Mechanism of Fluoride Activation of G Protein-gated Muscarinic Atrial K+ Channels*

Aluminum fluoride (AlF;) activates the heterotri- meric G protein G. (stimulatory G protein of adenylylcyclase) (Sternweis, P. C., and Gilman, A. G. (1982) Proc. Natl. Acad. Sci. U. S. A. 79,4888-4891) and GT (transducin), and for GT, Bigay et al. (Bigay, J., De-terre, P., Pfister, C., and Chabre, M. (1985) FEBS Lett. 191,181-185) have made the intriguing proposal that AlF; acts by mimicking the y-phosphate of GTP. The endogenous G protein (probably Gai-* or GaiSS (Ya-tani, A., Mattera, R., Codina, J., Graf, R., Okabe, K., Padrell, E., Iyengar, R., Brown, A. M., and Birnbau-mer, L. Nature 336, stimulates the muscarinic atrial K+ (K+[ACh]) channel is also thought to be activated by investigate the A1F; mechanism, applied potassium fluoride (KF) the cytoplasmic face of inside-out membrane patches excised guinea pig atria. We found that KF activated single K+[ACh] channel cur- rents in both a concentration-

mer, L. (1988) Nature 336, 680-682) that stimulates the muscarinic atrial K+ (K+[ACh]) channel is also thought to be activated by AlF; (Kurachi, Y., Nakajima, T., and Ito, H. (1987) Circulation 76, 105P). To investigate the A1F; mechanism, we applied potassium fluoride (KF) to the cytoplasmic face of inside-out membrane patches excised from guinea pig atria. We found that KF activated single K+[ACh] channel currents in both a concentration-and a Mg2+-dependent manner. Activation persisted following removal of KF, but unlike activation by guanosine 5'-(3-thiotriphosphate) (GTPyS), was fully reversed by removal of M8+. Evidence for A13+ involvement was that the A13+ chelator deferoxamine (500 PM) inhibited KF activation and that at low concentrations of KF (<1 mM), micromolar AlCL concentrations potentiated KF stimulation. The rate of activation produced by KF was far slower than the rate produced by GTP or GTPyS, and unlike these guanine nucleotides, the rate was unchanged in the presence of agonist. To test the yphosphate-mimicking hypothesis, we evaluated the requirement for GDP; and to accomplish this, it was necessary to establish a condition that ensured exchange of guanine nucleotides. This condition was satisfied by using the muscarinic agonist carbachol because both the rate and the extent of activation of the K+[ACh] channels produced by GTP were much faster in carbachol, and both were greatly slowed when GDP was added along with GTP. By contrast, the effects of KF were unchanged by carbachol in the presence or absence of GDP. Further evidence that GDP is not essential for activation by AlFa was provided by the observation that during carbachol activation and following extensive washing with GMP, guanosine 5'-0-(2-thiodiphosphate) at blocking concentrations had no effect on activation produced by KF. We conclude that AlF; activates the endogenous G protein, but the mechanism appears to be more complicated than implied by * This work was supported in part by National Institutes of Health Grants HL36930, HL39262, and NS23877 (to A. M. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. the y-phosphate-mimicking hypothesis.
Ion channels are thought to be G protein' effectors, a good example being the muscarinic atrial K+ (K+[ACh]) channel in guinea pig atrium (1). The responsible G protein is called Gk (1-3) and is probably either of two a subunits, Gai.* or Gai.3 (4). Like other heterotrimeric G proteins, Gk can be activated reversibly by G T P (5) or irreversibly by GTPyS (3,6); and in either case, there is an absolute requirement for M e . Another agent that activates G proteins is AlF; (7), and activation by AlF; generally introduced as NaF or KF has been used as evidence for G protein effector coupling (8). Adenylylcyclase can be either stimulated or inhibited by NaFactivated G, or Gi, respectively (7, [9][10][11]. NaF also activates transducin and the G proteins responsible for polyphosphoinositide hydrolysis (12-14); and pertinent to our interest in ion channels as G protein effectors, an abstract has been published indicating that AlF; may activate Gk (15). Activation by AlF; is especially intriguing because it has been proposed that AlF; activates the a subunit of G proteins by mimicking the y-phosphate of G T P (12, 13). NMR spectroscopy of Go and G, was interpreted as supporting this view (16). Moreover, the concept that A1F; may act as a y-phosphate in the presence of nucleotide diphosphates has been extended to other enzymes that bind phosphate or nucleotide diphosphates (17). We examined the action of KF on K'[ACh] channel currents in excised inside-out patches from guinea pig atrial myocytes. KF stimulated the currents in a concentration-dependent manner, and there was an absolute requirement for Mg2+. However, in these experiments, the mechanism appeared to be independent of GDP, leading us to question whether the y-phosphate-mimicking hypothesis applies to all in uiuo circumstances.

MATERIALS AND METHODS
Single atrial cells were dissociated from adult guinea pig heart by collagenase digestion (3). Single channel currents were measured with the gigaseal patch-clamp method (18). Patch pipettes had tip resistances of 5-10 megaohms and were filled with a standard K+ solution (140 mM KC1, 2 mM MgCI2, 5 mM EGTA, 5 mM HEPES (pH 7.3 with Tris base)). The bathing solution (internal solution for insideout patch recording) had the same composition, unless otherwise noted. For Mg2+ experiments, EGTA was replaced by EDTA, and divalent cation concentrations were calculated with an interactive program that used the dissociation constant of Fabiato and Fabiato The abbreviations used are: G protein, signal-transducing guanine nucleotide-binding protein of subunit structure apy; G, and G,, stimulatory and inhibitory G proteins of adenylylcyclase, respectively; GT, transducin; Gt, G protein that stimulates a class of K' channels; GTPyS, (19) for EDTA. MgCI, and KF were added to the bathing solution; when KF concentrations >10 mM were tested, KC1 was replaced by KF. At higher intracellular Mg2' concentrations (10-20 mM), the current-voltage relationships shifted by -5 f 2 mV ( n = 4), and this shift was not compensated for. In eight experiments, both EGTA and EDTA were eliminated from the bathing solution. Ai3+ was present in trace amounts in our solutions since no special precautions were taken (7); but in some experiments, AI3+ was removed by adding the Al"' chelator deferoxamine (14,20).
All test agents were applied by the concentration clamp method (21,22), which allowed rapid solution changes to occur within 10 ms. All experiments were done at room temperature (20-22 "C).
Single channel currents were displayed on a chart recorder and were stored in a video recorder. The data were analyzed using an IBM PC/AT 386 processor. The records were low pass-filtered at 1-2 kHz (-3 db) and digitized between 5 and 10 kHz. Transitions between closed and open states were determined using an automated interactive threshold detection program. Mean open times were obtained by fitting open frequency histograms to a probability density function, and mean amplitudes (i) were obtained by fitting amplitude histograms to a gaussian function. The single channel currents were summed in buffers of 2048 sample points to give the total current ( I ) . The opening probability ( P for N channels, NP for each record) was calculated as Z divided by i . T o measure concentration-dependent effects of test agents, we averaged NP for periods of 20 s during control and after addition of test agents. The rates at which these agents produced their effects were estimated from NPs averaged every 500 ms for periods between 30 s and 5 min. An exponential function was fit to the rising phase after allowing for a delay; and from the fitted value, a half-time ( t H ) was calculated (see Fig. 3A).
Parameter estimates for all data fits were obtained using a maximum likelihood estimator and the Marquardt-Levenberg algorithm for nonlinear least-squares curve fitting. Mean values ? S.E. are given in the text.
KF, AICI,, carbachol, atropine, theophylline, and deferoxamine were obtained from Sigma. Nucleotides were from Boehringer Mannheim. Other chemicals were obtained from Fisher. All solutions were made with glass-distilled water. The purity of GDPpS was determined by high performance liquid chromatography (Beckman System Gold 126 gradient system with a model 166 detector).  (23). The current-voltage relationships of KF-and GTP-activated channels were indistinguishable (Fig. IC, panel 3 ) , and the single channel slope conductances of inward channel current were 38 -+ 1 picosiemens ( n = 4 for each condition). Analysis of single channel currents in eight patches showed that mean open times in KF-and GTP-activated channels at -80 mV were 1.3 k 0.3 and 1.2 k 0.5 ms, respectively. The burst durations were also identical in both cases. In the experiment shown in Fig. IC, average burst durations were 4.9 ms for GTP-activated channels (panel I ) and 4.5 ms for KF-activated channels (panel 2 ) .

K F Activation
Influence of MgZ+ on Activation of K+[ACh] Channels by KF-The rate of activation of K+[ACh] channel currents by KF was unexpectedly slow compared to that by GTP or GTPyS. Upon further examination, we found that M$+ was an important cofactor for KF effects on the K+[ACh] channels. In the presence of higher M$+ concentrations, KF (10 mM) activated the channel currents within 10 s (Fig. 2 A ) . Compared to the rates of activation for GTP or GTPyS, these rates were still slower, and the concentrations of M e were much higher.
In the absence of KF, Mg2+ had no effects; and without Mg2+, KF was ineffective (Fig. 2B). In contrast to the channel activation produced by GTPyS (24)(25)(26), the KF effects were reversed by washing with M$+-free solution (Fig. 2C). However, KF activation was irreversible in the presence of M$+ for up to 10 min of washing in KF-free solution (Fig. 2B).
The concentration-dependent effect of M$+ on KF activation was examined at a fixed concentration of KF (10 mM) (Fig. 3). The extent of activation was normalized to the maximum response. Mg2+ acted in a concentration-dependent manner on both the extent and the rate of activation (Fig.  3B). The ECso was 5.9 mM. For comparison, the effects of M e on carbachol (10 p~) plus GTP (100 p~) activation were much smaller, and the ECso was 3 p~ (27).
Concentration Dependence and Magnitude of KF Activation-Having established the optimum M C concentration, we turned to the concentration response to KF. Both the magnitude and the rate of activation were concentrationdependent (Fig. 4). The extent of activation measured at steady state was normalized by the maximum response to GTP-yS (100 p~) and gave the concentration-response curve shown in Fig. 4B. The ECbO was 3.8 mM. When GTPyS (100 p~) was applied first and allowed to produce its maximum effect, subsequent addition of KF (10 mM) had no further effect (n = 5 ) , indicating that activation by GTPyS and KF was not additive. When the opposite sequence was used, GTPyS always produced a large increase beyond the maximum increase attained by KF. KF activated K+[ACh] channel currents to the same extent in the presence or absence of the muscarinic agonist carbachol (10 p~) in the patch solution (Fig. 4B). Addition of the muscarinic antagonist atropine (10 p M ) and the purinergic antagonist theophylline (100 p~) to the pipette solution did not affect the activation ( n = 4). The activation produced by KF was always less than the maximum activation obtained by GTPyS (Fig. 4B). By contrast, GTP in the presence of agonist produced the same level of activation as GTPyS (5). In addition, the rate of activation produced by KF was unaltered by carbachol, whereas the rate of activation produced by GTP was greatly enhanced by carbachol (5).
Requirement of Aluminum Ion for KF Activation-As micromolar concentrations of aluminum ions ( A P ) were required for Factivation of G, (7), we have tested APeffects on KF activation. The effect of KF (1 mM) was potentiated by addition of 100 p~ A1C13 (Fig. 5A) KF activation was faster a t higher Mg2+ concentrations. B, M$+ requires KF to produce activation, and KF effects persist after removal of KF in the presence of Mg'+ (2 mM). C, KF-produced channel activity decreased quickly after removal of KF and Mg2+. The channel activity was restored by GTPyS in the presence of Mg'+ (2 mM). Recording conditions were identical to those described for Fig. 1 ( A and B ) .

>10 mM at both Mg2+ concentrations ( n = 4).
The chelator deferoxamine (14) was also tested for its ability to reverse the AI3' effects. At 500 WM, deferoxamine inhibited the 10 mM KF-stimulated channel by -40% (Fig.  5 B ) , but a complete block could not be obtained. On average, Mg2+ at 2 mM deferoxamine (500 p~) inhibited the K+[ACh]  Fig. 1 ( A and E ) Fig. l (A and B ) . by GTP plus agonist or GTPyS for up to 5 min, but longer exposure or higher concentrations blocked channel currents ( n = 6). The results described here permit us to attribute the effects of KF to AlF;. However, we were concerned that the much slower rate of activation produced by K F acting as A1F; might have been due to the slow release of A13+ from the EGTA or EDTA used in our bathing solutions. Therefore, in eight experiments, we omitted EGTA or EDTA and used MgC12 at 2 mM (three experiments) and at 20 mM (five experiments). The nominal free Ca2+ concentration in these experiments was M, and K+[ACh] channel currents were unaffected at these levels. The activation rates produced by K F were unchanged from those observed in the presence of EGTA or EDTA (Fig. 5C) and were 196 k 20 and 17.5 & 14 s at the lower and higher concentrations of MgC12, respectively.
Comparison of Effects of Muscarinic Agonist and GDP on Rates of Activation of Gk Produced by AlK, GTPyS, or GTP-This set of experiments was prompted by the very slow rates at which concentration jumps of AlF; produced activation of the K'[ACh] channel even at optimum concentrations of Mg2+ and the lack of effect of carbachol on the extent of activation produced by AlF;. We found that carbachol also had no effect on the rate of activation produced by A1F; (Fig.  6). The rate was measured as shown in Fig. 3A. By contrast, carbachol had large effects on the rates produced by concentration jumps of GTP or GTPyS (Fig. 6). The rates for A1F; were, as noted earlier, markedly slower than those for either of the guanine nucleotides. The rates for GTPyS were slightly slower than those for GTP, possibly as a result of slower association with the a subunit (28).
The lack of any effect of carbachol on the rate of activation produced by AlF; suggested that A1F; might not be simply mimicking the y-phosphate of GTP as proposed by Bigay et al. (12,13), but it was also possible that considerable amounts of Gk might not have been cleared of GDP in our experiments.
To examine this, we tested whether addition of GDP slowed the activation rates produced by either nucleotide or AlF;. Fig. 6A shows that this was clearly the case for GTP and GTPyS, but was not the case for AlF;. In fact, GDP at concentrations as great as 300 p~ had no effect on activation produced by AlF; ( n = 8 ) (Fig. 7A). Furthermore, prolonged exposure of tens of minutes to guanine nucleotide-free bathing solutions in the presence of carbachol did not diminish the subsequent activation produced by AlF;.
A strong test of the y-phosphate-mimicking hypothesis is a block by GDPPS, which prevents A1F; from acting as a yphosphate. We found that in the presence of agonist, GDPDS at concentrations of 43-86 pM ( n = 4) did not impair activa- Recording conditions were identical to those described for Fig. 1 (A and B ) .
p M to the GDPPS-containing solutions. As shown in Fig. 7B, K F activation was unchanged in the presence of both GMP and GDPPS, and the average rate of activation in 20 mM M P for these experiments was 24 f 5 s.

DISCUSSION
Our results showed that intracellular application of KF activated single-channel K+[ACh] channel currents that had unitary conductances and mean open times identical to the currents produced by muscarinic agonists such as carbachol in the presence of GTP or GTPyS in the presence or absence of carbachol (3,5). Like the guanine nucleotides, there was an absolute requirement for Mg". There were differences, however. Activation by GTPyS was irreversible on the time scale of these experiments even after removal of Mg2+ (24,26); and whereas K F effects were also irreversible after washing out KF, unlike GTPyS, the KF effects were fully reversed by washing out in Mg2+-free solution. Activation by K F differed in several other ways from activation by guanine nucleotides. Most notably, the rate of activation was much slower and was independent of agonist. Moreover, the extent of activation was much less than the maximum possible activation.
The concentrations of M$+ required for K F activation were -1000 times higher than those required for physiological activation of the K+[ACh] channels by GTP and carbachol (ECBO = 3 p~) (26,27). This may account for the fast recovery from KF effects in Mg2+-free solution. The M$+ concentration required for full KF effects found in our experiments exceeded the cytoplasmic free Mg2+ concentration thought to range from 0.4 to 3.5 mM in heart (29,30).
Activation of G proteins by Frequired AI"+, and the active ligand was thought to be AlF; (7,11). In our experiments, AlC13 (10-100 PM) potentiated submaximum effects of KF. However, AICls had no further activation of K+[ACh] channel currents when the KF concentration was >10 mM. The A13+ chelator deferoxamine (500 WM) reversibly inhibited the K F effects. Similar effects have been reported in K F stimulation of ATP-sensitive K' channels in an insulinoma cell line (RINm5F) (20). We conclude that the activating agent in our experiments is AlF;.
As to its mechanism of action, we assume that AlF; acts upon the a subunit of Gk just as it does on the a subunit of Go (31). This is supported by the observation that recombinant ai-3, which is a likely candidate for G, k (4), can stimulate single-channel K'[ACh] channel currents after activation by AlF; (32). There is no evidence that AlF; acted at the receptor level since muscarinic and purinergic receptor antagonists had no effects on KF stimulation. Nor is a direct effect on the K+[ACh] channel likely. In four experiments in the presence of 10 p~ carbachol, 100 p~ GTP, and 2 mM M$+, which produced maximum activation of single-channel K+[ACh] channel currents, the addition of 10 mM K F had no effect on mean open time (1.3 k 0.1 ms) or the single-channel current amplitude at -80 mV (2.1 f 0.06 PA). In another four experiments in the presence of 100 W M GTPyS and 2 mM M$+, which maximally activated the K+[ACh] channel, addition of 10 mM K F again had no effect on mean open time (1.4 f 0.05 ms) or amplitude at -80 mV (2.1 & 0.03 PA). These results are consistent with our observation that activation of the K+[ACh] channel by K F or guanine nucleotides is indistinguishable (Fig. 1).
Our results do not seem to support the attractive y-phosphate-mimicking hypothesis of Bigay et al. (12, 13). These authors arrived at the conclusion that AlF; activation of transducin requires GDP by clearing transducin of GDP using excess photoactivated rhodopsin. This result lead to the hypothesis that AlF; interacts with the GDP-bound form of transducin and that AlF; mimicks the role of the y-phosphate of GTP. Support for this view comes from recent spectroscopic measurements (16). In our experiments, in the presence of agonist, additional GDP had no effect on AlF; activation, although additional GDP had large effects on activation produced by GTP. Therefore, the Gk present in the membrane patches that was activating the K'[ACh] channel was capable of releasing its GDP in the presence of GTP and presumably AlF;. If this is so, then AlF; was producing its effects on Gk proteins that, in the presence of agonist and absence of guanine nucleotides, were cleared of GDP. Another test was our attempt to bind Gk with GDPPS, which does not support the action of AlF; as a y-phosphate. Again, to attempt to clear GDP, we used GMP in great excess. Despite our best efforts, it is possible that some Gk, which did not interact with receptor yet was GDP-bound, was being activated by AIF;. This issue can be resolved when the stoichiometry of the muscarinic receptor, Gk, and the K+[ACh] channel is known. However, another possibility is that the mechanism by which AlF; activates Gk in these membrane patches differs from the mechanisms reported for membrane-free GT, Go, or G,.