Solution Structure of Parathyroid Hormone Related Protein (Residues 1–34) Containing an Ala Substituted for an Ile in Position 15 (PTHrP[Ala15]-(1–34))*

The structure of human parathyroid hormone (PTH) related protein (residues 1–34) containing an Ala substituted for an Ile in position 15 was studied by two-dimensional proton nuclear magnetic resonance spectroscopy. This mutant retains quite high levels of adenylate cyclase activity based on slightly reduced PTH receptor binding capacity. Three segments of helix were revealed extending from His5 to Lys11, Lys13 to Arg19, and from Phe22 to Thr33/Ala34, with a decided kink between the first two helices around Gly12. N- and C-terminal helices were stabilized by charged and hydrophobic side chain interactions between His5 and Glu30, Asp17 and both His9 and His25, and between Leu8 and Ala29, resulting in a globular molecule occupying a single conformation. While the structure of the entire mid-molecule region differed greatly from the structure of the native peptide, the structure of both N- and C-terminal regions remains essentially unaltered. The residues responsible for initiating signal transduction in the mutant are located in the vicinity of the residues responsible for receptor binding. The C-terminal amphipathic helix forming the receptor binding site exhibits reduced binding as a result of the closely applied N-terminal signal transduction-activating region. Although not contributing directly to receptor binding, the N-terminal region can sterically affect hormone binding through modifications to certain N-terminal side chains.

Parathyroid hormone related protein (PTHrP) 1 was discovered because of its expression by cancers, commonly of squamous origin, giving rise to the syndrome of humoral hypercalcemia of malignancy (HHM) (for reviews, see Refs. 1 and 2). In these cancer patients, PTHrP is secreted into the circulation and acts on classical PTH receptors in kidney and bone to conserve calcium and reproduce many of the features of hyperparathyroidism.
In normal physiology, PTHrP acts largely as a paracrine regulator in several tissues (1,2), including smooth muscle, at many locations where it acts via PTH/PTHrP receptors to relax smooth muscle. PTHrP is an example of a peptide possessing discrete structural domains for receptor binding and agonist activity. Small structural modifications are able to transform a fully functional receptor agonist into a strong competitive antagonist. Only the N-terminal domain (residues 1-34) of PTHrP shares any sequence homology with PTH (3). This domain exhibits full PTH-like adenylate cyclase agonist activity (4) although only 10 of the residues are identical with PTH, with 8 in the first 13. However, many residues in the putative primary receptor binding site (24 -31) have either conserved charged or hydrophobic side chains (5). Several binding and activity studies have been performed using both N-terminal and C-terminal truncated PTH and PTHrP peptides (4, 6 -10). These demonstrated that agonist activity is dependent on the presence of the intact N-terminal residues while receptor binding was dependent on the presence of the C-terminal residues. Hence, the signal transduction site is confined to the N terminus, whereas the bulk of the receptor binding site is located within the C-terminal region of the sequence 1-34. Part of this region (segment 30 -34) contains the residues necessary for the stimulation of DNA synthesis (11,12).
NMR studies have been carried out on both PTHrP-(1-34) and PTH- , in a range of solvents (5,(13)(14)(15)(16)(17)(18)(19)(20), as well as on the circulating fragment PTH-(1-37) (21). PTH and PTHrP both exhibit nascent structures in water, which are stabilized in dilute F 3 EtOH-d 2 . PTH is less well conserved, with the Nand C-terminal helices showing no evidence that they interact, unlike PTHrP-  in water (13). Thus PTHrP-  in dilute F 3 EtOH-d 2 more closely resembles PTH-  in water or salt buffer. Both exhibit a hydrophobic core formed by the interaction of the side chains of residues 15 and 23 and similar C-terminal helices (16,21). Since PTH-  and PTHrP-(1-34) exhibited a flexible hinge at residues 13/14 (5, 14 -21), no particular interaction between the N-and C-domains could be detected except for weak interactions in water detected with a very long mixing time, such as those between residues 2 and 31, and 8 and 28 as examples (13). These investigations not only revealed the presence of separate N-and C-terminal helices, but with the exception of PTH in the presence of 10% F 3 EtOH-d 2 , a reverse turn was found in the intervening segment. In this latter case, however, the C-terminal helical segment commenced at Ser 17 (5,14), and only a hinge region separated the two helices. F 3 EtOH/water mixtures possess altered hydrogen-bonding properties compared with water alone, and thus amide proton exchange rates are reduced, leading to improved resolution and an increase in NOE intensities (22). Moreover, this solvent increases helix content in those peptides that already have a propensity to form helices in water (23), which is the case with PTHrP (13) and PTH polypeptides (16,17,24). However, the mild hydrophobicity can have the effect of weakening hydrophobically stabilized regions of polypeptide structures (25). Thus, any interaction between the N-and C-terminal helices in PTH/PTHrP-(1-34) would be expected to be weakened under these conditions.
In an attempt to disrupt the relatively strong and consistent hydrophobic packing found between residues 15 and 23 in both PTH (16,21) and PTHrP (18,20), a new active analog of PTHrP-(1-34)-amide was synthesized with Ile 15 replaced by an Ala. The structure was examined to check for any changes to the hydrophobic packing that could then be quarantined from involvement in direct receptor binding. The intention was to determine the part or parts of the structure remaining unaltered that could be responsible for receptor binding. Since signal transduction is separate from receptor binding and since it appeared that the residues responsible for receptor binding, confined to the C-terminal half of the peptide, were not closely associated with the N terminus, then the antagonist-and agonist-specific sites could have been well separated. A comprehensive understanding of the structure-activity relationship of the PTHrP/PTH pair of molecules requires a detailed knowledge of their structure and mobility. In particular, the design of antagonists useful for the control of hypercalcemia of malignancy firstly requires that the stabilized structure of the receptor binding domain be determined. We have thus studied the stabilized structure of the active mutant PTHrP[Ala 15 ]-(1-34)-amide in 8% F 3 EtOH-d 2 and compared the principal structural features of the hormone with the tertiary solution structures of other homologous hormones to identify the residues responsible for receptor binding and their spatial relationship with residues responsible for signal transduction. 15 ]-(1-34)-amide was synthesized on an Applied Biosystems 430A peptide synthesizer using tbutyloxycarbonyl amino acid derivatives (26). The Glu, Ser, and Thr side chains were protected using dinitrophenol while the Lys side chains had chlorobenzyl protection, and Arg side chains were protected with the nitro group. Dinitrophenol deprotection of the peptide was performed with 20% ␤-mercaptoethanol in dimethylformamide. Cleavage from the benzhydrylamine resin was achieved with anhydrous HF containing 10% (v/v) anisole (27). The peptide was extracted with 60% (v/v) acetonitrile/trifluoroacetic acid solvent (0.1% v/v) in water, after which it was rotary evaporated. The aqueous solution was lyophilized, and the crude peptide was purified using preparative reverse-phase high performance liquid chromatography. The remaining trifluoroacetic acid was removed using G-10 gel filtration, and the peptide was lyophilized. Amino acid analyses of the peptide were in close accordance with the theoretical values. The C-terminal residue was amidated.
Specific Binding-PTH-receptor binding of PTHrP peptides to PTHreceptor positive UMR-106 -01 cells was assessed as described previously (29)  NMR spectra were obtained on a Bruker AMX-600 spectrometer operating in the Fourier transform mode with quadrature detection in both directions without sample spinning and with the use of a calibrated external temperature control unit. One-dimensional spectra were collected using a spectral width of 6024 Hz with pre-acquisition delay of 2 s. A total of 256 summed free induction decays were collected in a data block of 32 K. Signal-to-noise ratio was 670 with the dedicated proton probe.
Signal Assignments-Spin systems were assigned using standard procedures (32). The complete spin systems from all residues were identified using two-dimensional TOCSY and double quantum-filtered correlation spectroscopy spectra. The TOCSY spectra were obtained with a MLEV-17 pulse sequence (33, 34) using a mixing time of 100-and 2.5-ms trim pulses, which enabled connectivities from the amino acid backbone NH resonances to all the coupled side chain proton resonances to be measured. Spin-lock field strength was 9 kHz. Suppression of the water resonance was achieved using continuous low power (65 db) O1/O2 coherent irradiation throughout the relaxation (1.8 s) and mixing periods (35) with a stimulated cross-peaks under bleached ␣ modification to reduce cross-peak suppression under the solvent (36). Twodimensional phase-sensitive NOESY experiments (37,38) were used to make sequence-specific resonance assignments as well as for estimations of proton-proton distance constraints. These were recorded using the time-proportional phase-increments method in 1 (39) and also using the stimulated cross-peaks under bleached ␣ modification. A mixing time of 200 ms was used to provide the distance constraints table since it provided sufficient cross-peak intensity without spindiffusion effects becoming problematical. The sweep widths in both 1 and 2 were 6024 Hz. 4 K data points were acquired in 2 with 512-1024 data points collected in 1 , with each free induction decay consisting of 128 -160 scans. The data were zero-filled to 4 -8 K ϫ 1-2 K and then apodized using Gaussian multiplication in 2 and shifted sine-bell in 1 prior to Fourier transformation. Third-order polynomial functions were employed to correct base lines. Data were analyzed using XWIN NMR on a Silicon Graphics Indy workstation. Chemical shifts were referenced to trimethylsilylpropanesulfonic acid.
Distance Geometry Calculations-The 365 measured NOE crosspeaks, of which 195 were structure-determining interresidues, were separated into three conservative distance categories depending on contour intensity. Strong NOE were given an upper distance constraint of 0.30 nm, medium NOE a value of 0.40 nm, and weak NOE a value of 0.45 nm. Corrections for pseudoatoms were applied wherever stereospecific identifications were not obtained. A total of 2000 distance geometry structures were calculated from random starting configurations using the program DIANA 2.8 (40) on a Silicon Graphics Indy workstation. This technique utilizes least squares minimization in torsion angle space incorporating a variable target function, which considers progressively longer range constraints as the global structure is generated.
Dynamic Simulated Annealing-The best 20 distance geometry structures possessing the lowest penalty values were refined in X-PLOR 3.1 (41) utilizing a dynamic simulated annealing protocol (42) run on Silicon Graphics Indigo R4000 and Indy work stations. Nonbonded interactions were not considered in the DIANA distance geometry structures. Consequently, they possessed poor potential energies. 500 cycles of computationally inexpensive conjugate gradient energy minimization were applied in the initial stage of the simulation keeping the atom co-ordinates fixed to improve the distance geometry starting structures. Covalent geometry was constrained with parameters k bond ϭ 20 kJ⅐mol Ϫ1 ⅐nm Ϫ2 , k angle ϭ 2000 kJ⅐mol Ϫ1 ⅐rad Ϫ2 , k improper ϭ 2000 kJ⅐mol Ϫ1 ⅐rad Ϫ2 , and k cdihed ϭ 0 kJ⅐mol Ϫ1 ⅐rad Ϫ2 (41). The non-bonded interactions were modelled with the function repel ϭ 1 and the nonbonded repel term C rep ϭ 7.5 ϫ 10 Ϫ7 kJ⅐mol Ϫ1 ⅐nm Ϫ4 , which ignored electrostatic interactions. Interatomic distances, and the 1 angles were constrained by the experimental energy terms k NOE ϭ 2 kJ⅐mol Ϫ1 ⅐nm Ϫ2 and k cdihed ϭ 200 kJ⅐mol Ϫ1 ⅐rad Ϫ2 , with asymptote ϭ 0.1 (NOE restraint term). Stage 2 involved assigning high kinetic energies to the atoms from a Maxwellian distribution equivalent to heating the molecule to a temperature of 1000 K. At this stage, the dynamic trajectory of the molecule was followed for 75 ps in steps of 2 fs. The NOE restraint term was thereafter linearly increased to 1.0 over 25 ps in 2 fs steps. This procedure gradually increased the influence given to the covalent geometry in the calculations from zero to 100%. Stage 4 of the simulation involved cooling the system over 10 ps from 1000 to 300 K, with the repel term set to 0.9 and C rep ϭ 1.5 ϫ 10 Ϫ3 kJ⅐mol Ϫ1 ⅐nm Ϫ4 . Equilibration of the molecule then commenced at 300 K for 5 ps using Lennard-Jones and Coulomb potentials for non-bonded interactions setting repel ϭ 0. Each structure was then refined with 2000 cycles of energy minimization. Molecular graphics were processed using Insight II operating on a Silicon Graphics Indigo 2.

RESULTS
The IC 50 was measured for both native human PTHrP-(1-34)amide and the Ala 15 analog used in the structure analysis with values of 0.7 and 4.5 nM obtained, respectively, from the data in Fig. 1. The PTH-like activities of the peptides were assayed for stimulation of [ 3 H]cAMP formation in intact UMR-106 -01 cells pre-labeled with [ 3 H]adenine (4). The PTHrP analog exhibited only slightly lower relative adenylate cyclase activity than that of native PTHrP based on reduced binding (Fig. 1). Points are means of at least triplicates, and basal activity is shown as the asymptote at the bottom of the figure. Values for the EC 50 of 0.7 and 9.5 nM were obtained for native and mutant peptides, respectively. Activity was reduced only 2-fold in relation to the weaker binding exhibited by the mutant. While reduced, the binding is still a third of the level recorded for human PTH-(1-34)-amide in the same assay (15), indicating that little has changed in the binding site structure.
A TOCSY NMR spectrum of Ala-Val-Ser-Glu-His-Gln-Leu-Leu-His-Asp-Lys-Gly-Lys-Ser-Ala-Gln-Asp-Leu-Arg-Arg-Arg-Phe-Phe-Leu-His-His-Leu-Ile-Ala-Glu-Ile-His-Thr-Ala-amide in 8% F 3 EtOH-d 2 was recorded using a mixing time of 100 ms as described previously (5,16). Among the 34 spin systems in the peptide, the unique residues Ala 1 , Val 2 , Gly 12 , and Thr 33 were assigned sequence specifically from the spectrum. Each of the remaining spin systems was assigned to residue type by the TOCSY and double quantum-filtered correlation spectroscopy spectra. Coherence transfers were detected in the TOCSY as far apart as the seven bonds separating the NH backbone and the ⑀CH 2 protons of Lys residues. Several entry points were available in the sequence-specific assignment analysis that utilized NOESY measurements of the distances between adjacent residues, including the NH, H␣, and H␤ protons of residue (i) coupled to the NH proton of residue (i ϩ 1) (32). Some sequential connectivities are shown in Fig. 2 between H␣ i and NH iϩ1 , which enabled most resonance assignments to be determined. The complete list of chemical shifts is shown in Table   I. The numerous sequential NH resonances in Fig. 3 reveal the likely presence of a substantial quantity of helix. Connectivities from Phe 22 through to Ala 34 at the C terminus are certainly indicative of the presence of a long stretch of continuous helix within this segment. The overall pattern of sequential and non-sequential inter-residue backbone and side chain NOEs is summarized in Fig. 4. The relative intensity of the inter-residue NOE is represented by the line thickness. The shortest inter-proton distances are those with the strongest NOE and are represented by the thickest lines. Degenerate NOEs which are probably present are shown by the asterisks and dotted lines. These were used only for qualitative purposes and remained excluded from structure calculations. Weak or absent sequential H␣ i /NH iϩ1 cross-peaks combined with strong sequential NH i /NH iiϩ1 cross-peaks are indicators of the presence of ␣-helix (43), a conclusion supported by the presence of several medium range NOEs, such as H␣ i /NH iϩ3 and H␣ i /H␤ iϩ3 . On the basis of the patterns of these medium range NOEs, a large region in the sequence appears to contain helix. However, the pattern of long range NOEs indicates that the helix is bent at one point, at least enabling the ends to interact. Some of these interactions are between side chains of His 5 -His 26 , Leu 8 -His 26 , Leu 8 -Ala 29 , His 9 -His 25 , His 9 -Ile 28 , His 9 -Ala 29 , and Ser 14 -His 25 .
All 385 constraints were used in the distance geometry algorithm DIANA to generate 2000 structures from random starting conformations. These calculations yielded 20 structures, which satisfied all distance constraints within 5 pm. Each member of the family of 20 structures possessing the very lowest penalty functions displayed very good covalent geometry. These low penalty structures were refined further using   3. A section of the two-dimensional NOESY spectrum in Fig. 2 showing sequential HN i -HN i؉1 connectivities throughout the molecule.
clearly disordered, and so comparing the 20 best structures with the average structure over the backbone atoms for residues 4 -34 yielded a much smaller r.m.s. value of 0.44 Ϯ 0.11 Å. The segment 4 -33 reveals quite low local average r.m.s. deviation in backbone atoms, whereas for all non-hydrogen atoms, the value was higher with the Arg cluster 19 -21 being particularly unrestrained. In contrast, the residues around 8 and 28 are particularly well constrained as expected from the number of NOE constraints involving them. The / dihedral angle pairings in the segments Glu 4 -His 9 /Asp 10 , Lys 13 -Arg 19 , and Phe 22 -Ala 34 give the appearance of helical conformations.
A view of the backbone of 20 structures superimposed over the segment Glu 4 -Ala 34 is shown in Fig. 5. As was deduced from the dihedral angles, three segments of helix are present in the regions mentioned above. The N-terminal helices are separated by a kink around Gly 12 and ends in a distinct turn at the Arg cluster (residues 19 -21). A single stretch of helix then continues to the C terminus. The whole structure is stabilized by interactions between side chains in the the N-and C-terminal helices. The N terminus appears devoid of any regular structure, but the bulk of the molecule occupies a well defined conformational space under these solution conditions. This is best revealed in Fig. 6 in which the backbone of the average structure is shown with a ribbon and the several interacting side chains are shown represented in ball-and-stick form. The side chains shown include His 5 , Leu 8 , His 9 , Ser 14 , Leu 18 , His 25 , His 26 , Ile 28 , and Ala 29 .

DISCUSSION
A single major conformer of hPTHrP[Ala 15 ]-(1-34)-amide is revealed with no evidence of hinges found at residues 13/14 and 20/21 in different analogs (15)(16)18). The axis of the C-terminal helix in Fig. 6 forms an angle of 80°with respect to the axis of the N-terminal helix. The two helices appear to be stabilized primarily through electrostatic interactions between the side chains of Asp 17 , near the bend at the end of the central helix, and the side chain of His 25 , some 3 Å away, as well as an Asp 17 to His 9 interaction, about 0.5 Å further removed. Other interactions include Glu 30 separated from His 5 by 2.5 Å, and Glu 30 to His 26 along the same helix, nearly 3.5 Å away. The methyl groups of Leu 8 and Ala 29 form an apparently tight interaction that helps stabilize the molecule. Thus a total of eight residues, six of them charged, have inwardly oriented side chains while all others can be considered more surface accessible. The Cterminal helix exhibits a distinct hydrophobic face comprising the side chains of Phe 22 , Phe 23 , Leu 24 , Leu 27 , Ile 28 , and Ile 31 .
The three Arg side chains 19 -21 are clustered on the outer surface of the molecule at the tight bend separating the Cterminal helix from the central helix.
A fundamental purpose of the investigation was to determine whether alterations in the key hydrophobic interaction sites responsible for stabilizing the peptide affect receptor binding properties. Binding of the stabilized conformer to PTH receptors was found to be reduced about 6-fold with respect to native PTHrP-  or only about 3-fold with respect to PTH-(1-34) as a result of substituting Ala for Ile 15 even though the local structure encompassing the mutation site was found to be dramatically altered and all hinges were eliminated. The structure of the putative receptor binding site formed by the Cterminal helical domain remained unaffected. Thus, receptor binding must depend principally on the maintenance of the structure of the C-terminal helix 22-34 and the availability of a binding interface within this segment. The structure of the C-terminal helix is maintained even without the hydrophobic interactions found in several other studies, in particular those between residues 15 and 23 in both PTH (16,21) and PTHrP (13,15,18,20). Other interactions, besides the formation of a hydrophobic core, act to stabilize the C-terminal helix. In particular, the N-terminal helix was found to interact closely with several side chains in the C-terminal helix. Hence, the C terminus in the stabilized conformer is located close to the N terminus. The central helix (Lys 13 to Arg 19 ) and the turn formed at the Arg cluster serves to provide the connection between the N-and C-terminal helices. Interactions between the N-and C-terminal helices in most analogs appear transiently (13). Because of the lack of conformational rigidity, Ala 1 could still interact with the receptor several Å from the main receptor binding site even though Ser 3 and Glu 4 appear adjacent to the receptor binding site.
Several positions in the N-terminal helical segment appear to be far more important than others in maintaining biological activity. Substitutions at positions 7 and 9 show little change in activity (13,20) nor does alteration of the Met 8 sidechain in PTH (44). Modification of the conserved residues Ser 3 and Gln 6 in PTH- , also conserved in PTHrP, have quite different effects on activity and binding (45). While some substitutions are tolerated, others have profound effects, with receptor binding and adenylyl activity almost abolished. In particular, the substitution of Ser 3 by Glu 3 reduces both activity and binding by 3 log. However, simple deletion of the N-terminal six residues barely affects receptor binding while essentially abolishing activity in PTH (9) and PTHrP (10). Consequently, the presence of the N-terminal helical segment is unnecessary for maintaining strong receptor binding but modifications of side chains within it can act to sterically hinder binding. This lends support to the results in this paper which show that the locus of N-terminal binding is adjacent to the essential C-terminal receptor binding domain, and hence, the site on the receptor responsible for agonist activity is adjacent to the body of the receptor binding site for the C-terminal helix of PTHrP-(1-34) and PTH- .
Another concern in the current investigation is the question of what constitutes the essential receptor binding site domain on PTHrP. The use of both PTH and PTHrP truncated peptides in binding studies has indicated that residues 25-31 are essential for receptor binding (4,8). Some residues in the segment 14 -24 also have been thought to contribute to a secondary receptor binding site (8,46). Deletion of segment (1-6) from PTHrP-(1-34)-amide reduces binding capacity to PTH receptors in both bone cells and kidney membranes by 2 log to about 100 nM (47) while even PTHrP(14 -34) still binds at 10 M (48). The structure in Fig. 6 reveals how the absence of the bulk of the N-terminal helix would destabilize the C-terminal helix, thus reducing binding affinity by 2 log while agonist activity is reduced about 7 log (10). Structural comparisons between PTH-(1-34) and all the PTHrP-(1-34) analogs reveal that the majority of segment 22-34 in all the hormones forms an ␣-helix. The N-terminal part of this C-terminal helical segment interacts closely with residue 15 in the putative secondary binding region (16,20,21), and so the sites are immediately adjacent to one another. Since the C-terminal segment can, by itself, fully displace PTH-(1-34) and PTHrP-(1-34) from the receptor, it is more likely that segment 14 -24 simply stabilizes the C-terminal helical structure without being directly involved in receptor binding. Certainly, the 3-fold lower binding exhibited by the the Ala 15 analog compared with PTH indicates the central portion of the molecule plays, at best, a miniscule role in direct receptor binding.
Progressive C-terminal deletions reduce biological activity in PTHrP-(1-34)-amide by progressively reducing the length of the C-terminal helix and thus the stability of the receptor binding site. Removal of segment 30 -34 results in the osteogenic sarcoma cell cAMP and the chicken kidney adenylate cyclase activities declining by 90% (4) and the loss of all mitogenic activity in PTH (11,12). A much larger reduction (100fold) in binding affinity was observed in PTH (7). Additional deletion of residues His 25 to Ala 29 essentially abolishes agonist activity and binding (4). The structure presented here reveals that the site of agonist activity must be adjacent to the primary receptor binding site so that modifications to residues at the N terminus can cause steric blocking effects on the binding of the C-terminal helix to the receptor (45). The residues responsible for signal transduction do not possess any regular structure in the absence of interactions with the receptor. The mid-portion of the molecule accommodates dramatic structural differences while maintaining significant receptor binding capacity and agonist activity provided the N-and C-terminal helices are conserved and able to stabilize each other in the required conformation. This helps provide a structural basis for the design of effective PTHrP and PTH antagonists. We conclude that these must primarily mimic the structure of the conserved hydrophobic face of the C-terminal amphipathic ␣-helix.