Novel ω-Conotoxins from Conus catus Discriminate among Neuronal Calcium Channel Subtypes*

ω-Conotoxins selective for N-type calcium channels are useful in the management of severe pain. In an attempt to expand the therapeutic potential of this class, four new ω-conotoxins (CVIA–D) have been discovered in the venom of the piscivorous cone snail, Conus catus, using assay-guided fractionation and gene cloning. Compared with other ω-conotoxins, CVID has a novel loop 4 sequence and the highest selectivity for N-type over P/Q-type calcium channels in radioligand binding assays. CVIA−D also inhibited contractions of electrically stimulated rat vas deferens. In electrophysiological studies, ω-conotoxins CVID and MVIIA had similar potencies to inhibit current through central (α1B-d) and peripheral (α1B-b) splice variants of the rat N-type calcium channels when coexpressed with rat β3 in Xenopus oocytes. However, the potency of CVID and MVIIA increased when α1B-d and α1B-b were expressed in the absence of rat β3, an effect most pronounced for CVID at α1B-d (up to 540-fold) and least pronounced for MVIIA at α1B-d (3-fold). The novel selectivity of CVID may have therapeutic implications. 1H NMR studies reveal that CVID possesses a combination of unique structural features, including two hydrogen bonds that stabilize loop 2 and place loop 2 proximal to loop 4, creating a globular surface that is rigid and well defined.

Predatory marine gastropods of the genus Conus (cone snails) represent a diverse family of marine molluscs that use venom for prey capture. The venom of each species contains a unique array of more than 100 peptides, whose pharmaceutical potential remain largely unexploited (1). Different classes of conotoxins have evolved to target ion channels and receptors for the successful capture of fish, molluscs, or worms (2,3). One important class, the -conotoxins (Table I) isolated from piscivorous species, inhibits neuronal voltage-sensitive calcium channels (VSCCs) 1 found in mammals (4). -Conotoxins have selectivity for N-type (e.g. MVIIA and GVIA) or P/Q-type (e.g. MVIIC) VSCCs, making them widely used research tools for defining the distribution and role of neuronal VSCCs (4 -8). In addition to their use as research tools, animal studies have shown that -conotoxins that target N-type VSCCs have clinical potential in ischemic brain injury (9) and pain (10). However, -conotoxins presently available are not ideal therapeutics, despite their selectivity and potency and the dominant role of N-type VSCCs at synapses in carrying nociceptive information in the spinal cord (11). For example, intrathecal MVIIA causes a variety of neurological side effects of unknown origin (12). Another N-type VSSC selective -conotoxin, GVIA, dissociates slowly from N-type VSCCs (13,14) and, accordingly, may be difficult to administer in a clinical setting. Other -conotoxins are less selective for N-type VSCCs and are not considered useful therapeutic candidates.
A number of central and peripheral splice variants of the N-type VSCC have been reported (14). However, calcium channel probes with greater selectivity are required to help understand the role(s) played by the various forms of the N-type VSCC in normal and disease states. Here we report the isolation and characterization of four new -conotoxins (CVIA-D, see Table I) discovered in the venom of the piscivorous cone snail, Conus catus. One of these, CVID, has a number of novel structural features that underlie an enhanced ability to discriminate between N-type (␣ 1B ) and P/Q-type (␣ 1A ) VSCCs, as well as between splice variants of the ␣ 1B expressed in oocytes in the presence or absence of rat ␤ 3 .

EXPERIMENTAL PROCEDURES
Crude Venom Extraction-37 specimens of C. catus (average length 4 cm) were collected from the southern Great Barrier Reef. Crude venom duct contents were extracted with 30% acetonitrile/water acidified with 0.1% trifluoroacetic acid and centrifuged. Soluble material was lyophilized and stored at Ϫ20°C prior to use.
Sequencing of Native Peptides-The purified peptides (ϳ100 pmol) were reduced in the presence of TCEP, 50 mM ammonium acetate at pH 4.5 (37°C for 1 h), and then alkylated in the presence of maleimide (37°C, 1 h). The alkylated peptides were purified by RP-HPLC, applied to a Biobrene-treated glass fiber filter, and analyzed by Edman chemistry using an Applied Biosystems, Inc., model 470A protein sequencer. Alkylation of peptides with maleimide allowed their cysteine residues to be observed as phenylthiohydantoin-Cys-maleimide doublets (diastereomers), and Ser was confirmed by the presence of dehydroserine. The high number of serine, threonine, and cysteine residues present in each conotoxin resulted in a relatively rapid decrease in the amino acid yields in successive cycles; however, complete sequences were obtained in each case. The low serine signals were accompanied by the presence of large dehydroserine peak.
Gene Isolation and Characterization-Ducts from two specimens of C. catus were emulsified, and poly(A) ϩ -tailed mRNA was extracted using the QuickPrep mRNA purification system (Amersham Pharmacia Biotech). Strand-1 cDNA was 3Ј end-synthesized from the C. catus poly(A) ϩ mRNA templates using a NotI-d(T) 18 bifunctional primer (5Ј-AACTGGAAGAATTCGCGGCCGCAGGAAT 18 ) (Amersham Pharmacia Biotech) extended with Superscript II reverse transcriptase (Life Technologies, Inc.). The resultant cDNA templates were used to manufacture double-stranded cDNA using a RNase H/DNA polymerase procedure (cDNA Timesaver system, Amersham Pharmacia Biotech). Marathon adaptors (CLONTECH) were added to the 5Ј and 3Ј ends of the ds-cDNA molecules to complete the cDNA construction. An oligonucleotide PCR primer (R-301A 5Ј-atcatcaaaATGAAACTGACGTG) was designed from a consensus of aligned coding sequences from conotoxins with a 6-cysteine/4-loop framework comprising published (15,16) and unpublished data. 2 The primer included a translation start site (underlined) and 9 base pairs of 5Ј-untranslated sequence (lowercase). This forward PCR primer was designed to work in conjunction with the reverse ANCHOR primer (5Ј-AACTGGAAGAATTCGCGGCCGCAG-GAAT) that was homologous to the adaptor at the 3Ј termini of the ds-cDNA templates.
Peptide Quantitation-Peptides were quantified initially by triplicate amino acid analysis (19) and then by RP-HPLC (HP 1100) using an external reference standard for each peptide. Stability studies in physiological saline solutions showed that nonspecific binding of these peptides to glass or polypropylene was minimal.
Mass Spectrometry-Mass spectra were acquired on a PE-Sciex API III triple quadrupole electrospray mass spectrometer (MS) in positive ion mode (m/z 500 -2000, at 0.1-0.2-Da steps, declustering potentials of 10 -90 V, and dwell times of 0.4 -1.0 s). Data were deconvoluted (MacSpec 3.2, Sciex, Canada) to obtain the molecular weight from the multiply charged species. MS was used to confirm purity and to monitor peptide oxidation. Native and synthetic CVIAϪD were compared on a 5-m C18 column (0.21 ϫ 25 cm, Vydac) eluted with 0.05% trifluoroacetic acid for 2 min and then with a linear gradient to 80% solvent B (solvent B, 90% acetonitrile, 0.045% trifluoroacetic acid) over 80 min at a flow rate of 130 l/min, using a Hewlett-Packard 1100 pump and MS detection.
Radioligand Binding-Radioligand binding were performed on rat brain membrane incubated in the following (mM): 20 HEPES (pH 7.2), 75 NaCl, 0.1 EDTA, 0.1 EGTA, 2 leupeptin, 0.5 units of aprotinin, and 0.1% bovine serum albumin, as described previously (18,20,21). 125  On-rate and off-rate experiments were performed at 23°C in single 50-ml polypropylene tubes containing 125 I-CVID or 125 I-MVIIA (30 pM diluted with assay buffer). On-rate experiments commenced with the addition of rat brain membrane, and off-rate experiments commenced with the addition of excess CVID or MVIIA (10 Ϫ7 M) following a 30-min incubation of rat brain membrane with 125 I-CVID or 125 I-MVIIA (30 pM), respectively. At defined intervals, aliquots (300 l) were filtered (25-mm Whatman GF/B treated with 0.6% polyethyleneimine), and the filtrates were rapidly washed three times with ice-cold wash buffer. Nonlinear regressions were fitted to each experiment (n ϭ 3 data points per experiment) with Prism software (GraphPad, San Diego, CA).
cervical dislocation and exsanguinated. Prostatic halves of each vas deferens were mounted under 0.5 ϫ g tension in 5-ml organ baths containing physiological solution containing the following (mM): 119 NaCl, 4.7 KCl, 1.17, MgSO 4 , 1.18 KH 2 PO 4 , 25.0 NaHCO 3 , 5.5 glucose, 2.5 CaCl 2 , 0.026 EDTA, at 37°C and bubbled with 5% CO 2 in O 2 (pH 7.4). After 45 min of equilibration, electrical field stimulation every 3 min (single 55 V, 0.1-ms duration pulses generated by a Grass S44 stimulator) were applied via two platinum-stimulating electrodes straddling the tissue. Isometric contractions were measured using a force transducer (F-60, Narco Bio-System) and recorded digitally on a Power Macintosh computer using Chart version 3.5.6/s software and a MacLab/8 s data acquisition system (ADInstruments, Australia) at a sampling frequency of 200 Hz. Evoked contractions were abolished by tetradotoxin (0.1 M), indicating that they were of neurogenic origin. Conotoxins were added cumulatively at 18-min intervals, with allowance made for a small (0.16% /min) deterioration of the evoked response. CVID effect on ATP (3 ϫ 10 Ϫ3 M, Sigma) and norepinephrine (10 Ϫ4 M, Sigma) responses in the bisected epididymal portion of the rat vas deferens was also determined. Dose-response data (means Ϯ S.E.; n ϭ 5) were analyzed using Prism software.
Oocyte Injection and Recording-Oocytes were surgically removed from mature Xenopus laevis frogs anesthetized by submersion in 0.1% ethyl m-aminobenzoate (MS 222). Stage V-VI oocytes were prepared for injection by dissociation in Ca 2ϩ -free solution containing the following (mM): 96 NaCl, 2 KCl, 1 MgCl 2 , 5 HEPES (pH 7.4), plus 1 mg/ml collagenase (Sigma, type I) for 1 h at room temperature. Oocytes, maintained in storage solution containing (mM) 96 NaCl, 2 KCl, 1 CaCl 2 , 1 MgCl 2 , 5 HEPES, 5 pyruvate, plus 50 g/ml gentamicin (pH 7.4), were injected with 25-50 ng of total cRNA using a precision injector (Drummond "Nanojet") and incubated at 18°C for 3Ϫ5 days. Injected cRNA obtained using an Ambion mMessage mMachine kit comprised rat brain peripheral (␣ 1B-b ) or central (␣ 1B-d ) subunits of the N-type VSCC expressed in the absence or presence (1:1 ratio) of the rat ␤ 3 subunit provided by D. Lipscombe (12). Recordings were made in a nominally Ca 2ϩ -free solution containing (mM) 85 tetraethylammonium, 5 BaCl 2 , 5 KCl, 5 HEPES, adjusted to pH 7.4 with methanesulfonic acid perfused at a rate of ϳ10 ml/min. Depolarization-activated Ba 2ϩ currents were recorded using a GeneClamp 500B two-electrode voltage clamp amplifier (Axon Instruments Inc., Foster City, CA) at room temperature (21Ϫ23°C). Voltage and current electrodes were filled with 3 M KCl and had final resistances of 0.2-1.0 M⍀. VSCC currents in oocytes were evoked by a step depolarization to 0 mV from a holding potential of Ϫ80 mV using pClamp programs (Axon Instruments Inc). The linear membrane capacitative and leak currents were subtracted using a ϪP/4 pulse protocol. Membrane currents were sampled at 10 kHz and filtered at 2 kHz. Oocytes showing Ͻ15% change in peak current amplitude over a 10-min incubation period were used in these studies to avoid effects associated with current run down. Oocytes that exhibited a slowly developing Cl Ϫ current were discarded, and in control experiments on the remaining oocytes, BAPTA injection did not affect the Ba 2ϩ current, indicating that this current was not contaminated by Ca 2ϩ -activated Cl Ϫ current. Perfusion was stopped, and cumulative additions of -conotoxins were added to a Teflon chamber (0.8 ml volume) to achieve rapid mixing. Inhibition of peak Ba 2ϩ current amplitude was measured when the inhibitory effects of each concentration of MVIIA and CVID approached steady state values, typically requiring a 5-7-min incubation to achieve Ͻ2% change in peak current amplitude per min. Data (means Ϯ S.E.; n ϭ 3-5) were fitted with nonlinear regressions using Origin software.
1 H NMR and Three-dimensional Structure Calculations-All NMR experiments were recorded on a Bruker ARX 500 spectrometer equipped with a z-gradient unit, or a Bruker DMX 750 spectrometer equipped with an x,y,z-gradient unit. The H␣ chemical shifts were obtained from spectra of CVIA-D in 95% H 2 O, 5% D 2 O (pH 2.5-3.5) at 293 K, and restraints for three-dimensional structure calculations of CVID were obtained from spectra recorded at 280 and 293 K at pH 3 and 5.5 (18). Structures were calculated using the torsion angle dynamics/simulated annealing protocol in XPLOR version 3.8 (22)(23)(24)(25), as described previously (20).

Isolation and Synthesis of -Conotoxins from C. catus-The
four -conotoxins, CVIAϪD, were isolated from the crude venom of C. catus using activity-directed fractionation. Each comprised Ͻ1% of total venom peptide as determined by HPLC (214 nm) or MS detection (m/z 500 -2000). The amino acid sequences of CVIA and CVIB were cosequenced at a 3:2 ratio, whereas CVIC and CVID were obtained from separate fractions. Synthetic C-terminally amidated CVIA, CVIC, and CVID gave retention times and masses that were indistinguishable from the corresponding native peptides (Fig. 1), thus CVIA, CVIC, and CVID were determined to be C-terminally amidated. For each peptide, the oxidized form eluted earlier than the reduced form.
Gene Isolation and Characterization-PCR of the C. catus venom duct cDNA templates produced a DNA product of approximately 380Ϫ500 base pairs (data not shown). A relatively broad DNA band indicated the presence of a composite PCR product. Subsequent sequence analysis of more than 100 clones derived from the PCR product libraries confirmed this observation. The PCR products contained peptides of theand ␦-conotoxin types, as well as sequences from at least two other conotoxin families yet to be assigned. Within the -conotoxin sequences isolated were two that translated to putative mature peptides identical to the CVIA and CVID sequences. Despite an extensive screening of the two C. catus PCR product libraries, clones for the conotoxins CVIB and CVIC were not located, and the CVID clone was located in only one specimen. The complete nucleotide and predicted amino acid sequences derived for CVIA and CVID are shown in Fig. 2 and are aligned with the sequences for MVIIA and GVIA. Homology screening of public nucleotide and amino acid data bases with the CVIA and CVID sequences indicated that both sequences were unique.
Saturation binding studies indicated that 125 I-CVID and 125 I-MVIIA (67 Ϯ 11 and 37 Ϯ 12 pM, respectively) had poten-  (16), and MVIIA (Australian C. magus), 2 with start and stop coding sequences underlined. The first 14 residues of CVIA and CVID were generated as part of the PCR primer R-301A. However, subsequent 5Ј-rapid amplification of cDNA ends PCR indicated that this sequence was identical to the native coding sequence for both peptides (data not shown). The position of the cleavage site between the pre-pro and mature peptide sequences is indicated by an arrow. B, ClustalW alignment of the amino acid sequences for the conopeptides CVIA, CVID, MVIIA, and GVIA. The predicted amino acid sequence for MVIIA from A was the same as previously reported (55). The predicted signal (positions 1-21) and propeptide (positions 22-45) regions of the leader sequences are highlighted by dashed lines (56). An arrow indicates the position of the known cleavage site separating the pre-pro and mature peptides of the four sequences. The intercysteine (C) loop sequences for the mature peptides are also indicated. Conserved residues are dark shaded, and conservative changes in amino acids are light shaded. cies slightly lower than CVID and MVIIA in rat brain membrane, and both appeared to occupy a single class of receptor (Fig. 4). In paired experiments (n ϭ 4), 125 I-CVID (0.07 Ϯ 0.03 pmol/mg protein) bound to 31% fewer receptors in rat brain membrane than 125 I-MVIIA (0.10 Ϯ 0.04 pmol/mg protein) (p ϭ 0.049, paired two-tailed t test), and 125 I-CVID had half the nonspecific binding of 125 I-MVIIA (Fig. 4). Despite this difference in the density of receptor sites recognized, displacement assays did not reveal a component of 125 I-GVIA or 125 I-MVIIA binding that was not displaced by CVID. 127 I-CVID (IC 50 0.44 nM) also fully displaced 125 I-MVIIA binding to rat brain with a Hill slope of unity. However, saturation binding studies have limited power to resolve binding sites of similar affinity, and a lower affinity binding site comprising 30% of receptors would be under-represented with the concentration of 125 I-MVIIA used in the displacement assays (20% of its K d concentration). Displacement of higher concentrations of 125 I-MVIIA by MVIIA, GVIA, CVID, or 127 I-CVID also did not identify a significant resistant component, possibly due to the reduction in the signal to noise of the assay with increased labeled ligand concentrations (data not shown). The difference in B max did not arise from differences in binding kinetics; 125 I-CVID and 125 I-MVIIA had similar k on values (0.024 and 0.08 min Ϫ1 pM Ϫ1 , after adjusting for k off ) and k off values (0.56 and 0.53 min Ϫ1 , respectively). The K d values for 125 I-CVID and 125 I-MVIIA calculated from kinetic data (23 and 7 pM, respectively) were similar to the estimates obtained from saturation binding studies (67 and 37 pM, respectively).
Rat Vas Deferens-CVIA-D inhibited electrically stimulated rat vas deferens contractions (Fig. 5A). The rank order of potency was GVIA Ͼ CVID Ϸ MVIIA Ͼ CVIA Ͼ CVIC Ͼ CVIB (Fig. 5B and Table II). Inhibition of the nerve-evoked responses in rat vas deferens was positively correlated with displacement of 125 I-GVIA binding across the -conotoxins tested (log(rat vas deferens IC 50 ) ϭ 0.78(log( 125 I-GVIA IC 50 ) ϩ 0.1) (r 2 ϭ 0.92)). CVID and MVIIA gave Hill slopes significantly greater than unity (2.9 and 3.3, respectively), whereas CVIA-C and GVIA gave Hill slopes that were not significantly different from unity. The origin of these differences is unclear; however, differences in N-type versus P/Q-type VSCC selectivity do not appear to be a contributing factor. The nonlinear relationship between calcium influx and transmitter release (26) is expected to give rise to a Hill slope greater than unity. Base-line tension was not affected by this series of peptides, and CVID (3 ϫ 10 Ϫ7 M) had no significant effect on the size of ATP and norepinephrine-induced contractions in rat vas deferens (data not shown, n ϭ 3). A residual component of rat vas deferens contractile activity remained at saturating concentrations of CVID, MVIIA, and GVIA (14.8% (95% CI, 12-18%), 10.5% (7-14%), and 10% (6 -14%), respectively) (Fig. 5B).
Xenopus Oocytes-Two-electrode voltage clamp studies revealed that CVID and MVIIA were potent inhibitors of ionic currents through peripheral (␣ 1B-b ) and central (␣ 1B-d ) splice variants of the rat N-type calcium channels expressed in Xenopus oocytes in the presence or absence of rat ␤ 3 ( Fig. 6; Table  III). The voltage dependences of activation of these VSCC conductances were unaffected by CVID or MVIIA (data not shown). CVID and MVIIA had similar potencies to inhibit depolarization-activated inward Ba 2ϩ current arising from ␣ 1B-b or ␣ 1B-d coexpressed with rat ␤ 3 (Fig. 6, C and D). At the ␣ 1B-b , the potency of both MVIIA and CVID increased 10-fold in the absence of rat ␤ 3 . At the ␣ 1B-d , in contrast, CVID potency increased up to 540-fold, whereas MVIIA potency was only slightly increased in the absence of excess rat ␤ 3 . Inhibition of the inward Ba 2ϩ current by CVID was slowly reversed upon washout, and neither conopeptide inhibited the small endogenous calcium channel current (Ͻ5 nA) seen in oocytes that were not injected with cRNA (data not shown). For the ␣ 1B /␤ subunit combinations tested, a relatively small residual N-type current (ϳ10% of maximum peak current) remained in the presence of saturating concentrations of CVID or MVIIA (Fig. 6, C and D). The residual component had reversal potentials and inward current maxima typical of N-type VSCCs and were completely blocked by bath application of 100 M Cd 2ϩ (data not shown), suggesting that both peptides are unable to occlude completely current through open N-type VSCCs. Further studies are required to confirm that CVID and MVIIA are allosteric inhibitors of the N-type VSCC, as suggested for GVIA from single channel data (27). VSCC chimera studies indicate that GVIA may occlude the pore of the N-type VSCC (28). 1 H NMR Studies-The two-dimensional 1 H NMR spectra of -conotoxins CVIAϪD were each assigned using standard protocols (29). As with other -conotoxins we have examined (18,20,30), peaks were well dispersed, and there was little overlap. CVIA-D secondary shifts closely paralleled those of MVIIA (Fig. 7), whose structure is well defined (30), indicating that the overall fold and disulfide connectivity of these peptides are the FIG. 3. -Conotoxin potency at N-type and P/Q-type VSCCs in rat brain. A, potency at N-type VSCCs measured by displacement of 125 I-GVIA binding. B, potency at P/Q-type VSCCs measured by displacement of 125 I-MVIIC binding. Displacement data was best fit to a single-site competition model (shown). C, selectivity for N-type versus P/Q-type VSCCs in rat brain. Shown are the means Ϯ S.E. for pIC 50 (N-type) and pIC 50 (P/Q-type).
same. Secondary H␣ shift analysis has been shown to be a sensitive discriminator of disulfide-bonded isomers of other conotoxins (31) and can also be used to identify the location, but not the nature, of local conformational differences across a set of structurally related peptides (32). The secondary H␣ shifts of CVIA and CVIB, which have the same loop sizes as MVIIA, are almost identical to those of MVIIA. The secondary H␣ shifts of CVIC are also similar to those of MVIIA, apart from loop 4 where the loop sizes are different. In this case, the structure of CVIC would be better approximated by MVIIC, which has the same sized loop 4.
The greatest difference in H␣ secondary shifts compared with MVIIA was seen in loops 2 and 4 of CVID. Although differences in loop 4 are not surprising given that CVID has two additional residues, the differences in loop 2 are remarkable given the similarities between MVIIA and CVID at loop 2. The secondary shifts of residues 9 -14 in loop 2 of CVID follow the same pattern of those in MVIIA but are of greater magnitude, indicating that the structure of loop 2 in CVID may be more stabilized. This could stem from a long range interaction with loop 4, where the major differences in primary structure are located. Loop 2 has previously been the least defined region of -conotoxin structures, with residues of this loop characterized by relatively broad peaks in the 1 H NMR spectra, indicative of conformational exchange (30,33). This lack of structure definition has hindered attempts to understand the crucial role loop 2 plays in activity, function, and selectivity of -conotoxins, particularly that of the principal binding determinant Tyr 13 , as well as residues of secondary importance such as Leu 11 and Arg 10 in MVIIA (34). Because no line broadening is observed in loop 2 or elsewhere in CVID but is prominent in loop 2 of MVIIA, CVID may provide a significantly improved structural template for pharmacophore development. Given that the significant differences in secondary H␣ shifts for residues in loops 2 and 4 in CVID preclude accurate modeling of CVID from existing -conotoxins structures, and given its enhanced N-type selectivity and structural stability, we determined the three-dimensional structure of CVID using 1 H NMR spectroscopy, as described below.
Three-dimensional Structure of -Conotoxin CVID-CVID structures were calculated based on a total of 481 distance restraints derived from 159 intraresidue, 110 sequential, 184 medium and long range NOEs, 28 hydrogen bond restraints defining a total of 14 hydrogen bonds, and 23 and 10 1 dihedral angle restraints. A total of 47 of 50 structures converged to a consensus fold, with no NOE distance violation greater than 0.2 Å and no dihedral violations greater than 3°. Of these, the 20 lowest energy structures were chosen to represent the structure of CVID. The structures are exceptionally well defined, with a backbone pairwise root mean squared difference of 0.35 Å (calculated over all residues). The angular order parameters (S) for the and backbone dihedral angles had average values of 0.99, indicating a high degree of structure precision that is reflected in the low average root mean squared difference to the mean structure of 0.24 Å.
The CVID structures are shown superimposed globally over the backbone atoms, together with the lowest energy structure displayed in secondary structure mode in the same orientation (Fig. 8A). The same structures are shown rotated 90°y, to display side chains of residues in loops 2 and 4 which are important for binding (Fig. 8B). As for other -conotoxins, the secondary structure of CVID is dominated by a triple-stranded ␤-sheet, in this case incorporating residues 7Ϫ9 and 25Ϫ27 as the two peripheral strands and residues 19Ϫ21 as the central ␤-strand. Other secondary structural features include a  ␤-bridge composed of residues 1Ϫ2 and 14Ϫ16 and several ␤-turns. The ␤-turns that link the ␤-sheet and ␤-bridge regions comprise residues 3Ϫ6 (type II), 9Ϫ12 (type I), 10Ϫ13 (type I), 15Ϫ18 (type I), and 21Ϫ24 (type IV). The ␤-bridge has not previously been reported explicitly for other -conotoxins but is likely to exist in MVIIA, GVIA, and MVIIC and has been identified in -conotoxin PVIIA, which adopts the same structure as the -conotoxins but targets the Shaker potassium channel (35). Structural features of CVID that have not been described for other -conotoxins are the presence of two hydrogen bonds from the NH protons of Lys 10 and Leu 11 to the CϭO oxygen atoms of Gly 22 and Thr 23 , respectively. It is likely that these enhance the stability of loop 2 in CVID compared with other -conotoxins. Importantly, the backbone of Tyr 13 is stabilized in the ␣ L conformation, with the 1 side chain torsion angle at Ϫ60°. NMR data confirming this conformation for Tyr 13 in CVID include the presence of a strong intraresidue NH i -H␣ i NOE, together with a weaker H␣ iϪ1 -NH i and a 3 J NH-H␣ coupling constant of 7 Hz (36). Attempts to define the conformation of Tyr 13 in other -conotoxins have been ambig-

TABLE III Potency (nM) of CVID and MVIIA to inhibit Ba 2ϩ current through central (␣ 1B-d ) and peripheral (␣ 1B-b ) splice variants of the rat N-type VSCC in the presence or absence of rat ␤3 (n ϭ 3-5)
Subunit combination CVID MVIIA uous, and indeed Tyr 13 may adopt an averaged conformation in other -conotoxins. The enhanced structural stability of CVID is further reinforced by a comparison of the H␤ shifts of CVID and MVIIA (Fig. 9C). The H␤ protons of Cys 20 are degenerate in MVIIA indicative of disulfide bond isomerization, whereas the corresponding protons in CVID have disparate shifts indic-ative of an absence of any disulfide bond isomerization. The H␤ shifts differences of Cys 8 , Tyr 13 , and Asp 14 are also significantly greater in CVID than in MVIIA. This trend is followed over the temperature range 280Ϫ298 and the pH range 3Ϫ6 (data not shown). DISCUSSION By using assay-directed fractionation, we have isolated four new -conotoxins from C. catus, named -conotoxins CVIA-D. These four conotoxins were chemically synthesized and folded to be indistinguishable from the native peptides. C. catus is the sixth piscivorous cone snail to evolve mammalian active -conotoxins for prey capture. Analogous to other -conotoxins identified to date, CVIA-D have three disulfide bonds, fall within the mass range 2500Ϫ3600 Da (4), and are relatively hydrophilic. Inspection of the sequences in Table I reveals a conserved cysteine framework and three conserved residues Gly 5 , Tyr 13 , and Ser 19 (CVID numbering) among -conotoxins. Homology at loops 1-4 varies from 0 to 60% sequence identity among these -conotoxins, with the largest difference occurring in loop 4 of CVID. Cloning and sequencing of mRNA transcripts from the venom ducts of two specimens of C. catus confirmed CVIA and CVID are expressed in this tissue. Our inability to isolate mRNA transcripts encoding CVIB and CVIC was most likely associated with the relatively low abundance of these peptides. Variation in the levels of specific peptides between individual C. catus could also be involved. 2 Isolation of cDNA templates for CVIB and CVIC may require the analysis of additional C. catus specimens or the use of additional "PCR markers" that are more specific for the mature peptide sequences.
CVIA-D each fully displaced 125 I-GVIA and 125 I-MVIIC binding to rat brain membrane. CVID, GVIA, and MVIIA were similarly potent at rat brain N-type VSCCs. However, CVID was 100-fold more selective than MVIIA for N-type over P/Qtype VSCCs, and 125 I-CVID bound to 30% fewer sites in rat brain membrane than 125 I-MVIIA. Similar differences in B max were not observed in saturation binding of 125 I-CVID and 125 I-MVIIA to human brain membrane, suggesting that pharmacological differences exist in the N-type VSCCs present in rat and human brain. Further studies are required to determine the origin of these differences. CVIAϪD were reversible inhibitors of peripheral N-type VSCCs involved in neurally evoked transmitter release in rat vas deferens. Inhibition of neurotransmitter release was positively correlated with the displacement of 125 I-GVIA binding, consistent with N-type VSCCs being the major source of Ca 2ϩ influx contributing to neurotransmitter release in the rat vas deferens stimulated at low frequencies (37)(38)(39)(40). A correlation between inhibition of Ca 2ϩ influx and ligand displacement at the N-type VSCC was shown for a series of MVIIA and TVIA analogues (41) but not for a series of GVIA analogues (33). Potency estimates obtained using the vas deferens were lower than estimates obtained from radioligand displacement studies (Table II), consistent with previous observations (37,42). The relatively high levels of Ca 2ϩ in the physiological solution for the vas deferens compared with the binding assays is likely to contribute to this difference in potency, since Ca 2ϩ is a non-competitive inhibitor of -conotoxin binding (13). CVID had no effect on contractions of the vas deferens induced by exogenously applied ATP or norepinephrine or on voltage-sensitive sodium channels in rat autonomic neurons. 2 In Xenopus oocytes, CVID and MVIIA inhibited the inward current through central (␣ 1B-d ) and peripheral (␣ 1B-b ) splice variants of the rat N-type VSCC. In the presence of rat ␤ 3 subunit, chosen because it assembles with ␣ 1B (43) and is widely distributed in rat brain (44), MVIIA and CVID had 55) comprising both 3 families that were calculated with a force constant of 40 kcal/mol applied to the three disulfide bridge dihedral angles C␤ i -S␥ i -S␥ j -C␤ j (B). The disulfide bridge Cys 8 -Cys 20 (purple), and the side chains of important binding residues Arg 10 (Lys 10 in CVID) (blue), Leu 11 (yellow), and Tyr 13 (orange) are identified. The position of loop 4 is indicated in A and B. C shows a comparison of differences in H␤ shifts for selected residues of CVID (green) and MVIIA (gray, 30). Note that CVID has greater separations of H␤ shifts, particularly at Cys 8 , Tyr 13 , Asp 14 , and Cys 20 . similar potencies. However, the potency of MVIIA and CVID increased at ␣ 1B-d and ␣ 1B-b expressed in the absence of rat ␤ 3 . This increase was most pronounced for CVID at ␣ 1B-d (up to 540-fold), intermediate for CVID and MVIIA at ␣ 1B-b (ϳ10fold), and least pronounced for MVIIA at ␣ 1B-d (3-fold). Previous studies have shown that the ␤ subunit can influence the affinity of local anesthetics (45) and spider toxins (46) for ␣ 1A (P/Q-type) calcium channels expressed in oocytes and mammalian cells. Oocytes express low levels of an endogenous ␤ subunit (␤ 3xo ) with high homology to rat ␤ 3 that is essential for expression of functional channels, and overexpression of either of these ␤ 3 subunits causes the same facilitation of VSCC current activation and inactivation (47). We propose that expression of excess ␤ 3 can differentially modify -conotoxin pharmacology at central and peripheral splice variants of the N-type VSCC, presumably by binding to an intracellular domain of ␣ 1B to cause an extracellular conformational change that reduces -conotoxin binding.
The alternatively spliced ␣ 1B-d and ␣ 1B-b differ at extracellular loops IIIS3-S4 (ϩSFMG and ϪSFMG, respectively) and IVS3-S4 (ϪET and ϩ ET, respectively), indicating that relatively small changes in the sequence of ␣ 1B not only influence channel kinetics (insertion of ET in ␣ 1B-b slows activation kinetics (48)), but modify -conotoxin pharmacology. Since -conotoxins bind preferentially to the inactivated state than the resting state of N-type calcium channels (49), the observed effect is not through excess ␤ subunit that would act to increase the proportion of channels in the inactivated state (47). This auxiliary subunit effect may occur through the action of excess ␤ subunit at a second site that is present at the C terminus of ␣ 1A , ␣ 1B , and ␣ 1E (50), but a differential effect of rat and Xenopus ␤ 3 on the pharmacology of -conotoxins may also be involved. The potency of Aga IVA and funnel web spider toxin were reduced ϳ10-fold when ␣ 1A was coexpressed in mammalian cells with ␤ 3 or ␤2, rather than with ␤1b (46). The extent to which -conotoxin pharmacology at central and peripheral splice variants of ␣ 1B is influenced by different combinations of the five ␣2/␦ subunits and the four ␤ subunits requires further investigation, given the heterogeneity of ␣ 1B /auxiliary subunit combinations in the brain (43). The differential effect of ␤ subunits on -conotoxin pharmacology at central and peripheral splice variants of ␣ 1B may lead to the development of -conotoxins with selectivity for specific forms of the N-type VSCC present in different neural pathways. However, the precise location of specific combinations of ␣ 1B splice variants and different ␤ subunits have not yet been determined.
The ability of CVID to discriminate among different forms of the N-type VSCC may also have therapeutic implications. In a Freund's complete adjuvant model of inflammatory pain in rats, chronic intrathecal administration of CVID (AM336; 0.36 nmol/h) produced sustained antinociception with mild behavioral side effects (mean scores 0.5-1 on a scale of 7) over a 7-day infusion period. In comparison, MVIIA (Ziconotide; 0.1 for first 24 h and 0.13 nmol/h for next 6 days) produced similar initial levels of antinociception that returned to base line by day 6, whereas levels of motor side effects (mean scores 3-4) were greater and maintained throughout the 7-day infusion period (52). 3 These effects of MVIIA and CVID were completely reversed on ceasing conotoxin infusion. CVID has recently gained approval to enter clinical trials for the treatment of severe pain. 1 H NMR studies reveal that CVID adopts a unique fold, with loop 4 curved toward loop 2 to create a globular surface. In contrast, loop 4 is oriented parallel to loop 2 in MVIIA and MVIIC and away from loop 2 in GVIA (18,30,53,54). The presence of two hydrogen bonds between loops 2 and 4 in CVID favor loop 4 in this orientation, whereas hydrogen bonds between loops 2 and 4 have not been reported previously for GVIA, MVIIA or MVIIA. Early structural studies on MVIIA (30) raised the possibility of loop 2 instability, based on the observation of broadened lines in this region. This was supported by a recent study on the dynamics of MVIIA (55), where it was suggested that the motion of loop 2 was facilitated by isomerization of the disulfide bridge between Cys 8 and Cys 20 . In that study, the introduction of a force constant of 40 kcal mol Ϫ1 into the empirical force field restrained the 3 angle of this disulfide bridge to Ϯ90°resulting in two closely related families of structures (55). No such isomerization is evident in CVID, where a corresponding force constant of only 10 kcal/mol yielded a conformation of ϩ90°in all structures. In addition, the 1 angles at Cys 20 for MVIIA (ϩ180°) and CVID (Ϫ60°) differed significantly, as do the magnitude of H␤ 2 /H␤ 3 shift differences of AMX residues (Fig. 9C). The greater H␤ shift differences of Cys 8 , Cys 20 , Tyr 13 , and Asp 14 in CVID indicate that the side chains of these residues undergo less conformational exchange than MVIIA. The improved stability of this disulfide bridge in CVID together with the interloop hydrogen bonds may act to restrain loop 2 motion, providing an entropic advantage for binding. This overall stability of CVID is also likely to contribute to its selectivity, since loop 2/4 combinations contain many of the primary determinants for -conotoxin selectivity between N-type and P/Q-type VSCCs (18).
CVID and two closely related families of MVIIA (55) structures are compared to highlight important differences in the orientation of loops 4 and 2 (Fig. 9). Although Tyr 13 occupies similar spatial positions, the positions of the side chains of the important binding residues 10 and 11 appear reversed in the two peptides. According to Atkinson et al. (55), motion in MVIIA results in time-averaged NOEs that ultimately produce flaws in the resulting structure. Such motion does not appear to be present in CVID, since no line broadening is observed in loop 2 or elsewhere in the peptide. This, together with the increased number of distance restraints, particularly in the loop 2 region, ensures that the structure of CVID is represented by a single, well defined conformation. Interestingly, CVID lacks the secondary binding residues present in loop 4 of MVIIA (Arg 21 ) (34) or GVIA (Lys 24 and Tyr 22 ) (33) but utilizes a new -conotoxin/VSCC interaction through the relatively exposed Val 24 . 2 These structural features presumably act in concert to influence the selectivity of CVID, as well as providing a well defined template for pharmacophore development and the design of peptidomimetic inhibitors of N-type VSCCs.