This report was presented at The Journal of Physiology Symposium on Regulation of ion channels and transporters by phosphatidylinositol 4,5-bisphosphate (PIP2), Baltimore, MD, USA, 2 March 2007. It was commissioned by the Editorial Board and reflects the views of the author.
Corresponding author Bertil Hille, Department of Physiology and Biophysics, University of Washington School of Medicine, G-424 Health Sciences Building, Box 357290, Seattle, WA 98195-7290, USA. Email: email@example.com
Activation of phospholipase C (PLC) through G-protein-coupled receptors produces a large number of second messengers and regulates many physiological processes. Many membrane proteins including ion channels require the phosphoinositide phosphatidylinositol 4,5-bisphosphate (PIP2) to function. Activation of PLC can shut down their activity if it depletes the PIP2 pool strongly. Such a mechanism accounts for the muscarinic suppression of current in KCNQ channels. We describe a variety of methods used to show that these channels require PIP2 and that current in the channels is suppressed when receptor-activated PLC depletes PIP2. The methods include observing translocation of lipid-sensitive protein domains, overexpression of enzymes of phosphoinositide metabolism, engineering these enzymes to move to the plasma membrane in response to a chemical signal, and direct chemical analysis of phospholipids. These approaches are general and can be used to test for PIP2 requirements of other membrane proteins.
Phosphoinositides are minor phospholipids in all cellular membranes. Their primary role is signalling rather than membrane structure. There are seven phosphoinositide isomers differing in phosphorylation at the 3, 4 and 5 positions of the inositol ring, with each cellular compartment having a unique combination of phosphoinositides that serves almost as a ‘zip code’ directing protein activities in that membrane (Di Paolo & De Camilli, 2006). Phosphatidylinositol 4,5-bisphosphate (PIP2) is a marker of the plasma membrane. It comprises only a few percent of total cellular acidic lipids. PIP2 has long been famous as the precursor of two widely studied second messengers, diacylglycerol (DAG) and inositol trisphosphate (IP3), produced when PIP2 is cleaved by phospholipase C (PLC). However, more recently a direct PIP2 requirement has been recognized for the normal function of many intrinsic and peripheral proteins of the plasma membrane, including certain ion channels, transporters, trafficking molecules and nucleation proteins of the cytoskeleton (Hilgemann et al. 2001; Suh & Hille, 2005). This requirement may keep such proteins in an inactive state until they reach the cell surface. Among PIP2-requiring proteins is the ion channel formed by KCNQ2 and KCNQ3 subunits (Kv7.2 and Kv7.3). It mediates the M-current of sympathetic and other neurons. This report concerns the PIP2 sensitivity of the M-current channel and emphasizes studies from our laboratory primarily in expression systems.
The M-current was among the first whose modulation by G-protein-coupled receptors was recognized (Brown & Adams, 1980). It is reversibly suppressed by activation of M1 muscarinic receptors or of other receptors coupled to the G-protein Gq (Delmas & Brown, 2005). The task of determining the signalling pathway from M1 receptors to M-channels was complicated by the wealth of signals that result from activation of PLC by Gq (Fig. 1). Beyond the several messengers noted in the figure, arachidonic acid is metabolized to a host of prostaglandins, prostacyclins, and leukotrienes. Sorting out so many potential signals is a general problem whenever one studies a process regulated through Gq. Any hypothesis must be tested in multiple redundant ways to be convincing. For M-current, investigators tested various products of PIP2 hydrolysis in vain as candidate inhibitory messengers (Delmas & Brown, 2005), until finally a solution was recognized: inhibition of current was not due to production of second messengers but rather to depletion of the PIP2 phospholipid. The channels require PIP2 to function (Suh & Hille, 2002; Zhang et al. 2003; Winks et al. 2005), and PLC becomes active enough to deplete the PIP2 pool (Horowitz et al. 2005); hence current falls. Probably the major barrier to reaching this simple conclusion was a failure to realize that in some cells the pools of PIP2 can be so small and the activity of PLC so large that PIP2 can be consumed upon activation of receptors. That some channels and transporters shut off when PIP2 is removed by more extreme, non-physiological manoeuvers was already known (Hilgemann & Ball, 1996; Hilgemann et al. 2001). That PIP2 sensitivity underlies physiological signalling was not known.
This short report illustrates a set of methods that explores the hypothesis that PIP2 is required for a cellular process, using the M-current as an example. The methods, developed by other laboratories, make use of transfectable constructs based on lipid-processing enzymes and protein domains. Some of the methods focus on the recovery of M-current rather than on its inhibition. A key concept is that after receptors activate PLC and produce many distracting second messengers in a burst, the metabolism of each of those molecules and the resynthesis of PIP2 are independent steps that can be manipulated separately to dissect which molecule(s) carry the signal. To test the PIP2 hypothesis we must consider metabolic steps that are accessible and relevant. Proximal steps leading to synthesis and breakdown of PIP2 are summarized in Fig. 2. The cellular pool of phosphatidylinositol (PI) is as much as 50 times larger than that of PIP2. PI 4-kinase phosphorylates PI at the inositol 4-position to make phosphatidylinositol 4-phosphate (PIP). PIP 5-kinase then phosphorylates PIP at the 5-position to make PIP2. At the same time, two phosphatases, PIP2 5-phosphatase and PIP 4-phosphatase, dynamically oppose these reactions, so that on a timescale of seconds to minutes each phosphoinositide pool turns over and readjusts in size. The final step in Fig. 2 is the cleavage of PIP2 by PLC.
How can one observe and manipulate this system? Fluorescent translocation probes have been constructed that label PIP2/IP3 and DAG dynamically. The PH domain of PLCδ1 coupled to GFP has a high affinity for PIP2 and IP3 (Stauffer et al. 1998; Várnai & Balla, 1998; Hirose et al. 1999). It binds to PIP2 in the plasma membrane in resting cells and lights up the cell periphery until PLC is activated. Then it migrates to the cytoplasm. The C1 domain of PKC coupled to GFP has a high affinity for DAG (Oancea et al. 1998). It reports no DAG in resting cells and remains spread throughout the cytoplasm until PLC is activated. It then migrates to the plasma membrane. Both probes have been invaluable tools to show conditions that activate or block PLC in living cells. The phosphoinositides can be measured chemically as well by high-pressure liquid chromatography (HPLC) of glycerol headgroups (Nasuhoglu et al. 2002) or by mass spectrometry of lipid extracts (Wenk et al. 2003). Using the HPLC method on M1 receptor-expressing CHO cells showed that activation of PLC by a muscarinic agonist depletes total cellular PIP2 by 93% and PIP by 88% within 60 s (Horowitz et al. 2005; Li et al. 2005). Each of the enzymes in Fig. 2 has been cloned and can be overexpressed in cells to increase its activity. Two of the enzymes, PIP 5-kinase and PIP2 5-phosphatase, have also been engineered as constructs that can be brought to the plasma membrane at will by addition of a small molecule (see later). Finally, PI 4-kinases can be inhibited by phenylarsine oxide, the type III PI 4-kinase can be inhibited by high concentrations of wortmannin, and PLC can be inhibited by U73122 and edelfosine.
The dynamics of phosphoinositides and their interactions with M-current channels have enough steps and subtleties that a verbal description of individual steps might fail to recognize properties of the system of steps operating together. Therefore another necessary tool we would also include is kinetic modelling. We use an explicit mathematical model of the G-protein activation of PLC and the metabolic changes of phosphoinositide pools shown in Fig. 2 (Suh et al. 2004; Horowitz et al. 2005; Suh & Hille, 2006). Figure 3 is a sample of output from such a model showing a fast simulated decline of PIP2 and a slower simulated decline of PIP during application of a muscarinic receptor agonist. At the same time, equimolar amounts of IP3 and DAG are produced. The M-current is suppressed when PIP2 dissociates from the channel subunits and recovers as PIP2 is resynthesized. Modelling allows tests of our assumptions for self consistency.
PI 4-kinase is needed to maintain KCNQ current
Figure 4 plots the holding current from tsA-201 cells transfected with M1 muscarinic receptors and KCNQ2 and KCNQ3 channel subunits. The cells are held at –20 mV where the non-inactivating outward K+ current is tonically on. The KCNQ current is suppressed in 10–15 s when the muscarinic agonist oxotremorine M (Oxo-M) is added, and the current recovers within a few hundred seconds after the agonist is removed, recapitulating the classic M-current modulation discovered in sympathetic neurons by Brown & Adams (1980). The recovery does not occur if the whole-cell pipette lacks hydrolysable ATP (Fig. 4), and the muscarinic suppression does not occur if PLC is inhibited with U73122 or edelfosine (Suh & Hille, 2002; Winks et al. 2005; Horowitz et al. 2005). The following experiments are consistent with the hypothesis that the muscarinic inhibition of current is due to the depletion of PIP2 and that the ATP-dependent recovery of current is due to resynthesis of PIP2, starting from the large pool of PI.
We begin by manipulating the enzyme PI 4-kinase, which synthesizes the PIP2 precursor PIP (Fig. 2). If PI 4-kinase is overexpressed to raise its activity and speed the synthesis of PIP, the recovery of KCNQ current after Oxo-M removal occurs more rapidly (Fig. 5A). On the other hand, if PI 4-kinase is inhibited by wortmannin or phenylarsine oxide to stop supply of new PIP, current does not recover after Oxo-M removal (Suh & Hille, 2002; Zhang et al. 2003; Winks et al. 2005) (Fig. 5B). Such experiments show that the existing pool of PIP does not suffice to resynthesize PIP2. The HPLC experiments already mentioned showed that activating PLC depletes PIP as well as PIP2. For recovery, one must have continued flux from PI through PI 4-kinase to make more PIP. Indeed, when PI 4-kinase is inhibited by wortmannin or phenylarsine oxide, M-current begins to run down with a time constant of about 500 s (Fig. 5B). Apparently PIP2 is continually being broken down and remade even when agonists for Gq-coupled receptors are absent, so PIP2 will be depleted in 10 min if the 4-kinase is blocked. Whether the resting PIP2 breakdown is mostly by PIP2 5-phosphatase or by PLC has not been established.
PIP2 regulates KCNQ channels
Figure 6 shows experiments manipulating PIP 5-kinase and PIP2 5-phosphatase, two enzymes with PIP2 as their immediate product or substrate (Fig. 2). In published work, these enzymes have been overexpressed in KCNQ-expressing cells resulting in making the current highly resistant to suppression by Oxo-M or nearly eliminating the current, respectively (Winks et al. 2005; Li et al. 2005). Rather than overexpressing the full-length enzyme, we used a novel approach to activate engineered versions of the enzymes by translocating them to the plasma membrane on demand. The scheme, based on chemical dimerization by rapamycin or rapamycin analogues, is illustrated in Fig. 6A. Two protein domains, FRB and FKBP, have partial binding sites for rapamycin and can be dimerized by addition of rapamycin. Pairs of proteins that have been covalently joined to FRB or FKBP, respectively, can be drawn together by adding rapamycin. In the form developed by the laboratory of Tobias Meyer (Inoue et al. 2005), a membrane-anchoring domain is fused to FRB, so that FRB is permanently tethered to the plasma membrane. An enzyme of interest is then fused with FKBP, so that the enzyme will be drawn to the plasma membrane FRB anchor upon addition of rapamycin. For high concentrations of rapamycin, the translocation occurs in ∼20 s, as revealed using GFP-tagged proteins. Since the enzyme is originally free to move in a volume that might extend 5 μm from the plasma membrane but then is restricted through tethering to perhaps 50 Å, the effective enzyme concentration is raised 1000-fold at the plasma membrane.
We used this chemical-dimerization strategy with PIP 5-kinase and PIP2 5-phosphatase, translocating them to the plasma membrane with a rapamycin analogue (Suh et al. 2006). When the FKBP domain was joined with PIP 5-kinase to make more PIP2, the M-current increased slowly after dimerization (Fig. 6B). From kinetic modelling of all the reactions in Fig. 2, we deduce that the slowness of the current rise reflects (i) that the PIP2 binding sites on KNCQ channels are already well above 50% occupied so the PIP2 concentration must rise a lot to get closer to saturation, and (ii) that the size of the resting PIP pool is too small to supply all this increase at once so that reaching the new steady state must wait for slow replenishment of PIP via the PI 4-kinase pathway. The experiments with PIP 5-kinase show that PIP, its substrate, must be present in the plasma membrane.
On the other hand, when FKBP was joined with PIP2 5-phosphatase to dephosphorylate PIP2, the M-current fell precipitously after dimerization (Fig. 6B). Control experiments showed that this fall was accompanied by a rapid fall of PIP2 (seen with the PH-domain translocation probe), was not sensitive to PLC inhibitors and was not accompanied by production of a Ca2+ elevation or of DAG (tested with the C1-domain probe). Thus, inhibition of M-current does not require the downstream messengers generated when PLC cleaves PIP2. Furthermore, the translocated enzyme presumably generated a large bolus of PIP at the plasma membrane, yet the KCNQ current fell. This means that PIP cannot replace PIP2 as a permissive ligand for KCNQ channels. A more direct approach to the lipid specificity question involves exposing the cytoplasmic face of an excised patch of membrane to different phosphoinositides (Rohacs et al. 1999; Zhang et al. 2003). A very similar translocatable PIP2 5-phosphatase probe was developed by Várnai et al. (2006).
Outlook for the future
Deciding which messages mediate signals from Gq-coupled receptors to cellular outputs has been a daunting task in the past. Now we have a large number of tools suitable for use in single-cell electrophysiology on a microscope stage that allow us to test whether PIP2 is itself a major player in a specific cellular response. They should be used in combination to develop a convincing conclusion. They show that M-current requires PIP2 in the plasma membrane for activity and that M-current is suppressed when PLC cleaves PIP2. In unpublished work with these methods we have found that there is a component of the activity of Cav1.3 and Cav2.2 Ca2+ channels that requires PIP2. Many other membrane functions could now be screened in this way, and similar tools could be generated to explore possible requirements for other phosphoinositides.
We thank Lea Miller for technical assistance and Drs Ken Mackie, Takanari Inoue and Tobias Meyer for invaluable recent collaboration. We thank past and present members of our laboratory for valuable contributions to this project. This work was supported by NIH grants NS08174 from the National Institute of Neurological Disorders and Stroke (NINDS) and AR17803 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS).