Structural Conservation in Prokaryotic and Eukaryotic Potassium Channels

See allHide authors and affiliations

Science  03 Apr 1998:
Vol. 280, Issue 5360, pp. 106-109
DOI: 10.1126/science.280.5360.106


Toxins from scorpion venom interact with potassium channels. Resin-attached, mutant K+ channels from Streptomyces lividans were used to screen venom from Leiurus quinquestriatus hebraeus, and the toxins that interacted with the channel were rapidly identified by mass spectrometry. One of the toxins, agitoxin2, was further studied by mutagenesis and radioligand binding. The results show that a prokaryotic K+ channel has the same pore structure as eukaryotic K+ channels. This structural conservation, through application of techniques presented here, offers a new approach for K+ channel pharmacology.

Scorpion toxins inhibit ion conduction through potassium channels by occluding the pore at the extracellular opening. A single toxin protein binds very specifically to a single K+ channel to cause inhibition. The toxins are 35 to 40 amino acids in length and have a characteristic fold that is held rigidly by three disulfide bridges (1). They are active site inhibitors, because when they bind to the channel they interact energetically with K+ ions in the pore (2-4). The interaction between these inhibitors and the pore of K+ channels has been exploited to gain insights into the structure and function of K+ channels.

Studies employing site-directed mutagenesis of the Shaker K+ channel have mapped the scorpion toxin binding site to regions corresponding to the extracellular entryway of the K+ channel from Streptomyces lividans (the KcsA channel) (4-9). Although the amino acids of the K+ channel selectivity filter are highly conserved, the residues lining the entryway are quite variable. As if to mirror the amino acid variation at the binding site, the toxins are also highly variable in their amino acid composition. A given scorpion venom is a veritable library of toxins, apparently ensuring that a scorpion will inhibit a large fraction of K+ channel types in its victim. Studies on the specificity of toxin-channel interactions have led to the following understanding: The extracellular entryway to the K+ channel is relatively conserved in its three-dimensional structure but the precise amino acid composition is not conserved. The scorpion toxins have a shape, dictated by their conserved fold, that enables them to fit snugly into the entryway, but the affinity of a given toxin-channel pair depends on the residue match (or mismatch) on both interaction surfaces.

We have studied the interaction between the KcsA K+channel (5) and the scorpion toxin agitoxin2 (10). By producing, through mutagenesis, a competent toxin binding site, we show that the KcsA K+ channel pore structure and extracellular entryway are very similar to that of eukaryotic voltage-gated K+ channels such as theShaker K+ channel from Drosophila and the vertebrate voltage-gated K+ channels. By combining our extensive functional data on the toxin-channel interaction with the structures of both proteins we propose a highly-restrained model of the complex structure.

Guided by knowledge of the toxin receptor on theShaker K+ channel, we introduced three point mutations into the KcsA K+ channel that should render it sensitive to scorpion toxins (Fig. 1). Amino acids 61 and 64 were changed to their ShakerK+ channel counterpart and 58 was changed to Ala since a small side chain at this latter position favors binding (4,7). The mutant KcsA K+ channel was expressed inEscherichia coli, extracted from the membrane with the detergent decylmaltoside, and bound to cobalt resin through a carboxyl terminal hexahistidine tag (11). A 1-ml column, prepared with the K+ channel–containing resin, was used to screen the venom of the scorpion Leiurus quinquestriatus hebraeus, the source of numerous well-characterized ion channel toxins. Forty milligrams of venom was added to the column and, after washing, the K+ channel was eluted with an imidazole solution (12). The eluate was analyzed with MALDI-TOF mass spectrometry, focusing on the low mass range appropriate for scorpion toxins (∼4000 daltons). Passage of the venom over the K+channel column resulted in a dramatic enhancement of specific peaks (Fig. 2, A through C). Three of these corresponded in mass to the known K+ channel toxins agitoxin2, charybdotoxin, and Lq2 (Fig. 2, C and D). A fourth peak (Fig. 2C, asterisk) apparently represents a previously unknown toxin. The peak corresponding to chlorotoxin, a chloride channel inhibitor (13), did not bind to the column (Fig. 2, A and C).

Figure 1

Sequence alignment of the KcsA andShaker K+ channel pore regions. The numbering for KcsA is above the sequences. Structural elements are indicated (5). Asterisks, several Shaker K+channel amino acid locations where mutations influence agitoxin2 binding (4, 8, 9); arrows, the three KcsA K+ channel amino acids mutated in this study. The sequences are KcsA, S. lividans, accession number (acc) PIR S60172; and Shaker, Drosophila melanogaster, acc PIR S00479.

Figure 2

Mass spectra of scorpion toxins before and after purification on a K+ channel column. MALDI-TOF mass spectra of venom before purification (A) and after elution from a cobalt column in the absence (B) and presence (C) of attached mutant KcsA K+channel. The accuracy of the mass measurements (±0.3 dalton) permitted identification of most of the major peaks in the mass spectra searched from databases of known toxins of the L. quinquestriatus hebraeus scorpion (D). The KcsA-binding component labeled with an asterisk could not be assigned to a known scorpion toxin. The component labeled X (4193.0 daltons) binds nonspecifically to the column and was not identified. MALDI-MS was performed with the MALDI matrix 4-hydroxy-α-cyano-cinnamic acid (16).

Further quantitative analysis was carried out with agitoxin2. Radiolabeled agitoxin2 was prepared by producing the mutation D20C in the toxin (located far from its channel binding surface) and conjugating it with tritiated N-ethylmaleimide (14). A filter assay showed that labeled agitoxin2 binds to the mutant KcsA K+ channel with an equilibrium dissociation constant (K D) of about 0.6 μM (Fig.3A). In contrast, no binding to the wild-type channel could be detected. The total capacity of resin saturated with mutant channel, based on the specific activity of radiolabeled toxin and the known 1:1 stoichiometry (one toxin per tetrameric channel), is nearly 50 pmol of channel per microliter of resin. This value approximates the expected capacity of the resin and therefore implies that all of the channels in the preparation have a correct conformation.

Figure 3

Binding affinity of wild-type and mutant agitoxin2 to the mutant KcsA K+ channel. (A) Quantity of radiolabeled agitoxin2 bound to 0.3 μl of cobalt resin saturated with the mutant KcsA K+ channel as a function of the radiolabeled agitoxin2 concentration (17). Each point is the mean ±SEM of four measurements, except for the 0.03 μM and 1.5 μM concentrations, which are the mean ± range of mean of two measurements. The curve corresponds to equation: bound agitoxin2 = A × {1 +K D/[agitoxin2]}−1, with equilibrium dissociation constant K D = 0.62 μM and resin capacity A = 16 pmol. (B) Remaining bound fraction of radiolabeled wild-type toxin as a function of the concentration of unlabeled wild-type toxin or mutant toxins K27A or N30A (17). Each point is mean ± SEM of four measurements for wild-type agitoxin2 (squares) or mean ± range of mean of two measurements for K27A (circles) and N30A (triangles) agitoxin2 mutants. The curves correspond to equation: remaining bound fraction = {1 +K Dhot/[agitoxin2hot]} × {1 + (K Dhot/[agitoxin2hot]) × (1 + [agitoxin2cold]/K Dcold)}−1with labeled toxin concentration [agitoxin2hot] = 0.06 μM, wild-type toxin K Dhot = 0.62 μM, and competing toxin dissociation constant K Dcold = 0.62 μM (wild type), 81 μM (K27A), and 27 μM (N30A). (C) CPK model of the interaction surface of agitoxin2 (18). Side chains of functionally important amino acids are in red (4). This figure was prepared with the program GRASP (19).

Amino acids in a well-defined region of agitoxin2 form its functional interaction surface, as determined by the effects of alanine substitution on binding to the Shaker K+ channel [Fig. 3C (4, 8)]. Mutation of Lys27and Asn30 had the largest destabilizing effects. Lys27 is conserved in all members of this toxin family because its side chain apparently plugs the pore of K+channels (3). To confirm that agitoxin2 uses the same amino acids to interact with the mutant KcsA K+ channel, we studied the effects of the K27A and N30A toxin mutations with a competition binding assay (Fig. 3B). These mutations decreased the affinity for the toxin significantly (130-fold and 45-fold, respectively), as anticipated from the Shaker K+channel studies. In contrast, the D20C mutation (predicted to be on the back side of the toxin), even with a bulky N-ethylmaleimide adduct, did not influence affinity (Fig. 3, A and B). These results show that agitoxin2 binds in the same manner to both the mutant KcsA K+ channel and the Shaker K+channel. The affinity for the Shaker K+ channel is considerably higher (K D 1 nM), but we have only mutated three amino acids to mimic the site on theShaker K+ channel (Fig. 1).

These results demonstrate that the overall structure of the agitoxin2 receptor site is very similar on both the KcsA andShaker K+ channels. This conclusion justifies the use of energetic data borrowed from ShakerK+ channel studies to assist in the docking of agitoxin2 onto the KcsA K+ channel structure. Thermodynamic mutant cycle analysis has allowed the identification of numerous energetically coupled residue pairs on the interface [pairs of residues that are related by the fact that mutating one influences the effect (on equilibrium binding) of mutating the other (8)]. The four best defined of these residue pairs are displayed in matched colors on the KcsA K+ channel and agitoxin2 surfaces (Fig.4A). The three off-center residue pairs (blue, green, yellow) have the strongest mutant cycle coupling energies [>3 kT (4, 8)]. The central residue pair (red) is coupled by 1.7 kT and independent information places Lys27 (red residue on agitoxin2, Fig. 3A) over the pore (3, 4). Mere visual inspection suggests a unique orientation for the toxin on the channel (Fig. 4B). If the toxin is placed with its functionally defined interaction surface face-down in the groove formed by the turrets (5), with Lys27at the center, the colors match well in three dimensions. The toxin seems to fit perfectly into the vestibule of a K+ channel. The four-fold symmetry of the K+ channel provides four statistically distinguishable but energetically identical orientations available for a toxin to bind [(Fig. 4A) (15)].

Figure 4

Docking of agitoxin2 onto the KcsA K+ channel. (A) Molecular surface of the pore entryway of the KcsA K+ channel (left) and agitoxin2 (right). The colors indicate locations of interacting residues on the toxin and channel surfaces as determined by thermodynamic mutant cycle analysis of the Shaker K+ channel-agitoxin2 interaction (4, 8). The three pore mutations of the KcsA K+ channel used in this study (Q58A, T61S, and R64D) were introduced into the channel model with the program O (20). Indicated residues on the channel surface correspond to the positions of the Shaker K+ channel equivalent residues (Fig. 1), which couple to the indicated agitoxin2 residues. (B) The pattern of colors in (A) suggests the docking orientation shown by the main chain representation of agitoxin2 placed manually onto the pore entryway. The side chain colors match the colored patches in (A). The position of Gly10 is shown as a green band on agitoxin2. The mutant cycle coupling between residues atShaker 425 (mutant KcsA 58) and residue 10 of agitoxin2 comes about through substitution of a bulky side chain residue at either position (4, 7). Pictures were made with the program GRASP (19).

In summary, through a combination of structural and functional data we present a view of a K+ channel in complex with a neurotoxin from scorpion venom. The KcsA K+ channel is structurally very similar to eukaryotic K+ channels. This structural conservation, through application of techniques developed here, can be exploited to advance our understanding of K+ channel pharmacology.

  • * To whom correspondence should be addressed. E-mail: mackinn{at}


View Abstract

Navigate This Article