Direct evidence for (G)O6···H₂-N4(C)⁺ hydrogen bonding in transient G(syn)-C⁺ and G(syn)-m⁵C⁺ Hoogsteen base pairs in duplex DNA from cytosine amino nitrogen off-resonance R_1ρ relaxation dispersion measurements

Abstract

NMR relaxation dispersion studies have shown that Watson-Crick G-C and A-T base pairs in duplex DNA exist in dynamic equilibrium with their Hoogsteen counterparts. Hoogsteen base pairs form through concurrent rotation of the purine base about the glycosidic bond from an anti to a syn conformation and constriction of the C1′-C1′ distance across the base pair by ~2 Å to allow Hoogsteen type hydrogen bonding. Owing to their unique structure, Hoogsteen base pairs can play important roles in DNA recognition, the accommodation, recognition, and repair of DNA damage, and in DNA replication. NMR relaxation dispersion experiments targeting imino nitrogen and protonated base and sugar carbons have provided insights into many structural features of transient Hoogsteen base pairs, including one of two predicted hydrogen bonds involving (G)N7···H-N3(C)⁺ and (A)N7···H-N3(T). Here, through measurement of cytosine amino (N4) R_1ρ relaxation dispersion, we provide direct evidence for the second (G)O6···H₂-N4(C)⁺ hydrogen bond in G(syn)-C⁺ transient Hoogsteen base pairs. The utility of cytosine N4 R_1ρ relaxation dispersion as a new sensitive probe of transient Hoogsteen base pairs, and cytosine dynamics in general, is further demonstrated by measuring G(syn)-C⁺ Hoogsteen exchange near neutral pH and in the context of the naturally occurring DNA modification 5-methyl cytosine (m⁵C), in DNA samples prepared using chemical synthesis and a ¹⁵N labeled m⁵C phosphoramidite.

Graphical Abstract

Introduction

NMR experiments employing R_1ρ relaxation dispersion (RD) [1–4] have revealed that in duplex DNA, G-C and A-T Watson-Crick base pairs (bps) transiently adopt sparsely populated (~0.1% – 1%) Hoogsteen bps [5, 6]. Hoogsteen bps form through 180° rotation of the purine base about the glycosidic bond from an anti to a syn conformation [7, 8] (Figure 1A). Owing to their unique structure, Hoogsteen bps can play unique roles in DNA recognition by proteins [9, 10] and antibiotics [11, 12], in damage induction, accommodation [5, 13–15] and repair [16, 17], and in bypass of damaged bases during DNA replication [18, 19] (reviewed in [20] and [21]). A deep understanding of the inherent propensity of DNA duplexes to form Hoogsteen bps is important to decipher the occurrence and functional roles of these non-canonical bps in DNA biochemistry.

Figure 1.

Watson-Crick to Hoogsteen exchange and the DNA duplexes used in this study. (A) Watson-Crick G(anti)-C(anti) bps in B-DNA exist in dynamic equilibrium with G(syn)-C⁺(anti) Hoogsteen bps, that are formed by rotation of the guanine base about the glycosidic bond (red arrow) into a syn conformation. Filled green circles denote nuclei that have been used to probe the Watson-Crick to Hoogsteen exchange via R_1ρ RD measurements, while the yellow circle denotes the cytosine amino nitrogen probe used in this study. Rates and populations were obtained from RD measurements reported previously [6]. (B) m¹G destabilizes G-C Watson-Crick bps sterically (red dashes) and via the loss of a hydrogen bond acceptor, biasing the bp conformation towards Hoogsteen. (C) DNA duplexes used in this study with ¹³C, ¹⁵N labeled (A, C, T and G) or ¹⁵N only labeled (m⁵C) nucleotides denoted in bold. All other nucleotides are unlabeled. Black box denotes site at which formation of Hoogsteen bps was examined using NMR.

Transient Hoogsteen bps are proposed to be stabilized by two hydrogen bonds, the formation of which requires shortening of the C1′-C1′ distance across the bp through changes in the sugar and backbone conformation, which also leads to kinking of the DNA double helix [8, 15, 22]. ¹⁵N R_1ρ measurements targeting the imino nitrogen atoms guanine-N1 and thymine-N3, combined with deazapurine substitution experiments have provided direct evidence for the (G)N7···H-N3(C)⁺ and A(N7)···H-N3(T) hydrogen bonds in G(syn)-C⁺ and A(syn)-T transient Hoogsteen bps respectively [23]. The second hydrogen bond (G)O6···H₂-N4(C)⁺ and (A)N6-H₂···O4(T) in G(syn)-C⁺ and A(syn)-T transient Hoogsteen bps respectively, involves the amino groups of cytosine and adenine. Interestingly, these hydrogen bonds also form in Watson-Crick bps and therefore represent an interaction that is preserved during the Watson-Crick to Hoogsteen transition. While this hydrogen bond is observed in DNA duplexes containing N¹-methylated adenine (m¹A) and guanine (m¹G) which stabilize m¹A(syn)-T and m¹G(syn)-C⁺ Hoogsteen bps (Figure 1B) [5, 13, 15], there is as of yet no direct evidence that this hydrogen bond also forms in unmodified transient Hoogsteen bps. The methyl group could in principle alter base stacking and favor the formation of this hydrogen bond.

Recently we demonstrated the utility of guanine amino (N2) R_1ρ RD measurements [24] in characterizing (G)N2-H₂···O2(T) hydrogen bonds in Watson-Crick like G-T mismatches formed by tautomerization and ionization of the bases [25, 26]. Here, through the measurement of cytosine amino (N4) R_1ρ RD, we provide direct evidence for the second (G)O6···H₂-N4(C)⁺ hydrogen bond in transient G(syn)-C⁺ Hoogsteen bps. The utility of cytosine N4 R_1ρ RD as a new sensitive probe of transient Hoogsteen bps, and cytosine dynamics in general, is further demonstrated by measuring G(syn)-C⁺ Hoogsteen exchange near neutral pH and measuring Watson-Crick to Hoogsteen exchange in the context of the naturally occurring DNA modification 5-methyl cytosine (m⁵C), in DNA samples prepared using chemical synthesis and a ¹⁵N labeled m⁵C phosphoramidite.

Materials and Methods

Sample preparation

Unmodified and m¹G containing DNA oligonucleotides:

All unmodified DNA oligonucleotides were purchased from Integrated DNA Technologies with standard desalting purification, while the A₆-DNA^m1G10 single strand was purchased from Midland DNA technologies with cartridge purification.

Synthesis of ¹⁵N isotopically labeled m⁵C phosphoramidite:

Two variants of the ¹⁵N₃-m⁵C amidite were synthesized in the course of this project. First, a N⁴-acetyl (N⁴-Ac) protected variant was synthesized but low yields in the final two steps were observed due to the instability of the N⁴-acetyl group. Then, a revised protecting group strategy relying on N⁴-benzoyl (N⁴-Bz) protection was introduced and satisfactory yields were found in the final steps of the amidite synthesis. Thus, we only report the optimized synthetic procedure using N⁴-benzoyl protection in the Supplementary Information.

m⁵C containing DNA oligonucleotides:

The A₆-DNA^m5C15 single-stranded oligonucleotide was synthesized in-house using a MerMade 6 oligo synthesizer. Standard DNA phosphoramidites (n-ibu-dG, bz-dA, ac-dC, dT, Chemgenes) and a ¹⁵N₃-Ac-m⁵dC phosphoramidite were used with a coupling time of 1 min, with the final 5’-DMT group retained during synthesis. The oligonucleotides were cleaved from the supports (1 μmol) using ~1 mL of AMA (1:1 ratio of ammonium hydroxide and methylamine) for 30 min and deprotected at room temperature for 2 hrs. The single strands were then purified using Glen-Pak DNA cartridges and ethanol precipitated.

¹³C,¹⁵N isotopically labeled DNA oligonucleotides:

¹³C,¹⁵N isotopically labeled DNA single strands were synthesized as described by Zimmer and Crothers [27], using a chemically synthesized template (from IDT), Klenow fragment DNA polymerase (New England Biolabs) and ¹³C,¹⁵N isotopically labeled deoxynucleotide triphosphates (Silantes). The reaction mixture was centrifuged to remove excess pyrophosphate, and then subsequently concentrated to 1.5 mL using a 3 kDa molecular weight cutoff centrifugal concentrator (Millipore Sigma). 1.5 mL of a formamide based denaturing loading dye was then added to the reaction mixture, which was then heated at 95 °C for 5 min for denaturation. The mixture was then loaded onto a denaturing gel (20% polyacrylamide/8M urea) for resolution of the target oligonucleotide from other nucleic acid species. Gel bands corresponding to the pure target single strands were identified by UV-shadowing and subject to electroelution (Whatman, GE Healthcare) followed by ethanol precipitation.

¹³C, ¹⁵N isotopically labeled dCTP: Samples of dCTP were prepared by directly dissolving solid ¹³C, ¹⁵N isotopically labeled dCTP in a buffer of the desired pH.

Sample annealing and buffer exchange:

Single strands (following ethanol precipitation or as purchased) were re-suspended in water. Duplex samples were prepared by mixing equimolar amounts of the constituent single strands and annealing the sample by heating at T = 95 °C for ~5 min followed by cooling at room temperature for ~1 hr. Extinction coefficients for all single and double stranded oligonucleotides were estimated using the ADTBIO oligo calculator (https://www.atdbio.com/tools/oligo-calculator). Extinction coefficients for the modified single strands and duplexes were assumed to be the same as their unmodified counterparts. Modified bases are estimated to affect the extinction coefficient for the oligos used here by < 10% based on reference values in Basanta-Sanchez et al [28]. Following annealing, the samples were exchanged three times into the desired buffer using centrifugal concentrators (4 mL, Millipore Sigma). 10% by volume of D₂O (Millipore Sigma) was added to the samples prior to the NMR measurements.

Buffer preparation:

Sodium phosphate buffers for NMR measurements were prepared by the addition of equimolar solutions of sodium phosphate monobasic and dibasic salts, sodium chloride, and EDTA to give final concentrations (unless mentioned otherwise) of 15 mM (phosphate), 25 mM and 0.1 mM respectively. The pH of the buffers was adjusted by the addition of phosphoric acid, after which they were brought up to the desired volume, and filtered and stored for further usage.

NMR spectroscopy

NMR experiments were performed on a 600 or 700 MHz Bruker Avance 3 spectrometer equipped with triple-resonance HCN cryogenic probes. All experiments were conducted in at pH = 5.4 and at T = 25 °C in NMR buffer unless stated otherwise. The NMR data was processed and analyzed with NMRpipe [29] and SPARKY [30].

Resonance assignments:

Amino resonances were assigned using a combination of 2D NOESY and [¹⁵N, ¹H] HSQC experiments. Assignments for the C6/C1′ resonances were obtained as described previously [5].

Off-resonance R_1ρ RD measurements:

R_1ρ measurements were carried out as described previously for carbon [31] and amino [24] spins, using selective Hartman-Hahn transfers to excite signals corresponding to nuclei of interest. For the amino R_1ρ experiment, matched RF fields were applied on the amino nitrogen and the amino proton with the more intense signal in the 2D [¹⁵N, ¹H] HSQC. Magnetization corresponding to cytosine-N4/cytosine-C6/guanine-C1′ was allowed to relax under an applied spin-lock field for a maximal duration (< 120 ms for ¹⁵N and < 60 ms for ¹³C) chosen appropriately to achieve ~70% loss in signal intensity at the end of the relaxation period. The signal intensity was recorded for 4–7 delays equally spaced over the relaxation period. Spin-lock powers used for ¹³C and ¹⁵N R_1ρ measurements ranged from 150–2200 Hz and 150–600 Hz respectively (Table 5). Absolute offset frequencies were chosen ranging from 0 to 3.5 times the given spin-lock power (Table 5). Offsets greater than 3.5 times the spin-lock power were avoided owing to significant R₁ relaxation contributions [31].

Table 5.

Spin-lock powers (ω₁/2π, units Hz) and offsets (Ω/2π, units Hz) used in off-resonance R_1ρ RD measurements.

	[ω1/2π (Hz)] [Ω/2π (Hz)]
C15-N4 A6-DNA (pH = 5.4, T=25 °C)	[300] [−1050, −840, −630, −420, −210, −10, 10, 210, 420, 630, 840, 1050] [400] [−1400, −1120, −840, −560, −280, −10, 10, 280, 560, 840, 1120, 1400] [800] [−2800, −2240, −1680, −1120, −560, −10, 10, 560, 1120, 1680, 2240, 2800]
C15-C6 A6-DNA (pH = 5.4, T=25 °C)	[200] [−702, −624, −546, −468, −390, −312, −234, −156, −78, −10, 10, 78, 156, 234, 312, 390, 468, 546, 624, 702] [300] [−1053, −936, −819, −702, −585, −468, −351, −234, −117, −10, 10, 117, 234, 351, 468, 585, 702, 819, 936, 1053] [400] [−1404, −1248, −1092, −936, −780, −624, −468, −312, −156, −10, 10, 156, 312, 468, 624, 780, 936, 1092, 1248, 1404] [600] [−2097, −1864, −1631, −1398, −1165, −932, −699, −466, −233, −10, 10, 233, 466, 699, 932, 1165, 1398, 1631, 1864, 2097]
C15-N4 A6-DNA (pH = 6.8, T=25 °C)	[200] [−702, −585, −468, −351, −234, −117, −10, 10, 117, 234, 351, 468, 585, 702] [300] [−1050, −875, −700, −525, −350, −175, −10, 10, 175, 350, 525, 700, 875, 1050] [500] [−1752, −1460, −1168, −876, −584, −292, −10, 10, 292, 584, 876, 1168, 1460, 1752] [1000] [−3498, −2915, −2332, −1749, −1166, −583, −10, 10, 583, 1166, 1749, 2332, 2915, 3498]
G10-C1′ A6-DNA (pH = 6.8, T=25 °C)	[150] [−440, −400, −360, −320, −200, −160, −120, 100, 300] [200] [−600, −550, −500, −450, −400, −400, −350, −300, −250, −150, −50, 100, 200] [400] [−1150, −1000, −850, −700, −650, −600, −550, −500, −450, −400, − 400, −350, −300, −250, −200, −150, −100, 50, 200, 350, 500, 650, 800, 1000] [600] [−1800, −1400, −1200, −1000, −800, −700, −600, −550, −500, − 450, −400, −400, −350, −300, −250, −200, −100, 200, 400, 600, 1000, 1400, 1800]
m5C15-N4 A6-DNAm5C15 (pH = 5.4, T=25 °C)	[500] [−1750, −1400, −1050, −700, −350, −10, 10, 350, 700, 1050, 1400, 1750] [800] [−1750, −1400, −1050, −700, −350, −10, 10, 350, 700, 1050, 1400, 1750] [1400] [−4900, −3920, −2940, −1960, −980, −10, 10, 980, 1960, 2940, 3920, 4900]
m5C15-N4 A6-DNAm5C15 (pH = 6.8, T=25 °C)	[200] [−702, −585, −468, −351, −234, −117, −10, 10, 117, 234, 351, 468, 585, 702] [300][−1050, −875, −700, −525, −350, −175, −10, 10, 175, 350, 525, 700, 875, 1050] [500][−1752, −1460, −1168, −876, −584, −292, −10, 10, 292, 584, 876, 1168, 1460, 1752] [1500][−5250, −4375, −3500, −2625, −1750, −875, −10, 10, 875, 1750, 2625, 3500, 4375, 5250]

Fitting of R_1ρ RD data:

R_1ρ values for a given spin-lock power offset combination were obtained by fitting the peak intensities as a function of delay time to a mono-exponential function. Bloch-McConnell (B-M) simulations were used to fit R_1ρ values as a function of spin-lock power and offset to a two state exchange model [3, 25], with the uncertainties in the exchange parameters extracted using a Monte-Carlo scheme as described previously [32].

Density Functional Theory Calculations

Density Functional Theory (DFT) calculations were performed using Gaussian 09d (Gaussian Inc.) [33] to compute chemical shifts for G-C Watson-Crick and G(syn)-C⁺ Hoogsteen bps, and isolated cytosine nucleosides with and without protonation. The starting structure for the G-C Watson-Crick bp and the cytosine nucleoside were obtained from an idealized G-C bp in B-form DNA, which was constructed using the fiber module of the 3DNA suite of programs [34]. The starting structure for the G(syn)-C⁺ Hoogsteen bp and protonated cytosine nucleoside were obtained from the crystal structure (PDB ID: 1QNA) of TATA box binding protein in complex with DNA [10]. For all the calculations, two rounds of geometry optimization were initially performed using the B3LYP functional [35] and the 3–21G [36] and 6–311+G(2d,p) [37] basis sets, with the positions of all heavy atoms fixed during optimization. NMR chemical shifts were then computed using the GIAO method [38] on the configuration at the end of the second round of optimization. ¹³C and ¹⁵N chemical shifts were referenced to those of carbon and nitrogen in TMS and NH₃ respectively, which were computed at the same level of theory.

Results

Cytosine N4 R_1ρ RD provides direct evidence for (G)O6···H₂-N4(C)⁺ hydrogen bonding in transient G(syn)-C⁺ Hoogsteen base pairs

Our strategy to detect the (G)O6···H₂-N4(C)⁺ hydrogen bond in transient G(syn)-C⁺ Hoogsteen bps is to use the cytosine amino nitrogen (N4) as an R_1ρ RD probe. The choice of this nucleus is motivated by its direct involvement in the hydrogen bond, the well-known sensitivity of nitrogen chemical shifts to changes in hydrogen bonding [39, 40] and the ability to perform R_1ρ RD experiments on amino nitrogen atoms. The approach builds on that used in a prior study on G-T mismatches wherein guanine amino N2 R_1ρ RD was used to detect the (G)N2-H₂···O2(T) hydrogen bond in transient Watson-Crick like G-T mismatches [24, 26]. Although cytosine-N4 is hydrogen bonded in both G-C Watson-Crick and G(syn)-C⁺ Hoogsteen bps, differences in the cytosine-N4 chemical shift can be expected owing to protonation of the base and also possibly due to differences in stacking and hydrogen bonding.

To test whether the cytosine-N4 chemical shift is indeed a sensitive probe of the Watson-Crick to Hoogsteen transition, we first recorded 2D [¹⁵N, ¹H] HSQC spectra of A₆-DNA and A₆-DNA^m1G10 duplexes (Figure 1C) containing Watson-Crick G10-C15 and Hoogsteen m¹G10(syn)-C15⁺ bps, respectively (Figure 2A). In these samples, the DNA strand containing C15 was ¹³C, ¹⁵N isotopically labeled (Figure 1C). In m¹G, the N¹-methyl group biases the conformation of the bp towards Hoogsteen by sterically destabilizing the Watson-Crick conformation and by disrupting a Watson-Crick hydrogen bond (Figure 1B) [5]. The 2D [¹⁵N, ¹H] HSQC spectra of A₆-DNA and A₆-DNA^m1G10 show that that in both the G10-C15 Watson-Crick and m¹G10(syn)-C15⁺ Hoogsteen bps, cytosine-N4 is hydrogen bonded, as two separate amino proton signals are observed (Figure 2A). Overlaying the two spectra, we find that C15-N4 experiences a large downfield shift (Δω_C15-N4 = 10.6 ppm) on formation of a Hoogsteen bp (Figure 2A). Such sizeable downfield shifts have previously been reported for C(anti)-G(anti)-C⁺(anti) base triples in which the protonated cytosine forms Hoogsteen hydrogen bonds with the anti guanine [41]. These chemical shift perturbations are also in agreement with density functional theory (DFT) calculations [5] carried out on G(syn)-C⁺ Hoogsteen and G-C Watson-Crick bps, which predict Δω_C-N4 = 9.7 ppm (Methods section). The calculations show that this large shift is primarily driven by protonation of cytosine-N3 since even for an isolated cytosine nucleoside, Δω_C-N4 on protonation is calculated to be 13.4 ppm, which is similar to Δω_C-N4 = 10 ppm observed experimentally on protonation of cytosine nucleosides [42]. Nevertheless, we can expect that hydrogen-bonding does contribute to the chemical shift difference of cytosine-N4 between Watson-Crick and Hoogsteen bps as studies employing ¹⁵N labeling of the amino groups in the context of nucleic acids have shown that the amino nitrogen chemical shift can vary by 2–4 ppm due to changes in hydrogen bonding alone [43, 44], and because the N4 amino chemical shift of a protonated cytosine in a hydrogen bonded m¹G-C⁺ Hoogsteen bp is downfield shifted relative to isolated dCTP⁺ by ~ 6.8 ppm (Figure S1).

Figure 2.

Using off-resonance cytosine-N4 R_1ρ RD to probe a transient G(syn)-C⁺ Hoogsteen bp in duplex DNA. (A) Overlay of 2D [¹⁵N, ¹H] HSQC spectra of the amino region of A₆-DNA and A₆-DNA^m1G10. (B) Off-resonance R_1ρ RD profiles for C15-N4 and C15-C6 measured in A₆-DNA. Spin-lock amplitudes are color coded. Solid lines denote shared fits to the RD data using the Bloch-McConnell (B-M) equations assuming a two-state exchange process. The initial alignment of the magnetization during the B-M fitting was performed as described previously [3]. (C) Comparison of the difference in C15-N4 chemical shift (Δω_C15-N4 = ω_C15-N4(A₆-DNA^m1G10) - ω_C15-N4(A₆-DNA)) between Hoogsteen m¹G(syn)-C⁺ and Watson-Crick G-C bps from 2D HSQC spectra and between Hoogsteen G(syn)-C⁺ and Watson-Crick G-C bps from R_1ρ measurements (Δω_C15-N4 = ω_ES - ω_GS) (Table 1). Error bars for panels B and C represent the experimental uncertainty obtained while fitting the R_1ρ data, as determined by a Monte-Carlo scheme described previously [32]. All measurements were carried out at pH = 5.4 and T = 25 °C in NMR buffer. For the above measurements, the strand containing C15 in A₆-DNA and A₆-DNA^m1G10 was ¹³C, ¹⁵N isotopically labeled (Figure 1C).

Thus, based on the comparison of amino HSQC spectra of A₆-DNA and A₆-DNA^m1G10, if transient G(syn)-C⁺ Hoogsteen bps were indeed stabilized by a (G)O6···H₂-N4(C)⁺ hydrogen bond, we would predict cytosine-N4 to experience a ~ 10.6 ppm downfield change in chemical shift on forming the Hoogsteen bp. In contrast, if the protonated cytosine was not hydrogen bonded in the Hoogsteen bp, a smaller downfield shift of around 6–8 ppm is expected, based on prior studies exploring the sensitivity of the amino nitrogen chemical shift to changes in hydrogen bonding [43, 44]. To test the above prediction, we performed off-resonance R_1ρ RD measurements on C15-N4 in A₆-DNA at pH = 5.4 and T = 25 °C. As a control, we also measured off-resonance R_1ρ RD profiles for C15-C6, which has been shown to be sensitive to cytosine protonation accompanying Hoogsteen bp formation [5, 45].

We observed R_1ρ RD at C15-C6 as expected for exchange between G-C Watson-Crick and G(syn)-C⁺ Hoogsteen bps [5, 45]. The data could be satisfactorily fit to a two-state exchange model involving a dominant ground state (GS) and a sparsely populated excited state (ES) [46] (Figure S2). The exchange parameters obtained from this fit (p_ES = 0.942 ± 0.089 %, k_ex = k₁ + k₋₁ = 2194 ± 239 s⁻¹, where p_ES is the ES population, k_ex is the exchange rate and k₁ and k₋₁ are the forward and backward rates, respectively) were in very good agreement with values reported previously for the Watson-Crick to Hoogsteen transition for this G-C bp using other base and sugar RD probes (Table 1, Figure S2) [5, 6, 22, 45]. Indeed, we also observed R_1ρ RD for C15-N4 consistent with chemical exchange on the micro-to-millisecond timescale. A two-state fit of the data yield p_ES = 0.720 ± 0.028 % and k_ex = 1933 ± 147 s⁻¹ that are in excellent agreement with values obtained for C15-C6 and those reported previously for the Watson-Crick to Hoogsteen transition [5, 6, 22] (Table 1 and Figure S2). These results indicate that the ES detected in the C15-N4 RD measurements corresponds to a G(syn)-C⁺ Hoogsteen bp (Supplementary Discussion). Indeed, the RD data for C15-N4 and C15-C6 could be satisfactorily fit together when sharing the populations and exchange rates (p_ES = 0.740 ± 0.022 %, k_ex = 1866 ± 106 s⁻¹) (Figure 2B, Table 1).

Table 1.

Exchange parameters obtained from two-state fitting of off-resonance R_1ρ RD data for C15-N4 and C15-C6 measured in A₆-DNA at pH = 5.4 and T = 25 °C. For these measurements, an A₆-DNA duplex in which the strand containing C15 was ¹³C, ¹⁵N isotopically labeled was used (Figure 1C).

	C15-N4 (2-state Individual Fit)	C15-C6 (2-state Individual Fit)
pES (%)	0.720 ± 0.028	0.942 ± 0.089
kex (s−1)	1933 ± 147	2194 ± 239
Δω (ppm)	10.66 ± 0.18	2.01 ± 0.09
R1(s−1)	2.48 ± 0.08	2.32 ± 0.09
R2(s−1)	8.55 ± 0.3	30.35 ± 0.56
	C15-N4 + C15-C6 (2-state Shared Fit)
pES (%)	0.740 ± 0.022
kex (s−1)	1866 ± 106
Δω (ppm)	10.58 ± 0.16	2.19 ± 0.07
R1 (s−1)	2.48 ± 0.07	2.20 ± 0.07
R2(s−1)	8.59 ± 0.25	31.42 ± 0.25

Fitting of the RD data yielded a sizeable change (Δω_C15-N4 = ω_ES – ω_GS = 10.58 ± 0.16 ppm, where ω_ES and ω_GS are the chemical shifts of the nucleus in the ES and GS) in the C15-N4 chemical shift between the ES Hoogsteen and GS Watson-Crick bps which is in excellent agreement with the value Δω_C15-N4 = 10.6 ppm predicted based on the difference in C15-N4 chemical shift measured for m¹G(syn)-C⁺ Hoogsteen bps versus G-C Watson-Crick bps in A₆-DNA (Figure 2C). This agreement indicates that like the m¹G(syn)-C⁺ bp, the transient G(syn)-C⁺ Hoogsteen bp does indeed form a (G)O6···H₂-N4(C)⁺ hydrogen bond. Similar results were robustly obtained near physiological pH = 6.8 (Figure S3 and Table 2), which also showed the expected decrease in Hoogsteen bp population relative to pH = 5.4 (p_{ES,pH 5.4} = 0.740 ± 0.022 %, p_{ES,pH 6.8} = 0.039 ± 0.007 %). However, because of conformational exchange [45] between m¹G(syn)-C⁺(anti) and m¹G(anti)-C(anti) geometries of the m¹G-C bp at pH 6.8, where the population of the distorted Watson-Crick m¹G(anti)-C(anti) bp is nearly 28%, the amino signals are broadened out of detection, preventing comparison of the Δω with that obtained from R_1ρ, as done at pH 5.4 (Figure 2C).

Table 2.

Exchange parameters obtained from two-state fitting of off-resonance R_1ρ RD data for C15-N4 and G10-C1′ measured in A₆-DNA at pH = 6.8 and T = 25 °C. For these measurements, an A₆-DNA duplex in which both single strands were ¹³C, ¹⁵N isotopically labeled was used (Figure 1C).

	C15-N4 (2-state Individual Fit)	G10-C1′ (2-state Individual Fit)
pES (%)	0.035 ± 0.008	0.063 ± 0.016
kex (s−1)	1513 ± 701	1701 ± 899
Δω (ppm)	9.74 ±1.16	3.25 ± 0.50
R1 (s−1)	4.08 ±0.02	1.5 ± 0.10
R2 (s−1)	9.01 ±0.08	14.92 ± 0.19
	C15-N4 + G10-C1′ (2-state Shared Fit)
pES (%)	0.039 ± 0.007
kex (s−1)	1595 ± 611
Δω (ppm)	9.22 ± 1.04	3.59 ± 0.73
R1 (s−1)	4.09 ± 0.02	1.56 ± 0.08
R2 (s−1)	8.97 ± 0.08	15.13 ± 0.12

Using cytosine N4 R_1ρ RD to measure the Watson-Crick to Hoogsteen exchange at G-m⁵C base pairs

As an additional application, we used C-N4 R_1ρ RD to examine how methylation of the cytosine at the C5 position impacts the Watson-Crick to Hoogsteen exchange. 5-methyl cytosine (m⁵C) is one of the most abundant naturally occurring modified bases in DNA that plays important roles in modulating gene expression and regulation [47–49]. Modulation of the dynamics of the Watson-Crick to Hoogsteen exchange represents a potential avenue by which m⁵C could exert its biological effects.

For these studies, we synthesized an m⁵C phosphoramidite in which the base moiety was selectively ¹⁵N labeled at all ring nitrogen atoms (Figure 3A and Methods section). We first examined whether m⁵C affects hydrogen bonding by comparing 2D [¹⁵N, ¹H] HSQC spectra of A₆-DNA containing Watson-Crick G10-m⁵C15 (A₆-DNA^m5C15) and Hoogsteen m¹G10(syn)-m⁵C15⁺ (A₆-DNA^m5C15m1G10) bps at pH = 5.4 and T = 25 °C (Figure 1C and Figure 3B).

Figure 3.

Using off-resonance cytosine amino R_1ρ RD to probe Watson-Crick to Hoogsteen exchange in G-m⁵C bps at pH = 5.4 and T = 25 °C in NMR buffer. (A) Chemical structure of m⁵C with sites of ¹⁵N labeling highlighted with a red circle. (B) Overlay of 2D [¹⁵N, ¹H] HSQC spectra of the amino region of A₆-DNA^m5C15 and A₆-DNA^m5C15m1G10. (C) Off-resonance R_1ρ RD profile for m⁵C15-N4 in A₆-DNA^m5C15. Spin-lock amplitudes are color coded. Solid lines denote a fit of the RD data to the B-M equations assuming a two-state exchange process. The initial alignment of the magnetization during the B-M fitting was performed as described previously [3]. (D) Comparison of the difference in m⁵C15-N4 chemical shift (Δω_m5C15-N4 = ω_m5C15-N4(A₆-DNA^m5C15m1G10) - ω_m5C15-N4(A₆-DNA^m5C15)) between Hoogsteen m¹G(syn)-m⁵C⁺ and Watson-Crick G-m⁵C bps from 2D HSQC spectra, and between Hoogsteen G(syn)-m⁵C⁺ and Watson-Crick G-m⁵C bps from R_1ρ measurements on A₆-DNA^m5C15 (Δω_m5C15-N4 = ω_ES - ω_GS) (Table 3). (E) Comparison of the population of the transient Hoogsteen bp (p_HG) and the exchange rate (k_ex) for the Watson-Crick to Hoogsteen exchange for G10-C15 and G10-m⁵C15 bps in A₆-DNA and A₆-DNA^m5C15 at pH = 5.4 and T = 25 °C, respectively. Error bars for panels C, D and E represent the experimental uncertainty obtained while fitting the R_1ρ data, as determined by a Monte-Carlo scheme described previously [32].

The observation of two separate amino proton signals indicates that in both Watson-Crick and Hoogsteen conformations, the amino group of m⁵C is hydrogen bonded (Figure 3B). A large downfield shift in the m⁵C15-N4 chemical shift (Δω_m5C15-N4 = 11.4 ppm) is observed upon formation of a Hoogsteen bp involving an (m¹G)O6···H₂-N(m⁵C)⁺ hydrogen bond (Figure 3B). Thus, cytosine N4 R_1ρ RD should also be suitable for probing transient G(syn)-m⁵C⁺ Hoogsteen bps.

Next, we used off-resonance cytosine-N4 R_1ρ RD to examine how cytosine methylation impacts the Watson-Crick to Hoogsteen exchange. Indeed, we did observe off-resonance R_1ρ RD for m⁵C15-N4 in A₆-DNA^m5C15 at pH = 5.4 and T = 25 °C (Figure 3C). Fitting the RD data to a two state model yielded Δω_m5C15-N4 = 11.0 ± 0.2 ppm (Table 3), which is again in very good agreement with Δω_C15-N4 = 11.4 ppm obtained from the difference in m⁵C15-N4 chemical shift measured for m¹G(syn)-m⁵C⁺ Hoogsteen versus G-m⁵C Watson-Crick bps in A₆-DNA (Figure 3D). This agreement indicates that the transient G(syn)-m⁵C⁺ Hoogsteen bp also forms a (G)O6···H₂-N4(m⁵C)⁺ hydrogen bond.

Table 3.

Exchange parameters obtained from two-state fitting of off-resonance R_1ρ RD data for m⁵C15-N4 measured in A₆-DNA^m5C15 at pH = 5.4 and T = 25 °C.

	m5C15-N4 (2-state Individual Fit)
pES (%)	0.78 ± 0.046
kex (s−1)	2076 ± 192
Δω (ppm)	11.0 ± 0.16
R1 (s−1)	2.66 ± 0.04
R2(s−1)	9.33 ± 0.21

Importantly, the population (p_ES = 0.780 ± 0.046 %) and the exchange rate (k_ex = 2076 ± 192 s⁻¹) measured for transient G(syn)-m⁵C⁺ Hoogsteen were very similar to values measured for the unmodified G(syn)-C⁺ Hoogsteen (p_ES = 0.740 ± 0.022 % and k_ex = 1866 ± 106 s⁻¹) (Figure 3E), indicating that the m⁵C modification minimally impacts the thermodynamics or kinetics of Watson-Crick to Hoogsteen exchange. Similar results were obtained when performing experiments at near neutral pH in which the modification minimally impacts the exchange parameters, though as expected the population is reduced for both modified and unmodified samples (Tables 1–4, and Figure S4).

Table 4.

Exchange parameters obtained from two-state fitting of off-resonance R_1ρ RD data for m⁵C15-N4 measured in A₆-DNA^m5C15 at pH = 6.8 and T = 25 °C.

	m5C15-N4 (2-state Individual Fit)
pES (%)	0.037 ± 0.006
kex (s−1)	2323 ± 739
Δω (ppm)	10.13 ± 1.23
R1 (s−1)	1.97 ± 0.03
R2 (s−1)	7.45 ± 0.11

Discussion

One of the most important aspects of nucleic acid structure is the nature of base pairing which is defined by the number and types of hydrogen bonds between nucleotides. The measurement of R_1ρ RD data targeting imino nitrogen guanine-N1 and thymine/uracil-N3 [23, 50, 51] as well as amino nitrogen guanine-N2 [24] has made it possible to probe hydrogen bonding and base pairing in transient low-abundance ESs of DNA and RNA. Here, using RD measurements on cytosine-N4, we have obtained direct evidence in support of the second (G)O6···H₂-N4(C)⁺ hydrogen bond in transient G(syn)-C⁺ Hoogsteen bps in canonical duplex DNA. The excellent agreement between the C-N4 chemical shifts in transient G(syn)-C⁺ and m¹G(syn)-C⁺ Hoogsteen bps further validates the m¹G mutant as a mimic of the ES G(syn)-C⁺ Hoogsteen bp. Determining high resolution structure and dynamic ensemble of m¹G(syn)-C⁺ containing DNA duplexes may provide further insights into the G(syn)-C⁺ Hoogsteen ES, as recently done for m¹A(syn)-T Hoogsteen bps [15, 22].

The same strategy used in this work could in principle be applied to examine the (A)N6-H₂···O4(T) hydrogen bond in transient A(syn)-T Hoogsteen bps. However, performing R_1ρ RD measurements on the adenine amino group is typically hampered by low sensitivity in conventional 2D [¹⁵N, ¹H] HSQC experiments [52] owing to severe line broadening of ¹H signals due to rotation of the amino group on the intermediate chemical shift timescale [53]. This problem could be addressed by using sensitivity enhancement modules [54] or CPMG-INEPT magnetization transfer blocks [55]. Alternatively, one could rely on magnetization transfer to A-N6 from the directly bonded C6 using the relatively large ¹J_N6-C6 ~ 20 Hz [56]. More generally, cytosine amino R_1ρ RD experiments could be used to study conformational dynamics in other nucleic acid systems involving changes in hydrogen bonding and base protonation, such as triplexes [57] and i-motifs [58].

The exchange parameters measured for the Watson-Crick to Hoogsteen transition of the G10-C15 bp in A₆-DNA at pH = 5.4, T = 25 °C and 90% H₂O/10% D₂O in this study (p_ES = 0.740 ± 0.022 %, k_ex = 1866 ± 106 s⁻¹) differ slightly different from those determined previously [22] using the same buffer conditions but in the presence of 100% D₂O (p_ES = 0.40 ± 0.01 %, k_ex = 992 ± 52 s⁻¹). This hints at the presence of both kinetic and thermodynamic isotope effects for the G-C Watson-Crick to G(syn)-C⁺ Hoogsteen bp exchange; the former could be used to gain insights into the nature of the transition state for formation of a G(syn)-C⁺ Hoogsteen bp as done in prior studies of the enzyme Ribonuclease A [59], wherein rate-limiting conformational changes gating product release were suggested to involve proton transfer, based on the observation of H/D kinetic isotope effects. Finally, while our studies indicate that for the A₆-DNA sequence context m⁵C does not significantly impact the Watson-Crick to Hoogsteen exchange, we cannot rule out larger effects in other sequence contexts, such as CpG dinucleotides in which m⁵C is often found [48, 49]. Additional studies are required to test this possibility.

Highlights.

• Cytosine-N4 R_1ρ RD used to probe chemical exchange in DNA

• Detection of (G)O6···H₂-N4(C)⁺ hydrogen bond in transient Hoogsteen base pairs

• 5-methyl cytosine minimally impacts Watson-Crick to Hoogsteen exchange