Beyond Structural Biology to Functional Biology: Solid-State NMR Experiments and Strategies for Understanding the M2 Proton Channel Conductance

Abstract

In terms of structural biology, solid-state NMR experiments and strategies have been well established for resonance assignments leading to the determination of three-dimensional structures of insoluble membrane proteins in their native-like environment. It is also known that NMR has the unique capabilities to characterize structure-function relationships of membrane-bound biological systems beyond structural biology. Here, we report on solid-state NMR experiments and strategies for extracting functional activities on a sub-msec time scale. Specifically, we use the His37-labeled full length M2 (M2FL) protein of Influenza A virus embedded in synthetic lipid bilayers as an example to characterize the proton conduction mechanism and kinetics. The integral membrane M2 protein assembles as a tetrameric bundle to form a proton-conducting channel that is activated by low pH and is essential for the viral lifecycle. Our results present convincing evidence for the formation of imidazolium-imidazole hydrogen-bonds in the His37 tetrad at low pH and that these hydrogen bonds have a low barrier that facilitates the proton conduction mechanism in the M2FL protein. Moreover, it has been possible to measure hydronium ion exchange between water and the protons in the His37 NH bonds based on chemical exchange spectroscopy with minimized spin diffusion. The results identify an exchange rate constant of ~4000 s⁻¹ for pH 5.8 at −10°C.

TOC image

1. INTRODUCTION

One of the major frontiers in structural biology is membrane proteins, whose structure, dynamics and function are different from water-soluble proteins. They exist in a heterogeneous lipid bilayer environment exhibiting diverse conformations and conduct many essential biological processes, such as inter- and intra-cellular signal transduction, protein localization, and trafficking requiring synergistic effects between the proteins and their surrounding complex environments^1–3. In the past decades, high-resolution solid-state magic-angle-spinning (MAS) NMR has become a proven powerful technique for characterizing these insoluble membrane proteins and their interactions with the lipid environs^4–9. From a structural biology point of view, an assignment of resonances is mandatory in order to precisely determine high-resolution three-dimensional protein structures. However, spectral resolution in these proteins is a major hurdle because of the uniform low dielectric environment in the transmembrane (TM) domain that often yields highly uniform helical structure resulting in very little chemical shift dispersion. The high hydrophobic amino acid content can yield significant tertiary structural heterogeneity and the hydrophilic content can lead to multiple interactions with the heterogeneous lipid interface resulting in opportunities for exchange dynamics, which can broadens resonances. In the sample preparation for structural biology the interactions with various membrane mimetic environments can lead to spectral differences in the structural models generated^10–13.

Nevertheless, adapting the concepts used in solution¹⁴, solid-state MAS NMR experiments and strategies for the resonance assignments, an essential step towards structural determination of uniformly ¹⁵N and ¹³C labeled proteins, have been well established¹⁵, routinely requiring multi-dimensional solid-state MAS NMR techniques: 1) identifying ¹³C resonances of all amino acid types based on homo-nuclear ¹³C-¹³C correlation^16–22; 2) obtaining intra-residue assignment of ¹⁵N resonances via NCA experiments; 3) establishing inter-residue sequential assignments^{15, 23–26} to neighboring residues through N_i(CO)_i–1, N_i+1(CO)CA_i, or CA_i(CO)N_i+1, due to the fact that the backbone nitrogen of residue i covalently bonds with the carbonyl (C′) carbon of residue i-1 and the Cα of residue i providing a through-bond linkage between two sequential residues. Often, three-dimensional experiments, in a combined use of the above experiments, such as NCOCX and NCACX²⁷ are performed in order to improve the spectral resolution. In addition, long-range inter-residue interactions are required to provide structural long-range distance restraints, often achieved by homo-nuclear ¹³C-¹³C correlation experiments with long mixing times. Difference spectroscopy²⁸ has been proposed to subtract the spin diffusion resonances of relatively short intra-residue distances from the longer inter-residue distances, leading to a better identification of the inter-residue resonances in crowded two-dimensional (2D) ¹³C-¹³C chemical shift correlation spectra intrinsically associated with a long mixing time. Similarly, relaxation-compensated difference spin diffusion NMR²⁹ is also used to detect the long-range correlations. As NMR has the unique capabilities to characterize proteins with dynamic domains in native-like environments, the question is how to extract the functional information beyond the structural information. Here, we illustrate solid-state NMR experiments and strategies aimed at obtaining dynamics and exchange kinetics for proton channel conductance using the Influenza A full length M2 protein (M2FL) as an example.

The M2 protein is a 97-residue membrane protein with a 22-residue N-terminal and a 51-residue C-terminal segment connected by a single TM helix of 24 residues. It assembles as a tetrameric bundle to form a proton-conducting channel functioning at a slow rate (10²–10³/s) that is activated by low pH and is essential for the viral lifecycle^{30, 31}. The α-helical TM tetramer (residues 25–46) is responsible for the proton conductance that triggers the release of viral RNA into the host cells. This tetrameric TM domain is an important drug target^32–35. The four His37 residues reside near the center of the TM helix and are known to be the heart of the proton conducting channel, the key to the mechanism of proton transport³⁶. To the C terminus of the TM helix is an amphipathic helix (residues 47–62)^37–39 that has been shown to interact with the interfacial region of the lipid bilayer and is believed to be essential for membrane trafficking, localization and viral budding^{40, 41}. So far, several different constructs reconstituted in various lipid environments have been the subject to intensive structural studies in the past decade using both MAS^42–49 and oriented sample^{32, 50–53} solid-state NMR, as well as solution NMR^{37, 54–56} and x-ray crystallography^{57, 58}. It has been found that spectra from M2 vary dramatically depending on the M2 constructs and membrane mimetic environments used in the reconstitution, especially those associated with the conductance mechanism^{49, 59–62}, thus leading to different proton transfer mechanisms. For the conductance domain M2(22–62) in DOPC/DOPE lipids⁴³ and M2(18–60) in POPC and DPhPC^46–49, as well as for the M2FL protein in both DOPC/DOPE lipids and E. coli membranes⁴², a set of two ¹³C resonances for His37 was observed, suggesting the histidine tetrad exhibits a dimer of dimer conformation. Recently, in the pH titration study of the M2FL including ¹⁵N, ¹⁵N-¹³C and ¹³C-¹³C spectra of liquid crystalline lipid bilayer preparations, the derived first two pKa’s have identical values of 6.3±0.1, which were significantly lower than those obtained from different M2 constructs^{36, 48, 49, 59, 63}. Furthermore, the ¹H-¹⁵N heteronuclear correlation (HETCOR) spectra of the His37 sidechains show that the ¹⁵N resonances spread from 165 to ~200 ppm while their correlated ¹H frequencies extend up to 19 ppm⁶⁰. Such high ¹⁵N and ¹H frequencies conclusively indicate the formation of short imidazole-imidazolium H-bonds⁶⁴ in the histidine tetrad at low pH. This confirms the model of the so-called low-barrier hydrogen bond (LBHB), i.e., where the shared proton can transit across the potential energy barrier within each pair of His37 residues^{36, 51, 60}. One-dimensional (1D) pure chemical exchange measurements proposed recently⁶⁵ also confirm that the hydronium ions are in exchange with protons in the His37 NH bonds at the heart of the M2 proton conduction mechanism, with an exchange rate constant of ~1750 s⁻¹ at pH 6.2 and −10°C.

On the other hand, the M2 TM domain M2(22–46) in a virus-envelope-mimetic lipid membrane shows quite a different set of His37 resonances⁶³ that does not support amantadine binding. In the ¹H-¹⁵N HETCOR spectra of this preparation at a similar pH⁶¹, the ¹⁵N peaks were in the range of 160–180 ppm, correlating to ¹H chemical shifts of 8–12 ppm, similar to the backbone amide ¹H chemical shift range. No imidazole-imidazolium cross peaks were observed in the ¹³C-¹³C correlation spectra of the +2 charged channel state⁶³. Therefore, an alternative proton conducting mechanism^28,29 with the His37 shuttling protons through imidazole ring reorientations and exchanging protons with water, without forming any inter-monomer hydrogen bond between the His37 residues^{61, 63} was proposed. It is noteworthy that the inclusion of the cytoplasmic tail in the M2 construct (residues 21–97) in the virus-envelope-mimetic lipid membrane binds amantadine⁵⁹. Recently, it has also been shown that cholesterol stabilizes the amphipathic helix in the lipid interface⁶⁶.

Slow chemical exchange and spin diffusion are two main mechanisms that result in cross-peaks in homonuclear chemical shift correlation experiments. Chemical exchange represents kinetic processes within the molecules and can only be manipulated by temperature, not by any experimental techniques. The buildup of the cross-peak intensities in the exchange spectroscopy (EXSY)⁶⁷ solely depends on the chemical exchange rate constant. On the other hand, spin diffusion is primarily dependent upon the strength of through-space homo-nuclear dipolar interactions (proportional to 1/r⁶, where r is the distance between the two spins)⁶⁸. However, when spin diffusion is mediated by abundant spins (e.g. protons), such as in proton driven spin diffusion (PDSD)⁶⁸ and the rotary resonance conditions^{17, 19, 20, 13}C-¹³C spin diffusion is greatly enhanced and consequently becomes one of the most useful tools for obtaining homonuclear ¹³C-¹³C chemical shift correlation spectra, an important building block for obtaining carbon-carbon distance restraints for protein structural elucidation. Typically, a short mixing time (such as a few milliseconds) can generate cross peaks between directly bonded carbons to establish the intra-residue connectivity. Therefore, the chemical exchange information is severely buried in the ¹³C-¹³C correlation spectra. Here, we will use uniformly ¹³C,¹⁵N-labeled M2FL protein at pH 5.8 to demonstrate how to minimize spin diffusion induced cross peaks so that the chemical exchange induced imidazole-imidazolium cross peaks from the His37 tetrad could be observed in the homonuclear correlation spectra. We will then summarize our recent experiments to analyze the dynamic process in this sample.

2. EXPERIMENTAL METHODS

Protein expression, purification and reconstitution

M2FL (H_57,90Y) protein expression, purification and reconstitution in DOPC/DOPE liposomes are detailed in the literature^{28, 60}. It is important to note that in the uniform ¹³C and ¹⁵N labeling media for protein expression, unlabeled Phe, Tyr, and Trp amino acids were added to the bacterial cultures to suppress all of the aromatic amino acid resonances except those from the His37 tetrad.

Solid-State NMR Spectroscopy

The proteoliposome pellet containing approximately 10 mg of the M2FL protein was packed into a 3.2-mm thin-wall rotor (36 μL sample volume). Figure 1 shows the pulse sequence for 2D homo-nuclear correlation experiments with various irradiation schemes (PARIS¹⁹, no irradiation, and high power ¹H SPINAL64 decoupling⁶⁹) during the mixing time. The 2D ¹³C-¹³C correlation spectra and 1D ¹⁵N spin-echo spectra were acquired on a Bruker Avance 600.1 MHz NMR spectrometer using an NHMFL 3.2 mm Low-E triple-resonance biosolids MAS probe^{70, 71}. The sample spinning rate was controlled by a Bruker pneumatic MAS unit at 12.2 kHz ± 3Hz. The ¹³C magnetization was enhanced by cross polarization (CP) with a contact time of 1 ms, during which a ¹H spin-lock field of 50.0 kHz was used and the ¹³C B₁ field was ramped from 38 to 56 kHz⁷². The ¹³C 90° pulse length was 3.0 μs. A SPINAL64 decoupling sequence⁶⁹ with an ¹H B₁ field of 78.0 kHz was used during the t₁ and t₂ dimensions. TPPI was used for quadrature detection in the t₁ dimension⁷³. During the mixing time of t_m=20 ms, 12.2 and 78.0 kHz ¹H B₁ fields were applied for the PARIS and SPINAL64 irradiation, respectively. The acquisition times for the t₁ and t₂ dimensions were 4.53 and 7.74 ms, respectively. The data were zero-filled to a 4096 × 2048 matrix before Fourier transform and were processed with a Gaussian window function (LB=−50 Hz and GB=0.1) in both dimensions. 1D ¹⁵N spin-echo spectra were acquired with a Lee-Goldburg (LG) cross polarization (LGCP) sequence⁷⁴ using the contact time of 4 ms, during which the ¹Hs were spin-locked along the magic angle by the LG sequence. The ¹⁵N 180° pulse length was 8.2 μs and the echo time was set to a multiple of the spinning periods. A SPINAL64 decoupling sequence⁶⁹ with a ¹H B₁ field of 78.0 kHz was applied during the spin echo and data acquisition times.

Figure 1.

Pulse sequence used for homonuclear correlation experiments. During the mixing time t_m, three different irradiation methods are applied: PARIS (grey), no irradiation (i.e. PDSD), and SPINAL64 (purple).

The 2D ¹⁵N-¹⁵N and ¹H-¹⁵N correlation experiments and 1D dipolar-dephased ¹⁵N measurements were performed on a 63 mm mid-bore 800 MHz magnet equipped with a Bruker Avance NMR console. An NHMFL 3.2 mm low-E triple-resonance biosolids MAS probe⁷⁰ was used with a ¹H-¹⁵N double-resonance configuration optimized for ¹⁵N observation. The sample spinning rate was controlled by a Bruker pneumatic MAS unit at 13 kHz ± 3 Hz. In the 2D ¹⁵N-¹⁵N correlation experiments, the ¹⁵N magnetization was enhanced by LGCP with a contact time of 3 ms, during which the ¹Hs were spin-locked along the magic angle by the LG sequence and the ¹⁵N B₁ field was ramped from 38.0 to 54.0 kHz⁷². The ¹⁵N 90° pulse length was 5.0 μs. A SPINAL64 decoupling sequence⁶⁹ with a ¹H B₁ field of 83.3 kHz was used during the t₁ and t₂ dimensions. TPPI was used for quadrature detection in the t₁ dimension⁷³. No ¹H irradiation was applied during the mixing time t_m=50 ms. The acquisition times for the t₁ and t₂ dimensions were 2.4 and 10.3 ms, respectively. In the 2D ¹H-¹⁵N HETCOR experiments, the ¹H homo-nuclear decoupling was achieved using the phase-ramped frequency-switched LG sequence⁷⁵ in the t₁ dimension with an 83.3 kHz B₁ field, corresponding to 102.0 kHz decoupling amplitude along the magic angle. A short contact time of 200 μs was used to transfer ¹H magnetization to ¹⁵N for detection under the SPINAL64 decoupling⁶⁹ with a ¹H B₁ field of 83 kHz. During the short contact time, the ¹H magnetization was spin-locked along the magic angle to further minimize long-range ¹H->¹⁵N transfers. STATES was used for quadrature detection in the t₁ dimension⁷³. For the 1D dipolar-dephased ¹⁵N measurements as diagramed in the literature⁶⁵, the experimental parameters used were as follows: ¹H 90° pulse length of 3.0 μs, ¹⁵N 180° pulse length of 10.0 μs, and a total dephasing time of four rotor periods (i.e. 307.3 μs) was used in the REDOR-based dipolar dephasing before the LG spin-lock (LGSL); during the LGSL, a ¹H B₁ field of 53.2 kHz was applied at an offset of +37.6 kHz resulting in the effective spin-locking field of 65.2 kHz along the magic angle. After a given LGSL time t_SL, a short contact time of 200 μs was used to transfer the ¹H magnetization to its near-by ¹⁵N site for monitoring by ramping the ¹⁵N B₁ field from 38.0 to 54.0 kHz. The ¹⁵N signals were then acquired under the SPINAL64 decoupling⁶⁹ with a ¹H B₁ field of 83 kHz. The number of scans used to accumulate the signals was 20,480 with a recycle delay of 1s.

For both 600 and 800 MHz probes, the temperature calibrations were performed in separate experiments by observing ²⁰⁷Pb chemical shift in a dilute lead nitrate (~50%) sample at their respective spinning rates, and all experiments were carried out at the calibrated temperature of −10°C. The ¹³C chemical shifts were referenced to the carbonyl carbon resonance of glycine at 178.4 ppm relative to TMS, while the ¹⁵N chemical shifts were referenced to 34.1 ppm of the glycine ammonium peak, relative to TMS based on the known relative frequency ratio between TMS (¹H) and liquid ammonia (¹⁵N).

3. RESULTS AND DISCUSSIONS

¹³C-¹³C chemical shift correlation spectra

Figure 2 shows the aromatic/aromatic region of the 2D ¹³C-¹³C correlation spectra of the His37-labeled M2FL protein (pH 5.8) in DOPC/DOPE lipid bilayers with a mixing time of 20 ms. Since the ¹³C-¹³C spin diffusion is relatively efficient when mediated by abundant protons (such as PDSD⁶⁸, DARR¹⁷, and PARIS¹⁹), the connectivity for all carbons in close vicinity may be established with a short mixing time (e.g. 20 ms). However, when the protons are decoupled, the proton mediation is effectively cut off such that the spin diffusion from one carbon to others become insufficient, unless a rotational resonance condition is fulfilled⁷⁶. As documented in the SI (Figure S1), the cross peak intensities from Cβ to other carbons in the ¹³C-labeled Fmoc-valine sample were less than 1% of the diagonal peak when SPINAL64 was applied during the mixing time, while when both PARIS and PDSD were used in the mixing time, the cross peak intensities were much stronger. For the M2FL protein at pH5.8, the ¹³C-¹³C correlation spectrum (with PARIS irradiation) clearly shows the cross peaks between the neutral His37 (τ) and charged His37(+) residues, i.e. Cγ/ε1(τ)-Cγ/ε1(+) as indicated in Figure 2A. These assignments were adopted from our previous work⁶⁰, where the full ¹³C-¹³C correlation spectrum at pH 5.8 was shown. As the four His37 residues are tightly packed at the heart of the proton conducting channel, whether these cross peaks are solely from the dipolar coupling based spin diffusion or they have also some contributions from the kinetic processes between the His37 residues is yet to be answered.

Figure 2.

(Top) Schematics for the τ and charged histidine sidechains. (Bottom) Aromatic/aromatic region of 2D ¹³C-¹³C correlation spectra of the His37-labeled M2FL (pH 5.8) in DOPC/DOPE liposomes with a mixing time of 20ms, during which the PARIS (A) and SPINAL64 (B) irradiation were applied on the ¹H channel. The circled regions indicate the cross peaks between Cγ/ε1(τ) and Cδ2(τ) from the His37(τ) residues.

Figure 2B shows the ¹³C-¹³C correlation spectrum with the SPINAL64 decoupling during the mixing time. Obviously, the cross peaks between Cγ/ε1(τ) and Cδ2(τ) from the intra-His37(τ) residues disappear in the spectrum (circled regions), implying that the dipolar coupling induced spin diffusion is sufficiently suppressed. Since the inter-residue ¹³C-¹³C distances are longer than the intra-residue ¹³C-¹³C distances, it is expected that the dipolar coupling induced inter-residue ¹³C-¹³C spin diffusion will also be completely suppressed. As shown in Figure 2B, the cross peaks Cγ/ε1(τ)-Cγ/ε1(+) remain, although their intensities are weakened compared to Figure 2A. Thus, such an observation is clear evidence that the neutral His37(τ) and charged His37(+) residues are inter-changing chemically.

¹⁵N-¹⁵N chemical shift correlation spectrum

Since ¹⁵N’s gyromagnetic ratio is 2.5 times smaller than ¹³C’s, the ¹⁵N5¹⁵N dipolar interaction is approximately 6.25 times less compared to a ¹³C-¹³C pair at the same distance. In addition, ¹⁵N-¹⁵N distances are greater than many ¹³C-¹³C distances in protein structures and hence spin diffusion resonances are rarely observable⁷⁷. However, the molecular dynamic processes are not affected by a given spin type and therefore it is relatively easy to observe the chemical exchange processes through ¹⁵N-¹⁵N correlation spectra. Figure 3 shows the 2D ¹⁵N-¹⁵N correlation spectra of the His37-labeled M2FL protein in DOPC/DOPE lipid bilayers at pH 5.8 using a mixing time of 50 ms. Two cross peaks between the protonated nitrogens at ~166 ppm and the non-protonated nitrogens at 248.8 and 242.0 ppm are observed. As in a control experiment documented in the SI (Figure S2), no intra-histidine ¹⁵N cross peaks could be observed for both the neutral τ and charged histidine in the ¹⁵N-¹⁵N correlation spectrum with a mixing time of 50 ms, even without ¹H irradiation during the mixing time. In other words, there is no observable dipolar coupling induced ¹⁵N-¹⁵N spin diffusion cross peak within the histidine sidechain in a 50 ms mixing time experiment. Therefore, the observed cross peaks in Figure 3 originate from inter-residue slow exchange between the neutral His37(τ) non-protonated nitrogen and charged His37(+) protonated nitrogen sites.

Fig. 3.

2D ¹⁵N-¹⁵Ncorrelation (EXSY) spectrum of the His37-labeled M2FL (pH 5.8) in DOPC/DOPE liposomes. No 1H irradiation was applied on the ¹H channel during a mixing time of 50 ms. The red 1D spectrum was taken from the 2D spectrum along the red line at 166.3 ppm.

1D ¹⁵N spin-echo spectra

It is worth noting from the diagonal signals in Fig. 3 that the protonated ¹⁵N signals are dispersed from 160 ppm up to ~ 200 ppm, while the non-protonated ¹⁵N signals range from 250 ppm down to 235 ppm, which suggest that either the protonated and non-protonated nitrogen sites are in the process of motional coalescing (exchange), or the various imidazole-imidazolium hydrogen bonds may be present between the neutral His37(τ) and charged His37(+) residues giving rise to a range of frequencies. Shown in the 1D spectra of Figure 4, there are two distinct ¹⁵N spectral features in both the non-protontated and protonated ¹⁵N region. The signals at 249.0 and 166.0 ppm are relatively narrow, while the others appear to be broad or represented by a band of frequencies. For the non-protonated ¹⁵N resonances centered at 242 ppm, the signals extend to higher field (lower ppm) as low as 232 ppm. The protonated ¹⁵N signals at ~174.5 spread towards lower field (higher ppm) up to 200 ppm. Such a phenomenon is typically observed when the two resonances experience a motional averaging at a rate comparable to their isotropic chemical shift difference. The ¹⁵N signals observed here in Figure 4 appear to decay uniformly when the echo time was increased from 318 μs to 2.45 ms. In order to determine whether the broadening is inhomogeneous as a result of a distribution of chemically different neutral and charged His37 conformations, or homogeneous broadening due to the dynamic exchange at a rate comparable to the protonated and non-protonated ¹⁵N chemical shift difference, we measured ¹⁵N transverse relaxation time T₂ in the ¹⁵N spin-echo spectra using a series of echo times. Figure S3 shows the ¹⁵N spin-echo spectra at various echo times. The signal intensities at various chemical shift positions across the spectral range were monitored as a function of the echo times. By fitting the decay of these intensities mono-exponentially, we could obtain the T₂ values at these specific chemical shift positions. For the signals at ~ 249 and ~166 ppm (positions a and e), the measured T₂ values are 15.7 and 12.6 ms, respectively, longer than (at least comparable to) that for the backbone amide nitrogen (position f, 11.7 ms), which do not have a large degree of motion. While the signals at ~ 240, 185, and 175 ppm (positions b, c, and d) have T₂ values of 7.1, 7.1, and 8.2 ms, respectively. The shortest T₂ value of 7.1 ms translates to a natural linewidth Δν of ~45 Hz according to this equation Δν=1/πT₂, corresponding to 0.74 ppm line-broadening at an ¹⁵N frequency of 60 MHz. This natural line-width is much narrower than what appears in the spectra, clearly demonstrating that the broadening is inhomogeneous, meaning that there is a distribution of chemically distinguishable neutral and charged His37 conformations, rather than the motional induced broadening at a rate comparable to the protonated and non-protonated ¹⁵N chemical shift difference.

Figure 4.

Aromatic region of 1D ¹⁵N spin-echo spectra of the His37-labeled M2FL (pH 5.8). A total of 4 (Blue) and 15 (Purple) rotor periods were used for the spin echo times, corresponding to 318 μs and 2.45 ms, respectively.

¹H-¹⁵N HETCOR

Figure 5 shows the 2D ¹H-¹⁵N HETCOR spectrum of the His37-labeled M2FL (pH 5.8) in DOPC/DOPE liposomes. The natural linewidth characterization by T₂ measurements are confirmed by a few well resolved resonances in this HETCOR spectrum supporting the heterogeneous broadening of the resonances and confirming that in these (pH 5.8 and 6.2) preparations the broad band of resonance intensity is not due to exchange broadening. It is evident from the spectrum that the protonated nitrogen frequencies in the His37 residues correlate with water proton frequencies. Such cross peaks between the water resonances and protons in proteins/peptides in ¹H-¹⁵N HETCOR spectra have been observed previously^{61, 63, 78–81}. In particular, Figure 5 clearly shows a few well resolved resonances with exceptionally high ¹H frequencies (up to 19 ppm) and high protonated ¹⁵N frequencies (up to 190 ppm). Such high ¹⁵N and ¹H frequencies strongly support the formation of short imidazole-imidazolium hydrogen bonds⁶⁴. The linear correlation between the high ¹H and ¹⁵N frequencies reflect multiple hydrogen-bonded states between imidazolium and imidazole with a range of distances or geometry. It is noteworthy that the resonances at ~ 13 ppm as observed in the M2FL at pH 6.2⁶⁰ that are believed to reflect the NH protons hydrogen bonded to water do not appear here in Figure 5. This suggests that the lifetime of these states is significantly shorter at pH 5.8 compared to pH 6.2.

Figure 5.

2D ¹H-¹⁵N HETCOR spectrum of the His37-labeled M2FL (pH 5.8) in DOPC/DOPE liposomes at −10°C.

Recovery of dipolar-dephased ¹⁵N signals

In order to investigate how hydronium ions are interacting with the protons in the NH bonds of the His37 residues, we performed the newly proposed experiments⁶⁵ that probe the specific water-protein chemical exchange by spin-locking the ¹H magnetization along the magic angle and monitoring the recovery trajectory of dipolar-dephased ¹⁵N signals. As shown in Figure 6, the ¹⁵N spectra without ¹⁵N dephasing (black) show similar line-shapes and intensities at different ¹H LGSL time, t_SL. As a short contact time (i.e. 200 μs) was used to transfer the ¹H magnetization to ¹⁵N, only the ¹⁵N signals from the protonated ¹⁵N sites of His37 sidechains were cross-polarized (i.e. the τ state Nε2τ and the charged states Nδ1⁺ and Nε2⁺) resulting in the non-protonated τ state Nδ1τ (at ~ 250 ppm) being hardly polarized. Again, the observed ¹⁵N resonances are spread from 165 to ~200 ppm. When the ¹⁵N selective dephasing was applied, the protons of the Nε2τ sites and the charged His37 H5Nδ1⁺ and H-Nε2⁺ sites were dephased at the beginning of the LGSL and hence no ¹⁵N signals from these protonated sites would be expected. However, in the presence of the water-protein exchange, the protons from the τ state Nε2τ and the charged His37 H-Nδ1⁺ or H-Nε2⁺ (presumably H-Nε2⁺) could be re-polarized during the LGCP, such that the signals from these protonated sites could be observed at a short t_SL. As shown in the red spectra of Figure 6, the observed ¹⁵N signals were largely reduced at a short t_SL of 50 μs, especially the signals at ~166 ppm. As t_SL increased, the dipolar-dephased ¹⁵N signals from the protonated ¹⁵N sites gained more intensity, implying that their bonded protons gain magnetization during the LGSL. As ¹H spin diffusion is suppressed during the LGSL and any relayed transfer is largely eliminated, the observed gain can only be facilitated by chemical exchange between this particular proton and the hydronium ions.

Figure 6.

Expanded ¹⁵N spectra of the His37-labeled M2FL (pH 5.8) in lipid bilayers at −10°C without (black) and with (red) ¹⁵N-dipolar dephasing and spinlock time t_SL of 50 and 2000 μs.

A series of 1D dipolar-dephased ¹⁵N signals as a function of t_SL would permit the monitoring of chemical exchange processes between the hydronium ions and the protons in the His37 NH bonds. The recovery trajectory of these dipolar-dephased ¹⁵N signal intensities can be fitted⁶⁵ using the following equation to obtain their exchange rate constant k_IM:

$$I\left(t_{SL}\right)=I(0)+pM\left(0\right)\{1-\exp [-(p+1)k_{IM}{t}_{SL})]\}\exp \left(-t_{SL}/T_{1\rho }^H\right)/(p+1).$$ [1]

Here p is the hydronium ion concentration in the pool of water molecules, $T_{1\rho }^H$ is the spin-lattice relaxation time in the LGSL field for protons in water molecules and in the His37 NH bonds (with the assumption that they have the same spin-lattice relaxation time in the LGSL field), and I(0) represents an initial ¹⁵N signal intensity. This initial term I(0) should include a contribution from any incomplete HN dephasing by REDOR due to various NH bond lengths and re-polarized signals during the LGCP period according to the exchange rate constant and the LGCP contact time.

Figure 7 shows the plot of the dipolar-dephased ¹⁵N signals versus t_SL. In order to reduce the number of the fitting variables, we measured $T_{1\rho }^H$ in separate experiments, yielding $T_{1\rho }^H$ = 15.8 ms (Figure S4). We used this value as a constant to fit the recovery trajectory using Equation 1 to obtain (1+p)k_IM=4,000 ± 1,500 s⁻¹. The concentration p of the hydronium ions in the M2 channel pore is about 10⁻⁶ M with the assumption that the pH in the pore is the same as in the bulk environment. Thus, the exchange rate constant between hydronium ions and the protons in the His37 NH bonds for the M2FL is on the order of 4,000 ± 1,500 s⁻¹ for pH 5.8 at −10°C, which is about twice as fast as for the M2FL for pH 6.2⁶⁵, indicating that the increase of the proton dynamics at lower pH is directly proportional to the proton concentration. Again, this represents an average value over a number of different His37 states with various exposures to hydronium ions. It is anticipated that the exchange model^{60, 65} established at −10° should be relevant at the physiological temperature but the exchange rates should be faster.

Figure 7.

Normalized dipolar-dephased ¹⁵N integral intensities as a function of t_SL for the His37-labeled M2FL (pH 5.8) at −10°C. The red line represents the best fit curve with k_IM=4000 s⁻¹. The dashed green and blue lines are the curves for k_IM=2500 and 5500 s⁻¹, respectively.

Conductance mechanism

Despite the chemical complexity of the His37 states, a model for how the hydronium ions go through the His37 tetrad in the M2 proton channel has been established⁶⁵. Here we use this model to interpret our experimental results in detail. The “initial” NH protons in the His tetrad are colored green and blue for the imidazole-imidazolium hydrogen-bonded His C-D pair, respectively, as shown in the upper left panel of Figure 8. The hydronium ion based proton is colored red. When the hydronium ion is attracted by the non-protonated Nδ1τ site, the neutral His37 residue D becomes charged, so that both C and D residues are charged, consequently breaking the imidazole-imidazolium H-bond (upper right in Figure 8). The two imidazolium residues conformationally rearrange, due to steric clash and charge repulsion with the newly protonated Nδ1⁺H (D) site oriented toward the pool of externally exposed waters, while the original Nε2⁺H (C) and the newly formed Nε2⁺H (D) are both exposed to waters of the viral interior, as illustrated in Figure 9. The return of the His Nδ1⁺H (D) proton to the waters of the viral exterior (path I) results in a futile cycle, as illustrated by the red arrow and red circle in Figure 9, however, both His Nε2⁺H (C and D) protons are accessible by interior waters. The absorbance of either His Nε2⁺H proton by waters of the viral interior (green (II) and blue (III) arrows) result in successful conductance of a proton across the membrane. If the His Nε2⁺H (C) proton (the original imidazole-imidazolium H-bonded proton) is reabsorbed by water (path II in Figure 8), the imidazolium donates its H-Nδ1 (D) proton to reform the same His37 C-D pair with an imidazolium-imidazole hydrogen-bond utilizing a π state. This proton rapidly rearranges crossing the H-bond barrier to form a more stable τ-charge H-bonded pair, as in the original state and as illustrated in the lower right panel of Figure 8. When an interior water reabsorbs the proton of the newly formed Nε2⁺H (D), D becomes a π state and forms a new imidazolium-imidazole H-bond with H-Nδ1 of His A (path III in Figure 8) with the proton rapidly crossing the H-bond barrier to form a τ-charge hydrogen bonded pair, leading to a rotation of the imidazolium-imidazole bonding pairs⁵¹, as shown in the lower left panel of Figure 8. Consequently the imidazolium-imidazole bonding pairs are reestablished and ready to accept next hydronium ion. This process for His37 τ-states accepting protons from hydronium ions is essential for the buildup of the dipolar-dephased ¹⁵N signals as a function of t_SL.

Figure 8.

Chemical exchange model between hydronium ions and the protons in the His37 NH bonds. In the upper left is the +2 state of the His tetrad. The addition of a second charge to a His pair protonating the Nδ1 site from an external water breaks an imidazole-imidazolium hydrogen bond. Three paths are identified from this point. I) returns the Nδ1 proton to waters of the external environment. II) An Nε2 proton is given up to water from the interior resulting in conductance and requiring a proton transfer from the Nδ1 to Nε2 site. III) The other Nε2 gives up its proton to water resulting in a rearrangement of the imidazolium-imidazole pairs.

Figure 9.

Illustration of the His37 and Trp41 sidechain orientations upon the breaking of the imidazole-imidazolium hydrogen bond between a His37 C-D pair in the M2 proton channel. The paths given by numerals I, II, and III are the same as in Figure 8 and the arrows are used to illustrate the proton releasing paths.

It has been clear for decades that the proton conductance in the M2 channel was facilitated by the His37 residues, but the question has been how the His37 tetrad facilitates this process. With the imidazolium-imidazole bonding His-His pairs, the LBHB model was established, i.e. the proton is transferred through the breaking and reforming of the H-bonds between the two pairs of His37 dimers facilitated by hydronium attack^{36, 51, 60}. If the His37 does not form imidazole-imidazolium hydrogen bonds, but only directly hydrogen bonds with water, the proton shuttling mechanism is proposed, in which individual His37 residues shuttles protons by imidazole ring reorientations and proton exchange with water without the process of forming any inter-monomer hydrogen bond between the His37 residues^{61, 63}. Our experimental data presented here strongly supports the LBHB mechanism, as explained in the following.

First, the observation of high ¹H frequencies (up to 19 ppm) and high ¹⁵N frequencies (up to 190 ppm) in the ¹H-¹⁵N HETCOR spectrum (Figure 5) indicates the existence of the imidazolium-imidazole bonds between His-His pairs. The long T₂ values (> 7 ms) for the ¹⁵N resonances at ~240, ~185, and 175 ppm confirm that the broadening is inhomogeneous in nature corresponding to a distribution of chemically distinct His37 conformational states, rather than the exchange-induced homogeneous broadening.

Secondly, it is evident from the ¹H-¹⁵N HETCOR spectrum in Figure 5 that the ¹⁵N resonances at ~ 166 ppm are weakly correlated with water molecules as compared to the signals at ~ 175 ppm. In the dipolar-dephased ¹⁵N spectra in Figure 6, the dephased ¹⁵N signals at ~175 ppm almost recovered to their full intensities with a LGSL time of 2000 μs, while only a fraction of the signals at ~166 ppm regained their intensity. This suggests that the ¹⁵N sites resonating at ~166 ppm are discriminately less accessible by water. However, it is these signals at ~166 ppm that correlate with the non-protonated nitrogen sites at ~248 ppm in the ¹⁵N-¹⁵N EXSY spectrum (e.g. Figure 3). These observations are hard to explain with the proton shuttling mechanism, in which all ¹⁵N sites in either neutral or charged states should have equal accessibility by water, but in the LBHB mechanism unequal accessibility is easy to explain. As illustrated in path III in Figure 8, the hydronium ion from the exterior water goes to the Nδ1⁺H (D), while the initial (i.e. before the breaking of the imidazole-imidazolium hydrogen bond as in the upper left of Figure 8) His37 (D) was a neutral state Nδ1τ (D). Therefore, the ¹⁵N-¹⁵N exchange observed here, in fact, represents the dynamic process between the neutral and charged states in the same His37 residue. This argument is also supported by the ¹³C-¹³C EXSY spectrum in Figure 2B, showing the chemical exchanged resonances between the neutral and charged His37 residues. While in the path II of Figure 8, the proton in the Nδ1⁺H (C) is virtually not accessible at all by the hydronium ions. As the non-protonated and protonated ¹⁵N signals are well separated, it may be possible to characterize the dynamic process taking place only in path III by using 1D ¹⁵N pure exchange spectroscopy^{82, 83}, although the sensitivity is intrinsically low.

Thirdly, the recovery of the dipolar-dephased ¹⁵N signals as a function of LGSL time, as observed here, further discriminates the proton shuttling mechanism^{61, 63} in the M2FL proton channel in lipid bilayers. In our dipolar-dephased ¹⁵N measurements, all protonated ¹⁵N signals are initially minimized and should remain minimal during the LGSL except for those NH bonds whose protons are exchangeable with hydronium ions. As in the proton shuttling mechanism^{61, 63}, the non-protonated ¹⁵N site by accepting a proton from a hydronium ion originating from the viral exterior becomes protonated and thus is not capable of receiving next hydronium ion as it has to remain facing the viral exterior so that the other ¹⁵N site in the same His residue can be ready to release its proton to the viral interior. Furthermore, this proton transport seems to involve only the protonation and de-protonation of the neutral His37 residues. In other words, the protonated ¹⁵N signals from the charged His37 residues should not regain any intensity when dephased, but the protonated ¹⁵N signals from the neutral His37 residues are expected to recover fully. This is contradictory with our experimental data shown in Figures 6 and 7, where the protonated ¹⁵N signals at ~175 ppm from the charged His37 sites almost recover to their full intensities and only a fraction of the signals at ~166 ppm (partly from the neutral His37 residues) have regained their intensity. On the other hand, these experimental observations can be well explained by the LBHB model: 1) the reformed imidazole-imidazolium hydrogen bonding His37 pairs always allow the His37 tetrad to be ready for accepting next hydronium ion. 2) In the His37 tetrad, as illustrated in Figure 8, all the circled protons are dephased in applying the ¹⁵N REDOR dephasing sequence. But those in bold and color highlighted protons would regain their intensities as the hydronium ions are taking their positions through the exchange processes, including both protonated ¹⁵N sites from the neutral and charged His37 residues.

It is worth noting that in path II shown in Figure 8, only the protons in the imidazole-imidazolium H-bonds are in exchange with the hydronium ions. It could be understood as follows: due to the oligomeric helix packing effect, the reformed imidazole-imidazolium hydrogen bond in the lower right panel in Figure 8 may not have the same hydrogen bonding geometry as in the initial imidazole-imidazolium hydrogen bond in the upper left panel, leading to a distribution of hydrogen bonds with different strengths, thus chemically distinct resonances in the ¹H-¹⁵N HETCOR spectrum in Figure 5. This is supported by the presence of the slightly off-diagonal peaks of the protonated ¹⁵N signals in the ¹⁵N-¹⁵N correlation spectrum in Figure 3. This heterogeneity in imidazole-imidazolium resonance frequencies is also supported by several large hydrophobic residues in the helix-helix interface that could generate through various rotameric states slightly difference imidazole-imidazolium geometries resulting in the variety of observed resonance frequencies.

4. CONCLUSION

Unique functional insights associated with the His37 tetrad, the heart of proton conductance in the M2 proton channel, have been achieved by solid-state NMR experiments. It is confirmed through the existence of chemical exchange spectroscopy between the neutral and charged His37 residues using both ¹³C-¹³C and ¹⁵N-¹⁵N EXSY spectra that the exchange may involve an inter-conversion between the neutral and charged conformations of the His37 residues. Although broad protonated and non-protonated ¹⁵N resonance bands were observed, they have relatively long T₂ values, requiring that the band of resonances is inhomogeneous in nature, rather than exchange-induced, thus confirming the presence of chemically distinct conformational states of the His37 residues. Our experimental results provide conclusive evidence for the formation of imidazolium-imidazole hydrogen bonds in the histidine tetrad at low pH values and thus support the LBHB model for the proton conductance mechanism in the full length M2 protein. The recovery trajectory of the dipolar-dephased ¹⁵N signals allows us to determine the exchange rate constant of ~4000 s⁻¹ at pH 5.8 and −10°C between the hydronium ions and the protons in the His37 NH bonds at the heart of the M2 proton conduction mechanism. From a functional biology point of view, unique solid-state NMR experiments, as featured herein can be used to characterize detailed structure-function relationships of membrane-bound species at the water-protein interface, and in particular to understand the rates for hydrogen bond formation and breaking in biological systems during proton transport, such as in the M2 proton channel.

ABBREVIATIONS

MAS

magic angle spinning

TM

transmembrane

M2FL

full length M2

DOPC

1,2-dioleoyl-sn-glycero-3-phosphocholine

DOPE

1,2-dioleoyl-sn-glycero-3-phosphoethanolamine

POPC

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine

E. coli

Escherichia coli

DPhPC

1,2-diphytanoyl-sn-glycero-3-phosphocholine

H-bond

hydrogen-bond

PARIS

phase-alternated recoupling irradiation scheme

DARR

dipolar assisted rotational resonance

SPINAL

small phase incremental alternation

HETCOR

heteronuclear correlation

LBHB

low-barrier hydrogen bond

1D

one-dimensional

2D

two-dimensional

EXSY

exchange spectroscopy

PDSD

proton-driven spin diffusion

CP

cross polarization

LG

Lee-Goldburg

LGCP

Lee-Goldburg cross polarization

LGSL

Lee-Goldburg spin lock

REDOR

rotational-echo double-resonance