5‐Oxyacetic Acid Modification Destabilizes Double Helical Stem Structures and Favors Anionic Watson–Crick like cmo⁵U‐G Base Pairs

Abstract

Watson–Crick like G‐U mismatches with tautomeric G^enol or U^enol bases can evade fidelity checkpoints and thereby contribute to translational errors. The 5‐oxyacetic acid uridine (cmo⁵U) modification is a base modification at the wobble position on tRNAs and is presumed to expand the decoding capability of tRNA at this position by forming Watson–Crick like cmo⁵U^enol‐G mismatches. A detailed investigation on the influence of the cmo⁵U modification on structural and dynamic features of RNA was carried out by using solution NMR spectroscopy and UV melting curve analysis. The introduction of a stable isotope labeled variant of the cmo⁵U modifier allowed the application of relaxation dispersion NMR to probe the potentially formed Watson–Crick like cmo⁵U^enol‐G base pair. Surprisingly, we find that at neutral pH, the modification promotes transient formation of anionic Watson–Crick like cmo⁵U⁻‐G, and not enolic base pairs. Our results suggest that recoding is mediated by an anionic Watson–Crick like species, as well as bring an interesting aspect of naturally occurring RNA modifications into focus—the fine tuning of nucleobase properties leading to modulation of the RNA structural landscape by adoption of alternative base pairing patterns.

The fidelity of translation relies on the ability of the rigid ribosomal decoding center to recognize the shape of Watson–Crick base pairs.1 Mismatches can evade these fidelity mechanisms by spatial mimicry of the Watson–Crick like shape through tautomerization and ionization.2 The propensity to adopt such Watson–Crick like base pairs can be tuned by modifications of the canonical bases at the wobble position to rewire3 and change the efficiency of translational decoding.4 Uridine 5‐oxyacetic acid (cmo⁵U) is one such modification that based on X‐ray crystallographic and biochemical data has been proposed to increase the efficiency of translation of G‐ending codons by promoting formation of Watson–Crick like cmo⁵U^enol‐G base pairs.5 However, because hydrogen atoms could not be visualized by crystallography, the precise nature of the Watson–Crick like species could not be unambiguously determined.

Herein, we have synthesized a ¹⁵N isotopically labeled version of the cmo⁵U modification that allows the application of relaxation dispersion NMR to probe the formation of Watson–Crick like base pairs. Surprisingly, we find that the modification promotes the transient formation of anionic cmo⁵U⁻‐G base pairs via decreasing the pK _a of the nucleoside. Our results provide key mechanistic insights into how cmo⁵U influences base‐pairing properties and modulates decoding, and lay out a framework for studying the impacts of other RNA modifications.1

Our first efforts focused on the synthesis of a ¹⁵N₂‐cmo⁵U phosphoramidite. The synthetic access is loosely based on published syntheses but represents the first comprehensive description of the chemical synthesis of the cmo⁵U RNA building block. The synthesis started from ¹⁵N₂‐uridine, which was prepared according to published procedures by using ¹⁵N₂‐urea.6 In the next steps, the 5‐hydroxyl group was installed,7 followed by protection of the sugar hydroxyls and installation of the 5‐oxyacetic acid side‐chain as an ester functionality.8 Removal of the 2′,3′‐ketal protecting group was followed by the regioselective incorporation of the 2′‐tert‐butyl‐dimethylsilyl group by using a transient 3′,5′‐di‐tert‐butylsilyl protection. Then, after removal of the methyl ester functionality, the 4‐nitrophenylethylester was obtained via a Steglich esterification. The remaining steps were the introduction of the 5′‐(4,4′‐dimethoxy)‐trityl group and phosphitylation of the 3′‐hydroxyl group yielding the desired stable isotope modified cmo⁵U RNA building block. A detailed synthetic scheme and further experimental details on the synthetic steps can be found in the Supporting Information.

Then, the cmo⁵U modification was introduced into 20 nt RNA model system to probe changes in the overall thermodynamic stability. The cmo⁵U‐modified RNAs were characterized by anion‐exchange chromatography and mass spectrometry (Figures S1 and S2 in the Supporting Information). Four RNA constructs differing at the central U5‐A16 base pair were synthesized and subjected to UV melting analysis to give the melting temperatures and thermodynamic parameters of the 20 nt hairpins—the U‐A hp, the U⋅G hp, the cmo⁵U‐A hp and the cmo⁵U⋅G hp (Figure 1 a–c and Figure S3 in the Supporting Information). As was expected, the U‐A hairpin comprising only canonical base pairs showed the highest melting transition at 85 °C and the most negative free energy of −12.2 kcal mol⁻¹, closely followed by the U⋅G hp (T _m=83.3 °C, ΔG ^298 K=−11.3 kcal mol⁻¹). The cmo⁵U‐modified hairpins showed slightly decreased stabilities in‐line with previously conducted free‐energy calculations.9 The cmo⁵U‐A hp, although forming a canonical base pair exhibits an even lower thermodynamic stability than the U⋅G hp (ΔT _m ^{U⋅G/cmo5U‐A}=1.3 °C, ΔG ^{298 K,U⋅G/cmo5U‐A}=0.5 kcal mol⁻¹). The UV melting experiment gives information on the global thermodynamic stabilities of the stem‐loop folds, but no structural features and local residue‐specific dynamics can be delineated from the UV data.

Figure 1.

Thermodynamic stability of the unmodified and cmo⁵U‐modified RNAs. a) Secondary structures of the four RNAs under investigation. The inset shows the structure of the cmo⁵U modification. The nitrogen atoms highlighted in blue are ¹⁵N labeled. b) Bar graph of the melting temperatures of the hairpin RNAs. c) Bar graph of the free energies at 298 K of the hairpin RNAs.

However, solution NMR spectroscopy is perfectly suited to study structure and dynamics in biomolecules at atomic resolution.6, 10 Based on NMR data, the U‐A hp forms a standard A‐form RNA stem capped with an extra‐stable UUCG loop. The U⋅G hp was extensively investigated earlier and in the major populated ground state a U⋅G wobble base pair was observed.11 Basically, the same base pairing properties were observed for the cmo⁵U‐modified stem‐loop RNAs (Figure 2). The cmo⁵U‐A hp showed slightly shifted imino proton resonances compared to the U‐A wildtype sequence, which could be assigned using a ¹H‐homonuclear NOESY experiment (Figure 2 a, upper trace). Noteworthy, the N3H of resonance is shifted from 13.6 ppm (U‐A base pair) to 14.1 ppm (cmo⁵U‐A base pair) pointing toward a more de‐shielded magnetic environment. We further demonstrated the Watson–Crick like base pairing via an HNN‐COSY with residue‐specific ¹⁵N3 and ¹⁵N1 labeling of cmo⁵U5 and A16, respectively (Figure 2 a). The cmo⁵U⋅G wobble base pair was confirmed by ¹H‐¹⁵N‐correlation NMR spectroscopy making use of ¹⁵N3 and ¹⁵N1 labeling of cmo⁵U5 and G16 (Figure 2 b). Again, a 0.5 ppm de‐shielding effect for the cmo⁵U imino proton resonance (U⋅G hp U5 H3 11.6 ppm; cmo⁵U⋅G hp U5 H3 12.1 ppm) was induced by the 5‐oxyacetic acid modification.

Figure 2.

Structural features of cmo⁵U‐modified hairpins. a) Canonical cmo⁵U‐A base pair is observed in an HNN‐COSY experiment at 25 °C. b) ¹H‐¹⁵N HSQC data showing chemical shift signatures for the formation of a cmo⁵U‐G wobble base pair (at 5 °C). c) Structural model of the G4‐C17 and cmo⁵U5‐A16 base pairs in the cmo⁵U‐A hp. Interfering negative charges of the 5‐oxyacetate and the phosphate backbone are shown in dot representation.

We then addressed individual base pair exchange kinetics in all four hairpins by the application of CLEANEX‐PM experiments (Table S1 and Figure S4 in the Supporting Information).12 The slow to very slow imino proton exchange rates between 0.8 to 10 s⁻¹ for the U‐A hp are in‐line with the high thermodynamic stability of the stem loop structure and reveal no dynamic hotspot. In the U⋅G hp the wobble base pair represents a dynamic base pair opening hot spot with the U5‐N3H displaying the highest exchange rate with bulk water protons (22.15 s⁻¹) followed by its interaction partner G16‐N3H (10.14 s⁻¹). The destabilization effect introduced by the U⋅G wobble appears to be only localized as other imino proton exchange rates are only marginally affected (e.g., U18).

The water‐proton‐exchange NMR experiment of cmo⁵U‐A hp gave slightly to strongly elevated imino proton water‐exchange rates compared to its unmodified U‐A counterpart. The cmo⁵U modification strongly influences the stability of the 5′‐preceding base pair—in this case, the G4‐C17 base pair. We observed an almost tenfold increase in the water‐exchange rate of the G4‐N1H proton. A structural model of the cmo⁵U‐A hairpin gives a potential explanation for the destabilization of the 5′‐preceding base pair. The model reveals that the negatively charged 5‐oxyacetate side chain gets in proximity to the likewise negatively charged phosphate group of G4 (Figure 2 c). The repulsion effect induced by this unfavorable electrostatic interaction gives a rational for the tenfold weakening of the G4‐C17 base pair and the concomitant global fold destabilization, as has been seen in the UV melting curve analysis. In the thermodynamically most unstable hairpin, the highly dynamic hotspot at the cmo⁵U⋅G mismatch site was directly reflected in the imino proton water‐exchange rates. We found strongly elevated exchange rates at the mismatch site. Most prominently, the cmo⁵U5‐N3H resonance was strongly broadened at 25 °C preventing the determination of the exchange rate with water protons. The G16 N1H water exchange is sixfold enhanced compared to the U⋅G wobble base pair (59.88 vs. 10.14 s⁻¹). Again, the destabilization effect introduced by the cmo⁵U⋅G wobble is highly localized, because other imino proton water‐exchange rates are only marginally affected.

Finally, we addressed the postulated stabilization of the rare enolic uridine state by the cmo⁵U modification via ¹⁵N‐R_1ρ relaxation dispersion NMR experiments. We previously showed in the hpU⋅G sequence context (Figure 1 a) that the U⋅G wobble mismatch exists in dynamic equilibrium with two equally populated and rapidly exchanging U‐G^enol and U^enol‐G base pairs, with the anionic species falling below detection limits at neutral pH (p _B<0.05 %) and only becoming detectable (p _B≈0.3 %) at high pH>7.9.11

We examined how replacement of the U⋅G mismatch with a cmo⁵U⋅G base pair affects these dynamics. Strikingly, we found that at neutral pH, the modification significantly increased ¹⁵N RD at both cmo⁵U5‐N3 and G16‐N1 indicating that it promotes the exchange process to WC‐like mismatches (by ca. fivefold, Figure 3 a and Table S2 in the Supporting Information). As a negative control, no RD was observed for cmo⁵U‐A hp (Figure S5 in the Supporting Information). Unlike the data for unmodified U⋅G hp, which could be interpreted in terms of two‐state exchange between the wobble and two rapidly exchanging U‐G^enol and U^enol‐G base pairs, the RD data for the cmo⁵U⋅G hp called for a three‐state fit giving two ESs (Table S2 in the Supporting Information). ES1 (p _B=0.16±0.03 % and k _ex ^GS‐B=5949±709 s⁻¹) has Δω (cmo⁵U5‐N3)=16.0±4.8 ppm and Δω (G16‐N1)=40.4±4.1 ppm consistent with a mixture of rapidly exchanging cmo⁵U‐G^enol and cmo⁵U^enol‐G and species in a 70:30 ratio. ES2 (p _c=0.24±0.03 % and k _ex ^GS‐C=976±683 s⁻¹) has Δω (cmo⁵U5‐N3)=55.4±2.9 and Δω (G16‐N1)=−0.02±3.5 consistent with the formation of a cmo⁵U⁻‐G anionic base pair. This was confirmed through measurement of RD profiles at higher pH 8.0 (Table S2 and Figure S6 in the Supporting Information). Therefore, the modification increases the population of the anionic species by >twofold such that it becomes the dominant species in the ES at neutral pH (ca. 60 %), although minimally impacting the population of the tautomeric species (ca. twofold) or tautomeric equilibrium (from 50:50 to 70:30 cmo⁵U‐G^enol/cmo⁵U^enol‐G). These findings are in‐line with a recent computational study, in which no evidence for the stabilization of the enolic uridine state by the oxyacetic acid modification was found.13 They also mirror the effects of 5‐bromo‐2′‐deoxyuridine (5BrdU), which is a mutagenic agent in DNA that causes increased mis‐incorporation via an anionic species.14 We got further support for the preference of the anionic Watson–Crick like base pairing from the pK _a values determined for the cmo⁵U nucleoside (Figures S7 and S8 in the Supporting Information). By using pH‐dependent ¹³C NMR data, the pK _a of the 5‐oxyacetic acid side‐chain was found to be 2.8. Thus, at physiological pH the 5‐oxyacetic acid is negatively charged in‐line with the observed destabilization effect found for the cmo⁵U‐A RNA. The pK _a of the imino proton H3 in cmo⁵U is 8.7, that is, the proton is more acidic than in the unmodified uridine nucleoside (pK _a 9.2–9.6) by almost a factor of ten. This higher acidity favors the anionic Watson–Crick like base pair state, as has been seen in the relaxation dispersion data. Further, the higher acidity of the N3‐bound proton rationalizes the observed de‐shielding effect by the change in its electronic environment.

Figure 3.

Probing the excited states in cmo⁵U‐modified RNA. a) ¹⁵N‐R_1ρ relaxation dispersion experiments for N3 of cmo⁵U5 and N1 of G16 with a three‐state global fit at pH 6.9 and 10° C. b) Summary on ground/excited state equilibria for the unmodified U‐G and the cmo⁵U‐G wobble base pairs.

To conclude, we successfully synthesized a ¹⁵N stable isotope labeled variant of a cmo⁵U RNA building block that allowed to obtain unprecedented insights into the structure and dynamics of RNA carrying this naturally occurring modification via solution NMR spectroscopy. We found a generally destabilizing effect for double helical RNA stem structures by the cmo⁵U modification very likely due to the additional negative charge introduced by the oxyacetic acid sidechain. The destabilization is reflected in a lowered melting temperature and higher free energy values, and also slightly enhanced imino water proton exchange rates. We further probed the presumed stabilizing effect of the 5‐oxyacetic acid modification on the U‐enol tautomeric form by ¹⁵N‐R_1ρ relaxation dispersion NMR spectroscopy. We found that an anionic Watson–Crick like cmo⁵U⁻‐G base pair represents the major species in the excited state. Thus, we have evidence that in a previous work the cmo⁵U^enol‐G WC‐like base pair was misinterpreted and is in fact an anionic cmo⁵U‐G base pair.5a This anionic Watson–Crick base pair between the mRNA and the tRNA anticodon loop might be further stabilized by the controlled conditions regarding pH or metal ion positioning within the ribosome, thus giving an explanation for the expansion of the decoding capacity via the cmo⁵U modification by a lowered N3H pK _a value.10 The electron density map in the previous work of Weixlbaumer et al. is also compatible with the anionic base pair formed by N3 and O2 of cmo⁵U and N1H and N2H of G (Figure 3 b).5a

Further, the effects of cmo⁵U in RNA are reminiscent to that of 5‐BrdU in DNA, which also promotes formation of an anionic base pair by lowering the pK _a.14 Thus, cmo⁵U can be regarded as nature's version of 5‐BrdU. Although 5‐BrdU is a known chemotherapeutic agent that acts by promoting G point mutations, the function of cmo⁵U is the enhanced reading capability of G at the third codon position by the formation of an anionic base pair. To summarize, the results highlight an important aspect of the function of naturally occurring RNA modifications—the alteration of the folding landscape of RNA by fine‐tuning nucleobases properties leading to alternative base pairing patterns. We are further confident that solution NMR spectroscopy will give important insights into the function of naturally occurring RNA modifications, which is especially important in the light of the recent finding that RNA epigenetics plays an important role in many cellular processes.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

E. Strebitzer, A. Rangadurai, R. Plangger, J. Kremser, M. A. Juen, M. Tollinger, H. M. Al-Hashimi, C. Kreutz, Chem. Eur. J. 2018, 24, 18903.