Solid-State NMR Shows That Dynamically Different Domains of

Jul 15, 2014 - The highly mobile and immobile domains of DAGK and their water ... to this article, users are encouraged to perform a search inSciFinde...
0 downloads 0 Views 6MB Size
Article pubs.acs.org/JPCB

Solid-State NMR Shows That Dynamically Different Domains of Membrane Proteins Have Different Hydration Dependence Zhengfeng Zhang, Yanke Chen, Xinqi Tang, Jianping Li, Liying Wang, and Jun Yang* Key Laboratory of Magnetic Resonance in Biological Systems, State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics, Wuhan Center for Magnetic Resonance, Wuhan Institute of Physics and Mathematics, Chinese Academy of Sciences, Wuhan 430071, P. R. China S Supporting Information *

ABSTRACT: Hydration has a profound influence on the structure, dynamics, and functions of membrane and membrane-embedded proteins. So far the hydration response of molecular dynamics of membrane proteins in lipid bilayers is poorly understood. Here, we reveal different hydration dependence of the dynamics in dynamically different domains of membrane proteins by multidimensional magic angle spinning (MAS) solid-state NMR (ssNMR) spectroscopy using 121residue integral diacylglycerol kinase (DAGK) in 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC)/1,2-dimyristoyl-snglycero-3-phospho-(1′-rac-glycerol) (DMPG) lipid bilayers as a model system. The highly mobile and immobile domains of DAGK and their water accessibilities are identified sitespecifically by scalar- and dipolar-coupling based MAS ssNMR experiments, respectively. Our experiments reveal different hydration dependence of the dynamics in highly mobile and immobile domains of membrane proteins. We demonstrate that the fast, large-amplitude motions in highly mobile domains are not triggered until 20% hydration, enhanced at 20−50% hydration and unchanged at above 50% hydration. In contrast, motions on submicrosecond time scale of immobile residues are observed to be independent of the hydration levels in gel phase of lipids, and at the temperature near gel−liquid crystalline phase transition, amplitude of whole-molecule rotations around the bilayer normal is dominated by the fluidity of lipid bilayers, which is strongly hydration dependent. The hydration dependence of the dynamics of DAGK revealed by this study provides new insights into the correlations of hydration to dynamics and function of membrane proteins in lipid bilayers.



INTRODUCTION

Although the hydration dependences of lipid dynamics have been investigated, studies of the influence of hydration on the dynamics of membrane proteins in lipid bilayers are rare and incomplete12,13 due to the large complexity of membrane proteins in a heterogeneous system which contains hydration water, lipids, and protein components. By observing different dynamical transition temperatures of hydration water and purple membrane, neutron scattering has reported that the dynamics of purple membrane are not directly coupled with water dynamics below 260 K on the fast time scale.13 However, neutron scattering only allows general descriptions of the hydration dependence of dynamic properties averaged over the whole protein on the picoseconds to nanosecond time scale. In fact, membrane proteins embedded in lipid bilayers generally have domains with largely different dynamics, such as termini or loops with very high mobility and transmembrane domains with restricted mobility.14−18 These dynamically different domains may have different response to the hydration. In

Molecular motions, which give the conformation flexibility of proteins, are essential to the functions of proteins. Dynamics and functions of membrane proteins strongly depend on environmental factors such as temperature, lipid properties, and hydration1−4 since membrane proteins are embedded in or attached to lipid bilayers, which are composed of a mixture of many lipids and surrounded by solvent hydration. The influence of hydration levels on the dynamics of lipids has been extensively studied by experimental and computational methods such as neutron scattering,5 solid state NMR,6−9 and molecular dynamics simulations.10,11 Neutron scattering revealed that the picosecond internal molecular motions of the membrane are affected by the hydration interacting with the lipids.5 2H NMR reported that progressive hydration in lipid 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) in the liquid crystalline phase induces a change of headgroup conformation and increase of lipid mobility.8,9 In recent years, solid-state NMR 2D separated-local-field (SLF) experiments demonstrated that the overall order parameters SH−C of the headgroup of lipid generally decrease with the extent of hydration.6,7 © 2014 American Chemical Society

Received: March 27, 2014 Revised: July 14, 2014 Published: July 15, 2014 9553

dx.doi.org/10.1021/jp503032h | J. Phys. Chem. B 2014, 118, 9553−9564

The Journal of Physical Chemistry B

Article

was further studied by 1HN rotating-frame spin−lattice relaxation time (T1ρ) at various hydration levels. Based on characterization of water accessibility and measurements of dynamical parameters, the mechanism of hydration dependence of the dynamics of DAGK in lipid bilayers was revealed. The hydration dependence of the dynamics of DAGK on a broad time scale provides new insights into understanding the correlations between hydration and functions of membrane proteins in native-like environments. In addition, the hydration dependence of ssNMR spectra demonstrates the feasibility of using hydration-defined membrane protein samples in multidimensional MAS ssNMR experiments.

addition, since proteins are in a closely associated hydrationlipid-protein environment, hydration dependence of dynamics of membrane proteins should be influenced by lipid properties such as phases and ratios of lipid to protein. To our best knowledge, so far there are no experimental data available for addressing these fundamental issues. Escherichia coli diacylglycerol kinase (DAGK) is a good model protein suited for studying structure and functions of transmembrane proteins.19 Wild type DAGK is very stable and functions as a 40 kDa homotrimer. Its function is phosphorylation of diacylglycerol by Mg(II)−adenosine triphosphate (Mg-ATP) to form phosphatidic acid. As a model protein, expression, solubilization, purification, reconstitution, and molecular mechanism of DAGK have been extensively studied;20−28 and high-resolution 3D structures in dodecylphosphocholine (DPC) micelle and in monoolein lipidic cubic phase have been resolved by solution NMR29 and X-ray crystallography,30 respectively. Magic angle spinning (MAS) solid-state NMR (ssNMR) is a powerful tool for studying the dynamics and interactions of membrane proteins in lipid bilayers environment.31,32,16,33−38 The unique advantage of ssNMR on studies of the dynamics of proteins is that the function relevant motions on the nanosecond to millisecond time scales can be directly observed by motionally-averaged anisotropic interactions, whereas in solution NMR these anisotropic interactions are averaged by rapid isotropic molecular tumbling.39,40 Owing to the great progress of protein SSNMR spectroscopy in recent years, including applications of high or ultrahigh magnetic fields, multidimensional MAS ssNMR experiments, and homogeneous sample preparations, MAS ssNMR can provide residue sitespecific dynamics on a broad time scale from picosecond to second based on resonance assignments. In addition, MAS ssNMR is capable of distinguishing highly mobile and relatively rigid domains using scalar- and dipolar-coupling based experiments.18 In this study, we employed several MAS ssNMR techniques such as mobility filters by scalar- or dipolar-coupling based experiments, analyses of line shapes modulated by anisotropic interactions, and relaxation measurements, combining with differential scanning calorimetry (DSC) to investigate hydration dependence of the dynamics of membrane proteins using DAGK as a model system. We investigated the variation in resolution of MAS ssNMR spectra of U−13C, 15N DAGK as a function of hydration levels. A series of scalar- and dipolarcoupling based 2D and 3D chemical shift correlation experiments were used to assign the backbone resonances of highly mobile and immobile residues of DAGK, respectively. We refer to highly mobile residues detected by scalar-coupling (J-coupling) based experiments as “J-residues”, whereas immobile residues detected by dipolar-couplings based experiments as “D-residues”. Based on resonance assignments, the sitespecific water accessibilities of J- and D-residues were characterized by comparison of 1D-3D NMR spectra before and after H/D exchange. The 15N INEPT spectra demonstrated that the fast, large-amplitude motions of highly mobile residues are not triggered until 20% hydration, enhanced at 20−50% hydration and unchanged at above 50% hydration. The amplitudes of the motions on the submicrosecond time scale were monitored by motionally averaged chemical shift anisotropies (CSA) of carbonyl carbons (C′) and 1H−15N dipolar coupling in gel phase as a function of hydration levels. The rate of backbone motions on the microsecond time scales



EXPERIMENTAL SECTION Protein Expression and Purification. DAGK was expressed in E. coli BL21 (DE3) star cells using a pSD005 expression vector which contains a N-terminus His6-tag sequence of MGHHHHHHEL.41 To increase the yield of isotopic labeled protein, we used “dual media approach”42 for the expression. Colonies containing expression plasmid pSD005 were inoculated into LB medium containing 100 μg/ mL ampicillin and grown at 310 K overnight. Then 20 mL cell suspension was transferred into 4 L of LB medium and shaken at 310 K until OD600 reached 0.6−0.8; the cells were harvested and resuspended into 1 L of M9 medium without any glucose and NH4Cl and then collected by centrifugation at 6000g for 10 min. Cells were then suspended in 1 L of M9 (2 g/L 13C glucose, 1 g/L 15N NH4Cl), following by shaking at 310 K for 1 h. At this point, protein expression was initiated by 1 mM isopropyl-b-D-thiogalactoside (IPTG), and then the cell suspension was incubated under the same condition for another 5 h. Cells were harvested by centrifugation and suspended in lysis buffer (75 mM Tris, pH 7.7, 300 mM NaCl) and lysed by ultrasonication. After the removal of inclusion body by centrifugation (8000g for 15 min), Empigen BB detergent (Sigma-Aldrich) was added to the supernatant to a final concentration of 3% (v/v). The supernatant was tumbled for 1 h in order to solubilize DAGK from the cell membrane. Insoluble particles were removed by a further centrifugation step (40000g for 30 min). The resulting supernatant was mixed with 3 mL of Ni-IDA resin, following by gentle tumbling for 2 h. Then the Ni-IDA resin with bounded DAGK was loaded in a 10 mL column, and the flow-through was discarded. The column was then washed using 10 column volumes of lysis buffer with 1.5% Empigen BB and successively re-equilibrated using 12 × 1 column volumes of rinse buffer (20 mM NaPi, pH 7.2, 0.2% DPC (Avanti Polar Lipids Inc.). DAGK protein was finally eluted by 5 column volumes of elution buffer (250 mM imidazole, pH 7.8, 0.5% DPC). The final yield of pure 13C, 15N labeled DAGK is 20−30 mg/L. Reconstitution of Membrane Proteins. Liposome was prepared at a DMPC:DMPG molar ratio of 4:1. About 80 mg of DMPC and 20 mg of DMPG (Avanti Polar Lipids Inc.) were codissolved in 3 mL of CHCl3/CH3OH (v/v: 4/1) in a roundbottom flask. The mixture was dried under N2 stream and transformed into a thin lipid film on the sides of the roundbottom flask, and then it was placed under high vacuum overnight for further removal of residual organic solvent. 10 mL of double distilled H2O was added to the dried mixture, and the flask was vortexed in a shaker for 2 h at 310 K. A clear solution was prepared after several freeze (liquid nitrogen)−thaw (310 K)−bath sonicated cycles and extruding the solution through 9554

dx.doi.org/10.1021/jp503032h | J. Phys. Chem. B 2014, 118, 9553−9564

The Journal of Physical Chemistry B

Article

Figure 1. Resolution of the NCA spectra of U−13C, 15N DAGK in DMPC/DMPG lipid bilayer (molar ratio of lipid:protein = 1:15, LP15) as a function of hydration levels. 1D 1H single pulse (A), 2D NCA (B), and 15N slices (C) extracted from NCA spectra along the dashed line were plotted according to hydration levels. The average line width of 15N signals resolved in NCA spectra at 10%, 20%, 30%, and 50% hydration levels are 148, 105, 87, and 82 Hz, respectively. The experiments were carried out at 283 K. All 2D NCA spectra were acquired and processed by the same parameters.

polycarbonate filter membranes with 0.1 μm pores. Prepared liposome was added to purified DAGK in DPC to a lipid:protein molar ratio of 15:1 (LP15) and 100:1 (LP100). 20% n-octyl-β-D-glucopyranoside (OG, purchased from Anatrace) stock solution was added to a final concentration of 50 mM. After incubation at 310 K for 12−18 h, the solution was transferred into a 8000−12000 Da cutoff dialysis bag and dialyzed against 2 L of reconstitution buffer (20 mM NaAc, 10 mM MgSO4, 0.2 mM DTT, pH 4.5) at 291 K. To confirm the complete removal of DPC in the dialysis, we monitored the presence of DPC in the dialysis bags by 1H solution NMR spectra before dialysis, after 3 and 5 days dialysis (Supporting Information Figure S1). These spectra demonstrate complete removal of DPC after 5 days dialysis. The final liposome was pelleted by ultracentrifugation (548352g or 80000 rpm for 2 h) in type 90 Ti rotor (Beckman Instruments, Inc.). The pellet was packed into 4 mm zirconia MAS rotors.

Hydration of Samples. The liposome samples of DAGK pelleted by ultracentrifugation contain ∼200% water, which are referred to as “full hydration” in this study. Here the hydration level is defined as the ratio of water:(protein + lipid) (w/w) in samples. The lipid bilayers could be dried to a certain hydration level by lyophilization or gas stream, and then a certain hydration level could also be controlled by (1) explicit addition of water or (2) incubated at the desired relative humidity controlled by different saturated salt solutions such as CaCl2, CuCl2, KCl, and K2SO4, etc. Although different methods can be used to control hydration levels of lipid bilayers, a previous study has demonstrated that ssNMR spectra of membrane proteins do not depend on the methods used for control of hydration.12 In this study, the fully hydrated pellets in the rotors were first lyophilized until no evident decrease of sample weight was observed. The dried liposome samples were controlled to different hydration levels such as 10%, 20%, 30%, or 50%, etc., by explicitly adding water into the rotors, 9555

dx.doi.org/10.1021/jp503032h | J. Phys. Chem. B 2014, 118, 9553−9564

The Journal of Physical Chemistry B

Article

inhomogeneous conformations of the proteins and very poor resolution of the MAS ssNMR spectra.14,55−57 Our results are in agreement with these previous reports. The broad lines dramatically narrow with average 15N line widths of the resolved peaks being 148, 105, 87, and 82 Hz in NCA spectra of the samples with 10%, 20%, 30%, and 50% hydration, respectively. The line widths are not obviously reduced by further addition of water from 50% to 100%. The resolution of the NCA spectra of LP100 DAGK shows the similar hydration dependence (see Figure S2). The line widths of 13C, 15N signals in NCA spectra of the U−13C, 15N DAGK samples with 30− 50% hydration are comparable to that of other membrane proteins in lipid bilayers.17,43,58−60 To examine whether there is a difference between the structures of protein with 30% hydration and that with full hydration, we recorded 2D 13C−13C DARR and NCA spectra of U−13C, 15N DAGK with two hydration levels as shown in Figure 2. In spite of that a number of peaks are absent in the spectra of the fully hydrated sample probably due to lower sensitivity or higher mobility when compared to the spectra of 30% hydration, all the chemical shifts of peaks in both DARR and NCA spectra are almost identical, indicating nearly identical structures of the protein under two hydration states. Owing to its high resolution of the recorded spectra and

which were then centrifuged for several minutes in order to realize well mixing of water and dried liposome samples. Equilibrium of hydration water in the rotor after 1 h MAS spinning43 was evident by the unchanged line width of 1H NMR signals of lipid and a stable tuning of the NMR probe. Differential Scanning Calorimetry. DSC measurements of LP100 DAGK with different hydration levels were performed using a PerkinElmer DSC-7 instrument. The temperature range from 283 to 333 K was scanned at a rate of 1 K/min. MAS ssNMR Experiments. All the NMR experiments were carried out on a wide-bore Varian VNMRS 600 MHz (14.1 T) NMR spectrometer using a 4 mm triple-resonance T3-HXY MAS probe. The chemical shifts were referenced with respect to adamantane used as external referencing standards (40.48 ppm for the methylene carbon44). A series of dipolar- and scalar-coupling based experiments were conducted in this study. Among the dipolar-based 2D and 3D experiments, the polarization transfer between 13C and 15N was through SPECIFIC CP (spectrally induced filtering in combination with cross-polarization),45 and 13C−13C correlations was established by DARR (dipolar assisted rotational resonance)46 recoupling mixing. The scalar-coupling based experiments use refocused INEPT (insensitive nuclei enhanced by polarization transfer) to set up the initial polarization and use TOBSY (total through-bond correlation spectroscopy)47 to establish 13C−13C correlations. In 2D xCSA experiment, the method proposed by Gan et al. was used for recoupling of C′ CSA.48 In the 2D dipolarchemical shift (DIPSHIFT) experiments,49 R1817 was used for recoupling of heteronuclear dipolar coupling.50 The 1HN T1ρ was measured under 60 kHz 1H effective Lee−Goldburg (LG) spin-lock with the pulse sequence used by Hong et al.51 The xCSA and DIPSHIT experiments were carried under 8 kHz spinning rate, and T1ρ experiments were under 11.111 kHz. The temperature of NMR experiments at different MAS speeds was calibrated in separate experiments using the lead nitrate sample.52 Typical 90° pulse lengths of 3.5 μs (1H), 4.5 μs (13C), and 5.9 μs (15N) were used in MAS ssNMR experiments. The ∼65 kHz 1H radio-frequency (RF) field was used for decoupling during the acquisition and indirect evolution periods in the dipolar-coupling based 2D/3D experiments, while the ∼35 kHz 1 H RF field was used for decoupling in scalar-coupling based experiments. All the multidimensional data were processed in NMRpipe53 and analyzed in Sparky. The SPINEVOLUTION54 was used for simulations of C′ CSA and dipolar coupling line shapes.



RESULTS Hydration-Dependent MAS ssNMR Spectra. To investigate the hydration dependence of MAS ssNMR spectra, we recorded a series of scalar- and dipolar-coupling based NMR spectra of DAGK samples with various hydration levels. For the D-residues, we used 2D NCA experiments to monitor the resolution of the spectra. The 2D NCA spectra of LP15 and LP100 U−13C, 15N DAGK with 0−50% hydration levels acquired at 277 K are shown in Figure 1 and Figure S2 of the Supporting Information, respectively. Very broad lines of 13C and 15N signals in the spectra of lyophilized sample indicate the inhomogeneous conformations of the protein with almost complete dehydration. It is well-known that dehydration of the membrane proteins by freezing of bulk water results in

Figure 2. Comparisons of 2D spectra of 200% (red) and 30% (blue) hydrated LP15 U−13C, 15N DAGK in DMPC/DMPG lipid bilayers. Well overlap of the peaks in 2D 13C−13C DARR (A) and NCA (B) spectra at two hydration levels indicates that DAGK structures are virtually identical at 30% hydration and natively full hydration. 9556

dx.doi.org/10.1021/jp503032h | J. Phys. Chem. B 2014, 118, 9553−9564

The Journal of Physical Chemistry B

Article

preservation of the native structure, U−13C, 15N DAGK sample with 30% hydration was then used for a series of dipolarcoupling based 3D experiments for resonance assignments. Besides the D-residues as mentioned above, a number of Jresidues were detected by INEPT experiments. Very fast and large-amplitude motions on the submicrosecond time scales lead to isotropic “solution-state”-like spectra due to efficient averaging of anisotropic interactions. This enables observation of J-residues by NMR experiments based on scalar couplings such as INEPT and TOBSY. To investigate the influence of hydration content on the mobility of these J-residues, we conducted a series of 15N INEPT experiments of U−13C, 15N DAGK at various hydration levels. Only J-residues are able to be detected by INEPT experiments since rigid 1H will dephase rapidly in spin-echo delays without 1H decoupling in the INEPT experiments,18 giving rise to no signal. Moreover, the intensities of INEPT signals are effective indicators of their mobility. The 1H of higher mobility will produce a stronger intensity because more 1H magnetization can survive spin-echo delays without 1H decoupling and then results in stronger

Information). Experiments were performed at 283 K to ensure that samples are in gel phase but without freezing of water, maximizing sensitivity and resolution. Besides typical 3D experiments such as NCACX, NCOCX, and CONCA for backbone walk, another 3D experiment such as CAN (CO) CX61,62 was proved useful to connect sequential residues. Representative sequential assignments by 3D experiments are shown in Figure S3 of the Supporting Information. Further, we also assigned 16 J-residues (including 10 residues of the Histag) by scalar-coupling based experiments (Supporting Information Figures S4 and S5). All these J-residues were observed located on the N-terminus with coil secondary structure according to an analysis of their chemical shifts. All these assigned D- and J-residues of DAGK are labeled in Figure 4. Water Accessibility by H/D Exchange. To detect water accessibility of the protein, we conducted H/D exchange63,64and corresponding NMR experiments. When being incubated in the D2O solution, amide 1H of the protein can be exchanged to 2H partially or completely if the residues can be accessed by water, leading to a reduction of the sensitivities or disappearance of the signals in the 1H/15N INEPT or 1 H/15N CP spectra compared to those of without H/D exchange, while the intensities will not be affected by H/D exchange if the residues cannot be accessed by water. After three cycles of exchange of D2O, as shown in Figure 5, almost all the 15N signals in INEPT spectrum have disappeared, except a signal at ∼112 ppm with a reduced intensity, indicating that all the J-residues can be accessed by water. In addition, the water-accessibility of D-residues were studied by comparison of NCACX 3D spectra of U−13C, 15N DAGK before and after H/ D exchange (Supporting Information Figure S6), and the corresponding NCA 2D spectra are shown in Figure 5. In these 2D and 3D experiments, a short contact time of 0.2 ms was used for initial 1H/15N CP to prevent unwanted multibond 1 H/15N magnetization transformation. All these water-accessible D- and J-residues are labeled in cyan cycles in Figure 4. Hydration Dependence of the Dynamics of DResidues. DSC. Dynamics of membrane proteins in lipid bilayers is tightly correlated to the lipid phase. DSC is an effective tool to characterize the phase transition of lipid bilayers. To investigate the effect of hydration content on the temperature (Tm) of gel to liquid crystalline phase transition, we recorded DSC thermograms of LP100 DAGK samples at different hydration levels as shown in Figure 6. The DSC data display the strong dependence of Tm on the hydration levels, which is in good agreement with the previous studies.8,9 An elevated Tm of 14 K was observed from 100% to 10% hydration with half-width of ∼5 and 8 K for 100% and 10% hydration sample, respectively. All transitions profiles exhibit asymmetric shapes, which can be decomposed into multiple transition components,35 suggesting the coexistence of gel, liquidcrystalline, or intermediate phases at the broad transition temperature range. C′ CSA. C′ CSA is a useful probe to detect the amplitude of motions on the submicrosecond time scale. To investigate the effect of hydration levels on the amplitude of the fast motions, we measured C′ CSA by using 2D xCSA experiments at 277 and 309 K for LP15 and LP100 U−13C, 15N DAGK, respectively. xCSA is an elegant method for recouping of the CSA of U−13C, 15N-labeled protein at a medium MAS rate and medium 1H decoupling power (