the three tyrosines and the 6-amide protons of the two as- paragines to .... GVZA at 2.5 "C and pH 5.0. Residue (Hz). 6. 3JHNa. K2 s 3. 0 5. S6 s7. C8 s9. TI 1 s12.
Protein Science (1993), 2, 1591-1603. Cambridge University Press. Printed in the USA. Copyright 0 1993 The Protein Society
Solution structure of the calcium channel antagonist o-conotoxin GVIA
JACK J. SKALICKY,'y4 WILLIAM J. METZLER:*4 D. JOSEPH CIESLA,' ALPHONSE GALDES? AND ARTHUR PARDI'
' Department of Chemistry and Biochemistry, University of Colorado at Boulder, Boulder, Colorado 80309-0215 Bristol-Myers Squibb Pharmaceutical Research Institute,P.O. Box 4000, Princeton, New Jersey 08543-4000 The BOC Group Technical Center, 100 Mountain Avenue, Murray Hill, New Jersey 07974
(RECEIVED June 4, 1993; ACCEPTED July 23, 1993)
Abstract The three-dimensional solution structure is reported for w-conotoxin GVIA, which is a potent inhibitor of presynaptic calcium channels in vertebrate neuromuscular junctions. Structureswere generated by a hybrid distance geometry and restrained molecular dynamics approach using interproton distance, torsion angle, and hydrogenbonding constraints derived from 'H NMR data. Conformations of GVIA with low constraint violations converged to a common peptide fold. The secondary structure in the peptide is an antiparallel triple-stranded &sheet containing a &hairpin and three tight turns. The NMR data are consistent with the region of the peptide from residues S9 to C16 being more dynamic than therest of the peptide. The peptide has an amphiphilic structure with a positively charged hydrophilic side and an opposite side that contains a small hydrophobic region. Residues that are thought to be important in binding and function are located on the hydrophilic face of the peptide. Keywords: calcium channel blocker; w-conotoxin GVIA; Conus; NMR; protein structure; toxin
The marine gastropods belonging to the genus Conus are predatory fish-hunting cone snails that immobilize their prey using a highly evolved array of potent peptide toxins (see Gray et al. [1988] and Olivera et al. [1990, 1991bl for reviews). These so-called conotoxin peptides exhibit their effects in a concerted and specific manner by interfering with various components of the nervous system such as ion channels and neurotransmitter receptors. Conotoxins are classified according to their physiological targets, with the a-conotoxins blocking the neuromuscular nicotinic acetylcholine receptor, the p-conotoxins
blocking voltage-sensitive Na+ channels in skeletal muscles, and the w-conotoxins blocking neuronal presynaptic Ca2+channels. Because of their high specificity for a diversified setof targets, the conotoxins have becomeimportant tools for thestudy of the nervous system (Hillyard et al., 1992). The w-conotoxins are small peptides26-29 amino acids in length.The amino acid sequencesof many of these toxins have been determined (see Fig. 1). The w-conotoxins are highly cross-linked with a conserved disulfidebond arrangement (Nishiuchi et al., 1986). In addition, the glycine at position 5 is conserved in all the w-conotoxins characterized to date. Although, the remainder of the Reprint requests to: Arthur Pardi, Department of Chemistry andBioamino acids showno absolute sequence conservation, pochemistry, University of Colorado at Boulder, Boulder, Colorado sitions 2 and 25 are always occupied by either a lysine or 80309-0215. 4The contributions from these authors should be considered as an arginine. Moreover, a high concentration of residues equal. contain hydroxyl moieties, which is accentuated in many Abbreviations: COSY, correlated spectroscopy; RELAYED-COSY, relayed correlation spectroscopy; DQF-COSY, double quantum-filtered of the w-conotoxins by the substitution of y-hydroxyprocorrelated spectroscopy; TOCSY, total correlation spectroscopy, line for proline. PECOSY, primitive exclusive COSY; NOESY, two-dimensional nuclear Analysis of conotoxin cDNA clones indicates that the Overhauser enhancement spectroscopy; TOCSY-NOESY, combined tow-conotoxin GVIA is generated from a 73-residue preproDG, distal correlation-nuclear Overhauser enhancement spectroscopy; tance geometry; rms, root mean square. peptide (Colledge et al., 1992). The prepro region of the 1591
J. J. Skalicky et al.
1592 o-conotoxin GVIA is quite homologous to the prepro region of the related so-called King Kong peptide (Olivera et al., 1990; Woodward etal., 1990; Colledge et al., 1992). As pointed out by Colledge et al. (1992), the significance of homology between the prepro sequence of the King Kong peptidesand the prepro sequence inthe o-conotoxin GVIA cannot be interpreted unambiguously at this time. However, one can speculate that the prepro region of the peptides may be required for correct folding of the mature peptide. The large variabilityof amino acids inthe inter-cysteine regions of the conotoxin peptides is analogous to the hypervariable regions of antibodies (Woodward etal., 1990). Remarkably, the different o-conotoxins target the same receptor, the presynaptic neuronal Ca2+ion channel, yet display hypervariability of amino acid residues in the inter-cysteine regions. Perhaps thevarious o-conotoxins are targeting different microsites on the Ca2+ channel (Olivera et al., 1991a,b), or, alternatively, the backbone conformation of the various o-conotoxins may be very similar with the binding on the Ca2+channel driven to a large extent by the overall fold of the peptide and to a less extent by the exact amino acid sequence. Although o-conotoxin GVIA is widely usedas a Ca2+ channel probe, only a preliminary structure has been reported (Kobayashi et al., 1987). The small size of this peptide makes it quite amenable for study by NMR spectroscopy. In this report, we describe the three-dimensional conformation of o-conotoxin GVIA in aqueous solution as determined by a combination of NMR and computational techniques. This structure provides a physical basis for understanding many of the observed activities of GVIA.
1
15 5
10
20
Sequential resonance assignments Assignment of the 'H resonances in GVIA was accomplished by application of the sequential assignment procedure developed by Wiithrich (1986). Protons belonging to spin systems for each of the 27 amino acids were readily identified. Analysis of the spectra collected in H 2 0 (TOCSY, COSY, RELAYED-COSY) allowed the identification of 23 spin systems containing an amide NH (Fig. 2) and the corresponding spectra in 2H20 allowed the identification of the three hydroxyprolines and the N-terminal cysteine.The assignments of degenerate methylene protons were confirmed with the triple quantum spectrum. Spin system information was used to identify amino acid type when possible. The identities of the two arginines and one of the lysines wereconfirmed by connecting their side-chain methylene protons to their respective N'H and NcH2 resonances (Fig. 2). In addition, analysis of the NOESY spectra enabled the aromatic protons of the three tyrosines and the 6-amide protons of the two asparagines to be linked to their respective @-protons.Fif-
25
CKS~~SSCSETSYNCCR-SCNEYTKRCY*
GVIA GVIB GVI C SVIA SVIB GVIIA GVIIB MVIIA MVIIB MVIIC
CKSgGSSCSzTSYNCCR-SCNgYTKRCYG* CKSEGSSCSETSYNCCR-SCNEYTKRC* CRSSGSECGVTSI-CC-GRC--YRGKCT* CKLKGQSCRKTSYDCCSGSCGRS-GKC* CKSPGTflCSRGEIRDCCT-SCLLYSNKCRRY* CKSEGTOCSRGI4RDCCT-SCLSYSNKCRRY*
CKGKGAKCSRLYYDCCTGSCRS--GKC* CKGKGASCHRTSYDCCTGSCNR--GKC* CKGKGAECRKTMYDCCSGSCGRR-GKC*
I Disulfide linkages
*
I
C
b
.?i9
c
Fig. 1. Maximal amino acid alignment of w-conotoxin primary StrUCtures from three species of Conus snails. Standard one-letter amino acid abbreviations are used except for y-hydroxyproline, which is denoted Primary structures for w-conotoxins include GVIA (Olivera et a]., 1984); GVIB, GVIC, GVIIA, and GVIIB (Ohera et a]., 1985); SVIA and SVIB (Ramilo et al., 1992); MVIIA and MVIIB (Olivera et al., 1987); and MVIIC (Hillyard etal., 1992). The disulfide arrangement has been reported for synthetic GVIA (Nishiuchi et d.,1986). The asterisk indicates an amidated C-terminus.
r.
Results
8.'0
9.' 0
0 2
7.' 0
(PPW
Fig. 2. Contour plot of the amide protonto aliphatic proton regionof the TOCSY spectrum that was usedto assign amino acid spin systems. The spin systemsfor all residues that contain amideprotons, except for C1, are shown. For R17, K24, and R25, the backbone amide NH and side chain N'H or NrH are shown. The vertical lines connect the spin systems.
Solution structure of a-conotoxin GVIA teen of the 20 spin systems could be assigned to a specific amino acid type; the remaining 12 spin systems were either cysteine or serine. Sequence-specific assignments were made by analysis of the COSY, TOCSY, and the jump-return NOESY spectra in 90% H 2 0 (Figs. 2, 3). The three y-hydroxyprolines broke the sequential daNconnectivities into four segments. These segments are connected with dU6connectivities between eachy-hydroxyproline and its preceding residue. For each segment, at least one unique amino acid spin system (or pair of spin systems) could be identified, thusproviding numerous starting points for the sequential assignments. These included the unique glycine (G5) for the segment containing residues P4 through S9 (Fig. 3A); T11 and Y13-Nl4 for the segment containing residues p10 through N20 (Fig. 3B); and Y22-T23 and K24-R25 for the segment containing residues p21 through Y27 (Fig. 3C). K2 served as a marker for residues C1 through S3 once all other residues were assigned. Chemical shift assignments for w-conotoxin are listed inTable 1. Couplingconstants (3J,s and3JHN,) were measured using COSY and PECOSY spectra and are listed in Table 2. The NOES used in making the sequential connectivities are summarized in Figure 4.
I
I
CB
0
4'
0
B *
$4
0 g
'6
0
00 0
'
o
# I* 0 '
0
0 0
TI1
.O
&
'
e
0
*
Q
Secondary structure Several NMR spectralfeatures can be used to qualitatively describe protein secondary structure. The relatively large NH chemical shift range (2.4 ppm) and the presence of a single set of resonances for each amino acid indicate a single stable conformation. The well-resolved chemical shifts of the amide protons are characteristic of &sheet secondary structure (Wuthrich, 1986; Wishart et al., 1991). Stable secondary and tertiarystructures protect amide I o #O protons from exchange with solvent (Englander & Kallenbach, 1983). Seven slowly exchanging amide protons were observed after dissolving the peptide in 'H,O at 25 "C, 9.5 9 8.0 .0 8.5 7.5 7.0 pH 5.0 (Fig. 4); allthe protons completely exchanged with 'H20 within 4 h. Three additional amide protons (S6, 0 2 (PPW Y22, and T23) exchanged slowlyenough with 2H20 tobe Fig. 3. Contour plot of the combined NOESY and COSY spectraof wobserved at 10 "C and pH 5.0. conotoxin in H20. The sequential resonance assignments are illustrated The temperaturedependence of the amide NH chemivia daN connectivities for residues C1-S9 (A), P10-N20 (B),and p21cal shift (NH temperature coefficients) providesinformaY27 (C). In the COSY spectrum, positive and negative contours areplotted, but only positive contours are plotted for the NOESY spectrum. tion on the nature of the local environment for each NH The horizontal lines connect COSY and sequential NOESYcrosspeaks group (Jimenez etal., 1986). NH temperature coefficients for the C" protons, and the vertical lines connect sequential NOESY were measured at pH 5.0 from 5 to 35 "C and are shown and COSY crosspeaks for the amide protons. in Figure 5 (see Discussion). The chemical shifts of the nonexchangeable a protons do not significantly change from 5 to 35 "C, indicating that the global structure is not perturbed over this temperature range. Remarkably, the chemical shifts for 12 of the a protons and the slowly exQualitative analysis of nonsequential Ha-Ha, Ha-NH, changing amide protons do not change significantly even and NH-NH NOES shows the presence of an antiparalat 60 "C, indicating that the peptide had not unfolded at lel triple-stranded @-sheetin the peptide (Figs. 6, 7). This this temperature. @-sheetis comprised of a @-hairpin from residues R17 to
1594
J. J. Skalicky et al. Table 1. 'H chemical shift assignmentsfor the proton resonances of o-conotoxin GVIA at 25 "C in I O mM deuterated D,-sodium succinate (pH 5.0) a Residue
3.18 4.02
7.71
7.62
8.70
8.85
NH
c1 3.14, K2 s3 4.54 P4 3.67 G5 4.56, S6 s7 C8 s9 p10 T11 SI2 8.1 Y13 N14 C15 C16 R17 S18 C19 N20 p21 Y22 T23 K24
9.40 8.95 2.14 9.24 8.87 8.39 8.33 4.50 7.53 1 8.27 9.48 8.38 7.20 1.95 8.32 4.257.10 8.57
R25 C26 Y27 C-NH2
9.26
7.33 9.30
C"H
CpH
4.58 4.66 2.46,
protons
1.95R, 1.89'
CrH2 1.47, C6Hz 1.52, C'H2 3.05
4.22
4.76 5.01 4.94 2.102.38,4.46 4.15 4.42 3.27',4.56 4.92 4.67 4.56 2.06', 4.62 4.62 4.87 4.56 1.712.26,4.33 4.42 4.25 3.65
Other
CYH4.73, 3.88, C'H2 3.88,
3.75
OYH6.58
3.95R, 3.86' 3.93, 3.93 3.10R, 3.00' 4.00, CYH4.57, C6H2 4.02 4.02, CYH31.15 3.87, 3.87 2.96R 3.03, 2.76 3.WR, 2.47' 3.26R, 2.82' 1.70R 4.05, 4.02 2.97', 2.81R 3.50R, 1.34' 3.14, 3.14 2.40, 2.10 1.32R 3.69', 2.84R 2.89R
1.49', 5.25 4.97 3.19',4.51 7.41
C6H 7.11, C'H 6.86 N6H2 7.56, 7.01
C7H2 1.56, 1.60, C6H2N'H 3.19,
7.20
N6H2 7.43, 6.76 CYH 3.94, C6H2 4.504.50, C*H 7.30, C'H 6.89 CYH31.17 C7H2 1.33, 1.33, C6H2 1.74 1.68, C'H2 3.06, NH2 7.56 CYH2 1.66, 1.41, C*H2 3.16, 3.10,
N'H 7.19
C6H 7.31, C'H 6.85 NH2 8.08,
a The chemical shifts are referenced to the H 2 0 or residual HOD resonance at 4.80 ppm. The chemical shifts have errors of k0.02 ppm. Only assigned protons are listed. P is y-hydroxyproline. The amidated carboxyl-terminus is C-NH2. R and S designate stereospecific assignments.
Y27 and a third strand fromresidues G5 to C8. As seen in Figure 7, most of the cross-strand NOEs expected for this triple-stranded @-sheet wereobserved in the NOESY spectra.
fined using molecular dynamics with simulatedannealing as described in the Materials and methods. After refinement, 38 of the structures had no distance constraint violations greater than 0.2 A or total NOE constraint violations greater than 10 kcal/mol, using a square well NOE potential and a force constant of 60 kcal/mol A'. There were no torsionangle violations larger than 10" in these structures (see Table 3). Complete structural analysis was performed on the 20 structures that gave the lowest rms deviationsfrom the input distance constraints. An
Three-dimensionalstructure calculations A set of 50structures was embedded using the DG algorithm provided in XPLOR 3.0 (Brunger, 1992) and re-
10
5
cK s NH
-
G
s
a .
15
p
-
a
daN
d BN d NN
20
" " " "
.-.. 25
Fig. 4. Sequential connectivity diagram for w-conotoxin GVIA. The amino acid se-
quence OfGVIAis using given standard the single-letter abbreviations, except for the y-hydroxyproline residues, which are designated with E. NH, residues containing slowly exchanging amide protons. NOEs used to connect amino acid residues: d u N ,d O N , dNN,
=;
da6,
0.
1595
Solution structure of w-conotoxin G VIA
-
Table 2. 3JH,,,a and 3Jm8coupling constants and torsion angle constraints (+, xl) f o r w-conotoxin GVZA at 2.5 "C and p H 5.0
N
d
3JHNa
Residue
K2
(Hz) 8.0
s3 05 S6
6 - 120
k
40"
4.5 11.316.3 5.9
s7 C8 s9 TI 1 s12 Y13
6.0 8.7 8.8 8.8 1.4 7.5
N14 c15
7.9 6.0
C16 R17 SI8 C19
8.9
s m-s
3.95 3.86
3
m
10
S
2
Y22 T23 K24 R25
6.0 8.9 8.9 8.8
180 k 30"
-60
30" 4
2
8
e
+ 30" k
io
12
14
18
le
20
22
24
28
Residue Number
30"
+ 30"
3.27 2.96
4 11
w m-s
-60 k 30"
3.04 2.47 3.26 2.82 2.06 1.70
2 12 3 10 4
w-m
-60
10
S
2.97 2.81 3.50 1.34
3 12 12 3
-120 30" -120 k 30" -120 + 30"
1.49 1.32 3.69 2.84 3.19 2.89
12 3 2 12 3
5.8
4.5
w-m
10
S
~
~~
+ 30"
S
w
- 6 0 5 30"
S
w
5.9
6.0
Y27
-120
10
+ 30"
4.6
N20
C26
-120 -120 -120
1.95 1.89
s
-60 k 30" 180 + 30"
S
s
180 k 30"
S
m-s
Fig. 5. A comparison of the slowly exchanging amide protons in wconotoxin GVIA, amide proton solvent accessibility, and theamide proton chemical shift temperature coefficients. The top panel shows the average solvent-accessible surface areas ( A 2 ) for the amide protons in the 20 refined GVIA structures. The surface area for C1 is for only one of the amino protons. The lower panel showsthe amide temperature coefficients. The stippled bars indicate random coil temperature coefficients, and open bars indicate nonrandom coil temperature coefficients. The open circles indicate the amino acid residues having slowlyexchanging amide protons.
Examination of the family of structures determined for w-conotoxin GVIA (Fig. 8) clearly shows that some regions of the peptide are betterdefined by the NMR data than other regions. This frequently occurs in NMR-
180 k 30"
S
s
180t 30"
S
-60
+ 30"
~~
a CBH2are
the methylene proton chemical shifts. Indicates intramolecular NH-HP NOE intensity from 150 ms H 2 0 NOESY; NOE intensity is classified as w = weak, m = medium, or s = strong.
average refined structure was calculated from the20 structures; this structure was submitted to 1 ps of simulated annealing at 1,000 K and energy minimized. Figure 8 shows the 20 refined structures after superpositionof their backbone (N, C", C, 0) atoms, and Figure 9 shows the average energy-minimized structure. a-Conotoxin GVIA adopts a structure that can be characterized as a triple-stranded antiparallel @-sheetwith a +2x, -1 topology (Richardson, 1977). Residues R17-Y27 form a @-hairpinwith residues G5-C8 adjoining residues K24-Y27 to complete the three-stranded sheet (Fig. 8). The three strandsof the 0-sheet are connected by loops. The three-dimensional structure is stabilized with three disulfide bonds (Cl/C16, C8/C19, and C15/C26).
1
5
10
15
20
25
Fig. 6. Diagonal plot of the observed NOEs in w-conotoxin GVIA. The axes correspond to the amino acid residue numbers. The filled squares indicate NOEs between backbone protons, the circles indicate NOEs between backbone and side-chain protons, the crosses indicate NOEs between side-chain protons, and the letter S indicates a disulfide bond. Where a filled square, circle, cross, or letter S occupy the same location, the order of priority is S, filled square, circle, and crocq.
J. J. Skalicky et al.
1596 HH -
0
T
H
0
H
0
Fig. 7. Schematic representationof the secondarystructure of a-conotoxin GVIA. Backbone proton-backbone proton NOEs are indicated with double-headed arrows, and the slowly exchanging amides are circled. The secondary structures for the loop region from residues S9 to C16 and the N-terminus from residues C1 to E4 are not shown.
derived structures. The poorly defined regions can reflect either the dynamic nature forthese regionsof the peptide or the absence of tertiary NOEs to restrain them (see Supplementary material, Diskette Appendix). The antiparallel triple-stranded @-sheetis well defined in the NMR structures. Key constraints for defining this region were derived from NOEs between the a protons of residues
Table 3 . Structural statistics for w-conotoxin G VZA Average pairwise rms deviations (A) for select backbone regionsa Residues 1-27 Residues 17-27 3-8, Residues 17-27 1.82 Residues 9-16 Average numbers of experimental constraint violationsb and energy of final structures' Number of NOE violations >O. 1 A Number of torsion angle (6,x,) violations >IO" Average unconstrained energy (kcal/mol)
1.70 f 0.34 0.73 ? 0.19 0.38 t 0.08 i 0.45 2.6 i 0.9 None -87.3 & 8.0
a Refers to the average pairwise values t average deviations for the final 20 refined structures. The backbone atoms C", C, N, and 0 are superimposed. Average numbers of constraint violations from constrained MD refinement. Average unconstrained energy for structures calculated using standard XPLOR energy terms for bonds, angles, impropers, van der Waals, and electrostatics. The van der Waals energy term was estimated with a standard 6-12 potential. Electrostatics were estimated with the shifted option. See XPLOR 3.0 manual for more detailed information on energy terms.
S7/R25, P21/K24, and C26/C19; NOEs between amide and a protons of C8/R25 and N20/C26; and NOEs between theamideprotons of S6/C26, S18/Y27, and N20/R25 (Fig. 7).A rather unusual five-residue turn from N20 to K24 that reverses the peptide chain direction in the @-hairpinis also well defined by numerous NOEs involving the amide and a protons of residues P21-R25. In contrast, the N-terminal residues C1-p4 and the loop residues S9-Cl6 are less well defined, as there are sub~ a stantially fewer NOEs in these regions. The 3 J ~ values (6.0-7.9 Hz) for certain residues (S12-Cl5) are indicative of motional averaging around the 4 torsion angle. The conformation of the loop and turnregions of the peptides were identified by the NOE patterns, the J values, and the [(&, &), (43,$3)] torsion angles (Wiithrich, 1986; Lewis et al., 1973) inthe 20 final refined structures and the refined average GVIA structure. The large 3 J ~ ~ u coupling constant for T11 (8.7 Hz) and the strong NOE between amide protons of T11 and S12 indicate that residues S9-Sl2 form a type I @-turn.The y-hydroxyproline at position 2 of the turn can be accommodated in a type I turn (Lewis etal., 1973). In addition, the [(&, &), (43,q3)l torsion angles for residues P10 and T11 in the average structure are [(-61", -13"), (-90", -47")], respectively, which is consistent with a type I 0-turn. A second turn occurs from residues N14 to R17. The geometry of these residues does not fit a standard 0-turn. This is not surprising since the two central residues of the turn are cysteines that are involved in disulfide linkages that may deform the turn geometry. Superposition of residues N20-K24 shows a well-defined five-residue @-turn.Residues N20-T23 in the average structure have torsion angles [(42,q2),(43,rC.3)] of [(-71 ",-57"), (-40", -37")], consistent with a type 111 @-turn (Lewis et al., 1973). The small 3JHNa coupling constants observed for N20 and Y22 and a large NOE between amide protons of Y22 and T23 are also consistent with a type I11 turn. TheP21 residue is readily accommodated into a type I11 turn (Lewis et al., 1973). The turn is completed withanother residue, K24, resulting ina fiveresidue turn rather than the standard four-residue @-turn. Five-residue reverseturns have been previously observed (Sibanda & Thornton, 1985). The disulfides Cl/C16, C8/C19, and C15/C26 are well defined and are found in standard conformations (Srinivasan et al., 1990). The disulfides C8/C19and C15/C26 are inaccessible to solvent and contribute to a small hydrophobic core in the peptide. These disulfides are probably instrumental in stabilizingthe structure. The disulfide bond arrangement determined by Nishiuchi etal. (1986) wasconfirmed by the observation of tertiary NOEs between a and 0protons of Cl/C16 and C8/C19 and tertiary NOEs between /3 protons of C15/C26. Several other side chains in GVIA have well-defined conformations. In addition to the disulfides, the side chain of Y27 contributes to the small hydrophobic core. The side chains of N20, T11, and T23 appear to be in-
1597
Solution structure of a-conotoxin G VIA
11
11
volvedin intramolecular hydrogen-bonding networks. For example the 0 7 of T23 is within hydrogen-bonding distance of the K24 NH, and this amide proton is slowly exchanging, consistent with a hydrogen bond. Some of the secondary structural elements have nonstandard geometries. For example, the G5 residue has a conformation suggestive of a &bulge and sharply bends (60")the peptide chain. Since G5 is the only conserved non-cysteine residue among the w-conotoxins (Fig. l), it may be the only amino acid residue that can be accom~ a modated at this position in the structure. The 3 J ~ Values for the stretch of peptide from G5 to C8 show that those of S6 and S7 are smaller than expected for an ex-
Fig. 8. Stereo pair of the superposition of 20 refined o-conotoxin structures. Only the C", C, N, and 0 atoms are shown, and the superpositions were best fits of these atoms for residues p4-C8 and R17-Y27 in the P-sheet. Residue numbers are provided for reference.
tended strand, whereas those of C8 and S9 are normal for an antiparallel @-sheet.Likewise, small 3 J ~ values ~ a are observed for residues S18, C26, and Y27 in the &hairpin. Perhaps some of the nonstandard geometries are the result of the three disulfide bonds that strain the molecular structure. The degree of convergence of calculated structures was determined by analyzing the number of distance and torsion angle constraint violations and the pairwise rms deviations between backbone ( C a , C , N, 0 )atoms of the structures (see Table 3). Other criteria were also utilized to assess the quality of the structures. Unconstrained energy minimization (5,000 steps) of the final structures re-
Fig. 9. Stereo pair of the minimized average structure of wconotoxin GVIA. All heavy atoms are shown. Residue numbers are provided for reference.
Y
Y
1598
J.J. Skalicky et al.
sulted in an average total energy of -87.3 f 8.0 kcal/mol (see Table 3), indicating the pseudo-energy potentials of the distance and torsion angle constraints have not led to unrealistic geometries. The Lennard-Jones and electrostatic potentials are both large and negative, indicating that the distance and torsion angle constraints allowed for favorable interactions. These interactions were formed despite the fact that no Lennard-Jones attractive or electrostatic forces were explicitly included inthe initial structure calculations (see Materials and methods). Finally, Ramachandran plots illustrate that @ and )I torsion angles for most of the residues in the 20 refined structures and the average structure lie in sterically allowed regions (Fig. 10). Phi torsion angles for residues 12-15 in several of the refined structures and theaverage structure lie in positive C#J space; these same residues show the largest C#J and $ deviations (data not shown). The three-dimensional structure of GVIA accounts for several of the unusual features of the NMR spectra. Most striking, the proton of N20 is shifted upfield by 1.4 ppm from a random coil chemical shift (Wuthrich, 1986). Examination of the w-conotoxin GVIA structure indicates that this proton lies directly beneath the aromatic ring of Y27. Thus, the anomalousupfield shift of this proton is the direct result of the ring current effect. Two additional protons, theCY protons of K24 and R25, have an upfield shift of 0.70 and a downfield shiftof 0.90, respectively, from random coil chemicalshifts. These protons are also close to the Y27 aromatic ring.
os
5 0
23
-"OI
I
1Q
1
-120
I
I
I
I
-60
0
60
120
~
Fig. 10. Ramachandran plotfor one of the refinedw-COnOtOXin structures. The non-glycine residues are shown with solid squares, and the glycine residue is shown with a solid circle.
The proton of the OYH group in p 4 was assigned by intraresidue NOES to its fl, y, and 6 protons. The observation of this proton was surprising because the OH protons of serine, threonine, and hydroxyproline generally exchange rapidly with solvent and are usually not observed in aqueous solutions (Wuthrich, 1986). The average solvent accessibility (of the 20 refined structures) for the OYH atom in P4 is quite low (0.1 -t 0.02 A'), in contrast to the larger solvent accessibility for the other hydroxyproline OYH atoms (P10, 5.1 -1- 3.0 A 'and p21, 4.4 k 2.7 A'). Moreover, the P4 OYH proton is within 3.0 A of the hydrogen-bond acceptors on both the amidated C-terminus and the R17 guanidinium nitrogens. Thus the anomalousslow exchange of the P4 OYHproton is readily accounted for by the structure of GVIA.
Discussion
The GVIA structure suggests a role in function The utility ofconotoxins as probes to study the ion channels has resultedin the primary structure elucidation and characteriiation for many of these peptides (Olivera al., et 1991a). Figure 1 lists the known amino acid sequencesof the w-conotoxins. The most striking feature of the wconotoxin primary structures is howlittle is conserved except for thecysteine residues. The only other aminoacid residue that is invariant is G5. Relatively well conserved are the positively charged residues K2/R2 and K25/R25. Two views can be proposed for this apparent lack of sequence conservation. On the one hand,residues that are not evolutionarily conserved maynot be criticalfor either the structure or the function of the peptide. On the other hand, the diversity of the residues may enable the cone snails to attack a wider array of targets, thus ensuring that their prey does not escape them. The three-dimensional structure of GVIA shows an asymmetric distribution of nonpolar and charged side chains (Fig. 11). The basic residues form a distinctive positively charged hydrophilic face consisting of the N-terminus, K2, R17, K24, and R25. This is the peptide surface that is likely to be involved in binding to the Ca2+ channel, since it has been shown previously that the charged lysine residuesare important for GVIA binding to Ca'+ channels (Gray et al., 1988). Biotinylation of either lysine 2 or 24 leads to about a 10-fold decrease in binding, and biotinylation of both lysines results ina 100fold decrease inbinding affinity (Gray et al., 1988). Recent studies show that acetylation of the GVIA N-terminal amine or of the €-amino groups of K2 and K24 reduce binding and functional potency (Lampe et al., 1993). For the monoacetylated GVIA species, acetylation of the N-terminus causes the largest decrease in potency; however, K2 and K24 €-amino groups are also involved in binding the Ca2+channel (Lampe et al., 1993). Decreased
Solution structure of w-conotoxin GVZA
1599
Fig. 11. Stereo pair of theaveragew-conotoxin GVIA structure, where the charged side chains are shown inbold and the side chainsof the hydrophobic amino acids are shown with their der vanWaals surfaces. The amino acids were classified according to the criteria of Rose et al. (1985).
binding of acetylated or biotinylated GVIAs may result from either a decrease in net positive charge on the hydrophilic surface of the peptide, disruption of specific contacts between GVIA and the CaZ+channel, or a conformational change in the modified peptide. Since arginine can substitute for lysine at positions 2 and 25 (Fig. l), this argues for the importanceof the positive charge in binding. This is consistent with the observation that positively charged aminoglycoside antibiotics (neomycin, streptomycin, and kanamycin) and polylysine can specifically inhibit GVIA binding to rat brain membranes (Wagner et al., 1988). The polyamines spermineand spermidine are capable of blocking GVIA binding and arealso capable of blocking voltage-sensitiveCa2+channel function (Pullan et al., 1990). The opposite face the GVIA peptide contains the small hydrophobic core described above, as well as two other tyrosine residues. This face of the molecule is less likely to be involved in binding to the CaZ+channel, since iodination of the tyrosine residues leads to relatively little loss of activity (Abe & Saisu, 1987; Cruz et al., 1987). The large numbers of serine, threonine, and hydroxyproline residues are uniformly distributed over the peptide surface and arequite accessible to the solvent. These hydroxyl-containing amino acid residues may be important forpeptide solubility or, alternately, they may form hydrogen-bonding contacts with the Ca2+ channel. The role of the invariant G5 residue appears to be structural. At the G5 position there is a 60" turn in the peptide chain that would be difficult to produce with any residue besides glycine. Based upon the conservation of the disulfide bonds and theconserved glycine, the structure of other w-conotoxins of the GV* type are probably
quite similar to the structure determined here for GVIA. Most of the deletions and insertions would occur within the &hairpin structure of GVIA from R17 to Y27 (see Fig. 1). The three-dimensional structures for different wconotoxins therefore presumably differ primarily in the &hairpin. For example, addition of a single residue between R17 and S18 in GVIA can be accommodated easily by expanding the loop connecting the strands of the 0-hairpin. The only deletion observed outside of the 0hairpin region is for SVIA and this deletion is located in the loopat position 14. The structuraldifferences within the /3-hairpin would be predicted to not influence the overall structure and amphiphilic character of the molecule, and hence do not affect its function, but instead may modulate the binding affinity and selectivity among the w-conotoxins.
Amide protons as probes of structure anddynamics Amide proton solvent accessibilities, exchange rates, temperature coefficients, and intramolecular hydrogen bondingprovideinformation that complementsthe three-dimensional structure. The 10 slowly exchanging amide protons in GVIA have small temperature coefficients. Theseprotons have zero(or small for K24) solventaccessible surface areas (Fig. 5). Furthermore, each of these amide protons is within hydrogen-bonding distance of a suitable hydrogen-bond acceptor group. In contrast, none of the more rapidly exchanging amideprotons (lifetimes
