All presented compounds were obtained according to synthetic procedures described in detail in Supplementary Material.

A PCR product encompassing the firefly luciferase coding sequence under the T7 RNA polymerase promoter was used as the template for the IVT reaction. Transcription buffer, 25 ng/μL dsDNA template, 0.5 mM ATP/CTP/UTP, 0.1 mM GTP, 0.5 mM dinucleotide analogue of cap (cap:GTP molar ratio was 5:1), 0.5 U/µL Ribolock ribonuclease inhibitor and 1 U/µL RNA polymerase T7 (Thermo Fisher Scientific) were added to the reaction mix (Strenkowska et al., 2016). The transcription reaction was incubated for 1 h at 37°C, followed by the addition of 0.025 U/µL DNaseI (Thermo Scientific) and incubation for 20 min at 37°C. The prepared transcripts were purified using NucleoSpin^® RNA Clean-Up (Macherey-Nagel) according to the manufacturer’s instructions. The integrity of the transcripts was tested on a non-denaturing 1% agarose gel and the concentration was measured spectrophotometrically.

The expression level of luciferase mRNA bearing the synthesized cap analogues was measured in the Flexi Rabbit Reticulocyte Lysate System, Promega, (Strenkowska et al., 2016). The reaction mixture contained 40% Flexi RRL lysate, 0.01 mM amino acid mixture, 0.9 mM magnesium acetate (1.8 mM endogenous magnesium concentration in the lysate) and 190 mM potassium acetate. Samples were preincubated for 60 min at 30°C before the addition of mRNA. The translation reaction was carried out under the same conditions for another 60 min. The activity of the synthesized luciferase was measured using a Glomax luminometer.

The inhibitory activities of the new cap analogues were tested using Flexi Rabbit Reticulocyte Lysate, Promega, under conditions specified for cap-dependent translation (Kowalska et al., 2009). Briefly, the reaction mixture was pre-incubated for 60 min at 30°C, followed by the addition of a mixture of ARCA-mRNA encoding the firefly luciferase and the cap analogue under study. Samples were incubated for further 60 min at 30°C. Luciferase activity was measured using a Glomax luminometer (Promega). IC₅₀ values were determined by non-linear regression analysis of the experimental data using GraphPad Prism 8. IC₅₀ values are mean ± SD from at least three independent replicates.

The annealing product of two complementary primers: 5ʹ CAG​TAA​TAC​GAC​TCA​CTA​TAGGGA​AGC​GGG​CAT​GCG​GCC​AGC​CAT​AGC​CGA​TCA 3ʹ and 5ʹ TGA​TCG​GCT​ATG​GCT​GGC​CGC​ATG​CCC​GCT​TCCCTAT​AGT​GAG​TCG​TAT​TAC​TG 3ʹ was used as a template for short RNA synthesis. Thus obtained dsDNA contained the sequence of T7 promoter and G at position +1 (bold). Capping was performed co-transcriptionally during IVT. Transcripts were purified on Oligo Clean-Up and Concentration columns (Norgen Biotek) followed by DNAzyme trimming and subsequent purification (Grzela et al., 2022). 100 ng of analogue-capped RNA was digested with 2.5 µM hNudt16 and loaded onto 15% polyacrylamide/7 M urea TBE gel as described previously (Grzela et al., 2022). To assess both capping efficiency and the amount of decapped product, a densitometric analysis with the use of Image Lab software (Bio-Rad) was applied. Decapping was calculated as the per cent loss in the capped band after addition of the enzyme. The data represent the means ± SD from three experiments.

All statistical analyses were performed with GraphPad Prism 8 software. One-way ANOVA with post hoc Turkey test was used. Statistical significance in p-value is denoted in asterixes ****p < 0.0001, ***p < 0.001, and **p < 0.01 and the data are shown as mean ± SD.

Our current research concerns the synthesis and biological studies of three groups of N2-modified cap analogues. The first group consists of dinucleotides modified only within the exocyclic amino group of the m⁷G moiety (Figure 1A) with a modified triazole ring attached to N2 position via aliphatic linkers of different lengths. The second group consists of dinucleotide analogues modified both at the N2 position of 7-methylguanosine with aromatic substituents and within the 5ʹ-5ʹ bridge in the form of tetraphosphate or β-thiophosphate (Figures 1B, C). The last group consists of trinucleotides containing modified m⁷G moiety in the N2 position with aromatic substituents (benzyl, p-chlorobenzyl and isoxazole) and 2ʹ-O-methylated adenosine as the first transcribed nucleotide (characteristic for cap1) connected via 3ʹ,5ʹ-phosphodiester bond to guanosine (R²m⁷GpppA_mpG shown in Figure 1D).

The synthesis of N2-modified dinucleotide cap analogues has already been described (Piecyk et al., 2015; Grzela et al., 2022). Therefore, compounds 1-3 were obtained in an analogous procedure, with compound 1 synthesized again as a reference for biological studies.

Tetraphosphate analogues (4–5) were obtained using a method for the synthesis of a simpler analogue such as m⁷GppppN (Rydzik et al., 2009). This method requires the preparation of two 5ʹ-diphosphate subunits (GDP and m⁷GDP), one of which must be converted to an imidazole derivative using Mukaiyama method (Mukaiyama and Hashimoto, 1972) to increase the susceptibility to nucleophilic attack of the second subunit in the coupling reaction with a ZnCl₂ catalyst under anhydrous conditions (Kadokura et al., 1997). In the course of our research, N2-substituted 7-methylguanosine 5ʹ diphosphate (7) (R²m⁷GDP) and an imidazole derivative of guanosine 5ʹ diphosphate (im-GDP), were synthesized and subjected to a coupling reaction to obtain a series of tetraphosphate N2-modified dinucleotides.

Compound (6) was obtained analogous to the procedure reported for the synthesis of m⁷Gpp_spG (Kowalska et al., 2008). In the first step, 5ʹ monophosphate of N2-benzyl, 7-methylguanosine (bn²m⁷GMP), was obtained and converted to its imidazole derivative (im-bn²m⁷GMP). Thus prepared compound was subjected to a coupling reaction using triethylammonium thiophosphate in the presence of ZnCl₂ catalyst to obtain the diphosphate derivative bn²m⁷Gpp_S, which was further subjected to the coupling reaction with an imidazole derivative of GMP yielding the product in the form of two diastereoisomers. Since attempts to separate each isomer were unsuccessful, a mixture of diastereoisomers was used for biological experiments.

Trinucleotide analogues of cap (7–9) were obtained in a multistep synthesis, where in the final reaction, the two nucleotide subunits of im-R²m⁷GDP and pA_mpG were linked together in a coupling reaction analogous to the one used for preparation of dinucleotides. The im-R²m⁷GDP subunit was prepared according to previously described methods. The pA_mpG dinucleotide was synthesized in three steps using the amidophosphite method (Senthilvelan et al., 2021). In the first step, coupling reactions of the commercially available 5ʹ-O-dimetoxytrityl-N-benzoyl-2ʹ-O-methyladenosine 3ʹ-cyanoethyl phosphoramidite with N2-iso butyryl-2ʹ,3ʹ-isopropylideneguanosine were carried out according to the protocol (Eisenführ et al., 2003). The coupling reactions were conducted in acetonitrile and the presence of 5-(benzylthio)-1H-tetrazolium as activator while the oxidation was carried out using iodine to obtain the appropriately protected A_mpG dinucleotide. In a subsequent step, the protecting group was removed from the 5ʹ-OH position using dichloroacetic acid while the ketal protection was removed using trifluoroacetic acid. Thus the indirectly deprotected dinucleotide was phosphorylated at 5ʹ-OH position by the Yoshikawa method (Yoshikawa et al., 1969), and finally deprotection was performed using an aqueous ammonia solution to obtain the desired dinucleotide pA_mpG.

All dinucleotides and trinucleotides were isolated from the reaction mixtures and purified by ion-exchange chromatography described in Methods (Supplementary Material). The structure and homogeneity of each compound was confirmed by HPLC, high-resolution mass spectrometry with positive electrospray ionisation (HRMS-ESI) and ¹H and ³¹P NMR.

The protein biosynthesis process depends on the delivery of the information encoded on the mRNA to ribosomes by the eIF4E protein. The efficiency of this process is related to the affinity of eIF4E to the cap structure located at the 5ʹ end of the mRNA. The addition of a free cap analogue with high affinity for eIF4E impairs protein synthesis. Therefore, new compounds are routinely tested for their ability to inhibit translation. For this purpose, we used an extracellular translation system from rabbit reticulocytes (RRL). We added the ARCA-bearing mRNA encoding firefly luciferase to an RRL extract containing all factors necessary for protein synthesis and measured the bioluminescence level. To the subsequent sample, we added increasing amounts of newly synthesized cap analogues and observed changes in bioluminescence relative to the control sample without free cap analogue. Compounds bn²m⁷GppppG (4) and (4-Cl-bn)²m⁷GppppG (5) turned out to be the most effective inhibitors, with IC₅₀ of 0.43 ± 0.16 and 0.60 ± 0.29, respectively (Table 1; Figure 2A). These compounds are tetraphosphates of the derivatives described previously (Grzela et al., 2022), that have shown strong inhibition of protein biosynthesis. Here, an even stronger inhibitory effect was observed. This can be explained by the presence of an additional phosphate in the triphosphate chain that enhance the interaction between cap and eIF4E.

Compound bn²m⁷Gpp_spG (6) appeared as potent translational inhibitor as bn²m⁷GppppG (4) with an IC₅₀ of 0.45 ± 0.05 (Table 1; Figure 2A). For the trinucleotide analogues (bn²m⁷GpppA_mpG (7), (4-Cl-bn)²m⁷GpppA_mpG (8), (4-bn-isx)²m⁷GpppA_mpG (9)), the IC₅₀ values ranged from 1.38 ± 0.59 to 2.66 ± 1.0, while for compounds with longer linkers [(4-(diOCH₃-bn)-tz-CH₂)²m⁷GpppG (1), (4-(diOCH₃-bn)-tz-(CH₂)₂)²m⁷GpppG (2), (4-(diOCH₃-bn)-tz-(CH₂)₄)²m⁷GpppG (3)] from 0.85 ± 0.30 to 2.61 ± 1.0 (Table 1; Figure 2A). Although these scores are higher, still the studied compounds represent more potent inhibitors of translation than m⁷GppppG.

An extracellular translation system from RRL was used to assess the translational properties of mRNAs containing newly synthesized cap analogues. Analogue-capped mRNAs encoding luciferase were added to the extracts and protein activity was measured. The highest levels of luciferase synthesis were provided by mRNAs with trinucleotide derivatives bn²m⁷GpppA_mpG (7), (4-Cl-bn)²m⁷GpppA_mpG (8), and (4-bn-isx)²m⁷GpppA_mpG (9). The translation level of mRNA capped with these derivatives was 3.31, 3.25 and 3.72 times higher than that of m⁷GpppG-mRNA, respectively (Table 1; Figure 2B).

Transcripts bearing derivatives with a tetraphosphate bridge, bn²m⁷GppppG (4) and (4-Cl-bn)²m⁷GppppG (5), showed lower translational properties compared to their initial compounds bn²m⁷GpppG and (4-Cl-bn)²m⁷GpppG (Table 1). RNA capped with compounds that differ in linker length (4-(diOCH₃-bn)-tz-CH₂)²m⁷GpppG (1), (4-(diOCH₃-bn)-tz-(CH₂)₂)²m⁷GpppG (2), (4-(diOCH₃-bn)-tz-(CH₂)₄)²m⁷GpppG (3) show very similar level of translation, e.g., around 2.2–2.6 times higher than observed for m⁷GpppG-RNA (Table 1). RNA bearing cap analogue with a thiophosphate modification bn²m⁷GppspG (6) showed 2.6 times higher translation than m⁷GpppG-RNA (Table 1; Figure 2B).

To assess the incorporation of cap analogues, we used short RNAs that, after synthesis, were visualized on a UREA-PAGE gel. Such gels have a high capacity to separate products that differ in length by one nucleotide. The mRNA chain without the cap analogue has a length of 24 nt and is denoted by NC in Figure 3. When the cap analogue is attached, the short RNA migrates higher depending on whether a dinucleotide (25 nt) or a trinucleotide (26 nt) is attached. Residual products are also visible on the gel, which is due to the properties of the RNA polymerase, which has the ability to fall off the template before the product is completed or to further attach nucleotides off-template. Control caps: GpppG, m⁷GpppG, and ARCA showed expected incorporation efficiencies (Grzela et al., 2022) of 94.57%, 70.60% and 61.87%, respectively (Figure 3; Table 1). The trinucleotides (bn²m⁷GpppA_mpG (7) and (4-Cl-bn)²m⁷GpppA_mpG (8)) had a very high incorporation efficiency of 90.22% and 91.51%, similarly bn²m⁷GppppG (4) and (4-Cl-bn)²m⁷GppppG (5) (88.55% and 92.00%, respectively) while the percentage of incorporation of remaining dinucleotide caps into mRNA ranged from 78.00% to 85.77% (Figure 3; Table 1).

Next, we estimated the orientation of cap incorporation into RNA. For this purpose, we used an assay we developed with the enzyme hNudt16 (Grzela et al., 2022). In a certain concentration range, this enzyme performs the hydrolytic reaction only when the RNA at the 5ʹ end exposes unmethylated guanosine (Grzela et al., 2018; Chrabąszczewska et al., 2021). Given that the cap is asymmetric concerning methylation at position 7 of guanosine, the rate of hydrolysis reflects its orientation of incorporation.

Among the tested analogues, only 7.62% of the bn²m⁷GpppA_mpG-capped RNA and 6.73% of the (4-Cl-bn)²m⁷GpppA_mpG-capped RNA was hydrolyzed by the hNudt16 (Figure 3; Table 1). Such a result indicates the incorporation of the analogue only in the correct orientation. A similar level of hydrolysis was observed for bn²m⁷Gpp_spG-capped RNA (Figure 3; Table 1). However, the orientation of this compound incorporation cannot be assessed, since the presence of a thiophosphate modification in the triphosphate bridge prevents the hydrolysis reaction from proceeding. RNAs capped with compounds (4-(diOCH₃-bn)-tz-CH₂)²m⁷GpppG (1) and (4-(diOCH₃–bn)-tz-(CH₂)₄)²m⁷GpppG (3) undergo hNudt16 hydrolysis to a small extent (11.5%–12.4%), leading to the conclusion that these analogues are built into RNA mostly in the correct orientation, while capping with (4-(diOCH₃–bn)-tz-(CH₂)₂)²m⁷GpppG (2) is incorrect in about 30% (Figure 3; Table 1).

Given that the hNudt16 hydrolysis assay was developed for triphosphate-chain dinucleotides, we checked its fidelity for control tetraphosphate-chain dinucleotides. We used m⁷GppppG and m⁷Gppppm⁷G for the reaction. Hydrolysis of the first analogue showed the presence of GMP and m⁷GTP products. The second of studied analogues, m⁷Gppppm⁷G, did not undergo hydrolysis at the hNudt16 concentration tested. This confirms the validity of the statement that the enzyme preferentially attacks unmethylated guanosine. Therefore, we can conclude that the hydrolysis of dinucleotides with a tetraphosphate bridge proceeds in the same way as for triphosphate compounds.

The hydrolysis of RNA capped with analogues bearing tetraphosphate chain (bn²m⁷GppppG (4) - and (4-Cl-bn)²m⁷GppppG (5)) proceeded at a similar rate as that for m⁷GpppG-RNA, indicating two possible orientations of incorporation of these analogues into RNA chain (Figure 3; Table 1).