Functionality of capless IRES-containing mRNA vectors with double hairpins at the 5’ end

Previously, we demonstrated that an mRNA vector that lacked a 5’ cap expressed eGFP more efficiently than a canonical (capped) mRNA vector as long as it contained an IRES in the 5’ UTR to initiate translation and triple hairpins at both ends to protect it from exonucleases [20]. To simplify the mRNA vector design, we tested the vector’s expression efficiency if it contained a double, instead of a triple, hairpin at the 5’ end. Schematics of the experimental and control vectors are shown in Fig. 1. In the experimental mRNA vectors the encephalomyocarditis virus (EMCV) IRES was included in the 5’ UTR upstream of the ORF for eGFP to drive its expression. In all experimental vectors, hairpins were included at the 5’ end, whereas the 3’ end differed between vectors and contained either a hairpin or poly(A) as shown in Fig. 1. Vectors that did not have hairpins at the 5’ end or did not contain an IRES for eGFP expression were included as controls.

As expected, the control mRNA vector without hairpins and a poly(A) tail failed to express eGFP even though it contained an IRES (Vector A, Fig. 1). The addition of triple hairpins at both ends (Vector B), rescued eGFP expression. These results are consistent with our previous findings [20]. Using two, instead of three, hairpins at each end, (Vector C) gave equivalent eGFP expression. Each of the double hairpins at the 5’ end was 30 nt (a 13 bp stem with a 4 nt loop) and each of the double hairpins at the 3’ end was 34 nt (a 15 bp stem with a 4 nt loop). If these individual hairpins were shortened to just 16 nt at both ends (a 6 bp stem with a 4 nt loop) eGFP expression was significantly decreased (Vector D). Based on these results, we chose to use the longer (30 nt) hairpins at the 5’ end to study the effect of modifying the 3’ end on vector performance.

Replacing the 3’ double hairpin with a terminal 74 nt poly(A) tail significantly improved eGFP expression (Fig. 1, Vector E). Including a 30 nt poly(A) fragment 10 nt upstream from the terminal 74 nt poly(A) tail (a design used in some mRNA vectors to improve DNA template stability [23]) decreased eGFP expression slightly in MDCK cells (Fig. 1, Vector F). Removing the IRES (Fig. 1, Vector G) or removing the 5’ double hairpins (Fig. 1, Vector H) significantly reduced eGFP expression indicating that these are required components of a functional capless mRNA vector.

An IRES is required for protein expression from both endogenous and exogenous mRNAs harboring 5’ hairpins

The 5’ cap on mRNA is required for canonical translation. When mRNA is transcribed in vivo, it is capped at the 5’ end through the sequential actions of three enzymes, RNA triphosphatase, RNA guanylyltransferase, and guanine-N7 methyltransferase [24]. This is true whether the mRNA is transcribed from the host cell’s own DNA or whether it is transcribed from a DNA plasmid transfected into the cell. To determine whether the 5’ double hairpins would interfere with cap-dependent translation if mRNA was transcribed endogenously from a transfected DNA plasmid, we designed several DNA plasmid vectors for intracellular mRNA transcription and tested them in MDCK cells. All DNA plasmid vectors used for in vivo transcription were driven by a CMV promoter. Some of these DNA plasmids were designed to transcribe mRNAs with an unstructured 5’ end, whereas others were designed to transcribe mRNA that self-folded into a double hairpin at the 5’ end. The sequence of the 5’ double hairpin, where applicable, was the same as those in the in vitro synthesized mRNAs with the larger double hairpins described above (Fig. 1a, Vectors C and G). Schematics of the relevant parts of the DNA plasmids and their projected mRNA transcripts are shown in Fig. 2a. All DNA plasmid templates had the same transcription start site for the initiation of mRNA synthesis and the same SV40 cleavage site for polyadenylation (see Methods for details). The length of the 5’ UTR is predicted to be the same in all mRNA transcripts. Of note, the expectation is that all the mRNA transcribed from these DNA plasmids endogenously, will have a 5’ cap.

Figure 2b shows eGFP expression in MDCK cells 48 h post-transfection with equimolar amounts of each DNA expression plasmid. As expected, eGFP expression was robust in the control DNA Plasmid I. Adding a double hairpin to the 5’ end of the projected transcript (Plasmid II) prevented eGFP expression but adding an IRES upstream of the GFP ORF (Plasmid IV) rescued eGFP expression to about half maximum. Interestingly, eGFP expression was minimal from expression Plasmid III which generated an mRNA transcript that contained both a 5’ cap and an IRES but without 5’ hairpins. This suggested that there was some translation initiation conflict between the 5’ cap and the IRES that was prevented by the hairpins.

To directly compare cap-dependent and IRES-dependent protein expression from the same mRNA transcript, we modified all four DNA plasmids by adding the mCherry ORF and a short 20 bp linker upstream of the eGFP ORF in expression Plasmids I and II (to make Plasmids V and VI) or upstream of the IRES in expression Plasmids III and IV (to make Plasmids VII or VIII). In these new plasmids, mCherry expression required a functional cap, whereas eGFP expression required a functional IRES.

As shown in Fig. 2b, the first gene, mCherry, was only expressed when the transcribed mRNA did not have a 5’ double hairpin (Plasmids V and VII); in contrast, the second gene, eGFP, was expressed regardless of whether the 5’ end contained double hairpins or a cap as long as a functional IRES was present immediately upstream (Plasmids VII and VIII). This confirmed that the failure to express protein from plasmids that contained 5’ hairpins but lacked an upstream IRES (Plasmids II (eGFP), VI (mCherry), and VIII (mCherry)) was not due to transcript degradation, but rather a failure of translation.

What was not clear was whether the failure to initiate translation from the 5’ end of mRNA was due to a failure to cap a transcript that had likely already folded into a hairpin, or whether it was due to an inability of the 5’ cap located at the base of the hairpin to bind to the ribosome to initiate translation. To clarify if double hairpins in the 5’ terminus can directly inhibit expression from capped mRNA, we designed two DNA templates for in vitro mRNA synthesis with similar sequences as the projected mRNAs transcribed in vivo from DNA expression Plasmids VII and VIII (Fig. 2). The only differences between the in vivo and in vitro transcribed mRNAs were the sequences at the end of the 3’UTR and the length of the poly(A) tail (74 nt for the in vitro synthesized mRNAs, probably longer in the in vivo synthesized mRNAs). Using ARCA when synthesizing mRNA Vectors 1C and 2C (Fig. 3a) ensured that the transcript was capped regardless of its 5’ structure (i.e., regardless of whether it had double hairpins or not). We then compared eGFP and mCherry expression in MDCK cells 12 h after they were transfected with equimolar concentrations of mRNA.

Fig. 3: Double hairpins at the 5’ end interfere with cap-dependent translation of mCherry.

a mRNA vector schematics; © in vectors 1C and 2C indicates that the transcript has been capped using an Anti-Reverse Cap Analog (ARCA); A and B: individual hairpin sequences that comprise the double hairpin in mRNA vectors 2 and 2C. b eGFP and mCherry expression in MDCK cells 12 h after mRNA transfection with equimolar amounts of each vector (n = 4 independent experiments).

Cells transfected with uncapped mRNA lacking 5’ double hairpins (Fig. 3; mRNA 1) did not express mCherry, whereas cells transfected with similar mRNA which was capped (Fig. 3, mRNA 1C) did. However, if the transcribed mRNA had a double hairpin at the 5’ terminus, mCherry was not expressed regardless of whether it was capped (Fig. 3, mRNA 2C) or not (Fig. 3, mRNA 2). This indicated that the presence of the 5’ double hairpin directly interfered with the cap-dependent translation of mCherry. This explains the absence of eGFP expression in cells transfected with DNA expression Plasmid II and mCherry expression in cells transfected with DNA expression Plasmids VI and VIII in Fig. 2 and mRNA transcript 2C in Fig. 3. In contrast, eGFP, whose translation was initiated from an IRES in all the in vitro synthesized transcripts shown in Fig. 3, was expressed from all four mRNAs, although in varying levels.

Effect of polyadenylation and dephosphorylation of capless mRNA vectors on HA expression in vitro and anti-HA antibody induction in vivo

We next tested if a capless mRNA protected by a double hairpin at the 5’ end could express a membrane antigen at sufficiently high levels to be used as a vaccine. The antigen we chose was hemagglutinin (HA), the Influenza A surface protein responsible for virus binding to epithelial cell surface receptors [25] whose expression, endoplasmic reticulum membrane translocation, and proper folding includes additional steps beyond that required for eGFP translation [26]. In several of the mRNA vectors described in Fig. 1, we replaced eGFP with the full-length Influenza A HA. Because we planned to test the best-performing mRNA vector as an intramuscular vaccine in BALB/c mice, we transfected the mRNA vectors into myocytes and partially merged myotubes (MC) isolated from BALB/c mice in addition to the MDCK cells used in the previous experiments.

In contrast to our results with eGFP (Fig. 1), an mRNA vector with double hairpins on both ends (Fig. 4b, Vector C(HA)) expressed only a minimal amount of HA 12 h post-transfection in both MDCK cells and mouse myocytes (MC); its expression levels were barely above those detected in cells transfected with a vector with no hairpins at all (Fig. 4b, Vector A(AH)). However, replacing the 3’ hairpin with a 74 nt poly(A) tail markedly improved HA expression in both cell types (Fig. 4b, Vector E(HA)). Replacing the 3’ double terminal hairpin with a longer, segmented poly(A) tail (30 nt + 74 nt) did not improve HA expression over a vector with the single, shorter (74 nt) poly(A) configuration (Fig. 4b, Vectors E(HA) and F(HA)).

Fig. 4: Capless IRES-initiated mRNA vectors with double hairpins at the 5’ end require a poly(A) tail for HA expression in MDCK cells and mouse myoblasts and induce anti-HA antibody production when delivered as a vaccine.

a mRNA vector schematics. b HA expression in MDCK cells and mouse myoblasts (MB) 12 h after transfection with equimolar concentrations of the indicated mRNA vectors. ELISA data are normalized to the signal after transfection with vector E(HA) separately for MDCK cells and myocytes. (n = 4 independent experiments done at separate times, *P < 0.05). c HA expression in MDCK cells 12 h after transfection with equimolar concentrations of phosphatase-treated or -untreated mRNA vectors. ELISA data are normalized to the signal after transfection with the untreated mRNA vector E(HA). (n = 4 independent experiments done at separate times, *P < 0.05). d Anti-HA IgG titers three weeks after the last vaccination in individual mice immunized with a HA-mRNA vaccine that contained either the E(HA) or E(HA)-ph mRNA vectors. Control: data for non-immunized mice. For all comparisons, blood was collected 7 days before the first immunization and 21 days after the third immunization.

Capping of conventional mRNA synthesized in vitro may not be 100% efficient leaving a triphosphate at the 5’ end. This triphosphate can be recognized by RIG-I and leads to immune activation in cells [27]. Therefore, prior to use, some synthetic mRNAs are treated with phosphatases to remove the 5’ triphosphate to prevent this immunological response [27]. In capless mRNA, the 5’ end is buried and thus it is not clear whether the terminal triphosphate can trigger this immune response. To determine if dephosphorylation of the triphosphate on the 5’ end of the double-hairpin vector had any impact on protein expression in vitro, we treated all mRNAs with alkaline phosphatase prior to transfection (A(HA)-ph, C(HA)-ph, E(HA)-ph, F(HA)-ph). Dephosphorylation of A(AH), C(AH), E(HA), and F(AH) mRNA had no significant impact on HA expression in MDCK cells (Fig. 4c).

We then tested the best-expressing in vitro mRNA capless vector, (E(HA)), in mice to determine whether it was able to induce an antibody response against HA. 1.5 µg of mRNA were complexed with pre-formed lipid-based nanoparticles and delivered to mice using an in vivo-jetRNA+ transfection reagent (HA-mRNA vaccine). Three immunizations were performed per animal on days 0, 10, and 21 (Fig. 5a). Antibodies against HA were measured 1 week prior to, and 42 days after, the initial immunization. Vaccination with this capless double hairpin HA-mRNA vaccine generated significant HA antibody titers in all mice. Removing the 5’ triphosphate on this vector using phosphatase (E(HA)-ph) had no significant impact on anti-HA titers in mice (Fig. 4d).

Fig. 5: Immunization with a vaccine that includes a capless mRNA that has double hairpins at the 5’ end and encodes the EMCV IRES linked to the HA ORF protects mice against a lethal dose of Influenza A.

a Timeline of the experiment. b Schematic of the mRNA vectors used in the vaccines. c Anti-HA antibody titer 7 days before the first vaccination and then 21 days after the last vaccination (42 days after the first vaccination). d Survival data for vaccinated groups for the three weeks after challenge with the influenza A virus; no animals died after 3 weeks. e Weight loss in vaccinated mice after Influenza A (A/PR/8/34) virus infection ((50 × LD50) (LD50 = 95 PFU/mouse) for the first three weeks after challenge. f Clinical scores assigned for eGFP-mRNA vaccinated mice for the first ten days after a live Influenza A virus challenge. g Clinical score assigned for HA-mRNA vaccinated mice for the first ten days after a live Influenza A virus challenge. The following clinical scores were assigned: 0 = normal, 1 = slightly ruffled, 2 = ruffled fur, 3 = ruffled fur and inactive, 4 = hunched/moribund, and 5 = dead.

Vaccination with a capless IRES-containing mRNA that expresses HA protects mice against a lethal dose of Influenza A

We then tested whether a capless mRNA vector encoding the full-length Influenza A surface protein hemagglutinin (HA) linked to an EMCV IRES and having a double hairpin structure on its 5’ end (Fig. 4) could protect BALB/c mice against Influenza A. Because the mRNA Vector E(HA) generated the highest anti-HA titer in the previous experiment, we selected it for use in these experiments (Fig. 5B); mRNA vector E, expressing eGFP, was used as a negative control. 1.5 µg of mRNA were complexed with pre-formed lipid-based nanoparticles and delivered to mice using an in vivo-jetRNA+ transfection reagent. Three immunizations were performed per animal on days 0, 10, and 21 (Fig. 5a). The mRNA vectors used in the vaccine formulation are shown in Fig. 5b. Anti-HA antibody titers were measured in sera collected from each mouse 7 days before and 42 days after the first immunization (Fig. 5c). Seven out of eight mice injected with the HA-mRNA vaccine, but none of the mice injected with the control eGFP-mRNA vaccine showed high anti-HA antibody titers 21 day after the last immunization (Fig. 5c).

We then exposed mice to 50 times the lethal dose of live Influenza A virus (strain A/PR/8/34). While all the mice immunized with the control eGFP-mRNA vaccine died, only one of the eight mice vaccinated with the capless mRNA vaccine encoding HA died (Fig. 5d); the sole vaccinated mouse that died was the one that did not generate a high anti-HA titer following vaccination (Fig. 5c). In addition, most of the HA-mRNA vaccinated mice had no weight loss (Fig. 5e) and demonstrated minimal clinical impact after challenge (Fig. 5f, g).