Eur. J. Biochem. 209,765 - 771 (1 992)

(0FERS 1992

Solution conformation of human neuropeptide Y by 'H nuclear magnetic resonance and restrained molecular dynamics Herve DARBON ', Jean-Marie BERNASSAU2,Colettc DELhUZE *. Jacqucs CHENU ', Alain ROUSSEL' and Christian CAMBILLAU I

LCCMR, CNRS URA 3296, Facultt de mtdccine-Nord. Marseille, France SanoG-Recherche, Montpcllicr, France Sanofi- Recherche, Toulouse, France

(Received April 30/July 13, 1992)

-

EJB 92 0605

The solution structure of human neuropeptide Y has been solved by conventional two-dimensional NMR techniques followed by distance-geometry and molecular-dynamics methods. Thc conformation obtained is composed of two short contiguous ol-helices comprising residues 15 - 26 and 28 35, linked by a hinge inducing a 100' angle. l h e first helix (15-26) is connected to a polyproline stretch (residues 1- 10) by a tight hairpin (residues 11 - 14). The helices and thc polyprolinc stretch are packed together by hydrophobic interactions. This structure is related to that of the homologous avian pancreatic polypeptide and bovine pancreatic polypeptide. The C- and N-terminii, known to be involved in the biological activity for respectively the receptor binding and activation, are close together in space. The side chains of residues Arg33, Arg35 and Tyr36 on the one hand, and Tyrl and Pro2 on the other, form a continuous solvent-exposed surface of 4.9 nm2 which is supposed to interact with the receptor for neuropeptide Y -

Neuropeptide Y (NPY) is a 36-amino-acid neurotransmitter peptide (Fig. I), first isolated from porcine brain [l], but also found throughout the central and peripheral nervous system of many mammalian species, including man [2, 31. NPY has been identified within specific peptidergic neurons of central and autonomic nervous system. In the central nervous system, NPY stimulates food intake, produces cardiovascular depression, and inhibits the release of luteinizing hormone. In the periphery, it is a potent vasoconstrictor and a presynaptic inhibitor of neurotransmission (for a review, see [4]). NPY belongs to a family of homologous peptides, which includes peptide YY and the pancreatic polypeptides [3]. A member of this family, the avian pancreatic polypeptide (APP) has been crystallized and its structure determined by X-ray diffraction [5]. More recently, the structure of bovine pancreatic polypeptidc (BPP) has been determined by two-dimensional NMK in water [6]. Both APP and BPP have a verywell-defined C-terminal a-helix, involving residues 2 5 - 32. This helix and the C-terminal region are joined by a turn. The rather ordered conformation of residues 4- 8 is maintained by hydrophobic interaction between the helix and the Nterminal region. Both APP and BPP share a similar conformation. Models of NPY based on the crystal structure of the APP [5, 71 have been proposed [8,9]. Previous NMR studies as well as CD analysis [lo] have demonstrated that NPY, despite its Correspondence to H. Darbon, LCCMB, CNRS URA 1296, FacultC dc mkdecine-Nord, Bd. Pierre Draniard. F-13326. Marseille,

Cedex 15, France F a x : +3391698913. Ahhreviaiion,~.APP, avian pancreatic polypeptide; BPP, bovin pancreatic polypeptide; NPY, neuropeptide Y; NOESY, two-dimcnsional nuclear-Overhauser-cffcct spcctrum.

small sizc and the absence of disulfide bridge, adopted a wellordered three-dimensional structure in water. However, the structure of the NPY dimer in solution has been described [ll]. This structure is composed of a helix consisting of residues 11 36 and an unstructured mobile N-terminal segment. This structure disagrees with those of APP and BPP. In this paper, we present the three-dimensional structure of huinan NPY in water as a monomer, obtained by distance geometry, restrained energy minimization and restrained simulated annealing. This structure is close to those of APP and BPP, but is very different from that described by Cowley et al. [ll]. ~

MATERIALS AND METHODS

Peptide synthesis Human NPY was synthesized by standard solid-phase synthesis techniques using an automated peptide synthesizer with t-butoxycarboxylbenzyl chemistry [12] and p-methylben7hydrylamine-resin (0.42 mmol/g resin). After deprotection of the peptide and removal from the resin (low/high HF method), the peptide was purified by cation-exchange chromatography and preparative C1 reverse-phase HPLC, then characterized by analytical reverse-phasc HPLC and amino acid analysis. Sample preparation

Synthetic NPY was dissolved to a final concentration of 4 mM in either 90% H20/10% DzO, or 100% D 2 0 . The pH was adjusted to a glass-electrode reading of 3.2 (uncorrected for isotope effects).

766 KESULTS The sequential assignment (Fig. 2B and C)obtained for 3PP human NPY was essentially identical to that previously pro1 11 21 31 36 posed by Saudek and Pelton [lo] for porcine NPY (which has Fig. 1. Amino acid sequences of human NPY, APP and BPP. Leu at position 17, instead of Met). Only two out of the four proline residues, Pro5 and Pros, were found in both trans and cis conformations. The existence of two conformations was revealed by multiple resonance peaks for protons belonging NMR spectroscopy to the preceding residue. Consequently, Lys4 and Asn7 have 'H-NMR spectra were routinely recorded at 310 K on two chemical shifts for their Ca and amide protons (Fig. 3). a Bruker AMXSOO spectrometer. Two sets of spectra were The ratio of trans/cis conformation was estimated from the recorded either in D,O or in 90% H 2 0 , 10% D20, each relative intensity of their two Ha/HP cross-peaks, the Ha comprising double-quantum-filtered two-dimensional corre- chemical shift corresponding to the trans conformation being lation spectra, homonuclear Hartmann-Hahn spectra with a detected by the existence of HYAsn7/H6 Pro8 and Ha Lys4/ mixing time of 100 ms, and two-dimensional NOE spectra H6 Pros. This tmnslcis ratio was 0.50 and 0.80 for Pro8 and (NOESY) with mixing times of 200 ms and 350 ins. Solvent Pros, respectively. The resonances of the remaining residues suppression was achieved by irradiation of the H 2 0 peak indicated that they were in a single conformational state. The during the relaxation delay and during the mixing time in presence of sequential NOESY cross-peaks Ha XaajHh' Pro, together with the absence of sequential NOESY cross-peaks NOESY experiments. After the two-dimensional data matrices were zero-filled Ha Xaa/HaPro clearly indicated a trans Conformation for in Fl and multiplied by a shifted sine-bell in both dimensions, Pro2 and Pro13 (Fig. 3). The homogeneity of the resonances two-dimensional spectra were Fourier-transformed and base- belonging to all the residues except Asn7 and Lys4 strongly line corrected with standard Bruker software running on a suggests that the overall conformation of NPY is independent of proline conformation. For this reason, only all-tran.s conX32 station. formations were selected for structural calculations. Fig. 2 Resonance assignments for the 'H-NMR spectrum were show the fingerprint and the amide regions of the H 2 0 200carried out using the general strategy described by Wiithrich [13] with the use of the interactive part of the EASY program ms NOESY which was used for NOE collection. Several CHNHi,i13 (Fig. 2A) and sequential NH-NH (Fig. 2B) cross~41. peaks have been detected which unambiguously indicate the presence of a helical structure. NPY APP

YPSKPDNPGE CAPAEDMARY YSALRRYINL I T R Q X Y N H ~ SPSQPTYPGC DAPYZPLIRF YDULQQYLNV VTRHRYPH? APLEPEYPGD NATPCQYAQY AAELRRYINM LTRPRYwz

Structure calculations

NOE intensities, used as input for the distance geometry calculations, were determined from NOE spectra recorded in 90% H 2 0 , 30% D,O with a mixing time of 200 ms for all signals, except that involving C a H , very close or under the water signal. For these last signals, which may be bleached by water suppression, a NOE spectrum (200 ins mixing time) recorded in D 2 0was used. The peaks intensities were obtained by the peak-integration routine of the EASY program. The intraresidual and sequential NOE were partitioned into three categories (strong, medium and weak) that were converted into distance restraints ( < 0.25, < 0.3 nm and 0.05 nm

Violation average

maximum

nm

kJ/mol I 11 111

IV V

VI VII VIIl IX X

-2010 - 2900 - 3560 - 3650 -3850 - 3720 - 3800 -3610 - 3790 - 3700

RMS distance

142 163 151 147 142 151 151 142 147 142

798 831 924 7938 7434 8442 8400 8694 8106 8064

802 743 697 806 743 793 827 789 701 785

- 667

- 5323

- 575

5334 - 5061 - 5048 -5153 -5153 -5157 -5010 -4998 -5019

- 609 - 583

-617 - 655 - 693

646 -697 - 625 -

-

0 0 1

0 0 1 I 0 1 0

3.8 3.9 4.9 3.7 3.8 3.8 3.9 3.9 3.8 3.8

47 44 54 48 42 71 50 30 60 47

0 0.11 0.11 0.128 0.14 0.142 0.144

0.145 0.154 0.17

769 Table 3. Calculated energies and restraint violations during the refinement process of structure I. VdW, van dcr Waal's forces. Step I, DISMAN; step TI, restraincd energy minimization; step 111, simulated annealing; step IV, restrained energy minimization. Step

Total

Bond

Angle

Dihedral

VdW

No. of violations >0.05 nm

Electrostatic

kJ/mol ~.

I

IT I11 IV

- 1096 - 3246 -1870 -4010

210 139 181 143

2293 802 1151 802

-.

1562 722 1344 802

-319 -579 -508 -667

Violation average

maximum

pm ~~

-2650 -4351 -44099 -5325

4 1 0 0

5.1 4.2 3.8 3.0

A

72 51 47 31

Sequence position 1

B

5

lo

15

20

2s

30

3s

I

I

I

I

I

w

o 0

-

7'

Fig. 4. Stereo view of the best molecular NPY structures (only Cor are displayed) with lowest restraint energies (energy penalty due to the residual NMR restraint violations) superimposed for best fit on backbone heavy atoms, after distance geometry and minimization (A) and simulated annealing and minimization (B).

result is consistent with the findings of Saudek and Pelton [lo], who, on the basis of CD data, concluded that the peripheral elements of the a-helix are less constrained than the central part. Residues 15 - 35 form two a-helices connected by a kink at position 27. The angle between the two helices varies during the refinement process (Fig. 4). Two main differences arise during simulated annealing: the kink angle varies from 120 before, to 100O after, simulated annealing refinement, and the geometry of sequence 11 - 14, folded in a tight hairpin, evolves. Residues 11 - 14 were previously proposed by Saudek and Pelton [lo], based on the detection of short-range and medium-range NOE, to be part of the helix. It is, however, difficult to discriminate between a tight turn and an helical @

Fig. 5. Diagonal plot for NPY of the NOE observed in NOE spectra. ( W ) Position x , y indicates that one or more NOE were observed between backbone protons of the two rcsidues at sequence positions x and y. ('E)NOE between a backbone proton of one residue and a side-chain proton of an other residues. (n)NOE between side-chain protons of the two residues. Intraresidue NOE are not displayed. When two residues are connected by more than one NOE, only the one involving the largest number of backbone protons is shown.

peptide segment only from sequential and medium-range NOE as distance restraints are similar in both cases [13]. A hairpin conformation satisfies all the long-range NOE for this sequence. The helical stretch is thus shorter by four residues than that proposed by these authors. The hairpin connects the helix to the proline-rich N-terminal peptide whose conformation is poorly defined, due to the small number of NMR restraints (Fig. 6). The scarcity of these NOE within polyproline may be caused by rapid motion or local conformational flexibility. However, its spatial location with regard to the a-helices in ensured by inter-side-chain long-range NOE, with no corresponding interbackbone NOE being detected (Fig. 5). These two regions are closely packed through interaction of a central cluster of hydrophobic side chains including Pro5, Tyr20, Ala23, Leu24, Tyr27, Leu30 and Leu31 (Fig. 7).

770

35

u

c

30

8 z

25

20

15

10

Fig. 8. View of the C- and N-terminii parts of NPY showing the surfaces possibly involved in the interaction of NPY with its receptor (in blue) and responsible for bioactivity (in yellow).

5

I

5

ia

15

20

25

30

35

ScqlJCIlCC

Fig. 6. Distribution of intraresidue and sequential distance restraints (filled colnmns) and longer-range distance restraints (open columns).

Fig. 7. Longitudinal slice of NPY showing the internal hydrophobic cluster which stabilizes the helical structure of the peptide in register with the polyproline stretch.

DISCUSSION Before assigning a structure to experimental data, one should consider molecular aggregation. The line width mcas u r d in a one-dimensional exveriment. recorded with a concentrated or with a diluted simple (100-fold dilution), was

approximately 4 Hz, which is consistent with the absence of aggregation. Furthermore, no NOE could be attributed to intermolecular contacts. Structural calculations, taking the long-range NOE as intermolecular instead of intramolecular restraints, failed due to incompatibility with the short-range and medium-range constraints. The side chains, which are solvent-accessible in the structure determined, do not show any long-range NOE, which would have been a sign of dimerization. Although the helices are obviously amphipathic, in accordance with previous reports [lo], they are oriented in the structure determined in a way that leaves the hydrophilic side exposed to the solvent, whereas the hydrophobic side interacts with the N-terminal part of the molecule. Consequently, the hydrophobic side of the helix can not participate in dimer formation, although a recent paper described the solution structure of an NPY dimer [Ill. Even if we do not have any evidence for a dimerization of NPY, it is possible that such an association still exists, but in the same manner as does APP, i.e. via an interaction between some unburied hydrophobic residues such as Ile28 or Ile31 and, to a lesser extent, Leu24. The quality of NMR-derived structures should be assessed on two grounds: the degree to which the structures fits the experimental data and the variability of the calculated structures. The former can be assessed from consideration of the violations of the NOE-derived distance restraints. As shown in Table 3, the average violation diminishes from 5.1 pm for the distance-geometry structure to 3.0 pm after molecular dynamics. This corresponds to a fair improvement in the quality of the structure. The variability of structures can be measured by the average root-mean-square distance for the atomic positions of overlaid structures. In the case of the final structures of NPY, this value for the backbone atoms is in the range 0.1 -0.17 nm, suggesting converging structures (Tablc 2; however, the geometry of the proline-rich N-terminal is poorly defined due to the limited amount of available NMR data in that region). The NMR-derived restraints consisted exclusively of NOE. Coupling constants, which may be useful in assessing secondarv structure. could not be accuratelv measured in oneydimensioial spectra, due to overlap, nor in fwo-dimen-

77 1 sional spectra, due to low resolution. Furthermore, unambiguous assignment of protons engaged in hydrogen bonds was difficult to achieve: even if some amide protons were not exchanged for deuterium over a period of hours in accordance with Saudek’s investigations, they were detected in one-dimensional experiments, where signal overlap prevented assignment. Recording a two-dimensional NOESY with an accumulation time of less than 10 h gives a signal/noise ratio too low to be realistically assigned. NPY belongs to a family of homologous peptides, which includes peptide YY and the pancreatic polypeptides [3] (a member of this family, APP has been crystalized and its structure determined by X-ray diffraction). A solution structure of BPP has been recently determined by two-dimensional NMR [6].APP, BPP and our NPY structures (BPP and NPY share 55% sequence similarity) have a similar overall conformation. In BPP, Tyr20, Leu24 and Tyr27 make contacts with the Nterminal region, as in our NPY structure. The four residues at both the N-terminus and C-terminus are more disordered in both structures, as well as the turn (9 14 in BPP and 11 14 in our NPY structure). The BPP structure also shows a curved helix, although to a lesser extent than does our NPY structure. The kink in the helix of NPY may actually be a flexible hinge, since the angle between the two helices changes during molecular-dynamics refinement. Each step of the calculation gave converging structures consistent with the set of NMR restraints, as demonstrated by the low residual violations. The calculated energies of the sets of structures, obtained either by distance geometry or after simulated annealing rcfinement, are within the same order of magnitude, and the Ramachandran plots are of equally good quality. The two sets of structures thus represent equally probable solutions. Kinked heliccs havc been observed in a highly refined X-ray structure of the lipase from the fungus Mucor meihi [IS] (120” angle between two contiguous helices) or in myoglobin and flavodoxin (100” angle). The solution structure of paradaxin P-2, a 33-amino-acid peptide, also shows a broken helix [19]. Structure/activity-relationship studies have demonstrated the importance of the C-terminal tetrapeptide for binding of NPY to its receptor [9], as well as the N-terminus for biological activity [20]. The ainphipathic character of the r-helix hclps to stabilizc the relative orientation of the N- and C-termini. The helices thus do not interact with membranes as amphipathic helices generally do. This hypothesis is confirmed by our structure in which most of the long-range NOE involve the dense internal cluster of hydrophobic side chains. Fig. 8 shows the Connolly surface of residues 33-36 and 1 and 2, brought together by the hydrophobic interactions. These residues form a rather flat solvent-accessible surface of 4.9 nm2, by which the molecule may interact with its membrane receptor in accordance with the model proposed by McLean [20]. The organization of this surfaceis also in accordance with the findings of Allen et al. [8], proposing that both the C- and N-terminii regions of NPY are required for biological activity, and that receptor binding and activation may reside in the N- and C-terminii, respectively. ’ Tyrl and Tyr36 are distant from each other, but fully cxposcd to the solvcnt. A recent report has shown that desY1-NPY has no biological activity [21], further suggesting that Tyrl plays a direct role in interaction with the receptor, rather than the role of a conformation stabilizer. The importance of a similar solvent exposure of aromatic residues for -

-

pharmacological activity has been demonstrated in the case of scorpion toxins [22, 231. These small basic proteins indeed interact with their receptors through a solvent-exposed hydrophobic surface, rich in aromatic residues and surrounded by hydrophilic regions. In NPY, such a surface made by a similar association of hydrophilic (Arg33, Arg35) and hydrophobic (Tyrl, Pro2, Tyr36) residues is involved in the biological activity of the peptide. On the other hand, some structure/activity studies have demonstrated the full activity of a cyclic derivative of NPY having amino acids 1 - 4 directly linked to amino acids 25 36 through aminohexanoic acid [9]. Such a linkage is perfectly allowed by the NMR structure described, which provides a structural basis for the understanding of the biological activity of this ubiquitous neuropeptidc. We thank M. L. Ceccato for synthesizing human NPY.

REFERENCES 1. Tatemoto, K., Carlquist, M. &Mutt, V. (1982) Nature 296,659-

660. 2. Adrian, T. B., Allen, J. M., Bloom, S. R., Ghatei. M. A., Rossor. M. N., Roberts, G. W., Crow. T. J., Tatemoto. K. & Polak, J. M. (1983) Nature 306, 584- 586. 3. Allen. J. M. & Bloom, S. R. (1986) Neurochem. Int. 8. 1 - 8. 4. Lehmann, J. (1990) Drug Dcv. Res. 19, 329-351. 5. Blundell, T. J., Pitts, J. E., Tickle, I. J., Wood, S . 1’. & Wu, C . W. (1981) Prac. Nut1 Acad. Sci. USA 78,4175-4119. 6. Li, X., Sutcliffe, M. J., Schwartz. T. W. & Dodson, C . M. (1992) Biochemistry 31, 1245- 1253. 7. Glover, I., Hanceff, I.. Pitts, J., Wood, S., Moss, D., Tickle, I. & Blundell, T. J. (1983) Biopolymers 22, 293-304. 8. Allen, J. M., Nowotny, J., Martin, J. & Hcinrich, G. (1987) Proc.. Natl Acud. Sci. U S A 84, 2432-2536. 9. Beck-Sickingcr, A,, Gaida, W., Schnorrenberg, G., Lang, R. & Jung, G. (1990) Int. J . Pept. Protein R e s 3 6 , 522-530. 10. Saudek. V. & Pelton, J. T. (1990) Biochemistry 29,4509-4515. 11. Cowley. D. J.. Hoflack, J. M.. Pclton, J. T. & Saudek. V. (1992) Eur. .I. Biochem. 205, 1099- 1106. 12. Mcrrifield, R. R. (1963) J . A m . Chem. Soc. 85.2149-2154. 13. Wiithrich, K. (1986) M M R ofprolein undnuc/eic ucids John Wiley and Sons, New York. 14. Eccles, C., Giintert, P., Billeter. M. & Wuthrich; K. (1991) J . Biomol. N M R I , 1 1 1 - 130. 15. Braun, W. &Go, N. (1985) J . Mol. Riol. 186, 611 -626. 16. Briinger, A. T. (1990) X-PLOR v21 Manual. Yale Univ. Press, New Haven, CT. 17. Roussel. A. & Cambillau, C. (1989) in Silicon graphics geometry partner director); (Full 1989) pp. 77-78, Silicon Graphics, Mountain View, CA. 18. Brady, L., Brzozowski, M., Derewenda, Z. S., Uodson, E., Dodson, G., Tolley, S.. Turkcnburg, J. P., Christansen, L., Huge-Jensen, B., Norskov, L. & Mcnge, U. (1990) Nature 343, 167 - 770. 19. Zagorski, M. G., Norman, D. G., Barrow, C . J.; Iwashita, T., Tachibana, K. & Patcl. D. J. (1991) Riochemi.rtr.y 30, 80098017. 20. McI,ean, L., Buck, S . H. & Krstenansky, J. L. (1990) Biochembtrj 29,2016-2022. 21. Benchenkroun, M. T., Fournier, A,, St-Pierre, S. & Cadieux, A. (2991) Proceedings of the 12th American peptide symposium, p. 75, Cambridge, MA. 22. Fontecilla-Camps, J . C., Habersetzcr-Rochat, C. & Rochat, H. (1988) Pror.. Nut1 Acud. Sci. U S A 95, 7443-7441. 23. Darbon, H., Webcr, C. & Braun, W. (1991) Biochemistry 30, 1836- 1844.