(Received for publication, October 1, 1996, and in revised form, December 16, 1996)
From the Department of Chemistry, National Tsing Hua University, Hsinchu, Taiwan and the § Department of Biochemistry, Kaohsiung Medical College, Kaohsiung, Taiwan
The solution structure of toxin b, a long
neurotoxin (73 amino acids and 5 disulfides) from the venom of
Ophiophagus hannah (king cobra), has been determined using
1H NMR and dynamical simulated annealing techniques. The
structures were calculated using 485 distance constraints and 52 dihedral angle restraints. The 21 structures that were obtained satisfy the experimental restraints and possess good nonbonded contacts. Analysis of the converged structures revealed that the protein consists
of a core region from which three finger-like loops extend outwards.
The regular secondary structure in toxin b includes a double and a
triple stranded antiparallel sheet. Comparison with the solution
structures of other long neurotoxins reveals that although the
structure of toxin b is similar to those of previously reported long
neurotoxins, clear local structural differences are observed in regions
proposed to be involved in binding to the acetylcholine receptor. A
positively charged cluster is found in the C-terminal tail, in Loop
III, and in the tip of Loop II. This cationic cluster could be crucial
for the binding of the long neurotoxins to the acetylcholine
receptor.
Venom of snakes from the Elapidae and Hydrophyidae families
possesses proteins with pronounced pharmacological activities (1, 2).
Some of these proteins are potent cardiotoxins (3, 4), whereas others
are postsynaptic neurotoxins (5-7). The neurotoxins are classified
into two general groups, long and short neurotoxins (8, 9). Both
classes of toxins bind specifically to the nicotinic acetylcholine
receptor and block synaptic nerve transmission (10, 11). Binding of the
neurotoxins to the acetylcholine receptor leads to a complete closure
of the channel (11). The extremely tight, noncovalent association
between the receptor and neurotoxins (Kd ranging
from 109 to 10
11 M) in
comparison with that of acetylcholine (Kd of
10
6 M) makes them useful tools with which to
investigate the function of the neuromuscular synapse and its receptors
(10, 12). The long and short neurotoxins exhibit sequence homology and
similar overall topologies, characterized by a three stranded
antiparallel
sheet and three finger-like loops protruding from a
globular core (13). Long neurotoxins have four disulfide bridges like the short neurotoxins but possess an additional disulfide bridge in the
central loop of the molecule. In addition to insertion and deletion
within the main chain itself, long neurotoxins have an extra
polypeptide chain between residues 65 and 73 that gives rise to a
characteristic C-terminal tail (2). In long neurotoxins, the least
conserved regions tend to be found in the C-terminal tail and the first
loop (9). To date, x-ray and NMR structures of three long neurotoxins,
namely
-cobratoxin (14, 15),
-bungarotoxin (16), and LSIII from
Laticauda semifasciata (17), have been determined. The king
cobra (Ophiophagus hannah), which belongs to the elapid
family, is the world's largest poisonous snake (18). At least six long
neurotoxin isoforms have been isolated from this venom source, and
among these neurotoxin analogues, toxin b is the most toxic (18). Toxin
b is 73 amino acids long and contains 5 disulfide bridges. In this
paper, we report the solution structure of toxin b. We find that the
overall fold of toxin b is similar to that of the other long neurotoxin
structures. Consistent with previous studies, the secondary structure
is characterized by an antiparallel double and a triple stranded
sheet conformation. However, subtle but important difference(s) are
observed among the structures of
-cobratoxin (14, 15),
-bungarotoxin (16), and LSIII (17) with toxin b.
NMR Experiments
Toxin b was purified from the O. hannah (king cobra) venom (Sigma) as described previously by Chang and co-workers (18). For all the NMR experiments, 38 mg of the toxin were dissolved in 0.5 ml of water (9 mM). The samples were prepared in two different ways to attain varied conditions for the exchange of the amide protons. A nonexchanged sample with all the amide protons was prepared by dissolving lyophilized toxin b in a mixture of H2O (90%) and D2O (10%) at 20 °C. The fully exchanged sample, yielding a spectrum without labile amide protons, was obtained by dissolving the protein in D2O. The pH of the protein samples for NMR measurements was set at 3.0.
All NMR experiments were performed on a Bruker DMX-600 NMR spectrometer, and data were processed on a INDY work station using the UXNMR software. DQF-COSY1 (19), total correlation spectroscopy (50 ms mixing time) (20), and DQ-COSY (21) spectra were collected for through-bond interactions. Water-gated NOESY (22) spectrum was acquired with a mixing time of 150 ms. In all NMR experiments, sodium 3-[trimethylsilyl-2,2,3,3-2H]proponate was used as an internal standard.
Experimental Restraints
Distance RestraintsInter-proton distance restraints,
torsional angle restraints, and hydrogen bond restraints were derived
from the NMR data. Intensities of the NOESY cross-peaks were classified
into distances with three ranges, 0.18-0.27, 0.18-0.35, and
0.18-0.50 nm, corresponding to strong, medium, and weak NOE (23, 24).
Distance restraints were explicitly used for the hydrogen bonds
expected in the sheet and consistent with the observed exchange
rates and NOEs (25, 26). The hydrogen-deuterium exchange data for the
backbone amide protons revealed that there are slowly exchanging amide protons in the secondary structure region.
The
3JNH coupling constants were measured in the
DQF-COSY spectra from cross-peak multiplets parallel to the
2 axis. The dihedral angle was restricted to the range
of
90°,
40° for 3JNH
smaller than
5.5; for larger couplings with 3JNH
greater
than 10 Hz, the angle restraints were to the range of
140°,
100° (27). Based on the above criteria, 38 backbone dihedral angle
(
) restraints were included in the structure calculation. Using
analysis of the side chain coupling constants 3J
and intra-residue NOE, the
1 angle constraints were obtained (23). Stereospecific
assignments were achieved on the basis of the
3J
coupling constants and the intra- and
sequential inter-residue NOEs involving the amide, C
H,
and C
H protons (28). In all, 14 prochiral centers were
determined for the methylene protons on
carbons.
Three-dimensional Structure Calculations
The three-dimensional structures were calculated from the distance and angle constraints with a combination of the distance geometry and simulated annealing protocols using X-PLOR (29, 30). The target function contains several terms, one for bond lengths and bond angles, an improper torsional angular term, square well potential terms for NOE and restraints of torsional angles, and a soft repulsive term instead of van der Waals and electrostatic potentials (26). 50 initial structures were generated from a preliminary set of distance and angle constraints. The convergence of these initial structures was reasonable. The selection of the structures was based on the conditions that no distance constraints (per structure) exceeded 0.02 nm and that the NOE and total energies were minimal. The structure calculation is based on 485 distance constraints (158 long range, 50 medium range, and 101 sequential constraints, 36 for the backbone hydrogen bonding and 5 distance constraints for the disulfide bridges). Another 52 constraints of the dihedral angles were also included. The structures obtained upon distance geometry protocol were further refined by simulated annealing calculations using X-PLOR (30). To improve the ill-behaved structures, simulated annealing refinements were carried out on the energy-minimized structures. The refinement was based on satisfaction of each coordinate set in the family of generated structures. This is a slow cooling protocol with an additional feature of "softening" for van der Waals repulsion to enable atoms to move through each other. This routine consists of a 9-ps cooling dynamics followed by 200 cycles of Powell minimization. QUANTA (Molecular Simulations, Inc.) was used to generate, display, and analyze the structures.
Almost all the protons in the protein
have been assigned. The sequential resonance assignments were conducted
following a well established method (31, 32). The DQ-COSY recorded in H2O was of great value for unambiguous identification of
glycine spin systems, since a remote peak is observed at
2 =
NH,
1 =
+
1 (21). These peaks were located
in a specific well resolved region, and thus were easily identified.
Identification of the
H
(i)-HN(i)
or
H
(i)-HN(i+1)
cross-peaks in the DQF-COSY and the NOESY spectra (mixing period, 150 ms) allowed for the sequential assignments in toxin b (Fig.
1). For most of the residues, the sequential NOE could
be easily recognized, and in most cases, additional correlations with
the protons of the side chain were also observed. According to the
patterns of sequential NOEs observed for prolines, Pro7,
Pro15, Pro65, and Pro67 have been
found to adopt the trans conformation.
The protons of residues Asp16, Asp19,
Cys21, Asp28, Phe30,
Cys31, Ser32, Asp39,
Cys42, Cys57, Cys58,
Ser59, and Asn62 were stereochemically assigned
following the criteria based on the values of the
3J
coupling constants and intra-residue
NOEs (Fig. 2) (28, 29). The secondary structure elements
in the protein were identified from the sequential and long range NOEs
(C
H-C
H, NH-C
H, and NH-NH).
Identification of the NOE between
protons of the backbone using the
main chain-directed approach (33) also confirms the location of
antiparallel double and triple stranded
sheets in the toxin (Fig.
3). The amide protons of most of the residues involved
in the secondary structure formation are found to be protected from
deuterium exchange. Residues Cys27, Asp28,
Gly29, and Phe30 form a type II turn. The other
turns in the protein are not well characterized by the NOE constraints
and do not fit into the classical types. In addition, medium range NOEs
are observed between the C
H of Cys27 and the
NH of Gly35 and between the C
H of
Cys31 and the NH of Arg34. These results
suggest the presence of some local structure at the tip of Loop II.
Similar types of NOEs were observed in the same region in
-cobratoxin and have been attributed to the presence of
helix
(15). Interestingly, the exchange data show that the amide protons of
Phe30, Cys31, and Lys36 are
protected from exchange (Fig. 2). This result gives an additional clue
to the existence of local structure in this segment (residues 26-34)
located at the tip of Loop II in toxin b. However, based on our NOE
data we were unable to observe the helix conformation in this segment
of toxin b. The NOE connectivity pattern in this segment does not
satisfy all the criteria required for a typical helix.
Structure Determination
For the final calculation, 21 distance geometry structures were selected on the basis of their final
error in the distance geometry calculations. These structures were
calculated based on a total of 485 distance constraints (Fig.
4). The 21 structures were further refined with the
simulated annealing protocol. The resultant structures were overlapped
and are shown in Fig. 5. The energetic and geometric
statistics of these refined structures are shown in Table
I. All the structures are in good agreement with the
experimental restraints, with no structure having a NOE restraint
violation(s) greater than 0.2 Å or a dihedral angle restraint
violation(s) greater than 4° (Fig. 5).
|
The Ramachandran plot for the average of 21 structures showed that the
backbone dihedral angles of nearly all residues in the secondary
structure segment(s) lie in the allowed regions (Fig.
6). Those residues that have unfavorable -
angles
(backbone dihedral angles) were in poorly defined regions. Minimization of individual structures using CHARMm (33) gave negative Lennard-Jones energies (
291 ± 30 kcal/mol) with negligible shifts in atomic positions (backbone root mean square deviation < 0.15 Å),
indicating favorable nonbonded interaction.
The three-dimensional structure of
toxin b consists of three hairpin-type loops emerging from a globular
head. Among the three loops, Loop I (Fig. 5, residues 1-14, left
loop) and Loop III (residues 44-59, right loop) are
shorter. Loop II (middle loop), comprising residues 19-43,
is the longest. The sequence of residues 65-73 of the C-terminal end
constitutes the tail of the molecule. The three loops are tethered
together by four disulfide bridges, namely
Cys3-Cys21,
Cys14-Cys42,
Cys46-Cys57, and
Cys58-Cys63. The fifth disulfide bond from
Cys27-Cys31 is located at the lower tip of Loop
II. Loop I involves two strands (strands 1 and 2) of the double
stranded sheet. The antiparallel double stranded
sheet in toxin
b is very short, comprising residues 3-4 (strand 1) and 12-13 (strand
2). The double stranded
sheet are linked by a segment consisting of
residues 4-11 (Fig. 5). It is internally stabilized by hydrogen
bonding between residues 13 and 3. Loop I is partly hydrophobic
(residues 2-5) and partly exposed to the solvent (residues 8-14).
Loop II is the largest loop in the tertiary structure of all long
neurotoxins; it consists of a narrow hairpin stem with a bulky tip that
is stabilized in its structure by the disulfide bridge
Cys27-Cys31. The stem portion lodges a part of
the triple stranded antiparallel
sheet spanning residues 20-25 (in
strand 3) and residues 38-43 in strand 4 (Fig. 5). Although the tip of
Loop II appears mostly disordered, with high root mean square deviation
values, a closer examination reveals some local order. The overlapped
segment that constitutes the tip region (residues 26-34) showed a
distorted right handed turn (Fig. 7). Such local
structures have also been identified in the solution structures of both
-cobratoxin (15) and LSIII (17). The local structure in
-cobratoxin, however, has been designated as a "fluctuating"
helix (15). The structure in this region was found to be
conformationally variable. The mobility is believed to be caused by the
shortness of the helix (5 residues) and the high accessibility of the
region to the solvent in
-cobratoxin. However, we believe that the
nonavailability of the long range constraints could also account for
the flexibility in this region of toxin b. Loop III of the molecule is
stabilized by both the disulfide bridge between Cys46 and
Cys57 and the type II
turn formed by
Lys50-Pro51-Gly52-Val53.
A portion of strand 5 that constitutes Loop III is involved in
sheet formation with Loop II and constitutes the triple strand (Fig.
5). The tail at the C-terminal end is connected with Loop III by the
disulfide bond between Cys58 and Cys63. The
C-terminal end of the molecule is found to be parallel to Loop II due
to the formation of a type II
turn among residues 60-63. The
C-terminal end tail comprising residues 65-73 is poorly defined by the
available experimental data. There are very few long range constraints
in this part of the molecule. The hydrophobic core of toxin b is well
defined and is provided by residues that cluster around
Tyr22. The residues Lys24, Leu40,
Ala43, and Ala44 form a sort of cylinder from
which the hydroxyl group of Tyr22 sticks out. In addition,
Tyr4 along with Asn62, Cys63, and
Asn64 lie close to the hydrophobic patch formed around
Tyr22.
Comparison with Other Long Neurotoxins
Toxin b shows
reasonably high sequence homology with -bungarotoxin (16),
-cobratoxin (15), and LSIII (17). Comparison of the solution
structures of toxin b and the other long neurotoxins shows that the
overall fold in all the toxins is similar (Fig. 8). The
characteristic feature of all the long neurotoxins is the high
flexibility of the tip portion of Loop II, as indicated by large root
mean square deviations. The high flexibility could have a significant
effect(s) on the thermodynamics and kinetics of ligand receptor binding
of the long neurotoxins. The rate of association of the toxin to the
acetylcholine receptor could increase, since the high flexibility of
the loop could favor complex formation with the acetylcholine receptor
by lowering the free energy barrier. The length of Loop II is the
shortest in toxin b compared with that in
-cobratoxin,
-bungarotoxin, or LSIII (16-18). The length of this loop is
believed to be intricately connected with the binding affinity of the
neurotoxins to receptor (2). The longer the length of Loop II in the
toxin, the greater is the affinity to the receptor (9). The structure
of toxin b is strikingly different from that of the other long
neurotoxins at the tip of Loop II. In both
-bungarotoxin (16) and
-cobratoxin (15) there is a distinct short nascent helix formed by
the residues in segment 29-35. In toxin b, although there is local
structure in the same region of Loop II, it could not be ascribed to a
helical conformation (Fig. 7). Comparison of the solution structures of
-bungarotoxin,
-cobratoxin, LSIII, and toxin b reveals that except for the residues involved in the
sheet conformation, most
portions of these long neurotoxin molecules are flexible (Fig. 8). The
C-terminal tail region is poorly defined in all the long neurotoxins
whose structures in solution have been studied. However, excision of
the C-terminal tail by treatment with trypsin (34) and carboxypeptidase
P (35) has been shown to have no effect on the structure of this
toxin.
Possible Binding Region(s) of Neurotoxins to the Acetylcholine Receptor
Many residues located at the tip of Loop II are either
strictly conserved or only conservatively substituted among short and long neurotoxins. Chemical modification studies have clearly
demonstrated that many of the well conserved residues in Loop II have
an important role in the toxicity of the long neurotoxins (9, 11, 12). Trp27 is conserved in all long and short neurotoxins (36).
The indole ring of this conserved tryptophan is exposed to the solvent.
Modification of Trp27 in a long neurotoxin analogue (Oh-4)
from the king cobra venom showed a significant loss in the toxicity and
acetylcholine binding efficiency of the toxin (37). Lys24
is yet another residue that is well conserved among the neurotoxins (5,
6). The neurotoxins are believed to interact with acetylcholine receptor through charge-charge interaction (11). It is postulated that
the role of Trp26 is to orient the side chains of
Lys24 and Asp28 residues at the tip of Loop II
for effective binding to the acetylcholine receptor (38). In addition
to Lys24, the positive charge contributed by
Arg34 is also believed to be important for the receptor
binding (39). All the conserved residues in toxin b form a cluster at
the tip of Loop II and thus may have a critical role in binding to the acetylcholine receptor. The disulfide bond holding Loop II has also
been shown to be functionally important (40). Presumably, the cleavage
of this disulfide bond disrupts the positively charged cluster at the
tip of Loop II. In -cobratoxin, the mobile lysine residue at
position 49 has been postulated to have a role in the receptor binding
(42). In erabutoxin b, it has been proposed that Loop II and Loop III
together build a concave surface and that formation of such a surface
promotes facile binding of the toxin to the receptor (5). Formation of
such a concave surface(s) can also be seen in the GRASP representation
of toxin b (Fig. 9). Neurotoxins are believed to bind to
the acetylcholine receptor through multiple sites. In toxin a and
-bungarotoxin, the positively charged cluster formed by residues in
Loop II (Arg34, Arg37), the C-terminal tail
(Arg71, Lys70, Lys72), and Loop III
(Lys50, Lys56) is shown to be crucial for the
receptor binding (41, 42). The positively charged cluster can also be
visualized in the structure of toxin b (Fig. 9). Chemical modification
of all the arginine and lysine residues belonging to the positively
charged cluster leads to substantial loss in the antigenicity and
receptor binding ability of the long neurotoxins (43-45). These
results imply that the cationic centers located in Loops II and III and
the C-terminal tail (Fig. 9) collectively contribute to the multipoint
contact of the toxin to the acetylcholine receptor. The solution
structure of toxin b thus reveals most of the structural features
previously observed for the family of long neurotoxins. The tip of Loop
II shows a prominent local order and, together with the cluster of positive charges formed by residues in the C terminus, Loop II, and
Loop III, could constitute the active site involved in the acetylcholine receptor binding.
The atomic coordinates and structure factors (codes 1TXA and 1TXB) have been deposited in the Protein Data Bank, Brookhaven National Laboratory, Upton, NY.
We acknowledge the Regional Instrumentation Center (Hsinchu, Taiwan) for allowing us to use the 600-MHz spectrometer.