Solution structure, backbone dynamics, and stability of a double mutant single-chain monellin. structural origin of sweetness.

Single-chain monellin (SCM), which is an engineered 94-residue polypeptide, has been characterized as being as sweet as native two-chain monellin. Data from gel-filtration high performance liquid chromatography and NMR has proven that SCM exists as a monomer in aqueous solution. In order to determine the structural origin of the taste of sweetness, we engineered several mutant SCM proteins by mutating Glu(2), Asp(7), and Arg(39) residues, which are responsible for sweetness. In this study, we present the solution structure, backbone dynamics, and stability of mutant SCM proteins using circular dichroism, fluorescence, and NMR spectroscopy. Based on the NMR data, a stable alpha-helix and five-stranded antiparallel beta-sheet were identified for double mutant SCM. Strands beta1 and beta2 are connected by a small bulge, and the disruption of the first beta-strand were observed with SCM(DR) comprising residues of Ile(38)-Cys(41). The dynamical and folding characteristics from circular dichroism, fluorescence, and backbone dynamics studies revealed that both wild type and mutant proteins showed distinct dynamical as well as stability differences, suggesting the important role of mutated residues in the sweet taste of SCM. Our results will provide an insight into the structural origin of sweet taste as well as the mutational effect in the stability of the engineered sweet protein SCM.


INTRODUCTION
The native sweet protein, monellin, which was originally isolated from the berries of the West African plant Dioscoreophyllum cumminsii (1,2), consists of two separate polypeptide chains: an A chain of 45-residues, and a B chain of 50-residues.
Native two-chain monellin is approximately 70,000 times sweeter than sucrose and about 300 times sweeter than the dipeptide sweetener aspartame (3,4). Other sweet taste proteins, such as thaumatin, pentadin and mabinlin are also known (5)(6)(7)(8)(9). Among these sweet proteins, a curculin protein has demonstrated a sweet taste and shown tastemodifying activity (10). The crystal structure of native two-chain monellin has been determined as showing a β-sheet comprised of five antiparallel strands and a single 17residue long α -helix. The two chains were packed closely by hydrogen bonds and hydrophobic interactions (11). In addition, the crystal structure showed that the amino terminus of the A chain was connected to the carboxyl terminus of the B chain through intermolecular hydrogen bond networks.
The recombinant SCM proteins were expressed in Eschericha coli strain BL21 (DE3) containing the plasmid pET21. Transformed cells were propagated in E. coli nitrogen base containing 5% glucose and 0.5% ammonium sulfate at 30 ºC for 2 hours and grown in M9 media containing 2% glucose and 0.1% ammonium sulfate at 30 ºC for 48 hours. 15 N-labeled ammonium sulfate was used as the sole source of the nitrogen for uniformly 15 N-labeled SCM DR and wild type SCM. The cells were harvested by centrifugation at 3500 rpm for 25 min. Cells were stored at -80 ºC and used for purification procedures. Cell pastes were disrupted by a bead beater in 25 mM sodium phosphate, 5 mM EDTA, 150 mM NaCl, and 1 mM PMSF at pH 7.0. The cell lysates were collected by centrifugation at 12000 rpm for 215 min. After pH adjustment and centrifugation, the supernatants were diluted with 10 mM of sodium phosphate and loaded onto a CM-Sepharose column. The bound SCM was eluted with a salt gradient.
The collected protein solution was dialyzed and dried with a freeze-dryer for spectroscopic measurements. The protein concentration was determined using the Bradford method.

Fluorescence Spectroscopy
Fluorescence spectra were measured in 50 mM potassium phosphate buffer, at pH 7.0 and 25 ºC on F-4500 fluorescence spectrophotometer. Fluorescence emission spectra were recorded from 270 to 450 nm at each GdnHCl concentration using two different excitation wavelengths, 280 and 295 nm. The protein concentration in the cuvette was 30 µM and a path length of 1 cm was used. GdnHCl-unfolding experiments were carried out after the protein was incubated in solutions containing different concentrations of the denaturant for 24 hours, at 25 ºC. The refolding reaction of SCM was carried out under various conditions by diluting the denaturant concentration.
Data from the equilibrium denaturation were converted to plots of f U versus denaturant concentration using the equation [1]: where Y O is the observed signal at a particular GdnHCl concentration. Y F and Y U represent the intercepts, and m F and m U are the slopes of the native protein and the unfolded baselines. They were obtained by extrapolation of linear least-squares fits of the baselines.
To determine whether the two-state unfolding model was appropriate for analyzing the GdnHCl-induced denaturation data, f U values were fitted to equation [2]: In eq [2], f U is related to ∆G U by a transformation of the Gibbs-Helmholtz equation in which the equilibrium constant for unfolding in the folding transition zone, K U , is given by K U = f U /(1-f U ), for a two-state transition. It is also implicit in eq [2] that the free energy of unfolding is dependent linearly on denaturant concentration (17).

CD Spectroscopy
CD spectra were measured in 50mM of potassium phosphate buffer, at pH 7.0 and 25 ºC on a Jasco 720 spectropolarimeter. Far-UV CD spectra were monitored from mdeg sensitivity, response time of 1 s, and scan speed of 50 nm/min. The spectra were recorded as a 6 scan average value. The molar ellipticity was determined as: where c is the protein concentration (in g/ml), l the light path length in the cell (in mm), θ λ the measured ellipticity (in degrees) at wavelength λ, and M ar the mean molecular mass of amino acid of the protein determined from its amino acid sequence.  [4] Reversibility was examined by comparing the transition curves of a sample that was briefly heated to a temperature where the protein was completely unfolded.

NMR Spectroscopy
All NMR spectra were acquired on a Bruker DRX-500 spectrometer in quadrature detection mode, equipped with a triple-resonance probe with an actively shielded pulsed field gradient (PFG) coil. All two-dimensional experiments were performed at 298 K. Pulsed-field gradient techniques were used for all H 2 O experiments, 8 resulting in good suppression of the solvent signal. 15  duplicate spectra were recorded for T = 246 ms (T 1 spectra) and T = 56.8 ms (T 2 spectra). In order to eliminate the effects of cross correlation between 15 N-1 H dipolar and 15 N CSA relaxation mechanisms, 1 H 180º pulses were inserted during the relaxation time according to the published methods (27,28). 15 N-{ 1 H} steady-state heteronuclear NOE (XNOE) (29,30) data was also obtained using a relaxation delay of 5 s.

NMR Data Processing and Analysis
The NMR data were processed using the nmrPipe/NMRDraw software packages 9 (Biosym/Molecular Simulations, Inc.) and analyzed by the Sparky 3.60 software. The experimental data were extended by linear prediction and zero-filled to give 2048 R 512 data matrices and processed using gaussian multiplication and a shifted (π/3) sine bell function prior to Fourier transformation. The peak intensities in the twodimensional spectra were measured by peak heights using the Sparky program. The XNOE value for a given residue was calculated as the intensity ratio (I/I 0 ) of the 15

Experimental Constraints and Structure Calculations
Structures were generated using hybrid distance geometry and dynamical simulated annealing protocol with the CNS 1.0 program on a SGI Indigo 2 workstation.
Our methodology was similar to that used by Clore and Gronenborn (31,32) and their coworkers. Distance geometry (DG) substructures were generated using a subset of atoms in the peptide and followed a refinement protocol described in Lee et al (33). The target function for molecular dynamics and energy minimization consisted of a covalent structure, van der Waals repulsion, NOE and torsion angle constraints (34). The torsion angle and NOE constraints were represented by square-well potentials. Based on cross peak intensities in the NOESY spectra with mixing times of 100 and 150 ms, the distance restraints were then classified as strong, medium or weak corresponding to upper distance bounds of 2.7 Å, 3.3 Å, and 5.0 Å, respectively. An additional 1 Å was added to upper distance bounds for pseudoatom involving non-stereospecifically assigned methylene protons, methyl groups and the ring protons of phenylalanine residue (35). A lower distance bound of 1.8 Å was used for all NOE-derived distance restraints. Structures were calculated using 296 intra-residues, 245 sequential, 124 medium range and 331 long-rang NOE restraints. A total of 76 hydrogen bond restraints were also included in the calculations. Potential hydrogen bond donors were assigned from a 1 H-15 N HSQC spectrum recorded immediately after dissolving lyophilized H 2 O sample to 100% D 2 O solution. Hydrogen bonds were further identified from characteristic NOE patterns that were observed for residues in regular secondary structure, together with the solvent exchange data. From 2D DQF-COSY (36,37) and 15 N-edited HNHA spectra, 65 torsion angle restraints were also derived for backbone Φangles within elements of secondary structure based on 3 J HNα coupling-constants ( 3 J HNα >8, 120 (±50)°, 3 J HNα <6, 60(±45)°). Distance geometry (DG) substructures were generated using a subset of atoms in the peptide, and followed a refinement protocol described in Lee et al (33). All modeling calculations were performed within the InsightII program (Biosym/Molecular Simulations, Inc.) on a SGI Indigo 2 workstation.

Circular Dichroism
Circular dichroic spectra of both wild-type and double mutant proteins in the far-UV region were collected in 50 mM of sodium phosphate buffer solution at pH 7.0. The spectra suggest that the global folding of the two proteins are similar, showing a major β-strand and minor α-helical contents. However, a small difference at 217 nm was clearly detected, indicating that the structural change of the β-sheet region was due to double mutation of SCM. The additional minima observed at 206-208 nm have been ascribed from the contributions of aromatic side chains (38, 39) (Fig. 1).

Solution Structures and Sweet Taste
Spin system assignments were easily made by homonuclear 2D TOCSY and 15 Nedited 3D TOCSY-HSQC spectrum. All identified spin systems served as a starting point for complete sequence-specific resonance assignment procedure. Fig. 2 shows 2D 1 H-15 N HSQC spectrum with the assignments. A total of 50 substructures generated from distance geometry algorithms were used as starting structures in the simulated annealing stage. After simulated annealing calculations, the 20 structures (<SA> k ) showed no constraint violations greater than 0.5 Å for distances and 5° for torsional angles. These structures were used for detailed structural analysis. Table I Table II summarizes the structural statistics associated with 20 final <SA> k structures of SCM DR . The deviations from idealized geometry are also very small and satisfy ideal geometry. The <SA> kr structure clearly demonstrates the relative orientations of its major secondary structures, showing that the twisted β-sheet partially wrapped the beginning of the α-helix.
Especially, the bend of three strands (β2, β3 and β4) enable to have a close contacts with the α -helix. A best-fit backbone superposition of all final <SA> k structures with an average REM structure is displayed in Fig. 3. The angular order parameter for Φ, Ψ Five loops including an engineered one were also characterized. The side chain orientations of Tyr63 and Asp66 which are common to all sweet peptides and correspond to Phe and Asp residues of aspartame, were observed on the opposite side to the H1 helix and those that were mostly exposed to solvent.  Fig. 6B shows that the fractional change of unfolding suggests a twostate model for GdnHCl-induced denaturation of both SCM DR and wild-type protein.
However, the transition midpoint of the double mutant SCM has shifted slightly towards the left, implying that the double mutation induces destabilization of the protein.
In each case, the reversibility of the unfolding reaction was confirmed by obtaining a refolding curve through dilution of the protein from high GdnHCl concentration. The denaturation and renaturation curves are determined to be exactly superimposable.

Backbone Dynamics of SCM DR
The rotational diffusion anisotropy was estimated for each residue, yielding a ratio of the principal moments of inertia of SCM DR as 1.00:0.86:0.40. This data suggests that the global shape of SCM DR is a prolate ellipsoid. Using data from R 2 /R 1 ratios and structural coordinates of 66 selected residues, axial diffusion tensors, D /D , were also calculated. The starting values of the parameters τ m and D /D were 4.9 ns and 0.65 at 500 MHz, respectively. The standard values of R 1 , R 2 , and 15 N-{ 1 H} NOEs were fitted in the model-free system selected by an F-test. Fig. 7 demonstrates that the regions of the secondary structures showed higher S 2 values than those of loops, as we expected.
The average values of S 2 are above 0.94 for most residues in β-strands and above 0.98 for α-helical region. More than 80% of the residues in SCM DR have order parameter S 2 values greater than 0.8, indicating that the protein in general is relatively rigid. Five residues of Glu4, Ile5, Thr33, Tyr47, and Arg82 belonged to the loop regions showed relatively higher R 2 /R 1 ratios, originated from a significant chemical exchange contribution (Fig. 7A). In addition to this exchange contribution, residues of Glu4 and Ile6 exhibited a significant reduction in the 15 N-{ 1 H} NOEs and S 2 values of 0.7 and 0.69 (Fig. 7B). Thus, it can be supposed that both Glu4 and Ile5 residues might be involved in dynamical motions. Four residues of Ile5, Tyr47, Arg51, and Asp66 located in the loop regions contained τ e contributions, indicating enhanced backbone dynamics on a fast time scale.
In the previous report, we determined that SCM exists as a monomer conformation based on NMR and gel-filtration experiments. Solution structure suggested that Arg70, Arg86 residues have a close correlation with the degree of sweetness of SCM protein. A comparison of the secondary structure of SCM DR with wild type protein indicates that the structural difference between the two proteins can be observed mainly in β-strands. The first short β-strand composed of residue Glu2-Ile5 found in the wild type SCM was not detected in SCM DR . Amide hydrogen exchange data, backbone-backbone NOEs, and C α H chemical shift indices did not provide any evidence of support β-strand in this region. Therefore, it can be supposed that a mutation of Asp7 disrupts the first short strand, perturbating the stability of the network of β-strand. The main structural differences are the following: each β-strand encompasses six residues from Cys41 to Ile46 for a second β-strand, nine residues from Gly55 to Tyr63 for a third, and six residues from Arg70 to Glu75 for a fourth β-strand, whereas seven residues from Cys41 to Tyr47, eleven residues from Lys54 to Ala64 and seven residues from Phe69 to Glu75 constitute each third, fourth, and fifth β-strand in wild type SCM. Several reports have already proposed that a sulfhydryl group of Cys41 in the beginning of β3 strand could be critical for sweetness. In our solution structures of both wild type SCM and SCM DR , the side chain of Cys41 is located on the hydrophobic interface between β-sheet and α-helix. We might think that the side chain of Cys41 plays a role for sweetness because it maintains a bulge between Ile38-Cys41 responsible for structural organization, especially the H1 helix orientation and, furthermore, the tertiary structure of single chain monellin. We also proposed that hydrophobic and/or side chain-side chain interaction related to tertiary structure of SCM is in part responsible for sweetness. Mutational studies of SCM proteins have supported our structural data, showing that the size of Asp7 residue in the loop A is important for