TROSY NMR with a 52 kDa sugar transport protein and the binding of a small-molecule inhibitor

Abstract

Using the sugar transport protein, GalP, from Escherichia coli, which is a homologue of human GLUT transporters, we have overcome the challenges for achieving high-resolution [¹⁵N-¹H]- and [¹³C-¹H]-methyl-TROSY NMR spectra with a 52 kDa membrane protein that putatively has 12 transmembrane-spanning α-helices and used the spectra to detect inhibitor binding. The protein reconstituted in DDM detergent micelles retained structural and functional integrity for at least 48 h at a temperature of 25 °C as demonstrated by circular dichroism spectroscopy and fluorescence measurements of ligand binding, respectively. Selective labelling of tryptophan residues reproducibly gave 12 resolved signals for tryptophan ¹⁵N backbone positions and also resolved signals for ¹⁵N side-chain positions. For improved sensitivity isoleucine, leucine and valine (ILV) methyl-labelled protein was prepared, which produced unexpectedly well resolved [¹³C-¹H]-methyl-TROSY spectra showing clear signals for the majority of methyl groups. The GalP/GLUT inhibitor forskolin was added to the ILV-labelled sample inducing a pronounced chemical shift change in one Ile residue and more subtle changes in other methyl groups. This work demonstrates that high-resolution TROSY NMR spectra can be achieved with large complex α-helical membrane proteins without the use of elevated temperatures. This is a prerequisite to applying further labelling strategies and NMR experiments for measurement of dynamics, structure elucidation and use of the spectra to screen ligand binding.

• α-helical
• GLUT
• ligand binding
• membrane transport
• TROSY NMR

Introduction

Membrane proteins represent about 30% of the proteins in all organisms (Fagerberg et al., 2010; Liu & Rost, 2001; Wallin & von Heijne, 1998). They are involved in many vital biological processes, some of which are linked with human diseases (Kurze et al., 2010; Ng et al., 2012; Rosenbaum et al., 2009; Shukla et al., 2012; von Heijne, 2007; Watanabe et al., 2008) and they provide the molecular targets for up to 60% of validated drugs (Drews, 2000; Hopkins & Groom, 2002; Lundstrom, 2006; Overington et al., 2006) whilst remaining a principal target for discovery of new drugs (Bahar et al., 2010; Bakheet & Doig, 2009; Rask-Andersen et al., 2011). Despite this progress, membrane proteins represent less than 1% of structures in the Protein Data Bank owing to the various challenges associated with applying the main biophysical techniques used for high-resolution protein structure determination, X-ray crystallography and solution-state NMR spectroscopy, thus limiting the information available for structure-based drug design. The advent of transverse relaxation optimized spectroscopy (TROSY) (Pervushin et al., 1997) and related NMR experiments (Konrat et al., 1999; Riek et al., 1999, 2002; Salzmann et al., 1998, 1999a, 1999b, 2000; Venters et al., 2002; Wider & Wüthrich, 1999; Yang & Kay, 1999) combined with improved expression systems, larger NMR magnets, higher sensitivity probes and development of isotope-labelling strategies (Craven et al., 2007; Hefke et al., 2011; Kainosho et al., 2006; Maslennikov et al., 2010; Parker et al., 2004; Reckel et al., 2010; Takeda et al., 2010; Trbovic et al., 2005; Skrisovska et al., 2010; Sobhanifar et al., 2010) has led to a recent surge in the number of structures of polytopic integral membrane proteins determined by solution-state NMR spectroscopy (Kang & Li, 2011; Kim et al., 2009; Klammt et al., 2012; Maslennikov & Choe, 2013; Nietlispach & Gautier, 2011; Oxenoid & Chou, 2013; Patching, 2011). For α-helical systems however, solution-state NMR structures have only been determined for proteins with a maximum of seven unique transmembrane-spanning α-helices, with two exceptionally stable proteins: Sensory rhodopsin II (Gautier et al., 2008, 2010) and proteorhodopsin (Reckel et al., 2011). The large majority of solution-state NMR studies of membrane proteins are conducted at elevated temperatures (e.g. 50 °C with the rhodopsins), which are usually detrimental to maintaining functional integrity.

For several years we have been developing expression, labelling and sample preparation strategies to overcome the challenges for achieving high-resolution TROSY NMR spectra for even larger α-helical systems, which we can now demonstrate with the Escherichia coli sugar transport protein GalP (Macpherson et al., 1983). Like its close homologues XylE (Henderson & Baldwin, 2013; Sun et al., 2012) and GlcP_Se (Iancu et al., 2013), GalP (MW 52 kDa) probably has 12 transmembrane-spanning α-helices and is evolutionarily- and functionally-related to the series of human GLUT D-glucose transporters and similar series of sugar transporters vital for metabolism in plants and yeasts (Baldwin & Henderson, 1989; Büttner, 2007; Henderson, 1990; Henderson & Baldwin, 2013; Li et al., 2011; Maiden et al., 1987; Ozcan & Johnston, 1999; Pascual et al., 2004; Thorens & Mueckler, 2010). The GalP protein is especially suited for analysis owing to the high level of expression that can be achieved from minimal media in native inner membrane preparations (up to 50% of total protein) and its easy purification in milligram quantities (up to 10 mg per litre of cell culture) with the aid of a C-terminal His₆-tag (Appleyard et al., 2000; Patching et al., 2004, 2008a, 2008b, 2013; Spooner et al., 1994). Furthermore, there are two experimentally useful small-molecule antibiotic inhibitors of GalP, forskolin and cytochalasin B, with dissociation constants in the range K_D = 0.1–5 μM, as tools to explore binding of ligands (Martin et al., 1994, 1995; Robichaud et al., 2011; Walmsley et al., 1994a).

Materials and methods

General

Chemicals, reagents and media components of the highest available quality were obtained from Sigma-Aldrich Co., Fisher Scientific UK Ltd, Melford Laboratories Ltd, BDH Chemical Supplies or Difco Laboratories, unless where stated otherwise, and used without further purification. All media, buffers and other solutions were prepared using either deionized water or MilliQ™ water, and sterilized by autoclaving or by passage through 0.2 μm Minisart™ high-flow sterile syringe-driven filters (Sartorius) or using vacuum-driven 0.2 μm filters (Stericup™) from Millipore. Total protein determinations employed the method of Schaffner & Weissmann (1973) or a BCA assay using Pierce™ BCA protein assay reagent A from Thermo Scientific. SDS-PAGE was performed by the method of Laemmli (1970) refined for membrane proteins as described by Henderson & Macpherson (1986) using 4% stacking gels and 15% resolving gels in a BioRad Mini PROTEAN 3 apparatus. Acrylamide (40%) and bisacrylamide (2%) solutions were from BioRad Laboratories and SDS-7 protein molecular weight markers were from Sigma-Aldrich Co.

Growth and labelling conditions

Cultures were started from deep frozen (−80 °C) cell stocks that were first used to inoculate rich medium (2TY or LB) plates and then liquid cultures before transfer to minimal medium liquid cultures. An inoculum volume of 2% was used when transferring between successive liquid cultures. All media and flasks were sterilized by autoclaving or passage through filters of pore size 0.2 μm. All shaking cultures were grown at a temperature of 37 °C with aeration at 200 rpm (see below) and cell growth was monitored by measuring the absorbance at 600 nm. All plate and liquid cultures contained the antibiotic tetracycline at a concentration of 15 μg/ml to ensure continuous selection for plasmid-containing cells.

Uniform deuteration of the protein combined with uniform ¹⁵N-labelling was achieved by growth of E. coli strain JM1100 transformed with the plasmid pPER3(His₆) for expressing GalP with a C-terminal RGS-His₆ tag in a minimal medium (Ward et al., 2000) that contained successively increasing concentrations of D₂O (Apollo Scientific, >99.92% D or Cambridge Isotope Laboratories, 99.9% D) (0, 25, 50, 75, 100%) in small-scale (10 ml in tubes) cultures before transfer into final large-scale cultures with 100% D₂O along with d₇-D-glucose (Cambridge Isotope Laboratories, 98% D) and [¹⁵N]NH₄Cl (Cambridge Isotope Laboratories, 99% ¹⁵N). The deuteration procedure typically reduced the expression level from ∼50% (unlabelled) to ∼10% of total protein in membrane preparations and the purified protein yield by a similar factor to around 1 mg/litre (see below).

Specific labelling of tryptophan residues with ¹⁵N combined with perdeuteration was achieved by growth of the tryptophan auxotrophic host strain of E. coli CY15077 (E. coli Genetic Stock Centre, Yale University) transformed with the plasmid pPER3(His₆) (Patching et al., 2008a) in a modified M9 minimal medium that contained Na₂HPO₄ (6 g/litre), KH₂PO₄ (3 g/l), NaCl (0.5 g/litre), NH4Cl (40 mM), D-glucose (22.2 mM), MgSO₄ (2 mM), CaCl₂ (0.2 mM), L-tryptophan (40 mg/l) and tetracycline (15 μg/ml). Small-scale cultures (200 ml in 1-l flasks) were grown with successive increasing concentrations of D₂O up to 100% and then used to inoculate 5 l of the medium with 100% D₂O, d₇-D-glucose and [U-²H, U-¹⁵N₂]-L-tryptophan (Cambridge Isotope Laboratories, 98% D, 98% ¹⁵N) (40 mg/l) in a fermentor (New Brunswick Scientific BIOFLO 3000 bioreactor) operating in batch mode. The culture was grown at a temperature of 37 °C with agitation rate 200 rpm and air flow 10 l/min over ∼5 days to give a final cell density with A₆₀₀ 0.84 and then harvested to give a cell yield of 9.4 g. The cells were used to prepare membranes (see below) with a volume of 6 ml and with a total protein concentration of 13.53 mg/ml.

Specific labelling of methyl groups with ¹³C on isoleucine, leucine and valine (ILV) residues combined with perdeuteration and uniform ¹⁵N-labelling ([U-²H, U-¹⁵N, Ile-D₂, ¹³CH₃, Leu/Val-¹³CH₃, ¹²CD₃]GalP) was achieved by growth of E. coli strain JM1100 transformed with the plasmid pPER3(His₆) in a modified minimal medium (Ward et al., 2000) supplemented with labelled precursors to these amino acids. The medium contained the keto acids 2-keto-3,3-methyl-d₂-4-¹³C-butyrate and 2-keto-3-methyl-d₃-3-d₁-4-¹³C-butyrate (Sigma-Aldrich: ISOTEC, >99% ¹³C, >98% D) both at a concentration of 100 mg/l along with 100% D₂O, d₇-D-glucose (12.5 mM), [¹⁵N]NH₄Cl (2.5 g/l) and Na₂HPO₄ (3.4 g/l), KH₂PO₄ (1.6 g/l), MgSO₄ (66.7 mg/l), CaCl₂ (16.9 mg/l), thymine (20 μg/ml), L-histidine (40 μg/ml), MnCl₂.4H₂O (3.33 mg/l), FeSO₄ċ7H₂O (3.33 mg/l), thiamine (83 μg/l) and tetracycline (15 μg/ml). Small-scale cultures (10 ml in tubes) were grown with successive increasing concentrations of D₂O up to 100% and then used to inoculate a total of 3 l of the medium with the full complement of labelled components divided between four 2-l flasks (750 ml each). The cultures were grown over almost 80 h to give final cell densities with A₆₀₀ ∼0.6 and then harvested to give a total cell yield of 4.2 g. The cells were used to prepare membranes (see below) with a volume of 5 ml and with a total protein concentration of 8.41 mg/ml.

Membrane preparation and protein purification

Cells were thawed, suspended in Tris-EDTA buffer (20 mM Tris pH 7.5 with 0.5 mM EDTA), disrupted using a cell disruptor (Constant Systems) then mixed membranes (inner plus outer) were prepared as described previously (Ward et al., 2000), followed by washing and resuspension in Tris buffer (20 mM, pH 7.5), dispensing into aliquots, rapid freezing in liquid nitrogen and storage at −80 °C. Inner and outer membranes were not separated as previous attempts to separate membranes that were fully deuterated by ultracentrifugation on a sucrose gradient proved not to be successful owing to their higher densities compared with unlabelled membranes.

General procedure for protein solubilization and purification. Membrane preparations were solubilized for up to 4 h at 4 °C in a buffer containing 10 mM HEPES (pH 7.9), 20% glycerol, 300 mM NaCl, 1% DDM, and 20 mM imidazole at a protein concentration of 3 mg/ml followed by removal of insoluble material by ultracentrifugation (100 000 g, 1 h, 4 °C). Immobilized-metal affinity chromatography (IMAC) was performed by mixing the supernatant obtained above with Ni-NTA resin (QIAGEN) (1 ml of resin per 3 mg of protein) overnight at 4 °C, which was then packed into a column. All column steps were performed at 4 °C. Unbound material was collected followed by washing of the column with at least 40× column volumes of a buffer containing 10 mM HEPES (pH 7.9), 10% glycerol, 100 mM NaCl, 0.05% DDM and 20 mM imidazole. The His-tagged protein was eluted from the column using ∼7 ml (for a 1 ml column) of a buffer containing 10 mM HEPES (pH 7.9), 5% glycerol, 0.05% DDM and 200 mM imidazole, which was then concentrated to a volume of ∼300 μl by centrifugation using a concentrator with a MW cut-off of 100 kDa (Vivaspin 20, Sartorius). Using the same column, the protein was washed a minimum of five times with at least 5 ml of a buffer containing 10 mM KPi pH 7.5, 5% glycerol and 0.05% DDM, before concentrating to a volume of 200–500 μl, dispensing into aliquots, rapid freezing in liquid nitrogen and storage at −80 °C.

The tryptophan-specific ¹⁵N-labelled protein [U-²H, U-¹⁵N₂-Trp]GalP was purified from the whole of the membrane preparation using d₈-glycerol [(Cambridge Isotope Laboratories, 98% D) and [U-²H]DDM (synthesized in-house from d₇-D-glucose and d₂₅-n-dodecanol (Cambridge Isotope Laboratories)] in all buffers after packing the resin into the column. The volume of the final purified protein was 740 μl with a protein concentration of 9.98 mg/ml, i.e. a protein yield of 7.4 mg equivalent to 1.5 mg/l of cell culture; 2 × 300 μl aliquots were stored for use as NMR samples.

The ILV-specific ¹³C-labelled protein [U-²H, U-¹⁵N, Ile-D₂, ¹³CH₃, Leu/Val-¹³CH₃, ¹²CD₃]GalP was purified from the whole of the membrane preparation using d₈-glycerol in all buffers after packing the resin into the column. The volume of the final purified protein was 220 μl with a protein concentration of 2 mg/ml, i.e. a protein yield of 0.44 mg equivalent to 0.15 mg/l of cell culture; 2 × 100 μl aliquots were stored for use as NMR samples.

NMR samples

The tryptophan-specific ¹⁵N-labelled NMR samples contained [U-²H, U-¹⁵N₂-Trp]GalP at a concentration of 165 μM (8.41 mg/ml) along with 5% d₈-glycerol and 1% [U-²H]DDM in KPi buffer (10 mM, pH 7.5) plus 10% D₂O.

The ILV-specific ¹³C-labelled NMR samples contained [U-²H, U-¹⁵N, Ile-D₂, ¹³CH₃, Leu/Val-¹³CH₃, ¹²CD₃]GalP at a concentration of 51.8 μM (2.64 mg/ml) along with 5% d₈-glycerol and 0.05% DDM in KPi buffer (10 mM, pH 7.5) plus 10% D₂O. The inhibitor forskolin was added to these samples to a final concentration of 100 μM from a higher concentration stock solution in d₆-ethanol. The concentration of added ethanol did not exceed 1%.

Samples for the WaterLOGSY experiments had components at the following concentrations: Purified unlabelled GalP protein at 5 μM, forskolin at 100 μM, 1% d₆-ethanol, 0.05% DDM in KPi buffer (10 mM, pH 7.5).

NMR experiments

[¹⁵N-¹H]TROSY and [¹³C-¹H]-methyl-TROSY spectra were measured at 750 and 900 MHz on Varian Unity spectrometers using cryogenically cooled probes. All samples used Shigemi micro-tubes with a sample volume of approximately 300 μl. [¹³C-¹H]-methyl-TROSY spectra were acquired using a PEP-methyl-TROSY pulse sequence (Guo et al., 2008). At 750 MHz data were acquired with 1024 points in the ¹H dimension over a sweep width of 10 484 Hz and 124 increments in the indirect ¹³C dimension over a sweep width of 4600 Hz. For the 900 MHz data 1024 and 124 direct and indirect dimension points were acquired and the ¹H and ¹³C sweep widths were 12 608.4 and 5800.0, respectively. Data were processed using NMRPipe offset sine bell and sine bell squared windows in the direct and indirect dimensions; the offset used was 0.4. The spectra were extended in the indirect dimension using mirror image linear prediction and zero filled to give a final size of 4096 and 1024 points in the direct and indirect dimensions. Spectra were displayed and analyzed in CCPN analysis version 2.2. Chemical shift measurements were referenced directly and indirectly to an internal DSS standard.

¹H WaterLOGSY experiments (Dalvit et al., 2000) were measured at 500 MHz using a sample containing 5 μM GalP in DDM micelles with 200 μM forskolin; a reference spectrum was measured under the same conditions without GalP. The sample buffer and conditions were the same as those used for methyl-TROSY measurements excluding the use of a 10% D₂O:90% H₂O solvent. Experiments were measured at 298 K with a ¹H sweep width of 8000 Hz, 13 108 data points and 2048 transients. Water saturation was achieved using a 15 ms Gaussian pulse at the water signal frequency and a NOE mixing time of 2 sec. Spectra were processed with a 0.5 sine bell shifted window. Reference/control experiments were also performed on samples of buffer alone, protein alone and forskolin alone, all containing the detergent DDM.

Circular dichroism measurements

Far-UV synchrotron radiation circular dichroism measurements were performed using a nitrogen-flushed instrument on Beamline B23 at the Diamond Light Source, Oxfordshire, UK. Samples contained purified GalP at a concentration of 0.5 mg/ml (∼10 μM) in KPi buffer (10 mM, pH 7.5), 5% glycerol, 0.05% DDM in a cell with path-length 0.02 cm (nominal volume 30 μl) and were maintained at a temperature of 25 °C. Measurements were made over the wavelength range 180–260 nm using an integration time of 1 sec, increment of 1 nm and slit widths of 0.5 or 1.0 mm. Spectra are presented in units of mean residue ellipticity ([θ]MR) and all measurements had PMT values that were below 600 V.

Fluorescence measurements

The ligand-binding activity of the protein was measured by observing the direct quench in fluorescence intensity of intrinsic tryptophans by titration with the inhibitor cytochalasin B. Measurements were made using a Perkin Elmer LS-50B spectrofluorimeter with sample volumes of 2 ml held in a magnetically-stirred quartz cuvette with a path-length of 1 cm at a temperature of 25 °C controlled by flow from an external water bath. Fluorescence was excited at a wavelength of 280 nm and emission spectra were recorded in the wavelength range 300–400 nm. Excitation and emission slit widths were both set to 2.5 mm and each spectrum was the average of three scans. Samples contained purified GalP at a concentration of 100 μg/ml (2 μM) in KPi buffer (10 mM, pH 7.5), 5% glycerol, 1% DDM and were equilibrated in the instrument for 15 min before recording the first spectrum. Additions of cytochalasin B were made by introducing 2 μl aliquots of a relevant higher concentration stock solution in 100% ethanol and the sample was incubated for 2 min following each addition before recording the next spectrum. A full titration up to a cytochalasin B concentration of 50 μM increased the sample volume by less than 1% and introduced less than 1% ethanol into the sample. Titration experiments were performed on samples that had previously been incubated at a range of temperatures (4, 25 and 37 °C) over a range of times (0, 24, 48, 72 and 144 h).

Results and discussion

Sample conditions and protein stability

For this work the protein was purified and reconstituted using the non-ionic detergent n-dodecyl-β-D-maltoside (DDM), which we have found gives the best solubilization and purification yield and retains functional activity of GalP and other similar secondary active transporters (Ward et al., 2000). At a temperature of 25 °C in this detergent the secondary structure of the protein was stable for at least 48 h as shown by circular dichroism spectroscopy from a comparison of repeated scans over this period (Figure 1A) and any significant degradation at this temperature was observed only by exposure to significantly higher UV irradiation (Figure 1A, dashed spectrum). The activity of the protein was also stable at a temperature of 25 °C for at least 48 h as shown by monitoring its quench in intrinsic fluorescence intensity on excitation at 280 nm and titration with the inhibitor cytochalasin B, which had a starting apparent dissociation constant of K_D = 11.67 ± 1.82 μM in this assay (Figure 1B and Supplementary Figure 1, available online). At a temperature of 37 °C, the activity of the protein was reduced significantly over a period of 24 h; elevated sample temperatures for NMR measurements, as used with the rhodopsins (Gautier et al., 2008, 2010; Reckel et al., 2011) would therefore not be feasible with GalP. Though protein-detergent micelles derived using DDM tend to be larger than those using other detergents that are more commonly used for solution-state NMR studies with membrane proteins, such as the zwitterionic detergents DPC/DHPC/LDAO (Patching, 2011), this milder detergent was chosen as the membrane mimetic for GalP since retaining the stability of structure and activity by the protein for the duration of NMR measurements was a primary consideration.

Figure 1. Stability of purified GalP reconstituted in DDM micelles. (A) Overlay of far-UV circular dichroism spectra for purified GalP [10 μM in 10 mM KPi buffer, pH 7.5, 5% glycerol, 0.05% DDM] recorded at a temperature of 25 °C using slit widths of 0.5 mm with repeated scans every 20 min over a period of 48 h (left) and plots of the CD signal at wavelengths of 193, 210 and 223 nm against time (right). The black dashed spectrum is from a separate sample that was subject to rapid degradation over 100 min by exposure to significantly higher intensity UV light using slit widths of 1 mm. (B) Fluorescence spectra excited at 280 nm for purified GalP [100 μg/ml in KPi buffer (10 mM, pH 7.5), 5% glycerol, 1% DDM] recorded at a temperature of 25 °C following successive additions of the inhibitor cytochalasin B (to give concentrations of 0, 1, 2, 3, 5, 10, 15, 20, 30, 40, 50 μM) on aliquots removed at a number of time-points from a solution of protein held at a temperature of 25 °C (top). This experiment was repeated on protein samples held at a range of temperatures and the % quench in fluorescence intensity at 330 nm on titration with cytochalasin B used as a measure of protein activity (bottom).

Tryptophan residues using amino acid-specific labelling and [¹⁵N-¹H]TROSY

As would be expected with a protein the size of GalP reconstituted in DDM detergent micelles, NMR measurements performed on perdeuterated and uniformly ¹⁵N-labelled protein using the [¹⁵N-¹H]TROSY experiment at 750 MHz with the sample at a temperature of 25 °C produced spectra in which the large majority of signals were absent with only around 30 observed out of the expected 464 (Supplementary Figure S2, available online). This led us to pursue selective amino acid labelling of the protein, which would obviously be necessary to achieve highly resolved spectra. Selective labelling with ¹⁵N focused on tryptophan residues, some of which are directly involved in the sugar transport process as previously shown by sugar binding experiments and transport assays performed with mutants and by homology modelling (McDonald et al., 1995; Patching et al., 2008a, 2008b; Walmsley et al., 1994b). Labelling of the tryptophan residues with ¹⁵N combined with uniform deuteration was achieved by expression of the protein using an E. coli strain auxotrophic for tryptophan and growth in a modified minimal medium with L-[U-²H, U-¹⁵N₂]tryptophan at a concentration of 40 mg/litre (Supplementary Figure S3A, online). The expression in membranes was around 25% of total protein, which was purified and reconstituted with 5% d₈-glycerol and 1% [²H]DDM (Figure 2A and Supplementary Figure S3B, online) with a yield of 1.2 mg per litre (a total of 6 mg purified protein was obtained from 5 l of expression medium). NMR samples contained [U-²H, ¹⁵N₂-Trp]GalP(His₆) at a concentration of 165 μM and [¹⁵N-¹H]TROSY experiments at 900 MHz (Figure 2B) at a temperature of 25 °C reproducibly gave 12 resolved signals for the backbone ¹⁵N positions in tryptophan residues, which matches the number of tryptophan residues in the protein. Up to five resolved signals for side-chain ¹⁵N positions in tryptophan residues were also seen along with two large overlapped peaks of similar intensity, likely arising from the other unresolved tryptophans. The spectrum in Figure 2(B) was acquired in a 65-h experiment using 900 MHz, which was necessary to achieve spectra of this quality, since at 750 MHz the TROSY effect was inadequate, especially for backbone resonances. It should be noted that the absence of tryptophan signals, both backbone and side-chain positions, in the spectrum for uniformly ¹⁵N-labelled protein (Supplementary Figure S2) derives from a combination of the reduced TROSY effect and lower intrinsic sensitivity at 750 MHz and an acquisition time that was four-times shorter than that used for the spectrum in Figure 2(B).

Figure 2. Amplified expression and purification of tryptophan-labelled GalP and detection of signals for all 12 tryptophan residues. (A) Coomassie-stained SDS-PAGE analysis for amplified expression of [U-²H, ¹⁵N₂-Trp]GalP(His₆) in native membranes (left) and the purified protein reconstituted in [²H]DDM micelles (right). (B) [¹⁵N-¹H]TROSY spectrum at 900 MHz of [U-²H, ¹⁵N₂-Trp]GalP(His₆) (∼165 μM) in [U-²H]DDM (1%), [²H₈]glycerol (5%), KPi buffer (10 mM, pH 7.5) acquired over 65 h at 25 °C.

To the best of our knowledge, the tryptophan-labelled samples of GalP provide the best resolved [¹⁵N-¹H]TROSY spectra achieved for any membrane protein with 12 transmembrane-spanning α-helices and demonstrates the potential of the TROSY technique for elucidating structural features and examining dynamics or ligand binding with such large proteins. The results provide a useful strategy for large polytopic α-helical membrane proteins when sites of interest can be labelled with amino acid-specific isotopes using auxotrophs or cell-free synthesis combined with deuteration. The long recording times required to achieve acceptable signal to noise ratios at the typical protein concentrations that are used for large proteins are a major drawback, however, providing an incentive for us to identify more sensitive methods to achieve well resolved spectra as follows.

Methyl groups on isoleucine, leucine and valine residues using labelled precursors and [¹³C-¹H]-methyl-TROSY

To achieve higher sensitivity we turned to ¹³C-labelling of methyl groups on isoleucine, leucine and valine (ILV) residues combined with the [¹³C-¹H]-methyl-TROSY experiment developed principally by Tugarinov and Kay (Kay, 2011; Korzhnev et al., 2004; Tugarinov et al., 2003, 2004; Tugarinov & Kay, 2003, 2004a, 2004b), which achieves high-resolution NMR spectra with very large systems and that has been used to assist structure-determination of a number of membrane proteins (Fernández et al., 2004; Gautier et al., 2008, 2010; Renault et al., 2009; Yu et al., 2005). We grew our producing organism in minimal medium under conditions for uniform deuteration and ¹⁵N-labelling along with labelled precursors to ILV residues [(α-ketobutyrate-3,3-d₂-4-¹³C and α-ketoisovalerate-3-(methyl-d₃)-4-¹³C)] followed by purification and sample preparation with 0.05% DDM and 5% d₈-glycerol (Supplementary Figure S4, online). The specific labelling pattern that these precursors achieve is [U-²H, U-¹⁵N, Ile-D₂, ¹³CH₃, Leu/Val-¹³CH₃, ¹²CD₃]GalP(His₆), which minimizes inter-methyl relaxation and spectral crowding by labelling only one of the two methyl groups on Leu and Val residues (Ollerenshaw et al., 2005; Religa & Kay, 2010). A total of 1.5 mg of pure protein was produced from 3 l of cell culture and divided to produce two identical ∼50 μM NMR samples. [¹³C-¹H]-Methyl-TROSY experiments were performed on each sample at 750 MHz and 900 MHz, both over 21 h at 20 °C. The resultant spectra showed an expected improvement in resolution of signals, though not in sensitivity, at 900 MHz over 750 MHz (Figure 3). At 900 MHz signals were resolved for 30 (out of 36) isoleucine residues and for around 60 (out of 60 + 34) leucine and valine residues (Figure 3). The Ile region shows especially well resolved signals and a nearly complete peak count. The quality of these spectra are a clear demonstration that the methyl-TROSY experiment is capable of providing impressive sensitivity on sub-milligram amounts of large polytopic α-helical membrane proteins. These results show that even using relatively low concentration samples with low sample temperatures, there is a realistic prospect for obtaining good quality NMR spectra of many complex polytopic membrane proteins in detergents that support functional integrity of the proteins. In such proteins that contain many methyl groups, very high magnetic fields therefore still offer a clear advantage in signal resolution.

Figure 3. Improvement in resolution for detection of isoleucine, leucine and valine methyl groups at 900 MHz over 750 MHz. [¹³C-¹H]-Methyl-TROSY spectra at 750 MHz (left) and 900 MHz (right) of [U-²H, U-¹⁵N, Ile-D₂, ¹³CH₃, Leu/Val-¹³CH₃, ¹²CD₃]GalP(His₆) (51.8 μM) in DDM (0.05%), d₈-glycerol (5%), D₂O (10%), KPi buffer (10 mM, pH 7.5) acquired over 21 h at 20 °C. The asterisk (*) indicates a signal coming from the detergent DDM.

Complicating factors in the acquisition of methyl-TROSY spectra on membrane proteins can be interference from detergents or from ¹³C labelled cellular lipids carried over during purification. These give rise to strong signals that may interfere in the vicinity of the methyl region of the protein. In our case the signals from unlabeled detergent and ¹³C-labelled lipid precluded the acquisition of spectra using the original methyl-TROSY sequence. Instead we used the PEP “preservation of equivalent pathways” variant of the methyl-TROSY experiment (Guo et al., 2008). This sequence uses coherence selection gradients to suppress these signals. In the case of transferred lipids the PEP sequence intrinsically suppresses signals from ¹³CH₂ groups. Whereas for the detergent peaks the sequence avoids interferences that arise due to incomplete subtraction of ¹H signals from ¹H-¹²C groups, but does not filter out the 1.1% natural abundance ¹³CH₃ and ¹³CH signals from the concentrated DDM present in the sample, which are still visible in the spectra (labelled * in Figures 3 and 4 and in Supplementary Figure S5, online). It is possible that the protein purification procedure for GalP could be modified to minimize the transfer of lipids. A more complex purification procedure did however lead to lower protein yields and may compromise the functional integrity of the protein.

Figure 4. Detection of inhibitor binding to GalP in methyl-TROSY spectra at 900 MHz. (A) Ile C^δ1 region of [¹³C-¹H]-methyl-TROSY spectra at 900 MHz of [ILV-¹³CH₃]GalP(His₆) (51.8 μM) in DDM (0.05%), d₈-glycerol (5%), D₂O (10%), KPi buffer (10 mM, pH 7.5) in the absence (black) and presence (red) of forskolin (100 μM with 1% d₆-ethanol) acquired over 21 h at 20 °C. The largest chemical shift changes are indicated with numbered arrows. The asterisk (*) indicates a signal coming from the detergent DDM. (B) Histogram of chemical shift change metric (SQRT[(3*Δ¹H)²+ (Δ¹³C)²]) ordered according to the size of the chemical shift change upon binding forskolin for the Ile C^δ1 peaks. Changes for signals beyond those labelled 1–5 are considered as not significant.

To the best of our knowledge, these are the best resolved [¹³C-¹H]-methyl-TROSY spectra achieved for any membrane protein with more than seven putative transmembrane-spanning α-helices.

Use of methyl-TROSY spectra to detect ligand binding

Following acquisition of the well-resolved methyl-TROSY spectra at 900 MHz next we investigated whether the [¹³C-¹H]-methyl-TROSY experiment would detect the binding of a small-molecule inhibitor of the GalP protein. Forskolin at a concentration of 100 μM was added to the same sample used to obtain the 900 MHz spectrum in Figure 3 and then the [¹³C-¹H]-methyl-TROSY experiment was repeated. The resultant spectrum showed a pronounced chemical shift change in one Ile residue and smaller changes in four other Ile residues (Figure 4). Differences are also apparent in the Leu/Val region of the spectrum, but due to the greater overlap of signals in this region it is more difficult to identify individual chemical shift changes (Supplementary Figure S5, online). Binding of forskolin to GalP was confirmed by use of the ligand screening experiment WaterLOGSY (Dalvit et al., 2000), which showed a binding effect in samples containing the same components as those used in the methyl-TROSY experiments (Supplementary Figure S6, online). A complete characterization of ligand binding should ideally include a multiple point titration to allow verification of affinity and stoichiometry under NMR conditions; such an approach should be used when the stability and availability of the protein of interest and other circumstances allow it. Nevertheless, the methyl-TROSY spectra clearly indicate a single Ile residue in very close proximity to the inhibitor binding site. As forskolin is a competitive inhibitor similar shift changes may be shared with binding of the natural sugar substrates and demonstrate that the Ile region of the methyl-TROSY spectra provides an excellent active site probe for further dynamic and structural studies of the protein. Recently the crystal structures of the evolutionary related D-xylose transporter XylE from E. coli (Henderson & Baldwin, 2012; Sun et al., 2012) and the D-glucose transporter GlcP from Staphylococcus epidermidis (Iancu et al., 2013) have been solved, which are also close homologues of the human glucose transporter (GLUT1). The XylE structure reveals a single Ile residue (residue 171) on transmembrane helix 5 in the sugar binding site in van der Waals contact with the sugar (Sun et al., 2012). This Ile is highly conserved between XylE, GlcP (residue 161), GalP (residue 150) and sugar transporters in humans, yeasts and plants (Supplementary Figure S7, online) and is therefore an immediate candidate for the Ile that responds most strongly to the addition of forskolin.

With proteins of this size it is clear that amide-based correlation experiments cannot be used to transfer chemical shift assignments from the backbone to methyl-TROSY spectra. Definitive chemical shift assignments in the methyl-TROSY spectra could be made by residue mutation, which may be a useful strategy for investigating specific residues. More complete assignments can potentially be made utilizing PRE and NOE measurements as well as chemical shifts together in a combinatorial assignment procedure. This would require relatively few mutation-based assignments to arrive at a unique assignment for most methyl groups, especially if this procedure is informed by a good homology model or a crystal structure of the protein.

Conclusions

Using the GalP protein as a model system, this work has overcome the challenges for achieving high-resolution TROSY NMR spectra with large polytopic α-helical membrane proteins, a prerequisite to applying further labelling strategies and NMR experiments for measurement of dynamics and structure elucidation. This work has also demonstrated the feasibility of using these experiments to screen for the binding of small-molecule inhibitor or drug molecules interacting with large α-helical membrane proteins, therefore enhancing the application of NMR tools to the discovery of drugs.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

This work was supported by the EU EDICT consortium (contract 201924), the EPSRC (EP/G035695/1), the BBSRC (BB/G020043/1), the Wellcome Trust (093792/Z/10/Z) and the University of Leeds. PJFH received personal funding from the Leverhulme Trust.

Acknowledgements

The authors thank the HWB Biomolecular NMR Facility (University of Birmingham) for 900 MHz instrument time and Diamond Light Source Ltd (Oxfordshire) for instrument time on synchrotron CD Beamline B23 with assistance from Giuliano Siligardi and Rohanah Hussain. The authors also thank John O’ Reilly, Ryan Hope and David Sharples (University of Leeds) for assistance with fermentor cultures and Steve Baldwin (University of Leeds) for useful discussions.