Abstract
In neurons, loss of plasma membrane phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] leads to a decrease in exocytosis and changes in electrical excitability. Restoration of PI(4,5)P2 levels after phospholipase C activation is therefore essential for a return to basal neuronal activity. However, the dynamics of phosphoinositide metabolism have not been analyzed in neurons. We measured dynamic changes of PI(4,5)P2, phosphatidylinositol 4-phosphate, diacylglycerol, inositol 1,4,5-trisphosphate, and Ca2+ upon muscarinic stimulation in sympathetic neurons from adult male Sprague-Dawley rats with electrophysiological and optical approaches. We used this kinetic information to develop a quantitative description of neuronal phosphoinositide metabolism. The measurements and analysis show and explain faster synthesis of PI(4,5)P2 in sympathetic neurons than in electrically nonexcitable tsA201 cells. They can be used to understand dynamic effects of receptor-mediated phospholipase C activation on excitability and other PI(4,5)P2-dependent processes in neurons.
SIGNIFICANCE STATEMENT Phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] is a minor phospholipid in the cytoplasmic leaflet of the plasma membrane. Depletion of PI(4,5)P2 via phospholipase C-mediated hydrolysis leads to a decrease in exocytosis and alters electrical excitability in neurons. Restoration of PI(4,5)P2 is essential for a return to basal neuronal activity. However, the dynamics of phosphoinositide metabolism have not been analyzed in neurons. We studied the dynamics of phosphoinositide metabolism in sympathetic neurons upon muscarinic stimulation and used the kinetic information to develop a quantitative description of neuronal phosphoinositide metabolism. The measurements and analysis show a several-fold faster synthesis of PI(4,5)P2 in sympathetic neurons than in an electrically nonexcitable cell line, and provide a framework for future studies of PI(4,5)P2-dependent processes in neurons.
Introduction
We report the dynamics of phosphoinositide metabolism and signaling in single living neurons. We measure receptor-induced changes of phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2], phosphatidylinositol 4-phosphate [PI(4)P], inositol 1,4,5-trisphosphate (IP3), diacylglycerol (DAG), and Ca2+ and develop a kinetic description. Phosphoinositides are fundamental signaling components of all cellular membranes. PI(4,5)P2 located at the plasma membrane is required by many membrane proteins to fulfill their functions. In neurons, binding of SNARE proteins and synaptotagmin to PI(4,5)P2 allows vesicle docking and secretion (Di Paolo and De Camilli, 2006; Martin, 2015). Similarly, endocytosis (Cremona and De Camilli, 2001; Posor et al., 2015), actin filament assembly (Saarikangas et al., 2010), and ion channel activity (Hille et al., 2015) are regulated by PI(4,5)P2.
PI(4,5)P2 levels can be reduced through activation of receptors. Phospholipase C (PLC) activated by G-protein-coupled receptors and receptor tyrosine kinases cleave PI(4,5)P2 into IP3 and DAG, activating several signaling pathways (Michell, 1975; Berridge and Irvine, 1984). At least 80 types of PLC-coupled receptors are known (Pawson et al., 2014). Receptor tyrosine kinases acting through PI 3-kinases also use PI(4,5)P2 as a precursor to synthesize PI(3,4,5)P3. Once the stimulus activating these receptors is gone, PI(4,5)P2 levels are restored (Willars et al., 1998; Cantley, 2002). PI(4,5)P2 is synthesized in two steps involving sequential phosphorylation of phosphatidylinositol (PI) and PI(4)P by the action of lipid 4- and 5-kinases, respectively. The kinase steps are opposed by lipid phosphatases that convert PI(4,5)P2 back to PI(4)P (5-phosphatase) and PI(4)P to PI (4-phosphatase). Biochemical experiments show that the PI pool is by far the largest, comprising 90–96% of the total phosphoinositide, whereas PI(4)P and PI(4,5)P2 make up much of the remainder (Willars et al., 1998; Xu et al., 2003), suggesting that primarily PI(4)P levels limit PI(4,5)P2 synthesis.
In a scenario where neuromodulators are released from extrasynaptic or synaptic sites to bathe nearby neurons, the activation of PLC and subsequent hydrolysis of PI(4,5)P2 will arrest PI(4,5)P2-dependent processes. For example, in superior cervical ganglion (SCG) neurons a voltage-gated potassium channel can be turned off by PLC-mediated hydrolysis of PI(4,5)P2 (Brown and Adams, 1980; Suh and Hille, 2002; Zhang et al., 2003; Winks et al., 2005). This greatly increases the excitability of SCG neurons, facilitating firing of spikes while PI(4,5)P2 is low, a process that in the CNS would be akin to arousal. Despite the importance of PI(4,5)P2 resynthesis for the duration of such neuronal arousal, relatively little is known about the dynamics of PI(4,5)P2 replenishment in neurons.
The dynamics of PI(4,5)P2 synthesis have been measured carefully in several cell lines (Willars et al., 1998; Xu et al., 2003; Jensen et al., 2009; Falkenburger et al., 2010a,b). After depletion by PLC, 63% of PI(4,5)P2 is resynthesized within 120 s in these cells. Similarly in neurons, PI(4,5)P2 is depleted and resynthesized after PLC activation (Micheva et al., 2001; Winks et al., 2005). Although the kinetics of PI(4,5)P2 depletion have been characterized and modeled for sympathetic neurons (Winks et al., 2005), we still lack a quantitative description of the underlying dynamics of PI(4,5)P2 resynthesis.
We ask the following questions: how long does it take for a neuron to replenish the pool of PI(4,5)P2 at the plasma membrane? How do these dynamics compare with other cells, and how are they regulated? We decided to study these questions in superior cervical ganglion neurons as the expression of PI(4,5)P2-dependent ion channels and PLC-activating G-protein-coupled receptors are well described in these cells, thus making them an ideal model system to characterize the phosphoinositide metabolism of a primary neuron. Our results show a several-fold faster phosphoinositide metabolism in SCG neurons compared with nonexcitable cell lines, leading us to the hypothesis that PI(4,5)P2 dynamics are cell-type-specific and that neurons use faster kinetics of PI(4,5)P2 synthesis to fine-tune electrical excitability and other PI(4,5)P2-dependent processes. We develop a quantitative framework for the dynamics of physiological processes in neurons that depend on PI(4,5)P2.
Materials and Methods
Cells.
Animals were handled according to guidelines of the University of Washington Institutional Animal Care and Use Committee. Neurons were isolated from SCG of adult (7- to 12-week-old) male Sprague-Dawley rats by enzymatic digestion as described previously (Beech et al., 1991; Vivas et al., 2013). Isolated neurons were plated on poly-l-lysine (Sigma-Aldrich)-coated glass chips and incubated in 5% CO2 at 37°C in medium supplemented with 10% fetal bovine serum.
SCG neurons were transfected by nuclear injection 24 h after isolation using an Eppendorf 5242 pressure microinjector and 5171 micromanipulator system (Eppendorf). cDNA was dissolved in 1 mg/ml dextran-fluorescein solution (10,000 kDa; Molecular Probes) to yield a final concentration of 50–150 ng/μl of DNA and microinjected into the nucleus for 0.3 s at pressures of 80–120 hPa. TsA201 cells were cultured and transfected as described previously (Jensen et al., 2009). The following plasmids were used: TubbyR332H-YFP (provided by A. Tinker, University College of London, London, United Kingdom); human eCFP-PH (PLCδ1), eYFP-PH (PLCδ1), and CAAX-CFP (from K-Ras, provided by K. Jalink, The Netherlands Cancer Institute, Amsterdam, Netherlands); C1-YFP (from the C1 domain of PKCγ, provided by T. Meyer, Stanford University, Stanford, CA); the zebrafish voltage-sensitive phosphatase (Dr-VSP; provided by Y. Okamura, Osaka University, Osaka, Japan); and the IP3 LIBRA version III (LIBRAvIII; provided by A. Tanimura, Health Sciences University of Hokkaido, Tobetsu, Japan).
Electrophysiology.
The potassium current IKCNQ2/3 was recorded in whole-cell configuration as previously described for SCG neurons (Vivas et al., 2014) and tsA201 cells (Falkenburger et al., 2010b).
Lipid analysis by mass spectrometry.
Lipid analysis was a modification of previously described methods (Clark et al., 2011; Laketa et al., 2014). Each lipid determination required two sympathetic ganglia or the cultured cells from one 35 mm ∼70% confluent dish. Briefly, pairs of ganglia were mechanically minced, and neurons and tsA201 cells were pelleted by centrifugation for 3 min at 1000 × g and 4°C. Cell pellets were resuspended in water. Lipids were first extracted twice with N-butanol, followed by three extractions with chloroform. All extraction steps were performed on ice, and separation of aqueous and organic phases was achieved by centrifugation for 3 min at 20,000 × g and 4°C. Extracts were combined and solvents were evaporated under nitrogen. Lipids were methylated with (trimethlysilyl)diazomethane for 1 h at room temperature with suitable precautions. Lipids were separated by ultrahigh-pressure liquid chromatography (Acquity UPLC Protein BEH C4 column, 300Å, 1.7 μm, 1 × 100 mm, Waters) and detected by mass spectroscopy (Xevo TQ-S, Waters) with sodium formate infusion into the ionization chamber. To detect changes in phosphoinositide levels after activation of muscarinic receptors, whole SCG were cut into small pieces and treated for 1 min with 10 μm oxotremorine methiodide (Oxo-M) before the reaction was stopped by addition of ice-cold methanol/1 N HCl. Tsa201 cells transiently transfected with muscarinic receptors (M1Rs) were equally treated with Oxo-M for 1 min, followed by stoppage of the reaction ice-cold methanol/1 N HCl. Extraction of lipids was performed as described above. To normalize for cell numbers, 10% of each sample was removed before lipid extraction and genomic DNA was extracted with GeneJet Genomic DNA Purification Kit (Thermo Scientific) according to the manufacturer's instructions. Genomic DNA amounts were quantified by spectrophotometry and used for normalization of lipid signal intensities.
FRET and calcium photometry.
Optical measurements of calcium and Förster resonance energy transfer (FRET) were performed on single neurons by whole-cell photometry (not with images) as previously described (Falkenburger et al., 2010a,b, 2013; Dickson et al., 2013). Cells were illuminated by a grating-controlled monochromatic light source (Polychrome IV; TILL Photonics). For measurements of cytoplasmic-free Ca2+, neurons were loaded with 2 μm fura-2-AM (Invitrogen) dissolved in Ringer's solution containing 0.02% pluronic acid F-68. Fura-2 signals were reported as the fluorescence ratio with two wavelengths of excitation, F340/F380. FRET was measured as the ratio of corrected fluorescence from YFP and CFP after excitation of CFP molecules and is reported as FYFP/FCFP (FRETr).
Western blot analysis.
Cytoplasmic proteins from SCG neurons and tsA201 cells were isolated and purified with the Mem-PER Plus Membrane Protein Extraction Kit (Pierce Protein Biology Products, Thermo Scientific) according to the manufacturer's instructions. Membrane proteins were extracted as follows: cells were spun down at 300 × g and resuspended in 1 ml H2O. Cell suspensions were frozen in liquid nitrogen and thawed at 37°C in a water bath. This step was repeated twice before samples were centrifuged for 3 min at 300 × g at 4°C. The supernatant was removed and spun again at 20,000 × g for 20 min at 4°C. The pellet was resuspended in HEPES-lysis buffer (150 mm NaCl, 10 mm HEPES, 0.5% Triton X-100, pH 7.4). Protein concentration was determined by BCA protein assay (Pierce Protein Biology Products, Thermo Scientific). Twenty micrograms of protein was separated by SDS-PAGE using standard techniques. The primary antibody against rat and human PLCβ1 was used at a concentration of 1 μg/ml (Abcam, ab140746); anti-IP3R1 and anti-IP3R2 were used at a dilution of 1:500 (Alomone Labs, ACC-019 and ACC-116), anti-IRBIT (Abcam, ab178693) was used at a dilution of 1:10,000, and all were detected using a horseradish peroxidase-coupled secondary antibody (1:5000, Santa Cruz Biotechnology) and chemiluminescence (Pierce ECL Western Blotting Substrate, Pierce Protein Biology Products, Thermo Scientific). Recombinant protein of human phospholipase C β 1 (PLCβ1) was used for Western blot analysis (OriGene Technologies).
Mathematical modeling.
The Virtual Cell software environment (University of Connecticut Health Center, http://www.nrcam.uchc.edu) was used to develop a kinetic description of phosphoinositide metabolism in neurons. As a starting point and for comparison, we used a compartmental model (Dickson et al., 2013; Falkenburger et al., 2013) that had been developed to describe observations on the tsA201 cell line. Conditions that were changed to resemble the results obtained in SCG neurons are shown in Figures 3A,B, 5E, 8C,G, and Table 1. Model rate constants were chosen manually to fit experimental data. To compare modeled calculations to data obtained from FRET and electrophysiological recordings, we segmented the traces into onset and recovery of the effect and normalized to basal levels in each case to correct for incomplete recovery. Τhe modified Virtual Cell Model “hillelab: KruseVivasTraynorKaplanHille_PI-metabolism_SCG_2015” is publicly available at http://www.vcell.org/.
Rate constants and parameters for SCG neuron and tsA201 cell models
Data analysis.
We used IGOR Pro (IGOR Software, RRID:nlx_156887, WaveMetrics) and Excel (Microsoft) to analyze data. Statistics are given as mean ± SEM. Student's t test was used to test for statistical significance. p values <0.05 were considered significant.
Results
We start by explaining the logic of the experiments. We first make empirical measurements of lipid metabolism and then bring in kinetic reasoning to develop a clearer understanding of several differences in the underlying steps in neurons. The lipids are followed by electrophysiological and optical indicators and by chemical analysis. Figure 1A shows the steps to be analyzed. They include the lipid 4- and 5-kinases (enzymes are shown in red) that phosphorylate PI to PI(4)P and PI(4,5)P2, the lipid 4- and 5-phosphatases that reverse these reactions, and the PLC that hydrolyzes PI(4,5)P2 into DAG and IP3. We perturb lipid levels by activating enzymes. Thus, when PLC is activated by receptors, the pools of PI(4)P and PI(4,5)P2 both become depleted (see the next section) and recovery of PI(4,5)P2 is a two-step reaction that requires both the 4- and the 5-kinase. When we want to analyze the kinetics of the 5-kinase alone, we use a special voltage-sensing lipid phosphatase (VSP; Murata and Okamura, 2007) that dephosphorylates PI(4,5)P2 to PI(4)P (Fig. 1A). Subsequent recovery of PI(4,5)P2 then requires only the 5-kinase.
Faster kinetics of IKCNQ2/3 and PI(4,5)P2 recovery in SCG neurons versus tsA201 cells. A, Overview of the analyzed reactions of PI metabolism. Red, Enzymes. B, Schematic of closure of KCNQ2/3 channels due to PI(4,5)P2 hydrolysis. Left, Activation of phospholipase C after binding of the ligand Oxo-M to M1R. Right, Activation of a VSP by membrane depolarization. C, Normalized KCNQ2/3 tail current amplitude upon activation of M1 receptors by 10 μm Oxo-M measured with whole-cell recordings from SCG neurons and tsA201 cells. D, Time constants of tail current amplitude recovery after removal of Oxo-M. Numbers in brackets indicate number of experiments. E, Normalized KCNQ2/3 tail current amplitude upon activation of M1 receptors by 1 μm Oxo-M measured with whole-cell recordings from SCG neurons 1 h after isolation. F, Time constants of tail current amplitude recovery after activation of M1Rs with 1 μm Oxo-M at different time points after isolation of SCG neurons (black) or percentage inhibition of M-current (red). G, Same as in C but with activation of VSP. H, Time constants of tail current amplitude recovery after VSP activation. I, SCG neurons were transfected by intranuclear injection with plasmids encoding TubbyR332H-YFP and CAAX-CFP to measure changes in plasma membrane PI(4,5)P2 levels with ratiometric FRETr recordings. Top, Schematic of FRETr measurements before and after depletion of PI(4,5)P2. Bottom, Negative contrast confocal image of fluorescence intensity distribution of TubbyR332H before (left) and after (right) application of Oxo-M. J, Line scans of TubbyR332H distribution before (black) and after (orange) application of Oxo-M along the lines indicated in I. K, Normalized FRETr between TubbyR332H-YFP and CAAX-CFP upon application of Oxo-M to SCG neurons and tsA201 cells. L, Negative contrast confocal images of PH-PLCδ1-YFP fluorescence in a SCG neuron before and 20 s after stimulation with 10 μm Oxo-M. M, Line scans showing fluorescence intensity distribution of PH-PLCδ1-YFP along lines indicated in L before (black) and after (orange) application of Oxo-M. N, Mean normalized FRETr from SCG neurons transfected with PH-PLCδ1-CFP and -YFP upon stimulation of cells with Oxo-M.
Fast recovery of IKCNQ2/3 and PI(4,5)P2 in SCG neurons
In the first experiments, we monitored the levels of PI(4,5)P2 in the plasma membrane of SCG neurons using the endogenous PI(4,5)P2-dependent voltage-gated potassium channel KCNQ2/3 (KV7.2/7.3) as a reporter (Suh and Hille, 2002; Zhang et al., 2003; Winks et al., 2005). These channels underlie the M-current in SCG neurons (Brown and Passmore, 2009). We depleted PI(4,5)P2 by activation of M1Rs through a 20 s application of a muscarinic agonist, 10 μm Oxo-M (Fig. 1B, left). During agonist application, IKCNQ2/3 declined rapidly and almost completely, and upon removal of agonist, current recovered with a time constant of 42 ± 1 s at room temperature (Fig. 1C). Closure of these channels makes the neurons much more excitable (Brown et al., 2007), and their reopening terminates the period of raised excitability. The experiment was repeated for comparison in tsA201 cells transfected with KCNQ2, KCNQ3, and M1R. Now the current recovered with a much longer time constant of 123 ± 20 s upon removal of the agonist (Fig. 1C,D; Jensen et al., 2009). Thus, PI(4,5)P2 pools are replenished more rapidly in the neurons than in tsA201 cells.
The experiments shown in Figure 1, C and D, were conducted with neurons cultured for 48 h after isolation. To test whether the culture conditions alter the time constant of M-current recovery, we repeated the experiments at different culture times, ranging from 1 to 72 h after isolation. In addition, we used only 1 μm Oxo-M, which had previously been shown to cause half-maximal inhibition of M-current (Winks et al., 2005) and allowed us to observe potential alterations in M1 receptor signaling better. In freshly dissociated neurons, cultured only for 1 h, the application of 1 μm Oxo-M inhibited ∼50% of M-current and the time constant of current recovery was 44 ± 9 s (n = 9; Fig. 1E). Culturing neurons for up to 72 h did not change the fraction of current inhibition or the time constant of M-current recovery (Fig. 1F), showing that the chosen culture conditions do not alter M1R signaling or the time constant of M-current recovery. Similar to recordings obtained from neurons cultured for 48 h and treated with 10 μm Oxo-M, the time constant of current recovery was significantly faster than in tsA201 cells, suggesting that fast replenishment of PI(4,5)P2 pools is an intrinsic property of sympathetic neurons.
Are both lipid kinases in Figure 1A faster in neurons? To isolate the contribution of endogenous lipid 5-kinases, we expressed the VSP in SCG neurons. During a depolarizing voltage step to +100 mV, VSP dephosphorylates PI(4,5)P2 to PI(4)P, transiently depleting PI(4,5)P2, and KCNQ2/3 channel activity decreases sharply (Fig. 1B, right). After the brief activation of VSP, IKCNQ2/3 was strongly suppressed and then recovered with a time constant of 24 ± 5 s in SCG neurons and 10 ± 1 s in tsA201 cells (Fig. 1G,H; Falkenburger et al., 2010b). Apparently, the endogenous 5-kinase activity is lower in the neurons than in the tsA201 cells. Hence, we must hypothesize that the relative activity of the 4-kinase is markedly boosted in the neurons because the overall recovery of IKCNQ2/3 after PLC-activation is several-fold faster than in tsA201 cells. We test this notion in the next section.
To confirm the kinetic measurements of PI(4,5)P2, we repeated the experiments with two other indicators of PI(4,5)P2. The first was a YFP-tagged PI(4,5)P2 binding domain from Tubby (TubbyR332H-YFP; Hughes et al., 2007). To monitor plasma-membrane Tubby by a FRET assay (Fig. 1I), we coexpressed TubbyR332H-YFP (FRET acceptor) with a small plasma membrane localized CFP-tagged protein (CAAX-CFP, FRET donor). Under basal conditions, fluorescence of TubbyR332H-YFP was localized to the plasma membrane of SCG neurons (Fig. 1I,J) and the FRET ratio (FRETr FYFP/FCFP) was high. Then activation of M1R led to a translocation of the domain to the cytoplasm and a drop of FRETr (Fig. 1K), indicating PI(4,5)P2 hydrolysis. FRETr recovery during the washout had a time constant of 53 ± 15 s in SCG neurons and a time constant of 127 ± 25 s in tsA201 cells, corresponding well with our findings for IKCNQ2/3 recovery after receptor activation that overall PI(4,5)P2 synthesis occurs more slowly in tsA201 cells.
As a second FRETr-based assay for PI(4,5)P2, we repeated the experiments with CFP- and YFP-tagged versions of pleckstrin homology (PH) domains from PLCδ1. These experiments were done only in neurons. Similar to TubbyR332H, the PH-PLCδ1 domain probes were localized at the plasma membrane under basal conditions and FRETr was high, and the probes moved to the cytoplasm upon application of Oxo-M (Fig. 1L,M). The time constant of PH domain FRETr recovery was 47 ± 8 s (Fig. 1N), similar to the 53 s recovery of TubbyR332H FRETr and the 42 s recovery of IKCNQ2/3. Hence, the FRETr recordings with TubbyR332H and PH-PLCδ1 confirm that IKCNQ2/3 recovery tracks PI(4,5)P2 recovery in SCG neurons and that overall PI(4,5)P2 resynthesis is appreciably faster in SCG neurons than in tsA201 cells.
Accelerated synthesis of PI(4)P in SCG neurons
Our analysis of IKCNQ2/3 recovery after VSP stimulation implied that there is a faster lipid 4-kinase activity in SCG neurons. We used a recently developed fluorescent probe for PI(4)P (P4M; Hammond et al., 2014) to measure the kinetics of PI(4)P changes. In SCG neurons, a GFP-labeled version of P4M showed a significant component at the plasma membrane under basal conditions, which, as expected, became mainly cytoplasmic after application of Oxo-M to SCG neurons (Fig. 2A,B). In basal conditions, there also was intracellular labeling, likely at the Golgi. To measure the time course at the plasma membrane, we cotransfected SCG neurons with a YFP-tagged P4M-probe and the plasma-membrane localized CAAX-CFP. Under basal conditions there was FRETr between the two fluorophores as PI(4)P was present at the plasma membrane. Upon muscarinic stimulation, the FRETr decreased; P4M-probes were leaving the plasma membrane. Upon washout of the agonist, there was partial recovery with a time constant of 73 ± 15 s, reflecting resynthesis of PI(4)P (Fig. 2C). Similar experiments in tsA201 cells gave similar results except that the recovery of FRETr was threefold slower (Fig. 2C). With this direct measure we confirm that neurons synthesize PI(4)P faster than tsA201 cells.
Recovery of PI(4)P is faster in SCG neurons than in tsA201 cells. A, Negative contrast confocal images of the PI(4)P indicator P4M-GFP in a SCG neuron before (left) and after (right) application of 10 μm Oxo-M. B, Line scan showing P4M distribution along lines indicated in (A) before (black) and after (orange) Oxo-M application. C, Normalized FRETr measured from SCG neurons (black) and tsA201 cells (blue) transfected with P4M-YFP and CAAX-CFP upon application of Oxo-M for 20 s.
Determination of PLC and M1R expression levels in SCG neurons
We now develop a more quantitative description of the underlying events. To deconvolve the rates of enzymatic steps, we adapted our earlier kinetic model of phosphoinositide metabolism in tsA201 cells (Dickson et al., 2013; Falkenburger et al., 2013) to our observations in SCG neurons. An important initial step was to determine the relative amounts of PLC and M1R proteins in SCG compared with tsA201 cells. We isolated cytoplasmic proteins from both cell types and detected PLCβ1 in these lysates by Western blot analysis with an anti-PLCβ1 antibody. Comparison of PLCβ1-signal intensities showed lower expression in superior cervical ganglia than in tsA201 cells when referred to a total protein basis (Fig. 3A,B). Accordingly we set the level of PLCβ1 in the model for SCG neurons to ∼31% of the level used in the tsA201 cell model (Fig. 3B; Table 1).
Estimation of PLC and M1R expression levels in SCG neurons. A, Western blot of cytoplasmic lysates of tsA201 cells and SCG neurons with an anti-PLCβ1 antibody. Protein amounts of the two lysates were normalized for equal amounts; percentages indicate dilution steps. “PLCβ1” indicates detection of a recombinant PLCβ1 protein used as a positive control for the primary antibody. B, Percentage of PLCβ1 in SCG neurons compared with tsA201 cells as measured by analysis in A. Number in brackets indicates number of experiments. C, Left, Time course of KCNQ2/3 tail current amplitude recorded from a SCG neuron upon application of indicated concentrations of Oxo-M. Right, Current traces corresponding to time points indicated by letters on the left. Arrow indicates tail currents. D, IKCNQ2/3 inhibition evoked by 20 s applications of different concentrations of Oxo-M in SCG neurons (black) or tsA201 cells (blue). E, Modeled IKCNQ2/3 inhibition evoked by 20 s applications of different concentrations of Oxo-M for indicated densities of M1R. Dotted line indicates M1R density of 16 molecules μm−2. F, Comparison of experimental (black) and modeled (red) dose–response curves for I KCNQ2/3 inhibition by Oxo-M in SCG neurons. G, IC50 values for I KCNQ2/3 inhibition by Oxo-M in SCG neurons as measured experimentally (n = 5–25) and from our model.
The surface density of receptors was very high in our M1R transiently transfected tsA201 cell model, 500 receptors μm−2 (Falkenburger et al., 2010a). As M1R is endogenously expressed in SCG neurons, we anticipated that the surface density might be significantly lower there. Our method of estimation combined electrophysiological recordings and mathematical modeling, following the concept that the functional surface density of M1R defines the concentration-response curve of IKCNQ2/3 to graded concentrations of Oxo-M. We measured IKCNQ2/3 inhibition with 20 s applications of different concentrations of Oxo-M (Fig. 3C), finding an IC50 for IKCNQ2/3 inhibition of 1.2 ± 0.2 μm Oxo-M (Fig. 3D), one order of magnitude higher than that needed in receptor-transfected tsA201 cells under similar conditions (0.1 μm; Fig. 3D, data from Jensen et al., 2009). Our IC50 value for IM in SCG neurons is approximately twofold higher than a previously measured value of 0.7 ± 0.1 μm (Winks et al., 2005), a difference that is explained by the duration of Oxo-M application. We used only 20 s to permit for comparison to our previously measured IC50 in tsA201 cells, but not long enough for channel inhibition to reach a steady state at lower concentrations of Oxo-M. Winks et al. (2005) applied Oxo-M for >1 min, thereby measuring stronger IKCNQ2/3 inhibition at the lowest concentrations of Oxo-M.
We next used modeling to estimate the surface density of M1 receptors. We simulated the Oxo-M concentration-response curves for varying M1R surface densities (Fig. 3E) and estimated a mean M1R surface density of 16 ± 2 receptors μm−2 as the best fit to the experiments (Fig. 3F). This density predicts an IC50 for IKCNQ2/3 of 0.9 μm Oxo-M compared with the measured value of 1.2 ± 0.2 μm (Fig. 3G). It was used for all following simulations.
Determination of phosphoinositide levels by mass spectrometry
We had found a faster synthesis of PI(4)P from PI in SCG neurons. Such a speed up relative to tsA201 cells might be explained in two ways: (1) by an increased amount of PI relative to PI(4)P in neurons, or (2) by an increased specific activity of the lipid 4-kinase (Fig. 1A). We evaluated the first possibility by measuring the phosphoinositides directly using mass spectrometry.
Extraction of lipids, followed by derivatization, allowed successful detection of phosphoinositides. Note that mass spectrometry does not distinguish the positions of phosphorylation so here we refer to PIP and PIP2 without being able to distinguish regio-isomers. We found a ratio of PI/PIP of 60 ± 12 (n = 4) for SCG neurons, and of 58 ± 8 (n = 7) for tsA201 cells (Fig. 4A). We confirmed that these lipids were accessible to hydrolysis by phospholipase C through activation of muscarinic receptors. Application of 10 μm Oxo-M reduced PIP to 27 ± 8% and PIP2 to 16 ± 3% of control levels (n = 3; Fig. 4B). A similar reduction was observed in tsA201 cells transiently transfected with M1 receptors. PIP was reduced to 39 ± 8% and PIP2 to 41 ± 5% (n = 3; Fig. 4B). It should be noted that we extracted total cellular lipids from all cell types within the SCG. Histologic analyses by several different groups have shown that neurons occupy the majority of the volume of the SCG (Jacobowitz and Woodward, 1968; Ichikawa et al., 2009; Ke et al., 2015), suggesting that the reduction of PIP and PIP2 detected by our measurements is mainly caused by activation of muscarinic receptors of neurons. In addition, we tested whether the relative abundance of different types of phosphoinositides of SCG neurons and tsA201 cells showed any significant differences. The three phosphoinositide classes PI, PIP, and PIP2 made up the following percentages of the total: 96.2 ± 1.2%, 1.8 ± 0.6%, and 2.1 ± 0.6% in neurons and 96.9 ± 0.9%, 1.5 ± 0.3%, and 1.7 ± 0.7% in tsA201 cells, respectively (Fig. 4C). These values are similar to previously reported ratios of phosphoinositides in other cell types (Willars et al., 1998; Xu et al., 2003) and show no significant differences between neurons and tsA201 cells. We conclude that neurons have the same overall relative abundance of PI, PIP, and PIP2 as other cells and the faster synthesis of PI(4)P observed in neurons is not caused by a superabundance of PI.
Phosphoinositide levels determined by mass spectrometry. A, Ratio of total PI to total PIP as measured from SCG and tsA201 cells. B, Levels of PIP and PIP2 relative to controls from SCG and tsA201 cells after 1 min application of 10 μm Oxo-M. C, Relative amounts of total PI, PIP, and PIP2 as percentages of total phosphoinositides for SCG and tsA201 cells. Number in brackets indicates number of experiments.
Modeling of phosphoinositide metabolism requires high activity of the lipid 4-kinase in SCG neurons
Having determined levels of PLCβ1 and M1R for the model, and recognizing that the relative levels of PI, PIP, and PIP2 are not significantly different between neurons and tsA201 cells, we turned our attention to the rate constants of the lipid kinases and phosphatases, abbreviated 4K, 4P, 5K, and 5P in Figure 1A. These rate constants were adjusted to approximate the measured time courses for IKCNQ2/3 and the two FRET probes after M1R or VSP activation in neurons. The simulations then agreed reasonably with the experimental kinetics of inhibition and recovery of IKCNQ2/3 (Fig. 5A,B, red lines) as well as with the loss of FRETr of Tubby-R332H and PH-PLCδ1 and recovery of FRETr (Fig. 5C,D, red line).
Modeling PI metabolism of SCG neurons requires faster lipid 4-kinase. A, Time course of normalized IKCNQ2/3 amplitude upon stimulation of SCG neurons with Oxo-M. Comparison of experimental data from whole-cell recordings (black) and normalized IKCNQ2/3 amplitude predicted by simulation (red). Traces were segmented into onset and recovery phases and normalized to basal levels to correct for partial recovery. B, Same as in A, but for activation of VSP. C, Time course of normalized FRETr between TubbyR332H-YFP and CAAX-CFP upon stimulation of SCG neurons with Oxo-M as measured experimentally (black) and as predicted by simulation (red). Traces were segmented into onset and recovery phases and normalized to basal levels to correct for partial recovery. D, Same as in C, but for PH-PLCδ1 domains as FRETr reporters. The prediction plotted in red assumes no binding of PH-PLCδ1 domains to IP3. Note that the modeled traces approximate changes in FRETr as a cooperative square law of the membrane-bound fraction of PH-PLCδ1 domains as recently described (Itsuki et al., 2014). E, Comparison of rate constant of synthesis of PI(4)P and PI(4,5)P2 between SCG neurons and tsA201 cells, expressed as fold-changes, SCG/tsA201. F, Simulated time courses of normalized levels of PI(4)P in SCG neurons (red, “neuron model”) and tsA201 cells (blue, “tsA201 model”) upon stimulation with Oxo-M.
The major adjustments in the model were to increase the rate of phosphorylation from PI to PI(4)P by a factor of 4 and to decrease the rate of phosphorylation from PI(4)P to PI(4,5)P2 by a factor of 2 for neurons (Fig. 5E; Table 1). As in the experiments, the neuron model shows a strikingly faster synthesis of PI(4)P (Fig. 5F), which leads to an overall faster recovery of PI(4,5)P2 after transient depletion by PLC.
DAG turnover is faster in SCG neurons, but IP3 turnover is not
Next, we turned to the two second messengers generated by hydrolysis of PI(4,5)P2, DAG, and IP3. We measured their production and degradation in neurons transfected with fluorescent FRET-probes for IP3 (LIBRAvIII) and DAG (C1-YFP donor and membrane-bound CAAX-CFP acceptor; Fig. 6A). These probes indicated significant and rapid generation of IP3 and DAG during muscarinic stimulation (Fig. 6B,E) and a relatively fast decay with time constants of 73 ± 9 s for IP3-bound LIBRAvIII and 15 ± 5 s for DAG-bound C1-domains. We used these measured kinetics together with the affinities of the probes (Table 1), to adapt our model for SCG neurons. The chosen PLC, M1R, and PI(4,5)P2 levels (Table 1) turned out to be appropriate. We adjusted the rate constants of processes metabolizing DAG and IP3 to reproduce the observed kinetics (Fig. 6C,D,F,G; Table 1). The modeled kinetics of IP3 production and degradation are almost identical for SCG neurons and tsA201 cells (Dickson et al., 2013), suggesting a similarity of the underlying machinery (see Fig. 7A). On the other hand, although DAG production follows a similar time course in these two cell types, its degradation is faster in SCG neurons (Fig. 7B). Hydrolysis of PI(4,5)P2 and closure of PI(4,5)P2-dependent KCNQ2/3 channels also are kinetically indistinguishable between the two types of cells, suggesting that their endogenous PLCs have similar activity. Nevertheless, in the model, resynthesis of PI(4,5)P2 happens significantly faster in SCG neurons than in tsA201 cells (Fig. 7C) as was observed, so simulated KCNQ2/3 channel activity recovers much more quickly (Fig. 7D).
Kinetics of IP3 and DAG signals after M1R activation. A, Schematic for detection of DAG and IP3 by FRET recordings. B, Mean normalized FRETr of LIBRAvIII in response to 10 μm Oxo-M in SCG neurons (inverted scale, FCFP/FYFP). C, D, Overlay of modeled LIBRAvIII response (red) on the time course of normalized inverted LIBRAvIII-FRETr evoked by stimulation of SCG neurons with Oxo-M (black). Traces were segmented into onset (C) and recovery (D) phases and normalized to basal levels to correct for partial recovery. E, Same as in B, but SCG neurons were transfected with C1-YFP and CAAX-CFP to measure changes in DAG levels (FYFP/FCFP). Note different time scale. F, G, Same as in C, D, but for FRETr data and modeled C1-YFP/CAAX-CFP signal.
Summary comparison of modeled phosphoinositide metabolism and signaling in SCG neurons and tsA201 cells. A, Modeled time courses of IP3 in SCG neurons and tsA201 cells in response to a 20 s long 10 μm Oxo-M application. B–D, As in A, but for DAG (B), PI(4,5)P2 (C), and IKCNQ2/3 (D).
Modeling suggests that regulation of IP3R-availability and cytoplasmic Ca2+-buffering limit Ca2+ elevations from intracellular stores
As Oxo-M initiates fast and sizeable IP3 production in SCG neurons with calculated peak concentrations of ∼3 μm one might expect a significant release of Ca2+ from intracellular stores via IP3 receptors (IP3R). In tsA201 cells, ∼2 μm IP3 is sufficient to evoke almost maximal release of Ca2+ from stores, and peak IP3 concentrations of ∼13 μm do not lead to significantly larger Ca2+ release. Accordingly, we loaded SCG neurons with fura-2-AM to measure changes in cytoplasmic Ca2+ levels ratiometrically. In unclamped cells, we wanted to avoid action potential firing that would open voltage-gated calcium channels. Therefore, we bathed the cells with 100 nm tetrodotoxin (TTX). As others have reported (Cruzblanca et al., 1998; del Río et al., 1999; Zaika et al., 2007), upon stimulation of these neurons with 10 μm Oxo-M there was only a small Ca2+ rise, suggesting very little release of Ca2+ from intracellular stores (Fig. 8A). To check whether the Ca2+-increase is due to opening of low-threshold voltage-gated calcium channels, which could still occur with TTX, we repeated the experiment replacing TTX with 100 μm Cd2+, a blocker of voltage-gated calcium channels. We still observed a small cytoplasmic fura-2 signal under these conditions (Fig. 8B). Hence, it must be mediated by Ca2+ release from intracellular stores, and we had to consider why it is so small compared with the large increase seen with the same Oxo-M stimulus in tsA201 cells (Dickson et al., 2013).
Ca2+ signals after M1R activation in SCG neurons. A, Cytoplasmic Ca2+ in SCG neurons loaded with fura-2-AM and stimulated with 10 μm Oxo-M. Fura-2 signals, measured in the presence of 100 nm TTX. B, Same as in A, but with 100 μm CdCl2 as indicated instead of TTX. C, Western blot of membrane preparations of tsA201 cells and SCG neurons with anti-IP3R1 (top left) and anti-IP3R2 (bottom left) antibodies. Protein amounts of the samples were normalized for equal amounts. Right, Percentage of IP3R in SCG neurons compared with tsA201 cells. Number in brackets indicates number of experiments. D, Steady-state IP3 profile in a two-dimensional model of the cytoplasm of a neuron. IP3-production is assumed to occur at the origin. Concentrations at different distances depend on the diffusion coefficient of IP3 as well as the modeled activity of an IP3-5-phosphatase. E, Calculated cytoplasmic IP3 concentrations in one plane of a three-dimensional model of a SCG neuron upon uniform activation of M1Rs for the indicated time on the cell surface with a saturating concentration of Oxo-M. F, Calculated IP3 concentrations at two selected regions-of-interest (ROI) in the 3-D diffusion model in E. G, Western blot of lysate of SCG neurons with anti-IRBIT antibody. H, Scheme of the components of the model involved in Ca2+-release from intracellular stores. Gray, Components included in tsA201 cell model; Red, additional components added in SCG model. I, Overlay of modeled fura-2 response (red) on the time course of normalized fura-2 signal evoked by stimulation of SCG neurons with Oxo-M (black) in the presence of 100 nm TTX. J, Same as in I, but for simulation of fura-2 signal measured in the presence of 100 μm CdCl2 instead of TTX.
We tested alternative hypotheses with the model. First, we tested whether low IP3R levels can explain small Ca2+ signals. The assumed protein level of IP3Rs had to be lowered to only 0.005% of the previous value to match the observed Ca2+ signals (data not shown). However, Western blot analysis of neurons did not confirm such a reduction in IP3R protein levels (Fig. 8C). Instead, the protein amounts of IP3R1 and IP3R2 were reduced by only 60% in SCG neurons. Therefore, we considered other explanations.
It has been proposed that M1 receptors in SCG neurons are too far away from the ER membrane to activate IP3Rs (Delmas et al., 2002; Zaika et al., 2011; Zhang et al., 2013). This hypothesis requires a fast local degradation of released IP3. We developed a two-dimensional spatial model to ask how far IP3 might diffuse before it eventually is degraded. We started with the published diffusion coefficient for IP3 (283 μm2 s−1; Allbritton et al., 1992), and the experimentally determined degradation rate of IP3 (0.13 s−1, equivalent to a 7.7 s lifetime) in SCG neurons (Fig. 6B). Assuming that the IP3 degrading enzymes are uniformly distributed, this calculation showed that IP3 would diffuse beyond 50 μm before half of it is degraded (Fig. 8D). This distance is larger than the diameter of the soma of the neuron and several orders of magnitude larger than the expected distance between points of IP3-generation at the plasma membrane and the intracellular Ca2+ stores. Using a realistic geometry in a three-dimensional spatial model from an actual SCG neuron, the predicted profile of IP3 showed an almost uniform rise throughout the cytoplasm (Fig. 8E,F). A second three-dimensional spatial model in which we simulated muscarinic receptors on only one side of the cell did reveal an initial gradient of IP3 in the cytoplasm. However, within a couple of seconds IP3 again became uniform throughout the cytoplasm (Fig. 9A–C). We conclude from these simulations that under the observed rate of IP3 degradation, and no barriers impeding IP3 diffusion, IP3 should be able to bind to IP3Rs anywhere in the soma.
Gradients of IP3 during local activation of M1R in a simulated SCG neuron. A, The image shows one plane of a three-dimensional spatial model using the geometry of a SCG neuron. Outlined are the plasma membrane as well as the nuclear membrane. The red field is the area in which the activating ligand 10 μm Oxo-M is present throughout the simulation. B, Map of cytoplasmic IP3 concentrations for several time points during the simulation. The dark area inside the cytoplasm is the cell nucleus. C, Time course of IP3 concentrations for two regions-of-interest (ROI). Inset, Localization of ROIs. D, Simulated relative levels of Ca2+-bound and Ca2+-unbound NCS-1 isoforms in response to a simulated application of 10 μm Oxo-M. Simulations were performed for presence (black) or absence (red) of IRBIT. E, Simulated increase in Ca2+-bound NCS-1 upon muscarinic stimulation in the presence or absence of IRBIT.
A third possibility is that IP3Rs of SCG neurons, though present, are inhibited. IRBIT (IP3R-binding protein released with inositol 1,4,5-trisphosphate), an inhibitor that interacts with IP3Rs and is displaced competitively when IP3 binds, is expressed in the brain (Ando et al., 2003, 2006). We detected expression of IRBIT in superior cervical ganglia by Western blot (Fig. 8G), in agreement with previous reports showing expression of IRBIT by immunofluorescence (Zaika et al., 2011). When bound to an IP3R, IRBIT renders the IP3 binding domain unavailable for IP3. Therefore, higher IP3 concentrations are required to displace IRBIT and activate receptors. In addition, it has been shown that the intracellular Ca2+-buffering of SCG neurons is due to two pools of Ca2+-buffers with equivalent KD values of 100 nm and 1 μm (Wanaverbecq et al., 2003). We tested by simulation whether competition by IRBIT or a similar molecule together with strong buffering of released Ca2+ could account for the small amplitude of the fura-2 signal observed upon muscarinic stimulation. Addition of a species with the properties of IRBIT and of two Ca2+-buffering pools into our model (Fig. 8H; Table 1) sufficed to reproduce the experimental data (Fig. 8I,J). They predict an increase in free cytoplasmic Ca2+ of only 20 nm upon activation of M1R by a supramaximal concentration of ligand.
It has been shown that the Ca2+-binding protein neuronal calcium sensor-1 (NCS-1), which is highly expressed in SCG neurons, interacts with and stimulates lipid 4-kinases in its Ca2+-bound form (Zhao et al., 2001; Rajebhosale et al., 2003; Strahl et al., 2003; Gamper et al., 2004; Winks et al., 2005). We used our model to address the question whether the small rise in cytoplasmic Ca2+ upon muscarinic stimulation would be sufficient to convert NCS-1 into its Ca2+-bound form and stimulate lipid 4-kinases. We included a species with NCS-1's properties into our model and used the published KD of NCS-1 for free Ca2+ (Aravind et al., 2008). Our simulations show that stimulation with 10 μm Oxo-M for 20 s increases the amount of Ca2+-bound NCS-1 by only 2% (Fig. 9D,E). Interestingly, omitting the restriction of IP3-binding to IP3R by removing the species with IRBIT's properties from the model leads to a strong increase of Ca2+-bound NCS-1 by 75% (Fig. 9D,E), illustrating the important role of a protein with IRBIT's properties for Ca2+-dependent processes in SCG neurons. In conclusion, our simulations argue against a strong activation of NCS-1 by Ca2+ release from intracellular stores upon muscarinic stimulation, although they have not ruled out some role of NCS-1 for the fast phosphoinositide metabolism in SCG neurons.
Discussion
PI(4)P replenishment is fast in SCG neurons
We found that PI(4)P, the PI(4,5)P2 precursor, is restored faster in SCG neurons than in tsA201 cells. Because the PI pool is not larger, the most obvious explanation is that the enzymatic activity of the lipid 4-kinase is higher. The situation is subtle, because there are at least two sources of PI(4)P for PI(4,5)P2 synthesis, the PM and the Golgi. In tsA201 cells, 70% of PI(4,5)P2 is synthesized from the PI(4)P pool at the PM, and the remaining 30% of PI(4,5)P2 is synthesized from the pool at the Golgi (Dickson et al., 2014). Our model, which neglects the presence of two cellular compartments, predicts a fourfold faster PI(4)P synthesis than in tsA201 cells, in close agreement with the threefold increase in the rate of recovery measured experimentally. Our experimental setup allows us to measure PI(4)P recovery at the PM but does not allow us to discriminate between synthesis of PI(4)P from PI at the PM or delivery of PI(4)P from an intracellular compartment. Refined experimental designs will be necessary to analyze these processes in greater detail as studies have described mechanisms for an acceleration of lipid 4-kinases, as well as an important role for PI transfer proteins in supporting restoration of PI(4)P levels as discussed below.
Several mechanisms might increase the activity of lipid 4-kinases in SCG neurons. General mechanisms include altered expression levels or posttranslational modifications. Other mechanisms depend on protein interaction for directed localization. One is palmitoylation acting on lipid 4-kinase type II (Barylko et al., 2009; Zhou et al., 2014). Others include interaction of the lipid 4-kinase type III with PKC (Xu et al., 2014), TMEM150 (Chung et al., 2015), or with neuronal calcium sensor 1 (Paterlini et al., 2000; Zhao et al., 2001). We do not yet know whether any of these might apply here.
Faster PI(4)P synthesis at the PM requires a sufficient supply of the precursor, PI. Recent studies demonstrate an important role for PI transfer proteins, such as Nir2 (or PITPNM1), Nir3, and PITPNC1, in supplying PI to the PM from the ER and, under certain conditions, Golgi (Hardie et al., 2001; Garner et al., 2012; Chang et al., 2013; Kim et al., 2013, 2015; Chang and Liou, 2015). These proteins facilitate the transfer of PI to the PM at ER-PM junctions, and monitor PI(4,5)P2 depletion by binding to phosphatidic acid, which stimulates their recruitment to ER-PM junctions. These papers demonstrate an essential role of PI transfer proteins in PI(4,5)P2 recovery after its depletion, whereas they and other studies found only minor contributions of Golgi PI(4)P and vesicular transport for PI(4,5)P2 replenishment (Szentpetery et al., 2010). Our concept of faster synthesis of PI(4)P at the PM might have to be extended to include faster transfer of phosphoinositides to the PM.
Implications of fast PI(4,5)P2 replenishment in neurons
We used KCNQ2/3 channels as a tool to measure PI(4,5)P2. Focusing on the role of these channels we can see the relevance of fast synthesis of PI(4,5)P2 in excitable cells. KCNQ2/3 channels are the molecular determinant of M-current (Wang et al., 1998), which controls excitability in sympathetic and many central neurons (Brown and Adams, 1980; Marrion, 1997; Jentsch, 2000). Sympathetic neurons of the SCG receive cholinergic input from preganglionic neurons. Acetylcholine produces a fast and a slow response (Libet, 1964; Weight and Votava, 1970), with the slow response being mediated by activation of G-protein-coupled receptors that deplete PI(4,5)P2 (Suh and Hille, 2002; Zhang et al., 2003). Thus, whereas PI(4,5)P2 is depleted, M-current is inhibited and neurons are more excitable. Our analysis demonstrates that it takes ∼40 s for SCG neurons to resynthesize enough PI(4,5)P2 to reactivate most KCNQ2/3 channels, thereby restoring the resting membrane potential, ending the elevated excitability, and ending action potential firing. If instead, PI(4,5)P2 synthesis rates were as measured for nonexcitable tsA201 cells, elevated electrical activity of SCG neurons would last threefold to fourfold longer.
We have seen that phosphoinositide metabolism differs between SCG neurons and tsA201 cells. Are all neurons the same? Given the special needs of the nervous system, it is tempting to imagine that many neurons have faster phosphoinositide dynamics than nonexcitable cells. In sensory neurons and photoreceptors, for example, TRP and other channels require phosphoinositides (Rohacs, 2014; Hille et al., 2015), and therefore the speed of phosphoinositide restoration in these neurons controls the dynamic range for sensory input (Yadav et al., 2015). In the brain, modulatory areas control other brain areas by releasing neurotransmitters and hormones, such as acetylcholine, norepinephrine, or serotonin, acting on G-protein-coupled receptors that activate PLC and hydrolyze PI(4,5)P2. PI(4,5)P2-dependent KCNQ2/3, CaV2.2, and CaV2.1 channels are widely expressed in the brain, emphasizing the importance for rapid adjustment of phosphoinositide levels for coordinated ion channel activity in the brain.
PI(4,5)P2 depletion in neuronal and cardiac tissues
Studies on isolated neurons provide strong evidence for net PI(4,5)P2 depletion upon activation of muscarinic receptors, leading to closure of PI(4,5)P2-dependent ion channels such as KCNQ2/3 or CaV channels in these cells (Suh and Hille, 2002; Gamper et al., 2004; Brown et al., 2007; Hughes et al., 2007; Zaika et al., 2007; Hille et al., 2015). However, it is not known to what extent natural stimuli cause net PI(4,5)P2 depletion in intact neuronal tissues. Our simulations show that even a density of <20 M1R/μm2 is sufficient to cause net PI(4,5)P2 depletion at low agonist concentrations and to modulate KCNQ2/3 channel activity. Thus our simulations support the concept that even short exposure to low concentrations of agonists can cause net PI(4,5)P2 depletion in intact neuronal tissues.
In contrast to studies on neurons, studies on cardiac cells have failed to reveal evidence for net PI(4,5)P2 depletion, although the cells express PI(4,5)P2-dependent ion channels and PLC-activating G-protein-coupled receptors (Gertjegerdes et al., 1979; Cho et al., 2002; Nasuhoglu et al., 2002; Hilgemann, 2007). It has been shown that cardiac cells respond to activation of certain Gαq-coupled receptors with a slight increase in PI(4,5)P2 levels, which mechanistically has been explained by enhanced activities of lipid 4- and 5-kinases upon activation of protein kinase C downstream of PLC activation (Xu et al., 2014). These temporally increased lipid kinase activities ensure the uninterrupted activity of PI(4,5)P2-dependent ion channels. This stimulated PI(4,5)P2 synthesis becomes understandable if one takes into account that the two main repolarizing potassium currents in the human heart, IKr and IKs, are conducted by the PI(4,5)P2-dependent Herg and KCNQ1/KCNE1 channel complexes (Charpentier et al., 2010; Vandenberg et al., 2012). Any reduction in the activity of these ion channels could cause cardiac arrhythmias or even sudden death, which makes it necessary to remain sufficiently high PI(4,5)P2 levels in cardiac cells. This situation is different for neuronal cells, which can use net PI(4,5)P2 depletion as a mechanism to regulate ion channels and hereby neuronal activity. Perhaps in heart, the synthesis of phosphoinositides is significantly faster than in neurons or the maximum activity of PLC is significantly less.
Uncoupling of Ca2+ release and PI(4,5)P2 hydrolysis
Most cell types respond to elevated levels of IP3 by increases in cytoplasmic Ca2+ concentrations in the micromolar range via release of Ca2+ from intracellular stores (Berridge, 2009). It has been shown that activation of purinergic and bradykinin receptors in SCG neurons causes significant release of Ca2+ from intracellular stores, but does not cause net PI(4,5)P2 depletion (Zaika et al., 2007, 2011; Zhang et al., 2013). Activation of M1 receptors on the other hand causes net PI(4,5)P2 depletion, but circumvents Ca2+ release from intracellular stores despite substantial IP3 generation as shown in this and previous studies (Zaika et al., 2007, 2011). Although all of these receptors are coupled to the small G-protein Gαq and activate PLC upon binding of their ligand, the subsequently activated signaling pathways are different. It is an interesting hypothesis that a protein like IRBIT could be a differentiating factor for the type of signals evoked by activation of these different receptors. It has been shown that bradykinin receptors physically interact with IP3 receptors while M1 receptors lack this interaction (Delmas et al., 2002; Zhang et al., 2013). One could speculate that the physical interaction between bradykinin and IP3 receptors prevents the interaction of IRBIT with IP3Rs, which would create a pool of IRBIT-free IP3 receptors upon activation of bradykinin receptors. This receptor pool would then be available for IP3 to evoke a larger cytoplasmic Ca2+ increase compared with IRBIT-restricted muscarinic stimulation as reported by several studies (Gamper and Shapiro, 2003; Zaika et al., 2011). A similar mechanism might be present for purinergic receptors for which a similarly large increase in cytoplasmic Ca2+ has been reported as for stimulation with bradykinin (Zaika et al., 2011). The differentiated response to PLC-activating stimuli could be a mechanism for SCG neurons to react combinatorially to incoming signals. In comparison, tsA201 cells lack this property as activation of PLC by any Gαq-coupled receptor evokes release of Ca2+ from intracellular stores (Dickson et al., 2013). Apparently, neurons have a more nuanced response.
Footnotes
This work was supported by the National Institute of Neurological Disorders and Stroke of the NIH (R37NS008174), the Wayne E. Crill Endowed Professorship, and an Alexander von Humboldt-Foundation fellowship (M.K.); The Virtual Cell is supported by NIH Grant P41 GM103313 from the National Institute for General Medical Sciences; we appreciate use of the Mass Spectrometry Center of the School of Pharmacy. We thank Drs. Eamonn J. Dickson, Jill B. Jensen, Seung R. Jung, Duk-Su Koh, Jong B. Seo, and Haijie Yu for reading the paper; Gucan Dai, Eamonn Dickson, and Dale Whittington for help with establishing mass spectrometry experiments; Jennifer Deem and Stanley McKnight for experimental advice with IP3R Western blot analysis; Mika Munari for help with plasmid purification; all members of the Hille laboratory and colleagues in the Department of Physiology and Biophysics at the University of Washington for discussions and experimental advice; and Lea M. Miller for technical help.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Bertil Hille, Department of Physiology and Biophysics, University of Washington School of Medicine, Box 357290, Seattle, WA 98195-7290. hille{at}u.washington.edu