NMR Observation of Intermolecular Hydrogen Bonds between Protein Tyrosine Side-Chain OH and DNA Phosphate Groups

Abstract

Hydrogen bonds between protein side-chain hydroxyl (OH) and phosphate groups are one of the most common types of intermolecular hydrogen bonds in protein-DNA/RNA complexes. Using NMR spectroscopy, we identified and characterized the hydrogen bonds between tyrosine side-chain OH and DNA phosphate groups in a protein-DNA complex. These OH groups exhibited relatively slow hydrogen exchange rates and sizable scalar couplings between hydroxyl ¹H and DNA phosphate ³¹P nuclei across the hydrogen bonds. Information about intermolecular hydrogen bonds facilitates investigations of DNA/RNA recognition by the protein.

Graphical Abstract

Introduction

Hydrogen bonding is vital for structures and functions of biological molecules.¹ Specific patterns of hydrogen bonds define secondary structures of proteins and nucleic acids. Molecular recognition and catalysis by these molecules are often facilitated through stable or transient intermolecular hydrogen bonds. Therefore, for a mechanistic understanding of a particular biomolecular process, it is important to identify and characterize the hydrogen bonds directly involved in the process.

NMR spectroscopy is a powerful tool for investigations of hydrogen bonds.² Although hydrogen bonds are typically identified using the geometrical criteria of donor and acceptor groups in three-dimensional structures, some NMR methods allow direct identification of hydrogen bonds even without any structural information.³ These methods became available in the late 90s when scalar couplings across hydrogen bonds (referred to as “hydrogen-bond scalar couplings”) were discovered for nucleic-acid base pairs and protein secondary structures.^4–7 Since then, various types of hydrogen-bond scalar couplings have been identified.⁸ Hydrogen-bond scalar couplings provide not only structural information but also information about dynamics because these couplings are influenced by the transient distortion or breakage of the hydrogen bonds as well.^9–12

Observation of intermolecular hydrogen-bond scalar couplings is often more challenging than that of intramolecular hydrogen-bond scalar couplings because self-decoupling occurs when the residence time of the complex is not sufficiently long.¹² Nonetheless, intermolecular hydrogen-bond scalar couplings have been observed for nucleic acids (as reviewed in refs^13–14), protein–nucleic acid complexes,^12,15–22 and other protein–ligand complexes^23–24. Observations of intermolecular hydrogen-bond scalar couplings have been made for some protein side chains, including serine (Ser),²³ threonine (Thr),²³ arginine (Arg),²⁵ histidine (His),²¹ and lysine (Lys).^12,15–20

Another signature of hydrogen bonds detectable by NMR is slow hydrogen exchange.²⁶ Labile hydrogen atoms of NH/NH₂/NH₃⁺ or OH groups undergo hydrogen exchange with water. When these groups form stable hydrogen bonds, the hydrogen exchange is generally slower.²⁷ For example, although ¹H signals of Lys side-chain NH₃⁺ groups are typically undetectable due to rapid hydrogen exchange, their ¹H signals are detectable when these moieties interact with other moieties and form hydrogen bonds.²⁸

In this work, we used NMR spectroscopy to identify and characterize intermolecular hydrogen bonds between tyrosine (Tyr) side-chain hydroxyl (OH) groups and DNA phosphates in a protein-DNA complex. Although rapid hydrogen exchange often broadens ¹H resonances of OH groups beyond detection, two interfacial Tyr OH groups in contact with DNA phosphates in the fruit-fly Antennapedia homeodomain (Antp HD)–DNA complex exhibited clearly observable ¹H resonances, owing to relatively slow hydrogen exchange. For these OH groups, we were able to observe hydrogen-bond scalar couplings (^h2J_HP) between Tyr side-chain hydroxyl ¹H and DNA phosphate ³¹P nuclei. Because hydrogen bonds between OH and phosphate groups are highly common in protein-DNA and protein-RNA complexes, NMR data on these hydrogen bonds would be useful for a physicochemical understanding of how proteins recognize nucleic acids.

Experimental Methods

Preparation of the Antp HD–DNA complex

The Antp HD–DNA complex was prepared as previously described.²⁹ The protein was the 60-residue construct used in the crystallographic study by Pabo and coworkers,³⁰ rather than the 68-residue construct used in the NMR studies by Wüthrich and coworkers.^31–32 The DNA was the 15-bp DNA with the same sequence as that used for the crystallographic study.³⁰ A 370-μl solution of 1.1 mM complex of ¹³C,¹⁵N-lableled Antp HD and unlabeled 15-bp DNA in a pH 5.8 buffer containing 20 mM potassium succinate-d₄, 100 mM KCl, and 0.4 mM NaF (as a preservative) was sealed into a 5-mm coaxial NMR tube (Norell, Inc.). A 2-mm stem insert (Norell, Inc.) containing 100 μl of D₂O was used for the NMR lock.

Through-bond ¹H-¹³C correlation for Tyr side chains

¹H, ¹³C, and ¹⁵N resonances of backbone and non-aromatic side chains of the Antp HD–DNA complex used for the current sample had previously been assigned.^20,29 Aromatic ¹³C-¹H resonances were assigned using through-bond correlation signals in a three-dimensional (3D) HCCH-COSY spectrum, two-dimensional (2D) ¹H-¹³C HSQC and (HB)CB(CGCD)HD spectra^33–34 recorded using Bruker Avance III spectrometers equipped with a TCI cryogenic probe at a ¹H frequency of 750 or 800 MHz. A long-range HMQC spectrum was recorded to observe hydroxyl ¹H_η–¹³C_ε correlation signals via ³J_CH couplings.

Measurements of hydrogen exchange rates for Tyr OH groups

For the Tyr OH groups, hydrogen exchange rates (k_HX) were measured at 15°C and 25°C through one-dimensional (1D) CLEANEX-PM experiments³⁵ in which a ¹⁵N-filter scheme³⁶ was implemented to exclude signals from NH protons. This ¹⁵N filter was incorporated into the WATERGATE scheme³⁷ in the manner presented in Figure 1D of Ikura and Bax.³⁶ The experiments were performed at a ¹H frequency of 600 MHz using 7 mixing-time points: 0.4, 2.1, 4.2, 8.4, 14.7, 21.0, 33.6, and 62.9 ms. The k_HX rates were determined through fitting calculations with Eq. 1 in Hwang et al.³⁸ using MATLAB (MathWorks, Inc.). Each k_HX measurement took ~14 hours with 3744 scans per free induction decay (FID). Uncertainties in the k_HX rates were estimated from three replicates.

Figure 1.

NMR signals from Y8 and Y25 OH groups at the protein-DNA interface of the complex of ¹³C,¹⁵N-labeled Antp HD and unlabeled 15-bp DNA at 15˚C. (A) The Tyr OH signals in a 1D ¹H spectrum recorded with ¹⁵N decoupling. (B) Through-bond correlations between Tyr hydroxyl ¹H and ¹³C_ε nuclei observed in a long-range ¹H-¹³C HMQC spectrum. ¹H_ε−¹³C_ε correlations observed in a standard ¹H-¹³C HMQC spectrum recorded for the same complex is also shown. (C) Interactions between Tyr side-chain OH and DNA phosphate groups in the 2.5-Å resolution crystal structure of the Antp HD–DNA complex (PDB 9ANT)³⁰.

Observation of hydrogen-bond scalar couplings between Tyr OH and DNA phosphate groups

Hydrogen-bond scalar couplings (^h2J_HP) between Tyr side-chain hydroxyl ¹H and DNA ³¹P nuclei were measured at 15°C using a Bruker Avance III spectrometer equipped with a ¹H/¹³C/¹⁵N/³¹P QCI cryogenic probe operated at a ¹H frequency of 600 MHz. To observe ¹H-³¹P correlation signals arising from coherence transfer via ^h2J_HP couplings, a long-range ¹H-³¹P HMQC spectrum was recorded at 15°C with ¹⁵N GARP and ³¹P WALTZ16 decoupling schemes, essentially as previously described.²¹ However, a shorter delay for ¹H-³¹P coherence transfer (i.e., 10 ms between ¹H and ³¹P 90° pulses) was used to mitigate the relaxation loss due to rapid hydrogen exchange of the Tyr OH groups. The total experiment time for recording of the ¹H-³¹P HMQC spectrum was ~136 hours. The magnitude of each ^h2J_HP coupling between OH and phosphate groups was measured using a 1D version of the ¹H spin-echo ^h2J_HP modulation difference experiment²¹ that incorporate a ¹⁵N filter scheme involving two 90° ¹⁵N pulses separated by (2J_NH)⁻¹ in the manner of Clore et al.³⁹ The total length of the spin-echo scheme (T_SE) was set to 33.6 ms, which was ~3-fold longer than the apparent transverse relaxation time of each OH proton. The absolute value of the coupling constant, |^h2J_HP|, was determined with the following equation: I_a/I_b = cos(πJT_SE); where I_a and I_b represent the signal intensities in the spectra with and without the ^h2J_HP modulation, respectively. Uncertainties in |^h2J_HP| were estimated from 3 replicates. For each, the total measurement time was ~42 hours. Other details about the |^h2J_HP| measurements are given below.

Statistical analysis of protein side-chain OH–DNA phosphate hydrogen bonds

In total, 3213 crystal structures of protein-DNA or protein-RNA complexes solved at a resolution < 2.5 Å were retrieved from the Research Collaboratory for Structural Bioinformatics (RCSB) PDB server.⁴⁰ Using MATLAB, intermolecular hydrogen bonds involving O_5’, O_P1, O_P2, or O_3’ atoms of DNA/RNA phosphate groups were identified using a criterion of O/N…O distance < 3.2 Å. Intermolecular hydrogen bonds between phosphate and Arg guanidium, Lys NH₃⁺, Asn/Gln NH₂, or Ser/Thr/Tyr OH groups were counted for the protein-DNA and protein-RNA complexes.

Results

NMR signals from Tyr OH groups

We noticed ¹H NMR signals from Tyr OH groups when we analyzed a 3D ¹³C-edited NOESY-HSQC spectrum recorded for aromatic side chains of the Antp HD–DNA complex. The ¹H_ε nuclei of Tyr8 and Tyr25 exhibited relatively strong cross peaks of the nuclear Overhauser effect (NOE) with ¹H resonances at 9.67 and 9.94 ppm, respectively, which did not match with any assigned resonances of ¹³C- or ¹⁵N-attached ¹H nuclei within a short (< 4 Å) distance from the Tyr side chains. These ¹H resonances were observed in the 1D ¹H spectrum as well (Figure 1A). Since 1D ¹H spectra with and without ¹⁵N (or ¹³C) decoupling showed these ¹H resonances as singlet peaks, these resonances are not from ¹H atoms bonded to ¹⁵N or ¹³C. To test the possibility that these ¹H resonances are from Tyr OH groups, we recorded a long-range ¹H-¹³C HMQC spectrum using relatively long (10 ms) delays for ¹H-¹³C coherence transfer. If these ¹H resonances are from Tyr OH groups, the three-bond scalar couplings (³J_CH) between the hydroxyl ¹H and aromatic ¹³C_ε nuclei could be sizable (~5–7 Hz)^41–42. The obtained spectrum showed cross peaks between these ¹H resonances and the ¹³C_ε resonances of Tyr8 and Tyr25 (Figure 1B–C), which unequivocally indicate that the ¹H signals at 9.67 ppm and 9.94 ppm are from the OH groups of Tyr8 and Tyr25, respectively.

Hydrogen exchange of interfacial Tyr OH groups

In the crystal structure of the Antp HD–DNA complex, the OH groups of Tyr8 and Tyr25 formed hydrogen bonds to DNA phosphate groups (Figure 1D). The intermolecular hydrogen bonds explain why these OH groups clearly exhibit ¹H resonances although hydroxyl ¹H resonances are typically undetectable due to severe broadening caused by rapid hydrogen exchange.²⁶ In fact, the Antp HD in the free state did not exhibit detectable ¹H signals from the Tyr OH groups. To quantitatively assess the hydrogen exchange of the Tyr8 and Tyr25 OH groups, we performed a 1D CLEANEX-PM experiment³⁵, in which water magnetization was transferred to the OH groups through hydrogen exchange. Since the ¹H resonances of the Tyr OH groups were close to some protein NH resonances (in particular, Arg52 ¹H_ε and Trp56 ¹H_ε1 resonances; see Figure 1A), we incorporated a ¹⁵N filter scheme³³ to suppress the NH resonances. Figure 2A shows intensities of the ¹H signals from Tyr 8 and Tyr25 as a function of the mixing time for the transfer. Through nonlinear least-squares fitting to these data, we determined the hydrogen exchange rates k_HX for Tyr8 and Tyr25 OH groups. Figure 2B shows the best-fit curves with the determined values of k_HX rates. Interestingly, the hydrogen exchange of the Tyr25 OH group was ~50–100-fold faster than that of the Tyr8 OH group. The rapid hydrogen exchange of the Tyr25 OH group is consistent with its broader ¹H resonance (see Figure 1A–B). The crystal structure and hydrogen-bond scalar couplings in solution (see below) suggest that the hydrogen-bonding geometry is similar for these two Tyr OH groups. However, the Tyr8 OH group is more buried in the binding interface and its solvent accessible surface areas is remarkably smaller (Figure 2C). Therefore, the slower hydrogen exchange rate of the Tyr8 OH group could be due to less accessibility for water.

Figure 2.

Hydrogen exchange of the interfacial Tyr OH groups of the Antp HD–DNA complex. (A) Series of 1D ¹H spectra recorded in the ¹⁵N-filtered CLEANEX-PM experiment for the complex of ¹⁵N-lableled Antp HD and 15-bp DNA. (B) CLEANEX-PM data for determination of the hydrogen exchange rates k_HX for Y8 and Y25 OH groups in the Antp HD–DNA complex at 25°C (red) and 15°C (blue). The ratio of the signal intensity observed in the CLEANEX-PM experiment to the signal intensity in the reference spectrum is plotted for each mixing time.³⁵ The solid lines represent best-fit curves obtained through fitting as described by Hwang et al.³⁸ The k_HX rate for the Tyr25 OH group was 43 ± 3 s⁻¹ at 15°C and 106 ± 4 s⁻¹ at 25°C. The k_HX rate for the Tyr8 OH group was determined to be 0.74 ± 0.04 s⁻¹ at 25°C and 1.1 ± 0.6 s⁻¹ at 15°C. (C) Solvent accessible surface areas of the Tyr O_η atoms in the crystal structure (PDB 9ANT).

Direct evidence of Tyr OH–phosphate hydrogen bonds

To examine whether the interfacial Tyr OH groups in contact with DNA phosphates exhibit hydrogen-bond scalar couplings ^h2J_HP, we recorded a long-range ¹H-³¹P HMQC spectrum. As shown in Figure 3A, this spectrum clearly showed ¹H-³¹P correlation signals arising from coherence transfer via ^h2J_HP couplings for the Tyr8 and Tyr25 OH groups. Compared to typical ³¹P chemical shifts of B-form DNA (−4.6 to −4.0 ppm with respect to tetramethyl phosphate [TMP]),⁴³ the ³¹P chemical shifts of these signals were up-field shifted presumably due to the hydrogen bonds. This shift was also clearly observed in 1D ³¹P spectrum recorded for the sample (Figure 3B). Through the 1D spin-echo ^h2J_HP modulation difference experiment, we also measured the magnitude of the ^h2J_HP couplings. In this experiment, ¹⁵N-filtered 1D ¹H spectra were recorded after the spin-echo period with and without ^h2J_HP modulation (Figure 3C). The values of |^h2J_HP| for the OH groups of Tyr8 and Tyr25 were determined to be 3.0 ± 0.1 and 3.4 ± 0.3 Hz, respectively. These couplings were significantly larger compared to those between the Ser/Thr hydroxyl ¹H and FMN ³¹P nuclei in the flavodoxin-FMN complex (0.5 – 1.7 Hz)²³ despite similar O^…O distances (2.53 – 2.87 Å for Tyr and 2.49 – 2.79 Å for Ser/Thr) in the crystal structures. Although the hydrogen bonds may differ in the positionings of hydrogen atoms of the rotatable OH groups, this possibility cannot be examined in the current case because the crystal structures do not include hydrogen atomic coordinates. Nonetheless, the observed ^h2J_HP couplings represent direct evidence of the intermolecular hydrogen bonds in solution.

Figure 3.

Observation of hydrogen-bond scalar couplings ^h2J_PH between DNA phosphate ³¹P and Tyr hydroxyl ¹H nuclei for the Antp HD–DNA complex. (A) 2D long-range ¹H-³¹P HMQC spectrum. ¹H-³¹P correlation signals arising from coherence transfer via ^h2J_PH couplings were observed for the Tyr8 and Tyr25 OH groups. ³¹P chemical shifts are with respect to TMP. (B) 1D ³¹P spectrum recorded for the same complex. The positions of the ³¹P resonances observed in the long-range ¹H-³¹P HMQC are indicated by green arrow. (C) Data of 1D spin-echo ^h2J_PH modulation difference experiment incorporating a ¹⁵N filter scheme. The signals from the Y8 and Y25 OH groups in the subspectra recorded with and without ^h2J_PH modulation and a difference spectrum are shown. From these data, the values of | ^h2J_PH | for the Y8 and Y25 OH groups were determined to be 3.0 ± 0.1 and 3.4 ± 0.3 Hz, respectively.

Discussion

In this study, we observed the hydrogen-bond scalar couplings between the Tyr OH and DNA phosphate groups in the Antp HD–DNA complex. Although rapid hydrogen exchange typically precludes observation of hydroxyl ¹H resonances,²⁶ some OH groups of protein side chains and RNA nucleotides have been observed around 5.5–12.4 ppm in previous studies.^{23,41–42,44–48} For example, Löhr et al. observed hydrogen-bond scalar couplings between hydroxyl ¹H and FMN ³¹P nuclei for two Ser and two Thr OH groups in the flavodoxin-FMN complex.²³ Thus, measurements of hydrogen-bond scalar couplings are feasible for some OH groups.

Given this feasibility, statistical information about hydrogen bonding between protein side-chain OH and DNA/RNA phosphate groups may be of practical interest for researchers who study protein-DNA or protein-RNA complexes. A statistic study published in 2001 by Luscombe et al.⁴⁹ suggested that hydrogen bonds between protein side-chain OH and DNA phosphate groups are common among protein-DNA complexes. However, the total number of structures of protein-DNA complexes was far less at that time. Therefore, using the currently available 3213 crystal structures of protein-DNA or protein-RNA complexes solved at a resolution < 2.5 Å, we updated the statistics of intermolecular hydrogen bonds involving phosphate groups.

Table 1 show the frequencies of these intermolecular hydrogen bonds in the crystal structures of protein-DNA/RNA complexes. The statistical data confirmed that protein side-chain OH groups are one of the most common hydrogen bond donors to phosphate groups. Interestingly, although OH groups are not charged, the hydrogen bonds between OH and phosphate groups are even more common than those of NH₃⁺-phosphate ion pairs in protein-DNA complexes. On average, there are 2–3 hydrogen bonds between protein side-chain OH and phosphate groups per structure of protein-nucleic acid complex. Thus, hydrogen-bond scalar couplings between protein hydroxyl ¹H and phosphate ³¹P nuclei would be useful for structural and dynamic investigations of protein-DNA/RNA complexes.

Table 1.

Statistics of intermolecular hydrogen bonds of phosphate groups in PDB crystal structures of protein-nucleic acids complexes.

Protein-DNA complexes:		2607 a)
Donor	Acceptor	Hydrogen bonds b)
Arg guanidium	Phosphate	7629
Lys NH3+	Phosphate	3618
Asn/Gln NH2	Phosphate	1929 (1127/802) c)
Ser/Thr/Tyr OH	Phosphate	6475 (2399/2590/1486) c)
Protein-RNA complexes:		606 a)
Donor	Acceptor	Hydrogen bonds b)
Arg guanidium	Phosphate	6702
Lys NH3+	Phosphate	3065
Asn/Gln NH2	Phosphate	1016 (613/403) c)
Ser/Thr/Tyr OH	Phosphate	2219 (902/753/564) c)

^a)Total number of crystal structures with a resolution < 2.5 Å.

^b)Total number of hydrogen bonds identified using a criterion of the N/O…O distance < 3.2 Å.

^c)Numbers in parentheses indicate data for individual types of side chains.

As previously noted,⁸ hydrogen-bond scalar couplings are sensitive to a subtle difference in hydrogen-bonding geometry. Because OH groups are rotatable, their ^h2J_HP couplings should also depend on the bond rotation dynamics. Although these characteristics may cause difficulty in quantitative interpretation of ^h2J_HP couplings for OH groups, molecular dynamics simulations might facilitate interpretation of hydrogen-bond scalar couplings.

Conclusions

We demonstrated an NMR-based identification and characterization of the intermolecular hydrogen bonds between protein side-chain OH and DNA phosphate groups. The interfacial Tyr OH groups exhibited sizable ^h2J_HP couplings across the hydrogen bonds. These Tyr OH groups showed relatively slow hydrogen exchange rates as measured by the CLEANEX-PM method. These Tyr OH groups also exhibited NOE cross peaks with some being intermolecular NOEs (data not shown). Most likely, the same approach would be applicable to protein-RNA complexes as well. Because intermolecular hydrogen bonds between protein side-chain OH and nucleic-acid phosphate groups are common in protein-DNA/RNA complexes, we expect that the NMR-based investigations of this type of hydrogen bonds would facilitate biophysical investigations of many protein-nucleic acid complexes.