Hydroxylation-induced Stabilization of the Collagen Triple Helix

The collagen triple helix is one of the most abundant protein motifs in animals. The structural motif of collagen is the triple helix formed by the repeated sequence of -Gly-Xaa-Yaa-. Previous reports showed that H-(Pro-4(R)Hyp-Gly)10-OH (where `4(R)Hyp' is (2S,4R)-4-hydroxyproline) forms a trimeric structure, whereas H-(4(R)Hyp-Pro-Gly)10-OH does not form a triple helix. Compared with H-(Pro-Pro-Gly)10-OH, the melting temperature of H-(Pro-4(R)Hyp-Gly)10-OH is higher, suggesting that 4(R)Hyp in the Yaa position has a stabilizing effect. The inability of triple helix formation of H-(4(R)Hyp-Pro-Gly)10-OH has been explained by a stereoelectronic effect, but the details are unknown. In this study, we synthesized a peptide that contains 4(R)Hyp in both the Xaa and the Yaa positions, that is, Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2 and compared it to Ac-(Gly-Pro-4(R)Hyp)10-NH2, and Ac-(Gly-4(R)Hyp-Pro)10-NH2. Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2 showed a polyproline II-like circular dichroic spectrum in water. The thermal transition temperatures measured by circular dichroism and differential scanning calorimetry were slightly higher than the values measured for Ac-(Gly-Pro-4(R)Hyp)10-NH2 under the same conditions. For Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2, the calorimetric and the van't Hoff transition enthalpy ΔH were significantly smaller than that of Ac-(Gly-Pro-4(R)Hyp)10-NH2. We postulate that the denatured states of the two peptides are significantly different, with Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2 forming a more polyproline II-like structure instead of a random coil. Two-dimensional nuclear Overhauser effect spectroscopy suggests that the triple helical structure of Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2 is more flexible than that of Ac-(Gly-Pro-4(R)Hyp)10-NH2. This is confirmed by the kinetics of amide 1H exchange with solvent deuterium of Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2, which is faster than that of Ac-(Gly-Pro-4(R)Hyp)10-NH2. The higher transition temperature of Ac-(Gly-4(R)Hyp-4(R)Hyp)10-NH2, can be explained by the higher trans/cis ratio of the Gly-4(R)Hyp peptide bonds than that of the Gly-Pro bonds, and this ratio compensates for the weaker interchain hydrogen bonds.

The collagen triple helix is one of the most abundant protein motifs in animals. Each of the three polypeptide chains forms a polyproline-II like structure, and these structures twist around each other to form a right-handed superhelix (1,2). To form this structure, a sequence with Gly in every third position is required. The general sequence is -Gly-Xaa-Yaa-, and the Xaa and Yaa positions contain a high content of proline and hydroxyproline. Twenty-seven types of collagens and more than fifteen proteins that form a triple helix, but are not named collagens such as collectins, ficolins, and scavenger receptors (3)(4)(5), occur in humans. In vertebrates, most of the proline residues in the Yaa position of the -Gly-Xaa-Yaa-repeated sequence are post-translationally modified to 4(R)Hyp 1 by prolyl 4-hydroxylase (EC 1.14.11.2) (6,7). In invertebrates, collagens also play a crucial role. For example, Caenorhabditis elegans has more than 150 genes for collagenous proteins, mostly for cuticle collagens (8). Annelid cuticle collagen has 4(R)Hyp residues in the Xaa position and galactosylated threonine residues in the Yaa position of the repeated -Gly-Xaa-Yaa-sequences, which are not found in vertebrates (9). Mutations in individual cuticle collagen genes can cause exoskeletal defects that alter the shape of the animal. In general, the content of 4(R)Hyp is related to the melting temperature (T m ) of collagens (10,11), and the T m of dispersed collagen molecules in solution is a few degrees lower than the body temperature of the species from which the collagen was derived (12).
The knowledge of the structural basis of triple helix stabilization by proline and hydroxyproline residues has significantly improved recently. The pyrrolidine ring structure, including puckering, the peptide dihedral angles of , , and , and the propensity of trans conformation of X-Pro (X-4(R)Hyp) bonds play crucial roles in the stability. Raines and his colleagues (1,24) pointed out the importance of the inductive and stereoelectronic effects of the hydroxyl group of 4(R)Hyp. The importance of the pyrrolidine ring puckering of imino acids was shown with quantum and molecular mechanical analysis (25,26) and was supported by the "propensity-based" stabilization of proline in the Xaa position and 4(R)Hyp in the Yaa position and the destabilization of 4(R)Hyp in the Xaa position (27,28).
Recently, a number of high resolution structures of triple helical peptides have been reported (29 -34). These studies have shown that the conformation of proline in the Xaa position is C␥-endo (puckering down), and the proline (or 4(R)-hydroxyproline) in the Yaa position is C␥-exo (puckering up). In the crystals, each polypeptide forms a left-handed helix with 7/1 screw symmetry. Three peptides form a right-handed helix with 7/2 symmetry, when the Xaa and the Yaa are imino acids, or more loose 10/3-like helix symmetry in the imino acid-poor regions (35). The puckering of imino acids in H-(Pro-4(R)Hyp-Gly) 10 -OH was also measured by two-dimensional NMR in D 2 O at 10°C. The Xaa position Pro is C␥-endo (puckering down), and the Pro (Hyp) in the Yaa position is C␥-exo (puckering up) (36).
Brodsky and her colleagues (37) pointed out the importance of water molecules that interact with the hydroxyl group of 4(R)Hyp. The contribution of bound water to the stability of the collagen triple helix is controversial (38). There are 36 bound water molecules per asymmetric unit (21 residues) in H-(Pro-Pro-Gly) 10 -OH (27). Three water molecules per tripeptide unit may interact with the peptide chain directly. One interacts with the carbonyl oxygen of Gly, the other two are bound to the carbonyl oxygen of Pro in the Yaa position. There are two sharp peaks in the distribution of bound water molecules, one is at a distance of 2.7 Å from the nearest peptide atom, and the other is at 3.6 Å (39). The number of bound water molecules is similar in H-(Pro-Pro-Gly) 9 -OH (34). In contrast, the analysis of the peptide containing -Gly-Pro-4(R)Hyp-revealed 49 bound water molecules (40). However, these bound water molecules found in crystal structures exchange very rapidly with bulk solvent when these peptides are studied in solution by NMR (41), and their role in the stabilization is unclear.
Circular Dichroism-Circular dichroism spectra were recorded on an Aviv 202 spectropolarimeter using a Peltier thermostatted cell holder and a 1-mm path length rectangular cell (Starna Cells Inc., Atascadero, CA). Peptide concentrations were determined by amino acid analysis. The wavelength spectra represent at least an average of 10 scans with 0.1-nm resolution. The CD data were analyzed with SCIENTIST for Windows (MicroMath Research, St. Louis, MO) using the cubic equation for the all-or-none model described in Frank et al. (42).
Analytical Ultracentrifugation-Sedimentation equilibrium runs were performed in a Beckman Model E analytical ultracentrifuge (Beckman Instruments) equipped with a scanner. 12-mm Epon double-sector cells in an An-F Ti rotor were used. The peptides were analyzed in 20 mM phosphate buffer, pH 7.2, containing 150 mM NaCl. The peptide concentrations were adjusted to 0.2-1.2 mg/ml. Sedimentation equilibrium measurements were carried out at 4 -6°C with rotor speeds of 36,000 -40,000 rpm. Molecular masses were evaluated from ln A versus r 2 plots, where A represents the absorbance and r is the distance from the center of rotation. A partial specific volume of 0.71 ml/g was used for all calculations. The data were analyzed using a least square method with the SCIENTIST for Windows software (MicroMath Research).
NMR Spectroscopy-NMR spectra were recorded on a Bruker AMX-400 spectrometer, operating at 400.14 MHz. The 90°pulse width was 9 s, and a low power 2-s presaturation pulse was applied to suppress the H 2 O resonance. The spectra were recorded as 16,384 points for the one-dimensional spectra and as 1,024 ϫ 512 data point sets for the two-dimensional spectra. The NOESY data were collected with timeproportional phase incrementation in the indirect dimension, at mixing times between 30 and 120 ms, and a total recording time of about 10 h. Total correlation spectroscopy data were collected with various mixing times, ranging from 30 to 90 ms. The data were processed with Swan-MR or nmrPipe to 1024 ϫ 1024 real data sets after application of a 60°phase-shifted sin 2 function and Fourier transformation for the two-dimensional spectra; baselines were straightened with polynomials as needed. Spectra were referenced to 0 ppm via internal 2,2-dimethylsilapentane-5-sulfonate or via the water resonance (4.71 ppm at 30°C). Final visualization and analyses of the two-dimensional data sets were performed using NMRView (43) or Swan-MR.
The H/D exchange experiment was performed by recording a series of one-dimensional spectra after dissolving the peptides into D 2 O, buffered to pD 4.7 (uncorrected) with 100 mM acetate/acetic acid-d 4 . The peak intensities were evaluated by measuring the heights on the same absolute scale. This was found to give essentially the same results as when using the peak integrals.
Differential Scanning Calorimetry-The temperature dependence of the partial heat capacity was measured in an N-DSC II differential scanning calorimeter (Calorimetry Science Corp.). The heating rate was 7.5°C/h, and the data were collected using the software provided by the manufacturer. The data were analyzed using the cubic equation for the all-or-none model described in Frank et al. (42).

RESULTS
The CD spectra of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 in water at 4°C are shown in Fig. 1A together with the spectra of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 , Ac-(Gly-4(R)Hyp-Pro) 10 -NH 2 , and Ac-(Gly-Pro-Pro) 10 -NH 2 measured under the same conditions. The spectrum of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 shows a positive maximum at 225 nm and a negative minimum at 197 nm. In this regard, it most closely resembles the spectrum of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 , but both extrema are smaller than those of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 . The spectrum of Ac-(Gly-Pro-Pro) 10 -NH 2 is slightly red-shifted when compared with the first two spectra. The spectrum of Ac-(Gly-4(R)Hyp-Pro) 10 -NH 2 is significantly different with a positive maximum at 222 nm and a negative minimum at 200 nm. At 95°C the positive maximum disappears in all peptides except for Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 , which still shows a positive maximum at 225 nm. The negative minimum is at 203 nm for all peptides (Fig. 1B). The persistence of the positive maximum at 225 nm with increasing temperature is shown in Fig. 1C, which also indicates the shift of the negative minimum from 197 to 203 nm. The change in ellipticity at 225 nm as a function of temperature shows that all peptides with the exception of Ac-(Gly-4(R)Hyp-Pro) 10 -NH 2 show a cooperative transition (Fig. 2). The midpoint of the transition (T m ) of the heating scan of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 is at 74°C, the T m of the much broader transition of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 is near 80°C, and the T m of Ac-(Gly-Pro-Pro) 10 -NH 2 is at 47°C. The folding and unfolding transitions of collagens and collagen-like peptides are slow, and a significant hysteresis is observed with fast scanning rates and low peptide concentrations (44,45). For the evaluation of thermodynamic data, equilibrium transition curves were required. Fig. 3A shows the temperature dependence of the heat capacity for Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 and Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 measured at a 2 mM peptide concentration and with a heating/cooling rate of 7.5°C/h. When a high concentration and a slow heating/cooling rate were used, the unfolding/refolding curves were practically identical. The hysteresis observed at lower concentrations, even present in peptides with cross-links (46), disappeared under these conditions. Fig. 3B shows the CD transition curves measured at 235 nm under the same conditions. This wavelength was chosen because at 225 nm the CD signal fell outside the dynamic range of the instrument at this high concentration. The thermodynamic values of these transitions for the concentration-dependent all-or-none model (47) are given in Table I together with the values determined by differential scanning calorimetry. The van't Hoff enthalpy and entropy changes as well as the calorimetric enthalpy change were significantly smaller for Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 (⌬H cal ϭ Ϫ6.2 kJ/mole of tripeptide unit) than the calorimetric enthalpy change for Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 (⌬H cal ϭ Ϫ12.0 kJ/mole of tripeptide unit). The enthalpy change of Ϫ12.0 kJ/mole for Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 is similar to the values of Ϫ13.4 and Ϫ13.9 kJ/mole obtained for H-(Pro-4(R)Hyp-Gly) 10 -OH (47,48).

TABLE I Thermodynamic values for the thermal transitions of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 and Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2
The thermodynamic values were calculated from the CD transition curves and from differential scanning calorimetry (DSC) experiments. For CD and DSC the peptide concentration was 2 mM, and the rate of heating/cooling was 7.5°C/h. The values are given per mole of tripeptide unit.   10 -NH 2 and Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 in D 2 O solution (Fig. 4). Both data sets showed cross-peaks between nearly all proton resonance positions, clearly demonstrating that the peptides are triple helical at 30°C (49 -51). Thus the CD, calorimetric, and analytical ultracentrifugation data about these peptides are confirmed. A comparison between of the two data sets suggested that there might be additional broadening in the NOESY spectrum of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 . This stronger broadening might arise from less mobility of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 when compared with Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 .

4(R)Hyp-4(R)Hyp)
We therefore performed an H/D exchange experiment to test for potential unfolding of the helix by examining the rate of the Gly NH exchange (Fig. 5). The results of this H/D exchange

Peptide
Experimental mass Theoretical monomer mass Theoretical trimer mass

Da
Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 8300 Ϯ 500 2731.9 8195.7 Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 8979 Ϯ 500 2891.9 8675.7 Ac-(Gly-4(R)Hyp-Pro) 10 -NH 2 2932 Ϯ 700 2731.9 8195.7 Ac-(Gly-Pro-Pro) 10 -NH 2 7256 Ϯ 500 2571.9 7715.7 experiment show that at least two different Gly NH populations exist: a minor one with a faster decay, and a slower, dominant one. Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 retains a smaller fraction of the minor population at the initial time measured (1 min after mixing), and faster exchange of both populations (Table III). Thus, the H/D exchange data indicate that Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 has less overall structural integrity, insofar as H/D exchange would indicate. DISCUSSION The results show that the peptide Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 forms a more stable triple helix than the peptide Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 , when evaluated by the T m value. The steric restrictions imposed by the proline ring play a dominant role in the stability of the triple helix. The interchain hydrogen bonds between GlyNH . . . OC(Xaa) of an adjacent chain are also recognized as important for the stability. The additional stabilization observed by the hydroxylation of proline residues in the Yaa position to 4(R)-hydroxyproline was explained in two different models. In the first model the hydroxyl group of 4(R)Hyp is thought to form water bridges between backbone groups, and in the second model the inductive effect of the hydroxyl group influences the ring puckering of the pyrrolidine ring and changes the cis/trans ratio of the peptide bond. The triple helix of (Pro-Pro-Gly) 10 shows a preference of down puckering of the proline ring in the Xaa position and an up puckering of the proline ring in the Yaa position. Previous studies have shown that the -Gly-4(R)Hyp-Pro-tripeptide units destabilize the triple helix (19). This is explained by the preference of up puckering of the proline ring of 4(R)Hyp in the Xaa position, and, conversely, it also explains the stabilizing effect of 4(R)Hyp in the Yaa position. Therefore, we hypothesized that Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 should have an intermediate stability between Ac-(Gly-4(R)Hyp-Pro) 10 -NH 2 and Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 . However, the experimental data ( Figs. 2 and 3) indicate that Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 is a little more stable than Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 .
The unfolding of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 measured by CD and DSC shows that the thermal transition is broader than that of Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 . This is consistent with the smaller enthalpy change observed by CD and DSC for the transition of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 . However, this smaller enthalpy change is compensated by a smaller entropy change, which results in a higher T m for Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 . The presence of 4(R)Hyp in the Xaa position, with a "wrong" pyrrolidine ring puckering for that position, results in a less tight triple helix, in which the amide hydrogens of glycine exchange with solvent faster than in the triple helix formed by Ac-(Gly-Pro-4(R)Hyp) 10 -NH 2 .
Additional stabilization of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 might come from the higher trans/cis ratio of peptide bonds in the denatured state. The magnitude of the difference in the enthalpy and entropy changes observed for the two peptides could also indicate that the two denatured states are significantly different. For Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 the CD signal at 225 nm remains positive for the denatured state. Poly(4(R)-hydroxyproline) was shown to form a hydrogenbonded left-handed helix in aqueous solutions whose CD spectra had a negative peak at 205 nm and a positive peak at 225 nm (52). The negative peak of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 shifts from 197 to 203 nm with increasing temperature, while the positive peak remains at 225 nm, indicating that this peptide does not change the conformation drastically when transitioning to a monomeric, non-triple helical state. This is a similar finding as for the unfolded state of Ac-(Gly-3(S)Hyp-4(R)Hyp) 10 -NH 2 , a peptide that does not form a triple helix (49). For the structure of this peptide, the possibility of a polyproline II-like conformation has been suggested. This is
The stability of Ac-(Gly-4(R)Hyp-4(R)Hyp) 10 -NH 2 is also consistent with the stabilizing influence of 4(R)Hyp in the Xaa position, as long as the Yaa position is not proline (22,23). Recently, two reports were published showing that H-(4(S)Flp-Pro-Gly) 7 -OH (17) and H-(4(S)Flp-Pro-Gly) 10 -OH (18) form a triple helix in aqueous solution. If the pyrrolidine ring structure and inducing dihedral angles contribute significantly to the triple helix stability, -Gly-4(S)Hyp-4(R)Hyp-peptide should form a very stable helix, unless there is steric hindrance. Vitagliano et al. (27) predicted steric hindrance by modeling, if the proline in the Xaa position of -(Pro-Pro-Gly) n -triple helix is replaced by 4(S)Hyp. Quantum mechanical studies also come to this conclusion (25). However, this steric hindrance might be minimized by peptides with 4(S)Flp, which is smaller than 4(S)Hyp (17). Replacing the hydroxyl group with fluorine results in little steric hindrance.
The occurrence of 4(R)Hyp in the Xaa position is mostly found in collagens of invertebrates (9,53). No collagen sequences with Gly-4(R)Hyp-4(R)Hyp tripeptide units have been identified. However, it was shown recently that 4(R)Hyp can be incorporated in both the Xaa and Yaa positions of a collagenlike molecule in Escherichia coli when the NaCl concentration is high (54). Our results show that comparable stabilities should be obtained for collagen-like peptides expressed in this system. Collagen has been used extensively for tissue engineering purposes (55)(56)(57). For such purposes, the regulation of the triple helical stability is important, and collagenous peptides or recombinant collagen molecules with Gly-4(R)Hyp-4(R)Hyp tripeptide units might be useful.