Solution structure of a novel T-cell adhesion inhibitor derived from the fragment of ICAM-1 receptor: Cyclo(1,8)-Cys-Pro-Arg-Gly-Gly-Ser-Val-Cys

Residue Assignment

Two-dimensional TOCSY and ROESY spectra were used to assign each proton resonance and the proton connectivities. The fingerprint region of the TOCSY spectrum shows through-bond connectivities between amide protons and side chain protons (Figure 1a and Table I). The amide proton of Arg3 shows through-bond connectivities to its respective HCα, HCβ, HCγ, and HCδ. The amide protons of Ser6 and Cys8 show connectivities to their respective HCα and HCβ. Both Gly4 and Gly5 have connectivities between their amide protons to HCα. Gly4 shows two geminal HCα, while Gly5 shows only one HCα in the TOCSY spectrum. The amide proton of Val7 has connectivities to its respective HCα, HCβ, and HCγ. The N-terminal proton of Cys1 cannot be identified because of fast exchange with water protons. Because Pro2 lacks the amide proton, this residue does not show any peak in amide proton connectivity.

The two-dimensional ROESY spectrum of the NHHCα region was used to identify the sequential connectivities of neighboring residues (Figure 1b). The resonances for Gly4 and Gly5 were identified by their respective d_Nα sequential connectivities to Arg3 and Ser6, respectively. The complete sequential “backbone walk” of all residues in the peptide, which consists of 12 ROE cross-peaks, including d_αN(i, i + 1) and d_Nα(i, i) ROE cross-peaks is: d_αN of Pro2-Arg3 (peak 1), d_Nα of Arg3 (peak 2), d_αN of Arg3-Gly4 (peak 3), d_Nα of Gly4 (peak 4), d_αN of Gly4-Gly5 (peak 5), d_Nα of Gly5 (peak 6), d_αN of Gly5-Ser6 (peak 7), d_Nα of Ser6 (peak 8), d_αN of Ser6-Val7 (peak 9), d_Nα of Val 7 (peak 10), d_αN of Val7-Cys8 (peak 11), and d_Nα of Cys8 (peak 12) (Figure 1b). Cys1-Pro2 connectivity was not observed because of fast exchange of Cys1 amide protons with water.

Chemical Shift Assignments

Because the chemical shift of HCα depends on the backbone dihedral angles (φ and ψ) and on the local chemical environment, the chemical shift index (CSI), which defines the difference between the measured chemical shifts and the chemical shifts when the peptide is in a statistical coil structure, can be used to elucidate the secondary structure adopted by the peptide.29 The CSI values can be negative (−1) or positive (+1), depending on whether the measured HCα chemical shift of an amino acid is lower or higher than the HCα chemical shift of the amino acid in a random-coil chemical. A peptide with a helical structure must have four or more sequential residues with negative CSI values; β-strand peptides are indicated by three or more sequential residues with positive CSI values.

In cyclo(1,8)-CPRGGSVC, no four sequential residues have negative SCI values, indicating that the peptide does not adopt a helical conformation (see Figure 2). The possibility of the peptide adopting β-strand structure can also be ruled out because no three consecutive residues have positive CSI values. This suggests that the peptide adopts the turn structure or a random coil. However, the CSI values could not be used to indicate these two structures; thus, the ROE pattern and ³J_NHHCα scalar coupling constants are needed to determine the conformation of the peptide.

ROE Pattern and ³J_NHHCα Scalar Coupling Constants

The conformation of cyclo(1,8)-CPRGGSVC peptide was determined by evaluating the ROE connectivities (Figure 1c) and ³J_NHHCα scalar coupling constant (Figure 3 and Table II). The possibility of the peptide adopting helical structure could be ruled out because there were no characteristic ROE cross-peaks from d_NN(i, i + 1), d_αN(i, i + 3), d_αN(i, i + 4), and d_αβ(i, i + 3). In addition, there were no characteristic ROE cross-peaks for β-strand structure, which are strong d_αN(i, i + 1) cross-peaks but without d_NN(i, i + 1). The presence of d_NN(i, i + 1) and d_αN(i, i + 1) ROE cross-peaks indicates the propensity to adopt a turn structure. Four d_NN(i, i + 1) ROE cross peaks were found, that is, Arg3-Gly4, Gly4-Gly5, Ser6-Val7, and Val7-Cys8 (Figure 1c). The d_NN(i, i + 1) ROE cross-peak of Gly5-Ser6 was too weak to observe. This suggests that the peptide has two turns. The first is a type-I β-turn at Pro2-Arg3-Gly4-Gly5 (PRGG), which is supported by the successive d_NN(i, i + 1) ROE cross-peaks of Arg3-Gly4 and Gly4-Gly5. The dihedral angle (φ) calculated from ³J_NHHCα coupling constants (Table II) confirms that the PRGG sequence has the propensity to form a type-I β-turn. The calculated φ of Arg3 (i + 1) and Gly4 (i + 2) fall within the 30° deviation limit from the ideal φ for type-I β-turn conformation26; Arg3 has a calculated φ of −79.46° (ideal φ for type-1 β-turn is −60°) and Gly4 has a calculated φ of −70.28° compared with the ideal φ of −90°. The small values of ³J_NHHCα of Arg3 (6.65 Hz), Gly4 (5.48 Hz), and Gly5 (5.67 Hz) also suggest that the PRGG sequence adopts turn structure (Table II).

Because we identified Pro2-Arg3-Gly4-Gly5 adopts a type-I β-turn structure, we expected to see a d_αN(i, i + 2) ROE cross-peak between Arg3 and Gly5. However, our observation shows that this d_αN(i, i + 2) ROE cross-peak is very weak. Our effort to lower the contour level results in a very complex ROESY spectrum that might cause difficulties for the readers to see the important d_Nα(i, i) and d_αN(i, i + 1) cross-peaks for the sequential “backbone walk” (Figure 1b).

The second turn is observed at the Gly5-Ser6-Val7-Cys8 (GSVC) sequence, and the determination of its turn type is not as straightforward and conclusive as in the PRGG sequence. The presence of successive d_NN(i, i + 1) ROE cross-peaks of Ser6-Val7 and Val7-Cys8 suggests that the GSVC sequence may adopt a type-I β-turn as in the PRGG sequence. The Ser6 (i + 1) has a calculated φ of –89.25° that deviates less than 30° from the ideal φ of –60°, while Val7 has a calculated φ of –96.98°, close to the ideal value of –90° (Table I). The ³J_NHHCα values for GSVC are larger than the values in the PRGG sequence. Only one residue (Gly5) has ³J_NHHCα value less than 6.0 Hz, while Ser6 has ³J_NHHCα value of 7.83 Hz, and Val7 has ³J_NHHCα values larger than 8 Hz (Table II), suggesting that the GSVC sequence may not have a classical β-turn structure. Because the determination of GSVC secondary structure using ROE patterns and ³J_NHHCα scalar coupling constant did not give a conclusive result, molecular dynamics simulation of the peptide using ROE restraint was performed to help elucidate the conformation of the GSVC sequence.

Molecular Model of the Peptide Structure

The three-dimensional structure of the peptide was determined by restrained molecular dynamics (MD) simulation using ROE distance restraints and peptide bond (ω) dihedral angle restraint of 180° ± 10° to keep all peptide bond as trans, except for Cys1–Pro2 peptide bond. From the last 500 ps of MD trajectories, all interproton distances are deviated less than 0.5 Å from the upper boundaries of ROE distance restraints (Table III). The MD-simulated φ angles for all residues in the peptide are within 30° of the φ values calculated from the ³J_NHHCα, except Gly5 that has one φ value outside the 30° deviation limit, presumably because of the flexibility of this residue (Table II). Ten structures from the MD simulation were taken as an ensemble of the peptide structures in solution (Figure 4a).

The presence of a type-I β-turn at the PRGG sequence was confirmed by the φ and ψ angles from the last 500 ps trajectories of the MD simulation. The average φ and ψ angles of residues at the PRGG sequence are (φ°, ψ°): Pro2 (–78.0° ± 8.2°, 163.2° ± 6.3°), Arg3 (–88.6° ± 7.0°, –24.8° ± 11.25°), and Gly4 (–72.1° ± 12.1°, 15.2° ± 13.8°). Gly5 has two sets of φ angles (–169.6° ± 19.3° and 171.7° ± 6.3°) and ψ angles (–174.7° ± 4.3° and 167.4° ± 8.1°) (Table II), which are due to the flexibility of the Gly5 backbone. The simulated and J-coupling calculated φ and ψ angles of residues Arg3 (i + 1) and Gly4 (i + 2) are less than 30° from the ideal angles for the type-I β-turn structure.26 Another indicator of β-turn structure is that the distance from the Cα atom at residue i to the Cα atom at residue i + 3 must be 7 Å or less.26 The CαCα distance between Pro2 (i) and Gly5 (i + 3) is 6.9 ± 0.2 Å, which confirms the presence of a β-turn at the PRGG sequence (Figure 4b). The turn structure of the PRGG sequence is stabilized by a hydrogen bond between the backbone carbonyl group of Pro2 and the amide proton of Gly 5 (3.1 ± 0.3 Å) (Figure 4b), which suggests that the PRGG motif in this peptide has a closed type-I β-turn conformation. The involvement of the Gly5 amide proton in hydrogen bonding is also indicated by the small value of the Gly5 temperature coefficient (–3.9 ppb/K) (Figure 5).

The MD simulation result did not confirm the presence of a type-I β-turn structure at the GSVC sequence. The simulated (φ°, ψ°) angles of Ser6 (i + 1) and Val7 (i + 2) were (–148.2° ± 6.9°, 99.7° ± 6.4°) and (–144.6° ± 8.5°, –56.5° ±7.0°), respectively. These values fall outside the 30° deviation limit from the ideal values of φ and ψ angles of (i + 1) and (i + 2) residues for the type-I β-turn structure.26 Visual observation of the ensemble of 10 structures from the last 500 ps trajectories of the MD simulation showed no β-turn structure on the GSVC sequence. However, a “turn-like” structure could be found at GSVCC sequence (Figure 4). This turn-like structure involves five residues and is stabilized by a hydrogen bond between the backbone carbonyl of Gly5 and the amide proton of Cys1 with a distance of 2.2 ± 0.3 Å. Meanwhile, the CαCα distance between Gly5 and Cys1 is 6.4 ± 0.4 Å, which is close to the ideal CαCα distance for type-I (6.31 Å) and type-II (6.26 Å) α-turns.30 Because the GSVCC sequence is stabilized by a hydrogen bond and the CαCα distance between Gly5 and Cys1 falls within the ideal CαCα distance for a type-I or type-II α-turn structure, one might think that the GSVCC sequence adopts these two types of α-turn structure. However, simulated φ and ψ angles of GSVCC residues fall outside the 30° deviation limit from the ideal φ and ψ angles for type-I or type-II α-turn structure.30 The high ³J_NHHCα coupling constant values of Ser6 and Val7 suggest that the “turn-like” structure at GSVCC residues does not belong to the α-turn structure. It is possible that the “turn-like” structure around residues Gly5-Ser6-Val7-Cys8-Cys1 was randomly formed because of the side-chain cyclization between Cys1 and Cys8. Therefore, the cyclo(1,8)-CPRGGSVC has adopted the type-I β-turn conformation at the PRGG sequence and a “turn-like” structure at the GSVCC sequence.

Hydrogen Bonding Analysis of Amide Protons

In peptides, the amide proton chemical shifts are sensitive to temperature changes. The amide protons involved in intramolecular hydrogen bonding are less sensitive to temperature changes than amide protons that are solvent-accessed and not involved in intramolecular hydrogen bonding network. The involvement of amide protons in hydrogen bonding can be measured by calculating the temperature coefficient (Δδ/ΔT), which is the slope of amide proton chemical shift changes over different temperatures.31 Small values of Δδ/ΔT (−3 ppb/K or less) indicate amide protons that are involved in intramolecular hydrogen bonding and less exposed to the solvent. In contrast, high values of Δδ/ΔT (−5 ppb/K or more) suggest amide protons that are more exposed to the solvent.32

Our result indicates that Gly4 and Gly5 may have more involvement in intramolecular hydrogen bonding than the other residues (see Figure 5). The Δδ/ΔT values of Gly4 and Gly5 are −3.6 ppb/K and −3.9 ppb/K, respectively. The Δδ/ΔT values of the other residues are (ppb/K): −14.0 (Arg3), −8.3 (Ser6), −6.7 (Val7), and −8.7 (Cys8). This suggests that there are two possible intermolecular hydrogen bonds that involve the amide protons of Gly4 and Gly5. The backbone carbonyl groups that might form hydrogen bonds with Gly4 or Gly5 can be determined by observing the three-dimensional structure of the peptide determined by ROE-restraint molecular dynamics simulation.

Interestingly, the three-dimensional structure of the peptide shows two intramolecular hydrogen bonds between the backbone carbonyl group of Pro2 and the amide protons of Gly4 and Gly5 (Figure 4b). The first hydrogen bond between Pro2 and Gly5 is responsible for the stability of the β-turn of the PRGG motif as mentioned before. The second hydrogen bond between Pro2 and Gly4 (Pro2 COGly4 NH = 2.9 ± 0.4 Å) is suggested to strengthen the rigidity of the β-turn conformation at the PRGG sequence.

Peptide Synthesis

Cyclo(1,8)-CPRGGSVC was synthesized by standard Fmoc solid phase peptide chemistry on Fmoc-L-Cys(Trt)-PAL-PEG-PS resin (Applied Biosystems, Foster City, CA) using the automated peptide synthesizer (PerSeptive Biosystems, Framingham, MA). All couplings were carried out in DMF using HATU/DIEA as activator and 20% piperidine as deblocking solution. On completion of the synthesis, the peptide was cleaved from the resin using a cleavage mixture (reagent B) consisting of 95% trifluroacetic acid (trifluoracetic acid (TFA), 0.5% thioanisole, 0.3% ethanedithiol, and 0.2% anisole for 2 h at room temperature under nitrogen blanket. Then, the mixture of resin and reagent B was filtered, and the filtrate was added into cold ether to yield a white precipitate of the crude peptide. The peptide was purified using semi-preparative HPLC on a C-18 reverse phase column (Varian, Palo Alto, CA). Cyclization of the pure linear peptide was accomplished by air oxidation in ammonium bicarbonate buffer (0.05M, pH 8.5) overnight in high dilution (0.06 mM). After cyclization, the peptide solution was concentrated using a rotary evaporator to 10 mL followed by lyophilization. The lyophilized powder containing peptide (10%) and ammonium bicarbonate (90%) was then purified on semi-preparative HPLC using a C-18 column. The pure fractions of the peptide were pooled and lyophilized. The molecular weight of the peptide was determined by MALDI-TOF mass spectrometry.

NMR Spectroscopy

Two-dimensional NMR experiments were carried out on a Bruker Avance spectrometer operating at 800 MHz for protons. The temperature dependence of the amide proton chemical shifts was determined by one-dimensional proton NMR experiments between 298 and 323 K on a 400 MHz Bruker Avance spectrometer. Cyclo(1,8)-CPRGGSVC peptide (5 mg) was dissolved in 900 μL phosphate buffer (pH 5.0, 20 mM) and 100 μL D₂O. Sodium 3-(trimethylsilyl)-1-propiosulfonic acid (DSS) was used as an internal reference for chemical shift calibration in ROESY and TOCSY experiments. Meanwhile, tetramethylsilane (TMS) was used for an internal standard in temperature dependence one-dimensional NMR experiments. The ³J_NHHCα scalar coupling constants were calculated from the backbone amide region of F1 row of ROESY spectrum using Karplus equation:35

Total correlation spectroscopy (TOCSY) and rotating-frame nuclear Overhauser enhancement spectroscopy (ROESY) experiments were carried out by presaturation of water during relaxation delay. Proton resonance assignments and spin system identification were made using a TOCSY spectrum with mixing time of 62 ms. ROE assignments and identification of sequential ROE connectivities were made using ROESY spectra with mixing times of 200 ms. The ROE cross-peak volumes were classified as strong, medium, and weak, and they were correlated to distances of d_Nα(i, i), d_αN(i, i + 1), and d_NN(i, i + 1) for which the upper limits were assigned as 2.4, 3.5, and 5.0 Å, respectively. Spectra visualization was performed using CARA (Institute of Molecular Biology and Biophysics, ETH Zurich, Switzrland) and MestReNova (Mestrelab Research S.L., Spain) softwares.

Molecular Dynamics Simulation

NMR-restraint molecular dynamics simulation was performed in MOE (Molecular Operation Environment) software (Chemical Computing Group, Quebec, Canada) using AMBER99 force field. The potential energy terms for AMBER99 force field is expressed in an equation below36:

where the total potential energy is the sum of bond stretching, angle bending, torsional angle, van der Waals, and electrostatic energies. ROE distance restraints were applied with upper limits of 2.8, 3.5, and 5.0 Å for strong, medium, and weak ROE peaks, respectively. A peptide bond (ω) dihedral angle restraint of 180° ± 10° was applied to keep all peptide bond as trans, except for Cys1-Pro2 peptide bond because the ROESY spectrum does not show the presence of d_αα cross-peak between Cys1 and Pro2 as an indicator of cis X-Pro, and also the d_αδ cross-peak between Cys1 and Pro2 as an indicator of trans X-Pro. A constant force of 10 kcal/mol · Å was applied. The starting structure was simulated at high temperature (600 K) and NPT (constant pressure) in vacuo using dielectric constants of 1 and 80 for 30 ps to relax the structure. The final structure was then soaked in a waterbox (with periodic system) and was simulated at 300 K for 1000 ps (NPT), time-step of 0.1 fs, temperature and pressure response of 0.1, respectively. Data were collected every 1 ps. The stability of the simulated molecule was monitored by observing the backbone RMSD values. The RMSD values of the last 500 ps of molecular dynamics simulation showed equilibrium condition with lesser fluctuation compared with the first 500 ps of molecular dynamics simulation. The trajectories from the last 500 ps of molecular dynamics simulation were taken for analysis. The following criteria were used to determine the best fit solution conformation of peptide: (i) the proton should have an interproton distance <0.5 Å compared with the upper boundaries of distances from the ROE restraint; (ii) the conformation had φ angles within 30° of the calculated φ derived from ³J_NHHCα.25