Proc. Nati. Acad. Sci. USAVol. 85, pp. 3329-3333, May 1988Biochemistry

Purification, sequence, and model structure of charybdotoxin, apotent selective inhibitor of calcium-activated potassium channels

(ion-channel blocker/scorpion toxin/sequence homologies/snake neurotoxin)

GUILLERMO GIMENEZ-GALLEGO*t, MANUEL A. NAVIAt, JOHN P. REUBEN§, GEORGE M. KATZ§,GREGORY J. KACZOROWSKI§, AND MARIA L. GARCIA§¶Departments of *Growth Factor Research, tBiophysics, and Membrane Biochemistry, Merck Sharp & Dohme Research Laboratories, P.O. Box 2000,Rahway, NJ 07065

Communicated by Edward M. Scolnick, January 6, 1988

ABSTRACT Charybdotoxin (ChTX), a protein present inthe venom of the scorpion Leiurus quinquestriatus var. he-braeus, has been purified to homogeneity by a combination ofion-exchange and reversed-phase chromatography. Polyacryl-amide gel electrophoresis, amino acid analysis, and completeamino acid sequence determination of the pure protein revealthat it consists of a single polypeptide chain of 4.3 kDa. PurifiedChTX is a potent and selective inhibitor of the -220-pSCa21 -activated K+ channel present in GH3 anterior pituitarycells and primary bovine aortic smooth muscle cells. The toxinreversibly blocks channel activity by interacting at the externalpore ofthe channel protein with an apparentKd of 2.1 nM. Theprimary structure of ChTX is similar to a number of neuro-toxins of diverse origin, which suggests that ChTX is a memberof a superfamily of proteins that modify ion-channel activities.On the basis of this similarity, the three-dimensional structureof ChTX has been modeled from the known crystal structureof a-bungarotoxin. These studies indicate that ChTX is usefulas a probe of Ca2+-activated K+ -channel function and suggestthat the proposed tertiary structure of ChTX may provideinsight into the mechanism of channel block.

High-conductance Ca2+-activated K+ channels have beendescribed in a variety of electrically excitable and nonexcit-able cells (1). These channels provide a pathway by whichcell repolarization can occur after membrane depolarization,and consequently they have been implicated in the regulationof neuroendocrine secretion, in the control of muscle con-tractility, and in a number of other cellular processes.However, to assess the physiological role of Ca2 + -activatedK+ channels and attempt their purification, potent specificinhibitors of these channels are required.The venom of the scorpion Leiurus quinquestriatus var.

hebraeus is known to inhibit a number ofdifferentK + -channelpathways (2, 3). A minor component of the crude venom,termed charybdotoxin (ChTX), was discovered to block re-versibly a large-conductance (-200 pS) Ca2"-activated K+channel from rat skeletal muscle plasma membrane vesiclesthat had been reconstituted into planar lipid bilayers (4).Subsequently, ChTX was also found to inhibit low-con-ductance (=35 pS) Ca2"-activated K+ channels in neuronsfrom Aplysia californica, but not block Na+, Ca2+, transientK+, or delayed rectifying K + channels in this preparation (5).The properties ofChTX have been preliminarily characterized(6). The toxin was reported to be a protein with an apparentmolecular mass of 10 kDa and to inhibit Ca2+-activatedK + -channel function with a Kd of 3.5 nM. The same work alsoreports the amino acid composition of the protein and identi-fication of the N-terminal amino acid. However, some dis-

crepancies were noted in the determination of the molecularmass of ChTX based on amino acid composition and electro-phoretic mobility of the protein, which could be explained byinhomogeneity of the preparation.The present study describes the purification of ChTX to

homogeneity, the biological activity ofthe pure toxin, and thechemical characterization of this peptide in terms of aminoacid composition and sequence. The primary structureuniquely identifies this molecule and reveals its similarity toother toxins of species as phylogenetically distant as snakesand marine worms. Based on these similarities, a tertiarystructural model of ChTX has been generated from thepublished (7) x-ray coordinates of a-bungarotoxin depositedin the Brookhaven Protein Data Bank (8). At present, ChTXis the only agent that has been identified to cause potentselective block ofapamin-insensitive Ca2 + -activatedK + chan-nels, and hence it should be useful as a probe for studying theproperties of these channels. A preliminary report of thesefindings has been made in abstract form (9).

MATERIALS AND METHODS

Materials. Lyophilized venom of the scorpion Leiurusquinquestriatus var. hebraeus was obtained from LatoxanScorpion Farm (Rosans, France). GH3 cells were purchasedfrom the American Type Culture Collection, while primarycultures of bovine aortic smooth muscle were obtained asdescribed (10).

Electrophysiological Analysis. Both GH3 and aortic smoothmuscle cells were grown as described (11) and cultured for2-4 days on 25-mm glass coverslips before use in electro-physiological experiments. Single Ca2l -activated K+-chan-nel currents were monitored in outside-out excised mem-brane patches by conventional patch-clamp procedures (12).The samples to be assayed were added directly to anexperimental chamber in which the microelectrode contain-ing the excised patch was suspended.

Purification of ChTX. Lyophilized scorpion venom (480mg) was gently homogenized in 20 mM sodium borate (pH9.0), clarified by centrifugation at 27,000 x g for 15 min,passed twice through Millex-GV filters (pore size, 0.2 ,um)(Millipore) and loaded onto a Mono-S column (HR5/5;Pharmacia) equilibrated with the same buffer, at a flow rateof 0.5 ml/min. When the optical absorbance at 280 nm of theeluate decreased to within 0.06 optical density units abovethat of the elution buffer, the retained material was elutedwith a linear gradient of NaCl (0.75 M/hr). Fractions of thiscolumn containing ChTX activity were loaded onto an

Abbreviation: ChTX, charybdotoxin.tPresent address: Centro de Investigaciones Biologicas, C.S.I.C.,Velazquez 144, 28006 Madrid, Spain.ITo whom reprint requests should be addressed.

3329

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

3330 Biochemistry: Gimenez-Gallego et al.

Ultrapore RPSC reversed-phase column (Beckman) equili-brated with 10 mM trifluoroacetic acid, and eluted at a flowrate of 0.5 ml/min with a 0-20% linear gradient of isopro-panol/acetonitrile (2:1) over 30 min. Fractions to be assayedfrom this chromatography were made either 350 mM in NaClor 0.5% in bovine serum albumin, lyophilized, and laterreconstituted to their original volume with 20 mM sodiumborate (pH 9.0).

Polyacrylamide Gel Electrophoresis. Samples were heated(1000C, 2 min) with or without dithiothreitol (100 mM) in 3%NaDodSO4/63 mM Tris*HCl, pH 6.2. Gels (1.6 mm thick) ofa continuous concentration (25.7% polyacrylamide; acrylam-ide/bisacrylamide ratio, 37:1) were prepared and run for 24hr at 3.3 V per cm of gel as described (13). Gels were fixedas described (14) except that the first two steps were for 2 hreach with hourly changes of solution and the glutaraldehydetreatment was also carried out for 2 hr. Fixed gels were silverstained (15).Amino Acid Analysis and Extinction Coefficient Determina-

tion. The optical absorbance spectrum of protein sampleseluted from the reversed-phase column was digitized in aHewlett-Packard 8450A UV/VIS spectrophotometer. Ali-quots (o10-3 optical absorbance units at 280 nm) wereanalyzed for their amino acid composition as described (16)and the protein content was correlated with the recordedabsorbance.Amino Acid Sequence Determination. Pure ChTX was

alkylated, digested with Staphylococcus aureus V8 protease,and sequenced as described (17, 18). Endoproteinase Lys-C(Boehringer Mannheim GmbH) digestion was carried out at370C for 24 hr in 20 mM sodium phosphate (pH 7.6) at a ratioof 55 pFg of ChTX per enzyme unit. Phenylthiohydantoin-derivatized amino acids were detected by using an on-linephenylthiohydantoin analyzer (Applied Biosystems, FosterCity, CA; model 120A). Pyroglutamate aminopeptidase(Boehringer Mannheim GmbH) digestion was carried outaccording to described procedures (19) with a molar ratio ofenzyme to ChTX of 1:10 and adding all enzyme at thebeginning of the reaction.

RESULTS AND DISCUSSION

Protein Purification. The biological activity of ChTX hasbeen ascertained by electrical analysis of single high-conductance (=220 pS) Ca2'-activated K+ channels inisolated plasma membrane patches derived from either aorticsmooth muscle or GH3 pituitary cells possessing this activity.Measurements have been accomplished by using patch-clamp techniques with excised patches of membranes ori-ented with an outside-facing-out polarity. Inhibition ofCa2"-activated K+ channels by ChTX in these preparations resultsin an increase in silent periods between bursts of channelactivity, as has been described with reconstituted Ca2"-activated K+ channels from skeletal muscle (4, 6). Theappearance of such characteristic behavior was used tofollow ChTX activity during purification. The purificationprotocol that was developed is based on the procedures ofSmith et al. (6). However, the first chromatofocusing step inSP-Sephadex has been replaced with cationic-exchange chro-matography in a Mono-S column. In addition, both theorganic and the aqueous solvents in the reversed-phasechromatography have been changed. These higher-resolutionchromatographic steps have increased considerably the re-producibility and efficiency of the purification procedure. AtpH 9.0, 84% of the 280-nm absorbance of the filtered extractis not retained by the column and ChTX activity elutedbetween 300 and 340mM NaCl in two consecutive peaks thatwere collected separately (Fig. LA). The active proteinconstitutes 0.61% of the total protein injected onto the

0

1.c

L

L4.)

.0L 0 .

zT-m0U,mE: 5 10 15 5 10 15

EFFLUENT VOLUME (ml)

100

m

50 UUof

0

FIG. 1. Purification of ChTX. (A) Mono-S chromatography of 80mg oflyophilized scorpion venom treated as described. Buffer B was375 mM NaCl in 20 mM sodium borate (pH 9) and the gradient isdepicted as a dashed line. ChTX activity eluted where indicated bythe horizontal bars. (B) Reversed-phase chromatography of analiquot of fraction A of the Mono-S chromatography. Buffer B isisopropanol/acetonitrile (2:1). (Inset) A silver-stained NaDodSO4/polyacrylamide gel of reduced ChTX. Approximately 280 ng of thematerial obtained from the reversed-phase chromatography wastreated as described. Molecular mass standards are expressed inkDa.

column (51 mg). Reversed-phase chromatography of pool A(Fig. LA) resulted in a single major peak, which coelutes withChTX activity (Fig. 1B). In the case of pool B, two peakswere separated (data not shown); one with an identicalelution time, specific activity, and amino acid composition asthe one of Fig. 1B, while the other is an unrelated protein.Together, both active fractions represent 60o of the totalprotein content of pools A and B (Fig. LA). Typically, 100 mgof lyophilized venom yields 230 ug of ChTX.

Characterization of ChTX. NaDodSO4/polyacrylamidegels of purified ChTX with (Fig. 1B, Inset) or without (data

100

HI- 0

ZU0Z U

'a-Z z

I CL

75

50

25

- LOG CChTX (CM)

FIG. 2. Inhibition of Ca2+-activated K+ channels in GH3 cells byChTX. A 5- to 10-MQ microelectrode containing (in mM) 140 KCI,2 MgCl2, 7 EGTA, 6 CaCl2, and 10 Hepes (pH 7.3) was used to excisean outside-out membrane patch from a GH3 cell. The patch wasbathed in a medium consisting of (in mM) 135 NaCl, 5 KCI, 10 CaCl2,2 MgCl2, and 10 Hepes (pH 7.3), and held at a potential of + 20 mV.After recording channel activity, increasing concentrations of puri-fied ChTX were added to the bath and allowed to equilibrate.Inhibitory activity was monitored as the ratio of events per sec withChTX to events per sec in the control. Between additions of ChTX,the chamber was perfused with toxin-free medium, whereuponcontrol channel activity was regained, indicating that ChTX block iscompletely reversible.

Proc. Natl. Acad. Sci. USA 85 (1988)

a

Proc. Natl. Acad. Sci. USA 85 (1988) 3331

<CG U-PHE-TR sWVNERs-aSTR-TH-SER-YS ,L-Y-R-SRVI-CSGL*W-Eu-i FiN T-setFw-L-Ln-cs-mET-Fai-Y-YYSFA-CSTYR-

F (Q.-I i_A- _fw _f- _Am _am _am _~. _f. t--t. M-_ 80_m_ _ _ _ _ _tm _m _

ism JEW _. _10 _m _m _t M_ _ft A_ _~ _N _m Am _00 _m.*VS-i I, VS-3 I

UELC-4 MI -LC-5 1 i- cC-1 I- ELC-3 -

FIG. 3. The complete amino acid sequence of ChTX. <GL, V8, and ELC numbers refer to pyroglutaminase, Staphylococcus aureus V8protease, and endoproteinase Lys-C generated peptides. *, N terminus-blocked peptide that was treated on the sequenator glass fiber filter with25 Al of 11.7 M HC1. All residues identified by amino-terminal sequence determinations are denoted by single-headed arrows pointing to theright. Residues identified at the carboxyl terminus by timed digestion of 5 nmol of ChTX with 500 pmol of carboxypeptidase A (BoehringerMannheim GmbH) (16) are denoted with single-headed arrows pointing to the left. The observed and integer amino acid compositions of ChTX,determined as described in Materials and Methods, are as follows: Asx, 2.89, 3; Glx, 2.77, 3; Ser, 4.79, 5; Gly, 1.17, 1; His, 1.09, 1; Arg, 3.08,3; Thr, 4.05, 4; Tyr, 1.12, 1; Val, 1.98, 2; Met, 1.24, 1; Cys, 5.%, 6; Leu, 1.14, 1; Phe, 0.97, 1; Lys, 3.76, 4; Trp, not determined.

not shown) disulfide reduction by dithiothreitol display asingle band at 4.35 kDa. Because the molecular mass is notaltered by reduction, ChTX appears to be composed of asingle polypeptide chain. Since the lower limit of proteindetection in these polyacrylamide gels is =1 ng, the mostabundant contaminant in this preparation must be <0.25% ofthe total amount of ChTX present.The amino acid composition ofChTX is given in the legend

of Fig. 3. This amino acid composition resembles that ofSmith et al. (6) except in proline content, if it is assumed thatthe alanine residue in their determination is a contaminationand if the scaling is consequently done on the basis of eitherleucine, phenylalanine, or tyrosine. The extinction coeffi-cients of ChTX calculated from the amino acid analysis asdescribed in Materials and Methods are 1.52 (mg/ml)-lcm 1 at 280 nm and 15.05 (mg/ml)-1'cm-' at 215 nm.

Single-channel analyses of the action of pure ChTX revealthat the toxin reversibly blocks the =220-pS Ca2 '-activatedK+ channel ofGH3 cells by binding only at the external faceof the channel, as suggested by Miller et al. (4), withoutaffecting mean channel conductance or significantly modify-ing individual channel open times. According to previouswork (6), ChTX appears to block channel activity by a simplebimolecular inhibition process. Therefore, determination ofthe concentration of ChTX that results in a 50% probabilityof channel block should yield an apparent dissociationconstant (Kd) for ChTX under defined experimental condi-tions. As shown by the data of Fig. 2, pure ChTX blocks theCa2 + -activated K+ channel ofGH3 cells with a Kd of2.1 nM.In addition, purified ChTX (100 nM) has no effect on eitherrapidly inactivating or slowly inactivating Ca2+ currents oron rapidly inactivating K+ currents in GH3 cells, indicatingspecificity of action of the toxin. Titrations of ChTX activityperformed with isolated membrane patches obtained frombovine aortic cells give a nearly identical Kd value.Amino Acid Sequence. The complete amino acid sequence

of ChTX is given in Fig. 3. The 37-residue sequence is welldetermined, with all regions confirmed by overlapping pep-tides. Attempts to sequence ChTX from its amino-terminalend were unsuccessful, indicating a blocked N terminus. Nophenylthiohydantoin-derivatized amino acids were detectedafter three Edman degradation cycles of peptide V8-1,indicating that this peptide was derived from the amino-terminal end of the protein. The peptide was unblocked bytreating the sequenator glass fiber sample filter with HCO for30 min (Fig. 3, legend). The resulting sequence agrees withthe amino acid composition of this peptide except for twomissing amino acids, phenylalanine, and glutamic acid orglutamine, suggesting that the blocked amino-terminal resi-due is pyroglutamine. Equivalent results were obtained withpeptide ELC-4, although in this case the HCO treatment wasextended to 60 min. A sequence of37 residues, accounting forall residues of the amino acid composition of ChTX, exceptone glutamic acid or glutamine, was obtained when theprotein was treated with pyroglutaminase. A peptide wasgenerated by digestion with endoproteinase Lys-C, which

was shown to be devoid of lysine by amino acid analysis.Given the specificity of this enzyme, this peptide must bederived from the carboxyl-terminal end of the protein. Thesequence of this peptide correlates with the last residuesidentified by sequencing peptide V8-2 and pyroglutaminase-treated ChTX. The carboxyl-terminal sequence was con-firmed by digestion of the whole protein with carboxypepti-dase A (Fig. 3). The molecular mass of ChTX, calculatedfrom this sequence, is 4353 Da. Although two potentialasparagine glycosylation sequences (Asn-Xaa-Ser/Thr) existat positions 4 and 22 of ChTX, it is unlikely that this proteinis glycosylated since the yield ofasparagine at these positionsduring sequencing is not less than expected, and there is goodagreement in the molecular mass of ChTX as calculated fromelectrophoretic mobility and amino acid composition.Amino Acid Sequence Similarity. In a search ofthe National

Biomedical Research Foundation protein sequence databank,11 ChTX appears to be unique. ChTX was also com-pared with other known peptide inhibitors ofK+ channels. Ahigh level of similarity was detected between ChTX andnoxiustoxin, a toxin from the venom of the scorpion Centu-roides noxius that blocks the delayed rectifier K+ channel ofthe squid giant axon (20). The best alignment between ChTXand noxiustoxin (Fig. 4, I) is of 8.17 SD from random.Interestingly, ChTX displays little similarity with apamin, aninhibitor of a low conductance Ca2+-activated K+ channel(22), or with dendrotoxin and mast-cell degranulating pep-tide, both blockers of rapidly inactivating K+ channels(A-type current; ref. 23).The ChTX sequence has a remarkably high concentration

of basic residues between positions 21 and 28, making thisregion a potential site of interaction with the Ca2 +-activatedK+ channel. Therefore, a search of the National BiomedicalResearch Foundation data bank was repeated to detectpossible homologies with this region. Thirteen neurotoxinsbelonging to species as distant as scorpions, marine worms,and snakes were listed (Fig. 4, legend). Of these, the bestalignment is between ChTX and neurotoxin P2 of the scor-pion Androctonus mauretanicus var. mauretanicus (Fig. 4,I), whose mechanism of action is unknown (24). This align-ment is 4.1 SD from random, which corresponds to a 1 in32,000 chance that the alignment is nonsignificant. Alignmentof the 21-28 segment of ChTX with the equivalent region ofthe marine worm neurotoxins (25) indicated the cysteineresidues to be in register between amino acids 10 and 45. Thisalignment vs. one of the homologous worm sequences at 2.5SD above random is shown (Fig. 4, II). The probability is<0.62% that this value reflects zero correlation between thecompared sequences. It is noteworthy that the marine wormneurotoxins act by prolonging the repolarization phase of theaction potential in crustacean axon (25), a mechanism con-sistent with K+-channel block. The list of homologoussequences (Fig. 4, legend) also includes 10 well-characterized

IProtein Identification Resource (1986) Protein Sequence Database(Natl. Biomed. Res. Found., Washington, DC), Release 8.0.

Biochemistry: Gimenez-Gallego et al.

3332 Biochemistry: Gimenez-Gallego et al.

ChTX

1 (39)

2 (35)

IIChTX

3 (55)

IIIChTX

4 (73)

5 (71)

6 (71)

7 (73)

8 )73)

EI

T-I-I N -

*V S C-T T E C i-S-V C Q-R [ H-N-T S-R-G- -K-C-H-N K F-C R C-Y s

-V K C-T S- K Q C S-K-P C K-E L Y-G-S S ilGilK-C-N-N G I-C K C-Y N-Nj... . .I . .. .

-C: T -D-P-YT E C A-T-C CU' '-- SU''-L.*'- *'-- - - - -G-G R-G--K-C V-G-P-Q CN-R-I

E-F-T-N-V-S C T-T-S- -K-E C W-S-V C Q-R-L-H-N-T-SR C N-NR RC - - - - -R VY-S

P-A-C-E-N-N C R--Q-Y-D-D C -I-K C Q-G-K-W-A-G-K [-G-i-C A-A-H- C A-V-Q-T-I-S C N--

E-F N ]S

-P-S-S -A1VT8 UL

-D-V [ S-Q-I81-fD-I T S-K-D81-D-A - S-Q-T

-D-V-K-S-E-I

T-T-S-K-E-

P-P-G-E-N-L

A-D-G- -N-V

P-N-G- -R-V

P-D-G-Q-D-I

P-A-G-Q-D-I

W-S- -V-

Y-R-K-N-i

y-T-K-TiW

Y-T-K-T-i

Y-T-K-T-i

Y-T-E-T-W

Q-R-L-H-N -S-R-G-K

D-A-F-C- -S S-R-G-K

D-N-F-C- -A S-R-G-K

D-G-F-C- -S S-R-G-K

D-G-F-C- -S S-R-G-K

D-A-W-C- I-S-R-G-K

C-N-N-K-K

V-V-E-L-G

R-V-D-L-G

R-V-D-L-G

R-I-D-L-G

R-V-D-L-G

-R

A-A-T

A-A-T

A-A-T

A-A-T

A-A-T

y[~56

P Sl4t

P-T-47

P-T-48

P-K-48

P-I-

FIG. 4. The following scorpion (I), marine worm (II), and snaketoxins (III) are compared with ChTX: 1, noxiustoxin, Centruroidesnoxius; 2, neurotoxin P2, Androctonus mauretanicus var. maureta-nicus; 3, neurotoxin B-II, Cerebratulus lacteus; 4, a-bungarotoxin,Bungarus multicinctus; 5, long neurotoxin 2, Naja melanoleuca; 6,long neurotoxin 1, Naja naja; 7 and 8, long neurotoxin 2, and 1,Ophiophagus hannah. Numbers in parentheses indicate the numberof residues of each protein. When the comparison does not includethe full length of the protein, the residue number at the amino andcarboxyl terminus of the considered fragment appears above thesequence. Similarities were evaluated with the ALIGN program (21),using a 112-point-accepted mutation (PAM) matrix (I), 192-PAMmatrix (II), and 160-PAM matrix (III). The BIAS and GAP param-eters were iteratively optimized. Toxins included in II and III andneurotoxin P2 of Androctonus mauretanicus (I) were detected in asearch of the ChTX21-28 region in the National Biomedical ResearchFoundation sequence data bank, by using the IFIND program(Intelligenetics, Mountain View, CA). Only residues identical withChTX are framed. Additional identified toxins were as follows:neurotoxin B-IV, Cerebratulus lacteus; long neurotoxin 1, Najamelanoleuca; long neurotoxin 1, Laticauda semifasciata; long neu-rotoxin 1, Naja naja var. oxiana; long neurotoxin 1, Notechisscutatus; long neurotoxin 1, Acantophis antarticus. The single-letteramino acid code is used.

snake neurotoxins (26), which are all members of the longneurotoxin family. The alignments between these neuro-toxins and ChTX that score >3 SD from random are shown(Fig. 4, III). The highest score (3.4 SD from random)corresponds to the alignment between ChTX and a-bungarotoxin. The probabilities that the obtained scoresreflect zero correlation between ChTX and these snakeneurotoxins (Fig. 4, III) are between 1.3 x 10-3 and 4 x10-4. Interestingly, the region of greatest similarity betweenChTX and a-bungarotoxin (residues 23-27 of ChTX) corre-sponds to a highly conserved region of the long neurotoxinsthat is critical for the interaction at their receptor (26).The similarity of ChTX with toxins of species as phyloge-

netically distant as those listed in Fig. 4 is somewhat sur-prising. Perhaps all these toxins are versions of normalproteins with the nontoxic physiological function of modu-lating ion-channel activities. It may be that all of thesechannel modulating proteins evolved from a common ances-tral gene encoding a singular ion-channel modulator whosemode of action became diverse as ion channels evolved,acquiring different functional properties.

Three-Dimensional Structure of ChTX. Upon examinationof Fig. 4, a similarity between the hydropathic patterns ofChTX and the long neurotoxins becomes evident. This can beevaluated by the method of Sweet and Eisenberg (27), withthe solvent exposure index ("1-f," normalized to a mean of

FIG. 5. Modeled three-dimensional structure of ChTX. Thea-carbon backbone of ChTX (green) is shown superimposed on thatofa-bungarotoxin (red). The overlap between the structures is shownin yellow, where red and green mix. The conserved disulfide bridgesappear in white, while those exclusive to either a-bungarotoxin orChTX appear in orange and blue, respectively. Both structures areassigned the equivalent a-bungarotoxin residue numbers. The ami-no-terminal residues ofa-bungarotoxin and ChTX are labeled N1 andN10, whereas the corresponding carboxyl termini are labeled C74and C50, respectively. Positively charged residues (i.e., lysine,histidine, and arginine; K, H, and R, respectively) are in blue.a-Bungarotoxin coordinates were obtained from the BrookhavenProtein Data Bank (7, 8) and were displayed and manipulated withthe graphics program FRODO (30, 31) operating on an Evans &Sutherland PS330 system. Amino acid substitutions, deletions, andadditions were made according to Fig. 4, maintaining as closely aspossible the similarity between the structures.

zero and a SD from the mean of 1.0; ref. 28).** Thecorrelation coefficient (rH) between ChTX and a-bungaro-toxin is 0.71. The probability that this value reflects zerocorrelation is 9.9 x 10-8. This similarity in the solventexposure indicates a high similarity between the tertiarystructure of ChTX and the corresponding region of a-bungarotoxin. The three-dimensional structural similarity isalso supported by analysis of the minimum mutation distancebetween these two sequences (29), which is 1.03 mutationsper amino acid, -6 SD from random.Given this striking similarity between ChTX and region

10-48 of a-bungarotoxin, a model was constructed for thethree-dimensional structure of ChTX based on the solvedcrystal structure of a-bungarotoxin (in the following discus-sion of the model, both structures are assigned the equivalenta-bungarotoxin residue number according to Fig. 4; there-fore, the amino- and carboxyl-terminal residues of ChTXbecome Glu-10 and Ser-50, respectively). Fig. 5 illustrates ana-carbon plot of the ChTX model superimposed on thestructure of a-bungarotoxin. A large degree of overlap(shown in yellow, where red and green mix) is evidentbetween these two structures. The region that shares identitywith ChTX includes the whole central loop and some residuesof the first and third loops of a-bungarotoxin (26). Energyminimization (32) studies indicate that the proposed model isa stable molecular structure (R. Blevins, personal commu-nication). Only one disulfide bridge (residue 16 to 44; Fig. 5,white notation) can be conserved between ChTX and a-bungarotoxin. A second disulfide bridge (residue 23 to 48;

**By using the normalized exposure indexes, the correlation coef-ficient (rH) between the two homologous domains of rhodanase,cytochrome c2 of Rhodospirillum rubrum and cytochrome c551 ofPseudomonas aeruginosa, and cytochrome c2 of Rhodospirillumrubrum and cytochrome c of tuna are 0.43, 0.44, and 0.67,respectively.

Proc. Nad. Acad Sci. USA 85 (1988)

Proc. Natl. Acad. Sci. USA 85 (1988) 3333

Fig. 5, blue notation) was readily formed in the ChTX modelby a small reorientation of the C terminus. The remainingbridge (residue 17 to 27; Fig. 5, blue notation) did require areconstruction vs. the solved a-bungarotoxin structure tomake the sulfhydrils point toward each other. This results ina more conventional disulfide bond that relieves the strongdistortion of the structure by disulfide bond 29-33 in thisregion of a-bungarotoxin. The most striking differencesbetween both structures are the large C-terminal deletion of24 residues and a further N-terminal deletion of 9 residues.These deletions have the effect ofuncovering the central loopstructure of ChTX.

Circular dichroism spectra of the long neurotoxins revealthat these proteins contain a high proportion of (3-structure(26). In the solved a-bungarotoxin structure, there are threestrands of the polypeptide chain that adopt a triple-strandedantiparallel /3-pleated sheet conformation. Two of thesestrands are in the second loop, the one entirely conserved inChTX. Furthermore, the disappearance in ChTX of thedisulfide bridge equivalent to the one between residues 29and 33 of a-bungarotoxin should enlarge this P-region, as isthe case with the short neurotoxins. Hence, the relativeamount of B-structure should increase substantially in ChTX.The circular dichroism spectrum of ChTX agrees with thisprediction. Its maximum (195 nm), its minimum (217 nm), andthe ratio [601/[60215 (==2) are quite close to the theoreticalvalues of poly-L-lysine in /3-conformation (33). The percent-age of /3-structure content calculated from this spectrum (34)is -70%, a value that is considerably higher than the onecalculated for the long neurotoxins (26).

In the proposed model of the tertiary structure of ChTX(Fig. 5), there are two clusters of positively charged residues.The first is at the bottom of the central loop of the molecule.This includes three basic residues, Arg-30A, Arg-36, andLys-38, and possibly a charged His-32, all of whose sidechains are located on the same side of the molecule. Thesecond is found at the top of the molecule and includesLys-20, Lys-42, Lys-43, and Arg-45. There is, however, apossibility that Lys-42 forms a salt bridge with Glu-21. Thishypothesis is consistent with the fact that in noxiustoxin bothresidues are simultaneously replaced by two nonchargedamino acids. Interestingly, all residues of both clusters,except Lys-20, are on the same side of the molecule, givingChTX a distinct charge polarity on one face vs. the other. Thepositively charged face might be involved in the interactionby which, as it has been proposed, ChTX blocks Ca2"-activated K+ channels (4, 35). Obviously, the proposedmodel is a preliminary approach to the three-dimensionalstructure of ChTX. However, the findings presented in thisstudy form a basis for developing the structure-activityrelationship of the ChTX molecule and indicate that ChTXwill be a useful probe for elucidating the function of Ca2+ -activated K+ channels.

We thank William Randall for his assistance in obtaining circulardichroism spectra of ChTX, John Jacobs and Chris Dunwiddie foruseful discussions, and Kenneth A. Thomas for discussions andcritical reading of the manuscript.

1. Peterson, 0. H. & Maruyama, Y. (1984) Nature (London) 307,693-6%.

2. Abia, A., Lobaton, C. D., Moreno, A. & Garcia-Sancho, J.(1986) Biochim. Biophys. Acta 856, 403-407.

3. Castle, N. A. & Strong, P. N. (1986) FEBS Lett. 209, 117-121.4. Miller, C., Moczydlowski, E., Latorre, R. & Philips, M. (1985)

Nature (London) 313, 316-318.5. Hermann, A. & Erxleben, C. (1987) J. Gen. Physiol. 90, 27-47.6. Smith, C., Philips, M. & Miller, C. (1986) J. Biol. Chem. 261,

14607-14613.7. Agard, D. A. & Stroud, R. M. (1982) Acta Crystallogr. Sect. A

38, 186-194.8. Bernstein, F. C., Koetzle, I. F., Williams, G. J. B., Meyer,

E. F., Jr., Brice, M. C., Rodgers, J. R., Kennard, O., Shima-nouchi, T. & Tasumi, M. (1977) J. Mol. Biol. 112, 535-542.

9. Garcia, M. L., Gimenez-Gallego, G., Navia, M., Katz, G.,Reuben, J. P. & Kaczorowski, G. J. (1988) Biophys. J. 53, 151a(abstr.)

10. Ross, R. (1971) J. Cell Biol. 50, 172-186.11. King, V. F., Garcia, M. L., Himmel, D., Reuben, J. P., Lam,

Y. K., Pan, J. X., Han, G. Q. & Kaczorowski, G. J. (1988) J.Biol. Chem., 263, 2238-2244.

12. Hamil, 0. P., Marty, A., Neher, E., Sakmann, B. & Sigworth,F. J. (1981) Pflugers Arch. 391, 85-100.

13. Fling, S. P. & Gregerson, D. S. (1986) Anal. Biochem. 155,83-88.

14. Morrisey, J. H. (1981) Anal. Biochem. 117, 307-310.15. Wray, W., Boulikas, T., Wray, V. P. & Hancock, R. (1981)

Anal. Biochem. 118, 197-203.16. Gimenez-Gallego, G. & Thomas, K. A. (1987) J. Chromatogr.

409, 299-304.17. Gimenez-Gallego, G., Rodkey, J., Bennett, C., Rios Candelore,

M., DiSalvo, J. & Thomas, K. (1985) Science 230, 1385-1388.18. Gimenez-Gallego, G., Conn, G., Hatcher, V. B. & Thomas,

K. A. (1986) Biochem. Biophys. Res. Commun. 138, 611-617.19. Podell, N. N. & Abraham, G. N. (1978) Biochem. Biophys. Res.

Commun. 81, 176-185.20. Carbone, E., Prestipino, G., Spadavecchia, L., Franciolini, F. &

Possani, L. D. (1987) Pflugers Arch. 408, 423-431.21. Dayhoff, M. O., ed. (1978) Atlas of Protein Sequence and

Structure (Natl. Biomed. Res. Found., Washington, DC), Vol.5,Suppl. 3, pp. 345-358.

22. Romey, G., Hughes, M., Schmid-Antomarchi, H. & Lazdunski,M. (1984) J. Physiol. (Paris) 79, 259-264.

23. Bidard, J. N., Mourre, C. & Lazdunski, M. (1987) Biochem.Biophys. Res. Commun. 143, 383-389.

24. Rochat, H., Bernard, P. & Couraud, F. (1979) Adv. Cytophar-macol. 3, 325-334.

25. Blumenthal, K. M., Kein, P. S., Heinrikson, R. L. & Kem,W. R. (1981) J Biol. Chem. 256, 9063-9067.

26. Dufton, M. J. & Hider, R. C. (1982) CRC Crit. Rev. Biochem.14, 113-172.

27. Sweet, R. M. & Eisenberg, D. (1983) J. Mol. Biol. 171, 479-488.28. Rose, G. D., Geselowitz, A. R., Lesser, G. J., Lee, R. H. &

Zehfus, M. H. (1985) Science 229, 834-838.29. Dickerson, R. E. (1971) J. Mol. Biol. 57, 1-15.30. Jones, T. A. (1982) in Computational Crystallography, ed.

Sayer, D. (Clarendon, Oxford), pp. 303-318.31. Bush, B. L. (1984) Comput. Chem. 8, 1-11.32. Singh, V. C., Weiner, P. K., Caldwell, J. W. & Kollman, P. A.

(1986) AMBER (Department of Pharmaceutical Chemistry, Uni-versity of California, San Francisco), Version 3.0.

33. Greenfield, N. & Fasman, G. D. (1969) Biochemistry 8,4108-4116.

34. Yang, J. T., Wu, C. S. C. & Martinez, H. M. (1986) MethodsEnzymol. 130, 208-269.

35. MacKinnon, R. & Miller, C. (1987) Biophys. J. 51, 53a (abstr.).

Biochemistry: Gimenez-Gallego et al.