Allosteric Effects of Mg²⁺ on the Gating of Ca²⁺-Activated K⁺ Channels From Mammalian Skeletal Muscle

Ion channel proteins form the basis of all electrical signalling in both excitable and nonexcitable cells. Not only are they the effectors of the ion currents leading to transmembrane voltage changes, but must also serve receptor-like functions. They must be able to register voltage changes, detect internal or external ligands and even covalent modification by enzymatic processes. Thus, channels are not simply the sewer-pipes drawn in low-resolution diagrams of their structures, but rather are intricately constructed membrane proteins displaying a rich phenomenology. The advent of single-channel recording techniques (Sakmann & Neher, 1983) has made ion channels unique among proteins, in that we can now observe the behaviour of individual protein molecules. The literature is already replete with examples of the ease with which mechanistic information can be extracted from this unprecedented capability to observe a protein’s function.

Single channels may be studied in two different kinds of systems: in the cell membrane itself, using ‘patch recording’ methods, or in a reconstituted system consisting of a model phospholipid membrane into which channels are inserted. Each system has its own advantages: the former being unequivocally bound to physiological processes, and the latter exploiting the simplicity of a chemically defined system. The main interest of this laboratory lies in the relationships between protein function and the underlying molecular structure of ion channels, and we have thus relied solely on the ‘cleaner’ reconstituted system.

In this contribution, we do not review single-channel analysis or channel reconstitution methods. Recent surveys of these fields are available (Sakmann & Neher, 1983; Miller, 1986; Latorre, 1986). Instead, we present an example of the use of both single-channel and reconstitution approaches to answer a particular scientific question. The question concerns the interaction of a Ca²⁺-activated K⁺ channel with its activating ligand, internal Ca ion, and the modulation of this interaction by Mg²⁺. It is a study which will illustrate several aspects of this new technology: the degree of confidence offered by single-channel analysis for knowing exactly what is being measured, the ‘cleanness’ of using a single-channel protein at essentially infinite dilution in a reconstituted bilayer membrane, and the convenience and ease with which data can be collected, analysed and interpreted.

Many cellular processes respond to the concentration of free cytoplasmic Ca²⁺ through the intervention of specific Ca²⁺-binding proteins. Some of the best studied biochemical examples include activation of muscle by troponin and of numerous enzymes by calmodulin. In these and other systems, the activation of the target protein varies sigmoidally with [Ca²⁺]. This sigmoidicity makes physiological sense in that Ca²⁺ can act as a ‘switch’, in which small changes in concentration of the cation can cause large changes in protein activity; mechanistically, it is a consequence of the protein’s containing multiple Ca²⁺-binding sites which interact cooperatively.

This study is concerned with another cooperative Ca²⁺-binding protein: a Ca²⁺-activated K⁺ channel residing in the plasma membranes of many types of animal cells. This channel protein has not been purified, but its function can be studied in great detail via single-channel analysis, by directly observing the opening and closing of individual channels. Previous work (Barrett, Magleby & Pallotta, 1982; Methfessel & Boheim, 1982; Moczydlowski & Latorre, 1983; Magleby & Pallotta, 1983) has shown that the channel’s conducting, or ‘open’, conformation is favoured as cytoplasmic [Ca²⁺] increases, and that this activation process is cooperative. Opening probability increases sigmoidally with [Ca²⁺], giving Hill coefficients ranging from 1·2 (Moczydlowski & Latorre, 1983) to 3·6 (McManus & Magleby, 1985). These different degrees of cooperativity have led to proposals involving two, three or four Ca²⁺-binding sites on the channel protein.

It is known that several Ca²⁺-binding proteins also bind Mg²⁺, an ion abundant in the cytoplasm (about 10 mmol l⁻¹ total concentration, of which perhaps 1 mmol l⁻¹ is free). During our studies on Ca²⁺-activated K⁺ channels, we wished to see whether Mg²⁺ in the millimolar range would compete with or substitute for activator Ca²⁺ in the micromolar range. Instead, we found a remarkable effect, namely: that while neither competing with nor substituting for Ca²⁺, Mg²⁺ behaves like an allosteric effector by enhancing the cooperativity of Ca²⁺ activation; in addition, Mg²⁺ increases the apparent affinity of the channel for Ca²⁺.

In all experiments reported here, Ca²⁺-activated K⁺ channels from rat skeletal muscle were studied by fusing transverse tubule membrane vesicles, purified from hind leg muscle, with planar phospholipid membranes, using procedures introduced by Latorre, Vergara & Hidalgo (1982). Details of the method can be found elsewhere (Moczydlowski & Latorre, 1983; Miller, Moczydlowski, Latorre & Phillips, 1985). Briefly, bilayers were formed from neutral phospholipids (70% phosphatidyl-ethanolamine/30% phosphatidylcholine) on a hole in a plastic septum separating two aqueous chambers, as pictured in Fig. 1. The ‘internal’ aqueous solution, which is equivalent to the cytoplasmic side of the channel, was composed of 150mmol l⁻¹ KCl, 10 mmol l⁻¹ Mops-KOH, pH 7·2; neither Ca²⁺ nor Ca²⁺ buffers were added to this solution, and the free Ca²⁺ concentration (estimated from atomic absorption spectrophotometry) was about 3μmol l⁻¹. The external solution contained 150 mmol l⁻¹NaCl, 10 mmol l⁻¹ Mops-KOH, 0·1 mmol l⁻¹ EGTA; in some experiments, the external solution contained KCl instead of NaCl. An applied voltage was held across the membrane, and the resulting ionic current was continuously monitored. After a single Ca²⁺-activated K⁺ channel appeared in the membrane, fluctuations were recorded on video tape at a filter setting of 1–2 kHz, at various holding voltages. The data were later analysed by computer (Moczydlowski & Latorre, 1983). All voltages are reported according to the usual electrophysiological convention, with the external solution at zero voltage.

Fig. 2 illustrates typical records taken under different conditions of [Ca²⁺] and [Mg²⁺] in the internal solution. At 3 μmol l⁻¹ Ca²⁺, the channel is closed most of the time under these conditions; addition of 10 mmol l⁻¹ Mg²⁺ to the internal solution greatly increases the opening probability. Similar addition of Mg²⁺ to the external solution has no such effect. In the presence of 0·l mmol l⁻¹ EGTA, no channel opening is observed, either in the presence or absence of 10 mmol l⁻¹ Mg²⁺. Thus, while Mg²⁺ cannot by itself activate the channel, it appears to enhance the activation by Ca²⁺. This effect of Mg²⁺ is not an artifact due to Ca²⁺ contamination of our Mg²⁺ solutions. Atomic absorption analysis shows that at most 0·5 μmol l⁻¹ Ca²⁺ is introduced by addition of 10 mmol l⁻¹ Mg²⁺; this is sixfold less than the lowest Ca²⁺ concentration used here and would give rise to an insignificant error.

How does Mg²⁺ increase activation of the channel? In Fig. 3, we display Ca²⁺ activation curves of the channel opening probability, with and without Mg²⁺ present. It is clear that Mg²⁺ acts by shifting this curve to lower Ca²⁺ concentrations (i.e. it increases the apparent affinity of Ca²⁺ for the channel). The curve of Fig. 3 is typical for the effect of Mg²⁺ under various conditions: the concentration of Ca²⁺ giving 50% activation is lowered by a factor of 2–3 by 10 mmol l⁻¹ Mg²⁺.

The effects of Mg²⁺ described here are observed under a variety of conditions. We varied the membrane lipid composition by omitting the phosphatidylcholine from the membrane-forming solution, by altering pH from 6·8 to 7·5, by including 5 mmol l⁻¹ dithiothreitol in the aqueous solutions, and by using plasma membrane vesicles from different parts of the preparative sucrose gradient. None of these variations produced any discernible effects on the Hill coefficient of activation.

Since this channel’s opening probability is affected by voltage as well as [Ca²⁺], we examined the effect of voltage on the Hill coefficients in the presence and absence of Mg²⁺. Over the voltage range 20 to 60mV, we have not observed a reproducible variation of Hill coefficient in the absence of Mg²⁺, and only a slight trend for the parameter to decrease at higher potentials in the presence of Mg²⁺ (Fig. 4B).

We have chosen 10 mmol l⁻¹, the Mg²⁺ concentration used in most of these experiments, as a matter of convenience. The Mg²⁺ effect can be observed at 2 mmol l⁻¹, but it is much smaller, and there is no indication of the effect saturating at 10 mmol l⁻¹ (Fig. 4A). We avoided increasing Mg²⁺ above 10 mmol l⁻¹, however, out of worry about possible nonspecific effects of high divalent cation concentrations on this channel.

This report is intended as a qualitative description of an effect of Mg²⁺ on the high-conductance Ca²⁺-activated K⁺ channel from muscle. We find that Mg²⁺, while unable to open the channel by itself, greatly enhances the effectiveness of Ca²⁺ in opening the channel. The action of Mg²⁺ clearly occurs at sites other than those to which Ca²⁺ binds to activate the channel. Not only does Mg²⁺ fail to compete with Ca²⁺, but it actually potentiates Ca²⁺ activation. In addition, Mg²⁺ controls the apparent molecularity, or ‘cooperativity’, of Ca²⁺ activation. Neither of these effects would be expected if Mg²⁺ were acting at the Ca²⁺-binding sites involved in channel activation. We cannot offer a quantitative explanation for the effects of Mg²⁺, other than to say that we have established the existence of a modulatory Mg²⁺-binding site accessible from the cytoplasmic side of this membrane protein. We consider that there are several possibly important ramifications, both physiological and mechanistic, of the effect of Mg²⁺ on this channel.

First, the fact that Mg²⁺ modulates the activity of the channel may have physiological consequences worth noting. We now know that with Mg²⁺ present, this channel is a far better ‘switch’ than was previously thought; a Hill coefficient of 5 implies that a twofold increase in channel opening probability can result from only a 15% increase in Ca²⁺ concentration. These results also raise the possibility that the activity of the channel can be modulated in vivo by changes in free cytoplasmic [Mg²⁺], in addition to changes in [Ca²⁺]. This is merely speculation, since very little is known about free Mg²⁺ levels in muscle.

A fundamental question about the physiological relevance of these results immediately asks itself: are the concentrations of Mg²⁺ needed to give the effect, 5–10 mmol l⁻¹, too high to be meaningful in the cell? Indeed, this may be so, but there is a compelling reason to think that the Mg²⁺ effect may operate in cells at concentrations lower than those used in this study. Moczydlowski, Alvarez, Vergara & Latorre (1985) have shown that this particular channel senses the lipid surface charge in its immediate environment. In membranes containing negatively charged lipids, the channel is up to 10-fold more sensitive to Ca²⁺ than in neutral membranes, such as those used here. These authors argued persuasively that the channel’s Ca²⁺ activation site ‘feels’ the electrostatic surface potential set up by fixed negative charges on the membrane nearby. In general, living cells carry fixed negative charge on their membranes because of acidic lipids and proteins. Therefore, it is entirely reasonable to suppose that this channel is less sensitive to Ca²⁺ in neutral lipid bilayers than in living cells largely because of electrostatic surface potential effects. If such a situation applies to Ca²⁺, it may also apply to Mg²⁺. Thus, we suggest that the effects seen here at 10 mmol l⁻¹ Mg²⁺ may be expected to be seen in excised patches of cell membranes in the physiological range of 1 mmol l⁻¹ Mg²⁺. This is an easily testable suggestion via cellular patch recording. Regardless of the physiological relevance of these studies, a strong mechanistic conclusion can be drawn: that this channel contains more Ca²⁺ activation sites than has previously been believed. Since we have observed Hill coefficients as high as 5·8 in the presence of Mg²⁺, we suggest that there are at least six Ca²⁺-binding sites involved in activation of this channel. We feel that it is more physically palatable to propose that Mg²⁺reveals Ca²⁺-binding sites already present in the absence of Mg²⁺ rather than creating additional ones on the protein. Such an idea is easily accounted for qualitatively in terms of a general six-site model (Fig. 5). We can imagine a channel which binds up to six Ca²⁺ ions in both open (O) and closed (C) states (subscripts in Fig. 5 indicate number of Ca²⁺ ions bound). It is easy to adjust the equilibrium constants for each Ca²⁺-binding step to give Hill coefficients as low as 1 and as high as 6. The effect of Mg²⁺ could simply be to change several of these binding constants to shift the overall activation curve to more sigmoid behaviour.

A full kinetic analysis, such as those performed by Magleby & Pallotta (1983) and by McManus & Magleby (1985), would be needed to test such a model, and for this we have neither the capability nor the stomachs. The small amount of kinetic analysis we have performed (not shown) demonstrates that the effect of Mg²⁺ is seen on both open and closed channel lifetimes, and rules out none but the most simple and extreme models. We can say, however, that previous models proposing two, three or four Ca²⁺ activation sites are inadequate. Our results are consistent with McManus & Magleby’s (1985) kinetic studies suggesting that at least six closed states exist for this channel.

In the light of the complexity of the models needed to account for Ca²⁺ activation of this channel, it is tempting to take one of two conceptual steps: to give up trying to understand the gating of this channel through single-channel kinetic behaviour, or to try to simplify the situation by postulating that this channel is composed of six similar subunits, each of which binds Ca²⁺. We intend to take both paths, and consequently are eschewing further gating studies while attempting to purify the channel protein to test the subunit hypothesis directly at the biochemical level.

We are grateful to Drs Ramon Latorre and Gary Yellen for helpful suggestions in theory and practice involved in this work. JG and AK were supported by Dretzin and Brittner Graduate Fellowships of Brandeis University. This work was supported by NIH grant No. GM31768.