Direct NMR Evidence that Transient Tautomeric and Anionic States in dG•dT Form Watson-Crick like Base Pairs

Abstract

The replicative and translational machinery utilizes the unique geometry of canonical G•C and A•T/U Watson-Crick base pairs to discriminate against DNA and RNA mismatches in order to ensure high fidelity replication, transcription, and translation. There is growing evidence that spontaneous errors occur when mismatches adopt a Watson-Crick-like geometry through tautomerization and/or ionization of the bases. Studies employing NMR relaxation dispersion recently showed that wobble dG•dT and rG•rU mismatches in DNA and RNA duplexes transiently form tautomeric and anionic species with probabilities (≈0.01–0.40%) that are in concordance with replicative and translational errors. While computational studies indicate that these exceptionally short-lived and low-abundance species form Watson-Crick like base pairs, their conformation could not be directly deduced from the experimental data, and alternative pairing geometries could not be ruled out. Here, we report direct NMR evidence that the transient tautomeric and anionic species form hydrogenbonded Watson-Crick like base pairs. A guanine-to-inosine substitution, which selectively knocks out a Watson-Crick type (G)N2H₂•••O2(T) hydrogen bond, significantly destabilized the transient tautomeric and anionic species, as assessed by lack of any detectable chemical exchange by imino nitrogen rotating frame spin relaxation (R_1ρ) experiments. An ¹⁵N R_1ρ NMR experiments targeting the amino nitrogen of guanine (dG-N2) provides direct evidence for Watson-Crick (G)N2H₂•••O2(T) hydrogen bonding in the transient tautomeric state. The strategy presented in this work can be generally applied to examine hydrogen-bonding patterns in nucleic acid transient states including in other tautomeric and anionic species that are believed to play roles in replication and translational errors.

Graphical abstract

In their paper describing the structure of the DNA double helix, Watson and Crick hypothesized that spontaneous mutations could arise when bases adopt their rare energetically less favorable tautomeric forms¹ as this could allow mismatches such as dA•dC and dG•dT to pair up in a Watson-Crick (WC) like geometry. Topal and Fresco extended this idea to also encompass a wider range of tautomeric WC-like mismatches that together could account for a broader spectrum of transition and transversion mutations² as well as miscoding events during translation.³ They also hypothesized that the probability of tautomerization could be an important determinant of misincorporation probability during replication and translation.^2,3

Many decades later, it is now well-established that the replicative and translational machinery utilize the unique WC geometry (Figure 1a) to discriminate against mismatches.^4–8 In addition, there is considerable evidence that tautomeric⁹ as well as anionic¹⁰ WC-like mismatches play roles in replication^2,11–13 and translation errors.^3,14,15 WC-like mismatches have been observed in crystal structures of polymerase^9,11 and the ribosome^8,14,15 in catalytically active conformations. Chemical modifications that stabilize the tautomeric and anionic WC-like species also result in increased misincorporation and base substitution probabilities.^10,16–18 Recently, studies based on NMR relaxation dispersion (RD)^19–21 targeting the imino nitrogen of guanine (G-N1) and thymidine / uridine (T-N3 / U-N3) in DNA and RNA duplexes provided evidence for spontaneous transitions from wobble (WB) G•T/U mismatches (Figure 1b) toward short-lived and low-populated WC-like tautomeric (Figure 1c) and anionic (Figure 1d) mismatches.²² Consistent with the Topal and Fresco hypothesis, these transitions toward WC-like mismatches occur with probabilities (10⁻³–10⁻⁵) that are comparable to the probabilities of misincorporation.²²

Figure 1.

Wobble and Watson-Crick like dG•dT mismatches (a) Watson-Crick dG•dC base pair. (b) Wobble dG•dT base pair with dG-N2 highlighted with a red circle. (c) Enolic ES1 features a rapid equilibrium between major dG^enol•dT and minor dG•dT^enol (d) Anionic ES2 dG•dT⁻. A dynamic equilibrium with a minor dG⁻•dT that is not detectable by relaxation dispersion cannot be ruled out.²²

Although the NMR ¹⁵N imino RD data strongly suggest deprotonation of G-N1 or T/U-N3, whether or not the resulting enolic (G^enol and T^enol/U^enol) and anionic (T⁻/U⁻) bases do indeed adopt a WC-like base pair (bp) stabilized by hydrogen-bonding (H-bonding) remains to be established (Figure 1c/d). Characterizing H-bonds in short-lived low-populated species that form though a subtle rearrangement of protons presents a challenge to existing biophysical methods (Figure 1). Here, we developed two complimentary approaches and obtained evidence for WC-like H-bonding in transient tautomeric and anionic dG•dT mismatches.

The prior ¹⁵N NMR RD study²² showed that ground state (GS) dG•dT mismatches exist in dynamic equilibrium with two short-lived and low-abundance species referred to as ‘excited states’ (ESs).^19,22–25 The pH-independent ES1 is characterized by significantly downfield shifted dG-N1 (Δω_N1 = ω_ES1–ω_GS ≈ 36–40 p.p.m.) and dT-N3 (Δω_N3 ≈ 9–19 p.p.m.) chemical shifts.^11,22 This was interpreted as evidence for deprotonation of dG-N1 and T/U-N3 to form a major G^enol and minor T/U^enol tautomeric species that exist in fast exchange on the NMR timescale (Figure 1c). ES2 had a population (p_B) that increased with pH and featured a smaller downfield shift in G-N1 (Δω_N1 ≈ 1–10 p.p.m.) and much larger downfield shift in T/U-N3 (Δω_N3 ≈ 45–57). This was interpreted as ionization of T/U-N3 to form anionic T⁻/U⁻ (Figure 1d).²²

While the NMR data did not provide information regarding the conformation of these deprotonated bases, there is indirect evidence that they form WC-like bps (Figure 1c/d). First, the downfield shift in ES1 for the H-bond acceptor nitrogen (N1/3) was smaller than expected based on deprotonating isolated dGTP-N1 and dTTP-N3.²² This could in part be explained by H-bonding between N1 and N3 (Figure 1).^22,26 Second, the relative energetic stability of ES1 measured by NMR RD was in good agreement with values computed using density functional theory calculations for H-bonded WC-like tautomeric species.^26,27 However, without direct experimental evidence for H-bonding, alternative conformations²² could not be entirely ruled out.

In the WB dG•dT bp, the exocyclic amino group of guanine does not form a H-bond with dT though it may form a relatively weak water-mediated H-bond.²⁸ In contrast, the same amino group is predicted to form an N-H•••O H-bond in both ES1 and ES2 (Figure 1c/d), analogous to that observed in G•C bps (Figure 1a). It correctly positions the purine N3 and pyrimidine O2 H-bond acceptors for sequence-independent recognition by polymerases.²⁹ If the ESs do indeed adopt a WC-like geometry capable of replicative and translational errors, then they are expected to be stabilized by a WC-like N2H₂•••O2 H-bond. Omission of the amino group via a guanine to inosine substitution has been shown to have little effect (≈0.1 kcal mol⁻¹) on the stability of the GS dG•dT WB.³⁰ In contrast, by knocking out the WC-like N2H₂•••O2 H-bond, it is expected to destabilize ES1 and ES2 by at least ≈1 kcal mol⁻¹, and to diminish their abundance to levels (<0.01%) undetectable by RD NMR (Figure 2a).^2,11,22

Figure 2.

Guanine (dG) to inosine (dI) substitution suppresses ¹⁵N imino dT-N3 relaxation dispersion in dG•dT. (a) Removal of the exocyclic dG-NH₂ amino group selectively knocks out a H-bond in the proposed excited state Watson-Crick like mismatches (enolic dT•dI^enol, dT^enol•dI, and anionic dT•dI) without substantially affecting the H-bonds and stability of ground state wobble (dT•dI). (b) hpTG-CGC DNA hairpin construct used in RD measurements. (c) ¹⁵N RD profiles measured for dT-N3 in dT•dG (top) and dT•dI (bottom) at variable pH (inset). Solid lines represent best fit to RD data. Error bars represent experimental uncertainty (one s.d.).

We examined how a guanine (dG) to inosine (dI) substitution (Figure 2a) affects off-resonance R_1ρ RD profiles measured for dT-N3 in a dG•dT mismatch embedded in a DNA hairpin (hpTG-CGC) used in prior studies (Figure 2b).²² Comparison of NMR spectra for the two duplexes confirms that the dG to dI substitution has little effect on the GS WB conformation (Figure S1a) consistent with previous studies.^30,31 Additionally, inosine is expected to have a similar tautomeric behavior to guanine based on experimental³² and computational studies.^33,34 As expected, at pH = 6.9 in which ES2 is suppressed, we observed significant dT-N3 RD in dG•dT consistent with a single ES. Fitting the RD profiles to a two-state model yields exchange parameters (p_B = 0.169 ± 0.002% and k_ex = k₁ + k₋₁ = 2896 ± 96 s⁻¹, Δω_N3 = 18.4 ± 0.12 p.p.m.) that are consistent with the tautomeric ES1 (Figure 2). A second ES (p_B = 0.031 ± 0.001%, k_ex = 17030 ± 1230 s⁻¹, Δω_N3 = 49.4 ± 1.50 p.p.m.) is observed upon increasing the pH to 8.4 that is consistent with anionic ES2 (dG•dT⁻).²² In sharp contrast, the dT-N3 RD profiles measured for the dI•dT mismatch were flat at both pH 6.9 and 8.0 (Figure 2c) and over a range of temperatures (Figure S2b). These results indicate that a dG to dI substitution significantly destabilizes both the tautomeric ES1 and anionic ES2 by selectively knocking out a Watson-Crick type N2H₂•••O2 H-bond as assessed by the loss of detectable chemical exchange at the dT-N3.

To more directly probe the WC-like (G)N2H₂•••O2(T) H-bond, we adapted, with minor modifications (Supporting), the selective R_1ρ experiment first introduced for proteins³⁵ and then adapted for nucleic acid imino N1/3 nitrogens³⁶ to measure dG-N2 RD (Figure 1b). The selective R_1ρ experiment employs Hartmann-Hahn cross polarization^37,38 (CP) to selectively excite individual resonances. In dG•dC WC bps, the amino group is involved in a (G)N2H₂•••O2(C) H-bond (Figure 1a) and appears as a single NMR resonance with dG-N2 ≈ 72 p.p.m. In contrast, the WB dG•dT amino group is not H-bonded (Figure 1b) and the two equivalent protons appear as a single resonance with dG-N2 ≈ 74–76 p.p.m. Therefore, a transition from the GS WB to a WC-like ES, in which it forms a (G)N2H₂•••O2(T) H-bond (Figure 1b), is predicted to induce a significant downfield shift (Δω_G-N2 >2 p.p.m.) (Figure S2a) in the dG-N2 chemical shift which should give rise to measurable ¹⁵N RD.

RD measurements targeting amino nitrogens have not yet been reported in nucleic acids. However, the utility of ¹⁵N CPMG RD measurements targeting amino NH₂ groups in protein side chains has already been established.³⁹ These studies have highlighted additional considerations in RD experiments for AX₂ spin systems, including the need to account for potential contributions arising from N-H/N-H dipole-dipole (DD_N2Ha/DD_N2Hb) cross-correlated cross-relaxation.^39,40 Fortunately, because the angle between the two N-H dipoles in amino NH₂ groups is ≈120°, these contributions are predicted to be small for the constructs studied here and can safely be ignored.⁴¹ This was verified by observation of monoexponential decays and ¹⁵N multiplets with approximately 1:2:1 intensity ratios in a coupled HSQC spectrum (Figure S2c/d/S3).

As a negative control, we measured dG-N2 R_1ρ in a canonical dG•dC WC bp within a hairpin duplex (Figure S2e) under high pH conditions where chemical exchange due to transient HG bps is suppressed (Figure S2f). As expected, the RD profiles were all flat with no signs of chemical exchange (Figure 3a).

Figure 3.

dG-N2 RD profiles. (a) dG-N2 RD profiles measured in a Watson-Crick dG•dC base pair at pH 6.0 and 25°C showing absence of chemical exchange. (b) ¹⁵N RD profiles measured in a Wobble dG•dT mismatch for dG-N1, dT-N3, and dG-N2 in hpTG-CGC (see Figure 2b) at pH 6.9 25 °C. Solid line represents a global fit of the N1/N2/N3 RD data with shared populations and exchange rates. dG-N2 RD profiles measured in a Wobble dG•dT mismatch in hpTG-GGG at (c) pH 6.9, (d) pH 8.0. Solid lines represent individual fits of the N2 RD data.

Next, we measured dG-N2 RD in a site-specifically labeled dG•dT mismatch embedded within the hpTG-CGC construct (Figure 2b). Measurements were initially carried out at neutral pH so as to suppress ES2. In contrast to the dG•dC WC bp, we observed clear signs of dG-N2 RD in the dG•dT mismatch. Two-state analysis of the dG-N2 RD profiles (Supporting) yielded exchange parameters (p_B = 0.24 ± 0.04% and k_ex = 2486 ± 208) that are in good agreement with those obtained for ES1 (p_B = 0.17 ± 0.002% and k_ex = 2743 ± 41) based on N1/3 RD measurements.²² Indeed, the RD data measured for dG-N2, dG-N1, and dT-N3 at pH 6.9 could be globally fitted (Figure 3b), confirming that that they are reporting on the same GS-ES1 exchange process.

Importantly, the differences in the dG-N2 chemical shift between the GS and ES (Δω_G-N2 = +4.85 ± 0.19 p.p.m.) deduced by the 2-state analysis of the dG-N2 RD data are in excellent agreement with the experimentally observed differences (Δω_G-N2 ≈ +4 p.p.m.) between the dG-N2 chemical shift in H-bonded WC dG•dC and non-H-bonded WB dG•dT bps in the same sequence context (Figure S2a). We also carried out analogues dG-N2 RD measurements for a dG•dT mismatch in a second unique sequence context and distinct ES1/ES2 exchange parameters (Figure S2g). Once again we observed dG-N2 RD profiles that yield exchange parameters (p_B = 0.53 ± 0.037% and k_ex = 3087 ± 100) that are in good agreement with those obtained for ES1 based on N1/3 RD measurements (p_B = 0.37 ± 0.004% and k_ex = 2797 ± 41). The value of Δω_G-N2 ≈ +5.15 p.p.m. is consistent with H-bonding and in agreement with the experimentally observed differences (Δω_G-N2 ≈ +4 p.p.m.) between the dG-N2 chemical shifts in WC dG•dC and WB dG•dT bps.

dG-N2 is also predicted to be H-bonded in ES2. However, simulations show that due to the much faster GS-ES2 exchange kinetics (k_ex ≈ 50,000 s⁻¹ as compared to ≈ 2000 s⁻¹ for ES1), the ES2 contribution to dG-N2 RD is expected to be negligible (Figure S4) even at pH = 8.0, in which the population of both ES1 (0.23%) and ES2 (0.22%) are significant.²² In accord with this expectation, we observed insignificant changes in the dG-N2 RD profiles measured in the dG•dT mismatch in hpTG-GGG when increasing to pH 8.0 (Figure 3c/d).

In conclusion, our results provide direct experimental evidence that tautomeric and anionic dG•dT ESs adopt a WC-like geometry that mimics both the shape and H-bond patterns of canonical WC bps, and therefore satisfy the key requirements for inducing replication and translation errors. The methodology described in this work provides a new means for characterizing H-bond alignments and base-pairing patterns in transient nucleic acid conformational states.