Functional Human α7 Nicotinic Acetylcholine Receptor (nAChR) Generated from Escherichia coli

Abstract

Human Cys-loop receptors are important therapeutic targets. High-resolution structures are essential for rational drug design, but only a few are available due to difficulties in obtaining sufficient quantities of protein suitable for structural studies. Although expression of proteins in E. coli offers advantages of high yield, low cost, and fast turnover, this approach has not been thoroughly explored for full-length human Cys-loop receptors because of the conventional wisdom that E. coli lacks the specific chaperones and post-translational modifications potentially required for expression of human Cys-loop receptors. Here we report the successful production of full-length wild type human α7nAChR from E. coli. Chemically induced chaperones promote high expression levels of well-folded proteins. The choice of detergents, lipids, and ligands during purification determines the final protein quality. The purified α7nAChR not only forms pentamers as imaged by negative-stain electron microscopy, but also retains pharmacological characteristics of native α7nAChR, including binding to bungarotoxin and positive allosteric modulators specific to α7nAChR. Moreover, the purified α7nAChR injected into Xenopus oocytes can be activated by acetylcholine, choline, and nicotine, inhibited by the channel blockers QX-222 and phencyclidine, and potentiated by the α7nAChR specific modulators PNU-120596 and TQS. The successful generation of functional human α7nAChR from E. coli opens a new avenue for producing mammalian Cys-loop receptors to facilitate structure-based rational drug design.

Introduction

Human Cys-loop receptors are promising therapeutic targets for various neurological disorders and diseases (1–4). Structure-based drug design for these receptors requires their high-resolution structures (5). Although Cys-loop receptors contain only four major receptor types, including nicotinic acetylcholine receptors (nAChRs),² serotonin 5-HT3 receptors, glycine receptors, and GABA_A receptors, each receptor type often has multiple subtypes that form numerous functional distinct receptors. Among the human Cys-loop receptors, high-resolution structures have been obtained for only the β3 GABA_A and α3 glycine receptors (6, 7). Structures for other eukaryotic Cys-loop receptors include the mouse serotonin 5-HT3A receptor (8), the zebrafish α1 glycine receptor (9), the Caenorhabditis elegans GluCl (10, 11) and the muscle-type nicotinic acetylcholine receptor (nAChR) from Torpedo marmorota (12). The dichotomy between the small number of available structures and the relatively large receptor population in the superfamily indicates the technical difficulties for structural determination of these receptors. One of the greatest challenges for structural determination of Cys-loop receptors and similarly complex human membrane proteins is the production of a large quantity of well-folded functional proteins.

The α7 nAChR (α7nAChR) is one of the most abundant nAChR subtypes found in the brain (13, 14). It is also expressed in a wide variety of non-neuronal tissues (15). It has been implicated in diverse biological functions and is an important target for therapeutics (1). α7nAChR mostly forms a homo-pentameric ligand-gated ion channel (pLGIC) that conducts calcium and other cations, though heteromeric α7β2-nAChR has also been found in both heterologous expression systems and native neurons (16, 17). Structures of the extracellular domain (ECD) and transmembrane domain (TMD) of α7nAChR have been determined separately by x-ray crystallography (18) and NMR (19), but the full-length human α7nAChR has not previously been obtained from any species in a form suitable for structure determination (20, 21).

Currently available structures for Cys-loop receptors were obtained from proteins expressed in mammalian (6, 8) and insect (7, 9–11, 22) cell lines. We chose an alternate approach: to produce functional full-length human α7nAChR in E. coli. Production of recombinant proteins in E. coli is fast and inexpensive relative to other expression systems. In addition, E. coli may produce a more homogeneous population of purified proteins due to its limited ability for post-translational modifications (23). In contrast, the native pathway for α7nAChR expression in mammalian cells involves subcellular trafficking through multiple subcellular compartments, where specific chaperone proteins and post-translational modifications aid folding and assembly. Thus, α7nAChR purified from mammalian or insect cells contain a mixture of receptors with different post-translational modifications representing various stages of maturation, including glycosylation, palmitoylation, and alternate disulfide conformations (24–26). These modifications were reported to be essential for the proper assembly and navigation of functional α7nAChR to the mammalian cell surface (24). We hypothesize that these modifications are not needed for the proper folding and pentameric assembly of α7nAChR in the simpler expression environment of E. coli.

Here we report that α7nAChR purified from E. coli retains signature properties of native α7nAChR, including the ability to assemble into pentameric structures, to bind specific ligands, and to form functional ion channels that can be activated by agonists, inhibited by channel blockers, and enhanced by α7nAChR-specific positive allosteric modulators. The study demonstrates that E. coli is capable of producing human Cys-loop receptors. It also suggests that the post-translational machinery may not be as essential as previously thought for expressing functional complex membrane proteins.

Results

Essential Conditions to Obtain Full-length Human α7nAChR from E. coli

Our expression construct consists of DNA encoding the wild-type full-length human α7nAChR with the signal sequence replaced by that of pelB (a bacterial leader sequence) for expression in E. coli. An 8-histidine tag was added to the C terminus to aid in purification. All subsequent mention of α7nAChR refers to this construct. Initial attempts to express full-length human α7nAChR in E. coli resulted in low expression levels and the formation of inclusion bodies. These problems could not be solved by expression of α7nAChR as a fusion protein with MBP, as a chimera with the bacterial homolog ELIC (27) or with various modifications to the protein sequence including codon optimization, deletion of the intracellular loop between the TM3 and TM4 helices, truncation of the N and C termini, and specific mutagenesis of suspected problem sequences.³ The most significant improvement in the expression of α7nAChR was achieved by using osmotic shock (0.5 m sorbitol) at low temperature (15 °C) to induce E. coli native chaperones (28) and using the α7nAChR agonist choline as a chemical chaperone (29) during induction (Fig. 1a). After testing multiple E. coli strains, we found Rosetta(DE3)pLysS to be most suitable for α7nAChR expression. For the choice of detergents, we found that n-dodecylphosphocholine (DPC) was most effective for extracting α7nAChR from cell membranes. Both DPC and n-dodecyl β-d-maltoside (DDM) were suitable for purification. Fig. 1b shows the size exclusion chromatography (SEC) profile of purified α7nAChR, which is consistent with a pentameric assembly. Under these conditions, we can obtain 1.2 ± 0.2 mg pentameric α7nAChR from each liter induction culture (n = 6). Purified α7nAChR migrates on SDS-PAGE with an apparent molecular weight near 50 kDa (Fig. 1c), consistent with the de-glycosylated monomer of α7nAChR (30). The protein identity was further confirmed by mass spectrometry, which showed more than 51% coverage of tryptic fragments (Fig. 1d). Although the C terminus was not identified by mass spectrometry, the full-length α7nAChR was ensured by C-terminal His-tag purification. Representative mass spectra are shown in the supplemental data.

FIGURE 1.

Human α7nAChR generated from E. coli. (a) flowchart providing an overview of α7nAChR expression and purification; (b) size exclusion chromatography elution profile of purified α7nAChR; (c) Coomassie-stained SDS-PAGE (lane 1) and Western blot identifying purified α7nAChR (lane 2) compared with the uninduced control (lane 3); (d) sequence of human α7nAChR highlighted with residues identified by mass spectrometry (>51% coverage). Representative mass spectra are provided in the supporting materials.

Pentameric Assembly of Purified α7nAChR Visualized by EM

We used electron microscopy (EM) to analyze the oligomeric state of α7nAChR. To improve the quality of negative-stain EM images, we reduced excess detergent by exchanging DDM to the chemically similar lauryl maltose neopentyl glycol (LMNG), which forms micelles at very low concentrations (0.01 mm). In the negative-stain EM images, α7nAChR was observed as discrete particles (Fig. 2a). From 113 micrographs, 7300 particles were sorted into 30 classes by two-dimensional classification (Fig. 2b). Among them, at least 13 classes represent essentially different versions of pentameric particles of ∼6–7 nm in diameter with a pore in the middle (Fig. 2b, white box), accounting for 47% of the particles analyzed (Fig. 2c). The remaining classes may include various tilted views or aberrant particles. Under these sample conditions, α7nAChR appears to prefer an orientation showing the end-on view (i.e. top or bottom), precluding three-dimensional reconstruction.

FIGURE 2.

Negative-stain electron microscopy of purified α7nAChR demonstrates pentameric assembly. (a) representative negative-stain image of purified α7nAChR. Scale bar, 20 nm; (b) two-dimensional class averages of negative-stain particles. Classes clearly depicting pentamers (white boxes) account for 47% of the particles analyzed; (c) distribution of particles among the class averages.

Purified α7nAChR Retains Binding Sites for Bungarotoxin, PNU-120596, and TQS

Native human α7nAChR demonstrates nanomolar binding affinity for the antagonist bungarotoxin (31), and micromolar binding affinity for positive allosteric modulators specific to α7nAChR, such as 1-(5-chloro-2,4-dimethoxyphenyl)-3-(5-methyl-1,2-oxazol-3-yl)urea (PNUS-120596) (32) and 4-(1-naphthyl)-3a,4,5,9b-tetrahydro-3H-cyclopenta[c]quinoline-8-sulfonamide (TQS) (33). We measured binding of bungarotoxin to the purified α7nAChR immobilized on NiNTA plates using a chemiluminescent assay and found an affinity of 4 ± 1 nm (Fig. 3a), which is similar to the reported value of ∼5 nm measured from α7nAChR heterologously expressed in SH-EP1 human epithelial cells (31). Using surface plasmon resonance, we also measured binding affinities of PNU-120596 and TQS to purified α7nAChR immobilized on a NiNTA chip (Fig. 3, b and c). For our purified α7nAChR, PNU-120596 had an apparent disassociation constant (K_d) of 1.1 ± 0.1 μm, which is close to the previously reported half maximal effective concentration (EC₅₀) of 1.6 ± 0.4 μm (32). TQS has a relatively lower affinity than PNU-120596 with a K_d of 7.8 ± 3.2 μm, which is similar to the EC₅₀ of 6.2 ± 0.6 μm reported previously (33). The results provide convincing evidence that human α7nAChR purified from E. coli retains binding sites for these ligands as observed in native membranes.

FIGURE 3.

Purified α7nAChR retains binding sites for antagonists and positive allosteric modulators. (a) binding of the competitive antagonist bungarotoxin to purified α7nAChR. The data were obtained by chemiluminescence assay (n = 3) and fit by nonlinear regression (K_d = 4 ± 1 nm). (b) binding of positive allosteric modulators PNU-120596 (square) and TQS (circle) to purified α7nAChR. The data were obtained by surface plasmon resonance (n = 3) and fit by nonlinear regression. The K_d values for PNU-120596 and TQS are 1.1 ± 0.1 μm and 7.8 ± 3.2 μm, respectively.

Purified α7nAChR Remains Functional

We measured the functional integrity of purified α7nAChR using quantitative bungarotoxin pull-down experiments and electrophysiology measurements in Xenopus oocytes.

α-Bungarotoxin binding to α7nAChR has been used as a measure of functional α7nAChR at the cell surface (34). We found that 92.9 ± 0.3% of purified α7nAChR was pulled-down by α-bungarotoxin immobilized to Sepharose resin, as measured by absorbance at A₂₈₀ (n = 3). The result suggests that most of the purified α7nAChR is folded properly in an active state.

Oocytes injected with purified α7nAChR showed activation currents upon applying the agonist choline, acetylcholine, or nicotine in TEVC experiments (Fig. 4a). Note that choline is a selective agonist for α7nAChR (35). The observed currents from protein injection (6 ± 2 nA, n = 5) are smaller than currents from cRNA injection (150 ± 50 nA, n = 5). The smaller current is typical for purified channel proteins injected into oocytes (36–38). Oocytes tolerate injection of only a limited amount of reconstituted protein. We injected only 5 ng of purified α7nAChR compared with 25 ng of cRNA. Each cRNA molecule can result in hundreds of synthesized protein molecules (39). Therefore, it is anticipated that cRNA injection will generate many more channels and larger currents than protein injection. The currents were absent from oocytes injected with buffer only (Fig. 4a). Fast activation and desensitization are signature properties of α7nAChR (13). An expanded time scale demonstrates similar kinetics for currents resulting from protein and cRNA injections (Fig. 4b). Representative traces in Fig. 4c show that the currents were inhibited by channel blockers N-(2,6-dimethyl-phenylcarbamoylmethyl) trimethylammonium chloride (QX-222) (40) (79 ± 3% versus 74 ± 5% for protein and cRNA injections, respectively, n = 5) and phencyclidine (PCP) (41, 42) (92 ± 2% versus 93 ± 3% for protein and cRNA injections, respectively, n = 5). Protein- and cRNA-injected oocytes also showed similar potentiation by PNU-120596 (9 ± 3 versus 12 ± 3 for protein and cRNA injections, respectively, n = 5) and TQS (18 ± 5 versus 23 ± 4 for protein and cRNA injections, respectively, n = 5), both of which are α7nAChR specific positive allosteric modulators (32, 43). Representative traces are shown in Fig. 4d. These results demonstrate that human α7nAChR generated from E. coli can form functional channels that retain pharmacological characteristics of native α7nAChR.

FIGURE 4.

Reconstituted α7nAChR forms a functional ion channel. Channels generated by injecting purified α7nAChR (protein) or α7nAChR cRNA (cRNA) into oocytes respond similarly to (a) activation by the agonists choline (10 mm), acetylcholine (100 μm), and nicotine (100 μm). These agonists did not generate current in saline-injected oocytes. (b) channels from protein and cRNA injection showed similar kinetics to activation by acetylcholine (100 μm). (c) channels from protein and cRNA injection are inhibited similarly by the channel blockers QX-222 (100 μm) and PCP (100 μm), and potentiated by (d) PNU-120596 (10 μm) and TQS (30 μm). The bars over the traces indicate application of the indicated ligands. Vertical scale bars represent 5 nA and 50 nA for protein and cRNA injections, respectively. Horizontal scale bars represent 10 s for all traces except the one in (b) which is 1 s. Traces are representative of n = 5 independent oocytes.

Discussion

Among all the members of the Cys-loop superfamily of ligand gated ion channels, α7nAChR is known to be particularly difficult to express (20, 21). Its functional expression in mammalian cells is cell-type dependent, which has been linked to the requirement for specific chaperone proteins and post-translational modifications (24). Even in permissive cells, it has been estimated that only 62% of the total α7nAChR protein present in the cell is in a mature functional form (31). This may be a problem inherent to heterologous expression in eukaryotic cells, not only for α7nAChR but also for other homologous receptors. Indeed, those eukaryotic Cys-loop channels for which structures have been successfully determined have all been mutagenized or processed in vitro in an effort to improve monodispersity of the final product (6–11, 22). Production of functional human α7nAChR in E. coli suggests that specific chaperone proteins and post-translational modifications are not required for channel function, but instead are a consequence of sub-cellular trafficking in eukaryotic cells. Post-translational modification in E. coli is limited (23), which may improve homogeneity of the protein expression. Our results show that the E. coli chaperones induced by osmotic and cold shock as well as the chemical chaperone choline are sufficient for producing large quantities of well-folded α7nAChR.

In addition to the expression conditions, the oligomeric state of α7nAChR was sensitive to purification procedures. Under optimal conditions, the pentameric form was stable for 2 or 3 days at 4 °C. However, the isolated pentamer fraction had a tendency to aggregate under conditions of low ionic strength or detergent concentration; or to dissociate to smaller oligomeric structures under conditions of high ionic strength or detergent concentration. The tendency to form multiple oligomeric structures may be an intrinsic property of α7nAChR. Metabolically labeled α7nAChR obtained from mammalian PC12 cell culture through microscale purification was found to form multiple oligomeric structures (44). A mutated construct of the zebra finch α7nAChR expressed in HEK293F cells also showed micro-aggregation by negative-stain EM (45). Our study suggests that homogenous α7nAChR can be obtained by carefully controlling the purification conditions and time window.

E. coli readily expresses the bacterial homologues of Cys-loop receptors, such as GLIC (46, 47) and ELIC (48). Because of the traditional wisdom that post-translational modification is essential for producing functional eukaryotic channel proteins, using E. coli to express mammalian Cys-loop receptors is almost uncharted territory. Our results challenge that traditional wisdom and suggest that careful manipulation of expression conditions can allow production of functional human α7nAChR in E. coli. It is likely that other members of the Cys-loop receptor superfamily can also be produced from E. coli following similar protocols. Given the pharmaceutical interest in these receptors as therapeutic targets, perhaps their expression in E. coli should be revisited.

Experimental Procedures

Expression and Purification of Human α7nAChR from E. coli

Human α7nAChR (UniProtKB P36544: ACHA7_HUMAN) with the pelB leader sequence (removed during expression) and a C-terminal 8-histidine tag was expressed from the T7 promoter using the expression vector pTBSG1 (49). There was no modification to the native α7nAChR sequence. The protein was expressed in Rosetta 2(DE3)pLysS (Novagen) using the Marley protocol (50). Induction with 0.2 mm IPTG was for 16 h at 15 °C in LB broth containing 500 mm sorbitol and 10 mm choline. Cells from 1 liter induction medium were resuspended in 25 ml of buffer A containing 50 mm Tris, pH 8, 500 mm NaCl, 500 mm sucrose, 10 mm choline, 10% glycerol, and HALT protease inhibitor. All subsequent operations were at 4 °C. Cells were lysed using an Avestin Emulsiflex-C3 homogenizer. The cell lysate was adjusted to 0.33% DPC and 20 mm imidazole and incubated for 2 h. The insoluble fraction was then removed by ultracentrifugation (1 h 200 k × g) and the supernatant incubated with 3 ml of NiNTA resin (GEHealthcare) for 2 h, mixing by inversion. The resin was transferred to a column and washed with 10 column volumes (cv) of buffer A, followed by 10 cv buffer B containing 100 mm imidazole, 10% glycerol, 150 mm NaCl, and 0.05% DDM. α7nAChR was eluted with buffer C containing 250 mm imidazole, 10% glycerol, 150 mm NaCl, and 0.05% DDM. The purified protein was concentrated to 0.5 mg/ml using a Vivaspin 2 molecular weight cut-off 100,000 centrifugal filter (Sartorius, Germany), and the pentamer fraction was isolated by SEC using a S200 10/300 or S200 3.2/300 column (GEHealthcare). Protein purity and identity were assessed by SDS-PAGE and Western blotting analysis, respectively. Western blotting analysis was performed using a monoclonal mouse antibody to the C terminus of α7nAChR (catalogue number 60220–1-Ig, Lot number 10001012, ProteinTech). Antibody specificity was confirmed by the manufacturer and by comparison of E. coli extracts with and without α7nAChR expression.

Proteomics Analysis to Confirm α7nAChR Identity

Purified α7nAChR was digested using an in-solution digestion protocol as described previously (51, 52). Tryptic peptides of α7nAChR were injected into a 75 μm × 150 mm BEH C18 column (particle size 1.7 μm, Waters), separated using a Waters nanoACQUITY Ultra Performance LC^TM (UPLC^TM) System (Waters, Milford, MA) and subjected to LC-MS/MS analyses performed on a Thermo Scientific LTQ Orbitrap XL (Thermo Scientific) with a Finnigan Nanospray II electrospray ionization source in the data dependent mode using the TOP10 strategy (53). In brief, each scan cycle was initiated with a full MS scan of high mass accuracy (400–2,000 m/z; acquired in the Orbitrap Elite at 6 × 104 resolution setting and automatic gain control (AGC) target of 106), followed by MS/MS scans (AGC target 5,000; threshold 3,000) in the linear ion trap on the 10 most abundant precursor ions. Selected ions were dynamically excluded for 30 s. Singly charged ions were excluded from MS/MS analysis. MS/MS spectra were searched by using the SEQUEST algorithm against a composite database containing the International Protein Index (IPI) (human) protein sequences and their reverse sequences. Search parameters allowed for two missed tryptic cleavages, a mass tolerance of ±10 ppm for precursor ion, a mass tolerance of ±0.02 Da for product ion, a static modification of 57.02146 Da (carboxyamidomethylation) on cysteine and a dynamic modification of 15.99491 Dalton (oxidation).

Negative Stain Electron Microscopy and Image Analysis

To prepare samples for electron microscopy, purified α7nAChR was subjected to SEC in 10 mm HEPES pH 7.4, 150 mm NaCl, and 0.002% LMNG. Samples (3 μl) from the SEC peak fraction were absorbed to a glow-discharged 400-mesh carbon-coated copper grid. Samples were then stained with two drops of uranyl acetate (2%), blotted with filter paper, and imaged on a TF20 electron microscope (FEI, Hillsboro, OR) equipped with a field emission gun. Images were recorded at ×100,000 magnification on a 4k × 4k Gatan Ultrascan CCD camera (Gatan, Warrendale, PA). From 113 micrographs, a total of 7300 particles were picked using EMAN2 (54) and then extracted and classified using Relion 1.4 (55).

Binding Assays

A chemiluminescent assay was used to measure binding of biotinylated bungarotoxin to purified α7nAChR immobilized on 96-well NiNTA plates (ThermoFisher, Waltham, MA). Briefly, each well was incubated with 30 μl 100 μg/ml purified α7nAChR for 6 h at 4 °C, then washed three times with 50 mm sodium phosphate buffer pH 7.4, 0.025% DDM, 0.05 mg/ml asolectin, followed by another wash with 50 mm sodium phosphate, pH 7.4, 0.1 mg/ml asolectin. Biotin conjugated bungarotoxin (ThermoFisher, Waltham, MA) was added at the indicated concentrations in 50 mm sodium phosphate buffer and incubated overnight at 4 °C. Excess bungarotoxin was removed by three 200-μl washes with Tris-buffered saline, 0.05% Tween. Avidin-HRP (ThermoFisher, Waltham, MA) was added to each well at 1 μg/ml for 1 h and then developed using SuperSignal West Pico Chemiluminescent Substrate (ThermoFisher, Waltham, MA). Negative controls were treated identically, except no α7nAChR was present. For the α7nAChR-specific positive allosteric modulators PNU-120596 and TQS, steady-state surface plasmon resonance responses to ligand binding to purified α7nAChR as a function of ligand concentration were measured at 25 °C using a Biacore 3000 with the NTA sensor chip (GE Healthcare, Uppsala, Sweden). α7nAChR was immobilized to the NTA sensor chip with densities between 1000 and 2000 RU. Responses to ligand binding were measured in phosphate-buffered saline with 0.005% lauryldimethylamine-N-oxide and 1% dimethyl sulfoxide at a flow rate of 30 μl/min. After reference and buffer subtraction, the steady state regions of each sensogram for PNU-120596 and TQS for a series of concentrations up to the solubility limits of the respective ligand were averaged and used to plot the SPR response (n = 3). Dissociation constants were derived by non-linear regression analysis using a Langmuir isotherm equation.

α-Bungarotoxin-Sepharose Pull-down Experiments

α-Bungarotoxin Sepharose was prepared by reacting 1 mg of bungarotoxin with 200 μl of cyanogen bromide-activated Sepharose (GE Healthcare), and then blocking unreacted sites following the manufacturer's recommendations. Control-Sepharose was prepared by blocking reactive sites in the absence of α-bungarotoxin. For each experiment, 3.2 nmol of purified α7nAChR was incubated overnight with 20 μl each of the control-resin or α-bungarotoxin-resin. The amount of α7nAChR pulled down was measured by the absorbance at 280 nm and reported as the mean ratio of α-bungarotoxin resin/control resin based on three independent experiments.

Electrophysiology

Functional measurements of purified α7nAChR injected into Xenopus oocytes were performed using two electrode voltage clamp experiments (56). We injected additional oocytes with either saline or cRNA encoding α7nAChR as negative and positive controls, respectively. For cRNA injections, 25 ng cRNA for α7nAChR was co-injected with 25 ng of cRNA for RIC3 to increase surface expression of α7nAChR. For protein injections, the purified pentameric α7nAChR was reconstituted into vesicles composed of egg phosphatidylcholine, phosphatidic acid, and cholesterol in a 3:1:1 molar ratio. Detergent was removed using Biobeads SM-2 (Bio-Rad) following the manufacturer's recommendations. 5 ng of reconstituted α7nAChR was injected into Xenopus oocytes. After 1–2 days, channel function was measured in a 20-μl oocyte recording chamber (Automate Scientific) clamped at −60 mV with an OC-725C Amplifier (Warner Instruments). The recording solutions contained 96 mm NaCl, 2 mm KCl, 1.8 mm CaCl₂, 1 mm MgCl₂, and 5 mm HEPES, pH 7.0 and the indicated concentrations of the agonists acetylcholine, choline, nicotine, the channel blockers QX-222, and PCP, and the α7nAChR-specific positive allosteric modulators PNU-120596, and TQS. Data were collected and processed using Clampex 10 software (Molecular Devices).

Author Contributions

T. S. T. conducted most of the experiments and analyzed the results. F. J. D. A., C. L., and P. Z. performed the electron microscopy experiments. N. J. R. performed electrophysiology measurements. D. W. and K. X. performed the mass spectroscopy analysis. Y. X. and P. T. designed the project. T. S. T. and P. T. wrote the manuscript. All authors reviewed the results and approved the final version of the manuscript.