(Received for publication, November 15, 1996, and in revised form, February 6, 1997)
From the The structure-function relationships of the
N-type calcium channel blocker, The fish-hunting marine cone snails produce a range of polypeptide
toxins that rapidly immobilize their prey (1, 2). A number of such
toxins targeted at ion channels have been isolated from the venoms of
these cone shells, including several that selectively block N- and
P-type calcium channels (3). The toxin studied here, The amino acid sequence of GVIA (with the cystine bridges indicated by
lines) is as show below.
Department of Pharmacology,
ABSTRACT
INTRODUCTION
EXPERIMENTAL PROCEDURES
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES
-conotoxin GVIA (GVIA), have been
elucidated by structural, binding and in vitro and in
vivo functional studies of alanine-substituted analogues of the
native molecule. Alanine was substituted at all non-bridging positions
in the sequence. In most cases the structure of the analogues in
aqueous solution was shown to be native-like by 1H NMR
spectroscopy. Minor conformational changes observed in some cases were
characterized by two-dimensional NMR. Replacement of Lys2
and Tyr13 with Ala caused reductions in potency of more
than 2 orders of magnitude in three functional assays (sympathetic
nerve stimulation of rat isolated vas deferens, right atrium and
mesenteric artery) and a rat brain membrane binding assay. Replacement
of several other residues with Ala (particularly Arg17,
Tyr22 and Lys24) resulted in significant
reductions in potency (<100-fold) in the functional assays, but not
the binding assay. The potencies of the analogues were strongly
correlated between the different functional assays but not between the
functional assays and the binding assay. Thus, the physiologically
relevant assays employed in this study have shown that the high
affinity of GVIA for the N-type calcium channel is the result of
interactions between the channel binding site and the toxin at more
sites than the previously identified Lys2 and
Tyr13.
-conotoxin GVIA
(GVIA),1 is a 27-residue polypeptide from
Conus geographus (4) that selectively blocks N-type
voltage-gated calcium channels (5, 6), an activity that may confer a
number of useful therapeutic properties on GVIA, including
antihypertensive, analgesic, and neuroprotective activities, as
demonstrated for the closely related
-conotoxin MVIIA (MVIIA) from
Conus magus (7) .
O indicates hydroxyproline. The three-dimensional structure of
GVIA in solution has been solved by a number of groups (8-11). The
molecule is a member of the inhibitor cystine knot structural family
(12, 13), the members of which share a triple-stranded
-sheet and a
three-disulfide bridge cystine knot. The structure of the homologous
N-type calcium channel blocker MVIIA is essentially identical to that
of GVIA (14, 15), and the structure of the closely related P-type
calcium channel blocker from C. magus,
-conotoxin MVIIC
(MVIIC), is very similar (16, 17) .
A number of studies of GVIA have been carried out in an attempt to identify those residues that are essential for its high affinity interaction with the N-type calcium channel. The role of the ammonium groups was studied by selective acetylation, with the N terminus being found to be important for binding to rat hippocampal membranes and chick brain synaptosomes (18). Studies of GVIA biotinylated at Lys2 and/or Lys24 demonstrated that Lys2 was more important than Lys24 for binding to rat brain synaptic membranes (19). Studies of alanine substitutions at the basic residues (20) indicated that Lys2 was the only basic residue of importance for binding to chick brain synaptic membranes.
The roles of several structural residues in GVIA, including
Gly5, the half-cystine residues, and the N- and C-terminal
-strands, have also been investigated. An active analogue was
produced with the substitution of Gly5 with
D-Ala, implying that this residue was probably in a type II
-turn (21), as subsequently confirmed in the three-dimensional structures. Full-length analogues with single cystine bridges deleted
(21) or fragments with single disulfide bonds joining discontinuous
segments of the polypeptide chain (22) were found to be inactive in a
rat brain membrane binding assay. We reported recently that major
alterations to the topology of the molecule through deletion of
specific disulfide bridges or deletion of N-terminal or C-terminal
residues resulted in analogues that were inactive on the N-type calcium
channel in the rat isolated vas deferens (23).
To gain a more detailed understanding of the roles of individual amino acid residues in the biological activity of GVIA, we have undertaken an "alanine scan" of the molecule, in which each residue except the bridging half-cystine residues has been replaced individually by alanine. While this work was in progress, a chick brain binding study on a series of alanine-substituted analogues of GVIA was published by Kim et al. (24), in which only residues Lys2 and Tyr13 were found to be important for binding. That study also demonstrated that the post-translational modification from Pro to Hyp (hydroxyproline), for all three Hyp residues simultaneously, was unnecessary for activity. A similar investigation of MVIIA (25) yielded the result that Arg10, Leu11, and Arg21 were important for rat brain synaptosomal membrane binding, in addition to the conserved Lys2 and Tyr13 residues. The present study of GVIA differs from that of Kim et al. in several respects. First, the alanine-substituted analogues included separate substitutions of Hyp4, Hyp10 and Hyp21 with Ala. Second, the structures of the analogues were analyzed rigorously by one and two-dimensional 1H NMR spectroscopy to determine the extent of isomer purity and to assess conformational integrity. Third, in addition to binding studies, activity studies were performed with three functional organ bath assays (rat isolated vas deferens, right atrium, and small mesenteric artery), as well as in intact conscious rabbits. Our data provide a much more complete picture of the calcium channel binding surface of GVIA and point to the possibility of differences in subtype populations of the N-type channels in different tissues, and to differences between the epitopes important in binding and function.
N--Protected
amino acids were obtained from Auspep Pty. Ltd. (Melbourne, Australia),
Bachem, or Millipore (Sydney, Australia) and were of the
L-configuration except for Gly.
N-(9-Fluorenyl)methoxycarbonyl (Fmoc) amino acids included:
Cys (trityl), Arg (2,2,5,7,8-pentamethylchroman-6-sulfonyl), Ser
(t-butyl), Lys (t-butyloxycarbonyl), Tyr
(t-butyl), Thr (t-butyl), Hyp
(t-butyl), and Asn (trityl). All reagents and solvents
(peptide synthesis grade) were purchased from Auspep Pty. Ltd. The
polyethylene glycol-polystyrene resin was purchased from Millipore.
Peptide synthesis was performed using a semi-automatic NovaBiochem Gem peptide synthesizer at 0.13-0.2 mmol scale of polyethylene glycol-polystyrene resin. Fmoc amino acids (3 eq) were activated using 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate (3 eq) with diisopropylethylamine (6 eq) used to initiate the in situ reaction. Peptide resins were cleaved, and all protecting groups were removed, typically by treatment with a solution containing trifluoroacetic acid (10 ml), ethane-1,2-dithiol (0.25 ml), thioanisole (0.5 ml), H2O (0.5 ml), and phenol (0.5 g) (26). After 4-6 h, resins were removed by filtration and the crude peptide was precipitated with diethylether, recovered by centrifugation, washed three times with diethyl ether, dried in vacuo, dissolved in 20% MeCN (aq), 0.1% trifluoroacetic acid, and lyophilized.
Crude peptides were oxidized by dissolution in 10% dimethyl sulfoxide in ammonium bicarbonate (50 mM, pH 8) at 1 mg/ml (27). These oxidations were monitored by analytical reversed-phase HPLC.
HPLC was performed on a Hewlett Packard 1050 chromatograph with detection by UV absorption at 220 nm. Preparative HPLC was conducted on an Alltech Econosil ODS column (250 mm × 22 mm, 10-µm particle size), semi-preparative chromatography was performed on a Phenomenex Spherisorb ODS column (250 mm × 10 mm, 10-µm particle size), and analytical chromatography was conducted on a Hypersil ODS column (250 mm × 2.1 mm, 5-µm particle size) or on an Adsorbosphere ODS column (150 mm × 2.1 mm, 5-µm particle size). Peptides were eluted using either 0.1% trifluoroacetic acid in H2O (buffer A) and 0.1% trifluoroacetic acid in acetonitrile (MeCN) (buffer B) or 0.6% triethylammonium phosphate at pH 2.25 (TEAP 2.25) (buffer A) and TEAP in MeCN (buffer B). Flow rates were 0.3 ml/min for analytical HPLC, 4 ml/min for semi-preparative HPLC, and 10 ml/min for preparative HPLC. Hereafter, these mobile phase systems will be abbreviated to 0.1% trifluoroacetic acid H2O/MeCN and TEAP 2.25/MeCN.
The crude (oxidized) peptides were loaded batchwise on to the preparative column, and eluted with a linear gradient of TEAP 2.25/MeCN. The gradient applied to the preparative column was 3-16% over 46 min. The generated fractions were analyzed by analytical HPLC using gradient elution with 0.1% trifluoroacetic acid H2O/MeCN. Appropriate fractions (>95% pure) were pooled, diluted and loaded onto a 250 mm × 10-mm semi-preparative column for desalting. The peptide was eluted with a linear gradient of 0.1% trifluoroacetic acid H2O/MeCN. Fractions containing the purified peptide were pooled and lyophilized.
Matrix-assisted laser desorption ionization time-of-flight mass
spectrometry was used for confirmation of product identity and was
performed on a Finnigan Lasermat. The absorbing matrix used was
-cyano-4-hydroxy-cinnamic acid. Capillary zone electrophoresis was
used for confirmation of product homogeneity and was conducted on an
Applied Biosystems model 270A instrument, using the following conditions: applied potential, 30 kV; buffer, sodium citrate (20 mM, pH 2.5); capillary length, 72 cm; capillary diameter,
50 µm; temperature, 30 °C.
All polypeptide analogues were analyzed by 1H NMR spectroscopy on a Bruker AMX-600 spectrometer. Two-dimensional spectra were acquired on 2-10-mg samples of polypeptide in 90% H2O, 10% 2H2O at pH 3.4 and 298 K (400 t1 increments, 4096 data points, and 64-96 scans/t1 increment were employed). The TOCSY spin-lock and NOESY mixing times were 70 and 400 ms, respectively. The methodology was otherwise as described by Pallaghy et al. (28).
NOESY spectra of analogues were analyzed in terms of which non-trivial
(medium and long range) NOEs of the native molecule were lost in the
analogues and which, if any, new NOEs were observed that were not seen
in the native spectrum. The intensities of NOEs that were present in
both spectra were not monitored. A custom-written computer program,
NMRanalogue, was written to produce macros (suitable for input to the
spectral analysis programs Felix 2.3 and Felix 95) that allowed this
procedure to be partially automated for the series of NOESY spectra.
The lost NOEs were categorized as very weak, weak, medium, strong, or
very strong according to their intensities in NOESY spectra of the
native molecule. Only peaks of intensity "weak" or greater for
analogues of estimated sample quantity 7 mg were considered
significant. The native sample quantity was approximately 10 mg, and
only well shaped peaks were used in the native NOE
list.2 The designation of a native NOE as
"lost" for any given analogue was performed conservatively, so that
even a very distorted peak near the required position was considered to
be maintained. Insight II (version 95.0) was used for molecular
graphics.
Crude rat brain membranes were prepared by
the method of Cruz and Olivera (29). Individual brains were homogenized
using an Ultra-Turrax homogenizer (Janke and Kunkel) in 10 volumes of buffer (0.32 M sucrose, 5 mM HEPES-Tris, pH
7.4, 0.1 mM phenylmethylsulfonyl fluoride) and the
homogenate spun at 1000 × g for 10 min at 4 °C. The
pellet was resuspended in another 10 volumes of buffer and spun again.
The combined supernatants were then spun at 17,000 × g
for 60 min at 4 °C. The pellet was taken up in 100 volumes of buffer
as above but also containing 50 mg/liter lysozyme. This gives membrane
derived from about 200 µg of tissue/20 µl of homogenate. The
suspension was divided into 3-ml aliquots and stored at 70 °C.
Binding assays were conducted in 96-well microtiter plates with
0.65-µm filters in the bottom (Millipore multiscreen system), using a
modification of the method of Cruz and Olivera (29) and Haack et
al. (30), as also used by Kim et al.
(24).3 The binding buffer contained 0.32 M sucrose, 5 mM HEPES-NaOH (pH 7.4), 0.1 mM phenylmethylsulfonyl fluoride, 0.3% bovine serum albumin, and 50 mg/liter lysozyme. Wash buffer consisted of 150 mM NaCl, 5 mM HEPES-NaOH, pH 7.4, 1.5 mM CaCl2, and 0.1% bovine serum albumin.
Tracer solution was prepared as follows; 50 µCi of
125I-[Tyr22]GVIA (DuPont NEN) was dissolved
in 1 ml of water and divided into 50-µl aliquots, which were stored
at 15 °C. Tracer for each assay was prepared by diluting one
aliquot (2.5 µCi) into 3.5 ml of binding buffer for each 96-well
microtiter plate to be used in the assay. Incubation mixtures (total
volume 100 µl/well) consisted of 50 µl of binding buffer,
containing displacing ligand, 20 µl of membrane preparation, and 30 µl of diluted 125I-[Tyr22]GVIA tracer
solution, final concentration 0.13 nM. Nonspecific binding
was determined in the presence of 1 µM unlabeled GVIA. Assays were incubated at 4 °C for 90 min and were terminated by filtration under vacuum. Each well was washed three times with 200 µl
of wash buffer and then left under vacuum for 10 min to dry the
filters. Filters were punched out and counted in a
counter. Each
measurement was determined in quadruplicate within an experiment, and
each experiment was replicated at least three times. Data were analyzed
using the computer program GraphPad Prism.
Male Harlan Sprague Dawley rats
were killed by CO2 anesthesia followed by decapitation, and
the vasa deferentia, heart, and a portion of intestine with attached
mesentery removed and placed in a dish of cool Krebs solution
(composition in mM: Na+ 144, K+
5.9, Mg2+ 1.2, Ca2+ 2.5, HPO4 1.2, Cl
129, SO4
1.2, HCO3
25, glucose 11, EDTA 0.026, bubbled with 5% CO2 in oxygen) for further dissection.
Vasa deferentia were mounted in 5-ml organ baths, with the top of each tissue attached to an isometric force transducer (Grass FT03) and the bottom attached to a movable support and straddled with platinum stimulating electrodes. The vasa were stretched by a passive force of about 10 millinewtons and stimulated with single electrical field pulses (100 V, 0.2-ms duration) every 20 s. The resulting twitch responses were mediated by sympathetic nerves, being sensitive to inhibition by guanethidine (10 µM) or tetrodotoxin (0.1 µM), and were recorded on a chart recorder. GVIA (0.3-10 nM) caused a gradual concentration-related reduction in the size of the twitch response to electrical nerve stimulation in the rat vas deferens. Submaximally effective concentrations of GVIA elicited an initially rapid fall in the size of the twitch, and a continued very gradual decrease thereafter for over 90 min. This apparently slow equilibration meant that concentration-response curves with full equilibration of the toxin could not be constructed without interference from spontaneous fade of the twitch responses. To offset this problem, a protocol with fixed 20-min intervals between successive concentration increments was used. Cumulative concentration-response curves for GVIA or the analogues were constructed by addition of the peptides to the solution bathing the vas deferens in 10-fold concentration increments from 1 nM. A single concentration-response curve was constructed in each tissue.
Mesenteric ArteryUnder a dissecting microscope, an artery
(three branch orders proximal to the arteries that enter the intestine)
was carefully dissected free of the fat and connective tissue around
it, and a 2-mm-long segment was mounted on 40-µm wires in a
Mulvany-Halpern style isometric myograph and warmed to 37 °C. The
artery was incrementally stretched radially with about four steps, and
the force measured and the arterial circumference calculated at each
step, producing a diameter-force curve where the diameter is that of a
circle with the same circumference as the vessel at each level of
stretch. The diameter of the artery was then set to be 90% of the
diameter predicted for distending pressure of 100 mm Hg using standard calculations (31). The rat mesenteric artery set up under these conditions does not develop any spontaneous active contractile force.
Potassium depolarizing solution was applied for about 2 min to
maximally activate the artery, and washed out. The prejunctional 2-adrenoreceptors were blocked with the covalent
antagonist benextramine (3 µM for 5 min) in the presence
of prazosin (0.1 µM) to protect the postjunctional
1-adrenoreceptors. The antagonists were then washed out,
and noradrenaline (10 µM) was applied to confirm the washout of the antagonists. This procedure greatly increases the size
of the responses to sympathetic nerve stimulation. Each jaw of the
myograph was fitted with a platinum electrode about 1 mm away from the
artery. Sympathetic nerves in the wall of the artery were stimulated
with monopolar electrical field pulses of 0.25-ms duration at 30 V,
stimulation parameters that give responses that are blocked >95% by
tetrodotoxin. We have previously demonstrated that the responses are
abolished by guanethidine and are thus mediated by sympathetic nerves.
Stimulation of the sympathetic nerves produced contractile responses
that had a short (<1 s) latency and decayed rapidly after the end of
the train of stimuli. All responses were measured as the peak change in
force. Nerve stimulation was applied as sets of three trains of 75 pulses at 25 Hz with a 60-s interval between trains, and 30 min between successive sets of stimuli. GVIA or analogues were applied in a
cumulative fashion with 30 min of contact before each test
stimulation.
The right atrium was dissected free of the heart and placed in a 5-ml organ bath at 37 °C on a support having two fine platinum electrodes in contact with the atrium to collect the surface electrogram and another pair of electrodes for electrical stimulation. The spontaneous rate of contraction was continuously measured using the surface electrogram to trigger a period meter. Tachycardia responses mediated by sympathetic nerves were measured in response to sets of 4 electrical field pulses at 2 Hz, in the presence of atropine (1 µM) to abolish the effect of parasympathetic nerve stimulation. Increasing concentrations of GVIA or analogue were applied immediately following the second of two control stimulations, with the drug in contact with the tissue for 30 min before each stimulation.
Conscious Rabbit ExperimentsThe central ear artery and marginal ear vein of New Zealand White rabbits of either sex (weight 2.5 ± 0.1 kg) were cannulated under local anesthesia (1% lignocaine hydrochloride) for measurement of blood pressure (MAP) and for drug injections, respectively. The phasic blood pressure signal triggered a rate meter for the measurement of heart rate (HR). Phasic and mean arterial pressure and HR were recorded on a Grass polygraph. Following the minor procedures as outlined above, rabbits rested quietly for about 40 min in polycarbonate restrainers. The effects of intravenous administration of selected GVIA analogues were assessed on MAP, HR, and the baroreflex. The baroreflex was measured by eliciting alternate graded steady-state increases and decreases in MAP (± 5-35 mm Hg from base line) with phenylephrine and sodium nitroprusside, respectively (32). GVIA analogue potencies were assessed by comparing their effects to the effect of 10 µg/kg GVIA. The GVIA analogues were administered at an initial dose of 3 or 10 µg/kg. MAP and HR were monitored for 60 min and the dose of analogue increased (cumulative half-log10 dose increments) if its effect was less than that of 10 µg/kg intravenous GVIA (33, 34). The baroreflex curve was then reassessed 60 min after the highest dose of peptide was administered and the reflex parameters (gain, location, and plateaus) obtained by fitting the baroreceptor-heart rate reflex curve to a logistic equation (32).
Synthesis of all linear
peptides proceeded smoothly, as indicated by monitoring of the Fmoc
deprotection peak at each cycle. Oxidative refolding of linear peptide
analogues gave a complex mixture of products, but the structures
corresponding to the native fold were readily separated from non-native
structures by the TEAP 2.25/MeCN chromatography system (35) (Fig.
1A). Refolding of most of the linear peptide
analogues gave purified yields of cyclized peptides of approximately
5% of the crude material.
On the basis of HPLC analysis, only three of the oxidized analogues
(G5A, N20A, and T23A) failed to show a major product with native fold
that could be purified using the TEAP 2.25/MeCN chromatography system
(Fig. 1, B-D). The amount of material with native-like structure in the oxidized mixtures of G5A, N20A, and T23A was therefore
estimated by 1H NMR spectroscopy. The NMR spectrum of
native GVIA contains five backbone amide proton resonances in the
region 9.2 and 9.5 ppm (Fig. 2A), which were
also present in correctly folded Ala-containing analogues
(e.g. Fig. 2B) and thus served as a convenient
marker of the native fold. Disulfide isomers or other incorrectly
folded forms of the molecule could be distinguished from the
native-like fold on the basis of these resonances (9). Thus, in cases
where the purification scheme described above did not yield a
significant quantity of the desired isomer, one-dimensional
1H NMR spectra were employed to estimate the proportion of
"native-like" fold in the oxidized mixture, as illustrated for T23A
in Fig. 2C. The estimate was based on the average height of
the five downfield amide resonances indicative of native-like
structure, with the spectrometer calibrated using a sample of native
GVIA. By this means the contents of correctly-folded product in G5A,
N20A, and T23A were calculated to be, respectively, 0.1, 1, and 3% of
the oxidized peptide mixture (i.e. <0.15% of the crude
material). Substitution of D-Ala at position 5, instead of
L-Ala, resulted in a normal refolding pattern to give a
readily-isolated component corresponding to the native fold of the
peptide.
Conformational Analysis
Sequential assignments were derived
from two-dimensional TOCSY and NOESY NMR spectra for all of the
analogues except G5A, N20A, and T23A. The chemical shift deviations
from native were small enough to suggest that the native fold was
maintained for all of these analogues. The average of the magnitude of
the chemical shift deviations from native, ||, provides a
convenient single-parameter characterization of the structure and is
tabulated in Table I for the NH and C
H
resonances. The deviations of the C
H resonances were
typically half the size of the NH deviations and so were less sensitive
to conformational changes. The largest deviations occurred for the K2A,
O4A, G5dA, O10A, S12A, Y13A, R17A, O21A, and Y27A analogues.
|
Most Ala substitutions caused chemical shift changes primarily in
groups spatially close to the substitution, as typified by S3A in Fig.
3A. More widespread chemical shift changes occurred in the
K2A, S6A, S12A, Y13A, and R17A analogues and in all three Hyp Ala
analogues, O4A, O10A, and O21A. The changes in the Hyp
Ala
analogues were spread across the whole molecule (Fig.
3B) and, in the absence of significant NOE
changes (see below), can be ascribed to a small conformational
rearrangement throughout the molecule due to a different preferred
local conformation of the substituted non-cyclized amino acid.
In the K2A analogue, there appeared to be a more specific conformational change associated with Ala substitution, with chemical shift perturbations evident for resonances from Ser6 and Asn14 (normally in contact with Lys2), and more interestingly, Ser9, Ser12, and Tyr13 (Fig. 3C). The Ser12 and Tyr13 perturbations could reflect a local structural change transmitted via Asn14, but there were no NOEs in the native spectra between Ser9 and Lys2, Ser6, or Asn14 to account for the perturbation of the chemical shifts of Ser9.
NOE intensities, due to their r6 distance
dependence, are not as sensitive as chemical shifts to long distance
effects, and this is borne out by the small number of non-trivial NOEs
lost or gained in the various analogues, as summarized in Table I. A
total of 14 very strong, 26 strong, 51 medium, 57 weak, and 39 very
weak NOEs between non-adjacent residues in the amino acid sequence
(i-j > 1) defines the structure of native
GVIA.2 Of these, none of the 40 strong or very strong NOEs
was lost in any of the folded analogues, confirming the conclusion
based on chemical shift deviations that the tertiary fold was
maintained in all cases.
In K2A the long range effect evident from the chemical shifts was
clarified by the NOE data and was the clearest example of a small
conformational change among all of the correctly folded analogues. New
NOEs were observed between Ala2NH and
Asn14NH and between
Ala2C
H and
Asn14N
H (without corresponding native
Lys2/Asn14 NOEs) and the NOE between
Cys8 C
H and Ser12
C
H was lost. Visual inspection of the native structure
(Fig. 4) indicated that Asn14 had probably
moved relative to Lys2, Cys8, and
Ser12, pushing Cys8 and Ser12
further apart (and causing the loss of that NOE).
Another long range effect occurred in the S6A analogue, in which the
chemical shift of Cys16NH was perturbed, even though that
of Lys2, which is located between these two residues,
changed very little. This chemical shift change correlated with a new
NOE between Hyp4CH and
Cys16C
H, and it is likely that a small
conformational change in turn I (Ser3-Ser6) was
responsible for this change. In S12A there was a long range chemical
shift change of Cys19NH, transmitted across the gap between
-turns II (Ser9-Ser12) and IV
(Hyp21-Lys24), which have only two NOE
interactions between them (both involving Hyp10 and
Cys19). However, Ser12 and Cys19
are quite close in the structure (Fig. 4), even though there are no
NOEs between them. NOEs lost in the S12A analogue
(Ser12NH-Cys8C
H and
Ser12NH-Asn14C
H) were confined
to the site of substitution, but there was a new NOE
(Tyr22NH-Arg17C
H) in turn IV
near Cys19, suggesting that the long range chemical shift
effect on Cys19NH may be due to a small conformational
change.
In Y13A the chemical shift of Cys19 moved downfield by 0.24 ppm. This long distance effect might be due to loss of the ring-current effects in this analogue. The only NOE lost was near the substituted residue (Ser12NH-Asn14CH), but
there was a new NOE
(Ala13H
-Cys19NH) between the new
Ala residue and Cys19NH (which had a perturbed chemical
shift). This NOE may have been boosted in intensity due to the
substitution of the
-methylene of tyrosine with
-methyl of
alanine, and may not necessarily be indicative of a genuine
conformational change.
Specific binding of
125I-Tyr22-GVIA to rat brain membrane
constituted 65-75% of the total binding. Nonlinear regression
analysis indicated a single binding site with an apparent
pKD of 8.95 ± 0.11 and
Bmax of 43.7 ± 2.6 fmol/mg tissue.
Unlabeled GVIA displaced the labeled tracer with pIC50 of
9.7 ± 0.1. The binding affinities of the Ala-substituted
analogues are shown in Fig. 5D.
GVIA inhibited the twitch response to single-pulse stimulation of the sympathetic nerves of the vas deferens with a pEC50 of 8.9 ± 0.05, and inhibited the responses of the mesenteric artery and atria to trains of sympathetic nerve stimuli with pEC50 values of 8.8 ± 0.06 and 9.0 ± 0.09, respectively. In the vas deferens and atria, GVIA completely inhibited the responses to the stimulation conditions used. However, in the artery the maximal inhibition by GVIA was only about 90%, and the residual response was sensitive to tetrodotoxin (0.1 µM). Thus, there was a component of the response of the artery to sympathetic nerve stimulation that was resistant to GVIA in the artery but not the other assays. GVIA-resistant sympathetic neurotransmission mediated by P-type and Q-type calcium channels has been reported in vas deferens from the rat and mouse (36) and the guinea pig (37). This GVIA-resistant response was seen only with high frequencies of stimulation, either because the P and Q channels were recruited only at high frequencies or because their effect accumulated to functionally important levels at high frequencies (38). This GVIA-resistant component was too small to affect our estimates of N-channel blocking potencies in the artery experiments, and was not present in the rat vas deferens and atrium experiments in this study, where only single- or four-pulse stimuli were used.
In each assay the incorrectly folded isomers of G5A were inactive, but substitution of Gly5 with D-Ala did not result in any reduction in potency. Both N20A and T23A, tested only as impure mixtures of isomers, were weak in all assays and were applied at a sufficiently high concentration for a reliable potency estimate only in the binding and vas deferens assays. After correction for the fraction of properly folded product estimated from NMR, their potencies were similar to native GVIA.
K2A and Y13A showed markedly reduced potency compared with native GVIA
in both the binding assay and the functional assays (Fig. 5), but both
retained full activity (i.e. inhibited the sympathetic nerve
responses completely) when applied at a high enough concentration (Fig.
6). Arg17 was also significantly less potent
than GVIA in all assays (3-fold in the binding and the vas deferens,
10-fold in the artery and atrium), but for the other analogues there
was poor agreement between the binding assay and the functional assays.
An obvious discrepancy was that none of the analogues was substantially
more potent than native GVIA in the functional assays, but several were
in the binding assay (Fig. 5). There are some common features among the
different assays, with Arg17 being important in all four
and the regions of Hyp21-Tyr22 and
Lys24-Arg25 being important in the three
functional assays, but significant differences between the assays were
also apparent. A correlation analysis was performed (excluding the
results for K2A and Y13A to avoid the otherwise inevitable
overestimation of the quality of the correlation), and it was found
that the analogue potencies were significantly correlated in the three
bioassays (p < 0.01), but there was no significant
correlation between the potencies in the binding assay and the
potencies in the functional assays (Fig. 7). Two
analogues appeared to have some selectivity between the tissue assays.
Y22A was about 3-fold less potent than GVIA in the vas deferens but 30- and 100-fold less potent than GVIA in the atrium and artery, and O10A
was almost equipotent with GVIA in the vas deferens and 5- and 15-fold
less potent in the artery and atrium, respectively.
Conscious Rabbit Experiments
GVIA (10 µg/kg, intravenous) caused a fall in MAP of 15 mm Hg and a tachycardia of about 80 beats/min (33). The onset was slow, with a peak response after 35-40 min, and the effect was maintained for several hours. GVIA also decreased the baroreflex HR range from 175 to about 70 beats/min and increased the lower HR plateau of the baroreflex curve, which indicated an attenuation of both the sympathetic (upper) and vagal (lower) components of the curve. The effects of some of the analogues closely mirrored those of GVIA. G5dA and S6A each at a dose of 10 µg/kg (intravenous) caused similar changes in MAP and HR to those elicited by GVIA. The baroreflex curves after these two analogues were also similar to the curve after GVIA, with the HR range of the curves reduced to about 70 beats/min in each case. For all other analogues tested in this assay, a dose >10 µg/kg (intravenous) was required to cause similar changes in MAP and HR, as well as to the baroreflex, to those observed after GVIA. The approximate potency ratios compared with GVIA were: N14A, 3; K2A, 10; R17A, Y22A and K24A, 30; and Y13A, >100.
In this structure-activity study of GVIA, we have found that Lys2 and Tyr13 are the most important individual amino acids for affinity for the N-type calcium channel, a result that has been reported previously by others (24). In contrast to previous studies, however, the functional assays used in this study show that several other amino acid residues are also important for the high affinity of GVIA for the N-type calcium channel, and the detailed NMR investigation of the alanine-substituted analogue structures shows that the potency reductions produced by the Ala substitutions are primarily the result of the loss of side-chain interactions with the binding site rather than perturbations of the tertiary structure of the molecule.
Synthesis and Folding of GVIA AnaloguesOnly three of the
Ala-substituted analogues of GVIA (G5A, N20A, and T23A) failed to fold
into a native-like conformation in substantial yield. The inability of
G5A to fold into the same structure as the native toxin is not
surprising, as Gly5 is an essential component of the type
II -turn, which extends over residues 3-6 in the molecule (8-11).
Introduction of D-Ala at this position, in contrast to
L-Ala, resulted in a product that folded to a native-like
structure, presumably because D-Ala mimics the positive
-angle of Gly5 and thus supports a
-turn structure
(39). Surprisingly, Kim et al. (24) did not report any
difficulty in refolding G5A. N20A and T23A failed to fold into a single
major product, and our NMR studies indicated that only about 1% and
3%, respectively, of native-like structure were present in the
mixtures, giving overall yields of <0.15% of the crude materials. Kim
et al. (40) did report low yields of correctly folded N20A
and T23A. Residues 20 and 23 may be important for the initial folding
of the linear polypeptide chain as a precursor to formation of the
disulfide bridges, a possibility supported by a side chain to side
chain interaction between these residues evident in the refined NMR structures.2
The NMR studies confirmed that most Ala-substituted analogues of GVIA had conformations very similar to native GVIA, confirming that the modulation of activity by alanine substitutions can be ascribed to the altered chemical moiety rather than conformational changes. One possible exception to this is K2A, where some of the modulation of its activity may be due to a small conformational change. Even in this case, however, there were only three NOE changes out of the total of 192, so the conformational change was very small. An interesting feature of the NMR analyses was that the NOE changes were localized almost exclusively in the poorest defined region of the molecule, namely Cys8 to Cys15. It has been proposed that this part of the molecule may be genuinely flexible (9), and in the refined structures2 two families of conformations for this region of the molecule are present. This region may be more sensitive to substitutions at various positions throughout the molecule because of a lack of stabilizing interactions. Interestingly, Kim et al. did not report any perturbation of the structure of GVIA by the substitution of Ala for Lys2. This presumably reflects the fact that the circular dichroism methods used in their study are not nearly as sensitive to local conformational changes as NMR analysis.
N-type channel Blocking ActivitiesThe potencies of GVIA and the analogues measured in the functional assays are not direct measurements of their affinity for the N-type calcium channel because there is an unknown relationship between occupation of the N-type calcium channel by the blockers and the fractional inhibition of the responses we measured, and because the analogues were not allowed sufficient time for full equilibration within the tissues and at the channels. The onset of inhibition of N-type calcium channels by GVIA has a very slow time course, with over 90 min required for full expression of the effect of low concentrations in the rat isolated small mesenteric artery (41). Drug contact time in our functional in vitro assays was too short for full equilibration of each concentration of peptide. The under-equilibration was largest in the vas deferens assay because of a shorter contact time (20 min, compared with 30 min in the other assays) and because the vas deferens is substantially thicker than the other tissues and diffusion of the peptides into the tissue should be slower. The time for equilibration at a receptor is shorter for any drug when it is applied at a higher concentration, and thus any affinity reduction in the GVIA analogues would be underestimated more in the vas deferens assay than in the other assays. Additionally, the truncated contact time for GVIA and its analogues dictates that any decrease in affinity that is the result of a slowed association rate constant will result in an exaggerated decrease in potency, because it would also increase the time required for full equilibration. The corollary applies that any decrease in affinity that is the result of a faster dissociation rate constant will be underestimated. Without knowledge of the rate constants involved, it is not possible to determine the magnitude of these effects. These issues may not be relevant to the binding assays, which had 90-min incubation times (albeit at 4 °C).
There was a strong correlation between the potency of each analogue in the three bioassays, but, excluding positions 2 and 13 to avoid an artifactually inflated correlation coefficient, there was no correlation between the binding potencies and the functional potencies (Fig. 7). Furthermore, none of the Ala-substituted analogues showed a significantly increased potency compared with native GVIA in the functional assays, but several were 3-fold more potent than GVIA in the binding assays in this study and a previously published study by others (24). The in vivo experiments yielded data that were rather qualitative in nature because issues of drug distribution and metabolism can alter apparent potencies, but it was clear that the order of potency of the analogues tested in vivo resembled that obtained in the other functional assays. In particular, the activity of Y22A in vivo (about 30-fold less potent than GVIA) was within the range of its potency in the functional assays (between 3- and 100-fold less potent than GVIA) but not its potency in the binding assay (about 3-fold more potent than GVIA). The discrepancies between the functional and binding assays might be explained in several ways. First, the conditions of the binding assay may alter the conformation of the N-type calcium channel from its native (physiological) conformation, maintaining the strong interaction between the channel and the Lys2 and Tyr13 residues of GVIA but significantly altering the weaker interactions between the channel and the other residues. Our binding buffer contained isotonic sucrose and only the sodium ions from adjusting the pH with NaOH, and this near absence of small ions might well have altered the interaction between GVIA and the N-type calcium channel or the conformation of the channel. Indeed, the binding of GVIA to the channel has been reported to be modified by physiological concentrations of calcium, sodium, potassium and magnesium ions (42). The N-type calcium channel is voltage-gated, and thus by definition its conformation is sensitive to changes in membrane potential. In the binding assays the membrane potential is zero, with an identical ionic milieu on each side of the membrane (presuming that vesicles are not formed), so it might be expected that the interaction between GVIA and the channel could be different in the binding experiments from the functional experiments. It has been shown that the interaction of the scorpion toxin charybdotoxin with the voltage-gated potassium channel expressed in Xenopus oocytes was affected by interaction of the channel with the cytoskeleton (43), and similar interactions of the calcium channel could contribute to the differences observed here for GVIA. If the properties of the N-type calcium channel were significantly altered by the conditions chosen for binding experiments then any rational drug discovery program based solely on the binding data would be compromised.
Another possible explanation for differences among the assays is tissue-dependent subtypes of N-type calcium channels. The binding experiments were conducted with brain membranes, and the functional assays used peripheral tissues. Any subtle differences between the N-type calcium channels expressed in different tissues would show up as differences among the orders of potency of the Ala-substituted analogues in the different assays. This could also contribute to both the modest discrepancies in the analogue potencies among the three in vitro functional assays, and the more striking differences between the potencies in the functional assays and the rat brain binding assay in this study and the chick brain assay of Kim et al. (whose binding conditions were nominally the same as this study) (24, 44). The pairwise correlations between the results from the different functional assays were all similar in magnitude (Fig. 7). However, the correlation slopes suggest that the vas deferens was least sensitive to changes in analogue potencies and the artery assay the most sensitive. This could be the result of tissue differences in N-type calcium channels, but is more likely to be the result of a combination of the shorter than ideal toxin contact time and restricted diffusion of the analogues to their sites of action in the vas deferens, as discussed above. Tissue and species differences in the N-type calcium channel would be very important for potential therapeutic applications of N-type calcium channel blockers, but more detailed studies would be required to confirm such differences.
In all of the assays, the least potent of the correctly folded
analogues were Y13A and K2A, confirming that Tyr13 and
Lys2 are the most significant determinants of GVIA binding
affinity. Although our NMR studies showed that K2A had a slightly
different structure from native GVIA, the conformational change appears to be minor and we believe that the activity losses from Ala
substitutions at Lys2 and Tyr13 indicate that
these side chains are important in the interaction between native GVIA
and its binding site on the -subunit of the N-type calcium channel.
Other studies of Ala-substituted analogues have also identified these
residues as being important in GVIA (24, 44). Additionally, GVIA
biotinylated at Lys2 is about 10-fold less potent than
native GVIA in binding to rat brain membranes (30). A
structure-activity study of
-conotoxin MVIIA, a selective N-type
calcium channel blocker with a similar structure to GVIA, also pointed
to the importance of Lys2 and Tyr13 (25). As
the solution structures of GVIA and MVIIA are very similar (with only
small rearrangements due to a single insertion and deletion) (14, 15)
and the same residues, Lys2 and Tyr13 (25),
have been identified as being important, it may be assumed that GVIA
and MVIIA bind to the N-type calcium channel in the same way.
A study of chimeras of the -subunits of the N-type and a
GVIA-insensitive calcium channel pointed to a putative external loop
between the IIIS5-IIIH5 regions as being important to the activity of
GVIA. Substitution of Asn and Lys for each of two Glu residues caused a
substantial decrease in the association rate of GVIA (45). It is
possible that one or both of these negatively-charged Glu residues
interacts with the positively charged Lys2, either
decreasing the dissociation rate constant by acting as an energetically
favorable interaction between bound GVIA and the N-type calcium
channel, or increasing the association rate constant by contributing to
the favorable alignment of GVIA when it approaches the channel
(46).
Although Lys2 and Tyr13 are the most important
individual residues for affinity of GVIA, the functional assays
investigated in our study point to additional interactions that were
not evident from the binding assay. These additional residues are
distributed over a larger region of the molecular surface than that
encompassed by Lys2 and Tyr13 (Fig.
8). This means that the pharmacophore of GVIA is larger and more complex than has been suggested by others previously. The
interactions with the N-type calcium channel of GVIA and the structurally related MVIIA are probably similar. In the case of MVIIA,
Arg10, Leu11, and Arg21 are as
important as Lys2 and Tyr13 for the affinity of
MVIIA for the N-type calcium channel, although it should be noted that
these conclusions were based on rat (25) and chick (44) brain binding
data. Fig. 8 shows the structures of GVIA and MVIIA with key residues
highlighted. It can be seen that, in addition to the conserved
Lys2 and Tyr13, the locations of
Tyr22, Hyp21, and Hyp10 and to a
lesser extent Lys24 in GVIA are similar to those of
Arg21, Arg10, and Leu11 in MVIIA.
Thus, although these residues are not identical, it is likely that they
interact with similar regions of the calcium channel. The distributed
binding surfaces are consistent with a model in which the toxins block
the channel by interacting with both the pore and the vestibule of the
channel, as in the case of scorpion toxins and the voltage-gated
potassium channel (47). However, channel blockade due to toxin binding
remote from the pore cannot be ruled out from our data.
It is concluded that there are important differences between the structure-activity relationships for GVIA analogues when tested in binding and functional assays, which may result from either tissue-specific distribution of subtypes of the N-type calcium channel, or from the non-physiological nature of the conditions of the binding assay. The functional assay data show that there are several residues on GVIA that make important interactions with the channel in addition to the previously identified residues Tyr13 and Lys2.
We thank Peter Coles, Mark Ross-Smith, and Angela Mountain for expert technical assistance.