Mass Spectrometric Detection and Characterization of Atypical Membrane-Bound Zinc-Sensitive Phosphatases Modulating GABAA Receptors

Background GABAA receptor (GABAAR) function is maintained by an endogenous phosphorylation mechanism for which the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is the kinase. This phosphorylation is specific to the long intracellular loop I2 of the α1 subunit at two identified serine and threonine residues. The phosphorylation state is opposed by an unknown membrane-bound phosphatase, which inhibition favors the phosphorylated state of the receptor and contributes to the maintenance of its function. In cortical nervous tissue from epileptogenic areas in patients with drug-resistant epilepsies, both the endogenous phosphorylation and the functional state of the GABAAR are deficient. Methodology/Principal Findings The aim of this study is to characterize the membrane-bound phosphatases counteracting the endogenous phosphorylation of GABAAR. We have developed a new analytical tool for in vitro detection of the phosphatase activities in cortical washed membranes by liquid chromatography coupled to mass spectrometry. The substrates are two synthetic phosphopeptides, each including one of the identified endogenous phosphorylation sites of the I2 loop of GABAAR α1 subunit. We have shown the presence of multiple and atypical phosphatases sensitive to zinc ions. Patch-clamp studies of the rundown of the GABAAR currents on acutely isolated rat pyramidal cells using the phosphatase inhibitor okadaic acid revealed a clear heterogeneity of the phosphatases counteracting the function of the GABAAR. Conclusion/Significance Our results provide new insights on the regulation of GABAAR endogenous phosphorylation and function by several and atypical membrane-bound phosphatases specific to the α1 subunit of the receptor. By identifying specific inhibitors of these enzymes, novel development of antiepileptic drugs in patients with drug-resistant epilepsies may be proposed.


Introduction
Neuronal inhibition is essentially mediated by GABA type A receptors (GABA A R) forming anionic channels [1]. They are pentameric oligomers assembled with several subunit classes that may have multiple isoforms [1][2][3]. In adult rat brain, the most abundant subunits are a1, b2, and c2 typical for 60% to 90% of GABA A R [4]. The a 1 subunit is highly expressed throughout most brain regions especially in the cortex [5]. The function of these receptors can be modulated by reversible post-translational modifications such as phosphorylation-dephosphorylation [6][7][8]. Each subunit has four transmembrane domains and a large intracellular loop (I 2 ) between transmembrane domains 3 and 4 [2,9], containing consensus phosphorylation sites for both Ser/Thr and Tyr protein kinases [10][11][12]. The related modifications by several kinases have multiple effects such as direct modulation of the channel function [13], or receptor trafficking between synaptic sites and intracellular compartments [14,15]. In addition dephosphorylations by protein phosphatases (PP) have been shown to reverse the action of these kinases. For instance, PP1 and PP2B phosphatases dephosphorylate the b1-3 and c2 subunits [16,17], whereas PP2A dephosphorylates b3 subunits [18,19].
Recently, a new concept of GABAergic inhibition modulation by glycolysis has been described. The glyceraldehyde-3-phosphate dehydrogenase (GAPDH) a key glycolytic enzyme has been identified as a kinase for the GABA A R a1 subunits [20]. This endogenous kinase is directly tied to the receptor at the neuronal membrane and phosphorylates the a1 subunits at two identified serine and threonine residues. This phosphorylation of the a1 GABA A R subunit has been termed ''endogenous'' since it does not require any exogenous kinase or kinase activator [21].
We have previously shown that this phosphorylation prevents rundown of the GABAergic responses on acutely dissociated pyramidal neurons from rat cortex. In addition, an unknown membrane-bound phosphatase dephosphorylates the GABA A R a1 subunit and opposes the endogenous phosphorylation [22]. Protein phosphatases modulate the responsiveness of individual synapses to neural activity [16]. Ser/Thr protein phosphatases are expressed in many cell types and cellular compartments, and are regulated via several mechanisms [23]. They are classified into phosphoprotein phosphatases (PPP's) and metal-dependent protein phosphatases (PPM's), families defined by distinct amino acid sequences and 3-D structures.
Alterations in GABA A R expression and function lead to various neurological diseases including epilepsy, anxiety and schizophrenia [24]. Studies carried out on human brain tissue of patients with drug-resistant epilepsies show that both endogenous phosphorylation and function of GABA A R are deficient in the cortical epileptogenic zones [25]. Enhancing the GABA A R endogenous phosphorylation state by inhibiting unknown membrane-bound phosphatase(s) would maintain GABA A receptor function and therefore prevent seizures to occur. We propose this modulatory mechanism as a new target for the development of antiepileptic molecules active in drug-resistant epilepsies.
We have developed an electrophoresis autoradiographic method to measure the GABA A R endogenous phosphorylation/ dephosphorylation in washed cortical membranes [22]. However, this technique is not suitable to characterize the unidentified membrane phosphatase. The first goal of the present study was to elaborate an alternative approach to detect and characterize the membrane-bound phosphatases. Here, we describe a rapid and highly sensitive method for the quantification of phosphatase activity in washed brain cortical membranes. Further, we aimed at characterizing the kinetics and the pharmacological profile of these enzymes using the novel methodological tool proposed here. We show that the pharmacological profiles of these phosphatases are atypical and do not correspond to classical phosphatases. Our biochemical and functional results bring new insights on these membrane-bound phosphatases specific to GABA A R a1 subunit.

Synthetic peptides design
Four peptides were synthesized by Sigma-Genosys the amino acid sequences of which include the two identified endogenous phosphorylation sites of the I 2 loop of GABA A R a1 subunit (sequences are identical in the human, rat, bovine and mouse species). The minimal consensus [NXX(T/S)K] of GAPDHrelated phosphorylation site is present in several GABA A R alpha subunits (a1, a2, a3, a5), the I 2 loop sequences being different in each subunit. We were however able to design synthetic phosphopeptides specific for the a1 subunit as substrates. Indeed, making a BLAST with the designed peptide sequences (N-and Cterminal) against mammalian EMBL-EBI database entries showed sequence identity only for the GABA A R a1 subunit. The loop Nterminal (TM3 domain flanking) native peptide NYFTKR-GYAWDGK and its threonine-phosphorylated derivative NYF[pThr]KRGYAWDGK corresponds to the 13 amino acids 334-346 of the a1 subunit, whereas the loop C-terminal (TM4 domain flanking) native peptide EPKKTFNSVSKIDR and the serine-phosphorylated derivative EPKKTFNSV[pSer]KIDR correspond to the 14 amino acids 407-420 of the same subunit ( Figure 1). These peptides contain the minimum consensus sequence for phosphorylation by GAPDH, and are used as substrates to characterize the membrane-bound phosphatases of the GABA A R a1 subunit.

Washed cortical membrane preparation
Washed membranes were prepared mainly from bovine and rat brain cortex. The human cortex sample was frozen on dry ice in the operating room and then thawed for assay. The patient was a girl aged 16 years at the time of curative surgery. She was suffering from epilepsy associated to a non-malignant ganglioglioma. The tissue sample was taken outside the tumor in the posterior part of the right inferior temporal gyrus, with information and consent of the patient and of her parents.
The gray matter was homogenized in a Polytron system in 10 volumes of an ice-cold buffer containing 50 mM Tris pH 7.4, 0.32 M sucrose, 5 mM EDTA and 1 mM EGTA. The homogenate was centrifuged at 800-100 g for 10 min at 4uC. The supernatant (S1) was collected and centrifuged at 100 000 g for 30 min at 4uC. The pellet (P2) was osmotically shocked in 10-20 volumes of ice-cold water and sonicated. The microsome suspension was centrifuged at 100 000 g for 30 min at 4uC. The pellet (P3) corresponding to washed membranes was suspended in a buffer containing 10 mM Hepes (pH 7.4) and 10 mM KCl, and then centrifuged again. The last pellet (P4) was stored at 280uC.
Total protein concentrations were determined using the Bradford-based protein assay reagent (Bio-Rad) with Coomassie Brilliant Blue G-250. Calibration was done with a 20-500 mg/ml BSA range. Absorbance was measured at 570 nm.

Liquid chromatography-electrospray ionization mass spectrometry
Instrumentations. The HPLC-mass spectrometric analysis was performed using a LCQ Advantage ion-trap mass spectrometer (ThermoFinningan) equipped with electrospray ionization (ESI) source in the positive mode. The ESI source was coupled online with liquid chromatography (nano-or micro-LC) systems.
Nano-HPLC chromatography. Eluted peptides were separated with Ultimate 3000 system (Dionex). The peptide preparations were loaded under different conditions through an autosampler (Dionex) onto a reverse-phase C18 capillary column (packed C18 PepMap 15 cm L6180 mm ID63 mm dp, 100 Å -Dionex). Separation was achieved with a gradient elution program. The sample injection-loop was of 2 ml when the microcapillary nano-LC system was used. The peptide mixture (native and phosphorylated peptides) was eluted from the column with gradients in the mobile phase from A (5% acetonitrile/0.1% formic acid/94.9% water, v/v/v) to B (80% acetonitrile/0.1% formic acid/19.9% water, v/v/v). LC separation was resolved at the flow rate of 4 ml/min, by the following gradient conditions: 0-4 min 0-20% B, 4-14 min 20-30% B, 14-15 min 30-100% B, 15-17 min 100% B, 17-20 100-0% B. At the end of the run, the column was equilibrated in solvent A for 3 min. The elution solvents were filtered through a membrane filter (type HA 0.45 mm, Millipore).
Micro-HPLC chromatography. Chromatographic separations in micro-flow were conducted on a reversed phase C18 capillary column (150 mm L62.1 mm ID65 mm dp, 100 Å -Atlantis, Waters) at a flow of 250 ml/min. Injection loop was of 10 ml. Solvent and gradient are the same as in nano-LC. The chromatographic system is Surveyor MSPump (ThermoFinningan) coupled to the autosampler injector Surveyor (ThermoFinningan).
After separation the eluted peptides were directly electrosprayed into the LCQ mass spectrometer at a voltage of 5 kV. The capillary voltage was 3 V, and the temperature was kept at 280uC. The sheath (helium) and auxiliary gas (nitrogen) flow-rate was set at 35 and 15 (arbitrary units) respectively. Helium was used as a collision gas for fragmentation of ions. A full-scan mass spectrum (m/z 200-2000) was followed by fragmentation by CID (collisioninduced dissociation) of the most abundant peak from the full-scan mass spectrum (parent ion), using 25% of the normalized collision energy for obtaining MS/MS spectra. The absolute signal intensity of products ions was used to display the chromatogram allowing the quantification of samples and standards by integrating area of the specified peaks.
Optimization of LC-MS and LC-MS/MS conditions. The LC-MS conditions were optimized by monitoring chemical parameters of standard N-and C-terminal peptides dissolved in the medium used afterwards for phosphatase activity assays. The buffer medium contained Hepes only (10 to 50 mM; pH 7.3) since Tris decreased signal sensitivity. The Mg 2+ ions up to 100 mM did not produce any significant signal decrease. The enzymatic reaction was stopped with ice-cold 10% acetic acid instead of 0.1 M HCl, removing the signals. These conditions enabled us to obtain adequate media for assaying phosphatase activities through measurements of the two native and phosphorylated peptide forms. For the LC-MS/MS method, the collision gas (He) pressure and collision gas voltage were adjusted to get the highest signal of the ionized product for each peptide.
Data analysis. Chromatographic data acquisition, peak integration and quantification were performed using the Xcalibur LC-Quan software package. Mass analysis to determine the nature of fragments in MS/MS was done on the online EXPASY proteomic server using the 'ProteinProspector' tool for MS products (http://www.expasy.ch).

Phosphatase assays using phosphopeptides
Phosphatasic activity of bovine, rat and human washed cortical membranes was measured by using synthetic phosphopeptides as substrates. Reactions were performed in presence of 50 mg/ml (total proteins) of washed cortical membrane, 10 mM Hepes-Tris pH 7.4 and 1 mM MgCl 2 . Enzymatic reactions were initiated by adding 10 mM of phosphopeptide incubated for 10 min at 30uC and stopped by adding ice-cold acetic acid at a final concentration of 10% (w/v). Following a 800 g centrifugation during 10 min at 4uC, the supernatants were analyzed by LC-MS/MS. This centrifugation step was critical to remove membrane suspension that interferes with nano-LC and MS analysis. For these reasons, and to avoid micro-capillary column obstruction we have preferentially used the micro-LC column to characterize the pharmacological profile of phosphatasic activities.
Quantification of appearance of the native peptide and disappearance of the phosphopeptide in different conditions was performed using the calibration curve plotted using the Xcalibur software (P/N 64018-XCALI version 1.3). The calibration curve and linear regression were constructed using a series of standards, native and phosphorylated equimolar mixture, N-or C-terminal peptides at different concentrations 0.5, 1, 2, 5 and 10 mM (n = 2) in 10 mM Hepes pH 7.4, 1 mM Mg 2+ and 10% acetic acid, which was the same milieu condition as in the phosphatase assay after stopping reactions. The linear regression represents peak area of MS/MS chromatogram spectra depending on the peptide concentration ( Figure S1).
To analyze the impact of washed cortical membranes and incubation process on the native and phosphorylated peptides, calibration standards at different concentrations were prepared in the biological matrix using washed cortical membranes (50 mg/ml of total proteins) in 10 mM Hepes, 1 mM Mg 2+ and acetic acid 10% (w/v) incubated at 30uC for 10 min. For phosphopeptides, acetic acid was added before washed cortical membranes in order to prevent the dephosphorylation mechanism ( Figure S2). The total amount of the [native + phosphorylated] peptides retrieved after reaction incubations and LC-MS/MS analyses was constant and identical to the phosphopeptide amount added at the starting time ( Figure S3).

Enzyme kinetics
The best conditions of incubation time and total protein concentration of washed cortical membranes were determined for assaying phosphatase activity with 10 mM of pI 2 a1 phosphopeptide. Protein concentration varied from 1 to 300 mg/ml and incubation time from 0.25 to 30 min, at fixed temperature of 30uC as in previous studies [22]. Initial velocity of the enzymatic reaction was determined by linear regression (slope value). The optimal conditions were obtained with 50 mg/ml total membrane protein concentration and 10 min incubation time. The effects of the substrate effects were measured by varying the pI 2 a1N-P and pI 2 a1C-P concentrations between 3 and 75 mM. Saturation plots were generated via non-linear fit to the Michaelis-Menten equation. The affinity constant (K m ) and maximum velocity (V max ) were evaluated from this curve fitting using the Prism 5 computer program (GraphPad). Double-reciprocal plots with the Lineweaver-Burk equation were also drawn by linear fits using the same software.

Pharmacological profiling of phosphatase activities
Different effectors were tested by addition to the incubation medium at variable concentration: the dicationic ions Mg 2+ , Ca 2+ , Zn 2+ were used as chloride, the polycation spermine as hydrochloride, the anion F 2 as sodium fluoride (NaF), and the inhibitors of some known phosphatases okadaic acid and the two (+/2) enantiomers of p-bromotetramisole oxalate (p-Br-t). For the inhibitory effects, relative 50% inhibition concentration (IC 50 ) values were either determined graphically with a log-scale for the concentration axis or by the 'log (inhibitor) vs. response' nonlinear fit of Prism 5 software (GraphPad), considering activity in the absence of inhibitor as 100%. However, when Zn 2+ was tested with the N and C-terminal peptides as substrates, a step was observed that was considered as an intermediate reference value to determine two sequential IC 50 .
Whole-cell peak currents induced by GABA (100 mM; fast applications of 1 sec pulses every 3 min) were measured for each application and normalized to the maximal response. This maximal response was observed within 0-6 min following the seal of the patch since there was an initial run-up in some cases. The holding potential was 280 mV such that GABA evoked inward currents (Cl 2 equilibrium potentials were measured close to 240 mV).

Statistical analysis
Data points are expressed as mean 6 standard error to the mean (S.E.M.) of repeated measures (n). Non-linear regressions for the kinetics parameters (and corresponding errors), and statistical comparison between two linear regressions by the 'slope test' (case of N-terminal substrate) were assessed using the Prism 6 software (GraphPad). For cluster analysis, ANOVA and post-hoc tests, the JMP 10 software (SAS Institute Inc) was used. In all cases the alpha error was set at 5%.

Ethics statement
All animal experimental procedures were performed in accordance with the European Communities Council Directive (86/ 609/EEC) and were approved by the Animal Experimentation Committee of Paris Descartes University.
The sample of human brain cortex was taken from a registered collection called ''NeurochirEpilepsie'' (DC-2011-1378) affiliated with the biobank CRB-NSPN of Sainte Anne's Hospital Center (CHSA), with approval of the local ethical committee (CPP Ile-de-France II, September 12 th 2011).

Results
The first characterization of the GABA A R a1 subunits dephosphorylation were performed using 33 P-autoradiographic measurements on gel electrophoresis [22], a reliable method but too laborious for routine assays. An efficient method was therefore lacking for rapid assays of the membrane-bound phosphatase activity counteracting the endogenous phosphorylation by GAPDH. We therefore developed a novel tool.
Analytical tool for phosphatase assays specific to GABA A R a1 subunit using phosphopeptides and LC-MS/MS Designing synthetic peptides for the a1 subunit of GABA A R. Two fragments of the mammalian GABA A R intracellular loop I 2 of the a1 subunit (I 2 a1; 87 AA-length) including the two phosphorylated sites at threonine and serine residues were synthesized to study the corresponding phosphatases. The Nterminal peptide corresponds to the initial sequence of the I 2 a1 intracellular loop precisely adjacent to the TM3 domain, whereas the C-terminal peptide corresponds to the end-sequence of the I 2 a1 flanking precisely the TM4 domain ( Figure 1A insert). For each fragment, a native and a phosphorylated form were synthesized by Sigma-Genosys, with distinct molecular masses. We used these peptides as useful probes to investigate experimental conditions for the retention and separation of the native and phosphorylated forms by reversed-phase liquid chromatography (LC) and electrospray ionization (ESI) for a full scan mass spectrum.  Figure 1G, H). These peptides were eluted at 8.13 min for the native peptide and at 8.37 min for the phosphopeptide ( Figure 1G, H inserts). These methods allowed rapid, sensitive and specific simultaneous determinations of the native and phosphorylated forms, qualitatively and quantitatively.
We used initially as substrate for phosphatase assays the Nterminal phosphopeptide PI 2 a1N-P incubated with washed cortical membranes of bovine brain cortex as the enzyme source (Figure 2A, B). In one pilot test we used human brain cortex obtained during neurosurgery to assess that this method can be applied to human frozen tissue ( Figure 2C). A specific phosphatase activity was detected in all assays. Figure 2A, B, C shows the disappearance of phosphopeptide (PI 2 a1N-P) versus enzymatic production of native peptide (PI 2 a1-N) by varying concentration of total membrane proteins and incubation time. It shows the presence of an important membrane-bound phosphatase activity. The corresponding enzymes(s) recognized and dephosphorylated efficiently the a1-subunit substrates, and dephosphorylated the phosphopeptide. The enzyme activity was linear up to 50 mg/ml (total proteins) as shown in Figure 2A. The time course curves showed essentially a linear response up to 10 min ( Figure 2B). These parameters (concentration of enzyme and time incubation) were further applied to all the enzymatic assays and related pharmacological studies, using both N-and Cterminal phosphopeptides as substrates.

Enzyme kinetics of membrane-bound phosphatase activities
The phosphatase kinetics parameters were determined with either PI 2 a1N-P or PI 2 a1C-P as synthetic substrates ( Figure 2D, E). The results were obtained by fitting data into the Michaelis-Menten equation. The phosphatase activities showed an eightfold reduction in catalytic efficiency (V max /K m ) for the phosphoserine peptide (PI 2 a1C-P) compared the phosphothreonine peptide (PI 2 a1N-P), due to a combined K m increase and V max decrease ( Table 1).
The phosphatase activities have two apparent affinities for the N-terminal phosphopeptide as indicated by the Lineweaver-Burk double-reciprocal plot ( Figure 2D, insert). The two regression lines of the plot were significantly different as assessed by the slope test (p,0.01**). This apparent heterogeneity suggests the presence of at least two different species of membrane-bound phosphatases. The phosphatase activities are more efficient at the phosphothreonine peptide. At least one phosphatase species can recognize phosphoserine as a substrate (single apparent slope) ( Figure 2E, insert). This feature can be used to discriminate some of the membrane-bound phosphatases.

Pharmacological profiles of membrane-bound phosphatases
The phosphatase activities were inhibited by the ions Zn 2+ , Ca 2+ and F 2 , Zn 2+ showing the strongest inhibitory effect. Sodium fluoride, a common inhibitor of PP1, PP2A, and PP2B but not of PP2Ca and -b [26] inhibited phosphatase activities with an IC 50 , 2 mM. Okadaic acid, a potent phosphatase inhibitor of PP1, PP2A and PP2B with distinct IC 50 , also inhibited membrane-bound phosphatases.
Metal ion dependence of dephosphorylation. We had previously demonstrated that endogenous phosphorylation requires Mg 2+ ions in the incubation milieu, and that in the presence of these ions GABA A R dephosphorylation occurs spontaneously [22]. Indeed the activities of the PPM family require divalent cations such as Mg 2+ and Mn 2+ , Mg 2+ showing a stronger effect [27]. Other ions such as Ca 2+ and Zn 2+ competitively inhibit the phosphatases by blocking the Mg 2+ /Mn 2+ binding site [28]. Therefore, assays were performed to determine which divalent cations are effective in modulating the phosphatase activities using the N-terminal phosphopeptide as substrate.
Mg 2+ ions. Increasing Mg 2+ concentrations from 1 mM to 1 mM induced a moderate but not linear increase of the phosphatase activities (data not shown). A substantial phosphatase activity was still present without any addition of Mg 2+ , even when the chelators EDTA (5 mM) and EGTA (1 mM) were added (data not shown), indicating a heterogeneity of the dependence of the membrane-bound phosphatases to this ion. The other cations were tested in presence of 1 mM Mg 2+ .
Zn 2+ ions. Previous work [22] showed that Zn 2+ inhibited membrane-bound phosphatase activity at concentrations between 0.1 mM-1 mM. It was therefore important to confirm this observation by the present method using pI 2 a1N-P and pI 2 a1C-P as substrates ( Figure 3). All of the phosphatases activities were completely inhibited by 1 mM Zn 2+ using either substrate ( Figure 3A, 3E). Two apparent IC 50 for Zn 2+ were determined for each peptide: high affinity (H) at 1.7 mM (1.4-2.9 mM confidence interval) ( Figure 3B Figure 3D), of lower extent for the Cterminal substrate ( Figure 3H).
Okadaic acid inhibition. Okadaic acid has been shown to be a potent inhibitor for PP1, PP2A, PP4 and PP5 and a weaker inhibitor for PP2B. PP2C and PP7 are not or very slightly sensitive to inhibition by okadaic acid at micromolar concentrations [29]. The differences in the IC 50 can distinguish these phosphatases. Okadaic acid at 300 nM completely inhibits the membrane-bound phosphatases using the two phosphopeptides ( Figure 4A, B). Two apparent IC 50 for the inhibition of phosphatase activities by okadaic acid using pI 2 a1N-P were determined: a high affinity (H) at 0.15 nM and a low affinity (L) at 5 nM ( Figure 4A). Only one affinity was measured using pI 2 a1C-P with an apparent IC 50 at 0.18 nM ( Figure 4B). These data paralleled those obtained with Zn 2+ and confirm the presence of at least two species of membrane-bound phosphatases.
Inhibition of the phosphatasic activities by (+/2)-p-Br-t oxalate. The levamisole, (2) enantiomer of the p-bromotetramisole oxalate ((2)-p--Br--t) is a potent alkaline phosphatase inhibitor, whereas the same enzyme is completely insensitive to the (+) enantiomer (+)-p--Br--t) which is often used as a negative control [30]. We therefore tested both enantiomers with increasing concentrations ( Figure 4C). The membrane phosphatase activities were completely inhibited by both enantiomers at 1 mM, with IC 50 values of 0.1 mM for (+)-p--Br--t and 0.3 mM for (2)-p--Br--t, showing that the alkaline phosphatase is not involved in the dephosphorylation of the N-terminal phosphopeptide.
Inhibition by other ions. By the N-terminal phosphopeptide method, fluoride (F 2 ) reduced phosphatase activities by 45% at 1 mM and almost 90% inhibition was observed at 10 mM, indicating an IC 50 in the order of 2 mM. Ca 2+ inhibited phosphatase activities at 1 mM by only 30% and by slightly less than 60% at 10 mM indicating an IC 50 in the order of 5 mM, which is much less potent than inhibition by Zn 2+ (Figure 4D).
Polycations inhibition. Some mammalian phosphatases are inhibited by various polycations [26]. The effects of spermine (a tetracationic polyamine) vary from one enzyme to another. Some protein tyrosine phosphatases are stimulated [31][32][33], and other phosphatases are inhibited [34] [35]. For instance, spermine inhibits both PP1 and PP2A with similar potency [36]. A more precise reason to test especially spermine among other polycations is that its content is increased in epileptogenic zones [37] where a deficiency in GABA A R endogenous phosphorylation is also observed [25]. It appeared therefore relevant to examine whether this polyamine has an effect on the phosphatases responsible for a1-subunit dephosphorylation. Spermine at 1 mM moderately inhibited the phosphatase activities by only (10%) and 20% at 10 mM showing that spermine is a less effective inhibitor than Zn 2+ ions ( Figure 4D).

Effects of the okadaic acid on the rundown of GABA A currents in acutely isolated cortical neurons
We measured the rundown of the GABA-induced currents in pyramidal neurons acutely dissociated from rat cortex ( Figure 5A). Rundown of GABA A currents is a time-dependent response decrease due to a dephosphorylation process [7] (Figure 5B), which does not require the presence of the receptor agonist GABA, thus differing from the observed desensitization process during GABA applications. We had previously shown that the phosphatase inhibitor okadaic acid (10 mM) added to the pipette milieu has a variable effect on the GABA A current rundown in four neurons [20]. As illustrated in Figure 5C, the rundown could be abolished in some cells. To clarify this variable effect, okadaic acid was assayed on a larger number of rat pyramidal cells (n = 17). Normalized maximal responses of neurons displayed different effects with okadaic acid ( Figure 5D). We applied the hierarchical clusters test to the parameter of the 'mean normalized currents' (from t = 3 to t = 30 min) for all okadaic acid-treated neurons. Four groups of neurons were significantly distinguished by ranking rundown velocity as 'Very Rapid' -'Mid Rapid' -'Mid Slow' -'Very Slow'. In addition, the whole group of okadaic acid-treated cells was significantly different from the control cells group (p,0.01 **), using a two-tailed Student's t-test assuming unequal variances, in spite of the broad individual variation induced by addition of the phosphatase inhibitor. The 4 groups were considered separately using one-way ANOVA with Dunnett's post hoc test: the 'Very Rapid' group which was not sensitive to okadaic acid was identical to control (Ns), and the 3 other groups of neurons which responded to okadaic acid were statistically different from the control (p,0.01** for 'Mid Rapid', p,0.0001*** for both 'Mid Slow' and 'Very Slow'). This functional approach thus also shows heterogeneity in terms of phosphatase activities in pyramidal neurons.

Discussion
We have developed an analytical tool to characterize for the first time the membrane-bound GABA A -R a1-specific phosphatase activities, which were never previously identified or characterized. These phosphatases can be considered as novel targets for antiepileptic drugs research. Indeed, in recent clinical studies it was proposed to modulate the activity of protein phosphatases for therapeutic benefits [38].
Improved LC-MS-based techniques led to reliable alternatives for peptide analysis and protein quantification [39]. A known limitation in proteomic analysis of membrane proteins by mass spectrometry is the difficulty raised by their amphipathic nature [40]. Thus, detection and characterization of phosphatase activities in membrane preparations by LC-MS/MS analysis and phosphopeptides as substrate is a method not used until now, and its development was a challenge.
We used two fragments specific for the intracellular loop of a1 subunits, which can be used for all mammalians. The principle of the present phosphatase assay by LC-MS/MS is based on the measurement of relative MS signals of both phosphopeptides and native peptides resulting from dephosphorylation. We initially characterized these peptides by examining their specific MS/MS spectra of the related dominant ions. The advantage of this method was to analyze simultaneously native and phosphorylated peptides. Assaying washed brain membranes including human cortical membranes as a source of phosphatase activity we observed an important dephosphorylation by transformation of N-and C-terminal phosphopeptides into native peptides, showing thus that the membrane-bound phosphatases recognized our I 2 a1 probes as substrates. This result was crucial since it was a challenge to study such a complex membrane proteins preparation and was a prerequisite to characterize these phosphatases.
A comparative analysis of kinetics parameters of these phosphatases ( Table 1) showed that the N-terminal substrate was more efficient (higher specific activity and higher affinity) than the C-terminal one. For the N-terminal phosphopeptide the Lineweaver-Burk plot was not linear. This point was further studied using slope tests: two significantly different apparent affinities were observed. We noted a reduction of the phosphatase activity heterogeneity of C-terminal comparing to the N-terminal substrate. This was confirmed by the data obtained by okadaic acid inhibition, where we observed two distinct affinities for Nterminal substrate and only one apparent affinity for C-terminal. It has been shown that some phosphatases have indeed a better substrate preference for phosphothreonine than for phosphoserine residues in short peptides [41]. These observations strongly support the hypothesis of several phosphatases regulating the GABA A R function.
The protein phosphatases are generally classified according to their substrate specificity, ion requirement and sensitivity to inhibitors, and are divided into two major categories-protein: serine/threonine phosphatases (PSTPs) [42] and tyrosine phosphatases (PTPs) which include dual-specificity phosphatases (DUSPs) [43]. Two distinct PSTPs gene families have been described: PPM and PPP. The protein Ser/Thr phosphatases PP1, PP2A, and PP2B of the PPP family, together with PP2C of the PPM family, account for the majority of the protein serine/ threonine phosphatase activity. The PPM family is composed of Mg 2+ -dependent phosphatase including PP2C, pyruvate dehydro-  genase phosphatase, and PP2C-''like'' phosphatases [44]. These enzymes play a key role in neuronal plasticity, learning and memory [16], and in several neurological disorders such as amyotrophic lateral sclerosis and epilepsy [45]. An important issue is to identify the metal ions required for optimal phosphatase activity. Both PP1 and PP2A are active in the absence of divalent cations, whereas PP2B and PP2C have respectively an absolute requirement for Ca +2 and Mg 2+ [36]. The activity of membrane-bound phosphatases was tested in the presence of a variety of factors that can interfere with these enzymes. Addition of increasing concentrations of Mg 2+ exhibited moderate effects on activation of phosphatase activities, and in the presence of chelators a residual phosphatase activity was observed, revealing the presence of both Mg 2+ -dependent and Mg 2+ -independent protein phosphatases, indicating again the heterogeneity of these enzymes. A similar heterogeneity was reported for the atrial natriuretic peptide receptor (NPR-A) phosphorylated on 4 serine and 2 threonine residues by an unknown kinase(s) and dephosphorylated by two unknown phosphatases, the activity of which was dependent or not on Mg 2+ [46]. Ca 2+ , fluoride and spermine at millimolar concentrations inhibited the phosphatase activities. The weak inhibition by spermine was interesting since its content is increased in the epileptogenic tissue [37]. One may speculate that such inhibition could counteract the increased GABA A current rundown observed in neurons isolated from human epileptogenic cortical tissue. However, the very high IC 50 (.10 mM) of spermine inhibition makes such compensation unlikely, since millimolar spermine expressed as a percentage of that of control samples. (A) Total curve inhibition by Zn 2+ (0.1 mM-1 mM) with pI 2 a1N-P at 10 mM as substrate, the curve could be fitted to a two-step mechanism: 'H' high affinity (B), and 'L' low affinity (C). (D) At very low Zn 2+ concentrations (1 pM-200 nM) activation of PPase activity was observed. (E) Total curve inhibition by Zn 2+ (0.1 mM-1 mM) in presence of pI 2 a1C-P at 10 mM, two affinities was observed: 'H' high affinity (F), and 'L' low affinity (G). (H) PPase activity at very low Zn 2+ concentrations (1 pM-200 nM). The IC 50 values are mentioned with their 95% confidence intervals and calculated from nonlinear regression curve fits using GraphPad Prism 5. doi:10.1371/journal.pone.0100612.g003 concentrations were never observed in the human cortical tissue, even in pathological conditions. Inhibition observed with p-bromotetramisole oxalate (p-Br-t) allowed us to discard the alkaline phosphatase as an involved enzyme in the present mechanism. It has been shown that p-Br-t can inhibit other phosphatases such as PP1, PP2A, PP2B and PP2C with variable efficacy [30]. The same authors showed an inhibition by both enantiomers for PP2A and PP2C, however with different IC 50 in the millimolar range. They also observed that the (+) enantiomer is more potent than the (2) enantiomer for PP2C. Likewise, we showed that inhibition by (+)-p-Br-t was slightly more effective than that by (2)-p-Br-t for the membrane-bound phosphatase activity using the N-terminal phosphopeptide. However, the IC 50 we measured were 3-7 fold lower than those for PP2C and PP2A. Thus these classical phosphatases were very unlikely involved in the GABAergic a1 mechanism.
We had previously showed that sodium orthovanadate inhibits membrane phosphatases [22] and reduced rundown of GABA A currents [20]. This phosphate analog could not be used in our conditions for phosphatase characterization, because a complete ESI-MS/MS signal extinction occurred in positive mode occurred, likely due to the negative charge and non-volatility of the VO 4 32 anion. Indeed, all ESI-MS/MS manufacturers totally prohibit the use of phosphate buffers for the same reason [47].
It was of special interest to investigate the zinc-sensitivity of the membrane-bound phosphatase, since zinc oxide was known to present anti-epileptic effects and used for this purpose during the 19 th century [48]. The effect of Zn 2+ is variable depending on the phosphatases: these ions as well as Fe 2+ are activators at the catalytic site of PP1, considered as a metallo-enzyme [49], whereas Zn 2+ inhibits the PP2C class by competition with Mg 2+ /Mn 2+ at the metal binding site of the catalytic domain [26,28]. We had previously shown in washed brain membranes that complete inhibition of GABA A R dephosphorylation occurred with Zn 2+ using the autoradiographic method [22]. In the present study we found that Zn 2+ in the micro-to millimolar ranges was one of the most potent tested phosphatase inhibitors. Two apparent affinities were determined using both N-and C-terminal phosphopeptides. Free cellular Zn 2+ concentrations (picomolar range) are orders of magnitudes lower than the total cellular Zn 2+ content (nanomolar range) [50]. In these ranges, we observed a series of activations of phosphatasic activities for the N-terminal substrate. Our washed cortical membrane preparations are cytosol free, without proteins chelating Zn 2+ such as metallothioneins [51]. Therefore, the observed multiple activations of phosphatasic activities by zinc are rather reflecting different affinities of distinct phosphatases for these ions. These data indicate again the heterogeneity of the membrane-bound phosphatases dephosphorylating GABA A R a1 subunits. Unfortunately, the effect of Zn 2+ could not be tested directly on GABA A receptor currents in whole-cell patch-clamp of acutely isolated pyramidal neurons because Zn 2+ addition into the pipette milieu systematically induced breakdown of the nano-seal at the tip of the glass micropipettes (data not shown).
The organic inhibitors such as okadaic acid are very useful to distinguish some classes of protein phosphatases [52]. Okadaic acid inhibits various phosphatases in mammalian brain with very distinct IC 50 values: PP1 (270 nM), PP2A (2 nM), PP2B (3.6 mM), PP4 (0.2 nM), PP5 (1.4-10 nM) and PP6 (2 nM) whereas PP2C, PTP, PP7 acid phosphatase and alkaline phosphatase were not inhibited by up to 1-10 mM okadaic acid [36]. We showed that it strongly inhibited phosphatase activities but the apparent IC 50 values were different from those described for the classical phosphatases, suggesting that the membrane-bound phosphatases are not yet identified.
The rundown of the GABA A R-mediated responses in dissociated neurons involves an endogenous phosphorylation-dephosphorylation process [7,20]. Okadaic acid at 10 mM was tested on the GABA A response rundown. Four groups of cortical neurons were distinguished suggesting that each group could correspond to distinct phosphatase types. The group showing no inhibition indicates the presence in these cells of okadaic-resistant phosphatase activity. The 3 groups with different sensitivities to this inhibitor suggest the presence of at least two distinct classes of phosphatases, a further argument for their heterogeneity. In addition, okadaic acid at 10 mM inhibited most phosphatasic activities by the phosphopeptide method. This observation suggests that in the course of preparation of the washed cortical membranes, a loss of some GABA A R a1 subunit phosphatases weakly attached at the membranes occurred.
We therefore clearly propose heterogeneity of the phosphatases regulating GABA A R endogenous phosphorylation at the cellular level. In addition to the above-mentioned case of NPR-A, other examples of dephosphorylation of a single protein by multiple distinct Ser/Thr phosphatases are known, such as centrin [53] and [54]. The a1-subunit dephosphorylation by multiple phosphatases in GABA A R likely reflects alternate modulations of the inhibitory neurotransmission, and also different functional responses to ions and endogenous effectors in neurons. Since the involved phosphatases do not correspond to the classical enzymes, it will be of great interest to isolate these membrane-bound phosphatases and to identify them in a future study.

Conclusion and prospects
We have developed an analytical tool using LC-MS/MS method to study membrane-bound phosphatases specific to a1 GABA A R, including those prepared from human brain tissue. This method is highly sensitive and can be applied to crude extracts as well as to purified enzymes and would be useful for membrane phosphatases assays. Our main finding is the heterogeneity of the phosphatases dephosphorylating the GABA A R a1 subunit. Atypical pharmacological profiles were obtained for these phosphatases as shown by the zinc-sensitivity, excluding classical phosphatases even though sharing some of their properties. A challenge for future studies will be to purify these different membrane-bound phosphatases using the present LC-MS/MS method for monitoring. Identifying the membrane-bound phosphatases of GABA A receptors will improve our understanding of the molecular mechanisms modulating GABAergic function and their alterations in patients with drug-resistant seizures. These advances would also provide the basis for development of new antiepileptic drugs by targeting the phosphatases specific to the GABA A R a1 subunit.