seeded in a 12-well plate (500,000 cells/well) and left to adhere for 24 h prior to transfection. Briefly, in vitro transcribed eGFP mRNAs were complexed with TransIT-mRNA transfection reagents in OPTI-MEM serum-free medium according to manufacturer’s protocol, then gently added to wells. At 24 h post-transfection, cells were collected for both western blotting and flow analysis. First, cells were analyzed for functional GFP fluorescence by flow cytometry using a FACS Canto II flow cytometer (Becton Dickinson) run by BD FACSDiva software. 20,000 cells were measured per condition, and non-transfected HEK-293T cells were used as a negative control for GFP gating. Data were analyzed using FlowJo software (Becton Dickinson). Second, cells were lysed for western blot analysis of total eGFP expression. Cell lysates were prepared by resuspending cell pellets in ice-cold PBS containing protease inhibitor cocktail (EDTA-free, Cell Signaling Technology #5871S). Samples were then lysed by sonication using a 100 W QSonica XL2000 sonicator (3 × 1 s pulse, amplitude 1, 60 s resting on ice between pulses). Lysates were pelleted by centrifugation (20,817 r.c.f. × 30 minutes, 4 °C) and quantified using the Qubit 4.0 Fluorometer and Qubit Protein Assay Kit. Equal amounts of protein (15 μg) were loaded on 4%–12% Bis-Tris gels, separated using NuPAGE MES SDS running buffer (ThermoFisher Scientific) and transferred onto 0.45 μM nitrocellulose membranes using 1X NuPAGE^™ Transfer Buffer.

Membranes were blocked with StartingBlock^™ (PBS) Blocking Buffer for 30 min prior to incubating with primary anti-GFP antibody (1:1000 in blocking buffer) overnight at 4 °C. After washing 3 times with 1X Tris-Buffered Saline, 0.1% Tween^® 20 Detergent (TBST), membranes were incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit IgG (1:1000 dilution, Cell Signaling Technology) secondary antibody. Western blots were visualized for chemiluminescence using the LumiGLO^® Peroxidase Chemiluminescent Substrate and chemiluminescence was detected using the ImageQuant^™ LAS 4000 biomolecular imager (Cytiva) and quantified by densitometry using ImageQuant TL 8.2 image analysis software (Cytiva). To assess proper gel loading, membranes were stripped using OneMinute^® Western Blot Stripping Buffer, and reprobed with anti-β-Tubulin primary antibody (1:1000, Cell Signaling Technology).

Northern blot analysis of transfected eGFP mRNAs

Total RNA was isolated from HEK-293T transfected with eGFP mRNAs 24 h post-transfection using Trizol (Life Technologies) and the quality of the RNA was checked on a Bioanalyzer 2100 (Agilent) using the RNA 6000 Nano Kit (Agilent). For each sample, 8 μg total RNA was electroporated on 1.5% agarose–formaldehyde gel, and then transferred onto an Hybond-N membrane (Amersham Pharmacia Biotech) overnight. After transfer, the membrane was UV crosslinked and then stained with methylene blue to visualize the 18S and 28S rRNA using a ChemiDoc^™ Touch Imaging System (Bio-Rad). The blot was hybridized with ³²P-labeled eGFP DNA probe (5’-GGGTCTTGTAGTTGCCGTCGTCCTTGAAGA-3’) at 50 °C overnight. After washing with 2X SSC with 0.1% SDS buffer 3 × 15 min at 50 °C, the membrane was exposed to an Imaging Screen-K (Bio-Rad) for 20 min. Images were then analyzed by Typhoon Trio+ variable Mode Imager (GE Healthcare).

Cytokine expression analysis of mRNA-transfected immature MoDCs

Human immature MoDC were seeded in 48-well plates (700,000 cells/mL) prior to transfection with eGFP mRNAs using TransIT mRNA transfection reagent. At the given time point (24, 48, and 72 h), culture supernatants were removed and analyzed for cytokine secretion using Meso Scale Discovery V-Plex human biomarker assay kits using manufacturer’s protocol. Samples were kept frozen until use and were thawed on ice and sampled at a dilution factor of 1:2. Each test plate included addition of 3 (QC) controls, made using spiked samples that were prepared in-house and which had a predetermined target value. The measured QC values for each cytokine were within 25% of “expected” on plates having the test samples.

NanoString gene expression analysis

Human immature MoDCs transfected with eGFP reporter mRNAs were harvested and total RNA was extracted using Trizol^™ using the manufacturer’s protocol. RNA quality was assessed using TapeStation RNA ScreenTape (Agilent, Santa Clara, CA, USA). RNA (200 ng) was assessed for gene expression using the nCounter Human Inflammation v2 Panel (NanoString Technologies, Seattle, WA, USA) according to the nCounter XT CodeSet Gene Expression assay protocol. Resulting RCC files were examined with NanoString nSolver software for quality control analysis prior to analysis of gene expression.

hTLR3 NFkB-inducible SEAP assay

293/hTLR3 and 293-null cell were seeded at 30,000 cells/well in a poly-D-lysine coated 96-well black flat bottom plate and cultured 24 h prior to transfection with pNiFty-SEAP NF-κB-inducible secreted embryonic alkaline phosphatase (SEAP) reporter plasmid (Invivogen) using FuGENE^® 6 Transfection reagent (Promega) according to the manufacturer’s protocol. Cells were incubated at 37 °C for 24 hours before medium was exchanged with HEK-Blue^™ SEAP Detection medium (Invivogen) and SEAP expression was induced by transfecting cells with eGFP mRNAs (0.5–5 μg/mL) or positive control poly(I:C) (1 μg/mL) as a positive control. NF-κB-induced SEAP expression was monitored by reading the optical density (OD) at 620 nm over 36 hours.

Biotinylation of eGFP reporter mRNAs

In vitro transcribed eGFP mRNAs were 3’-biotinylated using the Pierce^™ RNA 3’ End Desthiobiotinylation Kit (ThermoFisher Scientific) according to the manufacturer’s protocol. Briefly, 50 pmol of mRNAs were heated 5 min at 85 °C to denature before kit components were added and the ligation reaction was incubated at 16 °C overnight. Reactions were diluted with nuclease-free water before chloroform:isoamyl alcohol extraction to remove RNA ligase. RNAs were then precipitated using 5M NaCl, glycogen, and ice-cold 100% ethanol. RNA pellets were washed with ice-cold 70% ethanol, then air dried and resuspended in 20 μL of nuclease-free water. To assess whether mRNAs were evenly biotinylated, dot-blot analysis was performed (Fig. S8). RNAs were diluted to 5 nM stocks with nuclease-free water and 2 μL of each RNA was spotted onto an Amersham Hybond-N+ membrane (GE Healthcare). Membrane was crosslinked twice for 30 s with 150 mJ/cm² in a UV 254 nm Stratalinker 2400 (Stratagene, San Diego, CA). After crosslinking, samples were blocked with 5% non-fat milk in PBS [pH 7.2] at room temperature for 30 min, and then probed with anti-biotin HRP-linked antibody (1:1000, Cell Signaling Technologies) in 5% non-fat milk in PBS at room temperature for 1 h. Membranes were washed four times with PBST and developed with the LumiGLO^® Peroxidase Chemiluminescent Substrate and chemiluminescence was detected using the ImageQuant^™ LAS 4000 biomolecular imager (Cytiva)

RNA-protein pulldown assay

Whole-cell lysates were prepared from HEK293T cells by resuspending cell pellets in ice-cold PBS containing protease inhibitor cocktail (EDTA-free, Cell Signaling Technology). Samples were then lysed by sonication using a 100 W QSonica XL2000 sonicator (3 × 1 s pulse, amplitude 1, 60 s resting on ice between pulses). Lysates were pelleted by centrifugation (20,817 r.c.f. x 30 minutes, 4 °C) and quantified using the Qubit 4.0 Fluorometer and Qubit Protein Assay Kit. Lysates were diluted to a concentration of 3 mg/mL in lysis buffer and stored at −80 °C until the time of assay.

RNA-protein pulldown assay was performed with Pierce^™ Magnetic RNA-Protein Pull-Down Kit according to the manufacturer’s protocol. First, magnetic streptavidin beads were added to tubes and washed twice with 20mM Tris [pH 7.5] before incubating with 50 pmol of biotinylated RNAs in 1X RNA Capture Buffer for 30 minutes at room temperature. After RNA binding, beads were again washed twice with 20mM Tris [pH 7.5], washed once with 1X Protein-RNA Binding buffer, then incubated with cell lysate solution (100 μg protein, 15% glycerol, 1X Protein-RNA Binding Buffer) for 1 h at 4 °C. After protein capture, beads were washed 3 times with wash buffer, then proteins were eluted by incubating beads with 50 μL of elution buffer for 30 minutes at 37 °C. Supernatant was removed, and proteins were precipitated with ice-cold acetone. Protein pellets were dried and stored at −80 °C for subsequent proteomic analysis.

MudPIT analysis of C or ac4C mRNA pulldown

Acetone precipitated protein samples that had been pulled down using capped C or Ac mRNA probes were analyzed by Multidimensional Protein Identification Technology (MudPIT), as described previously (Florens and Washburn, 2006; Washburn et al., 2001). Briefly, protein pellets were resuspended in 100mM Tris [pH 8.0] containing 8M urea. Disulfide bonds were reduced by the addition of tris(2-carboxyethyl)phosphine (TCEP) to a final concentration of 5mM and samples were incubated at room temperature for 30 min. Proteins were then digested overnight with trypsin at 37 °C with shaking at 800 RPM. Digestion reactions were quenched by the addition of formic acid to a final concentration of 5%. After trypsin digestion, peptide mixtures were pressure-loaded onto 100 μm fused silica microcapillary columns packed first with 10 cm of C18 reversed phase (RP) material (Aqua; Phenomenex), followed by 3 cm of 5-μm Strong Cation Exchange material (Luna; Phenomenex), followed by 2 cm of 5-μm C18 RP resin. The loaded microcapillary columns were placed in-line with a 1260 Quartenary HPLC (Agilent). The application of a 2.5 kV distal voltage electrosprayed the eluting peptides directly into Orbitrap-Velos Elite hybrid mass spectrometers (Thermo Scientific) equipped with a custom-made nano-LC electrospray ionization source. Full MS spectra were recorded on the eluting peptides over a 400 to 1600 m/z range in the Orbitrap at 60K resolution, followed by fragmentation in the ion trap (at 35% collision energy) on the first to fifteenth most intense ions selected from the full MS spectrum. Dynamic exclusion was enabled for 90 sec (Zhang et al., 2009). Mass spectrometer scan functions and HPLC solvent gradients were controlled by the XCalibur data system (Thermo Scientific).

RAW files were extracted into .ms2 file format (McDonald et al., 2004) using RawDistiller v. 1.0, in-house developed software (Zhang et al., 2011). RawDistiller D(g, 6) settings were used to abstract MS1 scan profiles by Gaussian fitting and to implement dynamic offline lock mass using six background polydimethylcyclosiloxane ions as internal calibrants (Zhang et al., 2011). MS/MS spectra were searched using ProLuCID (Xu et al., 2015) with a peptide mass tolerance of 10 ppm and 500 ppm for fragment ions. Trypsin specificity was imposed on both ends of candidate peptides during the search against a protein database combining 44, 166 human proteins (NCBI 2019–11-03 release), as well as 426 usual contaminants such as human keratins, IgGs and proteolytic enzymes. To estimate false discovery rates (FDR), each protein sequence was randomized (keeping the same amino acid composition and length) and the resulting “shuffled” sequences were added to the database, for a total search space of 89, 111 amino acid sequences. A mass of +15.9949 Da was differentially added to methionine residues.

DTASelect v.1.9 (Tabb et al., 2002) was used to select and sort peptide/spectrum matches (PSMs) passing the following criteria set: PSMs were only retained if they had a DeltCn of at least 0.08; minimum XCorr values of 1.8 for singly-, 2.1 for doubly-, and 2.3 for triply-charged spectra; peptides had to be at least 7 amino acids long. Results from each sample were merged and compared using CONTRAST (Tabb et al., 2002). Proteins that were subsets of others were removed using the parsimony option in DTASelect on the proteins detected after merging all runs. Proteins that were identified by the same set of peptides (including at least one peptide unique to such protein group to distinguish between isoforms) were grouped together, and one accession number was arbitrarily considered as representative of each protein group. NSAF7 (Zhang et al., 2010) was used to create the final reports on all detected peptides and non-redundant proteins identified across the different runs. Spectral and protein level FDRs were, on average, 0.20 ± 0.07% and 2.0 ± 1.4%, respectively. NSAF7 was used to create the protein table in Supplemental Table 2.

Induction of PKR phosphorylation by eGFP mRNAs

Prior to eGFP mRNA transfection, PKR expression was stimulated in HEK-293T cells by incubating in complete medium supplemented with 1000 U/ml interferon-αA/D (PBL Assay Science). Cells were seeded into 6-well plates (1 × 10⁶ cells/well) 24 hours prior to transfection with eGFP mRNAs. After transfection, cells were collected and lysed by sonication. Equal amounts of lysates (15 μg) were separated by SDS-PAGE prior to transfer onto nitrocellulose membranes. Membranes were blocked prior to incubation with either anti-PKR (1:1000, SCBT) or anti-pPKR (1:1000, Abcam) antibodies at 4 °C overnight. Membranes were washed with 1X TBST prior to incubation with secondary anti-rabbit IgG (1:1000, CST) antibody in 5% milk/TBST for 1 hour at room temperature. Membranes were visualized for chemiluminescence prior to stripping and reprobing with anti-GAPDH (1:1000, CST) antibody for loading control. Membranes were quantified and normalized by densitometry using ImageQuant TL 8.2 image analysis software.

SHAPE-MaP of eGFP reporter mRNAs

A previously published randomer workflow approach was used to carry out the SHAPE-MaP experiment (Smola et al., 2015). Briefly, 1 μg of in vitro transcribed C- and ac4C-containing eGFP RNAs were denatured at 95 °C for 2 minutes. After denaturation, RNAs were folded in 111 mM HEPES, pH 8.0, 111 mM NaCl and 11 mM MgCl₂ for 20 minutes at 37 °C. RNAs were then incubated with 10 mM 1-methyl-7-nitroisatoic anhydride (1M7) for 75 s at 37 °C. Controls with DMSO only (no SHAPE reagent) and SHAPE-modified fully denatured RNA (DC reaction) were prepared in parallel. After SHAPE modification, RNA was purified and used for the reverse transcription reaction with random primers. The cDNA products from the reactions and controls were purified using G-25 columns and converted to dsDNA using NEBNext^® Ultra^™ II Non-Directional RNA Second Strand Synthesis Module. Then, the sequencing library was prepared by tagging with adapters and multiplexing indices for Illumina MiSeq (2 × 150 cycles). Raw sequencing data was analyzed using the ShapeMapper software (Siegfried et al., 2014). SHAPE-Map reactivity data are illustrated in Figure 4c using a sliding window (median SHAPE) approach based on normalized reactivity (Mauger et al., 2019; Watts et al., 2009).

QUANTIFICATION AND STATISTICAL ANALYSIS

The number of replicates, fold-change, mean +/− standard deviation (SD), and p-values are reported in the figures and legends. Experiments were performed with three biological replicates from at least two independent experiments when possible. LC-MS experiments were performed using an n=3, with foldenrichments and −log10FDR calculated by analyzing dNSAF values using the statistical framework QPROT as described in STAR Methods. Statistical analysis of non-proteomic data (i.e. t-tests) were performed with Graphpad Prism software.

Supplementary Material

2

Table S1: Nanostring gene expression data for MoDCs treated with 1, 2, or reagent control. Related to Figure 3.

3

Table S2: LC-MS/MS data for proteins captured by biotinylated 1 and 2. Related to Figure 4.

4

Table S3: SHAPE-MaP analysis of eGFP CDS for 1 and 2. Related to Figure 4.

Significance

Modified RNA nucleobases play a central role in synthetic messenger RNA platforms applied for the prevention and treatment of disease. However, relatively few RNA modifications are known that can be incorporated into mRNA by in vitro transcription, enable protein production, and enhance immune evasion. Here we demonstrate that synthetic mRNAs containing the naturally-occurring RNA modification ac4C produce functional eGFP protein but reduce inflammatory gene expression normally induced by mRNA in patient-derived dendritic cells. Furthermore, we show how profiling the protein interaction landscape of synthetic mRNAs can lend valuable insights into their modification-dependent mechanistic effects. Our studies highlight ac4C as a unique medicinal chemistry element in synthetic nucleic acids that may prove valuable for tailoring their protein interactions and overall activity.

Highlights

• N4-acetylcytidine (ac4C) incorporated into synthetic eGFP messenger RNAs (mRNA).

• Synthetic ac4C mRNAs produce comparable eGFP as modified/unmodified mRNAs.

• Synthetic ac4C mRNAs are less inflammatory in immune cells than cytidine mRNAs.

• Cytidine acetylation alters the protein interaction landscape of synthetic mRNAs.