ShK-Dap, a Potent Kv1.3-specific Immunosuppressive Polypeptide*

From the Departments of Physiology & Biophysics, and Microbiology and Molecular Genetics, University of California, Irvine, California 92697-4560, ‡Bachem Bioscience, Incorporated, King of Prussia, Pennsylvania 19406, §Biomolecular Research Institute, Parkville 3052, Victoria, Australia, the iDepartment of Pharmacology and Therapeutics, College of Medicine, University of Florida, Gainesville, Florida 32610, the **Department of Respiratory, Inflammatory and Neurological Disorders, Zeneca Pharmaceuticals, Wilmington, Delaware 19850, and the ¶Department of Applied Physiology, University of Ulm, 89081 Ulm, Germany

Human T lymphocytes express a unique voltage-gated potassium (Kv) 1 channel encoded by the Kv1.3 gene (1). A homotetramer of Kv1.3 subunits forms the functional channel in T lymphocytes (1). Earlier studies showed that structurally dissimilar blockers of this channel suppressed mitogen-induced [ 3 H]thymidine incorporation and interleukin-2 production by T lymphocytes (1)(2)(3). More specific, high affinity blockers discovered in recent years have demonstrated convincingly that Kv1.3 blockers depolarize the T-cell membrane and attenuate the calcium signaling pathway that is vital for lymphocyte activation (1, 4 -9). Although Kv1.3 is found in B lymphocytes, macrophages, osteoclasts, platelets, and the brain, only in T lymphocytes does Kv1.3 channel activity seem to dominate the membrane potential (1,8). The critical role of Kv1.3 during T-cell activation, coupled with its functionally restricted tissue distribution, has stimulated a search for potent and selective Kv1.3 antagonists for potential use as immunosuppressants (e.g. see Refs. 8 and 9).
Many potent polypeptide inhibitors of Kv1. 3 have been isolated from scorpion venom. These polypeptides adopt well defined conformations constrained by 3 or 4 disulfide bonds and bind with extremely high affinity to a shallow vestibule at the external entrance to the Kv1.3 pore (10,11). The most selective of these, margatoxin (MgTX), suppresses T-lymphocyte activation in vitro and is immunosuppressive in vivo (9), suggesting the possibility of using MgTX as an injectable immunosuppressant. However, MgTX potently blocks the closely related Kv1.1 and Kv1.2 channels (12,13), which are expressed in the brain, peripheral nerves, and heart (14), raising concerns about potential cardiac and neuronal toxic side effects. Extensive efforts are therefore ongoing to identify other more selective and potent peptide and non-peptide inhibitors of Kv1. 3.
Recently, a 35-amino acid-residue polypeptide (ShK) from the sea anemone Stichodactyla helianthus was shown to block the Kv1.3 channel at low picomolar concentrations (15,16). Like scorpion toxins, ShK has a well defined conformation constrained by three disulfide bonds, minimizing possible structural changes upon its binding to the channel. However, the structure of ShK is significantly different from those of scorpion toxins (17,18). Using alanine-scanning mutagenesis, the channel-binding surfaces of ShK (15,16) and its closely related homologue, BgK (19), have been determined. Despite differences in the scaffolds, the sea anemone and scorpion toxins share a conserved diad of residues that is essential for block of potassium channels (16,19). This diad consists of a critical lysine (Lys 27 in the scorpion toxins and Lys 22 and Lys 25 in ShK and BgK) and a neighboring aromatic residue (Tyr 36 in ChTX, Tyr 23 in ShK, Tyr 26 in BgK) separated by ϳ7 Å (19). Lys 27 , in scorpion toxins, couples with the tyrosine (Tyr 400 in Kv1. 3,Tyr 445 in Shaker) in the potassium channel selectivity filter (11,20). A better understanding of the interactions between ShK and the Kv1.3 channel may guide the design of specific ShK mutants with the potential to be used clinically as immunosuppressants. Here, we describe a mutant polypeptide that shows selectivity for Kv1.3, inhibits T-cell activation in vitro, and is minimally toxic in vivo.

MATERIALS AND METHODS
Peptide Synthesis-Fmoc-amino acid derivatives were obtained from Bachem A.G. (CH-4416 Bubendorf, Switzerland). Solid-phase assembly was initiated with Fmoc-Cys(Trt)-2-chlorotrityl resin to minimize potential racemization of the C-terminal Cys residue (21). Automated stepwise assembly was carried out entirely on an ABI-431A peptide synthesizer (Applied Biosystems, Foster City, CA). Fmoc-Dap(t-butyloxycarbonyl) was substituted in place of Lys 22 in the assembly of the polypeptide. The Dap 22 -substituted polypeptide was cleaved and deprotected with reagent K (22) containing 5% triisopropylsilane. The ShK-Dap 22 analogue was solubilized, oxidized, and purified by reverse phase-high pressure liquid chromatography using the same method described previously for other ShK analogues (15). High pressure liquid chromatography-pure fractions were pooled and lyophilized. The structure and purity of the peptides were confirmed by reverse phase-high pressure liquid chromatography, amino acid analysis, and electrospray ionization-mass spectroscopy analysis. All other ShK analogues were synthesized, purified, and characterized as reported previously (15,16). Samples were weighed and adjusted to account for peptide content prior to bioassay.
Reagents-Cell lines stably expressing mKv1.1, rKv1.2, mKv1.3, hKv1.5, and mKv3.1 (7,12) were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and G418 (1 mg/ml). Human IK Ca channels were studied in activated human T cells as described previously (7). All the mKv1.3 mutants and mKv1.4 used in this study have been described previously (7, 10 -12). Rat Kv1. 6  I-ChTX (25 pM, 2200 Ci/mmol) was then added, and the reaction was allowed to proceed at 22°C for a further 20 min. The reaction was stopped by harvesting the membranes onto Packard GF/C Unifilter 96-well filter plates and by washing twice rapidly with ice-cold wash buffer (200 mM NaCl, 20 mM HEPES, pH 8.0). The filter plates were dried overnight, scintillation mixture (Packard Microscint-20; Packard Bioscience, Meriden, CT) was added, and the plates were counted in a scintillation counter (Packard Top Count). Specific binding was determined by subtracting nonspecific binding (defined by 100 nM unlabeled ChTX) from total binding. This binding assay was protein dependent, saturable (B max ϭ 916 Ϯ 37 fmol/mg protein), and of high affinity (K d ϭ 23 pM).
Mouse Acute Toxicity Determinations-Several doses of ShK or ShK-Dap 22 were administered by intravenous tail vein injection into 15-20-g Swiss-Webster male mice. Loss of righting ability (paralysis) was assessed over a 4-h period.
Activation of Human T Cells by Anti-CD3 Antibody-Mononuclear cells (MNCs) were isolated over a Ficoll-Hypaque density gradient (Sigma). The isolated MNCs were incubated (37°C, 5% CO 2 ) for Յ2 days in RPMI 1640 supplemented with 10% fetal calf serum, 1 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. The assay was conducted in a 96-well plate by first adding monoclonal anti-CD3 and various polypeptide concentrations to wells in triplicate. Anti-CD3 was titrated to dilutions that produced a 4 -25-fold increase in [ 3 H]thymidine incorporation. MNCs were resuspended in fresh media and then added to wells at a final concentration of 0.3 ϫ 10 6 cells/well (final volume 200 l). For determination of background uptake, anti-CD3 was not added to six wells in each plate, and the average [ 3 H]thymidine uptake from these wells subtracted from wells containing anti-CD3. Plates were incubated for 48 h, and [ 3 H]thymidine was added during the last 6 h. The contents of the wells were harvested onto glass fiber filters (Packard GF/C unifilters) using a multi-well harvester, and cells were lysed with water. Filters were air-dried overnight. Scintillation mixture (Packard Microscint-20) was added, and [ 3 H]thymidine incorporation was measured by counting in a scintillation counter.
Oocytes-cRNA was transcribed in vitro and injected into oocytes (Xenopus laevis purchased from NASCO, Fort Atkinson, WI) (10,11). Potassium currents were measured at room temperature using the two-voltage clamp technique (10,11), and data were analyzed using pClamp software (version 5.5.1, Axon Instruments, Burlingame, CA). Whole oocytes were held at -100 mV and depolarized to ϩ40 mV over 500 ms; time between pulses was 30 s. Capacitative and leak currents were subtracted prior to analysis using the P/4 procedure. The dissociation constant was calculated assuming a 1:1 binding of toxin to Kv1.3 as described (10,11).
Mutant Cycle Analysis-This method provides a simple way to evaluate the strength of interaction between any two pairs of protein residues (25). For each mutant cycle, we measured the potency (K d ) of ShK and each of its mutants on Kv1.3 and each of the channel mutants.
Three positively charged ShK residues, Lys 9 , Arg 11 , and Lys 22 , were replaced individually by the neutral residues alanine (Ala) or norleucine (Nle). In the mutant cycle studies, the wild-type polypeptide was compared against the corresponding neutral polypeptide. In addition, we replaced Lys 22 with the non-natural positively charged residues, diaminopropionic acid (Dap) or ornithine (Orn). These residues vary in their side chain lengths (Dap ϭ 2.5 Å, Orn ϭ 5.0 Å, Nle ϭ 5.0 Å, Lys ϭ 6.3 Å). The positively charged position-22 mutants (along with Lys 22 ) were treated as wild-type in the mutant cycle analysis and compared against the mutant polypeptide containing the neutral residue Nle 22 .
Four residues of the Kv1.3 channel were selected for mutagenesis: His 404 , Asp 402 , Tyr 400 , and Asp 386 . His 404 was replaced with the hydrophobic residue Val, and Asp 386 was replaced with Lys. As substitutions at positions 400 and 402 result in nonfunctional channels, we generated dimeric constructs containing one wild-type subunit and one mutant subunit. The resulting tetramers would be composed of (Asn 402 2 ,Asp 402 2 ) and (Val 400 2 ,Tyr 400 2 ). All of these channel mutants and the dimeric constructs have been used previously in mapping studies with kaliotoxin (10,11).
The mutant cycles for Kv1 The change in coupling energy, ⌬⌬G, for a given pair of ShK-Kv1.3 residues and their mutants was calculated using the formula ⌬⌬G ϭ kTln⍀, where ⍀ is a dimensionless value given by the formula For ⍀ values Ͻ1 the inverse was taken (10,11). Schreiber and Fersht (25) reported that ⌬⌬G values of Ն0.5 kcal⅐mol Ϫ1 (2 error) correspond to an inter-residue distance of Յ5 Å, and higher ⌬⌬G values match shorter inter-residue distances. We used a ⌬⌬G value of Ͼ0.8 kcal⅐mol Ϫ1 as an indicator of a close interaction (Յ5 Å) between a pair of peptide and channel residues. Although high ⌬⌬G values indicate tight interactions, residues that are physically close may be energetically "silent" and may not be detected by this method (26).
All of the peptide-mapping studies were performed on channels expressed in Xenopus oocytes, whereas the studies described in Fig. 1 were performed on channels expressed in mammalian cells. In general, there was good correspondence between the K d values measured on channels expressed in mammalian cells and oocytes, although ShK-Dap 22 was about 6-fold more potent in the oocyte system (K d ϭ 3.3 Ϯ 1.9 pM, n ϭ 12) compared with mammalian cells (see Fig. 1).
Structure Determination-Two-dimensional 1 H NMR spectra were recorded at 600 MHz on a ϳ2 mM solution of synthetic ShK-Dap 22 in 90% H 2 O, 10% 2 H 2 O (v/v) or 100% 2 H 2 O at pH 4.9 and 293 K, as described (17,27), but with water suppression using the Watergate scheme and a 3-9-19 selective pulse (28). Spectra were also recorded at 278 K in an attempt to sharpen backbone amide resonances from Ser 2 , Cys 3 , Met 21 , Dap 22 , and Tyr 23 . Chemical shifts for Dap resonances in the synthetic peptide GlyGlyDapGlyGly-OH were measured from onedimensional and total correlation spectroscopy spectra at 293 and 298 K in 90% H 2 O, 10% 2 H 2 O at pH 5.0, using 2,2-dimethyl-2-silapentane-5sulfonate as internal standard.
A figure (Fig. S1) included in the "Appendix" summarizes the sequential assignments, slowly exchanging amides, backbone coupling constants, and medium-range NOEs for ShK-Dap 22 , together with a table of 1 H chemical shifts (Table SI). Methods for obtaining distance and angle restraints, generating structures in DYANA (29), and refining the structures by restrained simulated annealing and restrained energy minimization in X-PLOR (30) were as described previously (17,27). The final NMR restraint list (from which values redundant with the covalent geometry had been eliminated by DYANA) consisted of 82 intraresidue, 82 sequential, 105 medium-range (͉i-j͉Ͻ5), and 74 long-range (͉i-j͉Ն5) upper bound restraints, no lower bound restraints, and 30 backbone and 6 side chain dihedral angle restraints. Of the 50 CHARMM-minimized structures, the best 25 were chosen on the basis of their stereochemical energies (i.e. excluding the electrostatic term). Of these, the best 20 were chosen on the basis of their Ramachandran plots and the consistency of their secondary structures with the NMR restraints. These structures and the NMR restraints on which they were based have been deposited in the Protein Data Bank (31) (code 1bei). Structures were analyzed using Insight II (Molecular Simulations Inc., San Diego) and MOLMOL (32). Hydrogen bonds were identified in MOLMOL using a maximum C-N distance of 2.4 Å and a maximum deviation of 35°from linearity.
Model of Kv1. 3 and ShK Docking-To create a model of the pore and vestibule of Kv1.3 (residues 380 -410), we relied heavily on the recent crystal structure of the bacterial K channel, KcsA (33,34), and on a molecular model of the Shaker channel (35). Residues Phe 425 , Lys 427 , Thr 449 , Gly 452 , Phe 453 , and Trp 454 from the Shaker model (35) were changed to the corresponding Kv1.3 residues Gly 380 , Asn 382 , His 404 , Thr 407 , Ile 408 , and Gly 409 , respectively, using Insight II. Modifications to the backbone and side chain dihedral angles were then made so that the local and global structure of the channel model better resembled the corresponding region of the KcsA channel (33). Following conjugate gradient minimization of the model using Discover (MSI), the closestto-average ShK structure (17) was juxtaposed with the channel so as to preclude steric contact between the two. The backbone atoms (N, C ␣ , and C) of Kv1.3 were fixed in space during the simulation, whereas the backbone fold of ShK was maintained by 16 medium-range and 3 long-range distance constraints. Inter-molecular distance constraints were added to the peptide-channel complex in conjunction with a 50 kcal mol Ϫ1 force constant in Discover so as to reflect data from mutant cycle analyses (see "Results"), with Lys 22 N (ShK) being kept within 5 Å of Tyr 400 C ␥ from each of the four Kv1.3 subunits and Arg 11 C being kept within 5 Å of a single His 404 N ␦1 . A lower limit of 6 Å was maintained between Arg 11 C and Asp 402 C ␥ to restrict any interaction between these two residues, which show no coupling (see "Results"). The complex was energy minimized using 10,000 steps of conjugategradient minimization, and then a 250-ps molecular dynamics simulation was performed in vacuo at 300 K with a 1-fs time step, a distancedependent dielectric, and a 15-Å non-bonded interaction cut-off. After equilibration of the complex, the conformation with the lowest van der Waals repulsive energy was chosen for further energy minimization, carried out as above.  Table I) and with 1:1 stoichiometry (Fig. 1B). Similar results were obtained for block of Kv1.3 channels in human peripheral blood T cells (data not shown).

ShK, a Potent Blocker of the Kv1.3 Channel in T Lympho
The ShK polypeptide inhibited 125 I-ChTX binding to its receptor in the external vestibule of hKv1.3.
10,500 Ϯ 900 (2) mKv1. 7 11,500 Ϯ 2340 act with a receptor in the external vestibule of the Kv1.3 channel that is identical or overlapping the receptor surface for the scorpion toxins.
To evaluate the selectivity of ShK for Kv1.3, we tested it against a panel of eight K ϩ channel targets (Table I). All the channels tested, with three exceptions, are Ͼ100-fold less sensitive to block by ShK compared with Kv1.3 (Table I). ShK, however, blocks mKv1.1, a cardiac and neuronal channel, with roughly the same potency as it does mKv1.3 (Fig. 1, C and D), and two other channels, mKv1.4 and rKv1.6, are also blocked in the picomolar range (Table I). Thus, ShK is not selective for Kv1.3, necessitating a search for an ShK mutant that might be more specific.
Identifying Polypeptide-Channel Interactions-Determination of the docking configuration of ShK in the Kv1.3 channel might help identify ShK mutants that exhibit Kv1.3 specificity. Guided by the solution structure of the ShK polypeptide (17) and by knowledge of the geometry of the pore and vestibule gained from studies with scorpion toxins (10,11,20), we generated complementary mutants of ShK and Kv1.3. Utilizing double mutant cycle analysis, we identified specific pairs of ShK-Kv1.3 interactions.
Three residues in ShK were chosen for mutagenesis: Arg 11 and Lys 22 on the surface, thought to interact with Kv1.3, and Lys 9 on the opposite surface (15)(16)(17)(18). We focused on four channel residues (His 404 , Asp 402 , Tyr 400 , and Asp 386 ) that have been shown previously to be important for scorpion toxin binding (10,11). His 404 (KcsA-Tyr 82 ) lies at the outer entrance to the ion conduction pathway (10,33). The ring of four His 404 residues is unique to Kv1.3, and compounds that target this ring might be selective for the lymphocyte channel (1, 7, 8). The highly conserved Tyr 400 (KcsA-Tyr 78 ) and Asp 402 (KcsA-Asp 80 ) in the critical signature sequence (GYGD) form part of the ion selectivity filter and couple with Lys 27 in the scorpion toxins (11,20,36). Asp 386 (KcsA-Arg 64 ) lies ϳ10 -14 Å from the center of the pore and interacts with Arg 24 in kaliotoxin and agitoxin-2 and with Arg 25 in charybdotoxin (10,36).
Examples of two mutant cycles are presented in Using a molecular model of Kv1.3 based on the known crystal structure of the KcsA channel (33), we used restrained molecular dynamics simulations to dock the ShK peptide into the channel (Fig. 4). In this configuration, Lys 22 protrudes into the pore, but its side chain does not make direct contact with the critical Tyr 400 and Asp 402 in Kv1.3 (Fig. 4); the corresponding residues in the KcsA channel face away from the channel pore (33). This docking configuration places Arg 11 in close proximity to His 404 in one channel subunit, and two of the remaining His 404 residues in the tetramer lie in close proximity to peptide residues Met 21 and Arg 29 . In addition, our model places Arg 29 near Asp 386 in the channel subunit adjacent to that which interacts with Arg 11 . Two lines of evidence support this placement. First, introduction of lysine at channel position 386 (D386K) causes a significant reduction in peptide potency ShK. Second, we also detect coupling between Asp 386 and Arg 29 (⌬⌬G ϭ 0.88 kcal⅐mol Ϫ1 , Fig. 3 legend).
This docking configuration, which resembles that of agitoxin-2 docked in the KcsA channel (36), was used to guide the identification of ShK mutants that exhibit Kv1.3 specificity.  Fig. 3.
ShK, ShK-Dap 22 , and MgTX Inhibit Human T Cell Activation with Similar Potency-We compared the ability of ShK, ShK-Dap 22 , and MgTX to suppress anti-CD3-stimulated [ 3 H]thymidine incorporation by human peripheral blood T cells. All three polypeptides inhibited mitogen-stimulated [ 3 H]thymidine incorporation to a maximum level of ϳ50 -60% (Fig. 6). However, the midpoint of inhibition (IC 50 ) for each toxin was below 500 pM, in keeping with their affinity for the Kv1.3 channel. Consistent with our results, an earlier study reported that peripheral blood T cells isolated from mini-pigs during intravenous MgTX infusion never showed more than a ϳ60% inhibition of mitogen-stimulated [ 3 H]thymidine incorporation in an ex vivo proliferation assay (9).
ShK-Dap 22 Does Not Exhibit Acute Toxicity following Intravenous Injection into Rodents-As an initial evaluation of the toxicity of ShK and ShK-Dap 22 , mice (n ϭ 5 in each case) were injected intravenously with each polypeptide. ShK toxin displayed a remarkably low toxicity when injected into mice, the median paralytic dose being approximately 0.5 mg per 20 g mouse, or 25 mg/kg body weight. ShK-Dap 22 was even less toxic; a 1.0-mg dose failed to cause any symptoms (hyperactivity or seizures) or mortality, and the median paralytic dose was ϳ200 mg/kg body weight.
Solution Structure of ShK-Dap 22 and Comparison with the  (Table II). Moreover, 91% of the residues havevalues in the generously allowed regions of a Ramachandran plot, Gly 33 being the only residue with a positive angle. The angular order parameters (S) (37) of the final 20 structures indicate that residues 2-21 and 23-35 are well defined locally, with S , Ͼ 0.8 (Fig. 7). Backbone r.m.s. difference values (Fig.  7B) also show that the structure is well defined over most of the molecule. Mean pairwise r.m.s. differences calculated over the backbone heavy atoms (N, C ␣ , C) and all heavy atoms, respectively, of the whole molecule were 0.63 Ϯ 0.15 and 1.41 Ϯ 0.23 Å, and for the well defined region (residues 2-21 and 23-35) 0.51 Ϯ 0.13 and 1.04 Ϯ 0.14 Å.
The main secondary structure elements of ShK-Dap 22 (Fig.  8, A and B) are two short ␣-helices encompassing residues 14 -19 and 21-24. The N terminus adopts an extended conformation up to residue 8, where a pair of interlocking turns commences; in 25% of the structures this pair of turns satisfies the criteria for a 3 10 -helix centered on residues 9 -10 (with an 1138 hydrogen bond found in all 20 structures). There is also a short stretch of helix between residues 29 and 32 (with a 32328 hydrogen bond in all 20 structures) that is a mixture of ␣and -helix. Backbone hydrogen bonds associated with these secondary structural elements account for many of the slowly exchanging backbone amide protons observed by NMR following dissolution in 2 H 2 O. Several other backbone amide protons found to be slowly exchanging were shielded from solvent. The main secondary structure elements of the two molecules are the same, but ShK-Dap 22 also has a recognizable helix near the C terminus involving residues 29 -32. In ShK, this region has a similar structure but does not satisfy the criteria for a helix. The only appreciable differences between the backbone dihedral angles of the two structures occur at Pro 8 (), Thr 31 (), and the three C-terminal residues (). Fig.  8C, the structures of ShK-Dap 22 and ShK are aligned over N, C ␣ , C, and C ␤ of residues 11-23, which includes the most important residues for potassium channel binding (15,16) (Figs. 3 and 4). In this view, the side chains of Arg 11 and Tyr 23 have similar orientations, although they have moved closer together. The distances from Tyr 23 C ␥ to Arg 11 C ␥ are 3.9 Ϯ 0.2 and 7.4 Ϯ 0.7 Å, respectively, in ShK-Dap 22 and ShK. The functionally more important distances from the centroid and phenolic oxygen of Tyr 23 to Arg 11 C are, respectively, 4.9 Ϯ 0.

DISCUSSION
In this study, we pursued three overlapping goals. First, using the ShK peptide as a structural template and applying thermodynamic mutant cycle analysis, we determined the spatial proximity of eight pairs of ShK and Kv1.3 residues. These data, along with those obtained from earlier mapping studies with scorpion toxins (10,11), guided our docking of ShK into the channel. This docking configuration might provide insights into the interaction of other members of this novel structural class of sea anemone peptides (e.g. BgK) and potassium channels. Second, we used the docking model to identify the Kv1. Peptide Toxins As Candidate Immunosuppressive Agents-The Kv1.3 channel is widely regarded as a novel therapeutic target for T-cell immunosuppression (e.g. Refs. 1,8). Due to its restricted tissue distribution and unique role in regulating lymphocyte function, selective and potent blockers of this channel might not have the toxic side effects of currently used drugs such as cyclosporin, FK-506, and rapamycin (1,8). Kv1.3-specific antagonists may therefore be therapeutically useful immunosuppressants. Several scorpion toxins potently and reversibly block this channel with IC 50 values in the low picomolar to nanomolar range and with 1:1 stoichiometry (10,11). By blocking Kv1.3, these polypeptides attenuate the calcium signaling response and inhibit mitogen activation of T cells in vitro (4 -9). The most potent and selective of these, MgTX, has also been shown to effectively suppress delayed-type hypersensitivity and alloimmune responses in vivo in micro-and mini-pigs, despite its inability to completely suppress T-lymphocyte activation in vitro (9). However, MgTX also potently blocks the closely related channels Kv1.1 and Kv1.2 (12,13), which are expressed in the brain and peripheral neurons (14), and is therefore potentially toxic. An equally potent but more selective peptide blocker of Kv1.3 might not exhibit these side effects. The structurally defined peptidic inhibitor, ShK-Dap 22  in Kv1.3 binding (Fig. 8). Are these differences significant, or do they reflect differences between the number and distribution of NMR-based restraints in key regions in the structure (Fig. 7A)? The 1 H chemical shifts of the two molecules are very similar, the only differences Ͼ0.1 ppm being for Met 21 NH (⌬␦ 0.25 ppm), Dap 22 , and residues 26 -28 (Table SI and Fig. S2 in "Appendix"). The 3 J HNC␣H coupling constants, which are dependent on backbone angles, also differed by Ͼ1Hz for residues 26, 27, and 29 (other residues in this category were 9, 10, 16, and 35). The backbone amide resonance of Dap 22 was not observed, and those of Met 21 and Tyr 23 in ShK-Dap 22 were broader than in ShK. As a result, there were fewer NOEs to these protons (Fig. S3 in "Appendix"), and this region of the structure is not as well defined in ShK-Dap 22 . Part of the reason for the broader Dap 22 NH resonance is that the intrinsic line width is greater, as found in the pentapeptide GlyGly-DapGlyGly; this presumably reflects the proximity of the side chain ammonium group of Dap to the backbone. However, this is unlikely to be the explanation for the flanking residues, suggesting that this region has greater conformational flexibility in ShK-Dap 22 . To confirm the difference between ShK-Dap 22 and ShK, we recorded a 2D NOE spectrum on a mixture of the two at pH 4.7 and 293 K. Resonance overlap prevented any comparison for Tyr 23 , but it was quite clear that the crosspeaks from Met 21 of ShK-Dap 22 were broader and weaker than those of ShK. As the chemical shift of Met 21 NH was also perturbed, it seems that there are some genuine differences in the local structure and dynamics of ShK-Dap 22 around the substituted residue. The backbone amides of ShK-Dap 22 also show slightly faster exchange than those of ShK (although respective rate constants are within a factor of 2), suggesting that the overall structure of ShK-Dap 22 may be slightly more flexible than that of ShK.
In other regions, particularly the N and C termini, the apparent structural differences (Fig. 8) stem partly from the presence of a few NOEs unique to one of the restraint sets. The ShK-Dap 22 structures are better defined than those of ShK at both termini, but it is important to note that there is some flexibility in these regions of both structures and that both may change when bound to Kv1.3. Finally, the close similarity between the structures of ShK-Dap 22 and ShK confirms that the structure of this sea anemone toxin is different from that of the homologous BgK toxin (19) with which it shares 13 residues. BgK contains two longer helices, involving residues 9 -16 and 24 -31, although its overall topology is similar to that of ShK.
Why Does ShK-Dap 22  Concluding Remarks-Although ShK-Dap 22 has the potential to be used clinically, improving its stability and enhancing its plasma half-life are important objectives. Achievement of these goals would be facilitated by knowledge of the docking configuration of ShK-Dap 22 in the Kv1.3 external vestibule. With this information, it might be possible to rationally substitute nonnatural amino acids at key positions in the polypeptide, introduce stabilizers of the toxin's interactive surface, and generate a "minimal" analogue that retains the channel binding surface of fulllength ShK-Dap 22 . A smaller analogue might also increase the oral availability of the compound, thereby enhancing its therapeutic usefulness. In conclusion, we have described a highly potent and selective antagonist of Kv1.3 that might be used for the prevention of graft rejection and for the treatment of autoimmune diseases.