Calcium sensitivity of BK-type KCa channels determined by a separable domain

Neuron,

Vol.

13, 671-681,

September,

1994,

Copyright

0

1994

by Cell

Press

Calcium Sensitivity of BKType bz, Channels Determ ined by a Separable Domain Aguan Wei,* Christopher Solaro,+ Christopher and Lawrence Salkoff ** *Department of Anatomy and Neurobiology +Department of Anesthesiology *Department of Genetics Washington University School of Medicine St. Louis, Missouri 63110

Lingle,*+

Summary High conductance, Ca*+-activated (BK-type) K+ channels from mouse (mSlo) and Drosophila (dSlo) differ in their functional properties but share a conserved core resembling voltage-gated K+ channels and a tail appended to the core by a nonconserved linker. W e have found that the channel subunit is physically divisible into these two conserved domains and that the core determines such properties as channel open time, conductance, and, probably, voltage dependence, whereas the tail determines apparent Caz+ sensitivity. Both domains are required for function. W e demonstrated the different roles of the core and tail by taking advantage of the functional differences between mSloand dSlo. Heterologous pairing of cores and tails from mSlo and dSlo showed that single-channel properties were alwayscharacteristic of the core species, but that apparent Caz+ sensitivity was adjusted up or down depending on the species of the tail. Thus, the tail is implicated in the Ca*+-sensing role of BK channels.

Introduction Both membrane voltage and cytoplasmic Ca*+ participate in the activation of large conductance, Ca*+dependent (BK) K+ channels (Barrett et al., 1982; Meecn, 1978; Blatz and Magleby, 1987; Latorre et al., 1989), butthe relation between thevoltage-dependent and Ca*-dependent functions of the channel is not clear (Pallotta, 1985; Moczydlowski and Latorre, 1983; McManus and Magleby, 1991; Miller, 1987). The cloningand expression of BKchannelsfrom mouse (mSlo; Butler et al., 1993) and Drosophila melanogaster (dSlo; Atkinson et al., 1991; Adelman et al., 1992) demonstrated that BK channels are related structurally to voltage-gated K’channels. Alignment of the predicted primary sequences and hydrophilicity profiles of mSlo and dSlo revealed a core domain similar to that of voltage-gated K+ channels (Wei et al., 1990; Salkoff et al., 1992; Jan and Jan, 1992) that has six putative transmembrane segments, termed Sl through S6, a string of regularly spaced positive charges in S4, and a highlyconserved region between S5and S6defining the pore. However, the predicted primary sequences of mSlo and dSlo peptides are approximately twice the length typical of voltage-gated K+ channels. This

additional length is due to an appended sequence on the carboxyl side of the core domain that is well conserved between mSlo and dSlo but not represented in voltage-gated K+ channels. Prominent in the C-terminal region are four hydrophobic regions, termed S7 through SIO, of unknown topology and a poorly conserved segment between S8 and S9. This structural organization clearly suggests that BK channels evolved from voltage-dependent KC channels or shared a common ancestor and raises the possibility that the additional C-terminal sequence may participate in functions unique to BKchannels, for example, Ca*+-dependent gating. Precedence for this type of organization is demonstrated by the family of cyclic nucleotide-gated channels which have a core that resembles voltage-dependent channels (Kaupp et al., 1989; Goulding et al., 1992) as well as additional C-terminal structure required for transduction of cyclic nucleotide binding into activation (Goulding et al., 1994, Biophys. J., abstract). To address the potential role of the core and C-terminal regions in the gating behavior of BK channels, we have taken advantage of two naturally occurring BK channel variants, mSlo and dSlo, which, despite extensivesequencehomologies,exhibit markedlydifferent gating properties. In particular, the dSlo channel variant used here requires substantially higher voltages and Ca2+ concentrations ([Ca*+]) for activation and opens for much briefer durations than does mSlo. W e conceptualized two domains in the channel subunit: a “core” domain that includes segments Sl-S8 and a”tail”domain that includes segments S9and SIO. The two domains are linked by the poorly conserved peptide segment between S8 and S9. cRNA constructs encoding either the core or tail domains from both mSlo and dSlo were generated. For both mSlo and dSlo, cRNA core constructs injected into Xenopus laevis oocytes did not produce functional channels; however, if cRNA encoding the homologous tail region was coinjected along with the core, normal channel expression and function were restored. By examining the properties of chimeric channels created by coinjection of heterologous combinations of core and tail regions, we have investigated the role of the core and tail domains in the apparent Ca2+ dependence of channel activation and in single-channel gating behavior. Our results suggestthat Ca*+-independent gating properties are primarily determined by the identity of the core domain, whereas the apparent Ca*+ dependenceof gating is influenced bythe tail domain. Results Two Separable Domains Within the C-terminal domain of Slo channels are four hydrophobic segments, termed S7-SIO, of unknown topologythat are highly conserved between mSloand

Neuron 672

Figure 1. Derivation of Core and Tail Expression from Full-Length mS/o and &S/o cDNAs

Constructs

Alignment is shown of the predicted amino acid sequences of mSlo (mbr5; Butler et al, 1993) and dSlo (A2,C2,E2,G5,10 splice variant; Adelman et al., 1992). Identically conserved residues are shown as white type on black background; hydrophobic segments Sl-SlO are underlined. Two main domains of contiguous sequence conservation are evident, roughly encompassing SlS8 (Core) and S9-SIO (Tail). Core expression constructs, denoted as mS/o Core and dS/o Core, were created by the introduction of frame-shift mutations at mSlo K648 and dSlo D662, which result in theaddition of short peptidesequencesflowercaseletters enclosed in brackets) and stop codons (circled U’s). Tail expression constructs, mS/o Tail and dSlo Tail, were created by the removal of Send sequences preceding S9 and the utilization of either a native methionine for initiation (mSlo M682 [circled]) or a synthetic methionine (circled) with a 2 residue linker, ds (in brackets), preceeding dSlo M691.

dSlo (Figure 1 and Figure 2). Moreover, the region delimited by segments S9 and SIO is the most highly conserved region of the entire channel peptide. Between segments S8 and S9 is a peptide segment of

much lower sequence and length conservation (Figure 1; mSlo 1634-K687, dSlo T650-M773; Figures 2A and 2B). The separation of S9 and SIO from the rest of the channel subunit by this peptide segment suggested that each channel subunit might include two discrete domains: a core domain that includes segments Sl-S8 and a tail domain that includes S9 and SIO. We hypothesized that the core associates with the tail to form a functional channel, even if the two are separate peptides. To investigate this possibility, two modified mS/o cDNA expression constructs were created, one encoding the core Sl-S8 domain (mS/o Core) and one encoding the C-terminal S9-SIO domain (mS/o Tail; Figure 1). injection of either mS/o Core or mSlo Tail alone into Xenopus oocytes produced no detectable currents, as assayed by either whole-cell (Figure 3A) or single-channel recordings (mS/o Core, n = 26 patches; mS/o Tail, n = 6 patches). However, coinjection of mS/o Core with mS/o Tail (mS/o Core + mSlo Tail) produced robust expression of currents that were virtually indistinguishable in magnitude from those produced by full-length mS/o (Figure 3A). The size of these currents suggested that similar numbers of functional channels were being assembled from the two separate peptides as from the intact mSlo peptide. The Ca2+ and voltage dependence of Slo current activation was measured by excising inside-out patches containing large numbers of channels from oocyte membranes and perfusing the cytoplasmic face of the patch membrane with CaB-containing solutions (see Experimental Procedures). Oocytes were injected with either full-length mS/o or mS/o Core + mS/o Tail. Figure 3B shows macroscopic currents activated by voltage steps in the presence of IO PM Ca2+. Full-IengthmSloand themS/oCore+mS/o Tailcombination produced similar currents, with virtually superimposable conductance-voltage (G-V) relations (Figure 3C). G-V relations were fitted with a single Boltzmann equation G(V) = (1 + exp[-(V - V,)/k]j-’ to obtain values for half-maximal activation voltage (V,) and slope (k; Table 1). A plot of VSO versus -log[Ca*+] (pCa) shows that activation of the two currents was similar over a wide range of Ca2+ concentrations (Figure 3D). Conductance and gating behaviors of channels encoded by mSlo and mSlo Core + mS/o Tail were compared by recording single-channel currents in insideout membrane patches held at constant voltage and [Ca2+]. Figure 4A provides an example of such experiments and shows continuous activity of both channels at +30 mV and 10 PM CaZC. Measurement of channel amplitudes and opening and closing events revealed that mSlo Core + mSlo Tail single channels displayed conductance and kinetic behavior similar to that of full-length mSlo single channels. Pairing Core and Tail from Different Species Having shown that the mSlo Tail was essential for channel activity, we investigated the role of the core

Separable 673

Domains

of mSlo and dSlo

v50

Expression 0 I

I

40 1

200 I

II

400

I

Constructs 600 I

I

800 I

I

1000 I,,

10pMCa”

n

34.9 2 0.7

15

182 f 9.2

4

1200

mSl0

non conserved dSlo

mSl0 Core

+

mSlo Tail

N

C

35.2 f 0.5

c

d.Slo Core

+

mSlo Tail

mSl0 Core

+

dSlo Tail

11

9

Figure 2. mS/o, dS/o, and Derivative cRNA constructs of VW values.

are schematically

cRNA Combinations represented

Expressed

by their predicted

and tail domains in the apparent Ca* dependence of channel activation and in single-channel gating behavior. To this end, we exploited the functional differences between mSlo and a naturally occurring

in Xenopus

hydrophilicity

Oocytes profiles.

See the text and Table 1 for the significance

Drosophila slowpoke channel variant (d.Slo; variant A2,C2,E2,G3,10; Adelman et al., 1992) that requires stronger depolarization and more Ca*+ for activation than does mSlo. In addition, single dSlo channels

Neuron 674

mSlo Core + mSlo Tail

mSl0

-

mSl0 ‘1750 pA/ 1750 pA

4

-rcoinjected

7

mSl0 Core

-

L.---

mSlo Tail -150 -100

2~4

-50 0 test potential

50 100 (mV)

150

3

6

I

100 msec Figure 3. Coexpression

of mSlo Core with

mSlo Tail Reconstitutes

Channels

with

Properties

Similar

to Full-Length

mSlo

(A) Whole-cell currents from voltage-clamped oocytes injected with cRNA transcribed from mS/o, mS/o Core + mS/o Tail (coinjected), mS/o Core alone, and mS/o Tail alone. Currents were recorded in response to voltage steps from -80 to +80 mV, in increments of IO mV, from a holding potential of -90 mV. (8) Representative macroscopic K+ currents recorded in inside-out patches from Xenopus oocytes injected with either mSlo or mS/o Core + mS/o Tail cRNAs. For each experiment, the patch was held at -40 mV and stepped from -60 to +VO mV in 10 mV increments for 240 ms. This protocol was repeated 4-16 times. Averaged, leak-subtracted currents activated by 10 uM internal Ca2’ are shown. Currents activated in the absence of internal Ca*+ were used to determine the amount of leakage current in each patch. (C) Averaged, normalized G-V relations for mS/o (open symbols) or mS/o Core + mS/o Tail (closed symbols) at 10 uM (circles and triangles) and 300 uM Cap+ (diamonds). Conductance was determined using the relation G = I/&‘,,,, - V,,,), where I is the average current measured at the end of the voltage step, V trrt is the test potential, and V,,, is the reversal potential, which, for all experiments, was 0 mV. Conductance values at all potentials were normalized to the maximum conductance activated at a particular [Caz’]. G-V relations were fitted (solid lines) with a single Boltzmann function G = ([l + exp-(V,,,, - V&/k])-‘, where VW is the voltage at which 50% of the conductance is activated, and k is the slope factor. Error bars show SEM for 3-15 experiments, Fitted parameters f 90% confidence limits are listed in Table 1. (D) Average VW, plotted as a function of pCa for mSlo (open symbols) and mS/o Core + mS/o Tail (closed symbols). Ca2+ concentrations tested were 4, 10, 30, 60, 100, and 300 uM. Error bars show SEM for 3-15 experiments.

open for much briefer durations than do single mSlo channels. Despite these functional differences, a high degree of sequence homology between mSlo and dSlo suggested that functional chimeric channels might be created by coexpression of the core domain of one species with the tail domain of the other. A comparison of the voltage and Ca* sensitivity as well as the gating behavior of such chimeric channels with those properties of the full-length mSlo and dSlo channels might allow us to associate Ca2+-dependent gating with either the core or tail domain. We created dSlo expression vectors analogous to the mS/o Core and mS/o Tail vectors by first isolating a full-length cDNA encoding the dSlo channel from a Drosophila head cDNA library. This cDNA was then modified to create the dSlo Core and dSlo Tail cRNA expression vectors (Figure 1).

As with the mSlo channel, injection of d.Slo Core into oocytes did not produce functional channels (n = 10 patches); however, coinjection of dS/o Core with dS/o Tail yielded channels that were functionally indistinguishable from full-length dSlo channels (Figure 4). Low expression levels of dSlo Core + dSlo Tail channels prevented us from obtaining the macroscopic currents necessary to accurately measure the voltage and Ca*+ dependence of this channel. Coexpression experiments pairing the core of one species with the tail of the other produced functional chimeric channels. Recordings of channels encoded either by mSlo Core + dSlo Tail or dSlo Core + mSlo Tail revealed that conductance and open time were characteristic of the species of the core domain (Figure4). Single-channel records from all constructs containing an mS/o Core revealed a qualitatively similar

Separable 675

Domains

of mSlo

and

dSlo

B

mSl0

dSlo Bll,

[&+I = 10&l 0

mSlo Core + mSlo Tail

dSlo Core + dSlo Tail

-6.Y,fl 0

dSlo Core + mSlo Tail

mSlo Core + dSlo Tail

1ollM

ht4

0 +60

I 242pqAmsec 10 $4

Caz+

+30 mV, 160//160

Figure 4. Qualitatively

Similar

40

mV

mr-V

_pA

m M K+

Patterns

40 m*ec

of Gating

Behavior

in Native and Chimeric

510 Channels

Correlate

with the Core

Domain

(A) Inside-out patches containing single mSlo, mSlo Core + mSlo Tail, or mSlo Core + dSlo Tail channels recorded at +30 m V and 10 u M Ca2+. (B) Inside-out patches containing dSlo, dSlo Core + dSlo Tail, or dSlo Core + mSlo Tail channels. Patches were held at -40 m V and steppedto+60mVinthepresenceof10~MCa2’(dSloanddSloCore+mSloTail)or6O~MCa’+(dSloCore+dSloTail).Leakage-subtracted current traces are shown. Traces without channel activity show current activated by the same protocol in each patch in the absence of internal Ca*+,

pattern of gating behavior (Figure4A). Although open interval histograms were complex, containing three components, qualitatively similar open interval properties were observed for all constructs containing an m.S/o Core when compared at identical Ca*+ concentrations and voltages. In contrast, closed interval histograms, which contained five to six components, showed longer duration closures for the mSlo Core + dSloTail channel.The mean duration of all openings longer than 100 ps measured at +30 m V and 10 P M Ca2+ was 3.1 m s for mS/o (n = 2), 2.6 m s for mS/o Core + mS/o Tail (n = I), and 2.1 m s for mSlo Core + d.S/o Tail (n = 3), indicating that the mean open interval durations do not differ appreciably among constructs containing the mSlo Core. This conclusion is supported by additional patches examined at other Ca*+ concentrations and voltages. Because openings of channels composed of dSlo Cores were too fast to be well resolved, even at 300 u M Ca*, open time and conductancecould not be measured accurately. However, extremely brief open durations were characteristic of all three dSlo Core constructs, irrespective of the species of the tail (Figure 48). Macroscopic currents recorded from inside-out patches containing chimeric channels showed that activation was shifted rightward or leftward along the voltage axis, depending on the species of the tail domain. Pairing the mSlo Core with the dSlo Tail produced a marked decrease in apparent Ca2+ sensitivity relative to full-length mSlo, as shown by a 30 m V

rightward shift in the V5,, of the G-V curve at 10 P M Ca2+ (Figure 5A; Table I). The reciprocal experiment, coexpression of dSlo Core with mSlo Tail, produced channels with a markedly higher apparent Ca*+ sensitivity relative to full-length dSlo, reflected by a leftward shift in theVsO of activation greater than 60 m V at 10 u M Ca* (Figure 58; Table 1). In both heterologous coexpression experiments, the respective shifts in the V5,, of the G-V relations were very large at low [Ca2+] but diminished with increasing [Cal’] (Figure 5; Table 1). Thus, at high [Ca*+] (300 PM), the G-V relation of each chimeric channel approached that of the fulllength channel of the core species. This functional convergence at high [Ca*] for channels having a common core but either the mSlo Tail or dSlo Tail implies that the Ca2’-binding step is reaching saturation, at which point the two tails become functionally equivalent. The functional equivalence also implies that other determinants of gating are independent of Ca2+ binding and are probably properties of the core. Thus, differences in the G-V relations at high [Ca2+] between channels having different cores but the same tail may be accounted for largely by differences in Caz+-independent gating steps conferred by the cores. The relation between VsO and [Ca*+] provides information about the relation of voltage- and Ca*+-dependent steps in BK channel gating. For eac:h of the BK channels we studied, plots of VsO versus pCa show a curvature that reflects an apparent decrease in the Ca*+ sensitivity of activation as [Ca*+] is raised (Figure

NWVJtl 676

A mSlo Core + dSlo Tall

mSl0

6). We interpret this flattening of the VSO-pCa relation as evidence that voltage-dependent steps leading to activation are distinct from Ca*+-binding steps (see Discussion).

Discussion

L2500 pA 40 msec

0.0 -150

-100

-50 0 50 test potential (mV)

100

150

B dSlo

dSlo Core + mSlo Tall l

-

750 pA 1000 PA L6 msec

+

1.0

0.0 -50

I 0

I 50

I 100

150

test potentral (mV) Figure 5. Heterologous Coexpression Experiments with mSlo and c/S/o Core and Tail cRNAs Suggest That Ca*+- and Voltage Dependent Steps in Activation Are Separable (A) Representative currents recorded in inside-out patches from oocytes injected with either mS/o or with mSlo Core + dS/o Tail, with 10 FM internal Ca2+. Averaged, normalized G-V relations for mS/o Core + dS/o Tail (closed symbols) at 10 PM (circles) and 300 PM Ca2+ (diamonds) are shown below. G-V relations for mS/o (open symbols) are reproduced from Figure 3C for comparison. Solid lines are single Boltzmann relations fitted to G-V plots. Fitted parameters * 90% confidence limits are listed in Table 1. (B) Representative currents recorded in inside-out patches from oocytes injected with either dS/o or dS/o Core + mS/o Tail. Patches were held at -40 mV and stepped for 16-24 ms from -20 to +I40 mV in 20 mV increments. Averaged, leak-subtracted currents activated by IO fiM CaZ+ are shown for each test potential. The top trace in each set, marked with a diamond, shows current activated in the same patch at +I40 mV, 300 uM Cal+.

In this study, we demonstrate that BK-type, Cal+activated K+ channels encoded by mouse (mS/o) and Drosophila (&lo) cDNAs are composed of two structural domains: an N-terminal core domain that includes regions homologous to sequences encoding voltage-gated K+ channel subunits and a C-terminal tail domain that is unique to BK channels and is well conserved between mouse and Drosophila. Coinjection of separate cRNA constructs encoding both the core and tail domains into Xenopus oocytes produced channels that are functionally indistinguishable from their full-length counterparts. Furthermore, chimeric channels could be formed by coexpressing core and tail domains from mSlo and dSlo together. Taking advantageof majorfunctional differences between mSlo and dSlo channels, we compared the single-channel gating behavior and Ca2+ and voltage dependence of chimeric Slo channels with those of native Slo channels and were able to conclude that major determinants of single-channel gating behavior were associated with the channel core, whereas apparent Ca*+ sensitivity was largely influenced by the channel tail.

Association

of Core and Tail Domains

The ability to reconstitute functional channels from the coexpression of separate core and tail cRNAs suggests that the core and tail peptides associate with high affinity and that their association does not depend on a contiguous linkage. This suggestion is supported most convincingly by the observation that channels encoded by coinjected core and tail cRNA constructs show no sign of wash-out in excised insideout patches. The structural regions mediating the high affinity association between core and tail peptides are likely to be well conserved, because of the ability to reconstitute heterologous channels with core and tail domains from mouse and Drosophila. Formation of functional BK channels from noncontiguous pieces is reminiscent of the reconstitution of Na+ channels from separate cRNAs encoding repeated domains I-III and IV (Stuhmer et al., 1989). The reasons for loss of function by truncation of the tail domain are more speculative. The core peptide may not be transported to the plasma membranewithout the tail. Proper posttranslational folding and assembly of functional channels may require the pres-

Averaged, normalized G-V relations with fitted Boltzmann relations (solid lines) for dS/o (open symbols) and dS/o Core + mSlo Tail (closed symbols) at 10 PM (circles) and 300 PM Ca*+ (diamonds) are shown below. Error bars represent SEM of 4-5 experiments. Fitted parameters * 90% confidence limits are listed in Table I.

Separable 677

Domains

of mSlo and dSlo

Table 1. Half-Maximal

Activation

(VsO) and Slope (k) Values

for Averaged,

Normalized

10 uM Ca*+ Construct

V50 b-W

mSl0 mSlo Core + mSlo Tail mS/o Core + dS/o Tail dS/o dS/o Core + mS/o Tail

34.9 35.2 66.3 182 115

* * f f f

k (mV) 0.7 0.5 2.8 9.2a 1.2

18.0 18.6 20.0 13.8 23.4

* f f * f

0.5 0.4 1.2 3.1’ 1.2

V, and k values with 90% confidence limits were obtained by fitting single Data points with superimposed fits are shown in Figure 3 and Figure 4. a These values were obtained by constraining the maximum conductance

ence of the tail domain. It is known that minor conformational changes in the cystic fibrosis transporter receptor can cause the trapping of otherwise functional channels in endoplasmic recticulum (Cheng et al., 1990). Alternatively, the core could be constitutively inactive without its association with the tail peptide. In this case, the tail peptide may act to derepress the channel-forming core domain in a Ca*+dependent manner. We cannot distinguish between these possibilities with our present results.

N

VW (mv)

(15) (11) (9) (4) (5)

-33.1 -29.7 -40.0 94.0 83.9

Boltzmann

mSlo

+

dSlo

mSlo Core + mSlo Tail

+

dSlo Core + mSlo Tall

-A

mSlo Core + dSlo Tall

200 s E8

150

> 100

50

6

5

4 PCs

6

2.0 2.3 0.7 1.7 1.8

24.7 24.6 19.1 32.6 25.9

to averaged, in the presence

f + * f *

R-0 1.7 2.0 0.7 0.8 1.1

VI (3) (3) (4) (5)

normalized

G-V relations.

of 300 uM Caz+.

Figure6. VW, Depends linear Fashion

250

:

k (mV)

* 5 f f f

relations

to that activated

A

-50

Relations

Relating the position of G-V relations to Ca*+ sensitivity is not a new idea. For BK channels studied in different tissues, the position of the G-V curve at a particular [Ca*+] is often taken as an indication of Ca*+ sensitivity, and large differences in this “apparent” Ca*+ sensitivity among BK channels have been reported (McManus, 1991). However, such differences could arise from variations in any number of steps in the channel gating process, many unrelated to Ca2+binding steps per se. We therefore have used the term “apparent Ca2+ dependence” to indicate our uncertainty about the precise basis for shifts in G-V curves. Apparent Ca2+ dependence is a complex manifestation of the properties of all transitions leading to and including channel opening. For example, a structural differencethat slows thechannel closing rate,without changing actual Ca*-binding steps, would be expected to shift G-V curves leftward at all Ca*‘concentrations. A simple difference in the channel closing rate could, in fact, account for someof the largedifference in apparent Ca*+ sensitivity seen between the dSlo and mSlo variants studied here. Thus, the radically different single-channel gating behavior observed for channels having mSlo versus dSlo cores limits the extent to which the heterologous species of tail is able to shift the apparent Cati sensitivity. In particular, we would not expect the substitution of the mSlo tail to shift the apparent Ca*+ dependence

Structural Separation of Voltage and Ca*+ Dependence The association of voltage-dependent function and Ca*+-dependent function with the core and tail domains, respectively, is supported by experiments with both chimeric and native Slo channels. Analysis of singlechannels that included mSlo cores showed that the tail domain, whether from Drosophila or mouse, had little influence on the single-channel conductanceorthedistributionsofopendurations.Aqualitative comparison of the gating behavior of channels formed with dSlo cores and either dSlo or mSlo tails also showed no obvious differences. Thus, only a single property is influenced by the identity of the tail domain: the position of the macroscopic G-V relations along the voltage axis.

-4+

G-V

300 uM Ca*+

4

5 PCB

3

on pCa

in a Non-

Average VW plotted as a function of pCa for full-length and mixed-species, coinjected constructs (mean f SEM; n = 3-15). Data for mS/o and mS/o Core + mS/o Tail are replotted from Figure 3D for comparison. For each construct, G-V curves measured at the lowest [Caz+] tested were fitted by fixing G,,. to the C,,, obtained at 300 uM Ca*+ in the same patch. For all other Ca2+ concentrations, C,,, was a free parameter. The horizontal line drawn at +I40 mV (right) indicates the most positive voltage tested and provides a lower limit of confidence for points above the line. CaZ+ concentrations tested were 4, 10, 30, 60, 100, and 300 PM.

Neuron 678

of dSlo completely to that of mSlo, since the rapid gating behavior conferred by the dSlo core will likely result in lower channel open probability at comparable levels of occupancy of Ca2+-binding sites. Likewise, we would not necessarily expect substitution of the dSlo tail to shift the apparent Ca2+ dependence of mSlo completely to that of dSlo. More detailed single-channel analysis will be required to confirm these interpretations. Thus, despite the uncertainties inherent in assigning mechanistic interpretations to shifts in G-V curves, two results lead us to suggest the simplistic view that the tail domain has an important and specific role in determining true CaZ+ sensitivity. First, the nature of the tail domain does not affect the singlechannel gating of either the mSlo or dSlo core. Second, at high [Ca2+], the dSlo and mSlo tails confer equivalent effects on a single type of core. This suggests that once Ca*-dependent steps determined by the tail domains are saturated, the channel gating behavior is determined by the core and is not modified by the species of the tail present. Furthermore, both results imply that the tail domain only affects gating steps preceding channel opening. An issue that remains unresolved is that the curvature in the VsO versus pCa relation appears to differ from previous descriptions of BK channel gating behavior (Moczydlowski and Latorre, 1983). This point is discussed in more detail below. Following the above arguments, we prefer the view that differences in apparent Ca2+ sensitivity between native mSlo and dSlo channels reflect different affinities of the channel protein for Cap or, if affinities for Ca2+ are similar, a difference in the ability of the channels to transduce Ca* binding into channel opening, or both of these possibilities. We interpret the fact that the dSlo and mSlo tails affected activation similarly at high [Ca”] as evidence that Ca2+-dependent steps leading to channel opening are approaching maximum activation and that transduction may be equally conferred by the two tails. Since the tail region contains the most highly conserved primary structure of the entire protein, it is quite possible that the structural determinants of transduction are indeed highly conserved. These results do not prove that CaLi binding takes place in the tail region, although it is a reasonable possibility. The tail region does not contain an EFhand Ca2+-binding consensus sequence, but it is rich in conserved aspartate residues, which are key residues in many Ca2+-binding domains (Nayal and Di Cera, 1994). Indeed, preliminary results mutating these conserved aspartate residues do suggest their involvement in determining Caz+ sensitivity of gating (Schreiber et al., unpublished data). Implications for BK Channel Gating Plots of VSOversus pCa for the BK channel constructs studied here revealed a trend for the G-V relations to become less dependent on Ca2+ as [Ca?‘] was

increased. An established model of BK channel activation proposes that the voltage dependence of activation arises from the voltage dependence of Ca2+-binding steps (Moczydlowski and Latorre, 1983). Specifically, the model postulates two voltagedependent Ca2+-binding steps separated by a voltageindependent conformational step corresponding to channel opening and closing. This model successfully accounted for the average open probability of BK channels over a wide range of voltages and Ca2+ concentrations and was also consistent with many features of the kinetic behavior of the channels. However, a prediction of this model, that the voltage of half-activation should vary linearly with pCa, is apparently in disagreement with the results described here. It has been shown that Ca2+ can block BK channels (Vergara and Latorre, 1983), and we were concerned that the curvature observed in the VsO-pCa relation might arise from blockade of Slo channels by Ca2+. We do not believe this is the case, however, for the following reasons: first, mSlo single-channel open probability remained above 90% at 300 PM Ca2+ and +30 mV; second, significant reduction of peak macroscopic currents was not observed until the [Ca2+] was elevated beyond 300 PM, only at positive membrane potentials; third, G-Vs measured at Ca*+ concentrations high enough to block current significantly would not be affected at voltages near VsO because blockade by Ca2+ is favored by positive potentials. The apparent equilibrium dissociation constant for the binding reaction of slow Ca2+ block, as reported by Vergara and Latorre (1983), is greater than 1 M at -25 mV;fourth,allcurrentsweremeasuredusingsymmetrical160 m M K+, under which conditions blockade by Ca2+ is minimized by competition between K’ and Ca2+ for the blocking site in the channel pore (Vergara and Latorre, 1983). To address further the possible origins of the discrepancy between our findings and those of Moczydlowski and Latorre (1983), we extended the range of [Ca2+] over which we have examined the VSO-pCa relation. We have found that 1 m M Ca2’ continues the observed flattening in the VSO-pCa relation, whereas raising [Ca*] to 10 m M results in a pronounced additional leftward shift of about 40 mV (n = 6 patches). However, the additional leftward shift at 10 m M Cal’ appears unrelated to any Ca2+-specific activation process since 10 m M MgL+, which is unable to activate mSlo channels by itself, appears to be about equally effective at producing the leftward shift in VsO. For example, the Vsa for 100 PM Ca2+ in the presence of 10 m M MgL’ is not significantly different from the VsO for 10 m M CaLi (V5,, = -70.4 + 1.7 mV; n = 5 for 100 f.rM CaZ+, 10 m M Mg’+; VsU = -72.0 +/- 3.6 mV; n = 7 for 10 m M Ca”). The ability of both MgL’ and Ca2+ to produce the additional leftward shift at 10 m M suggests that a portion of the effect involves screening of surface charges (Hille, 1992), perhaps by an attraction of CaL+ to the negative portions of the protein itself. mSlo (mbr5) contains 126 aspartate and gluta-

Separable 679

Domains

of mSlo and dSlo

mine residues, the majority of which are probably on the cytoplasmic side of the protein (97 of these residues are downstream of S6, which is thought to terminate intracellularly). Alternatively, the leftward shift may correspond to a previously described allosteric effect of Mg*’ (Golowasch et al., 1986). In either case, the additional leftward shift appears unrelated to Ca2+-selective steps involved in BK channel activation, which appear to saturate over the range of 100 PM to 1 m M Ca*+. Thus, our data show that, for mSlo channels expressed in Xenopus oocytes, the Vso begins to reach a limiting value over the range of 100 PM to 1 m M Ca*+. A possible explanation for the apparent discrepancy between the present results and the results of Moczydlowski and Latorre is that the essential observations are similar but that the measurements described here are more amenable to resolving curvature in the VSO-pCa relation. Although the bilayer work offered the advantage of direct estimates of single-channel open probability, the known channel to channel variability required that open probability estimates be made on individual channels in single bilayers over a wide range of voltages and Ca*+ concentrations. As a consequence, in single-channel experiments, it may be difficult to obtain the resolution that could demonstrate or preclude the existence of curvature in the dependence of V~O on pCa. How might curvature in the relation between V5,, and pCa arise? Qualitatively, this phenomenon indicates that some Ca*+-dependent process is reaching a limiting rate, such that additional increases in [Ca*+] are ineffective at shifting activation to more negative voltages. We have approached this problem through a series of simulations of BK channel behavior based on particular Markovian models. Curvature and flattening in the dependence of VSOon pCa of the type observed in our data is best produced by models in which Cap-binding steps are distinct from voltagedependent transitions. However, a decrease in slope in the VSo-pCa relation can also occur in models involving multiple voltage-dependent Ca*+-binding steps, if the most voltage-dependent binding steps occur earlier in the activation sequence. The primary conclusion from our simulations is that VsO varies linearly with pCa for models in which approximately equal voltage dependence is assigned to all Ca*+binding steps. In contrast, flattening in the Vso-pCa relation is more likely to occur if voltage-dependent steps are distinct from Ca*+-binding steps. We take this analysis as support for the idea that BK channel activation can best be accounted for by a category of models with separate voltage and Ca*+-dependent activation steps. Modular Construction and Evolution of Ion Channels Modular construction of ion channels from structural components encoding discrete functional domains is a recurrent theme in evolution. The S4 voltagesensing motif, the P or H.5 pore-forming motif, as well

as the N-terminal inactivation ball motif have been observed in a wide range of cloned ion channel genes (Jan and Jan, 1992) and in some instances can be functionally transplanted into heterologous channels (Hartmann et al., 1991; Logothetis et al., 1993). More recently, the cloning of channels encoding inward rectifier K+ channels (Ho et al., 1993; Ku bo et al., 1993) suggests that the S5-P-S6 motif should be viewed as a discretefunctional domain, related through evolution to voltage-gated K+ channels. Like mSlo and dSlo channels, cyclic nucleotidegated channels (Kaupp et al., 1989; Warmke et al., 1991; Briiggemann et al., 1993), represented by retinal and olfactory cCMP-gated cation channels, are organized intoacoredomain that resemblesvoltage-gated K+ channels and a seemingly appended carboxyl tail domain that is thought to bind cGMP. Although our results with mSlo and dSlo do not prove that Ca2+ binds to the tail domain, they do imply a modular construction in which the tail domain is the major determinant of apparent Ca*+ sensitivity and the core domain confers Ca*+-independent properties of channel gating.

Experimental

Procedures

mS/o and dSlo Expression Constructs mS/o expression constructs were derived from the full-length mbr5splicevariant (Butleret al., 1993), subcloned as a Clal-EcoRI fragment into pBScMXT, a pBlueScript (Stratagene) derivative that we modified to incorporate Sand 3’untranslated sequences from the Xenopus b-globin gene (Melton et al., 1984). mS/o Core was constructed from a cDNA variant from a mouse brain cDNA library (Clonetech) that contained an unexplained frame shift at K649. mSlo Tail was generated by making use of a unique Bell site immediately upstream of mSlo M682; the Sl-S&l portion of mbr5 was removed by digestion with Clal and Bell, followed by treatment with Klenow fragment to create blunt ends and by religation to itself. dsloexpression constructs werederived from a full-length cDNA isolated from a Drosophila head cDNA library (provided by P. Salvaterra), screened with a probe generated by polymerase chain reaction (PCR), using oligonucleotide primers based on published sequence (Atkinson et al., 1991; Adelman et al., 1992). This cDNA was found by sequencing to be the A2,C2,E2,C5,10 splice variant (Adelman et al., 1992). Three minor differences from the published sequence were found that resulted in amino acid substitutions: C77cTG(:) to S(TCG), D281(sAT) to N@AT), and N696(AA-C) to S(AGC). To achieve useful levels of functional expression, the na%e 5’ untranslated sequence was removed, and a consensus sequence for translational initiation (CCCCCCATGG; Kozak, 1987J was Inserted at the initiator methionine, by PCR mutagenesis. Native 3’untranslated sequence was replaced partially with 3’ untranslated Xenopus P-globin sequence from pBScMXT. dS/o Core was derived from the full-length dS/o expression construct by utilizing EcoRV, which cuts at D662 and in the 3’ untranslated sequence. Removal of the EcoRV-EcoRV fragment, followed by religation, resulted in a frame-shift mutation at D662, which added a short exogenous length of amino acid sequence, as noted in lower case singleletter amino acid code (ihrstan), followed by a stop codon. dS/o Tail was created by introducing a synthetic translational initiation sequence (encoding mds), immediately upstream of M691. This sequence was incorporated into a double-stranded adaptor fragment generated from two complementary oligonucleotides designed to be compatible with ends generated by EcoRl and Sphl. The Sportion of dS/o was removed by digesting with EcoRl and Sphl, followed by religation in the presence of the synthetic

Nl?llr0ll 680

adaptor junction

fragment. All constructs were verified points and PCR-generated portions.

by sequencing

Expression in Xenopus Oocytes Capped cRNAwas prepared by runoff transcription with T3 RNA polymerase using a commercial kit (mMessage mMachine, Ambion) and resuspended to 0.5-1.0 mglml. Mature stage IV Xenopus oocytes were prepared for injection as previously described (Wei et al., 1990). Approximately 50 nl of cRNA was injected per oocyte, and oocytes were then incubated for l-9 days in ND96 (96 m M NaCI, 2.0 m M KCI, 1.8 m M Car&, 1.0 m M MgCI,, 5.0 m M HEPES [pH 7.51) supplemented with sodium pyruvate (2.5 mM), penicillin (100 U/ml), and streptomyocin (100 mg/ml) before recording. Electrophysiology Whole-cell recordings were obtained 2-4 days after injection, by two electrode voltage clamp in ND96 at 24OC. Records were low pass filtered at 1 kHz and leak subtracted with capacitative transients removed. Currents were recorded in inside-out patches from oocytes l-9 days after injection with cRNAs. Patches used to construct G-V relations were obtained with electrodes of 2-5 MD resistance and typically contained large numbers (>50) of mSlo channels. In all experiments, the duration of voltage steps was adjusted so that the current was fully activated during the step. Because expression of the dSlo Core + mSlo Tail combination was lower than that of other experimental conditions, macropatches (patch pipette resistance 0.3-1.0 MD) were used to record currents for G-V analysis. To obtain patches containing a single channel, cRNA was diluted 50-to IOO-fold before injection. Analog data were recorded unfiltered onto videotape, then filtered at 6 kHz (Bessel lowpass filter;-3dB)duringdigitizationat30msintervals. Single-channel analysis employed standard half-amplitude detection methods to generate lists of open and closed intervals from activity recorded at 4-300 uM Ca*+ and at voltages from -30 mV to +60 mV. Gigaohm seals were formed on oocytes bathed in normal frog Ringer’s solution (115 m M NaCI, 2.5 m M KCI, 1.8 m M CaCI,, IO m M HEPES [pH 7.41) and transferred after excision to test solutions containing Ca*+. Pipette-extracellular solution was 140 m M potassium methanesulfonate, 20 m M KOH, 10 m M HEPES, 2 m M MgCh (pH 7.0). Test solutions bathing the cytoplasmic face of the patch membrane contained 140 m M potassium methanesulfonate, 20 m M KOH, IO m M HEPES (pH 7.0), and one of the following: 5 m M EGTA (nominally 0 uM Ca2+), 5 m M HEDTA (4 and IO uM Ca*‘), or no added Ca2+ buffer (30, 60, 100 and 300 PM Ca*). The amount of calcium methanesulfonate required to obtain a specific free Ca” concentration was determined using a computer program for similar chloride-based solutions, measuring the free Ca2+ in the chloride-based solutions with a Ca*+sensitive electrode (Orion), then adjusting the amount of added Ca2* in methanesulfonate-based solutions so that the electrode measurement matched that of the chloride-based solutions. Local perfusion of membrane patches were as previously described (Solar0 and Lingle, 1992). Modeling of Activation Behavior A program (written by C. L.) that generates the macroscopic currents expected for any stochastic model with up to 30 states was used to deriveexpected VW-pCa relations. Steady-stateoccupanties of channel states were determined for different models over voltages from -400 to +4QO mV (5 mV increments) and [Caz’] from 1 nM to 1 M. At each [Ca”], the conductance as a function of voltage was determined, and the voltage yielding the halfmaximal conductance (V,) for that concentration was determined by interpolation.

terization of our dS/o cDNA. Thanks also go to M. Schreiber, S. Tsunoda, M. Saito, and T. Jegla for many helpful discussions and to N. Fowler for excellent technical assistance. This work was supported by grants from the National Institutes of Health (L. S. and C. L.), Monsanto-Searle (L. S.), and the Muscular Dystrophy Association of America (L. S.). 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 USC Section 1734 solely to indicate this fact. Received

April

29, 1994; revised

June 15, 1994.

References Adelman, J. P., Shen, K.-Z., Kavanaugh, M., Warren, R. A., Wu, Y.-N., Lagrutta, A., Bond, C. T., and North, R. A. (1992). Calciumactivated potassium channels expressed from complementary DNAs. Neuron 9, 209-216. Atkinson, N., Robertson, G., and Ganetzky, B. (1991). A component of calcium-activated potassium channels encoded by the Drosophila s/o locus. Science 253, 551-553. Barrett, J., Magleby, K., and Pallotta, B. (1982). Properties of single calcium-activated potassium channels in cultured rat muscle. J. Physiol. 337, 211-230. Blatz, A., and Magleby, K. (1987). Calcium-activated channels. Trends Neurosci. 70, 463-468.

potassium

Brtiggemann, A., Pardo, L., Sttihmer, W., and Pongs, 0. (1993). Ether-d-go-go encodes a voltage-gated channel permeable to K’ and Ca2+ and modulated by CAMP. Nature 365,445-448. Butler, A., Tsunoda, S., McCobb, D., Wei, A., and Salkoff, L. (1993). mS/o, a complex mouse gene encoding “maxi” calciumactivated potassium channels. Science 267, 221-224. Cheng, S. H., Gregory, R. J., Marshall, J., Paul, S., Souza, D. W.. White, G. A., O’Riordan, C. R., and Smith, A. E. (1990). Defective intracellular transport and processing of CFTR is the molecular basis of most cystic fibrosis. Cell 63, 827-834. Colowasch, J., Kirkwood, A., and Miller, C. (1986). Allosteric effects of Mg*+ on the gating of Caz+-activated K+ channels from mammalian skeletal muscle. J. Exp. Biol. 124, 5-13. Goulding, E. H., Ngai, J., Kramer, R. H., Colicos, S., Axel, R., Siegelbaum, S. A., and Chess, A. (1992). Molecular cloning and single-channel propertiesofthecyclic nucleotide-gated channel from catfish olfactory neurons. Neuron 8, 45-58. Hartmann, H.A., Kirsch, C. E., Drewe, J. A.,Taglialatela, M., Joho, R.H., and Brown, A.M. (1991). Exchange of conduction pathways between two related K+ channels. Science 257, 942-944. Hille, B. (1992). Ionic edition. (Sunderland,

Channels of Excitiable Membranes, Massachusetts: Sinauer Associates

2nd Inc.).

Ho, K., Nichols, C., Lederer, W. J., Lytton, J., Vassilev, P., Kanazirska, M., and Hebert, S. (1993). Cloning and expression of an inwardly rectifying ATP-regulated potassium channel. Nature 362, 31-38. Hoshi, T., Zagotta, W., and Aldrich, R. (1990). Biophysical and molecular mechanisms of Shaker potassium channel inactivation. Science 250, 533-538. Jan, L. Y., and Jan, Y. N. (1992). Tracing Cell 69, 715-718.

the roots of ion channels.

Kaupp, U. B., Niidome, T., Tanabe, T., Terada, S., Bonigk, W., Stiihmer, W., Cook, N. J., Kangawa, K., Matsuo, H., Hirose, T., and Numa, S. (1989). Primary structureand functional expression from complementary DNA of the rod photoreceptor cyclic GMPgated channel. Nature 342, 762-766.

Acknowledgments

Kozak, M. (1987). At least six nucleotides preceding the AUC initator codon enhance translation in mammalian cells. J. Mol. Biol. 796, 947-950.

We wish to acknowledge gratefully A. Butler for isolation of the mS/o core cDNA and D. McCobb for isolation and initial charac-

Kubo, Y., Baldwin, T. J., Jan, Y. N., and Jan, L. Y. (1993). Prtmary structure and functional expression of a mouse inward rectifier potassium channel. Nature 362, 127-133.

Separable 6431

Domains

of mSlo and dSlo

Latorre, R., Oberhauser, A., Labarca, P., and Alvarez, 0. (1989). Varieties of calcium-activated potassium channels. Annu. Rev. Physiol. 57, 385-399. Logothetis, D. E., Kammen, B. F., Lindpaintner, K., Bisbas, D., and NadaCGinard, B. (1993). Cating charge differences between two voltage-gated K+ channels are due to the specific charge content of their respecitve S4 regions. Neuron 70, 1121-1129. McManus, regulation

0. B. (1991). Calcium-activated potassium channels: by calcium. J. Bioenerg. Biomembr. 23, 537-560.

McManus, 0. B., and Magleby, K. L. (1991). Accounting for the Caz+-dependent kinetics of single large-conductance Ca*+activated K+ channels in rat skeletal muscle. J. Physiol. 443, 739777. Meech, nervous

R. (1978). Calcium-dependent tissues. Annu. Rev. Biophys.

potassium activation Bioeng. 7, 1-18.

in

Melton, D., Krieg, P., Rebagliati, M., Maniatis, T., Zinn, K., and Green, M. (1984). Efficient in vitro synthesis of biologically active RNA and DNA hybridization probes from plasmids containing a bacteriophage SP promoter. Nucl. Acids Res. 72, 7035-7056. Miller, C. (1987). How ion channel proteins work. In Neuromodulation, the Biochemical Control of Neuronal Excitability, L. Kaczmarek and I. Levitan, eds. (New York: Oxford University Press), pp. 39-63. Moczydlowski, E., and Latorre, R. (1983). Cating kinetics of Ca2+activated potassium channels from rat muscle incorporated into planar lipid bilayers: evidence for two voltage-dependent CaZ+ binding reactions. J. Cen. Physiol. 82, 511-542. Nayal, M., and Di Cera, E. (1994). Predicting Ca2+-binding in proteins. Proc. Natl. Acad. Sci. USA 97, 817-821.

sites

Pallotta, B. (1985). N-bromoactetamide removes a calciumdependent component of channel opening from calcium-activated potassium channels in rat skeletal muscle. J. Cen. Physiol. 86, 601-611. Salkoff, L., Baker, K., Butler, A., Covarrubias, M., Pak, M., and Wei, A. (1992). An essential “set” of K’ channels conserved in 75, 161-166. flies, mice and humans. Trends Neurosci. Solaro, C. R., and Lingle, C. J. (1992). Trypsin-sensitive, inactivation of a calcium-activated potassium channel. 257, 1694-1698.

rapid Science

Sttihmer, W., Conti, F., Suzuki, H., Wang,X. D., Noda,M.,Yahagi, N., Kubo, H., and Numa, S. (1989). Structural parts involved in activation and inactivation of the sodium channel. Nature 339, 597-603. Vergara, channels evidence 568.

C., and Latorre, R. (1983). Kinetics of Ca2+-activated K+ from rabbit muscle incorporated into planar bilayers: for a Caa+ and Ba*+ blockade. J. Gen. Physiol. 82, 542-

Warmke, J., Drysdale, R., and Ganetzky, tassium channel polypeptide encoded locus. Science 252, 1560-1562.

B. (1991). A distinct by the Drosophila

poeag

Wei, A., Covarrubias, M., Butler, A., Baker, K., Pak, M., and Salkoff, L. (1990). Diverse K+ currents expressed by a Drosophila extended gene familywhich is conserved in mouse. Science248, 599-603.