[1] 4w Pfl/igers Arch (1988) 412:297--304

[1] 116w ATP-sensitive channels were observed in isolated inside-out membrane patches from rat cultured central neurones. Two types of ATP-sensitive K + channels were present in cortical neurones, one which had its open-state probability increased, the other its open-state probability decreased by application of ATP to the cytoplasmic membrane surface. Another, ATP-sensitive channel differing in ion conductance from all previously reported ATP-sensitive channels was also seen in patches from cortical neurones. This channel was nonselective with respect to Na +, K + and C1-ions and ATP produced a "flickery" type of block. The non-hydrolysable analogue, AMPPNP, did not mimic ATP and prevented ATP action. Preliminary experiments indicate that similar, but not, identical ATP-sensitive channels exist in cerebellar neurones.

[1] 250w The advent of the patch clamp technique for recording single channel currents has led to the identification and characterization of novel types of ionic channels in excitable membranes. Among these recently-discovered channels are a group of cation-selective channels that are sensitive to the concentration of certain nucleotides within the cell. The best known example is a potassium-selective channel the activity of which is inhibited by intracellular ATP (ATP-K + channel). This channel has been shown to exist in the plasma membrane of mammalian cardiac (Noma 1983;Trube and Hescheler 1984;Kakei et al. 1985) and amphibian skeletal (Spruce et al. 1985) muscle cells and in rodent primary cultured pancreatic/C-cells (Cook and Hales 1984;Ashcroft et al. 1984;Findlay et al. 1985;Misler et al. 1986), and a variety of insulin secreting cell lines (Sturgess et al. 1986a;Dunne et al. 1986;Light et al. 1987). In addition, it has recently been reported that a calcium-activated non-selective cation (Ca-NS +) channel present in the plasma membrane of an insulin-secreting cell line (Ashford et al. 1986a) is also sensitive to inhibition by intracellularly applied ATP (Ashford et al. 1986b;Sturgess et al. 1987a). However, unlike the ATP-K + channel, the Ca-NS + channel is more sensitive to the nucleotides ADP and AMP (Sturgess et al. 1986b). The ATP-K + channel has recently been shown to be present in the plasma membrane of human cultured/%cells (Ashcroft et al. 1987) and both the ATP-K + and the nucleotide sensitive Ca-NS + channel in human cultured insulin-secreting tumour cells (Sturgess et al. 1987b).

[2] 146w The presence of these nucleotide-sensitive ion channels in a variety of excitable cells provides the intriguing possibility of a close link between the metabolism of a cell and its excitability. For example, where the channel is active under resting conditions as in the case of/~-cells, it is argued (Cook and Hales 1984) that the ATP-K + channel activity is inhibited on metabolism of substrates such as glucose which increase cellular ATP levels (Ashcroft et al. 1973) resulting in membrane depolarization and hence insulin secretion. In contrast the ATP-K + channel in cardiac muscle cells is relatively inactive at normal physiological ATP levels but a hypoxic stimulus to the cell results in its activation as ATP levels decrease and so produces cell hyperpolarization and shortening of action potential duration (Noma and Shibasaki 1985). Such action would make the cell more refractory to stimuli and reduce contractile activity.

[3] 108w Because nucleotide-sensitive channels have been found in peripheral excitable tissue we considered it likely that similar channels exist in the plasma membrane of central neurones. Brain cells are known to be extremely sensitive to anoxia or a reduced supply of glucose and under either of these conditions a number of ionic movements occur which may be correlated with ATP depletion (Hansen 1985). Thus the present study, using the patch clamp technique, was undertaken in order to determined whether the plasma membranes of rat cultured central neurones contain nucleotidesensitive ion channels. A preliminary account of some of these data has been published in abstract form (Ashford et al. 1987).

[1] 271w In experiments where the external surface of inside-out patches excised from cortical neurones was exposed to a solution containing 5 mM K + (ie normal saline in the pipette) and the internal surface to one containing 140 mM K + and 1 ~tM free Ca 2+ (ie mimicking the cellular potassium gradient), various types of potassium selective channels were detected. Under these ionic conditions (ie no internal sodium and low calcium) three distinct types of non-voltage-activated K + selective channels could commonly by distinguishable in terms of conductance and sensitivity to ATP. On most occasions we observed only one type of K + channel in any single patch, although often more than one channel was active. Thus because of the heterogeneity of the cortical cultures one cannot discount the possibility that individual cells do not have all three channel types. Of the three K + channels, one which had a conductance of 30 pS did not exhibit any sensitivity to ATP (2-5 mM tested) and we did not investigate this channel further. The activity of both the others was affected by ATP applied to the intracellular surface of the membrane. The larger conductance channel (100 pS, n = 3), which had a reversal potential of approximately -70 mV under this physiological K + gradient, was found to have its open state probability increased by ATP at the cytoplasmic surface (data not shown). For example, in one experiment 5 mM ATP increased the open state probability (PoPEN) from a control value of PoPEN = 0.20 to PoPEN = 0.47 at a membrane potential of 0 inV. As the purpose of this investi-

[2] 28w A 3O 10 o jnBmlmmlL -,o i.t ..... -30 4pA I 2s B ~o~ r---, -80 -,~0 5 4 3 O 2 1 -1 -2 O ~o Em(mV)

[3] 468w Fig. 1. A Single channel current records obtained fron an inside-out membrane patch (apparently containing one channel) illustrating the activity of a potassium selective channel which has a conductance of 52.0 pS when measured between -20 and +20 mV. The recording pipette contained 135 mM NaC1 and 5 mM KC1 and the bath 140raM KC1 and 10-6M Ca z+. Openings are shown as upward deflections (outward current) and the numbers to the side refer to the patch membrane potential in inV. B Current-voltage relationship for the channel shown in A under the same ionic gradients. Note the pronounced outward rectification of this channel at depolarized potentials gation was to determine whether there existed a K +-selective channel, the activity of which was inhibited by ATP, we have at present not characterized this channel further. Single channel current records illustrating the activity of the third K + selective channel observed in patches from these cells are shown in Fig. 1 A. It can be seen that the channel activity is characterized by prolonged bursts of opening interspersed with numerous short closed periods and that there appears to be a subconductance state (denoted by arrow in Fig. 1 A). On plotting the single channel current amplitudes versus voltage (Fig. 1 B) it is apparent that this channel exhibits pronounced outward rectification (much as predicted by the Goldman-Hodgkin-Katz theory) at depolarized potentials, and that the reversal potential lies be-tween --60 mV and -70 mV under this asymmetric potassium ion condition. The slope conductance for this channel was 52.7___ 8.8 pS (n = 4), when measured between -20 mV and + 20 inV. Application of 1 mM ATP to the bathing solution (ie in contact with the intracellular surface) did not completely abolish channel activity, in contrast to the ATP-K § channels present in many other excitable tissues. In addition, ATP at concentrations of I -2 mM induced long closed periods between the bursts of openings (Fig. 2A). In one patch in which three identical K + channels were active, increasing concentrations of ATP (2-10 mM) caused a reversible dose-dependent reduction in channel open state probability, although even at 10 mM ATP this channel was not completely inhibited (Fig. 2 B). The non-hydrolysable analogue of ATP, AMPPNP was also found to inhibit channel activity reversibly (n = 3), in this case (Fig. 2 C) 3 mM being sufficient to almost abolish the single channel openings. Thus the ATP-K + channel in these cells is sensitive to millimolar levels of ATP applied to the cytoplasmic surface [2 mM ATP producing a 50-70% inhibition (n = 3) of POPEN]. Further investigation of this channel has been hampered because of its relative sparseness, being observed in only eight patches out of a total of 56 patches excised from cortical neurones which were used for analysis.

[4] 213w Although ATP-K + channels appeared to be few in number in patches excised from these cells, another channel, the activity of which was also inhibited by ATP, was much more common. This channel was observed in 26 out of the 56 patches used for analysis. Typical records illustrating the single channel currents are shown in Fig. 3A. The major points to note are, that under asymmetric ion conditions across the patch (irrespective of which side the high potassium or high sodium containing solutions are), the single channel current reversal potential is 0 mV and that the currents are larger at depolarized potentials than hyperpolarized potentials. This can be clearly seen from the current-voltage relationship illustrated in Fig. 3B (filled symbols) indicating that the conductance is continuously changing with voltage. Regardless of whether sodium or potassium was the major cation at either side of the membrane the mean single channel conductance for inward currents was 18.6 _+ 1.1 pS (n = 12) and for outward currents was 57.7 + 2.0 pS (n = 12) when measured between 0 mV and + 50 inV. These results indicate that the channel is either anion selective or non-selective for cations. In addition, a subconductance state could also be observed for this channel (arrowed in Fig. 3 A).

[5] 665w In order to differentiate between these possibilities the bathing solution was replaed with one containing predominantly 40 mM KC1, producing both a cation and an anion gradient. If the channel was cation selective the expected reversal potential would be + 32 mV and if anion selective a value of about -31 mV should be obtained. The single channel current data from this experiment are shown in Fig. 3 B and the resultant I-V relation in Fig. 3 C (open triangles). The data clearly show that surprisingly neither of the expected values occurred and the reversal potential was, in fact, unchanged by this procedure but that the currents were reduced in size at all applied potentials. This effect was observed in a total of four experiments and indicates that the channel does not select between chloride ions or the two A small K + selective channel (not sensitive to ATP) was often active in patches and resulted in an apparently noisy background closed state (see C in particular) major cations, Na + and K + and thus is best described as a non-selective channel. Additionally, in one experiment, the solution bathing the intracellular face of the membrane was changed for one containing 140 mM KC1 and 70 mM NaC1, and again the reversal potential remained at 0 mV and did not shift to either +_ 10 mV as predicted for a charge-selective channel. In three experiments changing the bathing solution calcium concentration to 1 mM, from 10 -6 M, appeared to increase channel activity in these patches. This effect has not yet been investigated in detail. Application of ATP (1 -3 mM) containing solutions to the cytoplasmic side of the isolated membrane patches resulted in a marked change in the activity of the non-selective channel. Figure 4A illustrates clearly the effects of 2 mM ATP on a patch at a membrane potential of + 50 mV. Unlike the ATP-K + channels, where addition of ATP reduces the open state probability by allowing fewer openings, the action of ATP on this channel is to break-up individual bursts of openings into a series of fast open and closed transitions reminiscent of the "flickering" ~type of block produced by drugs or ions that act by an open channel block mechanism (Hille 1985). This effect on the non-selective channel was observed on every occasion (n = 14) patches were exposed to ATP (1 -3 mM) and was reversible by washing away the ATP (Fig. 4A). Analysis of the channel kinetics on patches that contained only a single non-selective channel (n = 3) indicate that ATP acts to shorten the lifetime of the open state and increase the lifetime of the intraburst closed state (Table 1). The closed times between bursts also appear to be prolonged by ATP, but this has not been investigated rigourously. The overall effect of these actions on the channel kinetics is to produce a dose-dependent decrease in the open-state probability of the channel (Table 1). It was also noticeable that on removal of the ATP the channels appeared more active, for example patches in which 1-2 channels were operating prior to ATP addition often showed three active channels after ATP washout. This effect is illustrated in Fig. 4A for a single non-selective channel. The open state probability has been decreased 42% by 2 mM ATP compared to control but on wash has increased by 48% compared to the pre-ATP treated value, hence there appears to have been an activation of the single channel activity. These actions of ATP were observed at both positive and negative membrane potentials and with different species of cation at C Current-voltage relationships for this channel under the ionic conditions described in A ( 9 and B (A) respectively. Note that the single channel currents reverse, under both ionic conditions, at 0 mV and that there is pronounced outward rectification of the channel conductance at depolarized potentials. All data in A-C are from a single patch, which contained only one non-selective channel

[6] 490w ,.A[ ls Fig. 4. A Inhibition of the non-selective channel current by 2 mM ATP added to the bath solution. Note that the effect of ATP is reversible on wash. The patch membrane potential was + 50 mV and openings are denoted by upward deflections. The bath solution contained 135 mM NaC1, 5 mM KCI, 10-6 M Ca 2 § and the pipette solution, 140 mM KC1. The PoezN values were: control 0.62, 2 mM ATP 0.38, wash 0.92. B Illustrates from a separate membrane patch, the lack of effect that 2 mM AMPPNP has on the non-selective channel activity, and that after washing the AMPPNP out, 2 mM ATP is now also ineffective at producing block (compare bottom trace in B to middle trace in A). The patch membrane potential was +40 rnV and the pipette solution contained 140 mM KC1 and the bath solution 135 mM NaC1, 5 mM KC1 and 10 6 M Ca 2+. POPEN values obtained from this patch are as follows: control 0.76; 2 mM AMPPNP 0.82; wash (not shown) 0.88; 2 mM ATP 0.83 Table 1. Effects of ATP on channel kinetics of the non-selective channel in cortical neurones Patch [ATP] ZOP~N ZCLOSED POPEN (mM) (ms) (ms) 0 1.63 +__ 0.02 0.49 +__ 0.01 0.482 1.0 1.08 • 0.02 0.92 • 0.03 0.365 0 5.84 __+ 0.31 0.56 • 0.01 0.621 2.0 1.67 • 0.02 1.28 • 0.02 0.361 0 10.9 __+ 0.19 0.54 • 0.01 0.737 1.0 1.72 • 0.02 0.79 • 0.02 0.635 2.0 1.02 • 0.04 1.47 • 0.01 0.197 3.0 1.16 • 0.05 2.63 • 0.08 0.133 All time constant values in the table are • SEM from the best-fit exponentials to between 1500-2500 event transitions presented as mean values histograms obtained from . POPEN values obtained from this patch are as follows: control 0.73; 2 mM ATP 0.00; wash 0.69. B Effects of 2 mM ATP and 2 mM AMPPNP on a single non-selective channel from a separate membrane patch. The pipette solution contained 140 mM KC1 and the bath solution contained 135 mM NaC1, 5 mM KCI and 10 -6 M Ca 2+. Under these ionic conditions the conductance of the channel was 50 pS (n = 2) at negative membrane potentials. The patch membrane potential was -40 mV and single channel currents are shown as downward deflections. Note that the ATP and AMPPNP effects are indistinguishable and are reversible on wash (the AMPPNP wash is not shown), and that these nucleotides markedly reduce the amplitude of the currents by inducing a fast flickering type of block either membrane surface. The effects of ATP could not be mimicked by application of 2 mM AMPPNP, a nonhydrolysable analogue (Fig. 4 B) and interestingly, following the washout of the AMPPNP, on reapplication of ATP, it was found that the ATP (1 -2 raM) was now totally ineffective (n = 4) at blocking the single channel currents (Fig. 4 B).

[7] 188w Some preliminary experiments were performed on cerebellar neurones [these were probably granule cells (Moonen et al. 1982)] in order to give some indication of whether the ATPsensitive channels were confined only to cortical cells or were a more widespread phenomenon in central neurones. Figure 5A, B illustrates the effects of ATP on two different types of single channel recorded from patches isolated from these cells. A potassium selective channel was observed, the activity of which was completely inhibited by 2 mM ATP applied to the cytoplasmic face of the membrane (n = 2) and this action was completely reversible (Fig. 5 A). A second or subconductance state was also detected for this channel, as denoted by the arrow in the figure. Under asymmetric cation gradients and at potentials around 0 mV, the conductance of the main state was 60 pS and the subconductance state 26 pS. The current-voltage relationship for this channel (main conductance state) under a physiological cation gradient was virtually indistinguishable from the ATP-K § channel in cortex cells (data not shown), exhibiting outward rectification at depolarized potentials (which was also wellfitted by the Goldman-Hodgkin-Katz theory).

[8] 120w The second type of ATP-sensitive channel resembles the non-selective channel in cortical cells in terms of its ATP sensitivity, the openings are "chopped-up" into bursts (Fig. 5 B) and under an asymmetric ion distribution the reversal potential was 0 mV (n = 2). The noticeable differences are that this channel appears to rectify in the opposite direction (not shown) to the cortical neurone non-selective channel (ie inward currents are larger than outward currents), AMPPNP is equally capable of reducing channel activity and that the ATP and AMPPNP induce a much more intense flickering of the channel between open and closed states such that the full amplitude of the conducting state is not clearly resolved at the recording frequency (1 kHz).

[1] 330w Potassium selective channels, which can be inhibited by the intracellular application of ATP have been shown to be present in a variety of cell types (Stanfield 1987). The data presented in this study suggest that a similar channel is also present in the plasma membrane of central ncurones. This ATP-K § channel does not require the hydrolysis of ATP, as indicated by the inhibition produced by AMPPNP, in common with the ATP-K § in other tissues. However, it is apparent that there are differences between this ATP-K § channel and those in other cells. One important distinguishing feature is in the density of the channel in the plasma membrane as only in very few patches obtained from the cortical cells could this channel be identified. This contrasts with the much higher density of ATP-K + channels found in isolated patches obtained from pancreatic t-cells (Cook and Hales 1984;Rorsman and Trube 1985;Findlay et al. 1985;Misler et al. 1986), insulin-secreting cell lines (Sturgess et al. 1986a;Dunne et al. 1986;Light et al. 1987) and muscle cells (Kakei et al. 1985;Spruce et al. 1985). In addition, the conductance of the ATP-K § channel in cortical cells is larger (53 pS, under an approximately physiological gradient) in comparison to these other cell types (generally 14-35 pS under a similar gradient). The current-voltage relationship also exhibits outward rectification for the cortical ATP-K § channel as opposed to slight inward rectification at depolarized potentials in other tissues (ie in these cells the rectification is simply a consequence of the potassium concentration gradient). Lastly, the sensitivity of the channel to ATP in the cortical cells [the half maximal inhibitory concentration (Ki) in the region of 2 mM] is rather less than the ATP-K + channel in other cells, the Ki ranging from 12-15 txM in fi-cells (Cook and Hales 1984;Sturgess et al. 1986 a;Misler et al. 1986) and 135 laM in frog skeletal muscle (Spruce et al. 1987) to 0.5 mM in heart muscle (Kakei et al. 1985).

[2] 101w A second type of ATP-sensitive potassium channel was observed in excised membrane patches from cerebral cortical cells, one which had its open-state probability increased by the presence of ATP at the intracellular membrane surface. Similar effects of intracellularly applied ATP to K + selective channels in inside-out patches from a transformed insulinsecreting cell line (Light et al. 1987) and from guinea-pig ventricular cells have been reported (Kakei et al. 1985). In addition Benzanilla et al. (1986) have reported that intracellular application of ATP to the internally dialyzed squid giant axon increased the magnitude and altered the kinetics of the delayed rectifier.

[3] 354w The more frequently observed ATP-sensitive channel in inside-out patches obtained from the cortical cells appeared not to select for the normal physiological cations, Na + and K +, nor for C1-ions (its permeability to calcium has not yet been tested). Recently Chesnoy-Marchais and Evans (1986), using outside-out membrane patches obtained from Aplysia neurones, reported the existence of a non-selective ion channel. This channel was found to be permeable to a wide range of cations and anions and to have a conductance of 100 pS in 600 mM symmetrical chloride. Thus, it may be that such channels exist in a number of neuronal preparations, but would be extremely difficult to detect without single channel recordings. Inhibition of this channel by ATP is different from that of ATP-K + channels, in that a "flickcry" type of block of the channel open state is produced. This has been termed "intermediate" block by Hille (1985) as opposed to the effect of ATP on K + channels which has properties similar to a "slow" form of block. The ATP is unlikely to be breaking up the openings by the mechanism of open-channel block (Neher and Steinbach 1978;Hille 1985) because the ATP induces a reduction in the overall open-state probability of the channel, an effect not in keeping with this form of inhibition. Further evidence to suggest that there must be a distinct binding site for ATP, other than the open channel, is the observation that once AMPPNP had been applied and the patch washed thoroughly, ATP was no longer effective in inducing the "flickery" block. Thus the AMPPNP appears to remain bound in an irreversible or slowly reversible manner to the nucleotide binding site associated with this non-selective channel. As for the ATP-K + channel in pancreatic/~-cells (Findlay and Dunne 1986;Ohno-Shosaku et al. 1987) ATP has a dual effect on the nonselective channel in cortical neurones. Not only is there an inhibitory action on the open-state probability in the presence of the ATP, but also an increase in the open state after its removal. However, this effect was not observed if AMPPNP was applied to the intracellular membrane face.

[4] 71w Preliminary data from cerebellar neurones, probably granule cells indicate that similar, but probably not identical ATP-sensitive ion channels are present. For example, the ATP-K + channel appears to be more sensitive to ATP than its counterpart in the cortex and the channel resembling the non-selective type, has a much faster type of block (Hille 1985) induced by both ATP and AMPPNP indicating that hydrolysis is probably not a requirement for inhibition.

[5] 197w Thus it would appear that the plasma membranes of central neurones do indeed contain ATP-sensitive ion channels, although there are distinct differences from their counterparts found in peripheral tissues. Supporting evidence exists which indicates that these channels may be important in providing a link between cellular metabolism and excitability under conditions such as hypoglycaemia, ischemia, hypoxia and anoxia (Hansen 1985). It is well documented that under the above conditions there occur certain ionic movements in the brain which are believed to be associated with insufficient ATP generation. Two major phases of ionic movement can be distinguished in the cortex under these conditions. Firstly (phase 1) there is a loss of K + ions from neurones with a concomitant membrane hyperpolarisation and this is correlated with an 80-90% reduction in cellular ATP levels [from a resting level of approximately 3 mM (Hansen 1985)]. Following this in time there occurs, phase 2, a sudden, non-selective increase in the ionic permeability of the cell membrane (and hence depolarization) following a further fall in ATP levels. Indeed, conditions that preserve high levels of ATP within the cells delay the appearance of these changes and vice versa (for details see Hansen 1985).

[6] 155w It is therefore tempting to speculate that the changes described above in membrane ionic permeability of central neurones during reduction in ATP levels within the cell, produce initially an increase in K + permeability (phase 1) due to ATP-K + channel activation and latterly an overall increase in the permeability to all small ions (phase 2) due to ATP-non-selective channel activation. The differential sensitivity of these channels to ATP in the cortical cells as shown in this report does appear to fit this hypothesis. It is possible that the ATP-K + channel plays a physiological role in central neurones similar to that proposed by Noma (1983) for cardiac muscle cells. That is, if the cell is metabolically comprised and ATP levels begin to fall, this channel may activate and hyperpolarize the cell membrane thus making it refractory to further stimuli and so conserve ATP. However, the ATP-sensitive non-selective channel may only be of pathological significance.

[1] 149w Cell culture. Primary neuronal cultures were obtained from 2-7 day old neonatal rat cerebral or cerebellar cortices. The neuronal preparations were first chopped and placed in a enzyme solution which consisted of 5 mg dispace (Boehringer, Mannheim, FRG) and I mg DNase (Sigma, Poole, UK) dissolved in 5 ml of calcium and magnesium free phosphate buffered saline, and incubated at 37~ for 1 h. After 30 rain and i h the cells were agitated using a I ml Gilson pipette with sterile tip. The cells were then centrifuged at 800 rpm for 5 min and were resuspended in plating medium consisting of Dulbecco's modified Eagle's medium (Gibco, Paisley, Scotland), 5% foetal calf serum and 50 units/ml of penicillin and 50 gg/ml streptomycin. Cells were cultured in 35 mM Petri dishes at 37~ in an atmosphere of 5% CO2 -95% air and used for experiments after 1-10 days in culture.

[2] 76w The neurones were usually found associated with an underlying layer of glial cells (non-process bearing or epithelial-like) on the bottom of the culture dish. Cortical and cerebellar neurones were identified on the basis of appearance (phase-bright and process-bearing), size (< 10-12 gM in diameter) and also by immunocytochemical procedures being negative to staining with antisera to Glial fibrillary acidic protein (GFAP) but positive to RT 97, an antibody specific for neuronal filaments (Wood and Anderton 1981).

[3] 341w Solutions. The composition of the initial external saline used to wash the cells before single channel recording was (mM): NaC1 (135.0), KC1 (5.0), MgC12 (1.0), CaC12 (1.0) and HEPES (10.0), pH 7.2 with NaOH. The recording electrode was filled with either the same solution or one containing (raM): KC1 (140.0), MgC12 (1.0), CaC12 (1.0) and HEPES (10.0), pH 7.2 with KOH. The bathing solution during experiments consisted of (mM): KC1 (140.0) or NaC1 (135.0) and KC1 (5.0), MgC12 (1.0), CaClz (0.9), EGTA (1.0) and HEPES (10.0), pH 7.2 with either NaOH or KOH respectively, giving a final free calcium concentration of 1 gM. This protocol was used in order to reduce the number of different channel types observed and to facilitate the formation and maintenance of the excised inside-out membrane patches. In some experiments designed to determine ionic selectivity, the bath solution was changed to one containing (mM): KC1 (40.0), MgC12 (1.0), CaClz (0.9), EGTA (1.0) and HEPES (10.0) or: KC1 (140.0), NaC1 (70.0); MgC12 (1.0), CaC12 (0.9), EGTA (1.0) and HEPES (10.0). No corrections were made for liquid junction potentials (calculated to be < 1 mV). In experiments where high concentrations of ATP were used, the concentrations of CaC12, MgC1/and EGTA were altered in order to maintain the free concentrations of Ca 2 + and Mg 2 + ions at 1 gM and 1 mM respectively. Therefore for concentrations of 10, 2.0 and 1.0 mM ATP, the corresponding concentrations of EGTA were 10, 1.0 and 1.0 mM; CaC1/, 9, 0.9 and 0.9 mM and MgC12, 10, 2.6 and 1.8 mM. These values were calculated using the program described by Hesketh et al. (1983). ATP (Na § and K § salts, vanadium free) and AMPPNP were obtained from Sigma and Boehringer Mannheim respectively. Solutions were changed and ATP applied by superfusing the bath using continuous gravity-feed and expiration by suction at a rate of 0.1-0.2 ml s-1. All data in the text and figures are presented as mean values + SEM. All experiments were performed at room temperature, 22-25 ~ C.

[4] 148w Recording and analysis. Isolated patches, in the inside-out configuration (Hamill et al. 1981) were used in all experiments. Recording pipettes were pulled from borosilicate glass capillaries (Clark Electromedical Instruments, Pangbourne, Berks, UK), fire polished and filled with the appropriate electrolyte solution giving resistances of 8-12 MO. Single channel currents were recorded with an EPC-7 (List Electronic) or a Dagan 8900 patch clamp amplifier and stored on magnetic tape (Racal Store 4DS) for later analysis. Data were replayed into a chart recorder (Gould 2200) which filtered the records at 0.14 kHz and these traces were used for illustrative purposes and to measure single channel current amplitudes. Outward current (indicated by upward deflections in traces) refers to the flow of cations from the intracellular to the extracellular side of the membrane and the potential across the membrane patch is reported following the usual convention for membrane potential (ie inside negative).

[5] 101w The open state probability, current amplitude, and open and closed time distributions of single channel records were determined off-line using an analysis program incorporating a 50% threshold crossing parameter to detect events (provided by J. Dempster, University of Strathclyde) and run on a minicomputer (PDP11/23) or an IBM AT microcomputer. Stretches of data between 30 and 60 s in duration were replayed at the recorded speed, filtered at a bandwidth of 1 kHz (-3dB; 8 pole Bessel) and digitally sampled at 200 gS intervals using a Cambridge Electronic Design 502 interface (PDP 11/23) or a Data Translation 2801 board (IBM AT).

[6] 90w The value of open state probability (PoeEN) was obtained by measuring the total time a single channel spent in the open state and expressing this as a proportion of the total time of the data segment analysed taking into consideration the initial number of channels observed. In patches where there was only one channel observed (no "double events" for the non-selective channel at high values of POPEN), the open and closed lifetime distributions were measured and fitted with a single exponential probability density function using a non-linear least squares method.