Conformational Studies of Cyclic Peptide Structures in Solution from 'H-Nmr Data by Distance Geometry Calculation and Restrained Energy Minimization HANS SENN,' HANS-RUDOLF LOOSLI, ' MICHEL SANNER, ' and WERNER BRAUN*

' Sandoz Ltd., Preclinical Research, CH-4002 Basle, 'lnstitut f u r Molekularbiologie, Eidgenijssische Technische Hochschule, CH-8093 Zurich, Switzerland

SYNOPSIS

T h e three-dimensional structure of a cyclic bouvardin anaLgue, cyclo ( -Pro-MeTyr-AlaMeTyr-MeTyr-D-Ala-) has been determined by distance geometry calculation and restrained energy minimization from nmr data. The preparation ofthe input for the distance geometry calculations, the modification of the amino acid library, and the analysis of the structures were done with t h e aid of a recently developed software package, GEOM. A great variety of different initial structures were explored to check the uniqueness of the determined solution structure. Calculations with 500 different initial structures and two different strategies led to a uniquely determined backbone conformation with a root mean square deviations value of 0.4 A. The backbone structure consists of two @-turns,a 13-11turn at Pro'-MeTyr', and a @-VIturn at MeTyr4-MeTyr5.The efficiency of the two calculation strategies were compared in order to propose a n optimal means for performing distance geometry calculations with cyclic structures.

INTRODUCTION In typical drug design situations with polypeptides, the occurrence of cyclic structures and of chemically modified amino acid residues requires special care in performing distance geometry calculations from n m r data. A flexible software package, GEOM, specifically developed for support in the determination of three-dimensional structures of a great variety of covalent structures, has been tested and described.' We have used this tool t o determine t h e three-dimensional structure of the Bouvardin analogue 2 12, cyclo ( -Pro-MeTyr-Ala-MeTyr-MeTyr-D-Ala), where MeTyr designates the substitution of an amide proton by a methyl group. I n the structure calculation of this cyclic peptide, we did not assume any hypothetical initial model or available x-ray structure. Distance geometry calculations were per-

(c,

1990 J o h n Wiley & Sons, Inc.

ccc 0006-3525/90/10-111387-14

$04.00

Biopolymers, Vol. 29, 1387-1400 (1990)

formed with the program DISMAN,2i3where the residue library had been generated by GEOM. Previous studies with cyclic peptides 4-7 did not include a systematic variation of the initial structure, t o show t h a t t h e final result is not biased by the initial structure. Exhaustive methods for grid search can only be applied in practice to small rings because of the exponentially increasing number of conformations t h a t have t o be considered.8 By carefully checking the van der Waals radii and by using experimental data early in the tree search, the number of conformations to be looked a t can be dramatically de~ r e a s e d But . ~ it has still not been shown if the reduction of the number of conformations is sufficient for practical application with typical nmr data sets. T h e method we have applied is also not exhaustive, but it definitely explores a much greater variety of initial structures than has been previously possible.s T h e procedure is similar in spirit t o t h e approach followed by others, lo where sequential and shortrange information are carefully examined so as t o discard forbidden conformations. but it has the ad1387

1388

SENN ET AL.

vantage that most of the selections are made automatically by the DISMAN program. In our approach, the initial structures were chosen as random structures, which is the common approach for the determination of protein structures from nmr data.3.1'.'2 We employed two different strategies. In the first strategy, we a t first included the constraints for the cyclization of the peptide, calculated cyclic structures, and then included the experimental nmr data. In a second strategy, we used a combined data set of cyclic and experimental constraints. The performance of the distance geometry calculations for both strategies is compared. The best structures from both strategies were selected and refined by energy minimization. This approach differs from previous studies in that a quite systematic search for possible three-dimensional structures was carried out with well-defined objective criteria. The variation of the final backbone conformations with a root mean square deviation ( R M S D ) value of 0.4 A showed that the experimentally determined nmr data were sufficiently complete t o uniquely determine the backbone of the cyclic peptide 212.

MATERIAL A N D NMR SPECTROSCOPY The Bouvardin analogue 212, cyclo ( -Pro'-MeTyr2Ala3-MeTyr4-MeTyr5-~-Ala6) , was dissolved in freshly opened deuterated dimethylsulfoxide (99.96% D M S O - 4 ) of the highest available purity (30 mg/mL) . Nuclear magnetic resonance experiments were performed on a Bruker AM-360 spectrometer operating a t 360.13 MHz. The nmr data were processed on a n Aspect 3000 computer with standard Bruker software.

Resonance Assignment

The resonance assignment is based on connectivity information transmitted via homo- and heteronuclear scalar couplings and follows established proc e d u r e ~ . ' ~ -The ' ~ spin systems of the amino acids could easily be assigned with a conventional double quantum filtered (2QF) H,H-COSY e ~ p e r i m e n t . ' ~ . ' ~ The digitization was sufficiently high t o resolve also the cross peaks from the small J couplings between H B and H"s ( = H-C2, 6 ) of the three N-methyl tyrosines in the cyclo-peptide 2 12 and thus identify these residues unambigously. The two-dimensional ( 2 D ) H,H-COSY data matrix consisted of 600 slices of 4096 data points, which yielded, after zero filling to 2K in F1 and Fourier

transformation, a 2 MW matrix. The data were sinebell apodized and calculated as phase-sensitive ~ p e c t r a . 'A ~ spectral window of 3700 Hz was employed in both dimensions. Sequence-specific resonance assignment was obtained by the observation of interresidue C:/H,+, long-range couplings in the H,C-COLOC spectrum.'8-21 T h e assignment of the carbonyl carbons in the H,C-COLOC spectrum follows directly from cross peaks to protons whose chemical shifts and spin system identity are known from the 2QF COSY experiment. The 2D H,C-COLOC data matrix consisted of 128 4K spectra. The T1 dimension was zero filled to 512 words. The data were sine-bell apodized and Fourier transformed t o yield a 2D magnitude spectrum. A spectral window of 18500 Hz in F2 (I3C) and 3700 Hz in F1 ('H) was employed. Independent proof of the sequential assignments was obtained by the 2D nuclear Overhauser effect spectroscopy (NOESY) experiment where the sequence can be traced out by dipolar H,H through-space interactions." T h e assignment of the carbons was achieved by a H,C-COSY experiment with the help of the directly bound protons. The chemical shift of the unprotonated aromatic carbons ( C and C

Conformational studies of cyclic peptide structures in solution from 1H-Nmr data by distance geometry calculation and restrained energy minimization.

The three-dimensional structure of a cyclic bouvardin analogue, cyclo (-Pro-MeTyr-Ala-MeTyr-MeTyr-D-Ala-) has been determined by distance geometry cal...
921KB Sizes 0 Downloads 0 Views