The Ca2+‐activated cation channel TRPM4 is regulated by phosphatidylinositol 4,5‐biphosphate

Bernd Nilius, Frank Mahieu, Jean Prenen, Annelies Janssens, Grzegorz Owsianik, Rudi Vennekens, Thomas Voets

Author Affiliations

  1. Bernd Nilius*,1,
  2. Frank Mahieu,
  3. Jean Prenen,
  4. Annelies Janssens,
  5. Grzegorz Owsianik,
  6. Rudi Vennekens and
  7. Thomas Voets
  1. 1 Department of Physiology, Campus Gasthuisberg
  1. *Corresponding author. Laboratory of Physiology, Campus Gasthuisberg, Herestraat 49, KU Leuven, 3000 Leuven, Belgium. Tel.: +32 16 34 5937; Fax: +32 16 34 5991; E-mail: bernd.nilius{at}


Transient receptor potential (TRP) channel, melastatin subfamily (TRPM)4 is a Ca2+‐activated monovalent cation channel that depolarizes the plasma membrane and thereby modulates Ca2+ influx through Ca2+‐permeable pathways. A typical feature of TRPM4 is its rapid desensitization to intracellular Ca2+ ([Ca2+]i). Here we show that phosphatidylinositol 4,5‐biphosphate (PIP2) counteracts desensitization to [Ca2+]i in inside‐out patches and rundown of TRPM4 currents in whole‐cell patch‐clamp experiments. PIP2 shifted the voltage dependence of TRPM4 activation towards negative potentials and increased the channel's Ca2+ sensitivity 100‐fold. Conversely, activation of the phospholipase C (PLC)‐coupled M1 muscarinic receptor or pharmacological depletion of cellular PIP2 potently inhibited currents through TRPM4. Neutralization of basic residues in a C‐terminal pleckstrin homology (PH) domain accelerated TRPM4 current desensitization and strongly attenuated the effect of PIP2, whereas mutations to the C‐terminal TRP box and TRP domain had no effect on the PIP2 sensitivity. Our data demonstrate that PIP2 is a strong positive modulator of TRPM4, and implicate the C‐terminal PH domain in PIP2 action. PLC‐mediated PIP2 breakdown may constitute a physiologically important brake on TRPM4 activity.


Phosphoinositides (PI) are ubiquitously used signaling molecules in eukaryotic cells. Phosphatidylinositol 4,5‐biphosphate (PIP2) comprises about 3% of the total acidic membrane lipids and more than 99% of the doubly phosphorylated PI in a mammalian cell (McLaughlin et al, 2002). The list of ion channels that are regulated by PIP2 includes inward‐rectifier and voltage‐gated K+ channels, the two‐P domain K+ channels, voltage‐gated Ca2+ channels, cyclic nucleotide‐gated channels, intracellular Ca2+ ([Ca2+]i) release channels and the epithelial Na+ and Cl channels (for a recent review, see Suh and Hille, 2005).

More recently, it was shown that several members of the transient receptor potential (TRP) channel superfamily are also regulated by PIP2. TRPV1 is tonically inhibited by PIP2 (Chuang et al, 2001; Prescott and Julius, 2003), whereas TRP channel, melastatin subfamily (TRPM)5 (Liu and Liman, 2003), TRPM7 (Runnels et al, 2002), TRPM8 (Liu and Qin, 2005; Rohacs et al, 2005) and TRPV5 (Lee et al, 2005; Rohacs et al, 2005) are all activated in the presence of PIP2. Rohacs et al (2005) proposed a general role for the proximal C‐terminal TRP domain in PIP2 regulation of TRPM8 and other PIP2‐activated TRP channels, whereas Prescott and Julius (2003) identified a more distal C‐terminal region as a crucial determinant of PIP2 inhibition.

TRPM4 is a widely expressed member of the melastatin subfamily of TRP channels. It functions as a Ca2+‐activated and voltage‐dependent monovalent cation channel that depolarizes the plasma membrane and thereby modulates Ca2+ influx through Ca2+‐permeable pathways (Launay et al, 2002; Hofmann et al, 2003; Nilius et al, 2003). This physiological role was nicely illustrated in T‐lymphocytes, where TRPM4 provides a negative feedback on Ca2+ entry, thereby allowing the Ca2+ oscillations that control T‐cell cytokine release (Launay et al, 2004). A hallmark feature of TRPM4 is its rapid desensitization to [Ca2+]i (Nilius et al, 2004; Zhang et al, 2005). The mechanisms that underlie this desensitization are poorly understood. It has been shown that the Ca2+ sensitivity of TRPM4 is regulated by ATP, PKC‐dependent phosphorylation and by binding of calmodulin at the C‐terminus (Nilius et al, 2005a).

Here we show that PIP2 strongly enhances TRPM4 activity, by increasing the channel's Ca2+ sensitivity and shifting its voltage dependence of activation towards negative potentials. During revision of this article, the group of Liman (Zhang et al, 2005) published comparable observations. In addition, we provide complex mechanistic insights into the PIP2‐dependent activation of the channel and present data that suggest a close interaction between TRPM4 and enzymes involved in the PIP2 metabolism. Finally, we demonstrate that the TRP domain is not crucial for the PIP2 effect on TRPM4, and identify positively charged amino acids in a C‐terminal pleckstrin homology (PH) domain as important determinants of PIP2 action. Thus, our data provide further evidence that PIP2 is a general regulator of TRP channels, and implicate the C‐terminal PH domain of TRPM4 in PIP2 sensing.


Effect of the phospholipase C inhibitor U73122 and PIP2 on TRPM4 desensitization

A typical feature of whole‐cell TRPM4 currents activated by high [Ca2+]i is the rapid decay, which reflects a progressive decrease in the Ca2+ sensitivity of the channel (Figure 1A; see also Hofmann et al, 2003; Nilius et al, 2003, 2004, 2005a). We first examined whether Ca2+‐dependent activation of a phospholipase C (PLC) leading to PIP2 depletion may underlie desensitization of TRPM4 (as suggested in Rohacs et al, 2005). Addition of the aminosteroid derivative U73122, an inhibitor of receptor‐mediated PLC (Bleasdale et al, 1990; Smith et al, 1990), strongly attenuated the desensitization after activation by Ca2+ (Figure 1B). When U73122 was present in the extracellular solution before Ca2+‐induced activation, the typical outward rectification of the current was also strongly reduced (Figure 1A and B, right panels). Application of U73122 after maximal current decay led to a significant reversal of the desensitization process (Figure 1C; Irecovery/Imax=0.72±0.13, n=6, measured at +100 mV). U73343, the homolog of U73122 that is ineffective on PLC, did not change desensitization of TRPM4 (data not shown). Furthermore, inclusion of the none‐metabolizable PIP2 diC8‐PIP2 (from hereon referred to as PIP2) in the intracellular solution resulted in full recovery of the TRPM4 current after the initial decay (Figure 1D; Irecovery/Imax=1.06±0.14, n=8, +100 mV).

Figure 1.

Effects of the PLC blocker U73122 and PIP2 on whole‐cell currents through TRPM4. (A) Time course of the whole‐cell currents after activation by 3 μM Ca2+ via the patch pipette. Shown are the currents obtained from voltage ramps at −100 (triangles) and +100 mV (circles), respectively. The IV curves at the right side panel are from the currents indicated by solid circles. Arrows mark the onset of the whole‐cell mode. (B) Same experiment as in (A), but now in the presence of 10 μM U73122. Note the complete absence of desensitization and the nearly linear IV curves (right‐side panel, current points are marked at the time course). (C) Same experiment as in (A). After desensitization, U73122 was applied to the bath. Note the slow recovery of the current traces (IV curves from ramps as in (A). (D) Whole‐cell configuration was established with 3 μM Ca2+ and 10 μM diC8‐PIP2 in the pipette. After initial desensitization, the currents completely recovered (traces as the right‐hand side from the indicated points).

In inside‐out patches, Ca2+‐activated TRPM4 currents rapidly decay to a non‐zero steady‐state level, due to a time‐dependent reduction of the channel's Ca2+ sensitivity (Nilius et al, 2004). With 100 μM Ca2+, currents at +100 mV decayed to approximately one‐third of the maximal current after excision (Iss/Imax=0.35±0.09, n=5; see Figure 2A). Application of 10 μM PIP2 to the cytosolic side of the inside‐out patch resulted in a fast and full recovery of the current (Irecovery/Imax=1.07±0.15, n=5; Figure 2A) accompanied by an apparent loss of the voltage‐dependent kinetics (Figure 2B). Recovery from desensitization by application of PIP2 occurred with an EC50 of 5.1 μM (Figure 2C and D), similar to the EC50 reported elsewhere (Zhang et al, 2005). Washout of PIP2 led to slow current decay to a level similar to that obtained for desensitization in the absence of PIP2 (Iss/Imax=0.28±0.11, n=4; compare Figure 2C and E). When 10 μM PIP2 was applied to the bath before patch excision, desensitization was no longer observed (Iss/Imax=0.97±0.09, n=6; Figure 2E). Likewise, blockade of PLC activity using 10 μM U73122 fully prevented desensitization of TRPM4 in inside‐out patches (Iss/Imax=0.95±0.07, n=5; Figure 2E), whereas U73343 was ineffective (Iss/Imax=0.32±0.09, n=3). The effect of U73122 on the desensitization process was dose‐dependent, with a concentration for half‐maximal effect (EC50) of 0.31 μM (Figure 2F).

Figure 2.

PIP2 and U73122 counteract desensitization of TRPM4 in inside‐out patches. (A) TRPM4 current measured at −100 and +100 mV using the step protocol in an inside‐out patch excised in a solution containing 100 μM Ca2+. Note that desensitization reaches a steady‐state level. Application of 10 μM PIP2 reverses the desensitization. (B) Current traces obtained at the time‐points indicated in (A). Note the nearly complete disappearance of the time dependence after PIP2 application. Same scaling as in figure (A). Protocol used to determine the dose dependence of the action of PIP2 on TRPM4 (C) and resulting dose–response curve (D). The increase in outward current induced by a specific PIP2 concentration was normalized to that obtained with 50 μM PIP2. (E) PIP2 and U73122 applied before patch excision prevent desensitization of TRPM4. (F) Dose–response curve for the effect of U73122 on TRPM4 desensitization. Each data point is from at least four independent measurements. Data points represent the ratio between the steady‐state current and the maximal current measured immediately after application of Ca2+.

When after maximal desensitization Ca2+ was removed and subsequently re‐added, recovery was never observed. Likewise, U73122 was ineffective on desensitized inside‐out patches (data not shown). However, when during the Ca2+‐free period Mg‐ATP was present at the cytosolic side, currents through TRPM4 recovered to nearly the same size as the maximal current after excision (Nilius et al, 2005a), and this recovery process was further potentiated by U73122 (Supplementary Figure S1). We conclude that Mg‐ATP can reverse desensitization of TRPM4 by restoring the PIP2 levels in the membrane patch. Indeed, Mg‐ATP is known to activate phosphatidylinositol‐4‐kinases (PI‐4‐K), which regenerate PIP2 from PI(4)P (Balla, 2001; Hilgemann et al, 2001). These results also indicate that both PI‐4‐K and PLC activities remain preserved in TRPM4‐containing inside‐out patches.

To determine whether other PI have a similar effect on TRPM4, we tested seven different PI‐phosphates (Supplementary Figure S2). The following potency order was determined in inside‐out patches (300 μM [Ca2+]i): PI(4,5)P2>PI (3,4,5)P3=PI(3,5)P2=PI(3,4)P2=PI(5)P>PI(4)P=PI(3)P. Given that PIP2 is both the most potent and the most abundant PI‐phosphate, it is likely to represent the physiologically most important regulator of TRPM4.

It should be noted here that PIP2 or U73122 did not have any effect on inside‐out patches from nontransfected cells or on TRPM4‐containing patches in the absence of activating Ca2+, excluding nonspecific actions of these compounds (data not shown). Moreover, we confirmed that 10 μM PIP2 also caused recovery of TRPM4 endogenously expressed in umbilical vein‐derived EA cells (Supplementary Figure S3).

PIP2 and U73122 shift the voltage and Ca2+ dependence of TRPM4 activation

Many regulators of TRPM4 alter the voltage dependence of channel activation (Nilius et al, 2005a). Therefore, we examined the effects of PIP2 and U73122 on the voltage dependence of TRPM4. In the absence of U73122 and PIP2, hyperpolarizing voltage steps from a holding potential of 0 mV caused current deactivation, whereas activation was evident at positive potentials (Figure 3A). Nearly steady‐state current–voltage (IV) relationships obtained at the end of 400‐ms voltage steps were fitted with a function that combines a linear conductance multiplied by a Boltzmann activation term:

Figure 3.

PIP2 modulates the voltage dependence of TRPM4 in inside‐out patches. (A) Current measured in response to the indicated voltage protocol in the absence of PIP2. (B) Voltage dependence of steady‐state and tail current from the experiment shown in panel (A). (C) Voltage dependence of the open probability obtained from normalization of the tail current values to Imax. (D–F) Same as panel (A–C), but now in the presence of 10 μM PIP2. (G) Activation curves for TRPM4 in control conditions and in the presence of either 10 μM PIP2 or 10 μM U73122. Each data point is from 2–4 independent measurements. Solid lines represent fits using equation (3). (H) Average values for V1/2 and z obtained from experiments as in panels (A–F).

Embedded Image

where g is the whole‐cell conductance, Erev is the reversal potential, V1/2 is the potential for half maximal activation and s is the slope factor. The slope factor s is related to the effective gating charge z of the voltage sensor according to

Embedded Image

where R is the gas constant, T the absolute temperature and F the Faraday constant. As an alternative way to determine the voltage‐dependent gating parameters, tail currents at −100 mV were measured and normalized to the maximal tail current Imax (Figure 3A and B), yielding the voltage dependence of the steady‐state open probability of TRPM4 (Figure 3C). Normalized tail currents were fitted using the equation

Equivalent values of s and V1/2 were obtained by either method. As described in detail elsewhere, the s for TRPM4 is quite high, indicating that z is small (<1) (Figure 3; Nilius et al, 2005b).

Application of 10 μM PIP2 led to an almost complete loss of the time dependence of TRPM4 activation at positive potentials, a dramatic slowing of current deactivation at negative voltages and significant steady‐state inward currents (Figure 3D and E). The apparent steady‐state open probability was shifted towards negative potentials and was shallower than in the absence of PIP2 (Figure 3F–H). Virtually identical changes in voltage dependence were observed upon application of U73122 (Figure 3E and F and Supplementary Figure S4).

As described already in detail (Nilius et al, 2004), the decay of the TRPM4 current in inside‐out patches is faster and more complete at low [Ca2+]i, indicating that the Ca2+ sensitivity of the channel decreases after patch excision (Figure 4A). Inward currents are small (Figure 4B). Addition of 10 μM PIP2 to the patch restored the Ca2+ sensitivity. Under persistent PIP2 application, maximal TRPM4 current was already obtained at a [Ca2+]i of 10 μM. Inward currents are large (note the relatively large currents at 0.5 μM [Ca2+]i; Figure 4C and D). We quantified this effect by normalizing the steady‐state current to the peak current obtained immediately upon patch excision in the different Ca2+ concentrations. The resulting values were fitted by a dose–response curve of the form:

Figure 4.

PIP2 increases the Ca2+ sensitivity of TRPM4 in inside‐out patches. (A) Representative time course of TRPM4 currents at −100 and +100 mV in an inside‐out patch, illustrating the Ca2+ dependence of TRPM4 activation after desensitization. (B) Original traces from voltage steps obtained from different [Ca2+]i indicated in (A). (C) Same as in (A), but in the presence of 10 μM PIP2. (D) Original traces from voltage steps obtained from different [Ca2+]i indicated in (C). (E) Dose dependence of TRPM4 activation by Ca2+ in the desensitized state and after addition of 10 μM PIP2. The EC50 value for Ca2+ activation changed from 134 μM (control, open circles, nH=0.9; see also Nilius et al, 2004) to 1.3 μM (nH=1.0).

Embedded Image

where EC50 represents the concentration for half‐maximal TRPM4 activation and nH the Hill coefficient. As shown in Figure 4E, application of 10 μM PIP2 shifted the EC50 value for Ca2+ from 134±21 μM (control) to 1.3±0.2 μM. A similar 10‐fold increase in Ca2+ sensitivity was observed in the presence of U73122 (Figure 4E and Supplementary Figure S5). Interestingly, channel closure upon removal of Ca2+ was severely slowed down by PIP2 and U73122. In the absence of these compounds, perfusion with an EGTA‐containing Ca2+‐free solution led to a virtually immediate inactivation of current (time constant <1 s). In contrast, in the presence of PIP2 or U73122, removal of Ca2+ led to a slow decay of the currents, with exponential time constants in the range of 6–20 s, n=4 (see Supplementary Figure S6). This suggests that PIP2 decreases the rate of Ca2+ unbinding from the channel.

Inhibition of TRPM4 by various PIP2‐depleting protocols

Wortmannin, an inhibitor of PI‐4‐K, retards the replenishment of PIP2, leading to depletion of the intracellular PIP2 pool (Nakanishi et al, 1995). In the whole‐cell mode, inclusion of 50 μM wortmannin in the patch pipette reduced the amplitude of the Ca2+‐activated current and led to a faster current decay (Figure 5A–C). Moreover, whole‐cell TRPM4 currents were almost fully absent in cells pretreated for 25 min with 50 μM wortmannin (Figure 5D). Next, we tested whether the effect of wortmannin could be reversed by direct application of PIP2. Indeed, after a 25‐min pretreatment with 50 μM wortmannin, currents in inside‐out patches ([Ca2+]i=300 μM) were strongly decreased (Figure 5E–G), and subsequent application of 10 μM PIP2 led to a strong current increase (Figure 5E–G).

Figure 5.

Wortmannin and poly‐l‐lysin attenuate currents through TRPM4 and reversion by PIP2 in inside‐out patches. (A) Time course of whole‐cell currents measured as described in Figure 1 (circles +100 mV, triangles −100 mV). Activation of the current was established by 3 μM [Ca2+]i in the patch pipette. (B) Current activation as in (A). Wortmannin (WM, 50 μM, same calibration as in (A)) was included in the patch solution. Note the much smaller currents and the faster desensitization as compared with (A). Preincubation with 50 μM WM for 25 min completely prevented activation of TRPM4 currents with 3 μM [Ca2+]i (data not shown). (C) Pooled data from the maximal current obtained after breaking into the cells with 300 μM [Ca2+]i (supramaximal concentration for TRPM4; Ullrich et al, 2005). Note the decrease in the activated current by WM applied via the pipette and the nearly complete attenuation of the current for 25‐min preincubation. (D) From the time courses of current activation with 300 μM [Ca2+]i, the time to 50% desensitization was measured. This time was highly significantly shortened by application of WM into the pipette (t1/2 values not measurable for 25‐min preincubation). (E) Currents measured from inside‐out patches (300 μM [Ca2+]i). Before patch excision, cells were incubated for 25 min in 50 μM WM. Note that the first current after excision is very small (see panel F). Application of PIP2 restored the currents immediately. Typically, washing out of Ca2+ resulted in a delayed current decay (circles +100 mV, triangles −100 mV). Traces at the times indicated in the time course are shown at the right‐hand side. (F) Averaged current size under control conditions (no WM preincubation, 300 μM [Ca2+]i measured at +100 mV) and after 25‐min preincubation with 50 μM WM. (G) Comparison of the first current after excision (preincubation with WM) with the maximal current after 10 μM PIP2 application (cont: PIP2 application without WM, currents measured at +100 mV). (H) Time course of TRPM4 desensitization in inside‐out patches (300 μM [Ca2+]i). The time course at +100 mV is shown (same step protocol as in (A)). Poly‐l‐lysin (PLL) is added to the inner side of the membrane in an inside‐out patch. PLL (10 μg/ml) already blocked the current completely. Reapplication of 10 μM PIP2 recovered the current to nearly the same size as the steady‐state value. (I) Current traces for the indicated points in the time course shown in (H). (J) Concentration–response curve for the PLL block. Data were fitted by equation (4). The IC50 value is 0.6 μg/ml (nH=1.1). (K) Recovery from desensitization measured from experiments as shown in (H), in which 10 μM PIP2 was applied after complete block by PLL. Steady‐state current (e.g. trace b in panel H) was obtained from the stationary current divided by the maximal current (Iss/Imax)). The PIP2 recovered current was measured by normalizing the maximal recovered current with the maximal current immediately after Ca2+ application, Irecovery/Imax. Both values are not statistically significant.

Poly‐l‐lysine (PLL) is a positively charged macromolecule that acts as a scavenger of PIP2 (Lopes et al, 2002; Zhang et al, 2003). In inside‐out patches, TRPM4 currents were rapidly inhibited when PLL was applied during the steady‐state phase (Figure 5H and I). Inhibition by PLL was dose‐dependent, and half‐maximal inhibition was achieved at a concentration of 0.6 μg/ml (Figure 5J). TRPM4 currents were not restored upon washout of PLL, but subsequent application of PIP2 led to partial reactivation of the channel (Figure 5H and K).

Inositol polyphosphate 5‐phosphatases (5ptases) are enzymes that remove the phosphate group at the D5 position of PIP2, leading to significant depletion of intracellular PIP2 (Kisseleva et al, 2002). Using HEK cells with tetracyclin‐inducible expression of 5ptase IV, we found that induction of this enzyme strongly reduces the peak and steady‐state TRPM4 currents in inside‐out patches (Supplementary Figure S7). Remarkably, application of PIP2 reactivated TRPM4 currents in only ∼40% of the induced cells, suggesting a high 5ptase activity in the close vicinity of TRPM4 channels.

Finally, we tested the effect of hormonal stimulation of PLC‐coupled receptors on TRPM4 channel function. For this purpose, we used Chinese hamster ovary (CHO) cells permanently expressing the muscarinic receptor M1 (Buckley et al, 1989) and brief pulses of ionomycin (ION; 2 μM) to repetitively activate TRPM4 currents. In control conditions, a second pulse of ION applied 60–90 s after the first application led to a significant increase of the TRPM4 current at +100 mV (Figure 6A and C). In contrast, activation of the M1 receptor with 100 μM acetylcholine (ACh) after the first ION pulse led to a strong reduction of the current induced by the second ION pulse (Figure 6B and C). Under this condition, a third ION pulse applied 150–180 s after the washout of ACh revealed a partial recovery of the current (Figure 6B and C).

Figure 6.

Activation of the M1 receptor inhibits TRPM4 in M1 receptor‐expressing CHO cells. (A) Repetitive activation of TRPM4 by 2 μM ION. The IV curves (right) were obtained at the indicated time points (left). (B) Same protocol as in (A), but 100 μM ACh was applied between the first two ION applications. As a consequence, the second application of ION resulted in a much smaller current. A third ION application revealed partial recovery. (C) Averaged data from experiments as in (A) and (B). Current at +100 mV during the second ION application was normalized to that of the first ION application. The normalized current after a 3‐min recovery period (recov) is also indicated. (D) Control currents in inside‐out patches excised in 300 μM [Ca2+]i. Traces indicated in the time course are shown at the right (the same current calibration). (E) Same as in (D), but now obtained from cells preincubated for 60 s with 100 μM ACh. (F) Summary of the experiments on inside‐out patches shown in (D) and (E). Both the peak currents (obtained immediately after patch excision) and steady‐state currents were significantly decreased after preincubation with ACh (P<0.05, n=12). Shown are outward currents at +100 mV (columns up) and −100 mV (columns down).

We also tested whether M1 activation before patch excision would affect the current size in inside out patches. A 60‐s preincubation with 100 μM ACh significantly reduced the first current measured after patch excision in 300 μM [Ca2+]i (Figure 6D–F). Additionally, desensitization of the currents in pretreated cells was faster and more complete then in untreated cells (compare Figure 6D and E). Thus, PLC‐coupled receptor activation mimics the effects of PIP2 depletion on TRPM4.

The C‐terminal PH domain is a putative PIP2 interacting site

Rohacs et al (2005) recently reported that mutating positively charged residues in the TRP box and TRP domain of TRPM8 led to a 10–100‐fold reduction in the PIP2 sensitivity. Moreover, equivalent mutations in the TRP domain of TRPM5 and TRPV5 increased the sensitivity of these channels to wortmannin, suggestive of a reduced PIP2 sensitivity. They concluded that the TRP domain may be a general PIP2‐interacting site in different TRP channels. We investigated whether the corresponding mutations in TRPM4, K1059Q and R1062Q in the TRP box and R1072Q in the TRP domain affect the response to PIP2 in inside‐out patches (Figure 7A). In comparison to WT TRPM4, R1062Q and R1072Q displayed a more complete desensitization, with steady‐state outward current amplitudes that amounted to less than 5% of the peak current amplitude (Figure 8C–E), whereas K1059Q currents decayed to a similar steady‐state level as WT TRPM4 (Figure 8A, B and E). Surprisingly, application of 10 μM PIP2 after maximal desensitization resulted in robust reactivation of all three mutants (Figure 8A–D). The degree of recovery, quantified as Irecovery/Imax, was not significantly different between WT TRPM4 and the three TRP domain mutations (Figure 8E). These data indicate that the TRP box and TRP domain are not the main determinants of PIP2 action, and urged us to search for alternative PIP2 interacting sites in TRPM4.

Figure 7.

Schematic representation of mutations within putative PIP2‐binding regions of TRPM4. (A) Pairwise alignment of TRP domains of TRPM4 and TRPM5. Identical residues are shown in red within the yellow boxes, homologous residues in black within the green boxes. Note that the positively charged residues are conserved. Residues mutated in this study are marked by asterisks and are identical with the TRP box and TRP domain mutations used by Rohacs et al (2005). (B) Comparison of two putative PH domains in TRPM4 (dotted lines) with the corresponding region in TRPM5. Note the loss of the PH domains in TRPM5. Asterisks with numbers denote TRPM4 residues mutated in this work (for details, see Materials and methods).

Figure 8.

Effects of PIP2 on TRP domain mutants in inside‐out patches. (A) Time course showing WT TRPM4 current desensitization at −100 and +100 mV after excision in 300 μM [Ca2+]i (same protocol as in Figure 2), followed by reversal of desensitization upon application of 10 μM PIP2. To the right, raw current traces obtained at the indicated time points are shown using the same current calibration. (B–D) Same as in panel (A), but now for the indicated TRP domain mutants. (E) Averaged data showing the steady‐state current before PIP2 application (Iss/Imax) and the current after PIP2‐induced recovery (Irecovery/Imax), both normalized to the maximal current immediately after excision. Note that even the completely desensitizing mutants show unchanged recovery with PIP2.

The C‐terminus of TRPM4 contains two regions of positively charged residues that obey the consensus sequence of PH domains ([R/K]–X3−11–[R/K]–X–[R/K]–[R/K], where X is any amino acid; see; Figure 7B), which are known as PIP2 interaction sites (Harlan et al, 1994). The first, most proximal putative PH domain, R1136ARDKR1141, was previously shown to determine modulation of TRPM4 by decavanadate (Nilius et al, 2004). Neutralization of all four basic residues in this region resulted in a mutant channel (termed ΔR/K) that exhibited fast and complete desensitization in inside‐out patches (Figure 9A). Importantly, direct application of 10 μM PIP2 never led to any current recovery (n=6), whereas 30 μM PIP2 induced to a small but significant recovery (Irecovery/Imax=0.29±0.1; n=4). Likewise, U73122 was unable to recover of the ΔR/K mutant. In contrast, charge‐neutralizing mutations in the second putative PH domain did not affect the PIP2 sensitivity of the channel, although these mutant channels exhibited significantly faster desensitization than WT TRPM4 (Figure 9B, C and E). From this, we conclude that the first putative PH domain is specifically involved in interaction with PIP2 (Figure 9D and E).

Figure 9.

Effects of PIP2 on putative PH domain mutants in inside‐out patches. (A) Time course showing current desensitization at −100 and +100 mV for the ΔR/K mutant after excision in 300 μM [Ca2+]i (same protocol as in Figure 2), followed by partial reversal of desensitization upon application of 30 μM PIP2. To the right, raw current traces obtained at the indicated time points are shown using the same current calibration. PIP2 (10 μM), which was used in most other experiments, did not induce any recovery of this mutant (n=6, see E). (B) Same as (A), but now for the R1147Q mutant. Note the complete current recovery after application of 10 μM PIP2. (C) Same as (A), but now for the RKR1147XAAA mutant. Note the small and rapidly desensitizing initial currents. In this case, application of 10 μM PIP2 resulted in currents that were several‐fold larger than the initial peak current (i.e. Irecovery/Imax≫1; see panel E). (D) Same as (A), but now for the TRPM4/cM5 chimeric channel. Note the slow desensitization and the negligible recovery upon application of PIP2. (E) Averaged data showing the steady‐state current before PIP2 application (Iss/Imax) and the current after PIP2‐induced recovery (Irecovery/Imax), both normalized to the maximal current immediately after excision. PIP2 (10 μM) was used, unless noted otherwise.

As shown in Figure 7B, this first putative PH domain is not conserved in the closely related TRPM5. To test the consequence of this sequence divergence, we compared the PIP2 sensitivity of TRPM4 with that of TRPM5, and of a chimera combining the N‐terminus and transmembrane domains of TRPM4 with the C‐terminus of TRPM5 (TRPM4/cM5 chimera). Excision in 300 μM [Ca2+]i of inside‐out patches from cells expressing TRPM4/cM5 led to robust current activation followed by nearly complete desensitization, which, however, is slower than for TRPM4 (Figure 9D and E; see also Liu and Liman, 2003; Ullrich et al, 2005). Application of 10 μM PIP2 resulted in only a very small or no recovery of the current through TRPM4/cM5 chimera, compared to the complete recovery observed for TRPM4 under the same conditions (Figure 9D and E). Moreover, application of U73122 did not lead to any current recovery for TRPM5 or TRPM4/cM5 channels. Thus, the absence of the first PH domain in TRPM5 or the TRPM4/cM5 chimera correlates with the reduced PIP2 sensitivity of these channels.


Numerous intracellular signaling pathways depend on the activity of PLC, which hydrolyzes PIP2 into membrane‐bound diacyl glycerol and soluble inositol trisphosphate, both potent intracellular signaling molecules. Moreover, PIP2 itself acts as a regulator of signaling proteins in the plasma membrane, including a growing number of TRP channels (Chuang et al, 2001; Hardie et al, 2001; Runnels et al, 2002; Liu and Liman, 2003; Lee et al, 2005; Liu and Qin, 2005; Rohacs et al, 2005; Zhang et al, 2005). Although PIP2 is the most abundant PI‐phosphates, it has to be taken into account that all other PI‐phosphates can also contribute to the modulation of TRPM4, as shown in Supplementary Figure S2. Compartmentalization and location with respect to the active channel will probably determine which PI‐phosphate exerts the main effect. The action of PIP2 on TRP channels can be either stimulatory, as is the case for TRPM5, TRPM7, TRPM8 and TRPV5 (Runnels et al, 2002; Liu and Liman, 2003; Lee et al, 2005; Rohacs et al, 2005), or inhibitory, as is the case for TRPV1 and Drosophila TRPL (Chuang et al, 2001; Estacion et al, 2001). Here we add the monovalent cation channel TRPM4 to the list of channels directly regulated by PIP2. During revision of this article, qualitatively similar observations were published by Liman and colleagues (Zhang et al, 2005). In addition, we illustrate how TRPM4 activity can be modulated by several approaches that modify the cellular PIP2 levels, including PLC‐coupled M1 receptor activation, pharmacological inhibition of PLC activity, modulation of PIP2 metabolism and scavenging of PIP2. Finally, we present evidence that a distal C‐terminal PH domain rather than the proximal C‐terminal TRP domain is the major determinant of the PIP2 sensitivity of TRPM4.

We found that PIP2 is unable to gate TRPM4 when [Ca2+]i is buffered at low levels, indicating that PIP2 does not act as a direct TRPM4 activator. Instead, PIP2 acts as a modulator of the channel's sensitivity to both Ca2+ and voltage: increasing PIP2 levels cause a 100‐fold increase in Ca2+ sensitivity and a drastic shift of the voltage dependence of activation towards negative potentials, thereby strongly increasing the open probability of TRPM4 at physiological membrane potentials. We propose that the decay of TRPM4 currents in whole‐cell and inside‐out patches, which entails a loss of sensitivity to both Ca2+ and voltage, reflects the gradual decrease of cellular PIP2 levels due to a Ca2+‐dependent stimulation of PLC activity and/or diffusional loss. A possible scheme to explain the combined effects of Ca2+, voltage and PIP2 on TRPM4 activity is presented in Supplementary Figure S10.

The effects of PIP2 on the voltage dependence of TRPM4 activation are reminiscent of its effects on the voltage sensitivity of voltage‐gated channels from other channel families. For example, PIP2 increases the open probability of TREK1 and KCNQ1 channels by shifting the activation curve to more negative voltages (Lopes et al, 2005). In P/Q‐type Ca2+ channels, PIP2 has a dual effect on channel activity (Wu et al, 2002). First, it shifts the activation curve of P/Q‐type Ca2+ channels towards more positive potentials, thereby reducing channel opening at physiological potentials. Second, it causes a voltage‐independent stabilization of these P/Q‐type Ca2+ channels (Wu et al, 2002). In cloned and native N‐type Ca2+ channels, PIP2 strongly attenuated and reversed rundown of expressed channels (Gamper et al, 2004). Interestingly, we found that PIP2 did not only shift the midpoint (V1/2) of the activation curve of TRPM4 but also reduced its slope (s). This may suggest that PIP2 reduces the effective gating charge, which could be explained by screening of voltage‐sensing basic residues or by altering the electrical field around the voltage sensor (McLaughlin et al, 2002). Alternatively, PIP2 may alter the coupling between the voltage sensor and the channel gating. Regardless of the mechanism, shifts of the voltage‐dependent activation curve appear to represent a general activating principle whereby voltage‐dependent TRP channels respond to thermal or chemical stimuli (Voets et al, 2004; Nilius et al, 2005b).

At present, we can only speculate about the mechanism whereby PIP2 provokes a 100‐fold increase in the apparent Ca2+ affinity of TRPM4. We found that, in the presence of PIP2, washout of Ca2+ is dramatically slowed down, indicating that PIP2 hinders the unbinding of Ca2+ from its activatory binding site. One possibility would be that the negatively charged phosphate groups of membrane‐bound PIP2 constitute an integral part of a high‐affinity Ca2+‐binding site. Loss of PIP2 from the membrane would then result in a more rudimental, low‐affinity Ca2+‐binding site. Such a mechanism has, for example, been proposed to explain the increased Ca2+ affinity of synaptotagmin I, the putative Ca2+ sensor for rapid exocytosis, in the presence of PIP2‐containing phospholipid bilayers (Bai et al, 2004). Alternatively, PIP2 binding may have an allosteric effect on TRPM4's Ca2+‐binding site.

In previous studies, two distinct PIP2 interaction sites have been identified in TRP channels. First, a PIP2 interaction site comprising eight positively charged residues was found in the C‐terminus of TRPV1 (Prescott and Julius, 2003). Mutations to these residues strongly reduced the tonic inhibition by PIP2, leading to lowered thresholds for activation by heat or capsaicin. More recently, Rohacs et al (2005) identified three positive charges in the conserved TRP box and TRP domain as determinants of the interaction between PIP2 and TRPM8. Neutralization of these charges decreased the sensitivity of TRPM8 to activation by PIP2 and aggravated the inhibitory effects of PIP2 depletion. Equivalent mutations in TRPM5 and TRPV5 also increased the sensitivity of these channels to wortmannin, suggestive of a reduced PIP2 sensitivity, although these mutant channels were not directly probed with PIP2 (Rohacs et al, 2005). Surprisingly, when we made the corresponding mutations in the TRP domain of TRPM4, the activating effect of PIP2 appeared fully conserved. This indicates that, at least in TRPM4, the TRP domain is not the main PIP2 interaction site. We have to note, however, that two of the TRP domain mutants (R1062Q and R1072Q) exhibited accelerated desensitization, despite their normal PIP2 sensitivity. Moreover, in these two cases, desensitization could not be prevented by U73122. One possible explanation for these results would be that these mutations do not interfere with PIP2 binding, but rather alter the coupling between TRPM4 and PLC.

We identified two putative PH domains in the C‐terminus of TRPM4, and implicated the first (i.e. closest to TM6) PH domain in PIP2 activation. Neutralization of all four positively charged residues in this region resulted in a channel exhibiting very rapid desensitization and significantly reduced sensitivity to PIP2. As this PH domain is not conserved in the closely related TRPM5, we constructed a chimeric channel consisting of the N‐terminus and transmembrane domains of TRPM4 together with the C‐terminus of TRPM5, and found that this TRPM4/cM5 channel also displayed desensitization similar to TRPM5 (Ullrich et al, 2005) and has an extremely reduced PIP2 sensitivity. Interestingly, mutations to the first PH domain also abolish the effect of decavanadate on TRPM4 (Nilius et al, 2005a, 2005b), indicating that PIP2 and decavanadate act via a similar mechanism to alter the voltage‐ and Ca2+‐sensitivity of TRPM4. In contrast, mutations to the second C‐terminal PH domain were without effect on the channel's PIP2 sensitivity.

Interestingly, several lines of evidence suggest that some of the enzymes involved in PIP2 metabolism are tightly and functionally associated with TRPM4. Firstly, application of Ca2+‐free, Mg‐ATP‐containing solution to desensitized TRPM4 in inside‐out patches is able to fully reverse desensitization. This indicates that a PIP2‐regenerating kinase is present and functional in cell‐free patches containing TRPM4. Secondly, the reversal of desensitization by Mg‐ATP is potentiated by the PLC inhibitor U73122, suggesting that (Ca2+‐dependent) PLC remains active in inside‐out patches. Thirdly, some of the mutations to the TRP domain and to the second PH domain described in this work reduce the effect of PLC inhibition, without clear effects on PIP2 sensitivity. This indicates that residues in TRPM4 may have an effect on PLC activity. Finally, overexpression of the PIP2‐depleting enzyme 5ptase IV not only led to lower TRPM4 activity in the inside‐out patches but also reduced the effectiveness of exogenously applied PIP2 to restore channel activity. This latter result could be explained assuming a concentrated 5ptase activity in the close vicinity of TRPM4 channels. It is interesting to speculate that TRPM4 in the plasma membrane forms part of a signaling complex with PLC, PI‐4K and 5ptase, similar to the signalplex described in Drosophila rhabdomers, which consists of the cation channels TRP and TRPL together with components of the signal transduction pathways, such as PLC, PKC and calmodulin.

In conclusion, we have demonstrated that PIP2 enhances the activity of the Ca2+‐activated cation channels TRPM4 by decreasing its Ca2+ sensitivity and shifting its voltage dependence, and provided evidence that these effects require an intact C‐terminal PH domain. Activation of PLC‐coupled receptors is expected to have a dual effect on TRPM4 activity. On the one hand, PLC activity will lead to the production of IP3, which releases Ca2+ from intracellular stores, leading to a rise in [Ca2+]i and activation of TRPM4. On the other hand, PLC‐dependent hydrolysis of PIP2 will cause desensitization of TRPM4, thereby limiting its activity. PIP2‐dependent signaling emerges as a general mechanism for the regulation of diverse TRP channel functions.

Materials and methods

Cell culture

Human embryonic kidney (HEK) 293 cells and CHO cells expressing the human muscarinic M1 receptor were cultured as described in detail previously (Buckley et al, 1989). We used a T‐Rex tetracyclin‐regulated system (Invitrogen) for inducible expression of 5ptase IV in HEK293 cells (Kisseleva et al, 2002; Voets et al, 2004; Verbsky et al, 2005). To induce expression of 5ptase IV, 0.1 μg/ml tetracycline was added to the culture medium, and cells were used 12–16 h after induction. For transient expression of TRPM4, we used the bicistronic expression plasmid pdiTRPM4b, which carries the entire coding region for the human TRPM4b (accession number AX443227, kindly provided by Drs V Flockerzi and U Wissenbach) and for the green fluorescent protein, coupled with an internal ribosomal entry site sequence. HEK293 and CHO cells were transfected using TransIT‐293 transfection reagent (Mirus Corp., Madison, WI) and successfully transfected cells were identified by their green fluorescence (for details, see Nilius et al, 2003). Mutations were made using the standard polymerase chain reaction overlap extension technique (Ho et al, 1989).


The extracellular solution for cell attached measurements and the pipette solution for inside‐out patch clamp measurements contained (mM): 156 NaCl, 5 CaCl2, 1 MgCl2, 10 N‐(hydroxyethyl)piperazine‐N′‐2‐ethanesulfonic acid (HEPES), buffered at pH 7.4 with NaOH. Inside‐out patches were excised in a solution containing (mM): 156 CsCl, 1 mM MgCl2, 10 HEPES, pH 7.2 with CsOH. The Ca2+ concentration at the inner side of the membrane was adjusted between 100 nM and 5 μM by adding appropriate amounts of CaCl2 calculated by the CaBuf program ( to the 5 mM ethylene‐glycol‐O,O′‐bis(2‐aminoethyl)‐N,N,N′,N′‐tetraacetic acid (EGTA)‐containing solution. For Ca2+ concentrations between 10 μM and 1 mM, CaCl2 was added to an EGTA‐free solution. In all figures, ‘0Ca2+’ refers to a Ca2+‐free solution containing 5 mM EGTA. All internal solutions were ATP‐free. Experiments were performed at room temperature (22–25°C). DiC8‐PIP2 was purchased from Echelon Biosciences (Salt Lake City, UT) and wortmannin, PLL, U73343 and U73122 from Sigma (St Louis, MO).


Currents were measured in the whole‐cell and inside‐out configuration using an EPC‐9 patch‐clamp amplifier (HEKA Elektronik, Lambrecht, Germany). Patch electrodes had a DC resistance between 2 and 4 MΩ when filled with the different recording solutions. Currents were sampled at 2.5 kHz and filtered at 1 kHz. The step protocols consisted of a 400‐ms step to −100 mV from a holding of 0 mV, followed by a 250‐ms step to +100 mV applied at 0.5 Hz. The ramp protocol consisted of a 400‐ms ramp from −100 to +100 mV from a holding potential of 0 mV.

Data analysis

Electrophysiological data were analyzed using the WinASCD software (G Droogmans, Leuven, Belgium). Pooled data are given as mean±s.e.m. of n cells. Significance was tested using Student's paired t‐test (P<0.05).

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplementary Figures [emboj7600963-sup-0001.pdf]


We thank Drs K Talavera, D D'hoedt and C Frippiat (Leuven, Belgium) for very helpful discussions and Drs V Flockerzi and U Wissenbach (Homburg, Germany) for the TRPM4 plasmid. The 5ptase IV inducible 293‐Rex cells were kindly provided by Drs Majerus (St Louis, MO) and Douglas A Bayliss (Charlottesville, VI), and CHO cells expressing the human muscarinic M1 receptor were kindly provided by Dr NY Buckley (Leeds, UK). This work was supported by the Human Frontiers Science Programme (HFSP Research Grant Ref. RGP 32/2004), the Belgian Federal Government, the Flemish Government and the Onderzoeksraad KU Leuven (GOA 99/07, FWO G.0214.99, FWO G. 0136.00; FWO G.0172.03, Interuniversity Poles of Attraction Program, Prime Ministers Office IUAP).