Structure-Activity Relationships of the Kvβ1 Inactivation Domain and Its Putative Receptor Probed Using Peptide Analogs of Voltage-gated Potassium Channel α- and β-Subunits*

Certain β-subunits exert profound effects on the kinetics of voltage-gated (Kv) potassium channel inactivation through an interaction between the amino-terminal “inactivation domain” of the β-subunit and a “receptor” located at or near the cytoplasmic mouth of the channel pore. Here we used a bacterial random peptide library to examine the structural requirements for this interaction. To identify peptides that bind Kv1.1 we screened the library against a synthetic peptide corresponding to the predicted S4-S5 cytoplasmic loop of the Kv1.1 α-subunit (residues 313–328). Among the highest affinity interactors were peptides with significant homology to the amino terminus of Kvβ1. We performed a second screen using a peptide from the amino terminus of Kvβ1 (residues 2–31) as “bait” and identified peptide sequences with significant homology to the S4-S5 loop of Kv1.1. A series of synthetic peptides containing mutations of the wild-type Kvβ1 and Kv1.1 sequences were examined for their ability to inhibit Kvβ1/Kv1.1 binding. Amino acids Arg20 and Leu21 in Kvβ1 and residues Arg324 and Leu328 in Kv1.1 were found to be important for the interaction. Taken together, these data provide support for the contention that the S4-S5 loop of the Kv1.1 α subunit is the likely acceptor for the Kvβ1 inactivation domain and provide information about residues that may underlie the protein-protein interactions responsible for β-subunit mediated Kv channel inactivation.

Analysis of the sequences of cloned ␤-subunits taken together with mutagenesis studies indicates that the aminoterminal 30 amino acids of K v ␤1 contains a "ball" domain that is necessary and sufficient to rapidly inactivate K v 1.1 (12). This NH 2 -terminal inactivation domain of K v ␤1 shares primary amino acid sequence homology with the inactivation ball of the A-type K v channels K v 1.4 and the Drosophila Shaker channel. Point mutations in the S4-S5 cytoplasmic loop of the K v 1.4 ␣-subunit, which lies near the inner mouth of the channel pore, indicate that this may be the acceptor site for the NH 2 -terminal inactivation ball (18) and suggest that analysis of the interaction of NH 2 -terminal ball and S4-S5 "receptor" domains may provide clues to the structural requirements for N-type channel inactivation.
Recently, a method to study protein-protein interactions has been developed in which constrained random peptides are displayed on the surface of bacteria as functional fusions to the protein flagellin (19). This enables large libraries of random and diverse polypeptides to be screened and specific peptides selected and characterized following their binding to an immobilized target protein. Due to the immense diversity of these libraries and the relative ease at which sequences can be identified, it is possible to rapidly obtain information about the structural requirements of high-affinity protein-protein interactions (20 -25). Random peptide libraries displayed on the surface of bacteria and phage have been employed successfully to identify protein phosphatase-1-binding motifs (26) and potent rhodopsin-binding peptide sequences related to a COOHterminal G ␣t peptide (27), as well as for selecting peptide ligands for the erythropoiten receptor (28).
In the present study we used this bacterial peptide display library screening strategy to examine the interaction of K v channel ␣and ␤-subunit domains. In addition, we utilized an in vitro protein interaction assay to examine analogs of these ␣and ␤-subunits and identified key amino acid residues that are important for the interaction. We provide support for the contention that the S4-S5 loop of the K v 1.1 ␣-subunit is the likely acceptor for the K v ␤1 inactivation domain and provide information about residues that may underlie the protein-protein interactions responsible for ␤-subunit-mediated K v channel inactivation.

MATERIALS AND METHODS
Proteins and Peptides-A synthetic peptide (QILGQTLKASMR-ELGL) corresponding to amino acids (313-328) of the human K v 1.1 potassium channel ␣-subunit polypeptide was synthesized in both biotinylated and nonbiotinylated form (Genosys Biotechnologies, Woodlands, TX). A full-length K v ␤1 and a truncated version corresponding to * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
amino acids  were expressed in Escherichia coli as GST fusions (Amersham Pharmacia Biotech). The GST-K v ␤1 fusions were purified by affinity chromatography on a glutathione-Sepharose 4B column (Amersham Pharmacia Biotech) and cleaved from the GST by the addition of thrombin. All mutant peptides were synthesized as constrained 12-mers by Genosys Biotechnologies (Woodlands, TX).
Growth of the E. coli Peptide Library-The FliTrx random peptide bacterial library was obtained from Invitrogen (San Diego, CA); growth of the bacterial cultures and general panning methods are based on the system described previously (19).
Immobilization of K v ␤1 and K v 1.1 on Culture Plates-100 g of K v ␤1 or K v 1.1 was immobilized on 60-mm plastic Petri dishes. The plates were washed with 10 ml of sterile water and nonspecific sites blocked with 10 ml of blocking solution (1% non-fat dry milk, 150 mM NaCl, 1% methylmannoside, and 100 g/ml ampicillin in IMC medium) at room temperature for 1 h.
Selection for K v 1.1/K v ␤1 Binding-All incubations were performed at 25°C, and all other manipulations were at room temperature. An aliquot of the FliTrx library was grown to saturation for 15 h. 10 10 cells were added to IMC medium (Invitrogen) containing 100 g/ml ampicillin and 100 g/ml tryptophan to induce expression of the FliTrx plasmid and incubated for an additional 6 h. After library induction the culture was adjusted to a final concentration of 1% non-fat dry milk, 200 mM NaCl, and 1% methylmannoside. Subsequently, 10 ml of the induced cells were added to each culture plate for 1 h. Following binding each plate was washed three times with IMC/amp 100 medium containing 1% methylmannoside. The bound cells were removed by vortexing in a residual amount of buffer and incubated in 10 ml of IMC at 25°C for 14 -16 h. Up to five rounds of panning were performed, after which the cultures were plated out and individual colonies selected.
Plasmid Isolation and Sequence Determination-Plasmid minipreps were performed on the Qiagen BioRobot 9600 (Qiagen Inc., Santa Clarita, CA) and the resulting plasmid DNA sequenced on an Applied Biosystems 373A automated DNA sequencer (Perkin-Elmer/ Applied Biosystems, Foster City, CA) using the FliTrx forward and reverse sequencing primers (Invitrogen). The resulting dodecamer DNA sequences were translated using DNAstar (Madison, WI) and Clustal analysis used to align the sequences and generate a consensus sequence. The resulting consensus sequences generated from days 2, 3, and 5 were similarly aligned to the wild-type sequences. Gaps in the consensus sequences signify that no consensus residue was identified at this position.
Plate Assay-The full-length K v ␤1 protein was immobilized onto Microtiter plates (Corning Costar, Cambridge, MA,) for 1 h at 37°C at a final concentration of 10 ng/well. The wells were then washed twice with Hepes-buffered saline (10 mM Hepes, pH 7.4, 0.15 M NaCl, 3 mM EDTA, and 0.005% v/v Surfactant P20) followed by blocking with 1% bovine serum albumin in Hepes-buffered saline. For binding, 100 ng of a biotinylated K v 1.1 (S4-S5 loop) were added to each well and incubated for 1 h at 37°C, followed by detection with streptavidin, biotin-alkaline  phosphatase, and developed with p-nitrophenyl phosphate. Peptides were tested in the assay system for inhibition of binding. IC 50 values were calculated using the computer program Microsoft Excel 6.0 and p values calculated using the InStat 1.14 statistical program.

RESULTS
The panning procedure was applied to identify peptides that bind the K v 1.1 channel S4-S5 loop and K v ␤1 NH 2 terminus, using immobilized K v 1.1 (313-328) and K v ␤1 (2-31) as "baits." Following five successive rounds of panning the fusion peptides expressed by 50 randomly selected clones from each screen were characterized by DNA sequencing. Consensus sequences were identified by multiple alignment analysis, and in the search for K v 1.1 interactors, 92% of the clones analyzed overlapped in a core domain that can be readily found in the amino terminus of the human K v ␤1 subunit (Leu 12 -Lys 13 -Ser 14 -Arg 15 -Asn 16 -Gly 17 -Glu 18 -Asp 19 -Arg 20 -Leu 21 -Leu 22 -Ser 23 ) (Table I). Conversely, a consensus motif identical to the S4-S5 loop of K v 1.1 (Gln 317 -Thr 318 -Leu 319 -Lys 320 -Ala 321 -Ser 322 -Met 323 -Arg 324 -Glu 325 ) was observed in peptides that bind K v ␤1 (Table II). These results support the hypothesis existing in the literature that implicates the S4-S5 loop of K v 1.1 as the receptor for the inactivation gate and K v ␤1 as a regulator of channel gating properties (12) and suggests that the interacting domains are smaller than previously thought. Individual amino acid residues within the consensus regions are conserved to varying degrees, possibly reflecting their differing contributions toward binding. In the K v ␤1 analogs the most highly conserved amino acids were Gly 17 (65%), Arg 20 (55%), and Leu 21 (77%), whereas in the Kv1.1 analogs Arg 324 was seen in 70% of the clones and Ala 321 was represented in 46%.
Panning enriches for high affinity binders and peptides from earlier days may bind with a lower affinity but contain amino acids critical for the interaction. To identify these key residues, plasmids isolated from days 1, 2, and 3 of panning were sequenced. For the K v ␤1 analogs day 1 produced no consensus sequence and contained 50% stop codons (18/36). Subsequently, days 2 (33% stop codons, 12/36) and 3 (17% stop codons, 6/36) showed 41 and 66% sequence conservation relative to the day 5 consensus sequence, respectively, and yielded the following consensus sequence, Arg 15 -X-Gly 17 -Glu 18 -Asp 19 -Arg 20 -Leu 21 (Fig. 1). The decrease in the number of stop codons is indicative of the enrichment of specific binding peptides within the library. Similarly, the K v 1.1 analogs on day 1 showed no homology to the wild-type sequence with ϳ50% stop codons observed. The results of days 2 (22% stop codons, 8/36) and 3 (17% stop codons, 6/36) show a consensus of Thr 318 -Leu 319 -X-X-X-Met 323 -Arg 324 -Glu 325 -Leu 326 -X-Leu 328 , which has 62% sequence homology to day 5 (Fig. 2).
To investigate the influence of particular amino acids on the interaction of K v ␤1 and K v 1.1, a series of constrained mutant peptides were synthesized and tested for their ability to disrupt binding in a K v ␤1/K v 1.1 interaction assay (Table III). The wildtype sequences were analyzed and the IC 50 for K v ␤1 wild-type peptide was calculated to be 9 Ϯ 2 nM, the IC 50 for K v 1.1 wild-type peptide was 15 Ϯ 2 nM. The single mutants K v ␤1 R20A and L21A showed greatly reduced affinity (200-and 30-fold, respectively) when compared with the wild-type control sequence. Interestingly, these amino acids were conserved following just 2 days of panning, substantiating our belief that key residues would be conserved at these earlier days. K v ␤1 residues Arg 15 , Asn 16 , and Ser 23 were also observed at day 3, although their mutation had only a small 2-3-fold effect on binding. Among the mutants tested for K v ␤1 was the triple mutant K v ␤1 G17S,E18G,D19N, since it had been reported in the literature to enhance the K v ␤1 mediated K v 1.1 inactivation. Suprisingly, in this study the triple mutant had no inhibitory effect (Fig. 3). Of the single mutants tested for K v 1.1, R324A completely abolished peptide inhibition of K v ␤1/K v 1.1 binding, S322A and L328A demonstrated marked decreases in affinity (60-to 100-fold, respectively) (Fig. 4). T318A (100 Ϯ 2 nM) and E325A (100 Ϯ 3 nM) showed smaller 6-fold decreases in binding affinity. Other residues investigated showed no effect.

DISCUSSION
Voltage-activated potassium currents are vital for neuronal function and are carried by membrane channels composed of 4␣-and 4␤-subunits arranged as a hetero-octamer (31). Some ␤-subunits, such as K v ␤1, can convert noninactivating "delayed rectifying" currents of certain ␣-subunits (e.g. K v 1.1) into rapidly inactivating currents (12). In this study we used a random peptide library to investigate the interacting domains of K v 1.1 and K v ␤1. The N terminus of K v ␤1 (amino acids 2-31), when used as bait, bound peptides homologous to the S4-S5 loop of K v 1.1, whereas sequences related to K v ␤1 NH 2 terminus were observed when the S4-S5 loop of K v 1.1 was used as bait. This correlates well with studies that have implicated these regions as the sites of interaction between the K ϩ channel and NH 2terminal inactivation domains (12,18). Furthermore, we were able to restrict the interacting regions to small motifs, amino acids 12-23 in K v ␤1 and residues 317-325 of K v 1.1, which form core domains important for the interaction.
In the motif identified in K v ␤1, Arg 20 and Leu 21 were represented in 55 and 77% of the sequences analyzed at day 5. 69% of the clones had a basic residue (KR) at position 20, suggesting that a positive charge at this position may be important for structure/function of the K v ␤1. In contrast, a strong preference for a hydrophobic residue (AILFWVPM) was seen at position 21 (80% of the clones had a hydrophobic residue). In a protein interaction assay the K v ␤1 mutations R20A and L21A had the largest effect on binding, causing a 200-and 30-fold decrease in affinity, respectively. The other residues had little or no effect, and this is perhaps indicative of their having a more structural role in presenting the key residues in a conformation to maximize the interaction rather than participating directly in K v 1.1 binding. A sequence comparison of K v ␤1, K v ␤1.2, and K v ␤1.3 indicates that the NH 2 termini of these proteins are quite divergent (16,17). Interestingly, the key residues identified by our study are not found in K v ␤1.2, K v ␤1.3, K v ␤2, or K v ␤3, which could explain the lower effectiveness of the K v ␤1.2 in inactivating K v channels (12,30). Rettig et al. (12) have shown that the triple mutation G17S,E18G,D19N increased the inactivating ability of K v ␤1. In our study, this triple mutant was unable to inhibit the K v 1.1-K v ␤1 interaction even though this triplet was seen at early days of panning, suggesting that these amino acids play a role in the K v ␤1/K v 1.1 binding. One might speculate that this triplet is important for the structural integrity of the K v ␤1 interacting domain, and the conformational changes due to the mutation have less of an impact on the FIG. 1. K v ␤1 consensus sequence emergence during the panning process. Consensus sequences were obtained by alignment of binding peptide sequences following 2, 3, or 5 rounds of panning. No amino acid (or X) indicates no consensus residue was identified.
FIG. 2. K v 1.1 consensus sequence emergence. Sequences obtained from 2, 3, and 5 rounds of panning against immobilized K v ␤1 were aligned to the wild-type sequence (amino acids 313-328) of K v 1.1.
full-length sequence than the small constrained peptides used in this study. The results described here imply that the reported effects of this triple mutation are unlikely to be exerted purely at the level of the protein-protein interaction (12); the increase in net positive charge in the mutant ␤-subunit may also change the response of this subunit to changes in membrane voltage.
Our study also demonstrates that a minimal sequence motif of TLXXXMREL found in the K v 1.1 S4-S5 loop is important for the interaction of K v 1.1 and K v ␤1. Studies of Shaker K ϩ channel inactivation indicated that mutation of E395Q leads to a large decrease in the fast component of inactivation (18). The comparable mutation in our study gave a 6-fold increase in IC 50 , suggesting that its effect on inactivation may be due to a decrease in affinity of the inactivating ball peptide for the S4-S5 loop. Other substitutions, which decreased Shaker channel inactivation (T388S/A, S392C/A), also reduced the affinity of our constrained mutant peptides; S322A resulted in a 133fold increase in IC 50 , whereas T318A caused a more modest 6-fold increase.
In agreement with the mutation data in Shaker, our data also suggests that Leu 319 and Gly 327 play an insignificant role in the interaction. However, one anomaly is the large effect of the R324A mutation in our hands. This residue was conserved even at day 2 and was found in 70% of day 5 clones. Surprisingly, mutation of this residue to glutamine in Shaker was without effect. This may indicate the increased sensitivity of our assay to changes in conformation of a small peptide.
This study demonstrates that random peptide libraries are useful tools for the study of protein-protein interactions. Our results support and extend the literature indicating that the ␤1 NH 2 terminus is likely to mediate its action by binding the S4-S5 loop of K v 1.1. Furthermore, we have identified key residues which are likely to participate in the protein-protein interactions required for ␤-subunit-mediated K v channel inactivation.