We have further tested the hypothesis that receptor-mediated modulation of KCNQ channels involves depletion of phosphatidylinositol 4,5-bisphosphate (PIP2) by phosphoinositide-specific phospholipase C (PLC). We used four parallel assays to characterize the agonist-induced PLC response of cells (tsA or CHO cells) expressing M1 muscarinic receptors: translocation of two fluorescent probes for membrane lipids, release of calcium from intracellular stores, and chemical measurement of acidic lipids. Occupation of M1 receptors activates PLC and consumes cellular PIP2 in less than a minute and also partially depletes mono- and unphosphorylated phosphoinositides. KCNQ current is simultaneously suppressed. Two inhibitors of PLC, U73122 and edelfosine (ET-18-OCH3), can block the muscarinic actions completely, including suppression of KCNQ current. However, U73122 also had many side effects that were attributable to alkylation of various proteins. These were mimicked or occluded by prior reaction with the alkylating agent N-ethylmaleimide and included block of pertussis toxin–sensitive G proteins and effects that resembled a weak activation of PLC or an inhibition of lipid kinases. By our functional criteria, the putative PLC activator m-3M3FBS did stimulate PLC, but with a delay and an irregular time course. It also suppressed KCNQ current. The M1 receptor–mediated activation of PLC and suppression of KCNQ current were stopped by lowering intracellular calcium well below resting levels and were slowed by not allowing intracellular calcium to rise in response to PLC activation. Thus calcium release induced by PLC activation feeds back immediately on PLC, accelerating it during muscarinic stimulation in strong positive feedback. These experiments clarify important properties of receptor-coupled PLC responses and their inhibition in the context of the living cell. In each test, the suppression of KCNQ current closely paralleled the expected fall of PIP2. The results are described by a kinetic model.
Introduction
Recent work attributes the long-known muscarinic suppression of M-current K+ channels (Brown and Adams, 1980) to a receptor-driven depletion of the trace membrane lipid phosphatidylinositol 4,5-bisphosphate (PIP2) (Suh and Hille, 2002; Zhang et al., 2003; Ford et al., 2004; Winks et al., 2005). In this hypothesis, M-current channels require PIP2 to function, and activation of M1 muscarinic receptors results in destruction of PIP2 via PIP2-specific PLC. Our work seeks to strengthen the evidence for channel regulation by PIP2 through a deeper understanding of the parameters of PLC activation and pharmacology in living cells and by asking if M-current modulation parallels PLC activation and PIP2 hydrolysis under several conditions. Many cytoplasmic enzymes and cytoskeletal proteins become localized to membranes that contain PIP2 (Lemmon, 2003), and activities of many kinds of membrane proteins, including ion channels and transporters, depend on PIP2 levels in the plasma membrane (Hilgemann et al., 2001; Suh and Hille, 2005), so now there is renewed interest in understanding how PLC regulates phosphoinositide levels.
A large biochemical literature describes the signaling properties of phosphatidylinositol (PI)-specific PLCs in vitro (for review see Rhee, 2001). A few selected points given here are relevant for our work. This family of enzymes catalyzes the hydrolysis of PIP2 to inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). The β isoform, PLCβ, is activated by G protein–coupled receptors that couple through G proteins of the Gq family by direct action of GαqGTP (Smrcka et al., 1991), and PLCβ can also be activated by G protein βγ subunits (Rhee, 2001). Stimulation of Gq-coupled receptors in intact cells usually results in intracellular Ca2+ release mediated by gating of IP3 receptors, which may be followed by additional Ca2+ influx from the extracellular space. In test tube studies, the stimulation of PLC depends on the local calcium concentration [Ca2+] according to a broad, bell-shaped curve with a maximum between 10 and 100 μM Ca2+ (Ryu et al., 1987b). In spite of considerable efforts, the pharmacology of PLCs remains poorly developed and controversial. Many papers show that several isoforms of PLC are inhibited by the aminosteroid compound U73122 and not by its “inactive” analogue U73343 (Smith et al., 1990). It is also inhibited by the ether lipid analogue edelfosine (ET-18-OCH3) (Powis et al., 1992). Finally, a recent report says that several isoforms of PLC are activated by the benzenesulfonamide m-3M3FBS and not by its inactive analogue o-3M3FBS (Bae et al., 2003), but the specificity of this action has been disputed (Krjukova et al., 2004).
Here we are interested in the in vivo properties of PLC as they affect phosphoinositide metabolism and KCNQ current. We use three nondestructive, continuous monitors to follow the activity of PLC during stimulation by transfected M1 muscarinic receptors or by m-3M3FBS in intact tsA-201 cells, and we use a biochemical assay of phospholipids in Chinese hamster ovary (CHO) cells. The M1 receptor pathway uses Gq/11 in tsA-201 cells (Hirdes et al., 2004; Suh et al., 2004) and hence should selectively stimulate PLCβ isoforms. Activation of PLC is monitored with a confocal fluorescence microscope by observing two optical probes that track depletion of PIP2 and production of IP3 and DAG. These widely used translocation probes migrate between the cell membrane and cytoplasm during PLC activity. They are fusions of enhanced green fluorescent protein (EGFP) with the PIP2- and IP3-binding pleckstrin homology (PH) domain of PLCδ1 or with the DAG-binding PKC cysteine-rich homology-1 domain, C1A (PKC-C1; Oancea et al., 1998; Stauffer et al., 1998; Várnai and Balla, 1998). In addition, downstream effects of PLC are monitored optically with a Ca2+-sensitive fluorescent dye that reports the IP3-dependent release of Ca2+ from intracellular stores and Ca2+ entry from the medium. Further, we assay the relative sizes of various phospholipid pools by biochemical methods. Finally, parallel measurements are made of M current as expressed from KCNQ2 and KCNQ3 channel subunits (KCNQ2/3 current).
Our paper asks the following questions. Is PIP2 depleted by activation of M1 receptors, and to what extent do the different reporters of PIP2 monitor the same signal? What are the consequences and mechanism of inhibition by U73122? What are the parameters of activation by m-3M3FBS? Can PLC be activated at cellular resting levels of cytoplasmic [Ca2+]i? Is it speeded significantly by [Ca2+]i elevations, and is it slowed by [Ca2+]i reductions? Does this regulation affect whether PIP2 can be depleted significantly in the plasma membrane during receptor activation? Do changes of KCNQ2/3 current mirror changes of PIP2?
Materials And Methods
Cell Culture and Transfection
Human embryonic kidney tsA-201 (tsA) cells were cultured and transiently transfected using Lipofectamine 2000 (Invitrogen) with various cDNAs (Suh et al., 2004): M1-muscarinic receptor (1 μg, from Neil Nathanson, University of Washington, Seattle, WA), PH-PLCδ1-EGFP (PH-EGFP, 0.25 μg, from P. De Camilli, HHMI, Yale University, New Haven, CT), PKC-C1A-EGFP (C1-EGFP, 0.25 μg, from T. Meyer, Stanford University, Stanford, CA), PH-PLCδ1-mRFP (PH-mRFP, 0.25 μg, from C. Kearn, University of Washington, using mRFP from Roger Tsien, University of California, San Diego, CA), inositol 1,4,5-trisphosphate-5-phosphatase (IP3-5-P, 2 μg, from M. Iino, University of Tokyo, Japan), the channel subunits KCNQ2 and KCNQ3 (Kv7.2 and Kv7.3, 1 μg, from David McKinnon, State University of New York, Stony Brook, NY), RGS-2 (2 μg, Guthrie Research Institute, Sayre, PA), and GFP (0.1 μg) as a marker for transfection. TsA cells were maintained in DMEM (Invitrogen) supplemented with 10% FCS and 0.2% penicillin/streptomycin. In a few electrophysiological experiments we used the mouse AtT-20 neuroendocrine cell line, which expresses endogenous M4 muscarinic receptors and G protein–coupled inward rectifiers (GIRK channels). They were grown in DMEM supplemented with 10% horse serum and 0.2% penicillin/streptomycin. For biochemical experiments we used a line of CHO cells stably transfected with M1-muscarinic receptors (CHO-M1 from N.J. Buckley and D.A. Brown, University College, London, UK). They were grown in F-12K nutritional mixture with 10% bovine calf serum and 1% penicillin/streptomycin.
Reagents
The muscarinic receptor agonist oxotremorine-M (oxo-M) was used at 10 μM. Chemicals were purchased from Sigma-Aldrich, except U73122 (1-[6-[[17 β-3-methoxyestra-1,3,5(10)-trien-17-yl]amino]hexyl]-1H-pyrrole-2,5-dione) and U73343 (Research Biochemicals), BAPTA-AM and BAPTA (Molecular Probes), m-3M3FBS (2,4,6-trimethyl-N-(meta-3-trifluoromethyl-phenyl)-benzene-sulfonamide) and o-3M3FBS (Tocris), and edelfosine (Calbiochem).
Confocal Imaging and Analysis
TsA cells were imaged 24–48 h after transfection on poly-l-lysine–coated coverslips (Suh et al., 2004). The external Ringer's solution used for confocal observations and for current recording contained (in mM): 160 NaCl, 2.5 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES, and 8 glucose, adjusted to pH 7.4 with NaOH. Images were taken every 5 s on a Leica TCS/MP confocal microscope at room temperature and processed with Metamorph (UIC) and Igor Pro 4.0 (Wavemetrics, Inc.). To obtain averaged time courses, the fluorescence intensity F over a given region of the cytoplasm or nucleus was usually normalized to the average intensity for the 30 s before agonist application F0 (F/F0). Images are shown here in negative contrast so that the brightest fluorescence appears black.
Electrophysiology and Calcium Photometry
TsA cells were whole-cell clamped at room temperature 24–48 h after transfection (Suh et al., 2004). The pipette solution contained (in mM): 175 KCl, 5 MgCl2, 5 HEPES, 0.1 BAPTA, 3 Na2ATP, 0.1 Na3GTP, pH 7.4 adjusted with KOH. For the experiments of Fig. 8 (E and F), the pipette solution was sometimes supplemented with 10 mM calcium, 1 μM thapsigargin, 100 μg/ml pentosan polysulfate (an IP3 receptor blocker; Bezprozvanny et al., 1993), and more BAPTA as described. The special pipette solution with 10 mM CaCl2 and 20 mM BAPTA has an estimated free [Ca2+] of 128 nM. When measuring the rates of induction and recovery from muscarinic inhibition of the current, we applied test and control solutions rapidly to the 100 μl chamber (flow rate of 1.5 ml min−1) in the vicinity of the recorded cell. The solutions exchanged in ∼6 s. Recordings were performed using a HEKA EPC-9 amplifier with Pulse software (HEKA Elektronik). Data analysis was done using Igor Pro 4.0. Cells were held at a holding potential of −20 mV, and 500-ms test pulses to −60 mV were given every 4 s. Reported values (I−20 mV) are mean outward current taken at −20 mV just before each step to −60 mV. Recordings started ∼5 min after breakthrough. Some experiments studied inward rectifier currents in AtT-20 cells, which were recorded using the same pipette solution but with a bath solution where the Na+ and K+ concentrations were changed to 125 Na+ and 40 K+. For photometry, cells were preloaded with indo-1-AM (Molecular Probes) and not subjected to patch clamp. The ratio of background-corrected 405/505 nm emission elicited by excitation at 365 nm was used to calculate absolute [Ca2+]i levels (Shapiro et al., 2000).
Analysis of Acidic Membrane Lipids
Membrane lipids of CHO-M1 cells were analyzed as previously described (Nasuhoglu et al., 2002a). In brief, acidic phospholipids were extracted and deacylated to give the parent glycero-headgroups, which were then separated by HPLC and detected by suppressed conductivity. The relative quantity of each headgroup is expressed as the percentage of all acidic lipid headgroups.
Student's unpaired t test (2 tailed), or, when indicated, a one-way ANOVA with a Bonferroni post-hoc test for multiple comparisons, was used to test for significance. Where error bars are shown, they represent SEM
Online Supplemental Material
The supplemental material for this paper comprises one figure and the listing of a computer program . Fig. S1 shows changes in the voltage dependence of activation of KCNQ current caused by addition of U73122 or N-ethylmaleimide. The computer program describes the elementary model for translocation of PH-EGFP and C1-EGFP probes between cytoplasm and nucleus.
Results
Two Translocation Probes Report Activation of PLC via M1 Receptors
We start with control experiments to understand the properties of our two translocation probes. Fig. 1 A shows the M1 receptor–mediated activation of PLC as seen with the PH-EGFP probe (top) and the C1-EGFP probe (bottom) in separate cells before, during, and after a 3-min application of the muscarinic agonist oxo-M (10 μM). The closed green lines outline a cytoplasmic region of interest and the red lines, a nuclear region of interest. Since the PH-EGFP has affinity for PIP2 and IP3, it binds to the PIP2 in the plasma membrane at rest (dark cell outline in first frame), translocates to the cytoplasm as oxo-M stimulation of PLC causes IP3 to rise in the cytoplasm and PIP2 to fall at the membrane (darkened cytoplasm in second frame), and returns to the plasma membrane after oxo-M is washed off while IP3 is being hydrolyzed and PIP2 resynthesized (last frame). The rise and fall of cytoplasmic PH-EGFP fluorescence is reversible and repeatable (Fig. 1 B, green). Similar experiments with the C1-EGFP probe, which binds to DAG, show a reciprocal time course of translocation. That probe is uniformly distributed in the cytoplasm and nucleus at rest (dark cytoplasm in Fig. 1 A and symbols in Fig. 1 C), reporting little DAG on any membrane, but migrates to the plasma membrane as oxo-M stimulation of PLC produces DAG there, and returns to the cytoplasm when oxo-M removal allows DAG to decline. Oxo-M affects the two probes with approximately equal potency, giving midpoints for peak translocation near 0.25 and 0.28 μM oxo-M (Fig. 1, D and E). The initial migration takes 10–15 s (Fig. 1, B and C; points are 5 s apart), and the recovery takes 100–200 s.
The cytoplasmic concentration change of both probes sags during the 3-min-long application of oxo-M (Fig. 1, B and C) as if there were some kind of desensitization and reversal of the PLC response. This implied that there might be some unexpected features to PIP2 metabolism. However, further analysis suggests that the sag is at least partly caused by an additional slow exchange of probe molecules between the nucleus and cytoplasm. Compare the thin red traces of nuclear fluorescence in Fig. 1 (B and C) with the heavier green traces of cytoplasmic fluorescence. At rest, the nuclear and cytoplasmic concentrations are similar. During repeated stimulation and wash, the nuclear fluorescence starts to rise after the cytoplasmic fluorescence rises, and to fall after the cytoplasmic fluorescence falls. The same can be seen in the pictures of Fig. 1 A, where the nuclear and cytoplasmic fluorescence are about equal at rest (Fig. 1 A, top left), and early in the oxo-M treatment the nucleus (clear disk inside cell) remains much lighter than the cytoplasm, but after washout of oxo-M, the nucleus remains a little darker than before (Fig. 1 A, top right). Simple modeling (see APPENDIX and Fig. 11) shows that the nuclear fluorescence tracks the cytoplasmic fluorescence with an exponential relaxation time constant τ of ∼360 or 225 s (for PH-EGFP and C1-EGFP, respectively), equivalent to nuclear envelope permeability coefficients of 0.0025 or 0.004 μm s−1. Slow exchange with the nucleus accounts for perhaps half of the sag of the response during oxo-M, for most of the observed overshoot in the recovery of C1-EGFP after oxo-M is removed, and for the significant background level of cytoplasmic and nuclear PH-EGFP before oxo-M and of C1-EGFP during oxo-M. Thus the background level can be described as an equilibrium distribution of probe rather than as a nonspecific binding.
The nearly reciprocal movement of the two probes is highlighted in Fig. 1 F, which is from a simultaneous two-color fluorescence measurement in one cell. For this experiment, the PH domain is fused to a monomeric red fluorescent protein (mRFP), and the C1 domain probe is fused to green EGFP. As the red fluorescent protein was more subject to photobleaching with our illumination settings, it shows a superimposed gradual decline over the 400-s period. In a series of two-color experiments with these probes, the exponential time constants for initial translocation and the half times for fluorescence recovery for oxo-M are τtrans = 16 ± 2 s and t1/2 rec = 36 ± 6 s for PH-mRFP (n = 5) and τtrans = 19 ± 1 s and t1/2 rec = 87 ± 7 s for C1-EGFP (n = 5) (Fig. 1 G). The initial translocation time is similar, but recovery of the C1-EGFP probe is significantly slower than that of the PH-mRFP probe. However, there is no a priori expectation that the recoveries would be similar since the metabolic pathways that dephosphorylate IP3, synthesize more PIP2, and phosphorylate DAG are different and independent.
The Inhibitor U73122 Has Complex Actions
U73122 is a PLC inhibitor that is widely used as a quick test for the involvement of PLC in a signaling pathway. In our experiments, U73122 strongly uncoupled activation of PLC by receptors as expected. Fig. 2 (A and B) shows confocal experiments with U73122 and its relatively inactive analogue U73343 using the DAG probe (C1-EGFP). Again, oxo-M is applied twice for 3 min. The DAG probe translocates to the membrane during the first oxo-M application, but not after a 5-min application of U73122 (2.5 μM; Fig. 2 A; n = 10). The control compound U73343 seems inactive as it does not prevent the translocation (Fig. 2 B; n = 3). These robust effects seem readily interpreted as a simple inhibition of M1 receptor–stimulated PLC activation and DAG production. In these experiments, U73122 by itself tends to augment the cytoplasmic fluorescence of the DAG probe in resting cells as if it has released some of the probe from the membrane (Fig. 2 A).
Similar experiments with the IP3/PIP2 probe (PH-EGFP) suggest a more complicated picture (Fig. 2, C and D). Again, U73122 fully blocks the translocation induced by oxo-M; however, even without oxo-M, U73122 initiates a slow translocation of the probe (Fig. 2 C; n = 15 cells). This translocation is 30 times slower than that with oxo-M. As expected, the control compound U73343 does not block oxo-M action (Fig. 2 D; n = 7), although in some experiments at high concentration, U73343 by itself seemed to induce a slight translocation. Translocation occurs with U73122 alone even when the extracellular solution includes no calcium (unpublished data; n = 7). These observations suggest the working hypothesis that U73122 efficiently blocks activation by the Gq pathway, as is widely reported, but it is also a very weak agonist of PLC. Fig. 3 A examines the time course of these effects. U73122 (3 μM) was applied for 0, 0.5, 1, 2, 3, or 5 min. According to the PIP2/IP3 probe, receptor-induced activation of PLC is half blocked after just 0.5 min of 3 μM U73122, whereas the U73122-induced translocation of probe (closed symbol) takes significantly longer, and is half induced after 3 min of U73122.
Calcium photometry measurements are consistent with our working hypothesis (Fig. 2, E and F). A first application of oxo-M raises [Ca2+]i strongly. The initial calcium rise comes from intracellular stores, presumably by IP3-dependent release, but the later plateau requires extracellular calcium (unpublished data; n = 5 cells) and probably involves calcium entry that depends on store emptying. Induction of [Ca2+]i elevations by agonist is fully blocked after prior exposure to U73122 (Fig. 2 E; n = 5) but not after the inactive analogue U73343 (Fig. 2 F; n = 4). Nevertheless, U73122 alone raises [Ca2+]i a little, albeit very slowly (Fig. 2 E; n = 6).
Experiments with KCNQ2/3 current give a similar picture (Fig. 2, G and H). Fig. 2 G shows the expected suppression of current by oxo-M (measured at −20 mV) and its prevention by U73122. The initial control application of oxo-M for 30 s strongly suppresses current as PLC is activated and PIP2 is depleted (Suh and Hille, 2002; Suh et al., 2004). The suppression is reversible. The time course of this current modulation is reminiscent of the time course of PIP2 hydrolysis reported by the PH-EGFP probe (Fig. 2 C), although the recovery of current is consistently a little slower than the recovery of PH-EGFP distribution. Next, when U73122 is applied for 2 min, the current starts to increase rapidly, an unexpected result, and then to decline steadily below the initial level. Subsequent application of oxo-M is ineffectual, indicating again that M1 receptor coupling to PLC has been powerfully blocked (Fig. 2 G, n = 5). With a 5-min application of U73122, the initial augmentation is similar but the current declines considerably during the exposure to the inhibitor (Fig. 2 H, n = 12). Fig. 3 B examines the time course of these effects. U73122 (3 μM) was applied for 1, 2, 3, 4, or 5 min. The block of receptor coupling (open symbols) is almost complete after a 1-min exposure to U73122, and the U73122-induced depression of current (closed symbols) is still only partial after 3 min of U73122.
Why does the current increase and then decrease during U73122? We begin with the transient augmentation, which Stemkowski et al. (2002) have also described for endogenous M current in frog sympathetic neurons. We asked if it could be due to a changed voltage dependence of activation by measuring the amplitudes of tail currents at −70 mV after 500-ms depolarizations to different test potentials. Indeed, this protocol revealed that U73122 shifts the voltage dependence of KCNQ2/3 channel activation to more negative potentials by 14 ± 3 mV (n = 4; see Fig. S1 A). Such a shift would allow more channels to open at −20 mV since steady-state activation at that voltage is normally incomplete (only 52%) and would be increased to 80% by the drug (Fig. S1 A). The subsequent decline of current is consistent with the hypothesis that U73122 acts as a very weak agonist of PLC, resulting in a slow decline in the membrane PIP2 and hence of the current. Again, the rate of decline is ∼30 times slower than the rapid suppression by oxo-M, suggesting that PLC is turned on to only 3–4% of the maximum level achieved with receptor activation.
The actions of U73122 were sufficiently complex that we wanted to test another inhibitor of PLC. We tried the lipid-like, anti-neoplastic agent edelfosine (ET-18-OCH3). As is shown in Fig. 4 (A and B), this compound (10 μM for 30 min) entirely prevented the muscarinic translocation of the PH-EGFP probe and the muscarinic inhibition of KCNQ currents (at 15 min the action of edelfosine was strong but incomplete). Furthermore, by itself it did not induce the rise and inhibition of KCNQ current or the slow migration of the PH-EGFP probe seen with U73122. These actions of edelfosine are consistent with simple inhibition of PLC, and we interpret the additional effects of U73122 as side effects with other mechanisms.
Some Actions of U73122 May Represent Alkylation
The literature on U73122 contains numerous suggestions of side effects that could not be attributed to simple block of PLC (see DISCUSSION). We decided to investigate the initial augmentation of KCNQ current with U73122 as such a side effect. The only structural difference between U73122 and U73343 is one double bond (Fig. 2, A and B, insets) that makes U73122 chemically reactive. U73122 is an N-substituted maleimide, which like N-ethylmaleimide (NEM) should be able to alkylate cysteine −SH groups, tethering the entire U73122 compound covalently to the cysteine. U73343 is the corresponding unreactive succinimide. Both the augmentation of KCNQ current and the negative shift of activation with U73122 reminded us of augmentations and shifts induced by NEM acting on cysteine residues of KCNQ channels (Roche et al., 2002; Li et al., 2004). We verified that the shifts are quite similar (Fig. S1 B). Therefore we sought to test the hypothesis that U73122 can act as a hydrophobic- or membrane-targeted version of NEM, capable of reacting with −SH groups generally.
Fig. 5 A shows powerful augmentation of KCNQ2/3 current by 120 s of 100 μM NEM, as previously reported (Roche et al., 2002). The augmentation is consistently larger than that with U73122. Somewhat like U73122, NEM also slows and decreases the subsequent response to oxo-M; however, this effect is only partial. Fig. 5 B shows that prior NEM treatment prevents U73122 from augmenting KCNQ current and also from subsequently decreasing it, but prior NEM does not prevent U73122 from fully uncoupling receptor activation from PLC. Similarly, prior U73122 treatment prevents NEM for augmenting KCNQ current (Fig. 5 C). The effects summarized in Fig. 5 D suggest that both the augmentation and the subsequent decline of current with U73122 are side effects dependent on alkylating accessible −SH groups. As expected for such reactions, all actions of NEM and of U73122 were irreversible on the time scale of our experiments.
As a further test of the alkylation hypothesis, we asked if U73122 mimics another effect of NEM on signaling to ion channels. Previously we have shown that NEM readily blocks electrophysiological responses due to pertussis toxin–sensitive G proteins (Gi/Go), presumably by alkylating the same G protein cysteine residue that pertussis toxin ADP ribosylates (Shapiro et al., 1994). One such response is the pertussis toxin–sensitive opening of inward rectifier GIRK channels by muscarinic agonists acting on M2 receptors (Pfaffinger et al., 1985). Fig. 5 E shows that oxo-M acting via endogenous M4 receptors activates this inward current in AtT-20 cells, and that NEM completely uncouples the muscarinic effect as was expected from block of Gi/Go. Similarly, U73122 uncouples the M4 muscarinic action (Fig. 5 F; see summary Fig. 5 G) even though PLC is not involved. Again, U73122 mimics an effect of NEM, this time on pertussis toxin–sensitive G proteins. Treatment with U73122 also reduces the holding current somewhat.
Our last example of occlusion of U73122 actions by NEM returns to the translocation probes. Prior treatment with NEM eliminates the small augmentation of cytoplasmic fluorescence by U73122 seen with C1-EGFP (Fig. 5 H; n = 4; compare with Fig. 2 A) and the large slow augmentation seen with PH-EGFP (Fig. 5 I; n = 7; compare with Fig. 2 C). Thus by many criteria, U73122 and NEM mimic and occlude each others' actions. However, NEM did not occlude one action of U73122; namely, it did not prevent U73122 from fully uncoupling muscarinic activation of PLC as judged by the translocation assays (n = 11). In addition, by itself NEM did not induce translocation of either probe, even with a 5-min treatment (unpublished data).
Phosphoinositides Are Hydrolyzed
We undertook an analysis of acidic phospholipids of CHO-M1 cells during treatments with carbachol, U73122, and NEM. These cells express M1 receptors stably. Using deacylated headgroups, our method measures the relative abundance of six acidic lipids: phosphatidylinositol (PI), phosphatidylinositol 4-phosphate (PIP), PIP2, phosphatidylserine, cardiolipin, and phosphatidic acid (Nasuhoglu et al., 2002a). Fig. 6 reports PI, PIP, and PIP2 as percent of the total of the six acidic lipid classes measured. In unstimulated cells, PI is almost 50% of the total and PIP and PIP2 are only ∼1% each, truly trace phospholipids. A 1-min carbachol treatment strongly reduces PIP2 and PIP (to 7% and 12% of control, respectively) and slightly but significantly reduces PI (to 95% of control), consistent with a powerful muscarinic activation of PLCβ depleting the major cellular pools of PIP2. A 5-min treatment with 5 μM U73122 alone also reduces PIP2 and PIP, although not as much as carbachol, and it renders subsequent treatment with carbachol ineffectual. These results would be consistent with a weak activation of PLC by U73122, and a total block of carbachol action. Alternative interpretations are considered in DISCUSSION. NEM did not change the lipid composition in 2 min, but in further measurements not shown, it did reduce PIP2 and PIP by 35 and 50%, respectively, in 5 min. Also we mention that both PIP and PIP2 levels consistently rebounded ∼50% back toward control levels after 10 min in the continued presence of agonist.
Since the stable CHO-M1 cells used for chemical assays were different from the tsA cells used in the rest of our experiments, we checked the oxo-M–induced modulation of KCNQ2/3 current and the translocation of the PH-EGFP probe in them. The speed and amplitude of the muscarinic current modulation and of the probe migration were virtually identical to those in tsA cells, where the M1 receptor had been introduced by transient transfection (n = 5–8). We conclude that the quantitative parameters of KCNQ2/3 current inhibition and PI metabolism are indistinguishable in the two cell lines.
m-3M3FBS Activates PLC with a Delay
In search of other practical ways to activate PLC, we then studied the putative PLC activator m-3M3FBS and its control compound o-3M3FBS (Bae et al., 2003). We found that the meta compound causes migration of the PH-EGFP and C1-EGFP probes (Fig. 7, A and B), and it induces suppression of the KCNQ2/3 current (Fig. 7 C). By these criteria, it activates PLC. Judging from the relative steepness of the traces, the apparent activation by m-3M3FBS is relatively strong once it starts but is still a little weaker than for the maximal activation by oxo-M. The action begins abruptly after a variable delay of several hundred seconds and sometimes it reverses spontaneously for 200 s, only to start up again. The control compound o-3M3FBS induces no migration of the probes and no suppression of current beyond the usual rundown that would be seen in control conditions (Fig. 7, D–F). In our previous work, we have shown that coexpression of the Gq/11-specific regulator of G protein signaling RGS-2 blocks oxo-M–induced migration of PH-EGFP and suppression of KCNQ2/3 current in these cells (Suh et al., 2004). This is evidence that the normal muscarinic action via M1 receptors requires participation of the G protein Gq/11. Coexpression of RGS-2 does not block the activation of PLC by m-3M3FBS (Fig. 7, G–I). Hence the activator must work downstream of muscarinic receptors and Gq.
PLC Needs Calcium
We explored the [Ca2+]i dependence of the muscarinic activation of PLC by buffering cellular [Ca2+]i to different levels. The principal questions were whether PLC in live cells can be activated when Ca2+ is well below normal, whether the IP3-induced release of Ca2+ from intracellular stores exerts positive feedback to accelerate PLC, and whether modulation of KCNQ current has the same calcium dependence.
The experiments in Fig. 8, A, C, and E, each include three different intracellular conditions. In the top panels, [Ca2+]i is relatively unbuffered (control). In the middle panels, [Ca2+]i is strongly buffered under conditions intended to mimic and stabilize resting levels. In the lower panels, [Ca2+]i is strongly buffered using conditions intended to clamp it well below normal resting levels. In the photometry and confocal experiments (Fig. 8, A–D), which used intact cells, the Ca2+ buffering was achieved by preloading membrane-permeant BAPTA-AM and by pretreating with thapsigargin to allow any intracellular stores to empty. The [Ca2+]i level was manipulated by doing these incubations and the subsequent measurements either in 2 mM Ca2+–containing or Ca2+-free solutions (with EGTA). As expected, the [Ca2+]i measurements (Fig. 8, A and B) showed that [Ca2+]i rises briskly after the muscarinic stimulus when it is unbuffered, and it does not rise significantly when it is strongly buffered at either normal or subnormal levels. The confocal experiment (Fig. 8, C and D) showed that, relative to controls, the oxo-M–induced translocation of PH-EGFP still occurs but is slowed and reduced in final amplitude when [Ca2+]i is buffered at resting levels. Prior emptying of intracellular Ca2+ stores with thapsigargin without buffering internal Ca2+ with BAPTA/AM reduces subsequent oxo-M–induced translocation of PH-EGFP to a similar extent (Fig. 8 D). Translocation is nearly absent when [Ca2+]i is held at reduced levels (Fig. 8 C, bottom). Together, these experiments show that muscarinic activation of PLC requires some free Ca2+. PLC can be activated at resting [Ca2+]i levels, but activation is potentiated considerably (positive feedback) when IP3 is allowed to elevate [Ca2+]i above the resting level. This positive feedback occurs in <5 s.
Finally, we measured modulation of KCNQ2/3 current under these three calcium conditions. Since the electrophysiology experiments were done with whole-cell clamp, the Ca2+ buffers were introduced via the recording pipette (see MATERIALS AND METHODS and legend). The current measurements (Fig. 8, E and F) show that KCNQ2/3 current is quickly and completely suppressed after a 20-s exposure to oxo-M if [Ca2+]i is left unbuffered (τ = 7 ± 0.3 s, n = 5; inhibition = 98 ± 2%); it is more slowly and less completely suppressed if [Ca2+]i is kept at resting levels (τ = 14.3 ± 1.6 s, n = 6; inhibition = 75 ± 4%); and it is unaffected by oxo-M if [Ca2+]i is kept below normal (inhibition = 4 ± 2%, n = 5). Thus the calcium dependence of signaling to KCNQ current parallels that of PLC activation.
Does Ca2+ entry play a role in regulating PLC as well? An enhancement of the [Ca2+]i rise and of PLC activity by extracellular calcium is evident in experiments with longer exposures to oxo-M (Fig. 9). Consider the time course of [Ca2+]i during 180-s oxo-M exposures as extracellular Ca2+ is altered. Fig. 9 (A–C) shows [Ca2+]i responses to two successive oxo-M applications. In the control condition, oxo-M evokes a transient elevation of [Ca2+]i >1 μM that decays to a plateau of several hundred nanomolar until oxo-M is removed. The second application evokes a similar, often slightly less robust, response with the standard bathing medium (Fig. 9 A; n = 15) but a smaller transient and no plateau if done in Ca2+-free medium (Fig. 9 B; n = 5). If the medium Ca2+ is restored partway through the oxo-M exposure, the [Ca2+]i rises up again (Fig. 9 C; n = 5). These observations are the signatures of a late, store-operated Ca2+ entry. Apparently oxo-M induces an initial transient release of intracellular Ca2+ that in tsA cells is followed by opening of store-operated channels on the plasma membrane. These plasma membrane channels maintain the [Ca2+]i plateau, but shut quickly after agonist is removed. Fig. 9 (D–F) shows translocation of the IP3/PIP2 probe, PH-EGFP, in parallel experiments. The records resemble those of the [Ca2+]i transients. Namely, the translocation is decidedly attenuated and decays more rapidly when the bath has Ca2+-free solution (Fig. 9 F; n = 4), and it is augmented if Ca2+ is restored while oxo-M is still being applied (Fig. 9 E; n = 4). Thus, the activity of PLC is significantly increased not only by IP3-evoked Ca2+ release but also by the subsequent store-operated Ca2+ entry.
We used one more approach to test the calcium dependence of muscarinically activated PLC. By overexpression of the enzyme, type I IP3-5-phosphatase (IP3-5-P), which catalyzes the first step in IP3 degradation, we could abort the PLC-generated rise of IP3 concentration and consequently the IP3-induced elevation of [Ca2+]i. Fig. 10 (A and B) shows that expression of IP3-5-P effectively eliminates the muscarinic [Ca2+]i transient, presumably because IP3 is degraded as soon as it is made. Further, Fig. 10 (C–F) shows that expression of IP3-5-P slows and decreases the muscarinic translocation of the DAG probe, C1-EGFP, and the PIP2/IP3 probe, PH-EGFP. These two measures report that the PLC response is weakened but not abolished when IP3 buildup and Ca2+ release are prevented. They suggest that PIP2 is only partially depleted by 3 min of oxo-M when IP3-5-P prevents the positive feedback by Ca2+. With the PH-EGFP probe there is almost no translocation when IP3-5-P is expressed. Presumably, the combination of some unhydrolyzed PIP2 remaining in the plasma membrane and no IP3 accumulating in the cytoplasm means that the PH-EGFP probe remains membrane bound. The final experiment measures muscarinic suppression of KCNQ2/3 current for comparison with the observations on PLC. The suppression is slow and only partial in the IP3-5-P cells, as expected if PLC is slowed, PIP2 is cleaved more slowly, and much PIP2 remains.
Morphological Effects of our Reagents
We report here briefly several consistent additional effects of our reagents on the morphology of the tsA cells. When oxo-M was perfused onto cells, successive 5-s images show considerable agitation in the cell surface region. The membrane appears to be bubbling out in blebs and small extensions. Presumably this motion reflects local breakdown of the actin cytoskeleton and perhaps an arrest of endocytosis as plasma membrane PIP2 becomes depleted (Rohrbough and Broadie, 2005). A second set of observations is that application of U73122 slowly causes cells to round up irreversibly and occasionally leads to an abrupt enhancement of nuclear envelope permeability indicated by a sudden diffusion of the optical probes into the nucleus (<5 s); NEM induces blebbing; and edelfosine slowly induces cell swelling (although KCNQ current remains) and eventual lysis. NEM also strongly blocks diffusional exchange of probes between the nucleus and cytoplasm as if it closes nuclear pores.
Discussion
Comparison with Previous Observations on PIP2 and PLC
We have used three functional assays and one biochemical assay to monitor the effects of PLC on phosphoinositide metabolism. The concordance of the results obtained with all methods validates their use and strengthens our conclusions. All approaches agreed that activation of endogenous PLC, presumably mostly the β-isoform(s), by M1 receptors expressed in tsA and CHO-M1 cells strongly depletes both the functional plasma membrane PIP2 and the total cellular PIP2 in less than 1 min, producing IP3-dependent Ca2+ release and DAG. According to the translocation probes, the majority of free cellular PIP2 is confined to the plasma membrane, and DAG, which is initially not abundant, is produced primarily in the plasma membrane. The rapid and powerful depletion of PIP2 and considerable depletion of PIP agree well with biochemical measurements in the SH-SY5Y human neuroblastoma cell line (Willars et al., 1998) and contrast with observations in heart and cell cultures in which Gq/PLC-coupled receptors may be less prevalent (Nasuhoglu et al., 2002b). The results show that the cells we used have a large endogenous reserve of PLC activity that can couple to introduced receptors. We find that KCNQ2/3 channel activity closely follows the overall changes of PIP2 in response to a wide range of treatments. KCNQ current is therefore a sensitive and useful indicator of PIP2 that appears to report global PIP2 levels rather than a private “local pool” in these experiments.
We have published a kinetic model containing tentative estimates of the time course of phosphoinositide turnover following M1 receptor activation (Suh et al., 2004). The model considered plasma membrane PI, PIP, and PIP2 and four principal enzymes of phosphoinositide metabolism: a PI-4-kinase to make PIP from PI, a PI(4)P-5-kinase to make PIP2 from PIP, a PI(4,5)P2-5-phosphatase to reverse the former step, and a PIP2-specific PLC that hydrolyzes PIP2. Calculations with that model predict severe depletion of PIP2, with only 0.66% remaining for a 60-s exposure to 10 μM oxo-M, and moderate depletion of PIP, with 40% remaining. On the other hand, using a chemical assay, we now find that 7% of total cellular PIP2 and only 12% of the PIP remain under these conditions (Fig. 6). How can these differences be reconciled? We start with PIP2. Some fraction of the total cellular PIP2 may not be readily accessible to the PLC that is activated by M1 receptors and Gq because this fraction resides on membranes other than the plasma membrane or is very tightly sequestered or segregated from PLC. For example, PIP2-binding proteins and lipid rafts might be able to remove PIP2 from the accessible pool (Yin and Janmey, 2003) such that this PIP2 fraction is not hydrolyzed in the first minute. Nevertheless, in terms of this concept of distinct pools of lipids we have to conclude that at least 93% of the total cellular PIP2 and 88% of the PIP form large accessible pools at the plasma membrane, suggesting that any other functional pools that may exist are small. The large pool of PIP2 is reported by KCNQ current and by translocation of PH domain probes.
What about the greater-than-predicted loss of PIP, also reported by Willars et al. (1998)? In our model, a small depletion of PIP arises because the combination of PI(4)P-5-kinase and PI(4,5)P2-5-phosphatase acts in a cycle to make PIP partially dependent on PIP2 by feedback via PI(4,5)P2-5-phosphatase. The decline of PIP could be speeded and intensified in the model by speeding both enzymes of this cycle, making PIP more tightly coupled to PIP2. However, another more likely possibility is that PLCβ can hydrolyze PIP in addition to PIP2, contrary to many current descriptions. Early in vitro biochemical experiments with vesicles containing PI, PIP, or PIP2 led Wilson et al. (1984) and Ryu et al. (1987a) to propose that PLC hydrolyzes PI, PIP, and PIP2 in a Ca2+-dependent manner. If we neglect complications of separate pools, our model could deplete the PIP to a steady state of 12% (as observed) by assuming that PLC hydrolyzes PIP with a rate constant that is 1/7 of that for PIP2. The predicted depletion is relatively fast with a time constant of 14 s. (See APPENDIX for further analysis of the Suh et al., 2004, model.) If, however, we introduce an inaccessible pool of PIP, then the rate constant of PIP hydrolysis would have to be made more similar to that for PIP2.
Comparison with Previous Observations on Calcium and PLC
Our functional assays concur in finding that in the living cell, the activation of PLCβ by M1 receptors is slowed if [Ca2+]i is not permitted to rise above resting levels and is stopped if [Ca2+]i is kept well below resting levels. The activation is greatest if an IP3-dependent Ca2+ release occurs that is allowed to raise the [Ca2+]i to micromolar levels. Store-operated entry of Ca2+ from the outside also can contribute to maintaining PLC activity, but stops as soon as the receptor agonist is removed. The positive feedback by calcium is significant. When it is permitted, the activation of PLC may become 10–15-fold greater than when it is not permitted, according to our model. These properties make PLCβ a coincidence detector that responds strongly to simultaneous activation of receptors and rise of Ca2+ (Hashimotodani et al., 2005). For example, a postsynaptic membrane with colocalized Ca2+-permeable ionotropic glutamate receptors and Gq-coupled metabotropic glutamate receptors might activate PLC powerfully.
This calcium dependence agrees well with the reported in vitro calcium dependence curve for PLCβ (Ryu et al., 1987b; Smrcka et al., 1991; Biddlecome et al., 1996). It also agrees qualitatively with previous experiments reporting an arrest of receptor-stimulated phosphoinositide turnover when digitonin- or electro-permeabilized cells are exposed to EGTA, and an acceleration when exposed to solutions with elevated calcium (Fisher et al., 1989; Wojcikiewicz et al., 1994). The reports known to us indicate that the phosphoinositide PLC enzyme activity falls virtually to zero in the absence of Ca2+, and the rates of hydrolysis by Gq-activated PLCβ are strongly sensitive to Ca2+ in the region around the 100 nM typical resting concentration. The experiments with permeabilized cells agree that calcium concentrations achieved during cell signaling do not activate the β isoforms of PLC by themselves. However, calcium is a requirement for PLCβ activation by Gq. It remains to be established whether PLCs besides PLCβ become involved in the response amplification by Ca2+. The crystal structure of mammalian PLCδ has an essential Ca2+ ion in the active site that can coordinate with one of the phosphate groups of the substrate (Essen et al., 1996), and that isoform is an excellent candidate for direct activation by physiological Ca2+ signals, in particular by nearby plasma membrane Ca2+ channels (Kim et al., 1999).
Optical Probes as Indicators of PI Metabolism
The behavior of the PH-EGFP and C1-EGFP probes in our study fits well with their original description as probes of PIP2/IP3 and of DAG (Oancea et al., 1998; Stauffer et al., 1998; Várnai and Balla, 1998; Hirose et al., 1999). Both probes clearly have access to the nucleus, and they undergo a slow, NEM-sensitive exchange between nucleus and cytoplasm when the probe concentration differs between these two compartments. This exchange seems symmetrical and passive. It adds transient components to the time course of cytoplasmic probe, which so far as we know have not been recognized previously. The transients fit a simple three-compartment, passive model. The diffusion of unfused EGFP through nuclear pores of COS-7 cells has been studied by fluorescence recovery after photobleaching. In those cells, diffusional exchange of EGFP (27 kD) is characterized by a time constant and nuclear membrane permeability of τ = 164 s and Pnuc = 0.011 μm s−1 (Wei et al., 2003). For the larger C1-EGFP probe (34 kD) we found τ = 188 s and Pnuc = 0.005 μm s−1, and for PH-EGFP (41 kD), τ = 281 s and Pnuc = 0.0032 μm s−1 (see Fig. 11 and calculations in APPENDIX and computer program in online supplemental material). These diffusion properties fit well with the concept of lower permeability for a larger molecule, giving the expected permeability sequence EGFP > C1-EGFP > PH-EGFP.
There has been some argument whether PH-EGFP might be viewed as a pure PIP2 probe or as a pure IP3 probe, despite being able to bind to both compounds. One line of argument used to support a pure “IP3 hypothesis” was that the measured affinity for IP3 is 10–20 times higher than for PIP2 (Lemmon et al., 1995; Hirose et al., 1999). In our discussion of the kinetic model of Suh et al. (2004) in APPENDIX, we conclude that our results and other published work with the PH domain probe in cells are not compatible with the large difference in affinity measured by binding studies on purified components. A second argument for this hypothesis was that IP3 will displace PH domains from PIP2-containing cell membranes (Hirose et al., 1999), but in all cases the amount of IP3 needed exceeds that expected in cells (see APPENDIX). A third argument was that expression of the enzyme IP3-5-P can stop migration of the probe despite activation of PLC (Okubo et al., 2004), as we also found. It was presumed that the only change with IP3-5-P expression was that IP3 did not accumulate. However, we have now shown that this maneuver stops calcium feedback to PLC and in our case slowed it so much that PIP2 hydrolysis was no longer complete. Thus we suggest that experiments with IP3-5-P do not settle the argument. We prefer the viewpoint that the PH probe responds to both the PIP2 precursor and the IP3 product, and during an experiment their relative importance changes as their relative concentrations change. This approach is suggested by the results of Xu et al. (2003) and Winks et al. (2005).
Does m-3M3FBS Activate PLC?
We had hoped to develop other practical ways to activate PLC for our experiments. Agonist properties of m-3M3FBS had been discovered in a large chemical screen against responses of neutrophils (Bae et al., 2003). It was reported to enhance superoxide production, increase intracellular Ca2+, promote IP3 production, and to activate purified PLCs of the β, γ, and δ classes directly in vitro. The effects were not sensitive to pertussis toxin or to expression of RGS2. A later work found a Ca2+ rise but could not confirm activation of PLC either chemically or with the PH-EGFP translocation probe (Krjukova et al., 2004). In our work, we have seen all the effects expected for an activator of PLC (translocation of PH-EGFP, translocation of C1-EGFP, and decrease of KCNQ2/3 current) and conclude that 50 μM m-3M3FBS does activate PLC relatively strongly. However, it is slow to act, has a variable long latency, and a fluctuating response. The ebb and flow of translocation probes and current with m-3M3FBS suggest that conflicting signal systems are activated and the balance sometimes favors turning on PLC and sometimes, turning off. In the end, PLC turns on in our conditions. Perhaps in other cells and under different application conditions the putative turning-off process might win. Variability of response, photolability, fluorescence, and spectral overlap with the common Ca2+ indicator fura-2 (Jansen et al., 2004) could limit the applicability of this compound in experiments with intact cells.
U73122 and Alkylation
PubMed lists over 1,500 papers with U73122 in the abstract. The large majority use the inhibitor simply as a tool to check that some signaling pathway requires PLC. However, numerous papers report additional unexpected effects, raising question whether this agent can be used as a pharmacological tool without serious side effects. We select results from just four early papers. The initial brief announcement of U73122 from Upjohn reports that it inhibits partially purified PLC in vitro when the molar ratio of Ca2+:PI in the assay was <2, but “increased PLC activity” when the molar ratio was 4–12 (Bleasdale et al., 1989). There are no data or experimental details in that book chapter. A careful study in NG108-15 neuroblastoma-glioma cells and in dorsal root ganglion cells shows that U73122 blocks bradykinin-induced Ca2+ transients irreversibly with a steep dose–response curve and a half-effective dose IC50 of 200 nM for 20-min preincubations and that U73343 is without effect (Jin et al., 1994). However, these authors also find a slow rise of intracellular Ca2+ in Ca2+-free bath solutions and a total block of L-type Ca2+ channels. They mention the possibility of protein alkylation for such side effects. A study of the amount and isotopic labeling of phosphoinositide pools in platelets found 50% reduction of PIP and PIP2 with 15 μM U73122 and a reduction in the rate of incorporation of phosphate and glycerol; it proposes that U73122 may inhibit PI and PIP kinases and possibly also stimulate lipid phosphatases (Vickers, 1993). A patch-clamp study with pancreatic acinar cells shows that 10 μM U73122 can open plasma membrane ion channels and it can promote release of Ca2+ from intracellular stores (Mogami et al., 1997). They conclude that U73122 is not specific for PLC and mention the possibility of alkylation.
For U73122, our observations agree with strong inhibition of PLC, with a lack of specificity, and with some of the reported side effects. We find that a few micromolar U73122 blocks M1 receptor coupling to PLC virtually completely in 1–2 min, confirming that the compound is an efficacious inhibitor of G protein–coupled PLC activation. At the same time, in a selected subset of our observations, U73122 initiates a slow decline of KCNQ2/3 current, a slow translocation of the PH-EGFP probe, a slow rise of intracellular Ca2+, and in 5 min, a decrease of the PIP and PIP2 pools of the cell. In RESULTS we presented the hypothesis that U73122 activates PLC to ∼3% of its maximum activity while uncoupling signaling from Gq-coupled receptors. This hypothesis is consistent with the subset of results mentioned so far, but it does not explain for example why U73122 fails to induce a slow translocation of the C1-EGFP probe to the membrane. We can consider several alternative hypotheses. For example, inhibition of lipid kinases by U73122 (Vickers, 1993) would explain the translocation results, the decline of KCNQ2/3 current, and the fall of PIP and PIP2 pools accompanied by a rise of PI; however, it does not explain the slow rise in Ca2+. An activation of lipid phosphatases might look similar. Finally, the highly lipid-soluble cation U73122 might sequester anionic PIP2 much as the lipid-soluble local anesthetic benzocaine has been proposed to inhibit PIP2-sensitive GIRK K+ channels by sequestering PIP2 (Zhou et al., 2001). Sequestration of PIP2 could lead to a decline of KCNQ2/3 current and a translocation of the PH-EGFP probe. Hydrophobic interactions might even release some C1-EGFP from the membrane if DAG is also sequestered by U73122. To speculate further, IP3 receptors of the endoplasmic reticulum may be held in check by contact with PIP2 on the plasma membrane (Lupu et al., 1998; Glouchankova et al., 2000), and if the available PIP2 is reduced by sequestration, IP3 receptors could become active, releasing Ca2+ from internal stores. Although it will not be anchored to proteins in the membrane by alkylation, the control compound U73343 is a hydrophobic cation as well and should share some of these effects, as it does weakly. The sequestration hypothesis does not explain why PIP and PIP2 pool sizes fall with U73122. We feel that U73122 is a complex agent with multiple molecular actions, so it is not required that a single mechanism explain all the results. Perhaps all the mechanisms mentioned in this paragraph play some part in the overall effect.
Now we consider alkylation. U73122 and its control compound differ in that the active compound can alkylate and the control compound cannot. Our experiments comparing the two reactive maleimide compounds NEM and U73122 convince us that alkylation explains many side effects. There is a mutual mimicry or occlusion of effects. Either compound by itself quickly increases KCNQ2/3 currents, shifts their voltage dependence, and blocks Gi/o-coupled signaling to GIRK channels. Prior reaction with NEM prevents U73122 from either rapidly increasing KCNQ2/3 current or slowly decreasing it, and from inducing translocation of the PH-EGFP probe. Prior reaction with U73122 prevents NEM from augmenting KCNQ2/3 current. U73122 acts at 20-fold lower concentration than NEM, as might be expected for a maleimide with a lipophilic tether. When the reaction occurs, the entire aminosterol molecule (Fig. 2 A) will be coupled to the protein target, e.g., to the KCNQ channel subunit, or to Gαo, etc. NEM may mimic U73122 in another respect. Longer exposures to NEM depress the pool of cellular PIP2 and PIP, which might reflect inhibition of lipid kinases. For example, NEM and other sulfhydryl reagents block the stimulatory effects of ATP on ion transport activity that are now known to represent lipid kinase activity (Hilgemann and Collins, 1992).
A question we lack evidence about is whether U73122 alkylates PLC to inhibit it. We suggest that the answer is yes, since block is irreversible and the control compound U73343 is not able to alkylate. However, although NEM slows receptor-coupled PLC (Fig. 5, A and D) it definitely does not stop it (in 2 min), and prior NEM does not prevent subsequent U73122 from fully arresting receptor-coupled PLC (Fig. 5 B). Thus 2.5–5 μM U73122 must react quickly with at least one cysteine of PLC that 100 μM NEM reacts with much more slowly or not at all. Perhaps the sterol moiety of U73122 guides it to this target.
The dose–response relation for U73122 action is often steep and seemingly variable. Several factors probably contribute to the variability: the compound is difficult to dissolve and needs vigorous vortexing in a hydrophobic vehicle; it precipitates from the stock solution in the refrigerator; maleimides are unstable once placed in water; and finally alkylation is irreversible, so there is no equilibrium binding. The reaction rate is likely to be directly proportional to the inhibitor concentration and fairly temperature dependent. Thus 20 min at 200 nM (Jin et al., 1994) is likely to be similar to 2 min at 2 μM, provided the temperature is the same. In brief, many factors make U73122 a difficult compound to work with. To avoid excessive side reactions, the product of concentration and duration of treatment should be kept as low as possible. In our hands edelfosine (ET-18-OCH3) seemed a cleaner reagent although it was clearly cytotoxic and slow to act and may well have its own side effects.
Conclusions
We draw several conclusions. First, under a variety of new test conditions, the size of KCNQ current continues to be a consistent reporter of the cellular PIP2 pool. Second, the strict Ca2+ dependence of receptor-activated PLCβ within the living cell subjects the enzyme both to strong positive feedback via its own IP3 pathway and to strong enhancement by nearby coincident Ca2+ signaling in other systems. Third, the pharmacological tools for inhibiting and activating PLC must be used with caution. Fourth, the PIP2 levels in cells transfected with M1 muscarinic receptors undergo rapid and extensive depletion during receptor activation, whether monitored indirectly via KCNQ channel activity, fluorescent indicators of PIP2 and DAG, or directly via total phosphoinositide measurements. Clearly, in the cell lines we used, receptor-activated PLCs have rapid access to the major PIP2 pool of the cell. Work in other systems has suggested that the PIP2 used in IP3 signaling may instead be a small and highly compartmentalized pool of total cell PIP2, a conclusion reached for example by using RNAi-mediated knock-down of individual lipid kinases (Wang et al., 2004). It remains to be seen how such contrasting impressions of PIP2 signaling will be resolved.
Appendix
Diffusion of Probes into the Nucleus
An Elementary Model for Nuclear–Cytoplasmic Exchange of Probes
We observed that the translocation probes enter the nucleus slowly and recognized that this slow exchange would add transients to the measured values of cytoplasmic probe concentration. An elementary kinetic model was formulated to simulate the translocation of probes between plasma membrane, cytoplasm, and nucleus. The goal was to ask how adding the nuclear compartment alters the expected time course of cytoplasmic probe concentration changes.
In rough accord with images in Fig. 1 A, the cell was taken as a sphere with radius rC = 6 μm, and the nucleus, as a sphere with radius rN = 4.4 μm. This makes the area of the nuclear membrane AN = 273 μm2 and the volumes of the cytoplasm and nucleus, VC = 0.55 pl and VN = 0.36 pl, respectively. Flux FPr (units, mol s−1) of probe from cytoplasm to nucleus was assumed to obey the standard passive permeability equation (Hille, 2001) FPr = PNAN(PrC − PrN)*10−15, where PN is the nuclear membrane permeability to probe (units, μm s−1), PrC and PrN are free cytoplasmic and nuclear probe concentrations (units, M), and 10−15 converts from liters to cubic micrometers. The rate of change of nuclear probe concentration would be dPrN/dt = FPr/VN and similarly for the cytoplasmic concentration.
To simulate reversible release and binding of probe to a membrane ligand (e.g., to free PIP2 or DAG lipid, Lf), the probe and lipid were assumed to undergo a conventional bimolecular reaction at the membrane with rate constants kon and koff (units, M−1 s−1 and s−1), such that the rate of change of membrane-bound probe (PrM) is dPrM/dt = konPrCLf − koffPrM. To avoid introducing surface concentration units, the lipid-bound probe and the free lipid are treated as though in solution throughout the cell volume, although both are actually only on the membrane. This has no effect on the calculated time courses but simplifies the units. In these units, the equilibrium dissociation constants (Kd = koff/kon) for the probe–lipid complexes were taken as 1.2 μM for PIP2 and PH-EGFP and 3 μM for DAG and C1-EGFP. These apparent dissociation constants are overestimates in the sense that they do not take into account the possibility that most of the PIP2 or DAG lipid may already be involved in other protein–lipid interactions. The known interaction of PH-EGFP with IP3 was ignored, a shortcoming that has no influence on the nuclear–cytoplasmic exchange time that is the focus of the model. In the following section, Kds for PH-EGFP are used in a much more complex model, and a lower Kd is needed for PIP2 because there the IP3 competition is included.
Finally, the ligand (PIP2 or DAG lipid) was assumed to be produced at a rate kprod (zeroeth order, M s−1) and destroyed with a rate constant kdest (first order, s−1) so that the total rate of change of the free membrane lipid is dLf/dt = kprod − kdestLf − dPrM/dt. The total amount of probe (0.5 μM) was set below the total amount of lipid (5 μM for PIP2), so that binding to the probe would not strongly deplete the lipid. Activation of muscarinic receptor was simulated by increasing kdest (for the PIP2/IP3 probe) or increasing kprod (for the DAG probe) during the oxo-M application. The complete program for the model is available in the online supplemental material . The model was solved by Euler (first-order) integration using time steps of 25 ms in the Igor Pro programming environment (Wavemetrics).
Results of the Simple Exchange Model
Fig. 11 summarizes the main results after free parameters were manually adjusted to simulate the observations of Fig. 1 B. The model does simulate transients in the changes of cytoplasmic probe concentrations that continue well after the assumed lipid concentration changes brought on by agonist addition or removal are finished. Thus, PIP2 is depleted during the two oxo-M applications, so the PH-EGFP probes translocate from the membrane to the cytoplasm, and then some of these probe molecules equilibrate into the nucleus (PN = 0.0032 μm s−1, exponential exchange time, τ ∼281 s). Similarly, membrane DAG rises during oxo-M so the C1-EGFP probes translocate from the cytoplasm to the membrane, and additional probe molecules then equilibrate from the nucleus (PN = 0.005 μm s−1, exponential exchange time, τ ∼188 s). The transients and the nuclear changes during onset and recovery resemble the measured time courses well enough to show that most if not all of the observed overshoots and decays can be understood as exchange of probe between cytoplasm and nucleus without postulating desensitizations and rebounds in the signaling from M1 receptors.
Relationship of the PH Probe and C1 Probe Work to our Larger Kinetic Model of Phosphoinositide Metabolism
We discuss briefly the large kinetic model of Suh et al. (2004), which was tentatively formulated to describe muscarinic suppression of KCNQ current in 10–20 s, recovery of current after removal of agonist in ∼200 s, and perturbation by cytoplasmic Mg2+ depletion and nucleotide analogs. That larger model was implemented in the Virtual Cell environment of the National Resource for Cell Analysis and Modeling, University of Connecticut Health Center (http://www.nrcam.uchc.edu). The revised working model with control values of rate constants and initial conditions is available at that web page for public use and modification.
In DISCUSSION we indicated a need to add cleavage of PIP by PLC to this model to account for the strong measured muscarinic depletion of PIP (Fig. 6). We consider here incorporating the translocation probes and the overexpression of the IP3-phosphatase, IP3-5-P, into the model. Since the model “cell” starts with 5 μM PIP2 in the plasma membrane (referred to total cell volume), a sudden activation of PLC that hydrolyzes all the PIP2 would produce 5 μM DAG and 5 μM IP3. The PH probe binds to PIP2 and IP3 and the C1 probe binds to DAG, a species not present in the original model. Therefore we added DAG as a species formed by PLC and added its breakdown/metabolism with a rate constant k = 0.013 s−1, chosen to produce a time constant of 75 s for decline of DAG after agonist withdrawal as in Fig. 1 G. The rate constant for IP3 breakdown by endogenous IP3-5-P was k = 0.066 s−1, chosen to produce a time constant of ∼15 s for decline of IP3 (not a value we measured).
Probes were assumed to be expressed at 0.5 μM concentration and to undergo reversible bimolecular interactions with their ligands. Published calorimetric measurements give dissociation constants Kd of 1.7 and 0.2 μM for the interactions of the PH domain with PIP2 (in phospholipid vesicles) or IP3 (in solution) (Lemmon et al., 1995), and surface plasmon assay measurements give 2.1 and 0.09 μM for the same ligands (Hirose et al., 1999). Calorimetry and plasmon assays give Kds of 10 and 6 nM for the DAG interaction of the C1A domain of PKCγ (Ananthanarayanan et al., 2003). All these Kds are expected to depend on the phospholipid composition of the vesicles used in the assay (Bittova et al., 2001).
Model simulations using the literature values for affinities of the PH domain gave a modest mimicry of the translocation seen in control cells. Half of the probe was bound to PIP2 on the “plasma membrane” at rest and virtually all was released within 20 s of “oxo-M application.” Then we tried to mimic the experiments with expression of IP3-5-P. We slowed PLC 15-fold (to represent the absence of any Ca2+ rise) and speeded the IP3-5-P in the model by 100-fold, which together reduced peak IP3 accumulation below 30 nM. The modulation of KCNQ current and the appearance of DAG were slowed appropriately as in the experiments of Fig. 10 (C–F). However, unlike the experiment of Fig. 10 G, these changes did not prevent a strong translocation of PH domain to the cytoplasm during “oxo-M application.” The problem was that the affinity of the PH probe for membrane PIP2 was so low that any cleavage of PIP2 would result in significant probe translocation. This was readily fixed by lowering the Kd for the PH domain–PIP2 interaction 20-fold to 0.1 μM, approximately matching the Kd for IP3. A related discrepancy with literature Kds is revealed in published experiments determining what concentration of cytoplasmic IP3 is needed to displace PH domain probes from the plasma membrane in intact cells. To displace half of the PH probe from the plasma membrane, Hirose et al. (1999) needed 13 μM IP3 in MDCK cells, and Winks et al. (2005) needed 16 μM IP3 in sympathetic neurons; and van der Wal et al. (2001) found that release of 10 μM IP3 displaces only a small portion of the PH probe, 100 μM IP3 displaces much more, and hydrolysis of PIP2 induced by activation of bradykinin receptors displaces even more. Using the published Kds in our large model predicted half release of PH domains already by 0.6 mM IP3, quite unlike these experiments, whereas using the revised Kd for PIP2 (0.1 μM) predicts half release by 7.8 μM IP3. Thus the PIP2 affinity in native membranes must be significantly higher than that in the phosphatidylcholine-phosphatidylserine vesicle environment used in binding assays. As an alternative approach, one might consider raising the PIP2 concentration in the model from the chosen value of 5 μM (corresponding to 5,000 μm−2 in the plasma membrane for a cell of 0.6 μm−1 surface-to-volume ratio). However, this approach is inconsistent with the expected chemical content of the plasma membrane. If the inner leaflet of the plasma membrane is 50% lipid, there would be room for 700,000 phospholipid molecules per square micrometer. Raising the surface density of PIP2 20-fold would make the inner leaflet 14% PIP2. In conclusion, we consider that the effective in-cell Kds of the PH probe for PIP2 and IP3 are about equal and near 0.1 μM.
Model simulations using the literature Kd values for the C1 domain were also not acceptable. The probe was fully associated with the plasma membrane rather than the cytoplasm in basal conditions because of its high affinity for the DAG formed there by the low, steady basal activity of PLC. The discrepancy could be fixed by raising the Kd 300-fold to 3 μM, which allowed 75% of the probe to be in the cytoplasm at rest. This might indicate that most of the basal DAG in the membrane is sequestered by other avid DAG-binding proteins (and the published Kd is correct) or that adding GFP to the C1 domain greatly reduces its DAG affinity. An alternative approach would be to slow the basal PLC rate drastically so that there is less basal DAG. In turn, that would require that PIP2 synthesis in the model be made a regulated process with a basal rate reduced by the same factor as the basal PLC rate was reduced. We lack measurements of DAG and have no basis to make choices among these possibilities so far.
Acknowledgments
We thank Paulette Brunner and Greg Martin for help in the Keck Imaging Center, Stuart McLaughlin for discussion, and Lea Miller and Chengcheng Shen for expert technical assistance.
This work was supported by National Institutes of Health grants NS08174, AR17803, NS07332, DA00286, DA11322, and HL067942.
Olaf S. Andersen served as editor.
References
L.F. Horowitz, W. Hirdes, and B.-C. Suh contributed equally to this work.
L.F. Horowitz's present address is Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N., Seattle, WA 98109.
W. Hirdes's present address is Institut für Angewandte Physiologie, Universitätsklinikum Hamburg-Eppendorf, Universität Hamburg, D-20246 Hamburg, Germany.
Abbreviations used in this paper: CHO, Chinese hamster ovary; DAG, diacylglycerol; EGFP, enhanced green fluorescent protein; IP3, inositol 1,4,5-trisphosphate; mRFP, monomeric red fluorescent protein; NEM, N-ethylmaleimide; oxo-M, oxotremorine-M; PH, pleckstrin homology; PIP2, phosphatidylinositol 4,5-bisphosphate.