Regulation of Cloned Atp–Sensitive K Channels by Phosphorylation, Mgadp, and Phosphatidylinositol Bisphosphate (Pip₂): A Study of Channel Rundown and Reactivation

ATP-sensitive K (K_ATP) channels are comprised of a pore-forming protein of the Kir6 family and the sulfonylurea receptor SUR. Patch-clamp recordings of both native and cloned K_ATP channels show decay of channel activity upon patch excision. This phenomenon, referred to as “rundown,” is avoided or even reversed by addition of MgATP, but not MgAMPPNP. It was initially postulated that channel dephosphorylation was responsible for rundown (Misler et al. 1986; Ohno-Shosaku et al. 1987; Ribalet et al. 1989). More recent studies, however, have shown that native and cloned K_ATP channels are reactivated by phosphatidylinositol bisphosphate (PIP₂) after rundown (Furukawa et al. 1996; Hilgemann and Ball 1996; Fan and Makielski 1997, Fan and Makielski 1999; Baukrowitz et al. 1998; Shyng and Nichols 1998). These findings, together with the observation that very high concentrations of wortmannin, an inhibitor of PIP₂ production, prevents K_ATP channel reactivation by MgATP (Xie et al. 1999) have led to the interpretation that reactivation by MgATP is due to PIP₂ resynthesis. In addition, elevation of internal Ca²⁺ greatly accelerates rundown and MgATP reverses this effect (Ohno-Shosaku et al. 1987; Furukawa et al. 1996), consistent with the idea that Ca²⁺-induced rundown involves activation of Ca²⁺-dependent phospholipase C that hydrolyzes PIP₂ (Hilgemann and Ball 1996).

However, it has also been noted that prolonged periods of rundown led to irreversible channel closure, so that limited or no reactivation occurred with reexposure to MgATP or PIP₂ (Furukawa et al. 1996). This indicates that rundown involves other factors besides PIP₂ hydrolysis. Protein kinase–mediated phosphorylation is one possible candidate, since protein kinase C inhibition has been shown to increase ATP sensitivity (Hu et al. 1996; Light et al. 1996) similar to rundown (Deutsch et al. 1994; Terzic et al. 1994; Baukrowitz et al. 1998). In addition, we observed that rundown occurred more rapidly in the presence of the sulfonylurea receptor, SUR1, than in its absence (John et al. 1998), suggesting that SUR1 regulates the rundown process. Since SUR1 confers MgADP sensitivity to cloned K_ATP channels (Tucker et al. 1997; John et al. 1998) and rundown is accompanied by a loss of sensitivity to nucleotide disphosphates (Findlay 1987; Bokvist et al. 1991; Terzic et al. 1994), binding of MgADP to SUR could play a role.

To assess the roles of PIP₂, MgADP and phosphorylation in controlling the activity and rundown of K_ATP channels, we performed excised inside-out patch-clamp experiments of the Kir6.2-GFP+SUR1 complex and the pore-forming protein Kir6.2-GFP alone. Specifically, we investigated how these factors regulate the ATP sensitivity of the two channel types and the coupling between SUR1 and Kir6.2. A model is proposed to account for Kir6.2+SUR1 and Kir6.2 channel regulation by MgADP, PIP₂, and protein kinases.

The techniques used in the present studies for cDNA expression and patch clamp recording have been recently described in detail (John et al. 1998) and are only briefly outlined here.

HEK293 cells were transfected with cDNA for Kir6.2 linked to green fluorescent protein (GFP) at the COOH terminus (Kir6.2-GFP). All wild-type cDNAs were subcloned into the vector pCDNA3amp (Invitrogen). cDNAs used to make the GFP chimeras were subcloned into the pEGFP vector. Both vectors used the cytomegalovirus promoter. The transfections were carried out using the calcium phosphate precipitation method (Graham and van der Eb 1973). Expression of proteins linked to GFP was detected as early as 12 h after transfection. Patch-clamp experiments were started ∼30 h after transfection. HEK293 cells were cultured in DMEM high glucose medium supplemented with 10% (vol/vol) fetal calf serum, penicillin (100 U/ml), streptomycin (100 U/ml), and 2 mM glutamine and divided once a week by treatment with trypsin.

Currents were recorded in HEK293 cells using the inside-out patch-clamp configuration, with the pipette solution containing (mM): 140 KCl, 10 NaCl, 1.1 MgCl₂, 10 HEPES, pH 7.2 with KOH. The bath solution consisted of (mM): 140 KCl, 10 NaCl, 1.1 MgCl₂, 10 HEPES, 5 EGTA, 0.5 CaCl₂, pH 7.2 with KOH. ATP was added directly to the bath as MgATP. PIP₂ (Boehringer) was sonicated immediately before use. The protein kinase inhibitor H-7 and the phosphatase inhibitor okadaic acid were obtained from Calbiochem. The specific protein kinase C inhibitor peptide 19-36 was purchased from Peninsula Laboratories.

The data, filtered at 2 kHz with an eight-pole Bessel filter, was recorded with a patch clamp amplifier (EPC 7; List) and recorded on videotape at a fixed frequency of 44 kHz after digitization with a digital audio processor. For analysis, the data was sampled at a rate of 5.5 kHz. When discrete current steps could be resolved, channel activity was expressed as NP_o. To express channel activity as a function of ATP concentration, NP_o was estimated at each ATP concentration from data samples of 15 s duration. Since channel activity varied widely from patch to patch, NP_o values were normalized to values measured in the absence of ATP. When single channels could not be resolved, steady state current values were used instead of NP_o to assess the effects of adenine nucleotides and other interventions on channel activity.

Analysis of single-channel data was carried out using the approach described in detail previously (Ribalet et al. 1989), which consists of fitting the open- and closed-time histograms with a sum of exponentials. Since in most cases more than one level of channel opening was observed, we could not construct closed-time histograms. However, open-time histograms could be constructed, since membrane patches with low activity at the second level of opening could be obtained.

Kir6.2-GFP+SUR1 channel activity increased rapidly when inside-out patches were excised into a MgATP-free bath solution with Ca²⁺ concentration <10⁻⁷ M. Their activity then typically ran down in a fast and a slow phase (Fig. 1 A). The amplitude of the fast phase was variable. For instance, in six patches, channel activity decreased by 50% within 105 ± 42 s and was inhibited by >80% from the initial level by the start of the slow phase. In contrast, in seven other patches, there was no significant phase of fast rundown. In the majority of patches (n = 38), however, the amplitude of the fast phase varied between these two extremes. After fast rundown, channel activity declined more slowly, leading to nearly complete inhibition after several hours.

The amplitude of the fast phase of rundown correlated with the channel ATP sensitivity. In six patches with a pronounced fast phase of rundown, the concentration of ATP causing half-maximal channel inhibition (IC₅₀) after fast rundown averaged 23 μM, whereas in four patches without a significant fast phase of rundown, the IC₅₀ averaged 210 μM. MgADP reactivated the ATP-inhibited channels regardless of whether the initial ATP sensitivity was low or high (Fig. 1 B), and sulfonylureas blocked channel activity (not shown), indicating functional coupling between Kir6.2 and SUR1. Thus, in two patches with pronounced fast rundown and with an averaged IC₅₀ for ATP-dependent inhibition of 15 μM, MgADP reactivated ATP-inhibited channels to a similar extent (67%) to that in the absence of prominent fast rundown.

We observed that MgADP was a powerful inhibitor of the fast phase of rundown. In 12 of 14 patches tested, the addition of 150 μM MgADP to the bath before patch excision prevented fast rundown (Fig. 2 A). In the two patches that showed fast rundown with MgADP present, MgADP did not reactivate ATP-inhibited channels and evoked channel inhibition per se. As discussed later, this occasional observation may be explained assuming that channel dephosphorylation and uncoupling of Kir6.2 and SUR1 occurred at the time of patch excision. In the majority of patches (86%) in which MgADP prevented channel rundown, MgADP also reactivated ATP-inhibited channels by decreasing their sensitivity to ATP. The degree of this well-known effect of MgADP (see, for example, Misler et al. 1986; Findlay 1987) varied from patch to patch, but fell in two categories. In the first category (Fig. 2 D, left bar), the averaged channel sensitivity (IC₅₀) to ATP was 151 ± 57 μM in the presence of MgADP and 29.5 ± 14 μM in its absence. In the second group (Fig. 2 D, right bar), the respective values were 380 ± 50 μM (Fig. 2 B) and 171 ± 47 μM (Fig. 2 C). Furthermore, upon MgADP removal, the patches in the first group with high sensitivity to ATP exhibited fast rundown (>30%), whereas patches in the second group with low ATP sensitivity exhibited almost no fast rundown (<10%). These two sets of values for ATP sensitivity in the presence and absence of MgADP were used to generate part of the model presented in the discussion. Finally, the ability of MgADP to prevent fast rundown was only observed if MgADP was present at the time of patch excision (Fig. 2 A). If MgADP was added after fast rundown had begun, it was no longer protective.

The slow phase of rundown was also characterized by an increase in ATP sensitivity, with the IC₅₀ of all patches averaging 14 ± 4 μM after 1–2 h of slow rundown. Unlike fast rundown, however, the increase in ATP sensitivity was accompanied by an almost complete loss of MgADP-dependent reactivation of ATP-inhibited channels (Fig. 1 C) and of inhibition by sulfonylureas (data not shown). This loss of MgADP-dependent reactivation (Findlay 1987; Bokvist et al. 1991; Terzic et al. 1994) and sulfonylurea-induced inhibition (Gillis et al. 1989) observed during rundown had already been reported for native K_ATP channels and suggests functional uncoupling of Kir6.2-GFP and SUR1, since SUR1 confers MgADP and sulfonylurea sensitivity to Kir6.2+SUR1 channels (Tucker et al. 1997; John et al. 1998).

In summary, we found that spontaneous rundown of Kir6.2-GFP+SUR1 involved two phases. The initial fast phase begun immediately after patch excision was associated with marked increase in ATP sensitivity, probably attributable to the decrease in channel activity (Koster et al. 1999), and could be prevented in most patches by excising the patch into MgADP-containing solution. Right after fast rundown, MgADP still reactivated ATP-inhibited channels. In contrast, the slower phase of rundown, which evoked further decrease in channel activity and increase in ATP sensitivity, led to loss of MgADP-dependent reactivation. The loss of MgADP-dependent reactivation was irreversible and coincided with loss of sulfonylurea sensitivity, suggesting functional uncoupling of Kir6.2-GFP and SUR1.

Addition of Ca²⁺ (10 μM) to the bath solution blocked K_ATP channel activity within seconds, and recovery was minimal after Ca²⁺ removal. For these reasons, Ca²⁺ application has been used as a model to study K_ATP channel rundown (e.g., Ohno-Shosaku et al. 1987; Furukawa et al. 1996). Accordingly, we investigated the effects of Ca²⁺ on channel ATP sensitivity and coupling of SUR1 to Kir6.2-GFP. Fig. 3 A shows that addition of Ca²⁺ to the bath solution caused almost complete channel inhibition within seconds, which was poorly reversible upon Ca²⁺ removal. MgATP blocked the residual current, and reactivation of the current was apparent when MgATP was removed after several minutes (Fig. 3 A). AMP-PNP also suppressed the residual current after Ca²⁺ application, but did not reactivate the channels, as was previously shown (Ohno-Shosaku et al. 1987). Thus, reactivation by MgATP after Ca²⁺-induced rundown involves a phosphorylation process.

The degree of reactivation by MgATP after Ca²⁺ application was dependent on the duration of exposure to Ca²⁺. After short exposures (1 min or less), channel activity could be completely restored, but, after a 5-min exposure to Ca²⁺, reactivation by MgATP never exceeded 10–15% of control activity. After short Ca²⁺ application, the ATP sensitivity of the channels increased (Fig. 3B and Fig. C). In five patches with an initial IC₅₀ of 134 ± 42 μM ATP, exposure to 10 μM Ca²⁺ for ∼1 min decreased the average IC₅₀ to 12 ± 4 μM. Such brief exposures to Ca²⁺ had modest effects on MgADP-dependent reactivation of ATP-inhibited channels, as illustrated in Fig. 3 D, where open circles represent data obtained after Ca²⁺-induced rundown and closed circles data before Ca²⁺ application (Fig. 2 D).

In contrast, after a 5-min exposure to Ca²⁺, the reversal of ATP-dependent inhibition evoked by 200 μM MgADP decreased from 57 ± 27% before (Fig. 3 B), to 12 ± 7% after (C) Ca²⁺ application. Prolonged Ca²⁺ exposure also prevented channel inhibition by sulfonylureas (data not shown). Thus, short exposure to Ca²⁺ evoked increased channel sensitivity to ATP, while prolonged Ca²⁺ exposure caused functional uncoupling of Kir6.2-GFP and SUR1, which was irreversible.

According to the PIP₂ hypothesis (Hilgemann and Ball 1996), Ca²⁺ induces K_ATP channel rundown by activating Ca²⁺-sensitive phospholipase C, which hydrolyzes PIP₂ to levels that no longer stimulate channel activity. Consistent with this hypothesis, Fig. 4 A shows that PIP₂, like MgATP (Fig. 3 A), reactivated channels that were first inhibited by Ca²⁺. However, reactivation by PIP₂ was dependent on the duration of Ca²⁺ exposure. The addition of 50 μM PIP₂ to the bath evoked within 1 min a robust increase in channel activity when Ca²⁺ was applied for 1 min or less (Fig. 4 A), but after prolonged Ca²⁺ exposure (>5 min), the channel activity remaining after Ca²⁺ removal was insensitive to PIP₂ (Fig. 4 B). Spontaneous rundown had similar effects on channel regulation by PIP₂. After fast spontaneous rundown, PIP₂ could stimulate channel activity, but, after slow spontaneous rundown, PIP₂ was ineffective (data not shown).

To further characterize the effects of PIP₂, we examined the channel sensitivity to ATP after Ca²⁺-induced rundown, and then after PIP₂-mediated channel reactivation (Fig. 4, C–E). After brief Ca²⁺-induced rundown, addition of PIP₂ reactivated the channels within 1 min and shifted the IC₅₀ from 14 ± 4 to 185± 54 μM. After 3–5 min, PIP₂ further decreased the ATP sensitivity to 476 ± 92 μM (n = 5) without changing the current amplitude. These results are consistent with previous reports by Shyng and Nichols 1998 and Baukrowitz et al. 1998. After prolonged Ca²⁺-induced rundown (>5 min), channel activity did not increase significantly with PIP₂ (Fig. 4 B), but prolonged exposure (∼10 min) still caused a dramatic decrease in channel sensitivity to ATP, with the IC₅₀ increasing to 5.4 mM in three patches (Fig. 4 E, ▴). In the absence of any significant change in channel activity, this dramatic decrease in channel sensitivity to ATP evoked by PIP₂ after prolonged exposure to Ca²⁺ suggests a direct effect of PIP₂ on ATP binding affinity.

PIP₂ also affected MgADP-dependent reactivation of ATP-inhibited channels. In five experiments, before application of PIP₂, channel reactivation by MgADP averaged 42 ± 14% of control (Fig. 4 C). In contrast, after a 5-min exposure to PIP₂, which dramatically decreased the channel sensitivity to ATP, MgADP-dependent reactivation was suppressed to 8 ± 4% (Fig. 4 D). Under similar experimental conditions, sulfonylurea-dependent inhibition was also dramatically suppressed (Koster et al. 1999; Ribalet 1999). In summary, PIP₂ restored channel activity and reversed the increase in ATP sensitivity evoked by Ca²⁺-induced rundown, but also prevented channel regulation by MgADP and sulfonylureas after prolonged applications.

In addition to rendering the channel insensitive to inhibition by ATP and reactivation by MgADP, PIP₂ inhibited the ability of Ca²⁺ to induce rundown. Before PIP₂ exposure, block of channel activity by Ca²⁺ persisted after removal of Ca²⁺ (Fig. 5 A and 3 A). However, after exposure to PIP₂, channel activity recovered spontaneously when Ca²⁺ was removed (Fig. 5 B), even after a more than 10-min exposure to Ca²⁺ (n = 12). This latter result suggests that Ca²⁺ may directly block K_ATP channels. Lastly, PIP₂ exposure almost completely prevented any further rundown of channel activity, which decreased by <10% within 30 min (n = 8). Together, these data indicate that reactivation by PIP₂ does not restore the channel to the same state as before rundown, suggesting that spontaneous rundown of Kir6.2-GFP+SUR1 must involve other factors besides PIP₂ hydrolysis. Experiments were performed to identify these factors.

Given earlier evidence of K_ATP channel regulation by protein kinases (Hu et al. 1996; Light et al. 1996; Liu et al. 1996), we examined whether rundown may involve dephosphorylation of the channel or a closely associated protein. Accordingly, we tested the effects of protein kinase inhibitors on channel rundown and regulation by ATP and MgADP. Fig. 6 shows that the nonselective protein kinase inhibitor H-7 modestly accelerated rundown of Kir6.2-GFP+SUR1 channel activity. After addition of 100 μM H-7, channel activity began to decay after 1–2 min and slowly decreased by 22 ± 7% after 10 min (n = 6) (Fig. 6 A). After removal of H-7, addition of 100 μM MgATP completely and reversibly blocked channel activity. Okadaic acid, a phosphatase inhibitor, prevented channel inhibition by 100 μM H-7 (Fig. 6 A), indicating that active phosphorylation–dephosphorylation cycling is occurring in the excised patches, similar to previous conclusions of Light et al. 1996 for native cardiac K_ATP channels.

To identify the protein kinase involved, we studied the effects of selective protein kinase inhibitors. The selective protein kinase A inhibitor PKI had almost no effect (data not shown), but the selective PKC inhibitor peptide 19-36 (5 μM) evoked a 13 ± 8% decrease in channel activity after 10 min (n = 5) (Fig. 6 B), similar to H-7. These data are consistent with findings in native K_ATP channels reported previously (Ribalet and Eddlestone 1995; Hu et al. 1996; Light et al. 1996; Liu et al. 1996) implicating PKC in K_ATP channel regulation.

To further characterize rundown induced by protein kinase inhibitors, we tested the effects of H-7 and peptide 19-36 on ATP and MgADP sensitivity. In five patches with minimal initial fast rundown, H-7 decreased the IC₅₀ for channel inhibition by ATP from 175 ± 34 to 8.5 ± 5.5 μM (Fig. 6 E). These data are consistent with previous results indicating that PKC activation decreased the ATP sensitivity of native K_ATP channels (Hu et al. 1996; Light et al. 1996).

With H-7 or peptide 19-36, the increase in ATP sensitivity, which occurred within ∼5 min, was concomitant with the loss of MgADP-dependent reactivation of ATP-inhibited channels. As shown in Fig. 6 D, the presence of MgADP did not prevent channel inhibition by the PKC inhibitor peptide 19-36. In fact, when MgADP was present, channel inhibition by H-7 or PKC inhibitor peptide 19-36 reached 48 ± 12% after 7 min (n = 4), greatly exceeding the inhibition observed in the absence of MgADP (13–22%) for H-7 and peptide 19-36. This reflected the concomitant loss of MgADP stimulation, unmasking the inhibitory effect of ADP. When MgADP was removed, channel activity increased to level similar to that after PKC inhibitor peptide 19-36 treatment in the absence of MgADP (Fig. 6 D, compare with Fig. 4 B). Even 5–10 min after removal of the protein inhibitor, reapplication of MgADP still caused channel inhibition rather than stimulation (Fig. 6 D), indicating that the loss of MgADP stimulation was poorly reversible. After 7 min in the presence of H-7, reactivation by 300 μM MgADP reached only 8 ± 5% of control (Fig. 6 C). At this time, however, PIP₂ could still fully restore channel activity (Fig. 7 B). Much longer (10–15 min) exposures to protein kinase inhibitors were then necessary to suppress PIP₂-dependent channel stimulation.

In summary, it was not possible to assess whether, during prolonged spontaneous rundown, the loss of MgADP and of PIP₂-dependent channel activation involved more than one process. The experiments performed with the protein kinase inhibitors now clearly indicate that the two phenomena can be temporally dissociated and are therefore different. The demonstration that dephosphorylation-induced rundown causes uncoupling of SUR1 from Kir6.2 also supports the hypothesis that the uncoupling evoked by Ca²⁺-induced rundown involves a dephosphorylation process very likely mediated by a Ca²⁺-activated phosphatase.

The effects of MgADP and PIP₂ are similar in that they both stimulate channel activity and thereby decrease ATP sensitivity. The observation that PIP₂ prevents channel regulation by MgADP (Fig. 4 D) raises the possibility that the two modulators may act competitively to regulate channel activity. To test this possibility, we examined the effects of PIP₂ in the presence of MgADP. As shown in Fig. 7 A, when fast rundown was prevented by MgADP, PIP₂ had no stimulatory effect. In contrast, after rundown induced by H-7 or PCK inhibitor peptide 19-36, PIP₂ had potent stimulatory effects even in the presence of 150–300 μM MgADP (Fig. 7 B). The latter result supports the hypothesis that functional coupling of SUR1 to Kir6.2 is not required for channel stimulation by PIP₂. From these data, we may conclude that PIP₂ and MgADP do indeed compete to activate a common regulatory mechanism. However, another possibility exists (discussed later) whereby MgADP and PIP₂ control independent regulatory mechanisms, each capable of producing maximum channel activation.

To determine the extent to which rundown and reactivation of K_ATP channels are properties of the Kir6.2-GFP or the SUR1 subunit, we investigated homomeric Kir6.2-GFP channels, which we have previously shown behave similarly to wild-type Kir6.2 channels (John et al. 1998). In contrast to Kir6.2-GFP+SUR1 channels, homomeric Kir6.2-GFP channels were found to exhibit only a slow phase of rundown when excised into MgATP-free bath solution with low Ca²⁺ (Fig. 8 A). The time constant of decay averaged 780 ± 292s (n = 5).

Similar to Kir6.2-GFP+SUR1 channels, rundown of Kir6.2-GFP channels was markedly accelerated by [Ca²⁺] > 5 μM (Fig. 8 A); 1–2 μM had almost no effect and 20 μM had a maximum inhibitory effect, consistent with the PIP₂ hydrolysis mechanism postulated for native K_ATP channels (Hilgemann and Ball 1996). Also, as with Kir6.2-GFP+SUR1 channels, recovery of channel activity after Ca²⁺ removal varied from patch to patch and was related to the time of exposure to Ca²⁺. After brief exposure to Ca²⁺ (<1 min), channel activity partially recovered, while exposure to Ca²⁺ for >5 min resulted in complete loss of channel activity.

We measured the ATP sensitivity of Kir6.2-GFP channels immediately after patch excision (before rundown) and after Ca²⁺-induced rundown. Immediately after patch excision, the sensitivity to ATP was low with an IC₅₀ of 219 ± 43 μM (Fig. 8B and Fig. D). After a 5-min exposure to 10 μM Ca²⁺, there was little channel reactivation upon Ca²⁺ removal, and the IC₅₀ for ATP decreased to 12.6 ± 2.1 μM (Fig. 8C and Fig. D). We therefore conclude that Ca²⁺-induced rundown and its effect on ATP sensitivity are, at least in part, an intrinsic property of the Kir6.2 channel subunit.

After a brief Ca²⁺-induced rundown, PIP₂ reactivated Kir6.2-GFP channels (Fig. 9 A). In addition, similar to Kir6.2-GFP+SUR1 channels, PIP₂ markedly decreased channel sensitivity to ATP, decreasing the IC₅₀ after Ca²⁺-induced rundown from 12.6 to 465 μM (Fig. 9 B).

As with Kir6.2-GFP+SUR1, H-7 and peptide 19-36 caused slow inhibition of Kir6.2-GFP (Fig. 9 C). In five experiments, inhibition averaged 19 ± 8% of control after 10 min. Channel activity partially recovered after addition and withdrawal of MgATP (Fig. 9 C). These results indicate that protein kinases have a direct stimulatory effect on Kir6.2-GFP, in addition to facilitating coupling between Kir6.2-GFP and SUR1. However, in contrast to Kir6.2-GFP+SUR1 channels, the sensitivity of Kir6.2-GFP channels to ATP was not affected by protein kinase C inhibitors even though their activity decreased, possibly because of the nonlinear relationship between P_o and ATP sensitivity (see discussion).

We next compared the ability of PIP₂ to reactivate Kir6.2-GFP channels and Kir6.2-GFP+SUR1 channels after combined protein kinase inhibition and brief Ca²⁺-induced rundown. With Kir6.2-GFP alone, short exposures to protein kinase inhibitors for 2–3 min followed by brief Ca²⁺ exposure were sufficient to prevent channel reactivation by PIP₂ (Fig. 9 D). In contrast, Kir6.2-GFP+SUR1 channels subjected to the same protocol were readily reactivated by PIP₂ (Fig. 9 E), and prolonged exposure to protein kinase inhibition was required to finally disrupt PIP₂ regulation. Thus, the heteromultimeric channel is less sensitive to treatments that cause slow rundown, suggesting that SUR1 may protect Kir6.2 from dephosphorylation.

To search for additional evidence of dual regulation of Kir6.2-GFP channels by protein kinases and PIP₂, we examined the effects of channel reactivation by MgATP and PIP₂ on single Kir6.2-GFP channel kinetics. The rationale is that reactivation by MgATP should facilitate both protein kinase–mediated phosphorylation and PIP₂ resynthesis, and may therefore have different effects than PIP₂ alone. We were unable to analyze Kir6.2-GFP+SUR1 single-channel activity since patches contained too many channels, and, even with Kir6.2-GFP channels, patches typically had more than one channel. Therefore, we restricted our study to the analysis of open-time distributions after Ca²⁺-induced rundown left only a few active Kir6.2-GFP channels in the patch.

Fig. 10 shows representative single--channel traces and open-time distributions before and after application of either MgATP (Fig. 10A and Fig. B) or PIP₂ (C and D) to reactivate the channels. The open-time histogram obtained before and after MgATP was well fitted by two exponentials. Before MgATP, the time constants of the two exponentials were τ₂ = 0.6 ± 0.2 ms and τ₂ = 2.7 ± 0.6 ms and, after MgATP, the time constants were τ₂ = 1 ± 0.3 ms and τ₂ = 3 ± 0.3 ms (n = 4) (Fig. 10 E). According to the data shown in Fig. 10B and Fig. E, the rebound in channel activity evoked by MgATP was primarily due to a 5.6 ± 1.8-fold increase in the frequency of short open times. In contrast, the frequency of the burst of openings increased only by 1.2 ± 0.5-fold.

Similarly, after PIP₂, the open-time histogram was well fit by two exponentials with time constants of τ₂ = 0.8 ± 0.3 ms and τ₂ = 3.7 ± 1.1 ms. After PIP₂ application, the frequency of the burst of openings increased by 5.4 ± 2.3-fold, while that of short openings increased only moderately by 1.3 ± 0.6-fold (Fig. 10, C–E). Thus, PIP₂ appeared to facilitate channel transition to a prolonged active state, while MgATP primarily facilitated transition to a short open state. The different kinetic effects of MgATP and PIP₂ on Kir6.2-GFP channels' open-time distributions are consistent with two MgATP-dependent processes, mediated by protein kinase and PIP₂, and they suggest that the major control sites are located on the pore-forming Kir6.2 subunit.

The major conclusion of this study is that rundown of K_ATP channels involves multiple factors, including MgADP, PIP₂, and protein kinases. This conclusion is consistent with previous studies describing regulation of native and cloned K_ATP channels by protein kinases (Ribalet and Eddlestone 1995; Hu et al. 1996; Light et al. 1996; Liu et al. 1996), PIP₂ (Furukawa et al. 1996; Hilgemann and Ball 1996; Fan and Makielski 1997, Fan and Makielski 1999; Baukrowitz et al. 1998; Shyng and Nichols 1998), and nucleotide diphosphates (Findlay 1987; Bokvist et al. 1991; Terzic et al. 1994). Our findings can be summarized as follows. (a) Spontaneous rundown of Kir6.2-GFP+SUR1 involves a fast phase lasting close to 1 min that caused an increase in ATP sensitivity and that is prevented by MgADP. A slower phase then followed, lasting over tens of minutes to hours that caused a loss of sensitivity to MgADP, sulfonylureas, and PIP₂. (b) Brief exposure to Ca²⁺ (<1 min) increased only the ATP sensitivity, while prolonged exposure (>5 min) suppressed the channel sensitivity to MgADP, sulfonylureas, and PIP₂. (c) Protein kinase inhibitors had similar effects, but the loss of MgADP-dependent reactivation preceded the loss of PIP₂-dependent stimulation. (d) Kir6.2-GFP alone exhibited only a slow phase of rundown, suggesting that fast rundown involves SUR1 and slow rundown is a property of Kir6.2.

The fast phase of spontaneous rundown is specific to Kir6.2-GFP+SUR1, and therefore involves SUR1. Since MgADP in the bath prevented fast rundown (Fig. 2 A), we postulate that, in the absence of bath-applied MgADP, dissociation of MgADP from SUR1, with consequent loss of MgADP-dependent stimulation, is responsible for fast rundown. Previous studies consistent with this hypothesis have shown that the rate of closure of phosphorylated channels (∼1 min) upon removal of nucleotide diphosphates (Terzic et al. 1994) is similar to that of fast rundown (Fig. 1 A).

Fast rundown is characterized by an increase in ATP sensitivity. Such an increase in ATP sensitivity results, at least in part, from a decrease in channel activity since a correlation has been demonstrated between P_o and the channel sensitivity to ATP. Thus, decreasing the mean open time would shift the equilibrium towards the closed state to which ATP binds preferentially, thereby increasing the apparent affinity and sensitivity of the channel to inhibition by ATP (Shyng et al. 1997; Trapp et al. 1998; Tucker et al. 1998; Babenko et al. 1999; Fan and Makielski 1999). Accordingly, the decrease in channel activity and the accompanying increase in ATP sensitivity observed during fast rundown may be due to dissociation of the adenine nucleotide and loss of MgADP-dependent stimulation. As a corollary, the high ATP sensitivity of Kir6.2+SUR1 and native K_ATP channels (IC₅₀ ≅ 15μM) reported in many excised patch-clamp studies may result from fast rundown and loss of MgADP-dependent stimulation. This hypothesis is supported by the present data as well as previous reports demonstrating the effect of MgADP on ATP sensitivity, indicating that the physiological ATP sensitivity of Kir6.2+SUR1 and native K_ATP channels in the presence of intracellular MgADP may lie between 100 and 400 μM. This “low” ATP sensitivity would be compatible with the role of ATP-sensitive channels as metabolic sensors designed to respond to physiologically relevant variations in cellular ATP levels.

In contrast to fast rundown, slow rundown was observed with Kir6.2-GFP+SUR1, as well as Kir6.2-GFP alone. Thus, the slow phase is, at least in part, a characteristic of Kir6.2-GFP. Moreover, slow rundown of Kir6.2-GFP alone had at least two components, since brief Ca²⁺ application left the channels reactivatable by MgATP and PIP₂, whereas prolonged Ca²⁺ application did not. Lastly, slow rundown of Kir6.2-GFP+SUR1 exhibits a third feature involving functional uncoupling of Kir6.2 and SUR1, since MgADP-dependent reactivation of ATP-inhibited channels was gradually lost during slow rundown of the heteromeric channel.

With Kir6.2-GFP alone, brief exposure to Ca²⁺ facilitated spontaneous slow rundown by accelerating channel closure and increasing ATP sensitivity. Direct application of PIP₂ or MgATP reversed these effects (Fig. 9). These results are consistent with recent studies of PIP₂ on native and cloned K_ATP channels (Furukawa et al. 1996; Hilgemann and Ball 1996; Fan and Makielski 1997, Fan and Makielski 1999; Baukrowitz et al. 1998; Shyng and Nichols 1998) and suggest that one component of spontaneous slow rundown of Kir6.2-GFP channels involves PIP₂ hydrolysis. In accordance with this hypothesis, PIP₂ hydrolysis ought to be a slow process that requires hours. This may indeed very well be the case at submicromolar free Ca²⁺ concentrations, since our data showed that prolonged channel inhibition by Ca²⁺ was fully reversible when membrane patches had been previously exposed to excess exogenous PIP₂ (Fig. 5 B).

The decrease in ATP sensitivity induced by exogenous PIP₂ could very well be explained by the increase in channel activity evoked by PIP₂ assuming that ATP binds preferentially to a closed state of the channel (Shyng and Nichols 1998) and that PIP₂ shifts the equilibrium away from the closed state, thereby decreasing the apparent affinity and sensitivity of the channel to inhibition by ATP. According to this hypothesis, when P_o is near-maximal, small absolute changes (e.g., from 0.90 to 0.99), which are experimentally undetectable, could have major effects on ATP sensitivity. However, after slow spontaneous rundown or prolonged exposure to Ca²⁺, channel activity is at a low level and application of PIP₂ does not significantly increase the activity of these rundown channels, yet markedly decreases their ATP sensitivity. This data is not consistent with the former mechanism, and indicates a direct effect of PIP₂ on ATP binding affinity (Fan and Makielski 1999). Therefore, we conclude that PIP₂ decreases the ATP sensitivity by affecting ATP binding affinity as well as channel gating.

Protein kinase inhibition facilitates the slow rundown process of Kir6.2-GFP and Kir6.2-GFP+SUR1 channels, preventing reactivation of both channels by PIP₂ and changing the ATP and MgADP sensitivity of the latter channels. These results suggest that a second component of slow rundown is related to protein dephosphorylation, which is consistent with previous studies investigating the effects of protein kinases on K_ATP channel function. These studies did not specifically address rundown, but consistently showed that PKC desensitized K_ATP channels to ATP (Ribalet and Eddlestone 1995; Hu et al. 1996; Light et al. 1996; Liu et al. 1996). The increase in ATP sensitivity observed with Kir6.2-GFP+SUR1 channels during protein kinase inhibitor-induced rundown may be attributed in part to a decrease in channel activity due to loss of MgADP and/or PIP₂ stimulation. However, the decrease in channel activity was small (Fig. 6A and Fig. B) compared, for instance, with that measured during fast rundown, and may not account fully for the substantial increase in ATP sensitivity observed. This apparent discrepancy may be related to the nonlinear relationship between P_o in the absence of ATP and channel sensitivity to ATP (Trapp et al. 1998; Enkvetchakul et al. 2000). Interestingly, with Kir6.2-GFP alone, the small decrease in channel activity induced by PKC inhibitors did not cause a significant increase in ATP sensitivity. However, changes in P_o are most effective at altering ATP sensitivity when P_o is near maximal (∼0.9), as may be the case with Kir6.2-GFP+SUR1. The P_o of Kir6.2, on the other hand, is considerably lower (<0.5) (Tucker et al. 1997; John et al. 1998), and small changes in this P_o evoked by protein kinase inhibition are thus expected to have little effect on ATP sensitivity. These findings indicate that changes in ATP sensitivity associated with PKC inhibition are mediated primarily by changes in P_o. Alternatively, protein dephosphorylation may also directly decrease the affinity of Kir6.2 for ATP.

Recent studies with Kir6.2 mutants have suggested two types of interactions between ATP and Kir6.2, which involve two distinct regions of the pore-forming protein (Drain et al. 1998; Tucker et al. 1998). Mutations in one region, just outside the M2 transmembrane domain (around amino acids 170–180) cause a decrease in ATP sensitivity together with an increase in P_o. It has been suggested that this region comprises a gate, which interacts with ATP to close the channel. Mutation-induced weakening in this interaction would cause an increase in P_o and an apparent decrease in ATP sensitivity. Mutations in a second region, farther away from the M2 domain (amino acids 334 and 337, which may also include K185), also decrease the channel sensitivity to ATP, but without change in P_o. It has been postulated that this region is part of the ATP binding site. In this case, the mutations induce a decrease in ATP sensitivity by decreasing the channel affinity for ATP (Drain et al. 1998; Tucker et al. 1998). These observations are consistent with our findings and explain how interaction of PIP₂ and protein kinase with one or the other of these regions may affect the channel sensitivity to ATP by changing P_o or the binding of ATP to the channel. Interestingly, T180, which is part of the region that controls ATP binding, is a consensus site for PKC-dependent phosphorylation. We postulate that the effect of PKC inhibitors on ATP sensitivity, which we present here, may involve dephosphorylation of T180.

We cannot totally exclude that the decrease in channel activity evoked by protein kinase inhibitors may be due to nonspecific inhibition of phosphatidylinositol-kinase causing decreased PIP₂ synthesis. However, channel inhibition by the specific protein kinase inhibitor peptide 19-36, as well as the less specific inhibitor H-7, indicates that protein dephosphorylation is very likely to decrease channel activity during slow rundown. Because H-7 and peptide 19-36 suppress the PIP₂-dependent regulation of Kir6.2-GFP+SUR1 as well as of Kir6.2-GFP alone, we postulate that the inhibitors act by deactivating a PIP₂ regulatory site on Kir6.2, which may include residues R176 and R177 (Fan and Makielski 1997). It should also be noted that when SUR1 was coexpressed with Kir6.2-GFP, the combined effects of protein kinase inhibitors and Ca²⁺ on PIP₂-dependent regulation were less prominent, suggesting that SUR1 protects Kir6.2-GFP from dephosphorylation. This may contribute to the low P_o of functional homomeric Kir6.2 channels expressed without SUR1.

In addition to preventing reactivation by MgATP and PIP₂, dephosphorylation of Kir6.2-GFP+SUR1 channels also caused functional uncoupling of SUR1 from Kir6.2-GFP, since both MgADP-dependent reactivation and sulfonylurea inhibition were abolished. Thus, with Kir6.2-GFP+SUR1, the decrease in channel activity associated with slow rundown may have multiple causes, including loss of PIP₂-dependent stimulation as well as uncoupling of Kir6.2-GFP and SUR1.

Dephosphorylation-induced loss of MgADP-dependent reactivation of ATP-inhibited channels may be due to direct uncoupling of SUR1 from Kir6.2-GFP or decreased MgADP binding to SUR's nucleotide-binding fold (NBF), in which case uncoupling would only be apparent or functional. The protein phosphorylation site that regulates coupling between Kir6.2 and SUR1 is probably different from the phosphorylation site that controls PIP₂ regulation, since with Kir6.2-GFP+SUR1 channels subjected to protein kinase inhibitors, the loss of MgADP sensitivity and of PIP₂ stimulation occurred with different time courses (Fig. 7 B). If dephosphorylation-induced uncoupling is assumed to be primarily functional, it is then conceivable that the phosphorylation site controlling this process would be located on SUR1 rather than on Kir6.2. The hypothesis that phosphorylation of SUR1 favors binding of MgADP and thus functional coupling is supported by the observation that the K1384M mutation in SUR1's second NBF completely suppress MgADP-dependent reactivation of ATP-inhibited channels (Gribble et al. 1997). This lysine is part of a sequence identified as a PKC phosphorylation site (S X K) found in all SURs as well as CFTR second nucleotide-binding fold. Thus, slow rundown of Kir6.2-GFP+SUR1 may involve uncoupling of SUR1 from Kir6.2-GFP, as a result of dephosphorylation of one or both subunits, together with PIP₂ hydrolysis.

Previous kinetic studies have shown that K_ATP channels may exist in different states, with interconversion between states regulated by such factors as nucleotide disphosphates (Terzic et al. 1994; Alekseev et al. 1998) and PIP₂ (Trapp et al. 1998; Fan and Makielski 1999). Such gating schemes can be extended to account for most of our observations on channel rundown and reactivation. Fig. 11 depicts one possible scheme.

Fig. 11 summarizes how the channel activity can be modulated by PIP₂ and MgADP binding to specific sites on Kir6.2 and SUR1, respectively, and by dephosphorylation. The channel rundown seen experimentally is described by several channel behaviors that occur in parallel or sequentially after patch excision. With phosphorylated channel, fast and slow rundown occur due to loss of MgADP and PIP₂, respectively (pathways A and C). It is assumed that MgADP dissociation is fast, while PIP₂ hydrolysis is slow, and that States 1–3 have similar P_o, which are near maximum, whereas state 4 has a low P_o. The assumption that States 2 and 3 have higher P_o than State 4 is based on the observation that PIP₂ and MgADP stimulate rundown channels. In addition, the P_o of State 1 would not be higher than that of States 2 or 3 because MgADP had no further stimulatory effect on patches that exhibited only slow rundown upon patch excision (transition to State 3) and PIP₂ had no obvious stimulatory effect either when patches were excised in the presence of MgADP. The highly variable degree of fast rundown among excised patches can be explained by assuming differing levels of channels in States 1 and 2 before patch excision. With patches with most channels in State 1 (PIP₂ and MgADP regulated), there would be no fast rundown upon dissociation of MgADP, as PIP₂ remains bound, whereas with patches with most channels in State 2 (MgADP regulated), fast rundown would be prominent as MgADP dissociates.

For slow rundown, one pathway involves PIP₂ hydrolysis (pathway C), another may involve channel dephosphorylation leading to uncoupling of SUR1 from Kir6.2 (pathway B). In Fig. 11, we show pathway B to follow MgADP dissociation (pathway A). We have no experimental evidence in favor of this specific scheme and the two pathways are shown as sequential only for clarity. In fact, the observation that in two patches channel rundown occurred within 2–4 min after patch excision even in the presence of MgADP, and ADP was inhibitory afterward, suggests that process A (dissociation of MgADP) and process B (dephosphorylation-induced uncoupling) can occur simultaneously. In other words, MgADP would not protect from dephosphorylation-induced uncoupling. This hypothesis is supported by our observation that dephosphorylation, induced experimentally by protein kinase inhibition, causes pronounced channel inhibition in the presence of MgADP. This marked inhibition was presumably a result of SUR1 uncoupling from Kir6.2, which unmasked the inhibitory effect of ADP binding to the ATP inhibitory site on Kir6.2. Such an effect would be most prominent when intracellular PIP₂ is low and has little effect on channel activity (transition from State 2 to 4 to 6). In contrast, when PIP₂ controls channel opening (high P_o), uncoupling due to dephosphorylation may have little effect on channel activity and sensitivity to ATP (transition from State 1 to 3 to 5). This hypothesis is supported by our observation that protein kinase inhibitors had almost no effect on channel activity after PIP₂ treatment (data not shown). However, after inducing loss of MgADP-dependent reactivation, long exposures to the protein kinase inhibitors also caused deactivation of the PIP₂ regulatory site on Kir6.2. Thus, dephosphorylation-induced rundown may involve two irreversible processes, one relatively fast and mediated by uncoupling of SUR1 from Kir6.2, and a slower one, involving deactivation of the PIP₂ regulatory mechanism. The latter indicates that loss of PIP₂ (pathway C) may, therefore, occur as a result of hydrolysis as well as dephosphorylation of Kir6.2. Finally, complete dephosphorylation of Kir6.2 may lead from either State 5 or 6 to a final state characterized by complete and irreversible channel closure (not shown).

Our data also indicate that prolonged exposure to PIP₂ causes a dramatic decrease in ATP sensitivity, without noticeable change in channel activity. This decrease in sensitivity to ATP evoked by PIP₂ is observed before as well as after prolonged and irreversible rundown. Before irreversible channel rundown, the main decrease in sensitivity to ATP occurs long after channel stimulation has reached saturation (Shyng and Nichols 1998). Due to the extreme nonlinearity observed between ATP sensitivity and P_o when P_o is near maximum (Trapp et al. 1998; Enkvetchakul et al. 2000), PIP₂-induced decrease in ATP sensitivity may be explained, in this case, by minimal increase in P_o. However, such a mechanism cannot account for the dramatic decrease in ATP sensitivity observed with prolonged exposure to PIP₂ after irreversible channel rundown, since P_o remains very low. To account for this effect, we propose, in accordance with Fan and Makielski 1999, that PIP₂ has also a direct inhibitory effect on ATP binding to Kir6.2. This effect due to prolonged exposure to PIP₂ would occur independently of the channel state (Fig. 11 B). After slow rundown, the decrease in ATP affinity would lead the channel into States 5′ and 6′. A similar effect of PIP₂ may occur before slow rundown that would lead to States 3′ and 4′. It remains to be determined whether this effect of PIP₂ on adenine nucleotide binding to K_ATP channels is physiological, since K_ATP channels must be closed by intracellular ATP. It may therefore be postulated that intracellular PIP₂ levels are too low to cause adenine nucleotide desensitization. Alternatively, interaction of SUR with Kir6.2 may prevent this effect of PIP₂ in native channels. For these reasons, process D is not shown to have any effect on States 1 and 2.

In summary, MgADP, PIP₂, and subunit phosphorylation have complex regulatory effects on Kir6.2+SUR1 and native K_ATP channels. They all stimulate channel activity and decrease ATP sensitivity to physiological levels. However, MgADP appears to have short-term effects, while PIP₂ and subunit phosphorylation regulate channel activity over longer time courses. Thus, MgADP is more likely to regulate acute cell function such as glucose-induced β-cell membrane depolarization, which occurs on a time scale of minutes, whereas PIP₂ and phosphorylation may be more important in long-term regulation relevant to physiological adaptation and/or pathophysiological conditions. Finally, with physiological MgADP present, intracellular PIP₂ may set the channel sensitivity to ATP near 400 μM when high and 150 μM when low.

The authors thank Dr. Y. Lu for her expert assistance in preparing the various DNA constructs and Dr. J.R. Monck for his very constructive comments.

This work was supported by a Research Grant from the American Diabetes Association to B.Ribalet, by National Institutes of Health grants RO1HL36729, R37HL60025, and P50HL52319, and Laubisch and Kawata Endowments to J.N. Weiss, and by a Grant-in-Aid from the American Heart Association (Greater Los Angeles Affiliate) to S.A. John.