Backbone-methylated Analogues of the Principle Receptor Binding Region of Human Parathyroid Hormone

We have used backbone N-methylations of parathyroid hormone (PTH) to study the role of these NH groups in the C-terminal amphiphilic α-helix of PTH (1–31) in binding to and activating the PTH receptor (P1R). The circular dichroism (CD) spectra indicated the structure of the C-terminal α-helix was locally disrupted around the methylation site. The CD spectra differences were explained by assuming a helix disruption for four residues on each side of the site of methylation and taking into account the known dependence of CD on the length of an α-helix. Binding and adenylyl cyclase-stimulating data showed that outside of the α-helix, methylation of residues Asp30 and Val31 had little effect on structure or activities. Within the α-helix, disruption of the structure was associated with increased loss of activity, but for specific residues Val21, Leu24, Arg25, and Leu28 there was a dramatic loss of activities, thus suggesting a more direct role of these NH groups in correct P1R binding and activation. Activity analyses with P1R-delNT, a mutant with its long N-terminal region deleted, gave a different pattern of effects and implicated Ser17, Trp23, and Lys26 as important for its PTH activation. These two groups of residues are located on opposite sides of the helix. These results are compatible with the C-terminal helix binding to both the N-terminal segment and also to the looped-out extracellular region. These data thus provide direct evidence for important roles of the C-terminal domain of PTH in determining high affinity binding and activation of the P1R receptor.

Parathyroid hormone (PTH) 1 is a major regulator of extracellular calcium, functioning mainly through kidney, bone, and intestinal receptors (1,2). The hormone and certain analogues strongly stimulate bone growth in both animals (3) and humans (4,5). PTH and its paracrine equivalent, parathyroid hormone-related peptide (PTHrP), bind to a G-protein-linked seven-transmembrane receptor (P1R) (6) and activate both adenylyl cyclase (AC)/cAMP-dependent protein kinase and phospholipase-C (PLC)-dependent and -independent protein kinase C pathways (7,8). Deletion of even two N-terminal residues leads to a total loss of AC and PLC signaling, as well as the anabolic activities (9). In contrast, the PLC-independent protein kinase C activity requires the C-terminal region of hPTH   (10,11).
The principal receptor-binding domain of PTH-(1-34) (12) and PTH-(1-31)NH 2 is a C terminus ␣-helix bounded by residues 17-29 (13); this domain (14) interacts with the long extracellular N terminus sequence of P1R (8,15). A deletion analogue of the P1R (P1R-delNT) that lacks the N-terminal domain (Ref. 16 and citations, therein) can activate the AC and PLC signaling pathways on stimulation with PTH-(1-34) (17), although with much reduced potency, whereas N-terminal PTH analogues as short as PTH-(1-11) exhibit equivalent potencies on P1R-delNt and the intact P1R (16,17). The presence of PLC-independent protein kinase C activity determinants in the C-terminal region of PTH-  implies that this region also binds to the extracellular loop/transmembrane domain region of the receptor, but only one report directly supports such an interaction (18).
The osteogenic activity of a C terminus-truncated PTH analogue, hPTH-(1-31)NH 2 (19), is similar to that of the well studied PTH-   (20). This analogue has a helix-bend-helix structure in aqueous media at near-physiological pH and ionic strength (14), similar to that found in previous studies of PTH-(1-34) (13). A C-terminal amphiphilic ␣-helix, extending from Ser 17 to Gln 29 , is the major structure within this receptorbinding region, and this structure is thought to bind via its hydrophobic face to the receptor (21)(22)(23). Thus, binding and AC-stimulating activities are very sensitive to replacement of residues on the nonpolar face of this helix, including Leu 24 , Leu 28 , and to a lesser extent Val 31 , whereas residues on the polar face are less sensitive to substitution (22,24).
In general, the specificity of hormone binding to and activation of the receptor is likely determined not only by side-chain interactions between the two molecules but also by specific H-bonding of backbone CO and NH groups (25,26). Thus a complex pattern of H-bonding involving side-chain and backbone groups is expected to be important for the strength and specificity of the overall interaction (27). Backbone methylation has been used to probe the role of backbone amide nitrogens in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. receptor binding and activation (28). These methylations have been used successfully to modulate binding specificity of an agonist for the C5A anaphylatoxin receptors (29), and to create subtype-specific somatostatin receptor agonists and antagonists (30,31). Methylation of a backbone NH not only prevents formation of a H-bond at that point, but it also has an effect on secondary structures, such as ␣-helices or ␤-turns, that may have existed in the region of the modification.
In this study, we have methylated each backbone NH, in turn, of the C-terminal helix-containing region of PTH (1-31), residues Ser 17 to Val 31 , to delineate the role of these groups in receptor binding and activation. We measured the effect of N-methylation on the helix using CD spectroscopy and measured the AC-stimulating and binding activities of each analogue using P1R-expressing cells. Strong effects were seen with several of the methylations. We also measured the AC-stimulating activities of each analogue in COS-7 cells expressing P1R-delNT (17) and showed that the pattern of activity against this receptor is not the same as that observed with the intact P1R. This suggests that the C-terminal region of hPTH binds to the extracellular loops of P1R as well as to the N-terminal extracellular sequence, which is consistent with the membrane-bound protein kinase C-stimulating activity reported for the C-terminal 28 -34 region of PTH-(1-34) (11).

MATERIALS AND METHODS
Peptide Synthesis-All methylated analogues were based on the sequence of hPTH-(1-31)NH 2 (SVSEIQLMHNLGKHLNSMERVEWL-RKKLQDV-NH 2 ). Peptides were synthesized using an Fmoc protocol as described previously (32). Fmoc-␣-methyl-Leu and Fmoc-␣-methyl-Asp were purchased from Chem-Impex and Fmoc-␣-methyl-Val from Novabiochem (La Jolla, CA). The remaining chemicals were purchased from Aldrich. The remaining methylations were done during synthesis by slight modification of the method of Miller and Scanlan (33). Briefly, after addition of the amino acid to be methylated and removal of the Fmoc, synthesis was stopped and the ␣-amino group was then protected with o-nitrobenzenesulfonamide by treatment with the corresponding sulfonyl chloride in CH 2 Cl 2 containing collidine for 3 h. This was followed by deprotonation of the secondary amine with 1,3,4,6,7,8-hexahydro-1-methyl-2H-pyrimido[1,2-a]pyrimidine and methylation with methyl p-nitrobenzenesulfonate for 3 h. The o-nitrobenzenesulfonyl protection was selectively removed by treatment with ␤-mercaptoethanol and 1,8-diazabicyclo(5.4.0)undec-7-ene in N,N-dimethylformamide for 3 h. Completion of methylation was monitored by the Kaiser test for free amine. If methylation was incomplete, as indicated by a positive Kaiser test, the o-nitrobenzenesulfonamide was added again and methylation repeated. For coupling of the next amino acid, HATU was used in place of 2-(1H-benzotriazol-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate. The main problems observed occurred at this step and were especially serious for the coupling of Fmoc-Glu(tBu) after methylation of the Arg 20 residue.  (21), instead using N-␣-Me-Leu at residue 24. Coupling of the next amino acid, Fmoc-Trp(Boc), was done using HATU at 48°C until a negative Kaiser test was obtained. Removal of allyl-and alloc-protecting groups and cyclization were performed after the addition of the final Boc-Ser(tBu).
Circular Dichroism-Spectra were obtained on a JASCO J-600 spectropolarimeter at 20°C. Four spectra were averaged and the data smoothed by the JASCO software. The instrument was calibrated with ammonium (ϩ)-10-camphorsulfonate. Peptide concentrations were calculated from the absorption at 280 nm, using an extinction coefficient of 5700 M Ϫ1 for the single tryptophan.
Data Are Expressed per Peptide Bond-The ellipticities at 222 nm for ␣-helices of different lengths were extrapolated from the data of Chin et al. (34) for short fixed Ala ␣-helices, using a quadratic equation to fit their data as follows: [] 222 ϭ 6448 -5844N p ϩ 252N p 2 , where N p ϭ number of helical peptide bonds. This is compared with evaluation of the mean residue ellipticity of a helix of N p peptide units from the empirical formulation of Yang et al. (35) as follows: , where x is a wavelength-dependent constant that has the value 4 at 222 nm and [ H ] (ϱ) ϭ Ϫ43,000 degrees cm 2 dmol Ϫ1 (36).
In the Chin et al. (34) formulation, the x is effectively dependent on N p and results in ellipticities for helical regions with as little as two residues.
Cell Cultures-HKRK-B7 cell lines were derived from the porcine kidney cell line LLC-PK1 and were stably transformed to express fulllength human P1R to ϳ950,000 receptors/cell (37). P1R-DelNt, a human P1R construct with most (residues 24 -181) of the N-terminal extracellular domain deleted (38) was used to transiently transfect COS-7 cells and was used to assay interactions that occur specifically to the extracellular loop/transmembrane domain region of the receptor.
Adenylyl Cyclase Activity-AC activities were measured using a direct enzyme-linked immunosorbent assay for cAMP (Amersham Biosciences) as described elsewhere (37). The cAMP experiments comparing the effect of methylation at Leu 24 in the linear and lactamcontaining (cyclic) PTH-(1-31) scaffold were performed using radioimmunoassay to quantify cAMP, as described earlier (38). Activities reported are each the average of three separate experiments.
Receptor Binding-Radioligand binding assays were performed as previously described (38). In brief, a 125 I-labeled radioligand (ϳ100,000 cycles/min/well of a 24-well plate of 125 I-[Nle 8,21 ,Tyr 34 ]PTH-(1-34)NH 2 ) was incubated with whole cells expressing the wild-type P1R in the absence or presence of varying concentrations (3 ϫ 10 Ϫ9 -1 ϫ 10 Ϫ5 M) of unlabeled peptide. After incubation for 4 h at 15°C, the binding mixture was removed by aspiration, the cells were rinsed three times with binding buffer, lysed in NaOH, and the entire lysate was counted for ␥-irradiation.
The CD spectra suggested major losses of ␣-helix in some methylated analogues, particularly for those with a methylation near the center of the Ser 17 -Gln 29 ␣-helix. Fig. 2 shows three of the spectra obtained after subtraction of that of hPTH-(1-17)NH 2 , and these clearly show that the apparent ␣-helix, when methylation is near the center of the helix at Leu 24 , is much less than for Ser 17 and Gln 29 at the N-and C termini of the helix. Fig. 3 shows the helix parameter [] 222 for hPTH-(1-31)NH 2 for each methylated residue position, after subtraction of the spectrum for hPTH-(1-17)NH 2 , expressed as the ratio to the [] 222 for the unmethylated form, Ϫ11,880 degrees cm 2 ⅐dmol Ϫ1 . Minimum values are seen near the center of the helix, with up to 80% loss at Trp 23 and progressively smaller effects when the methylation is at either end of the helix. When the methylation is external to the helix, at residues Gln 29 and Val 31 , no loss of helix was observed, but ϳ34% of the helix was lost when methylation was at Asp 30 . Similarly, when the methylation was at the N terminus of the helix, the apparent loss of helix was ϳ15%, based on the [] 222 values.
Backbone methylation leads to disruption of the helix around the immediate point of methylation, although the exact extent is unknown (40). The loss of CD results from a combination of fewer residues in a helical conformation, the reduced mean residue ellipticities of shorter helices, and other conformational changes also known to affect the CD (41,42). We first approximated the helix length effect using a constant dependence factor x of four residues (see "Materials and Methods") ( Fig. 3) (35). This results in no CD signal for helices four residues long and shorter. To estimate the effect, we assumed disruption of the helix for four peptide bonds to the N and C termini on either side of the point of N-methylation. The results were expressed as ratios to the same calculation for the unmethylated 12-residue ␣-helix (Fig. 3). This model resulted in predicted ellipticities of zero for Val 21 to Lys 26 , contrary to the observed presence of the helix after these residues were N-methylated. The more recent model of Chin et al. (34), using a dependence of x on helix length (34), resulted in a profile much more similar to the experimentally observed one. This suggests that, even in the presence of backbone N-methylation, short helical regions are present in the 12-residue-long helix in hPTH(1-31)NH 2 .
Binding of N-Methylated Analogues to HKRK-B (P1R)-The methylations fell into three broad categories with respect to their effect on analogue binding to P1R (Table I). The first category included four methylations that had little effect on the binding, those on Arg 20 , Gln 29 , Asp 30 , and Val 31 . The second category included methylations that resulted in analogues with moderate to weak binding behavior, those on Ser 17 , Met 18 , Trp 23 , Glu 22 , Lys 26 , and Lys 27 . The last category, and perhaps the most interesting, included those that resulted in peptides that had lost almost all capacity to bind the receptor with binding constants Ͼ100,000 nM. These included methylations at Val 21 , Leu 24 , Arg 25 , and Leu 28 . The severe reductions in apparent binding affinity seen with these methylations suggest that the NH groups of these amino acids may be involved in critical H-bonds with the receptor.
Adenylyl Cyclase Activities-The adenylyl cyclase activities of the analogues (Table I) The high signaling activity of the N-methylated Arg 20 analogue is particularly noteworthy, because the side chain of this arginine is critical for PTH bioactivity, tolerating very little modification (46). The essential loss of activities of analogues with methylations of Val 21 , Leu 24 , Arg 25 , Lys 27 , and Leu 28 suggests that the ␣-amino groups of these residues are involved in critical contact with the receptor. The lack of effect of methylation of Val 31 , Asp 30 , and Gln 29 is not surprising given that there is little difference between the AC-stimulating activities of hPTH-(1-28)NH 2 and hPTH-(1-31)NH 2 (47).
Adenylyl Cyclase Activities with P1R-delNT-This truncated receptor construct (17) lacks nearly all of the long, N-terminal sequence of the P1R that has been associated with binding to the C-terminal region of PTH (1-34) (48). The assumption that the C-terminal helix of hPTH-(1-31) binds only to the extracellular N-terminal domain of P1R implies that the effect of individual N-methylation of the residues in this region would have no effect on interaction with delNT-P1R. However, the results of Fig. 4 clearly show that some of the methylations diminished cAMP signaling responses induced by the analogues, tested at a single high concentration (1 M) on this truncated receptor. Furthermore, although the absolute AC responses measured on this receptor cannot be directly compared with those determined on the intact receptor (ED 50 values were not determined for the truncated receptor, and different cAMP assay formats were used in the two experiments, radioimmunoassay versus enzyme-linked immunosorbent assay), clear differences in the patterns of effects that the methylations had on activity at the two receptors could nevertheless be discerned. Thus, three methylated analogues, Ser 17 , Trp 23 , and Lys 26 , show particularly low activity with P1R-delNT but have moderate to high activities with the intact P1R. This may indicate that the NH groups of these residues form H-bonds with the extracellular loops or transmembrane domains of the receptor (P1R and P1R-delNT). None of these residues is asso-  (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17) . deg, degrees. ciated with critical H-bond formation when the intact P1R is used, because their relatively weak effect is masked by the much stronger interactions that occur in the presence of the N-terminal domain of the receptor. Thus, although the C terminus of PTH-(1-31) clearly interacts with the extracellular N-terminal domain of the receptor, it must also be interacting with one or more of the extracellular loops or transmembrane domains of the receptor.
[N-Me-Leu 24 ]c(Glu 22 -Lys 26 )hPTH-(1-31)NH 2 -This lactam analogue was constructed to test whether loss of ␣-helix in the C-terminal domain was the major factor leading to the severe loss of binding and AC-stimulating activity. The lactam bridge between the side chains of Glu 22 and Lys 26 in this analogue has been shown in other PTH-(1-31) peptides to stabilize local helicity and to increase binding affinity and potency by severalfold (21,32). The CD spectrum of the analogue, compared with that of N-Me-Leu 24 -PTH- , showed that, although there was little increase of [] 222 , the spectrum clearly had a more helical structure but not with the same intense helical minima of 209 and 222 nm and the classical helical shape seen with the unmethylated analogue c(Glu 22 -Lys 26 )hPTH-(1-31)NH 2 (Fig. 5). It suggested a more classical helix but with a disruption at the point of methylation.
Comparison of the activities of the analogues in HKRK-B7 cells indicated that the lactam greatly increased the AC-stimulating activity of the N-Me-Leu 24 analogue. This increase in signaling potency was accompanied by at least some increase in binding affinity, in that inhibition of radioligand binding could be detected (Table II). The data on these analogues thus suggest that a large portion of the loss of binding and AC-stimulating activity caused by the Leu 24 methylation is directly related to the loss of ␣-helical structure.

DISCUSSION
The backbone methylations described here led to two main conclusions. First, the CD data indicated a general loss of structure that could be explained in large measure by a simple model assuming localized disruption near the N-methylated residue. Second, although there was a general loss of activity throughout the region of the helix, there were specific methylations that resulted in a dramatic loss of binding and AC activity, and the patterns of losses differed depending on whether the intact receptor or delNT-P1R was used for the measurements.
We know that there is disruption of helical structure at the site of a specific backbone methylation and that this disruption affects neighboring peptide bonds (40,49). These effects were incorporated into our analysis of the CD data. There is a paucity of data concerning the extent of helix disruption caused by a backbone NH methylation, but the effects on the hormonereceptor complex caused by some of our methylations appeared to be more severe than what would result from just the loss of one H-bond, and numerous other stabilizing interactions were expected to be lost. This is important, because the loss of only one H-bond would contribute, from the known free energy  19 18,000 (1000) 80.5 (6.5) Arg 20 280 (90) 24.8 (0.2) Val 21 105,000 (39,000) Ͼ500 Glu 22 41,000 (4,000) 150.3 (10.1) Trp 23 12,000 (500) 124.0 (2.0) Leu 24 Ͼ100,000 Ͼ500 Arg 25 Ͼ100,000 Ͼ500 Lys 26 56,000 (5000) 49 (5.9) Lys 27 52,000 (2000) Ͼ500 Leu 28 Ͼ100,000 Ͼ500 Gln 29 620 (  change, an approximate 10-fold loss of K a (50), far less than what we observed for several of the methylations. Our CD analysis was also limited to the effects of chain length on the CD signal and could not take into account the known effect of distortion of the ␣-helix from the classical , angles of Ϫ60°, Ϫ40° (42), and the relaxation of the trans conformation of the peptide bond at the point of methylation to a cis-trans equilib-rium (28). We also cannot be certain that PTH binds to the receptor in a conformation very similar to that which occurs in solution, in particular for the C-terminal domain studied here. However, a transfer nuclear Overhauser effect study of pituitary adenylyl cyclase activating polypeptide (PACAP) provides evidence that this hormone binds to the receptor with a conformation very much like the average solution one in its Cterminal domain (51). The receptors for PACAP and PTH are related, and the hormones share similar structural features, thus suggesting that PTH would likely behave similarly.
The model in Fig. 6 shows the location of effects of methylation on binding and activity for P1R and delNT-P1R. Previous data implicates the C-terminal amphiphilic ␣-helical domain of PTH-  in binding to the long, extracellular N terminus of P1R. Deletion of the P1R extracellular N-terminal sequence drastically limits PTH binding. Substitution of a photoactivatable group for Trp 23 identifies a close proximity to the Nterminal P1R segment, residues 23-40 (52). The dramatic (Ͼ150-fold) effects on P1R-binding affinity seen with NH methylations near Trp 23 at Leu 24 and Arg 25 provide evidence to support an important role for this interaction in the overall stability of the complex.
Relatively little evidence has implicated the same region of PTH in binding to the extracellular looped-out region of the receptor. Other photochemical cross-linking data implicate Lys 27 of the hydrophobic face being proximal to Leu 261 of EC-1 (18) and Glu 19 as proximal to the extracellular end of transmembrane-2 (53). The data presented here with P1R-delNT, together with these chemical cross-linking studies (18,52), strongly suggest that the C terminus of PTH-(1-34) does indeed interact with the extracellular looped-out region of the receptor. It is striking that the NH substitutions that most strongly affect interaction with P1R-delNT occur on the opposite face of the helix from those that most strongly affect interaction with the intact P1R (Fig. 6). This suggests that the polar face of the helix binds to the extracellular loops, whereas the nonpolar face binds to the N-terminal domain. It is also noteworthy that, of the 13 possible backbone H-bond donors in the 17-29 helical segment, seven are implicated in binding to some part of the receptor. A recent Ala scan study has identified Arg 20 , Val 21 , Trp 23 , Leu 24 , and Arg 25 as important residues in PTH-P1R recognition. 2 Collectively, these results and those of other previous reports (22,24,46) identify a hot spot from Arg 20 to Leu 28 that is critical to PTH-P1R complex formation.
The distance between the NH of Ser 17 and the NH of Lys 26 , the bounding backbone NHs implicated with delNT-P1R, is ϳ16 Å. Although the structure of the P1R is largely unknown, residue pairs from different extracellular loops, using the Protein Data Bank crystal structure coordinates of rhodopsin, could be bridged by this distance. Thus, a possible scenario is that the C-terminal helix bridges two loops, thus facilitating changes in the relative orientations of the transmembrane segments that result in G-protein binding. On the hydrophobic face, the backbone NHs of Val 21 and Leu 28 are slightly closer together at 11 Å. We should emphasize that side-chain substitution and methylation at the site of a particular residue need not give similar effects on binding and activity. For example, there is a high degree of stringency for the Arg 20 side chain (46), but methylation of the Arg 20 ␣-amino group has relatively little effect on binding and signaling activity in comparison to other residues, such as Glu 22 , which can be substituted more freely.
Combined with an earlier implication of extracellular loop binding to PTH (18), the data in this paper suggests that the principal binding domain C terminus of PTH binds in a cooperative manner to both the N terminus region of the receptor and one or more of the extracellular loops. Binding of PTH to either P1R-delNT (54) or the extracellular N terminus sequence alone of P1R (15) is very weak (ϳ M). The experiments reported here thus help to delineate the nature of the overall binding and activation process.  (38). e ED 50 values were calculated by fit of data by nonlinear regression (38).
FIG. 6. Model views of hPTH-(1-31)NH 2 Ser 17 -Gln 29 helix with critical residues. The helix is viewed along the long axis from the N terminus (A) and from the side (B). Backbone amide hydrogens that, when methylated, result in greatly diminished activities with P1R (those of Val 21 , Leu 24 , Arg 25 , and Leu 28 ) are shown in green, and the corresponding ones that diminish activity with P1R-delNT (those of Ser 17 , Trp 23 , and Lys 26 ) are highlighted in orange.