Isoform-specific Inhibition of TRPC4 Channel by Phosphatidylinositol 4,5-Bisphosphate*

Full-length transient receptor potential (TRP) cation channel TRPC4α and shorter TRPC4β lacking 84 amino acids in the cytosolic C terminus are expressed in smooth muscle and endothelial cells where they regulate membrane potential and Ca2+ influx. In common with other “classical” TRPCs, TRPC4 is activated by Gq/phospholipase C-coupled receptors, but the underlying mechanism remains elusive. Little is also known about any isoform-specific channel regulation. Here we show that TRPC4α but not TRPC4β was strongly inhibited by intracellularly applied phosphatidylinositol 4,5-bisphosphate (PIP2). In contrast, several other phosphoinositides (PI), including PI(3,4)P2, PI(3,5)P2, and PI(3,4,5)P3, had no effect or even potentiated TRPC4α indicating that PIP2 inhibits TRPC4α in a highly selective manner. We show that PIP2 binds to the C terminus of TRPC4α but not that of TRPC4β in vitro. Its inhibitory action was dependent on the association of TRPC4α with actin cytoskeleton as it was prevented by cytochalasin D treatment or by the deletion of the C-terminal PDZ-binding motif (Thr-Thr-Arg-Leu) that links TRPC4 to F-actin through the sodium-hydrogen exchanger regulatory factor and ezrin. PIP2 breakdown appears to be a required step in TRPC4α channel activation as PIP2 depletion alone was insufficient for channel opening, which additionally required Ca2+ and pertussis toxin-sensitive Gi/o proteins. Thus, TRPC4 channels integrate a variety of G-protein-dependent stimuli, including a PIP2/cytoskeleton dependence reminiscent of the TRPC4-like muscarinic agonist-activated cation channels in ileal myocytes.

The seven members of the "classical" family of mammalian TRP 4 (TRPC) proteins are closely related to the prototypical Drosophila TRP and TRPL (dTRPs) both structurally and functionally as they are commonly gated by phospholipase C (PLC) activation (1). However, the molecular scenarios of TRPC activation downstream of PLC are diverse and may involve relevant lipids such as diacylglycerol (DAG) in the case of TRPC2, -C3, -C6, and -C7, inositol 1,4,5-trisphosphate formation, and/or Ca 2ϩ store depletion in the case of TRPC1 and -C3 (1). The activation mechanism for the TRPC4/5 subgroup remains most elusive as none of the above appears to be involved (2). Notably, TRPC4/5 are most closely related to dTRPs, and recombinant dTRPL is known to be inhibited by the PLC substrate phosphatidylinositol 4,5-bisphosphate (PIP 2 ), although PIP 2 hydrolysis alone cannot account for dTRPL activation (3). Moreover, in trp mutants PIP 2 depletion in vivo causes inhibition, rather than potentiation, of the dTRPL channels, which then remain insensitive to light until PIP 2 is resynthesized (4). Among mammalian TRPs, several members of the TRPV (V1, V5) and TRPM (M4, M5, M7, and M8) families have recently been shown to be regulated by PIP 2 (5,6), whereas the C termini of several TRPCs (C1, C5, C6, and C7) and TRPV1 bind to anionic lipids directly (7). In TRPC6, this binding, in competition with calmodulin (CaM), removes its inhibitory action and thus facilitates channel activation (7).
Here we tested the hypothesis that PIP 2 can regulate TRPC4 channel activity. TRPC4 is widely expressed in different tissues, particularly in brain, vascular endothelium, and gastrointestinal smooth muscles (8). Native functions of TRPC4 include its role in the endothelium-dependent vasorelaxation and control of endothelial permeability (9,10). Its biophysical properties also resemble muscarinic cation current (mI CAT ) of the gastrointestinal tract, which is similarly regulated by the PLC system via M 3 muscarinic receptors (11). Indeed, evidence is growing that TRPC4 is a critical part of mI CAT ((12) and our observations 5 showing the lack of mI CAT in TRPC4-deficient mouse ileal myocytes).
Two most abundant TRPC4 variants are a "full-length" TRPC4␣ and shorter TRPC4␤ (or TRPC4⌬) lacking a stretch of 84 amino acids (781-864 in the case of mouse TRPC4) in the cytosolic C terminus (⌬84AA). The functional implications of their differential tissue expression (13,14) are intriguing, and they call for further investigations into the differential regulation of TRPC4 isoforms. They form heteromultimers whereby TRPC4␣ exerts a dominant negative effect ascribed to an autoinhibitory function of the ⌬84AA region (15). However, the nature of this inhibition remains unknown.
In this study we found that PIP 2 inhibited TRPC4␣, but not TRPC4␤, suggesting a role of the ⌬84AA region in PIP 2 interaction with the channel. This was revealed by direct in vitro binding assay. Furthermore, we show that this PIP 2 -dependent regulation involves cytoskeleton and TRPC4 C-terminal PDZinteracting domain through which TRPC4/5 are known to associate with NHERF/ezrin-radixin-moesin (ERM)/actin cytoskeleton (16,17). Importantly, we found that several other phosphoinositides, which bind to the C termini in other TRPCs (7), including PI(3,4)P 2 , PI(3,5)P 2 , and PI(3,4,5)P 3 , as well as highly negatively charged inositol hexaphosphate (IP 6 ) had no effect or even potentiated TRPC4␣ indicating that the inhibitory action of PIP 2 was highly specific and was not simply a negative charge effect. Overall, PIP 2 depletion appears to be a required step in the activation of TRPC4␣ but not TRPC4␤. Additional requirement of TRPC4 channel activation include intracellular Ca 2ϩ , pertussis toxin (PTX)-sensitive G i/o proteins, and perhaps other undefined components of PLC signaling. The multifaceted gating requirement suggests that TRPC4 channels are capable of integrating a variety of G-protein-dependent stimuli, including a PIP 2 /cytoskeleton dependence much like the TRPC4-like muscarinic acetylcholine receptor (mAChR)-activated cation channels in ileal myocytes. Integration of these stimuli might enable fine-tuning of channel activation as required for coordinated functions of the endothelium or smooth muscles in the intestine and the bladder, as well as neural signaling in the nervous system.

EXPERIMENTAL PROCEDURES
Cells and Treatments-HEK293 cells stably expressing mouse TRPC4 (mTRPC4) isoforms were grown under culture conditions as described (16) and seeded in 35-mm dishes 2 days prior to patch clamp recordings. For cells coexpressing M 5 AChR, the receptor cDNA in pIREShyg2 vector (BD Biosciences) was transfected to the stable TRPC4 cell lines, and transformants were selected and maintained in the culture medium supplemented with 100 g/ml hygromycin B. Male guinea pigs (300 -400 g) were killed by exposure to carbon dioxide (Schedule One of the UK Animals Scientific Procedures Act 1986). Single smooth muscle myocytes from the longitudinal muscle layer of the guinea pig ileum were isolated after collagenase (type 1A, 1 mg ml Ϫ1 at 36°C for 25 min) treatment (18). Cells were treated with cytochalasin D (5 M) for 2 h at 37°C, PTX (100 ng/ml) for 16 -18 h at 37°C, and wortmannin (20 or 30 M) for 10 -60 min at 20 -23°C.
Total RNA Isolation and Reverse Transcription-PCR-Purified single guinea pig ileal myocytes were used for protein and RNA analysis. Myenteric ganglia were removed from cell suspension using cell filtration through 80-m nylon net filters (Millipore). Total RNA was extracted from ileal myocytes using TRIzol reagent (Invitrogen). Reverse transcription was performed using SuperScript II reverse transcriptase (Invitrogen) and random hexamers. PCR was performed using sense primer 5Ј-CTGCAAATATCTCTGGGAAG-3Ј and antisense primer 5Ј-CTACAATCTTGTGGTCACGTAGT-3Ј, for predicted products encompassing guinea pig TRPC4 cDNA derived from exons 5 to 9 and the coding sequence for ⌬84AA with predicted products of 1422 and 1170 bp for the ␣ and ␤ isoforms, respectively.
Protein Preparation and Western Blot-Microsomal membrane proteins were prepared from wild type mouse brain, TRPC4-deficient mouse brain, and smooth muscle cells isolated from guinea pig ileal longitudinal muscle. Protein samples (150 g per lane) were separated on 8% SDS-polyacrylamide gels, blotted, and incubated in the presence of the TRPC4 antibody (14). To control protein loading, the filter was stripped and incubated in the presence of a CaV␤2 antibody that recognizes the type 2 ␤ subunit of voltage-activated Ca 2ϩ channels (19). Experiments were repeated twice with identical results.
Coimmunoprecipitation-Cell lysates from untransfected HEK293 cells and stably transfected cell lines were immunoprecipitated using goat anti-actin and goat anti-ezrin antibodies (both were from Santa Cruz Biotechnologies) in a buffer containing 150 mM NaCl, 20 mM Tris-Cl, 1 mM EDTA, and 0.5% Triton X-100, pH 7.4. The precipitated proteins were subjected to SDS-PAGE, transferred to nitrocellulose membrane, and immunoblotted using a polyclonal anti-TRPC4 antibody against an N-terminal peptide of TRPC4 (16).
In Vitro PIP 2 Binding Assay-The C termini of TRPC4␣-(733-974) and TRPC4␤-(733-890), as well as ⌬84AA-(781-864), were synthesized in vitro as maltose-binding protein fusion proteins using the transcription-and translation-coupled rabbit reticulocyte lysate system (Promega) in the presence of [ 35 S]Met and [ 35 S]Cys. The products were incubated with PIP 2 -agarose beads (Echelon Inc.) in 260 l of a binding buffer that contained 120 mM KCl, 20 mM Tris-HCl, pH 7.4, for 1 h at room temperature. Bound proteins were collected by a brief centrifugation at 1,000 ϫ g for 1 min and washed twice with 1 ml of the same binding buffer. Protein separation and exposure to x-ray film are as described previously (16).
Intracellular Ca 2ϩ and Membrane Potential Measurements-Control HEK293 cells and stable cell lines were seeded at 100,000 cells/well in wells of 96-well plates and allowed to grow for 18 -24 h. Cells were washed with an extracellular solution and loaded with Fluo4-AM (2 M) as described (20). [Ca 2ϩ ] i was measured using a fluid handling integrated fluorescence plate reader, FlexStation (Molecular Devices). Drugs were diluted in extracellular buffer at three times the desired final concentrations and delivered to the sample plate by the integrated robotic 8-channel pipettor at the preprogrammed time points. For membrane potential measurements, the FLIPR 5 V. Tsvilovskyy, M. Freichel, and V. Flockerzi, unpublished observations. TRPC4 and PIP 2 APRIL 11, 2008 • VOLUME 283 • NUMBER 15 membrane potential dye was diluted in the extracellular solution and added to cells following the manufacturer's protocol (Molecular Devices). The excitation/emission wavelengths used for Fluo4 and the membrane potential dye were 494/525 and 530/565 nm, respectively.
Patch Clamp Recordings and Data Analysis-Whole-cell currents were recorded using borosilicate patch pipettes (2-3 megohms) and an Axopatch 200A amplifier (Molecular Devices, Union City, CA) interfaced to Digidata 1322A with the pClamp 9 support. Holding potential was Ϫ40 mV. Parallel continuous data acquisition was performed using MiniDigi 1A (Molecular Devices) and AxoScope 9 software. TRPC4 currents in HEK293 cells were activated by carbachol application or, alternatively, by infusing 200 M GTP␥S via pipette. mI CAT in ileal myocytes was activated by applying carbachol at a submaximally effective concentration of 50 M.
Data were analyzed and plotted using Clampfit 9 (Molecular Devices) and Origin 7 (Microcal, Northampton, MA). Cationic conductance activation curves were obtained from steady-state I-V relationships measured by voltage steps or slow (6-s-long) voltage ramps from 80 to Ϫ120 mV (both producing similar I-V curves, e.g. Fig. 1, E and F) and fitted using a modified Boltzmann function as shown in Equation 1, where Gv and Gs are voltage-dependent and -independent components, respectively (Gs is typically Ͻ15% of Gv); Va and Vi are the potentials of half-maximal activation and inhibition, respectively; s a and s i are corresponding slopes; and Is is the noninactivating component. This formalism accounted for both region of negative slope conductance at potentials less than Ϫ40 mV and channel inactivation (or block) at potentials more than Ϫ20 mV. Fitting was restricted to potentials of Ͻ40 mV because at more depolarized potentials the blockade was somewhat relieved, resulting in an overall characteristic doubly rectifying I-V shape or N-shaped conductance curve.
Stock solutions of Me 2 SO-soluble drugs were made at the following concentrations: 2 mM U-73122 and U-73343, 5 mM cytochalasin D, 10 mM wortmannin. diC8-phosphoinositides (Echelon Research Laboratories, Salt Lake City, UT) were dissolved in deionized and deoxygenated water at 0.4 mM or directly in the pipette solution (if applied at concentrations higher than 20 M) by a 30-min sonication on ice. All other drugs were from Sigma.
All values are expressed as means Ϯ S.E. A t test (for two groups) or analysis of variance (for multiple comparisons) followed by Dunnett's test to compare all data versus control were used, and differences were considered statistically significant with p Ͻ 0.05.

RESULTS
Activation of TRPC4 Is Dependent on Voltage, PLC, and PIP 2 -TRPC4 channel currents and mI CAT in native gastrointestinal myocytes share many biophysical and regulatory properties (2,11). By examining the expression of TRPC4 in single guinea pig ileal myocytes with myenteric ganglia removed by cell filtration, we found that both TRPC4␣ and -␤ isoforms were expressed at the mRNA and protein levels (Fig. 1A). Importantly, Western blotting showed comparable levels of expression of the two isoforms. Thus, in this study we first investigated regulation of TRPC4␣ and -␤ channels stably expressed in HEK293 cells. We then tested relevance of our findings under identical conditions for functional regulation of mI CAT , which we used as a native counterpart.
We established HEK293 cells stably expressing murine TRPC4␣ or TRPC4␤ (Fig. 1B). These isoforms differ in the ⌬84AA stretch in the C terminus, which is present in TRPC4␣ only and which contains two CaM-binding sites as shown schematically in Fig. 1B, bottom inset. In fluorescence membrane potential or intracellular Ca 2ϩ concentration measurements, TRPC4 activity was readily detectable as membrane depolarization in response to carbachol (Fig. 1C) or a more prolonged agonist-induced Ca 2ϩ signal indicative of an enhanced Ca 2ϩ entry (Fig. 1D). This occurred through an endogenous G q/11coupled mAChR, most likely the M 3 type as an M 3 antagonist, p-F-HHSiD, inhibited the responses with a pA 2 value of 7.4 (derived from Schild regression analysis; data not shown).
In ileal myocytes, carbachol induces robust current responses (Fig. 1E, top inset), which have been shown to occur through synergistic activation of G i/o -coupled M 2 and G q/11coupled M 3 mAChR (21)(22)(23). This current shows a characteristic doubly rectifying I-V relationship (Fig. 1E, black curve), which is similar to that of TRPC4 current, especially in case of the TRPC4␤ isoform (e.g. Fig. 1E, gray line), which shows lesser inactivation at positive potentials (see below).
Gating of several TRPVs and TRPMs is efficiently regulated by both membrane potential and phosphoinositides (PIs) (24), whereas voltage-dependent aspects of TRPC gatings are less well appreciated. Although the origin of TRPC4 voltage dependence is not known, the current shows prominent regulation by membrane potential. This was particularly evident in the voltage-step experiments (Fig. 1F). Thus, stepping the membrane potential negatively from the holding potential of Ϫ40 mV resulted in an instantaneous current increase because of increased driving force followed by rapid current relaxations to levels that were smaller than the holding current at Ϫ40 mV. Tail currents were also prominent as shown in Fig. 1F (magnified inset) suggesting channel deactivation at negative potentials as well as additional activation at positive potentials. In the same cell, an identical I-V relationship could be obtained by applying a 6-s duration voltage ramp (Fig. 1E, compare gray line, ramp protocol, and circles, voltage step protocol). This justified the use of the slow voltage ramps for evaluation of the steady-state voltage-dependent properties of TRPC4. In this and in the majority of other experiments we mainly analyzed TRPC4 currents induced by intracellular GTP␥S because carbachol was rather inefficient (see below).
We quantitatively analyzed this voltage dependence by generating TRPC4 conductance curves and fitting them to Equation 1, which can simultaneously approximate either channel activation (between Ϫ120 and about Ϫ20 mV) and block or activation (positive to about Ϫ20 mV) (Fig. 1G). In one well studied case, TRPV6, strong inward rectification occurs because of intracellular Mg 2ϩ channel block (25), but its origin in TRPC4 remains unclear.
Both heterologously expressed TRPC4 channels and native mI CAT are activated downstream of PLC, but the prime activator in the PLC pathway remains unknown in both cases. The main products of PLC activation, inositol 1,4,5-trisphosphate and DAG (as well as DAG metabolites), do not activate TRPC4 or mI CAT to any appreciable extent (2,11). Therefore, in this study, we focused on the role of the PLC substrate, PIP 2 , and tested the hypothesis that PIP 2 may represent a physiologically important brake on TRPC4 activity.
In the fluorescence membrane potential assay, both TRPC4␣ and -␤ cell lines displayed similar sensitivity to carbachol with EC 50 values of ϳ7 M (Fig. 2, A and B). However, in whole-cell recordings, carbachol failed to induce any significant current in TRPC4␤-expressing cells (not shown). By contrast, about two-thirds of TRPC4␣-expressing cells responded to carbachol by generating inward currents of variable amplitudes (Fig. 2, C and D). This discrepancy between fluorescence membrane potential assay and the whole-cell recording could be due to the fact that the undisturbed cytosolic environment in the former assay allowed some positive feedback mechanisms, which could be missing in the whole-cell configuration (e.g. limited Ca 2ϩ -dependent TRPC4 potentiation because of [Ca 2ϩ ] i buffering), to amplify channel activation and that small and noisy current responses recorded in voltage clamp experiments were nevertheless sufficient to produce significant membrane depolarization owning to a very high input resistance of these cells.
To study the effect of PIP 2 on agonist-induced TRPC4␣ current, we included 20 M dioctanoyl analog of PIP 2 (diC8-PIP 2 ), the more water-soluble short form of PIP 2 , in the recording pipettes. Under these conditions, none of the 12 tested TRPC4␣-expressing cells produced any substantial response to carbachol (Fig. 2, C, bottom trace, and B). Mean current density with PIP 2 was significantly lower than without PIP 2 (p ϭ 0.015). This finding is in support of the above hypothesis that PLC-mediated breakdown of PIP 2 is involved in TRPC4 activation. However, the fact that not all control cells were responsive to the agonist prompted us to  Intriguingly, overexpression of the G q -coupled M 5 AChR failed to improve activation by carbachol of both TRPC4 isoforms in the whole-cell experiments. In fluorescence membrane potential assay, the EC 50 to carbachol was reduced 15and 51-fold for TRPC4␣ and -␤, respectively, without a change in the maximal response (Fig. 2, A and B), showing that the receptor is expressed and can be coupled to TRPC4 activation. The fact that it did not exert an effect on eliciting TRPC4 current in whole-cell recording indicated that PLC activation may not be the sole mechanism of TRPC4 gating (see below), at least under the whole-cell conditions used for recording mI CAT . The above mentioned reasons would, of course, make the membrane potential assay more sensitive compared with voltage clamp recording conditions.
To bypass the undefined components required for TRPC4 gating, we used GTP␥S, which nonselectively activates all heterotrimeric G-proteins (i.e. bypassing receptors). Infusion of 200 M GTP␥S produced robust long lasting currents (termed I TRPC4 ) in cells expressing both TRPC4␣ (n ϭ 14) and -␤ (n ϭ 7) (43.5 Ϯ 5.8 pA/pF at Ϫ40 mV, n ϭ 21), but not in control cells (2.3 Ϯ 0.7 pA/pF, n ϭ 5) (Fig. 3A). The current peaked, on average, at 336 Ϯ 39 s (variations seen in individual cells were from 120 to 720 s) and showed slight desensitization during the next 10 min. Thus, in all subsequent experiments GTP␥S was used to activate I TRPC4 as this approach allowed us to investigate both TRPC4 isoforms in the most reliable manner. To account for variations in cell size (C M 8 -41 pF, on average 19.0 Ϯ 1.3 pF in TRPC4␣-and 16.6 Ϯ 1.1 pF in TRPC4␤-expressing cells, n ϭ 19 -28), as well as time-dependent variations in culture, we normalized I TRPC4 by C M and performed all test and relevant control measurements by alternating them, usually on the same day. Steady-state I-V relationships were established by applying 6-s voltage ramps from 80 to Ϫ120 mV seen as vertical current deflections at 30-s intervals, e.g. Fig. 3, B-E, left panels.
Consistent with previous reports for both I TRPC4 and mI CAT (22,23,26), the PLC blocker U-73122 strongly suppressed both TRPC4␣ and -␤ currents as representative examples in Fig. 3, B and D, show (n ϭ 4; p Ͻ 0.02 or better). Its inactive analog U-73343 had no detectable effect (Fig. 3, C and E; n ϭ 4). How-  ever, it has more recently been shown that TRPC5, and perhaps TRPC4, can be activated by breaking a disulfide bridge near the selectivity filter (27). Given that the pyrrole-2,5-dione moiety of U-73122 is highly thiol-reactive, whereas the corresponding pyrrolidin-2,5-dione of U-73343 is not (28), it cannot be excluded that U-73122 may act more directly on the TRPC4 channel rather than through PLC. Isoform-specific Inhibition of TRPC4␣ but Not TRPC4␤ by PIP 2 -Similarly to the above-described experiments with carbachol, diC8-PIP 2 (20 M) was applied via pipette together with GTP␥S. This almost completely prevented GTP␥S-induced activation of TRPC4␣ ( Fig. 4A; p Ͻ 0.02). In controls, infusion of diC8-PIP 2 alone had no effect. Notably, the small remaining current showed more deactivation at Ϫ120 mV as its I-V relationship at negative potentials became more U-shaped (see amplified I-V curve shown in the inset of Fig. 4C).
In contrast, in TRPC4␤-expressing cells diC8-PIP 2 had no effect (Fig. 4, B and D), immediately suggesting that the ⌬84AA stretch is an important determinant of PIP 2 interaction with TRPC4␣. We therefore hypothesized that this domain may bind to PIP 2 , leading to channel closure as depicted by the mechanistic model shown later in Fig. 5D. We tested the interaction of TRPC4␣ C terminus (Ct, 733-974), TRPC4␤-Ct-(733-890), and the ⌬84AA region with PIP 2 . As a positive control, we used the PH domain of PLC␦ (PLC␦-PH), which binds PIP 2 with a high affinity. The binding of TRPC4␣-Ct, but not TRPC4␤-Ct, to PIP 2 agarose was clearly detectable, although it was much weaker compared with PLC␦-PH (Fig. 4E). ⌬84AA fragment showed weak binding compared with TRPC4␣-Ct suggesting that this domain may be only a part of a larger PIP 2binding domain. Plausibly, such relatively low PIP 2 affinity permits TRPC4␣ sensing a moderate PIP 2 change, as opposed to a high affinity regulatory site which would require a more severe PIP 2 depletion.
The PIP 2 -mediated Inhibition Is Dependent on the Association of TRPC4 with Actin Cytoskeleton-PIP 2 interacts with cytoskeletal proteins and anchors various signaling molecules to the plasma membrane (29,30). Proteins of cytoskeletal remodeling represent important targets of PIP 2 signaling. On the other hand, TRPC4, along with PLC␤, is associated with actin cytoskeleton through binding to a PDZ domain protein NHERF (also known as EBP50), which in turn interacts with the actin-binding ERM proteins (16,17). In related TRPC5, deleting the PDZ-binding motif did not have any effect on channel activation or biophysical properties, but overexpressing NHERF introduced a significant delay in current activation of the wild type channel (31).   n ϭ 9). B, in ⌬TTRL mutant, the effect of PIP 2 was also completely prevented (n ϭ 10). C, coimmunoprecipitation of TRPC4␣ with ERMs and actin. Cell lysates from untransfected control HEK293 cells and stable lines expressing human TRPC3, C-terminal hemagglutinin-tagged murine TRPC4␣ (C4␣-HAc), wild type TRPC4␣, or ⌬TTRL mutant were immunoprecipitated using goat anti-actin (middle) and goat anti-ezrin (lower) antibodies and then immunoblotted using a polyclonal anti-TRPC4 antibody. TRPC4␣ but not C4␣-HAc or ⌬TTRL was precipitated by the anti-actin and anti-ezrin antibodies. Upper panel is a Western blot of total cell lysates performed using the anti-TRPC4 antibody. D, proposed mechanistic model showing that PIP 2 binding to the ⌬84AA stretch (and likely to adjacent sites) present only in TRPC4␣ stabilizes its closed conformation and that interactions of TRPC4␣ with NHERF/ERM/F-actin are required for PIP 2 -dependent channel inactivation.
We therefore asked whether the association with actin cytoskeleton has a role in PIP 2 regulation of TRPC4␣. First, cells were treated with cytochalasin D, a potent cell-permeable fungal toxin that inhibits actin polymerization and thus disrupts actin microfilaments. This completely prevented diC8-PIP 2 inhibition of TRPC4␣ (Fig. 5A). Second, the last four amino acids of TRPC4␣, which constitute the PDZ recognition sequence, were deleted (⌬TTRL mutant); the mutant was stably expressed in HEK293 cells, and the current was studied as for the wild type TRPC4␣. The current of ⌬TTRL mutant channel showed identical biophysical properties to the wild type channel (compare Figs. 5B and 4C, and also to Ref. 31 in case of TRPC5), but, again, its PIP 2 sensitivity was completely lost (Fig.  5B).
These results, taken together with the coimmunoprecipitation results showing that wild type but not C-terminal hemagglutinin-tagged or ⌬TTRL mutant channel interacts with actin and ezrin (Fig. 5C), strongly imply that an intact multiprotein TRPC4␣⅐NHERF⅐ERM⅐actin complex, in addition to the ⌬84AA stretch, is required for PIP 2 inhibition of TRPC4␣. Thus, our mechanistic model explaining PIP 2 -mediated channel closure includes both the ⌬84AA stretch and these important protein-protein interactions (Fig. 5D).
Despite the complete loss of PIP 2 inhibition, no current or little spontaneous current was observed in both cases. Thus, relief from the PIP 2 -blocked state may not be the sole mechanism of TRPC4␣ activation (see below). PIP 2 Inhibits mI CAT Activation in an F-actin-dependent Manner-Because mI CAT appears to be largely mediated by TRPC4, it was interesting to see whether the PIP 2 dependence of TRPC4␣ gating could be interposed between the M 3 /PLC system and mI CAT activation. When carbachol was applied 4 -5 min after breakthrough with patch pipettes containing 20 M diC8-PIP 2 , no or small mI CAT was detectable. Therefore, to follow the time course of the PIP 2 effect, carbachol was applied shortly after establishing whole-cell access (Ͻ1 min as required for the series resistance compensation and background I-V curve measurement). In this case, mI CAT was normally activated by the agonist but then rapidly declined to a small steadystate level (Fig. 6A, compare with control response shown in gray). Similar to TRPC4␣, F-actin disruption by cytochalasin D prevented PIP 2 inhibition (Fig. 6B, n ϭ 5).
Similarly to PIP 2 action on TRPC4␣, the I-V relationships acquired during diC8-PIP 2 action became pronouncedly more U-shaped (Fig. 6, C and D, compare with Fig. 4C, inset) resulting in almost complete current inhibition at Ϫ120 mV. PIP 2 Depletion Is Not Sufficient for Full TRPC4␣ Activation; Some Additional Mechanisms of Channel Gating-Results presented so far revealed that although PIP 2 is an important determinant of TRPC4 gating, its depletion may not be the sole mechanism of the channel activation. Indeed, no or little spontaneous current was observed in all cases when PIP 2 inhibition was lacking (e.g. TRPC4␤ isoform, or disrupted cytoskeleton interaction; Fig. 4B and Fig. 5, A and B).
Control unstimulated (e.g. no GTP␥S in the pipette) TRPC4␣-expressing cells showed small currents characterized by a linear I-V relationship (Fig. 7A). Wortmannin-treated cells showed larger noisy currents with a characteristic TRPC4 I-V relationship (Fig. 7, A and B, squares), yet the amplitude of these responses (e.g. 2.0 Ϯ 0.3 pA at Ϫ40 mV; n ϭ 12) was Ͻ2% of GTP␥S-induced currents. In contrast, TRPC4␤-expressing cells did not show any significant increase of the background currents after wortmannin treatment (n ϭ 4, p ϭ 0.15).
Poly-L-lysine (PLL) was used as a PIP 2 scavenger inhibiting the activity of the PIP 2 -activated TRPM4 and TRPM8 channels (33,35). However, applied at 50 g/ml via pipette (without GTP␥S), PLL did not cause any significant TRPC4␣ current activation even during prolonged, up to 20 min, recordings (Fig.  7A). The I-V relationship recorded with PLL in the pipette solution was again reminiscent of TRPC4 (Fig. 7B, circles), but current density at Ϫ40 mV was only 2.4 Ϯ 0.9 pA (n ϭ 4).
Taken together, these data demonstrate that releasing TRPC4␣ from PIP 2 block is necessary but not sufficient for its full activation. On the other hand, intracellular Ca 2ϩ is required for TRPC4 activation (2). Although this mechanism was unlikely to make any significant contribution under conditions of strongly buffered [Ca 2ϩ ] i , as used in our experiments, we nevertheless studied its implications for TRPC4␣ and -␤ isoforms. This was especially interesting because the isoforms differ in two CaM-binding sites located within the ⌬84AA sequence (Fig. 1B). For both isoforms, I TRPC4 was abolished by Ca 2ϩ -free internal solution with 10 mM BAPTA (Fig. 7C). This is similar to mI CAT for which internal Ca 2ϩ is also required (36).
Furthermore, mI CAT is blocked by PTX treatment (37,38), an effect ascribed to G o protein involvement (39,40). We therefore hypothesized that the failure of carbachol to reliably induce I TRPC4 could be due to the lack of a G o -coupled mAChR in HEK293 cells. Consistent with this, treatment with PTX abolished GTP␥S-induced I TRPC4 (Fig. 7D). The effect was again similar for both TRPC4 isoforms. All the above described effects were statistically significant (p Ͻ 0.02 or better). Nearly complete current inhibition observed in all cases (Fig. 7, C and  D) indicated that functional G i/o proteins are necessary for I TRPC4 activation. In a separate study, we show that coexpression of M 2 AChR resulted in activation of large and sustained TRPC4 currents by carbachol under whole-cell conditions for all cells. Further investigation of the pathways linking G i/o -coupled receptors to TRPC4 activation is undertaken and is beyond the scope of this study.
Other PIs and TRPC4␣ Gating-Montell and co-workers (7) have recently demonstrated direct binding of several PIs to other members of the TRP superfamily, including TRPC1, -5, -6, and -7. In some cases this binding displaced inhibitory CaM, but in a related TRPC5 neither the PIs nor IP 6 displaced CaM. We therefore asked whether other PIs could also inhibit TRPC4␣. The effect of PIP 2 on TRP and other ion channels is often inter-preted in terms of electrostatic interactions between the negatively charged headgroup of PIP 2 and basic amino acid residues, often found in the C termini of various ion channels (6,41,42). Eight common PIs exist, which differ in the number and position of phosphorylation of the inositol ring and which can convert into each other through the physiological action of various lipid kinases and phosphatases (41). From the point of view of the role of the negative charge density and specificity of interactions with TRPC4␣, of particular interest were the two similarly charged isoforms of PIP 2 , PI(3,4)P 2 and PI(3,5)P 2 , as well as IP 6 with a higher negative charge resulting from the six phosphates.
Taking into account differences in the binding affinities (7), IP 6 and PIs were applied (similarly to PIP 2 ) at the following concentrations: IP 6 and PI(3,4,5)P 3 at 20 M, PI(3,4)P 2 and PI(3,5)P 2 at 100 M, and PI(4)P at 200 M. Current densities at three different test potentials (Ϫ60, Ϫ120, and 80 mV) were compared (Fig. 8A). No I TRPC4 inhibition was seen in these  Current scale is omitted because complete I-V relationships are shown. E, normalized and averaged activation curves obtained under these various conditions. Compared with control (circles) significant differences were seen in case of IP 6 (squares) and PI(3,5)P 2 (triangles); with other PIs the curves are shown by the gray lines overlapping with control. experiments, highlighting the specificity of the inhibitory action of PIP 2 , which thus appears to be not simply a negative charge effect.
Intriguingly, in two cases significant potentiation of I TRPC4 occurred in a highly specific voltage-dependent manner. Thus, IP 6 increased the current at negative but not at positive potentials, whereas with PI(3,5)P 2 a much larger current at positive potentials was observed. The time-dependent changes of I-V relationships of these two examples are further illustrated, in comparison with control conditions, in Fig. 8, B-D.
We also examined the normalized and averaged activation curves obtained under these various conditions (Fig. 8E). Again, compared with control (Fig. 8E, circles) significant differences were seen in the case of IP 6 (squares) and PI(3,5)P 2 (triangles). The possibility that TRPC4 voltage dependence can be altered depending on the lipid environment shows that TRPC4 is a novel example of the increasing number of TRPs showing important connections between lipid-sensing and voltage-dependent regulation (6,24,43).
Interestingly, during the initial TRPC4␣ activation the I-V relationship was similar to control both with IP 6 (Fig. 8C) and with PI(3,5)P 2 (Fig. 8D), but the above described characteristic changes developed somewhat later. For example, in Fig. 8D larger outward current developed and the I-V relationship at positive potentials became progressively more linear only after the current at negative potentials fully stabilized. This may suggest that interaction of these ligands with TRPC4␣ can take place only after substantial PIP 2 depletion, or some other components required for TRPC4 activation are fully engaged. Clearly, much further work is needed to define PIP 2 , PI(3,5)P 2 , and IP 6 binding properties and sites of interactions, but at present we can conclude that PI(3,4,5)P 3 , PI(3,4)P 2 , and PI(4)P do not functionally regulate I TRPC4 and are thus unlikely to interact with TRPC4.

DISCUSSION
This study adds TRPC4 to the growing list of TRP and other channels (e.g. Kir, KCNQ1, and P/Q-type Ca 2ϩ channels) regulated by PIP 2 (41). Although PIP 2 and related phosphoinositides exert positive effects on the activity of many TRP channels, such as TRPV5 (32), TRPM4 (35,44), TRPM5 (45), TRPM7 (46,47), TRPM8 (33,34), and TRPC6 (7,48), it is inhibitory to few others, such as dTRPL (3) and TRPV1 (42, 49). Ironically, the inhibitory effect of PIP 2 on TRPV1 has been challenged by more recent studies showing the activating effect of PIP 2 on TRPV1 in inside-out patches (50,51). Therefore, among mammalian TRP channels, the PIP 2 -induced inhibition appears to be unique for TRPC4␣. The only other example is the closely related Drosophila channel, TRPL (3). Three lines of evidence support the notion that relief from PIP 2 inhibition is required but not sufficient for full activation of TRPC4␣ as follows: (i) cytochalasin D treatment (Fig. 5A), (ii) deletion of the PDZbinding domain (Fig. 5B) both prevented PIP 2 dependent TRPC4␣ inhibition, whereas (iii) TRPC4␤ was naturally lacking this inhibition (Fig. 4, B and D), yet in all cases no significant spontaneous channel activity was observed. Wortmannin and PLL also caused a minimal TRPC4␣ activation (Fig. 7, A and B). On the other hand, PIP 2 interaction with TRPC4␣ in terms of channel gating appears to be strong enough to override all other activating stimuli as under conditions of constant intracellular diC8-PIP 2 supply carbachol or GTP␥S failed to induce any significant current (Fig. 2C and Fig. 4, A and C).
One of the most interesting and important outcomes of this study is uncovering of a previously unsuspected complexity of PIP 2 interaction with an ion channel, which, as we show here, requires integrity of a multiprotein complex rather than a single domain. Furthermore, although PIP 2 may be more directly associated with ERMs and actin cytoskeleton (29,30), the intricate connection between PIP 2 -induced TRPC4␣ inhibition and the intact F-actin network is mediated solely through the C-terminal binding of the channel to the PDZ domain scaffolding protein NHERF, indicating that this type of regulation is rather specific. These results have interesting functional implications as actin cytoskeleton is a highly dynamic structure, which can undergo a considerable remodeling during receptor activation. This can potentially relieve TRPC4␣ from PIP 2 inhibition even at a constant PIP 2 level. Taken together, our results reveal considerable complexity of TRPC4 activation whereby some factors are required (i.e. permissive) in nature (e.g. PIP 2 depletion, a certain minimal [Ca 2ϩ ] i level), whereas others are required for channel gating such as activated G i/o proteins (Fig. 7, C and D).
Thus, TRPC4 channels can integrate a variety of G-proteindependent stimuli. In some tissues, such multiple coupling can be naturally realized through the expression of differentially coupled receptors for the same agonist, such as acetylcholine in visceral smooth muscles acting simultaneously on M 2 and M 3 mAChRs. In this study, we showed that both TRPC4 isoforms were about equally expressed in ileal myocytes (Fig. 1A) and that mI CAT was inhibited by PIP 2 , and in a cytoskeleton-dependent manner (Fig. 6). Because TRPC4␤ is insensitive to PIP 2 , the strong inhibition of mI CAT may seem puzzling, yet this is as expected if both isoforms form native channels, assuming that even one ␣-subunit in the tetramer is sufficient for its PIP 2 inhibition. If both isoforms are expressed equally and combine randomly, the chance that all four subunits are TRPC4␤ will only be 1 of 16.
Our present results extend novel insights into the common logic of TRP regulation by PIP 2 and other PIs and a voltage connection recently provided by Nilius and co-workers (6,24,43) to TRPC4 regulation. Testing the effects of IP 6 and other PIs, we found that the inhibition was specific to PIP 2 , thus suggesting that it was not simply a negative charge effect. Moreover, IP 6 with its highest surface negative charge selectively potentiated I TRPC4 at negative potentials, whereas one of the isoforms of PIP 2 , PI(3,4)P 2 , was without effect and the other one, PI(3,5)P 2 , potentiated I TRPC4 at positive potentials (Fig. 8).
Interestingly, the ⌬84AA segment displayed weaker binding to PIP 2 than the entire C terminus of TRPC4␣ (Ct, 733-974) (Fig. 4E), suggesting that residues outside of the ⌬84AA region also contribute to the lipid binding. Together with four amino acids upstream from the beginning of the ⌬84AA region is a fragment, 776 KSQSEGNGKDKRK 788 , that confers the (R/K)X(3,11)(R/K)X(R/K)(R/K) (X is any amino acid) role of the TRPM4 PIP 2 -binding site, believed to represent the "pleckstrin homology" (PH) domain (32). However, of the two maltosebinding fusion proteins that contain this fragment, Ala 765 -Ala 804 and Glu 775 -Pro 801 , neither showed detectable binding to PIP 2 -agarose in the in vitro binding assay (data not shown). Therefore, PIP 2 binding most likely involves other regions of the ⌬84AA segment, where there are two clusters of positive charges at the C-terminal half. In other TRPs, a modular C-terminal domain containing eight positively charged residues in TRPV1 (42), three positive charges in the conserved TRP box and TRP domain in TRPM8 (34), and the C-terminal PH domain in TRPM4 (35) have been demonstrated to be PIP 2interacting sites. For TRPC6, the PI-binding motif overlaps with the previously identified CaM/inositol 1,4,5-trisphosphate receptor binding (CIRB) site (53), and it contains six positively charged residues with three being important for binding to PIs (7). In case of TRPC4, PIP 2 binding could involve discontinuous regions because the ⌬84AA segment showed weaker binding than the full C terminus. There are additional positively charged residues, 6 upstream and 14 downstream, from the ⌬84AA region, with some clustering on either side. Exactly how the TRPC4␣ C terminus folds to form the PIP 2 -binding pocket is unknown. However, based on what is known about PIP 2 binding to PH domains, positive charges and some hydrophobic residues are important (54). These residues may not be all contained within the ⌬84AA region, and they do not have to be continuous. It is possible that deletion of the ⌬84AA segment, such as in the case of TRPC4␤, changes the conformation of the whole C terminus which eliminates its ability to bind PIP 2 .
PH domains are commonly known to bind PIs. These motifs are about 100 residues long with minimal sequence homology but conserved three-dimensional structures of widely variable affinities and selectivities for PIs. For example, PLC-␦ 1 PH domain binds PI(3,4,5)P 3 even somewhat stronger than PIP 2 or PI(3,4)P 2 , whereas pleckstrin PH domain binds PI(3,4)P 2 only weakly, and the PH domain of Sos does not bind PI(3,4,5)P 3 (52). The ⌬84AA region does not confer the main feature of a typical PH domain in that it contains mostly ␣-helices. However, there exist two CaM-binding sites, implicating that PIP 2 may compete with CaM for channel regulation like in the case of TRPC6 (7). On the other hand, the site of CaM and PIs competition for TRPC6 is at the CIRB site, which is more upstream of the ⌬84AA region and common to all TRPCs (53) (Fig. 1B). Unlike the CIRB site, the two CaM-binding sites within the ⌬84AA regions are unique to TRPC4␣ with the more C-terminal one being homologous to a similar site in the closely related TRPC5 (55). Given that CaM binding to TRPC5 C terminus was not competed off by PIs (7), competition between CaM and PIP 2 for binding to TRPC4␣ is also unlikely. Therefore, the exact mechanisms and binding sites for the lipids regulating TRPC4 remain to be established, but at least in case of PIP 2 it may be narrowed to the ⌬84AA stretch and, very likely, adjacent sites in the cytosolic C terminus because the inhibition, as well as the binding, was absent in TRPC4␤ lacking this sequence (Fig. 4, B and D).
In conclusion, we suggest a novel mechanistic model of TRPC4 gating, whereby TRPC4 interaction via its C-terminal PDZ-binding domain with the adaptor NHERF, ERM proteins, and cortical actin is necessary for keeping the ⌬84AA stretch close to the inner surface of the plasma membrane, thus stabilizing its binding with PIP 2 (Fig. 5D). According to the model, these multiple interactions stabilize the inactive conformation of TRPC4␣, whereas depletion of PIP 2 and/or cytoskeleton rearrangement relieve the channel from this inactivation. However, this alone is not sufficient for TRPC4␣ activation as other factors (e.g. G i/o and intracellular Ca 2ϩ , or additional components of PLC signaling) need to synergize with PIP 2 removal for an efficient channel opening.