NMR structure of the "ball-and-chain" domain of KCNMB2, the beta 2-subunit of large conductance Ca2+- and voltage-activated potassium channels.

The auxiliary beta-subunit KCNMB2 (beta(2)) endows the non-inactivating large conductance Ca(2+)- and voltage-dependent potassium (BK) channel with fast inactivation. This process is mediated by the N terminus of KCNMB2 and closely resembles the "ball-and-chain"-type inactivation observed in voltage-gated potassium channels. Here we investigated the solution structure and function of the KCNMB2 N terminus (amino acids 1-45, BKbeta(2)N) using NMR spectroscopy and patch clamp recordings. BKbeta(2)N completely inactivated BK channels when applied to the cytoplasmic side; its interaction with the BK alpha-subunit is characterized by a particularly slow dissociation rate and an affinity in the upper nanomolar range. The BKbeta(2)N structure comprises two domains connected by a flexible linker: the pore-blocking "ball domain" (formed by residues 1-17) and the "chain domain" (between residues 20-45) linking it to the membrane segment of KCNMB2. The ball domain is made up of a flexible N terminus anchored at a well ordered loop-helix motif. The chain domain consists of a 4-turn helix with an unfolded linker at its C terminus. These structural properties explain the functional characteristics of BKbeta(2)N-mediated inactivation.

Large conductance K ϩ channels (BK 1 or MaxiK channels) are key modulators of excitability in many types of cell (1,2).
They are formed from four identical ␣-subunits encoded by the Slo gene and are activated by membrane depolarization and/or increase in intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i (3)(4)(5)(6)(7)(8)). This dual activation is unique among the large family of K ϩ channels and provides a direct feedback mechanism to regulate Ca 2ϩ influx.
In many tissues, the activation gating of BK channels is modulated by accessory ␤-subunits, a family of membrane proteins (KCNMB) closely associated with the ␣-subunit (7). Four KCNMB proteins have been identified (KCNMB1-4), and they all share a prototypic topology of two transmembrane domains with intracellular N and C termini (9 -13). Functionally, each of these KCNMB proteins distinctly changes the rates of channel activation and deactivation as well as the apparent sensitivity of the channel for Ca 2ϩ (9).
In addition, one of the ␤-subunits, KCNMB2 (␤ 2 ), was found to confer rapid and complete inactivation to the BK channel complex (11,12) in a manner similar to that observed in chromaffin cells of the adrenal gland or in hippocampal CA1 neurons (14,15). Analysis of this KCNMB2-mediated inactivation gating showed that it closely resembled the famous ball-andchain-type inactivation of voltage-gated K ϩ channels (Kv): (i) it is determined by the N terminus of KCNMB2; (ii) it occludes the open channel pore and competes with the pore-blocking agent tetraethylammonium (11,12); (iii) recovery from inactivation is speeded up by an increase of the extracellular K ϩ concentration (11).
Moreover, the N-terminal stretch of the KCNMB2 N terminus (19 amino acids) was shown to be a functional entity, i.e. its fusion to the N terminus of KCNMB1 (␤ 1 ) conferred rapid inactivation to this non-inactivating ␤-subunit, and it occluded BK channels as a synthetic peptide very similar to the "pore plugging" observed for the synthetic inactivation domains (ID) derived from various ␣-subunits and one ␤-subunit of Kv-type K ϩ channels (11).
The three-dimensional structure of Kv-derived IDs was analyzed with NMR spectroscopy in solution and revealed a wide range of structural variability for these proteins. Thus, the ID of Kv3.4 was found to exhibit well defined and compact folding although the backbone lacks secondary structural elements. In contrast, the ID from Shaker B and the inactivating N terminus of Kv␤1.1 (amino acids 1-62) showed no uniquely folded structure but rather behaved like random-coil peptides (16 -19).
To gain structural insight into KCNMB2-mediated inactivation of BK channels, we investigated the solution structure and function of the hydrophilic N terminus of KCNMB2 using NMR spectroscopy and giant patch clamp recording on a synthetic peptide (BK␤ 2 N) corresponding to the entire cytoplasmic N terminus of KCNMB2 (amino acids 1-45).

MATERIALS AND METHODS
Electrophysiology-BK channels were expressed heterologously in Xenopus oocytes as described elsewhere (20). Giant patch recordings were made at room temperature (ϳ23°C) 3-7 days after injection of capped cRNA encoding hBK␣ (GenBank TM accession no. U23767). Pipettes used were made from thick-walled borosilicate glass, had resistances of 0.3-0.6 megaohms (tip diameter of about 20 m), and were filled with (in mM) 5 KOH, 115 NaOH, 10 HEPES, and 0.5 CaCl 2 , pH adjusted to 7.2 with MES. Currents were sampled at 10 kHz and corrected for capacitative transients with an EPC9 amplifier (HEKA electronics, Lamprecht, Germany) with the analog filter set to 3 kHz (Ϫ3 db). The fast application system used is described elsewhere (17) and allowed for a complete solution exchange in less than 2 ms. BK␤ 2 N was dissolved in K int solution and applied via one barrel of the application system. K int was composed as follows (in mM): 119 KOH, 1 KCl, 10 HEPES, 1 EGTA, pH adjusted to 7.2 with MES. The amount of CaCl 2 required to yield a free Ca 2ϩ concentration of 10 M was calculated according to Fabiato (36) and added to the EGTA solution under pH meter control. Thereafter, pH was readjusted to 7.2 with KOH.
Rates of inactivation were determined as described previously (17). Briefly, k off was determined from the time constant of the wash-off ( off ) as k off ϭ 1/ off . k on was then calculated as k on ϭ (1/ on Ϫ k off )/[peptide], with on the time constant for wash-in. Affinity for the peptide-receptor interaction was calculated as k off /k on . All values throughout the paper are given as mean Ϯ S.D. of n experiments.
Peptide Synthesis and Sample Preparation-The BK␤ 2 N protein was synthesized by standard solid-phase synthesis and purified by high pressure liquid chromatography. The mass was confirmed by mass spectrometry.
5.4 mg of BK␤ 2 N were dissolved in 500 l of 90% NMR Spectroscopy-Homonuclear NMR spectra were acquired on a Bruker Avance 600 spectrometer at either 293 or 288 K with a spectral window of 11.5 ppm. Standard pulse sequences were used to record NOESY (21) (mixing times between 100 and 250 ms), CLEAN-TOCSY (22) (isotropic mixing time of 80 ms), and DQF-COSY (23) spectra with 4096 data points in F 2 and 512 increments in F 1 . All two-dimensional 1 H NMR spectra employed the method of time-proportional phase incrementation for quadrature detection in the F 1 dimension (24). Water suppression was achieved either by presaturation or by the WATER-GATE technique (25).
NMR data were processed with the Bruker XWINNMR software using shifted squared sine window functions prior to Fourier transformation. The final matrix size was 4096 ϫ 1024, except for the DQF-COSY spectrum, which was transformed to 16384 ϫ 1024 (corresponding to a digital resolution of 0.42 Hz/point in the F 2 dimension) to extract 3 J HNH␣ coupling constants through a fit of the COSY crosspeaks to two antiphase Lorentzian lines. The programs AURELIA (26) and XEASY (27) were used for analysis of two-dimensional spectra.
Structure Determination-NOE distance constraints were derived from a 250-ms NOESY spectrum in H 2 O and a 200-ms NOESY spectrum in D 2 O solution, both recorded at 288 K, pH 3.0. Unambiguously assigned NOESY cross-peaks were integrated manually with XEASY, and the resulting volumes were converted into proton-proton upper distance limits with the program CALIBA (28) using five different classes of NOEs. Constraints for the backbone dihedral angle ⌽ were obtained from the 3 J HNH␣ coupling constants Յ 6 Hz. In these cases, a ⌽ angle between Ϫ85°and Ϫ35°was imposed. Structures of BK␤ 2 N were calculated with the program DYANA (version 1.5) (29) employing a simulated annealing algorithm in the torsion angle space. Structures from preliminary DYANA calculations were used to recalibrate the

BK␤ 2 N Inactivated BK Channels in a Ball-like Manner-
The functional characteristics of BK␤ 2 N were tested in inside-out patches from Xenopus oocytes expressing non-inactivating homomeric BK channels. As shown in Fig. 1A, BK␤ 2 N induced rapid inactivation of the BK ␣-subunit when present at the cytoplasmic side of the patch. Moreover, BK␤ 2 N-mediated inactivation occurred only at open channels. Despite the longlasting presence of BK␤ 2 N, channels first opened upon depolarization before they were inactivated by the peptide (Fig. 1,  A and B).
The time course of inactivation was strongly dependent on the BK␤ 2 N concentration (not shown) and exhibited mild voltage dependence (Fig. 1B). Thus, the time constant as obtained from a monoexponential fitted to the current decay ( inact ) changes e-fold with a change in membrane potential of 117 mV, which is equivalent to a valence (z) of 0.21.
These results suggested that very similar to IDs derived from Kv␣ or Kv␤1.1 subunits, BK␤ 2 N blocks BK channels in a "balllike manner" via interaction with a receptor site on the ␣-subunit that becomes accessible once the channel is in the open state. Therefore, interaction between BK␤ 2 N and the channel ␣-subunit was more closely investigated by the "fast application" technique. This technique allows for complete solution exchanges at inside-out patches in less than 2 ms and enables separate determination for on-and off-rates of channel-peptide interaction (17). Fig. 1C shows rapid application and wash-off of BK␤ 2 N at a concentration of 27 M. Channels were activated prior to peptide application by a voltage-step to 0 mV at a [Ca 2ϩ ] i of 10 M. Inactivation occurred with a time constant of Ϸ17 ms (16.9 Ϯ 1.5 ms, n ϭ 8), identical to that induced by the continuously present BK␤ 2 N (Fig. 1B). Wash-off of BK␤ 2 N, which should reflect unbinding of the peptide from the receptor, exhibited a time constant of Ϸ850 ms (851.5 Ϯ 69.3 ms, n ϭ 8) and could be well fitted with a monoexponential (Fig. 1C). This was an indication that interaction between BK␤ 2 N and its receptor on the ␣-subunit could be described as a bimolecular reaction as suggested (11) with on-and off-rates (k on , k off ) of 2.0⅐10 6 (Ms) Ϫ1 and 1.2 s Ϫ1 (Fig. 1D), respectively. The affinity (IC 50 ) of BK␤ 2 N for its receptor as calculated from these rates is 0.59 M (Fig. 1D), which is very similar to the value obtained from a steady-state concentration-inhibition relationship (not shown).  (31). ⌬␦H␣ was calculated as the difference between the experimentally determined ␦H␣ and the random-coil ␦H␣. Lower panel, bar diagram of NOE constraints by residue. Intraresidual contacts are shown in gray, sequential contacts in white, and medium range contacts in black.
Interestingly, a comparison among the inactivation rates of various IDs shows that k off of BK␤ 2 N closely resembles that of the compactly folded Kv3.4-ID, whereas k on of BK␤ 2 N is very similar to that of Kv␤1.1 or the Shaker B-ID, IDs that both lack ordered three-dimensional structure in solution ( Fig. 1D and Ref. 19). Next we investigated the structural properties of BK␤ 2 N in solution with NMR spectroscopy.
Assignment and NOE Connectivities of BK␤ 2 N-NMR experiments were performed under various conditions in aqueous solution at pH 3.0 and in a physiological salt solution at pH 6.0 (see "Materials and Methods"). The 1 H NMR resonances of BK␤ 2 N were completely assigned by two-dimensional NMR methods in the low pH solution and verified under physiological salt and pH conditions.
As illustrated in Fig. 2A, NOE contacts between nonadjacent amino acids (i,iϩx) indicative for structured domains were only observed on the sequence stretch roughly extending from Ser 10 to Leu 31 . This "core domain" exhibited NOE patterns typically observed with ␣-helices. Thus, connectivities between the ␣-proton of one amino acid and the amide (␣N(i,iϩ3)) or ␤-proton (␣␤(i,iϩ3)) of the third amino acid following were observed. Most residues throughout this stretch show sequential contacts between backbone amide protons (dNN NOEs). Moreover, 3 J HNH␣ coupling constants were determined in this region for 8 of 22 residues. Five of these J-couplings showed values between 5.3 and 6 Hz, indicative of helical conformations, and three were between 6 and 8 Hz.
The NOE-based indication of secondary structural elements in BK␤ 2 N was corroborated by results of an H/D exchange experiment, where one-dimensional spectra were recorded 15, 30, 45, and 65 min after dissolving lyophilized BK␤ 2 N in D 2 O. Thus, a number of amide protons (HN) including those of Ile 21 , Gln 23 , Ile 25 , Asp 29 , and Leu 31 were identified in the first and second one-dimensional spectrum (see supplemental material). Resonances of the HNs of Ile 25 and Leu 31 were present even in the spectrum recorded 65 min after dissolution, indicating significant protection from exchange with the solvent as would be expected for hydrogen bonding in a helical conformation. This view is further supported by the deviations of the ␣-proton chemical shifts from random-coil values (31). As shown in Fig.  2B (upper panel), the ␣-protons of all residues in the core region are shifted up-field as typically seen in helical structures (32).
Together, the NOE pattern, the H/D exchange, the J-couplings, and the chemical shifts suggest that BK␤ 2 N consists of an ordered mostly helical core domain flanked by flexible N and C termini.
Structure of BK␤ 2 N in Solution-A total of 728 experimentally determined NOE constraints (average of 16.2/residue; Fig.  2B, lower panel) together with the restraints for dihedral angles and stereospecific assignment of protons were used to calculate the solution structure of BK␤ 2 N (Table I). After structure calculations using the simulated annealing protocol of DYANA (29) in the torsion angle space 24 structures with the lowest values of the target function and without NOE violations of Ͼ0.4 Å were selected as the final family of BK␤ 2 N structures (for structural statistics see Table I). Fig. 3 shows 18 representatives of this family of best structures, superimposed either between residues 10 and 17 (Fig.  3A) or residues 18 and 31 (Fig. 3B). Both superpositions display reasonable convergence with similar r.m.s. deviation values to the mean structure (0.65 Ϯ 0.22 and 0.48 Ϯ 0.27 Å for the backbone atoms in Fig. 3, A and B, respectively), whereas superposition on the entire range was not meaningful because of the divergent orientation of residues Glu 17 -Arg 19 that connect both stretches. The Glu 17 -Arg 19 linker thus divides BK␤ 2 N into an N-and a C-terminal domain. The N-terminal domain consists of a disordered part made up of residues 1-10 and a loop-helix motif formed by amino acids Ser 11 -Asp 16 (Fig.  3A). Superposition of residues 10 -16 revealed r.m.s. deviations from the mean structure of 0.60 Ϯ 0.17 Å for backbone atoms and of 1.14 Ϯ 0.17 Å for all atoms. The C-terminal domain is made up of an extended helical structure formed by residues 20 -31 and a flexible C terminus (residues 32-45; Fig. 3B). Within the helical structure, residues 22-30 form a regular ␣-helix that is preceded by one turn of a 3 10  When correlated with functional properties, it is only the N-terminal domain that is required for occlusion of the channel pore, as seen in experiments with this domain fused to the KCNMB1 subunit or applied to BK channels as a synthetic peptide (11). Accordingly, this domain was termed the ball domain (Fig. 4). In contrast, the C-terminal domain, which links the ball to the transmembrane segment of KCNMB2, may be regarded as the chain domain (Fig. 4). DISCUSSION BK␤ 2 N inactivates BK channels with characteristics known from the "pore plug-in" described for inactivation domains of Kv-type K ϩ channels. Accordingly, BK␤ 2 N-mediated pore occlusion exhibits shallow voltage dependence and is competed by the pore-blocking agent TEA (11,13). As determined from NMR experiments, BK␤ 2 N presents with a unique solution structure; it consists of two domains connected by a flexible linker, the ball domain, made up of a disordered N terminus anchored at a loop-helix motif, and the chain domain, a 4-turn helix with an unfolded region at its C terminus.
As shown in Fig. 1B, the structure of BK␤ 2 N is accompanied by functional properties that are unique with respect to those of Kv-derived IDs. Thus, the association rate of BK␤ 2 N with its receptor on the channel is very similar to that observed with the unstructured IDs of Kv␤1.1 or Shaker B, but considerably slower than that determined for the compactly folded Kv3.4-ID (17,19,33). The dissociation rate of BK␤ 2 N, on the other hand, is more than 10-fold lower than that of Kv␤1.1-ID or the ID of Shaker B and even about 2-fold lower than that of Kv3.4-ID.
These correlations between structural and functional properties are consistent with our earlier observations that well ordered IDs exhibit a faster k on and a much slower k off than unfolded domains (17,19). The latter seems to be caused either by the higher number of molecular contacts (hydrogen bonds, etc.) formed between the folded domain and the receptor or by the higher flexibility of the unfolded IDs, which destabilizes the ID-receptor interaction.
Together with the observation that the actual pore block is realized by the N-terminal 19 or 26 residues (11), BK␤ 2 Nmediated inactivation may be imagined to occur as follows. The ball domain (Fig. 4) will approach the open channel and, in a second step, bind to its receptor, which finally results in occlusion of the channel pore. Channel approach and binding are reflected by k on and are determined by the flexible part of the ball domain as suggested recently for Kv␤1.1-mediated inactivation (34). Unbinding of the ID from its receptor, as reflected by k off , should be controlled by the structured part of the ball domain. As k off of BK␤ 2 N is the lowest of all ID peptides investigated to date, the ball-receptor interaction must be particularly strong. Interestingly, the structured region of the ball domain contains a cluster of charged residues suggesting that hydrophilic interactions may be an important determinant for the BK␤ 2 N-BK␣ interaction. This premise would be in line with work by Toro et al. (35) who investigated the interaction between the ID from Shaker B with BK channels and concluded that the ID receptor of BK channels in the inner vestibule of the channel may contain hydrophilic residues and a "pocket" that favors binding of helical structures.
The molecular identity of the BK ID receptor, however, must remain open at this point as well as the question of how far BK␤ 2 N enters the channel pore and whether interactions between charges on the ball domain and the channel wall are involved in ID receptor interaction. FIG. 4. Ball and chain domains of BK␤ 2 N. N-terminal 34 residues of the best BK␤ 2 N structure in a ribbon representation emphasizing the secondary structural elements between residues 11-17 and 20 -30. Functional analysis delineated residues 1-18 as the pore-occluding ball domain, and residues 20 -45 represent the chain domain linking the ball to the transmembrane core of the KCNMB2 protein. To emphasize the flexibility of the very N terminus (residues 1-10), two representatives of the family in Fig. 3 are shown.

FIG. 3. Solution structure of BK␤ 2 N exhibits two highly ordered domains.
A, backbone superposition (N, C␣, and C atoms) of the family of best structures (with lowest target function) of BK␤ 2 N between residues 10 and 17 (highlighted in red). B, backbone superposition shown as in A but between residues 18 and 31 (highlighted in red). Both panels depict 18 of the 24 best structures of BK␤ 2 N (see text). The N and C termini are indicated.