Conformation af endothelin in aqueaus ethylene glycol determined by W-NMR and molecular dynamics simulations

Reecivcd I8 Dccembcr The solution cenfotmrttian of u 21 ariduc

1990; rcvixed vcrsian rrceivctl24

hnuwy

1991

(ET-1 ) in wntrr-rthyknc plyol hnr, been dctemlincti by twodmcnThe N+xminus (rcsiducr I-4) nppct~ to undergo cen~~rmntiunel nvcrnginy and no single structure cnnoirtcnt with the NMR canatminta eakl lx rcsnnd for this rcpion. Residues J-8 form rl turn, und rcriducr 9-16 cxiu~ in n hrlicul conrormntion. A flcxihlc ‘hinge’ txtwecn nsiducr 8=9 nlksv~ vnrinus arirnt~ttionr al thr turn relative to the helix, Another ‘hinge’ at residue 17 connects the cxtrnded Gtcrminur to the bicyclic core region (rcxidura; I-I 5). Residues important for binding and biutupicnl ttctivity form u cuntiguout surfi~cc on ant side UT the helix. with the two disullidea cvtcndiny from the ahsr &Jr of the helix.

rianttl

~KNMR

apxtruxcupy

and

vnsuconstrictur pptidc cndathclin-l

canstnined

molecular

Endothclin;

NMR:

dyntrmics

simulations.

Molecular

dynamics:

I. INTRODUCTION In this communication, wc summarize our recent NMR experiments and molecular dynamics simulation studies aimed at deriving the !hrcc-dimensional structure of endothelin-1 (ET-l) in aqueous ethylcnc g!gcol solution. ET-I, a 21-residue peptide with two disulfide bonds is a member of a new class of peptidic agents that includes mammalian endothelins and the sarafotoxin snake venom peptides [1,2]. Originally isolated from porcine endothelial cells [l), ET-l has been shown to be a major endogenous vasoconstrictor in mammals 131. ET-l may also have other physiological or pathological functions, such as hemodynamic regulation and modulat.ion of the cardiovascular system [3,4]. Increasing evidence for the involvement of endothelins in human disease has prompted a major effort in structural elucidation and pharmacological evaluation. A hypothetical structural model of endothelin and circular dichroism measurements both suggested some degree of helicity [S]. Prelim.inary NMR studies [6,7] in aqueous solutions indicated a heli.; I! region between residues 9-15. The two different NIvIR structures proposed for ET-1 in dimethyl sulfoxide [S,9], are quite different and neither shows a regular helical region. Correspondence address: S.R.

Krystek Jr., Department of Macromolecular Modeling, Bristol-Myers Squibb Research Institute, PO BOX 4000, Princeton, NJ 08543.4000, USA, or N.H. Anderson, Department of Chemistry, University of Washington, Seattle, WA 98195, USA. Fax: (1) (609) 683 6607

Thrcc.tlinlsnr;ion;lt

structure

In this study a novel algorithm, DISCON (= DIStancc CONstraints) [lo], has been used to obtain experimental interproton distances with tighter bounds than those used in typical protein NMR studies [l l], The experimental distance constraints were then employed in three molecular dynamics programs (CONGEN, DISCOVER, and XPLOR) to derive models of the three-dimensional structure of ET-l. We found that portions of ET-1 have stable persistent structure, while other structural regions undergo conformational averaging. The persistently ordered regions include the side chains implicated in physiological function. In particular, our structural model identifies a contiguous surface area on the molecule that may be responsible for receptor binding. 2. EXPERIMENTAL 2.1

I Experitnenrul methods

Synthetic ET-I was obtained frotn Penninsula Labs (Belmont, CA). NMR samples contained 1.5-3.2 mM of peptide in 60% v/v ethylene glycol-da/Hz0 with trifluoroacetic acid (2.5 mM) or KHIPOJIKZHPOJ (1G mM final). Samples in D20 media were obtained by repeated addition of 99.99% D20 and lyophilization. All NMR spectra were recorded at 500.145 MHz using Bruker instruments, Phase sensitive 112,131 NOESY spectra were collected and processed as previously described [14] with the following exceptions: all samples were spinning, and data was apodized using a skewed 60” shifted sine bell in both dimensions. Six separate experiments were carried out at 40-75% ethylene glycol, varying the mixing time, temperature and pH. TPPIXOSY spectra [15] were collected at each

Published by Elsevier Sciewe Publishers 8. V.

ET- I displnyrr well-rcsolvcd rharg NMR red~ntu~cc~ in QJ”to nquearua cthq”lcnc glycal. A downfield portion of the 160 ms NOESY RF 294 K is shown in Fig. E, All resmnncttn of &T-l spccrra could be identified by a orpplicwtion of rclafivelr srrnightforward the COSYPNCHSY (i-c, spin correlation and nuclear CJverhauscr effect) strategy used for protein rtructr~res [22]. The use two temperature?; (295 K, 309 K) for the NOE spcccro~opy was essential for the resolution af the extensive shift coincidences in the NM region, The aromatic & hydrogcnr gave, in each case, tm NQE eonm nectivity to the corresponding Her and/or I-Q3 resonances which were thus distinguished UWIIIbiguoutly, The corresponding HIS/I-IN eonnectivitiss appcnr in Fig. 1, The ~ftme figure diagrflmo the (i, i + 1) dcr~ conncetivirics from LyrY thru His “. Unambiguous (&i + 1). cl&d connecriviries link Cys’-,Lys” and 18--f rp *I. The resulting chart of sequential ctrwncctivicics npgears in Fig. 2, Tnblc I contains rhc eompletc resonance essignmenr, The five NCIESY experiments afforded 220 conformationally dependent constraints with upper and lower bounds and more than 40 con-

of

!95K (160msl v13

whcrc r” 11141 c: I’“ij+ fOrCCIl,;,&’

where force ,,,(,, is ;I user defined value (k is the Roltrninnn canstant. 7 is tlrc nbsalu~c tcmpcmture, S is rhe NOE scnlc factor and A is an crrot estimate). The skcwcd potential in CQNGEN and DISCOVEIZ allow. ed for inclusion 01’ all constraints into a single set. The dynamics simulation protocol in CONGEN consisted of 50 ps of constam temperature dynamics (3OOK) with a NOE maximum force constant of 10.0 kcal~mol-‘a A*’ duriny the dynamics and a finnl NOE force constant of 2,o kcnl~mol-‘e A-‘“’ d Wing the 200 rtcps of conjugate gradient minimirarion which also effected disulfidc closure. The DlSCOVEK proccdurc consisted of 50 ps molec~ler dynamics, which included 30 ps at the maximum tempcraturc (900K). The NOE force constant at Its maximum value was 89.4 kcal* mol”’ aA-‘*, the NOE force constant was 7,45 kcal*mol-‘a A-2 during the 200 steps of conjugate gradient minimization, For XPLOR dynamics simulations [21] the interproton constraints were divided into six groups namely (I) well determined dirN(i + I), d&i + I), dNN(i + I) connectivities and shorter d&i), d,,N(i + 3) and d&i +3) connectivities; (2) other intra-residue connectivities and less Well determined SCqUCntial &NI dNN and d(jN connectivities (i, i-l-n, n < 4); (3) inter-residue connectivities across the disulfide linkages; (4) all constraints involvin& methyl groups; (5) all other intra-residue constraints; and (6) a set of distance constraints to accomplish disulfide closure and to favor the regular torsional angles & 90” for the CB-S-SCB cystine :iJe chain atoms. Molecular dynamics was carried out at SOOK. Sca!e factors, initially set at 1.0 were uniformly increased by a factor of 1.15-l ,60 every 30 fs until they reached the maximum values of 60 kcal.mol-‘*A-” for constraint sets (3) to (6), and IO0 kcal~mol-‘~dr-2 for the constraint sets (1) and (2). The scale factors were periodically reduced by a randomly varied factor and then reincremented to their maximum values. Afterwards, structures were subjected to 200 steps of Powell minimization with NOE force constants at 20-40 kcal~mol-‘~A-2. for ensh cotwraint

and C= kTS/2d2

7.44

$1

I

63

I 9.21 E

’

I

, 1

7.44 +-

&(Ppm)

Fig. I. NH/NH region of the NOESY spectrum (ml = 160 ms) of ET-1 (pH 3.2, at 295K). Backbo:‘e NH positions aie solid or dashed lines labeled by residue number, The missing dplpl connectivities are shown as empty boxes, the continaous solid lines trace thedNN connectivities from SerJ through His’“. Aromatic CH positiorls ace shown by labeled dotted lines.

213

Fig. 2. Summary uF the scqucnrinl connccrivilicr. NO8 inmsitics mcshown by brrr thiekmxx: for ~hc (i. i + I) eennecrlvl~ica, in aprn bar inclientra it polcnlially mnbignmio ~~oipnmcnr, rn x intlisms (I shiR coincidcnee preventr devmion: r0r (i, i 4 n; n - 2-5) sonnrcrivili~n, Q da&xl linr: or bnr intliczms :\ prehablr nsai~nmenr (~scd 8%&Ilcsr weigh~ctl conrrroin1 in XPLOR), a doffed line indicates rht:t n rpceil’ic connectivity could be abrcurctl by ulhcr pc;lks but ir narimcd nbmr (ml cscludstl from nil cmmim list%).

straints with only lower bounds. The data set was rcduc. ed by deletion of constraints involving 19y, 20y, and 210 protons, due to our lack of prochiral assignments, and conservative conversion of @CHzunits co wildcard entries yielding 137 entries with upper and lower bounds and 31 entries with only a low bound, For the

better-defined short distances (rii< 3, IA), bounds were as tight as f 0,2 i\. The average range, < r”ij-r’iiS ) for the full set of constraints was 0,74 A. Molecular dynamics simulation runs produced 68 final NMR structures using XPLOR, 9 structures using CONCEN, and 18 structures using DISCOVER. Six

Table I Chemical shift assignment for cndothclin at pH. 3-P Residue

Chemical shift @pm) NH

Ly

B,R’

Cl s2 c3 s4 s5 LG M7 D8 K9 El0 Cl1 v12 Y13 F14 Cl5 H16 L17 D18 119 120 w21

ex 8487 8.18 9.10 7.62 8.70 8.06 7.36 8.20 8.44 7.59 8.15 7.89 8.27E 8.64’ 7.92 7.92 8.35 7.72 7.94’ 8.08”

4,331’ 4,70 5.0s 4.27 4856 4.13 4.40 4.72 3.86 4.14 4.22 3.46 4.22 4.16 4.58 4.44 4.25 4.59 4.10 4.I6 4.63’

3.30’ 3.80 3.21,2,49 3.95,3.90 3.96,3,66 1.64,1.60 2,17,1,89 3.21’,2.77’ 1.85,1,81 2.14 3.20,3.12 2.05 2.99,2.98 3.19,3.18 3.26,2.94 3.33’,3.32’ 1.62,1,53 2.84”,2.72’ 1.70 I.77 3.29,3.18

others

y 1.68; d 0.91,0,82 y 2.6lJ.45; c 2.08 y 1.53,1.43; 6 1,68,1.64: E 2.98,2.95 y 2.49’ y 0.98,0.83 y 6.84; B 6.66 6 7.32; t 7.35; S 7.28 82 7.25b; c 18.58” y I .52; 6 0,82,0.81 y, 1.33,l.M); yz 0.60; 6 0.78 y, 1.37,l.OS; y* 0.80; 6 0.79 61 7.17; E, 7.56; r3 7.05 rz 7.35; 72 7.11

“All shifts are 6 values at 295K reported to the nearest + 0.01 ppm for 60% v/v ethylene glycol-da H20. Resonances that undergo a substantial upfield shift upon adjusting the pH to 7 are highlighted: bA6>0.20, ‘Ah= 0.08-0.20 ppm. Some of the NH signals are shifted downfield at neutral pH, these are not highlighted.

214

1qw

” t I (3(&u 79t43t -72ttet 51tiw - Mi[ZI) 61(411 -46(M) -S(Ml -9t (591 -3$(4%) -61(63) -2qtw -22(98) i4(SlI

I[ I a4 42ldB) I I I(611 -431143 4(ltSU Il(W fJ(63) -44f64) lJQI1 f *I M(54) -?l(FJ) =4~txl, 73tt 13) atIm) -3(V)

-l03(&M) -23(501 -5b(J9) -28(27) -75(18) -Jl(l2) -Gt(7) -2llttat -98(28)

-iOl(25) -Jf(l II -65(17) -72(11) -G9(40) -30(41) -53(12) -25(Q) -?9(44)

-7$(29) -4l@2l -63(31) -16(2?) -S4(4lf -42(X) -43(241 -37(25) -G&(37)

-4(a) -113(97) -223104) -39(95) -9(521 -37(101) lG(G3) -47(GJ) 58(56) -102(18) 62(8l) -lOR(39) -1 lO(8l) “Avcrngc

of

I6 .rclectcd

structures

W&t)

-4~t\(111 IZl{xll -%(Z II I Swtl -RtttUt 7J(701 -Sl~Xll RZsi) ” tomu) -5?(l6) -13(??) S(99) =m(tt I ) 2i1(31)

- 18(30) -144(27) -33(761 -9X(46) tt(75) -39(lnt) 52(6(1) -128(28) 74(15) -I 16(18) 66(51) -I IJ(30) 136(4)

- I3(28) -124(31) -6x3 -BY(J

t1 II

-30(20) -4tt(R2) l7(114) -56(7G) 21(11) -80(70) 39(G4) -54(80) “146(70)

Jhi7fI -lOl((itQ 119(741 -?f@W 25W 4(9%1 M(7 I] -liw~~ 9(3Ll :I lY(36~ -3909) -113(27) 71~l66) -I I(931 4(351 -81(3 I) -5?(13) -c&t(8) -24(t 2) -6S(l8) -43( 10) -53(13) -27( I Et)

-99(27) -6(29) -127(29) .70(17) -85(491

-12(51) -35(&w -25(9S) -93(71) 32(M) -PO(N) -4917 I) -41(M) -120(77)

(see ml).

highly divergent starting structures were used, at least 4 of these were used in each protocol. None of the structures produced were disregarded in the analyses. Table II shows the variation in dihedral angles for all structures generated. It can be seen that the backbone conformation between residues 9- 17 is essentially identical from the 3 protocols while the N- and C-termini show wider variations. To allow for a normalized comparison of structures generated with the use of the CONGEN, DISCOVER and XPLOR force fields, selected models were relaxed to the local minimum using steepest descent minimization in CONGEN. Four CONGEN-, 5 DISCOVERand 7 XPLOR-generated structures were selected based on: (1) an initial low average violation measure (c 0. I2 A in XPLOR, and co.25 A in CONGEN and DISCOVER), and (2) acceptable energetics and fit to long range constraints after relaxation. Comparison of

the 16 selected structures is the subject of Tables II and III and Fig. 3. The complete backbones for 5 selected structures appear in Fig, 4. Tables II and Fig. 5B indicate that ail 3 protocols produced a helical region between residues 9-17, however, the helix is not regular. All 3 protocols place the backbone torsion angles *II and 914 at anomalous values near zero and open the helix at $15 (circa -12OO). For the N- and C-terminal regions data presented in Table II and Fig. 3 clearly suggest conformational averaging. However, a consensus turn-like structure emerges between residues 5-g (Fig. §A), Data in Table II also point to the S/9 amide lmit as a locus of conformational fluctuation (\k~ and o9). Similarly, the wide variance in +,/9 at Eeu” suggests -that this residue acts as a ‘hinge’ facilitating different orientations of the tail with respect to the rest of the molecule. The C-terminus does, however appear to be structured (cf. Fig. 1). 215

Table III reveals that XPLOR-1.5 achieves its low levels of violation via large distortions of bond angles and large violations of repulsive van der Waal terms, After rt common relaxation, all 3 protocols give structures with comparable fit to experimental constraints.

1.0

z: z

0.5

0.0 1 2 3 4 5 6 7 6 9 101112131415161716 I IIll I III II II III II 4 5 6 7 6 9 10 1112 13 14 15 18 17 1819 M 21

Range Fig, 3, Average root-mean-square dcvia:ion (RMSD) of backbone coordinates for the 16 selected structures taken pairwise. Using a window of 4 residues the RMSD of overlapping segments displaced from each other by one residue was calculated for each range.

216

Novotny ct al, [23] and Chichc et al. 1241showed that the ratio of nonpolar/polar sidechain solvent acccssiblity is a sensitive measure distinguishing native proteins from incorrectly folded models, with native proteins showing the maximum number of polar atoms exposed to the solvent. The non-polar/polar side-chain acccssiblc surface ratio for ET-1 structures generated by DISCOVER are more favorable than those of CONGEN and XPLOR suggesting that DISCOVER generated structures that more closely resemble native proteins. The ET-1 structural models dcrivcd in this paper have important physiological implications. It is known that the correct intact disulfide bonds are required for receptor binding and vasoconstrictor activity of ET-1 [25,263. Other residues important for binding and biological activity arc Asp”, and Glu”’ Phe”, and possibly His ” and or Asp” [I ,27-291. It’is interesting to see that the 2 disulfides are on one side of the helix (9-17) while the other side of the helix forms a contiguous area containing most of the above mentioned residues. Another side-chain implicated in biological activity is Trp” [25,23,29]. In our models, this residue is separated from the above active surface patch by flexible ‘hinge’ connecting the C-terminal tail to the rest of the molecule. In some accessible conformers this may place the C-terminus in a defined contact with the bicyclic core. A denser web of stereospecifically defined connectivities will be required to resolve this question,

Fig, 5. Views (backbone atoms only) of the consensus structural elements of the 16 selected models. (A) Residues 3-9 are displayed 5-8 superimposed showing the predominance of a turn-like conformation in this region of the peptide. (B) Residues X-17 are displayed 9-16 superimposed, demonstrating the presence of a helical region.

residues with residues

with

Volum¢ 281, number 1,2

FEB$ LETTERS

REFERENCES

[I'll (:loru, (J.M., N , ~ ,

April 1991

M,, S.l~.m~r... D.K., l~ru.mer, ^.T.,

K~rphl~, M, ~.d ¢Irolzenb~r.. A.M, (l~tl6) EMBO J. ~, [1] Ya..lll~awa, M,, K.rlh~tra, H,, Ktm~:r~t, 5., l"o~obe. Y.. Kob~y,~hl. M,, Mil~i, Y,, Y~l,~tkl, Y,, Go(o, K, ~nd Mcl~aki, T, (1988) Nal.re 33;I, 411-415, 12] Kloos, Y., Amb~tr, I,..~kolov~ky, M., Ko~;hv~. E., Wollbw¢, ~. and [hlolah, A, (1988) Science 242, 268~270, 13] Yamtilsctwa, M, ~t,tl Masakl. T. (1989] TIPS 1(3, Y'/4-)?~. [4] N~yler, W.G. (i91~9) J, Appl. PhDiol, 4, 49~1=~03, ISl Perkins, T,D.J,, tlider, R,C. ~nd Badow, D.J. (1990) Á~h J P~pllde Protein Re~, 36. 128=1,t~l, [6] Relly. M.D. ,nd D.nbnr, J,B, (1990) Abslrac~i IUPAB Sttl¢llil¢ Symposium Whlsller, Brldsh Coh:mbizt. [?l Dals~trno, D,, Senior, M., Slicer, L. trod Cit~iek~lamannil. S. (1990) Ab~lrttet, E,s(ern An;tlyll~:al Sympo~ium, Io~;,, 5onl~rset, NJ. [8] S=t.dek, V,, HoflwJk, J, ai~d Pdton, J,T. (1989) I:EBS LetL 257, 145-148, [9] Endo, S,, l.ookm H.. I~hibtt.~hi,Y,, Kh.da, C, Mixtt|a, E, a.d Fuji.o, M. (1989) FEBS Lell, 25~, 149=1~;4. [10} Andersen, N.H,, L=ti, X, .nd M;tr~¢hner, T, (1990) NOESYSIM/DISCON Documenla|lon, ~:opyrighled by the University of Wad~inszon; Lt~h X.. Chen, C. =rod Andersen, N,H,, J, Mzlgn, Reson, (in preparation), [ I l l Clore, G,M,, Appella, E., Y,'m~a(lm M,, Ma|sushio~a, K and Grot~enborn, A,M, (I.990) Biochemistry 29, 1689-r/o0, [12] Mueller, L. and Ernst, R.E, (|979) Mol. Phys. 38. 963-992, [13] Stat¢s, D,J,, Haberkorn. R.A, .',,ndRnben. D,J, (198;=)J, Magn, R~son, 48, 286-292, [14] Andcrs~:n, N.H,, Eaton, H,L, and Nsuycn, K.T, (1987) MaS, Resort, Chem 25, 1025-1034, [15] Marion, D, ~ml Wuthrich, K. (1983) Biocllern, Bicpi~ys, Rcs, Commun, 113,967-974. [16J Br~¢col~ri, R.E. and Karplus, M (1987) Biopolymers 26, 137-168,

218

$~amin~lh~n. S:. , , d K~rplm. M. i;1983) J. Comptm Chem. 4. [I 9] [taz~lino, D,A,, N~vo;rly. J, ~nd t.tru~¢ok'rt, R,E. ~,l~ll~¢l'ipz in proper,lion. [2(1] DISCOVER,/INSIGHT Mot~¢tllar Moddin8 Sofl:ware, version 2J. I~Io~ymTedmoioltles, S~. Di¢~lo, CA, [21) ,~PLOR Mole~ultt~" Moddlo8 Soflv.-~lre, ver~loa I.$, Polysen Corp., W~ltl~ln, MA. [22] W.|llri¢ih K, (1986) NME or Prol~il~ls,nd Nn¢leh: Acids, Jolm Wiley .nd Son~. New Yt~rk, [231 Novotny, J,, Ra~hin, A,,A, ;|rid Brtl~'olerf, R.i~. (1988) Pro[ein~: SirlnIlllr¢, Ftln¢llOn, rind (gelit~li¢'~4, 19-30, (24J Clfich~, L., G~boriatm, C.. Hei(x, ^.. Mornom J,.P., C~s(ro, B..rid Kollman. P.A, (1989) Proteins: S~ru~ure. F.nedom and [2~] Kim~lrm S., Ka~tly~, Y., S~w~mur~. T., Shinmi, O., Susitm Y.. Y~m;tsi~;~,wa, M,, Goto. K. m~d M~ts~tki. T, (1988) Bio~ch~m, Blophy~, Rc~. Commz:n. t56, 1182-1186, [26} K.masaye, S.-I ,, Kuroda. H,, Nakajima. K., Walzt,abe, T.X,. Kimurm T,, M ~ k i , T. ~nd Sak~tkibara, S, (198B) Int, .I. PepI, Protein Res, 32,519-526, [2"/] N~tkajinlm K,. K¢ibO,S,. Kt~r~aS~ye.S..I,, Nishio. I,|,, T~uncmi, M,, Inui, T,, Kuroda, H,, Chino, N,, Wazanabe, T,X., Kh'aura, T, and Sakakibarm S, (1989) Bio~h©m. Biophys, Re~, Commun, 163, 424-429. [28] Kitazumi, K,, Slfiba, T,, Nishiki, K,, l.':urukawm Y,, Taka~aki. C, m~d T.~s.'zk.'t,K, (1990) FEBS LeH 260, 269-2"/2, [29} Johanscn, N,L., Lundt. B,F., Mad~cn. K,, Olscn, U,B,, Suzd:tk, P,, Thogcrsen, H, and Weiss, J,U, (1990) Abstract. 2lst Eu,'(~pean Peptidc Symposium,