Characterization of the ligand-binding site of the serotonin 5-HT3 receptor: the role of glutamate residues 97, 224, AND 235.

Ligand-gated ion channels of the Cys loop family are receptors for small amine-containing neurotransmitters. Charged amino acids are strongly conserved in the ligand-binding domain of these receptor proteins. To investigate the role of particular residues in ligand binding of the serotonin 5-HT3AS receptor (5-HT3R), glutamate amino acid residues at three different positions, Glu97, Glu224, and Glu235, in the extracellular N-terminal domain were substituted with aspartate and glutamine using site-directed mutagenesis. Wild type and mutant receptor proteins were expressed in HEK293 cells and analyzed by electrophysiology, radioligand binding, fluorescence measurements, and immunochemistry. A structural model of the ligand-binding domain of the 5-HT3R based on the acetylcholine binding protein revealed the position of the mutated amino acids. Our results demonstrate that mutations of Glu97, distant from the ligand-binding site, had little effect on the receptor, whereas mutations Glu224 and Glu235, close to the predicted binding site, are indeed important for ligand binding. Mutations E224Q, E224D, and E235Q decreased EC50 and Kd values 5-20-fold, whereas E235D was functionally expressed at a low level and had a more than 100-fold increased EC50 value. Comparison of the fluorescence properties of a fluorescein-labeled antagonist upon binding to wild type 5-HT3R and E235Q, allowed us to localize Glu235 within a distance of 1 nm around the ligand-binding site, as proposed by our model.

The question of how ligand binding to certain receptor proteins eventually gates ion channels is of central importance in cellular signaling but still unresolved at a molecular level. In order to enhance our understanding of the molecular mechanisms, we used a combined biomolecular and biophysical approach to study a serotonin-gated ion channel.
The 5-hydroxytryptamine (5-HT 1 ; serotonin) type 3 receptor (5-HT 3 receptor (5-HT 3 R)) is the only ligand-gated ion channel found among the serotonin receptors. Its gene structure (1) and amino acid sequence are similar to those of other members of the Cys loop receptor family including the nicotinic acetylcholine (nAChR), ionotropic ␥-aminobutyric acid, and glycine receptors. They are composed of five homologous or identical subunits, each comprising four predicted transmembrane regions and a large extracellular N-terminal domain containing the ligand-binding site. Among them, the nAChR is most closely related to the 5-HT 3 receptor. Both receptors form ion channels that are permeable to cations and share about 20 -30% amino acid sequence identity.
So far, three different 5-HT 3 receptor subunits, A, B, and C, have been cloned as well as a short splicing variant of the A subunit. Expression of only A subunits results in functional ion channels of similar properties as for 5-HT 3 receptors in native tissues, suggesting that this receptor is active as a homopentamer (2). Expression of solely the B (3) or C (4) subunits did not result in functional receptors; however, their coexpression with the A subunit yields in functional receptor proteins with slightly different channel properties.
Site-directed mutagenesis and biochemical studies, combined with amino acid sequence alignments, have identified amino acid residues and sequence regions (the so-called loops A-F; see Fig. 1) in the N-terminal extracellular domain implicated in the ligand-binding site of the nicotinic acetylcholine and 5-HT 3 receptors (for reviews, see Refs. 5-7 and references therein).
For the nAChR, these experiments together with the recently resolved three-dimensional structure of the acetylcholine-binding protein (AChBP) (8) indicate that the binding site is located at the interface between two adjacent subunits and that the binding loops A-F are forming the binding pocket. According to this model, loops A-C would form the binding site on one subunit, whereas loops D-F of the adjacent subunit would contribute to a lesser extent to ligand binding. In the case of the 5-HT 3 receptor, only few details are known about the ligand-binding site. Published data indicate that certain residues, especially in loops C and D (see Fig. 1A), are important for ligand binding.
In earlier work from our group, the binding of a fluoresceinlabeled, high affinity competitive antagonist 1,2,3,9-tetrahydro-3-[(5-methyl-1H-imidazol-4-yl)methyl]-9-(3-amino-(N-fluorescein-thiocarbamoyl)-propyl)-4H-carbazol-4-one (GR-flu) to the 5-HT 3 receptor has been characterized by steady state (9,10) and time-resolved (11) fluorescence spectroscopy. These studies indicated that receptor-bound GR-flu senses a local environment, which is ϳ0.8 pH units more acidic than the bulk, suggesting that acidic residues, glutamate or aspartate, are located close to the binding site of GR-flu. Recently, Glu 128 has been shown to participate in the binding of ligands to the 5-HT 3 receptor, since mutations of this residue strongly affected ligand binding and channel activation (12). Interestingly, this residue is conserved in all primary sequences of both subunits, A and B, of the 5-HT 3 receptor and is located in a primary sequence region corresponding to loop A (Fig. 1), shown to be important for ligand binding in the nicotinic acetylcholine and 5-HT 3 receptors (12)(13)(14)(15)(16). The negatively charged glutamates at the binding sites were proposed to interact with the positively charged nitrogen atoms in high affinity ligands for the 5-HT 3 receptor. However, using nonnatural amino acids, it was shown that Trp 182 of the 5-HT 3 receptor forms a cation-binding with the amino group of 5-HT (17). Moreover, other acidic residues might be implicated in ligand binding. In nAChR, ␦Asp 180 and ␦Glu 189 of mouse (18) and ␣Asp 200 of Torpedo (19) have been shown to play a role in ligand binding. These residues are conserved in all known 5-HT 3 receptor subunit A sequences, corresponding to Asp 203 , Glu 212 , and Glu 235 , respectively. However, the exact role of acidic residues in the binding of ligand to the 5-HT 3 receptor is not clear, because chemical modification of these residues resulted only in a slight inhibition of radioligand binding (20).
In this paper, we investigate the role of glutamate residues in ligand binding to the mouse subunit A of 5-HT 3 receptor, hereafter denominated as 5-HT 3 R, at three different positions, Glu 97 , Glu 224 , and Glu 235 , by replacing them with glutamine and aspartate using site-directed mutagenesis. Glu 97 , present in loop D, is conserved in both 5-HT 3 R subunits A and B but neither in the 5-HT 3 R subunit C nor in nAChR subunits. Glu 224 and Glu 235 , located in loop C, are present in all 5-HT 3 R subunits A but absent in the B and C subunits. An acidic residue at a position corresponding to Glu 235 is conserved in all nAChR subunits. Amino acid residues in both loops have been shown to influence ligand binding (21). Here, wild type (WT) and mutant 5-HT 3 R proteins were transiently expressed in HEK293 cells and were characterized functionally by whole cell patch clamp and radioligand binding experiments; the presence of receptor proteins in the plasma membrane was investigated by immunocytochemistry and receptor binding of the GR-flu. The question whether the Glu residues investigated here create the local acidic environment sensed by receptorbound GR-flu was studied by fluorescence spectroscopy. A structural model of the N-terminal domain of 5-HT 3 R based on the structure of AChBP was conceived in order to interpret our results.
We present experimental evidence that, in contrast to Glu 97 , Glu 224 and Glu 235 are deeply involved in ligand binding of the 5-HT 3 R, because a decrease in both ligand affinity and efficacy of 1-3 orders of magnitude was observed when these residues were mutated. Moreover, for the first time, direct biophysical measurements are presented, restricting the location of Glu 235 within 1 nm of the ligand-binding site of the 5-HT 3 R. The structural model corroborated these results.
Site-directed Mutagenesis of 5-HT 3 R cDNA-Full-length cDNA encoding the short splicing variant of the murine 5-HT 3 R subunit (1) was cloned as a 1.5-kb SmaI/NotI DNA fragment in the eukaryotic expression vector pCMV␤ (Clontech, Palo Alto, CA) by replacing the ␤-galactosidase reporter gene. Site-directed mutagenesis was performed using the QuikChange TM kit (Stratagene). Glutamates 97, 224, and 235 were mutated to glutamine and aspartate using mutagenic oligonucleotides that introduced as well a silent mutation to create a new restriction site to facilitate mutant screening (Table I of Supplementary Material). Mutations were confirmed by restriction pattern analysis and verified through sequence analysis.
Transient Transfection of HEK293 Cells-Human embryonic kidney cells (HEK293 cells) were cultured in Dulbecco's modified Eagle's medium supplemented with 2.2% fetal calf serum in a humidified 5% CO 2 atmosphere at 37°C. Adherently growing cells were subcultivated once or twice a week at ratios between 1:10 and 1:100.
To investigate the function of mutant receptors in cells by electrophysiology, immunolocalization, or binding the fluorescent ligand GRflu, HEK293 cells (60 -80% confluent), growing in either six-well plates or 35-mm cell culture dishes, were transfected using Effectene TM lipofection according to the protocol of the manufacturer (Qiagen, Hilden, Germany) with 0.2 or 1.0 g of plasmids containing the coding region of WT or the mutant 5-HT 3 R. After 4 h, the transfection solution was replaced by fresh medium. Experiments were performed 24 -55 h after transfection.
To investigate mutant receptors in either cell membranes or solubilized in detergent micelles (radioligand or GR-flu binding), HEK293 cells were seeded (1 ϫ 10 5 cells/ml) into 150-cm 2 flasks. 16 -20 h after splitting, cells were transfected with calcium phosphate-precipitated DNA as described (22). For each transfection, 75 g of WT or mutant 5-HT 3 R cDNAs diluted in 1.5 ml of CaCl 2 (250 mM) was added to an A shows the residues 1-245 of the mouse 5-HT 3 R comprising the signal sequence (residues 1-23, indicated by dots); the first transmembrane domain would start at residue 246. Underlined are the binding loops A-F; indicated by plus signs are residues for which experimental data suggest importance for ligand binding (12, 16, 21, 28 -30). The glutamate residues investigated by mutation in this paper are indicated by asterisks, and sequences corresponding to the oligonucleotides used for mutagenesis are shaded. In B, amino acid sequences of loops D and C of different nAChRs and 5-HT 3 Rs are aligned. The studied glutamates (asterisks) and the corresponding conserved anionic residues in the other receptor subunits are indicated (gray shading). equal volume of 140 mM NaCl, 1.4 mM Na 2 HPO 4 , 50 mM HEPES, pH 7.05, at 23°C. After a 1-min incubation at room temperature, the transfection solution was added to the cell culture medium. After a 4-h incubation at 37°C, the transfection solution was replaced by fresh medium. 48 h after transfection, the cells were washed twice with PBS, detached from the surface by "hitting the flasks," and harvested by centrifugation (3200 ϫ g, 5 min).
Electrophysiology-HEK293 cells were transiently transfected as described above but cotransfecting 0.2 g of DNA of cytosolic green fluorescent protein (Clontech) to identify cells expressing the 5-HT 3 R. Standard patch clamp measurements were done in whole cell configuration employing an EPC-9 patch clamp amplifier (HEKA Elektronik Dr. Schulze GmbH, Lambrecht, Germany). For data acquisition and storage, the software PULSE 8.3 (HEKA) was used. Heat-polished borosilicate glass pipettes (resistances of 2-5 megaohms) were filled with 140 mM CsCl, 10 mM EGTA, 10 mM HEPES, whereas the external buffer solution was 147 mM NaCl, 12 mM glucose, 10 mM HEPES, 2 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , both adjusted to pH 7.3 with NaOH. The ground electrode was connected to the bath via a 1 M KCl agar bridge. Ligands dissolved in the external buffer solution were applied with a RSC-200 perfusion system (Bio-Logic, Claix, France). During patch clamp experiments, the cells were continuously perfused. All experiments were performed at room temperature, and the membrane potential was kept at Ϫ60 mV. Recorded inward currents are displayed downward. At least six patches of each mutant protein from at least two different transfections were examined.
The Marquardt-Levenberg algorithm of Igor Pro (Wavemetrics Inc., Lake Oswego, OR) was used to evaluate the data by fitting to the equations, and where I represents the peak current at a particular ligand concentration, I max is the maximal peak current achievable, I 0 is the peak current in absence of any antagonist, EC 50 and IC 50 are the half-maximal effective and inhibitory concentrations, respectively, and n is the Hill coefficient.
For antagonist-agonist competition experiments, cells were perfused with granisetron-containing solutions for at least 2 min before adding 5-HT at a concentration inducing 90% of I max of the respective mutant in the absence of granisetron.
Binding Assay Using Radioligands-The affinity of 5-HT 3 R mutant proteins for radioligands and the total amount of ligand-binding sites were determined as described (23) with the modification that samples containing ϳ0.2 pmol of 5-HT 3 R were incubated for 60 min at room temperature in 10 mM HEPES, pH 7.4, with increasing concentrations of [ 3 H]GR65630 (up to 20 nM) or [ 3 H]mCPBG (up to 120 nM) in a final volume of 0.2 ml. Nonspecific binding was determined in the presence of 1 M quipazine (pK ϭ 9.0 Ϯ 0.3 (9)). All experiments were performed in triplicate.
The affinity of the receptor for granisetron was determined by competition binding assays. Samples comprised ϳ0.8 nM [ 3 H]GR65630 or 20 nM [ 3 H]mCPBG in 10 mM HEPES, pH 7.4, and 5-HT 3 R-containing membranes to produce 100 -2000 cpm of specifically bound radioactive ligand; they were incubated at increasing concentrations of the competitor in a final volume of 0.2 ml for 60 min at room temperature.
The dissociation constant K d of the radioligand, the total concentration of 5-HT 3 R (expressed as concentration of binding sites), and Hill coefficients (n) were evaluated by fitting experimental data as follows.
[L] bound ϭ [5- Binding inhibition curves were fitted to the following equation.
[B] and [B o ] are the concentrations of bound radioligand in the presence and absence of unlabeled competitor, respectively, and IC 50 is the concentration of competing ligand that displaced 50% of the specifically bound radioligand. The dissociation constant of inhibition K i of competitors was estimated from the Cheng-Prusoff equation (23), where [L] and K d are the concentration and dissociation constant of radioligand, respectively.
Binding of GR-flu to Living Cells-HEK293 cells grown on glass coverslips in six-well plates were transfected as described above. 40 h after transfection, the coverslips were transferred into a sample holder and covered with 400 l of phosphate-buffered saline. Then, during image recording using an LSM 510 confocal microscope (Zeiss, Oberkochen, Germany) equipped with appropriate excitation and emission filters for fluorescein and using identical settings for all samples when not stated otherwise, GR-flu was added to reach either 1.5 or 15 nM final concentration. Nonspecific binding of GR-flu to cells, measured in the presence of 1 M quipazine, was negligible.
Binding of GR-flu to Detergent-solubilized 5-HT 3 R-HEK293 cells harvested from one T150 flask were resuspended in 1 ml of 10 mM HEPES, 0.5 mM EDTA, pH 7.4, pelleted by centrifugation (10 min at 3200 ϫ g at 4°C), resuspended in 1 ml of total volume of the same buffer, and finally homogenized for 10 s with an Ultra-Turrax T25 (IKA, Staufen, Germany). Membranes obtained by this procedure were pel- leted by centrifugation for 10 min at 3200 ϫ g and 4°C and resuspended in 1 ml of 10 mM HEPES, 1 mM C 12 E 9 , pH 7.4, and homogenized for 10 s with an Ultra-Turrax T25. Nonsolubilized matter was removed by centrifugation for 30 min at 50,000 ϫ g and 4°C.
Detergent-solubilized receptor was incubated for 30 min in a solution containing 1 nM GR-flu, 10 mM HEPES, and 1 mM C 12 E 9 , pH 7.4, in the dark at room temperature. Total fluorescence intensity and fluorescence anisotropy were measured using an Analyst multiwell fluorescence reader (Molecular Devices, Inc., Sunnyvale, CA). Fluorescence signals were corrected for background fluorescence and nonspecific binding of GR-flu in the presence of 1 M quipazine.
Localization of 5-HT 3 Rs in Cells by Immunofluorescence-Microscopy glass cover slides with attached cells transiently expressing either WT or mutant 5-HT 3 Rs were washed three times with PBS and incubated in ice-cold 4% paraformaldehyde in PBS for fixation of cells. To immunolabel the N-terminal extracellular domain, pAb120 antiserum, containing antibodies directed against the first 13 residues of the mature 5-HT 3 R (24), was used at 1:1000 dilution in Tris-buffered saline. Intracellular receptor expression was determined by incubating cells in the presence of 0.3% Triton X-100 for membrane permeabilization. Primary antibodies were incubated for 1 h at room temperature. Biotinylated anti-rabbit IgG (Vector, Burlingame, CA) and fluorescein isothiocyanate-labeled avidin D (Vector) were used to detect bound antibodies following the manufacturer's instructions. Coverslides were mounted in Vectashield mounting medium (Vector); cells labeled with fluorescent antibodies were observed using an LSM 510 confocal microscope.
Structural Model of Extracellular N-terminal Domain of 5-HT 3 R-The N-terminal domain of the murine 5-HT 3 R (SwissProt entry P23979, residues 14 -217 of the mature protein; i.e. until 6 residues before the start of the first transmembrane domain) was aligned to the sequence of the AChBP (residues 1-198 of the recently solved crystal structure (Protein Data Bank entry 1I9B.pdb). Optimal alignment resulted in three gaps of 2, 1, and 3 residues, respectively, between residues Asp 63 -Gln 64 , Ala 135 -Thr 136 , and Asp 153 -Pro 154 in the sequence of the AChBP. The alignment was submitted to SWISS-MODEL for the generation of a homology model of the N-terminal domain of the 5-HT 3 R based on the template of the structure of AChBP using Swiss-PbdViewer 3.7 as interface (25). The structural model thus obtained was inspected using either Swiss-PbdViewer or Rasmol version 2.7.

RESULTS
Electrophysiology-The functional properties of WT and mutant receptor proteins were investigated by whole cell patch clamp measurements. Typical electrical responses upon application of 5-HT with peak currents around 1-4 nA at saturating 5-HT concentrations (Fig. 2) indicated that WT and mutant receptor proteins were functionally expressed at similar levels.
As an exception, in the case of E235D only small currents were observed at millimolar concentrations of 5-HT. From doseresponse curves (Fig. 3), the EC 50 (Table I) and the Hill coefficient n for 5-HT were evaluated. Whereas both mutant proteins of Glu 97 showed no difference compared with WT, the EC 50 values for E224Q, E224D, and E235Q were increased by a factor of 4, 10, and 20, respectively. In the case of E235D, no EC 50 could be determined, since only small currents of some 100 pA could be recorded repetitively at 20 mM 5-HT, close to the solubility limit. Neither with 5-HT nor with the agonist mCPBG, which has a higher affinity to the WT receptor than 5-HT, could saturating ligand binding conditions be reached. Control experiments verified that nontransfected HEK293 cells did not respond electrically upon application of 5-HT or mCPBG concentrations up to 20 mM. This suggests that the EC 50 of E235D is shifted more than 1000-fold to higher concentrations. The Hill coefficient n was similar for all mutant (except E235D) and WT receptors and ranged from 1.5 to 1.9. In contrast, the kinetics of channel activation and desensitization of E235Q were considerably slower than for Glu 97 and WT (Fig. 2).
Antagonist-agonist competition experiments yielded similar results ( Fig. 4 and Table I). After incubation for 2-5 min with increasing concentrations of the antagonist granisetron, the response to 5-HT was recorded. The 5-HT concentration applied evoked 90% of I max of the respective mutant in absence of granisetron. For mutant proteins of Glu 97 , granisetron inhibited 5-HT-induced currents equally well as for the WT; for E224Q, E224D, and E235Q, the IC 50 of granisetron was shifted to higher concentrations by a factor of 3, 9, and Ͼ100, respectively. The Hill coefficient varied between 1.0 and 1.6. For E235D, this experiment could not have been carried out in a

FIG. 4. Granisetron competition for 5-HT binding measured with whole cell patch clamp.
After incubation with increasing concentrations of granisetron, the maximal electrical response upon the addition of 5-HT (concentration corresponding to 90% I max of the respective mutant) is recorded for 5-HT 3 R WT (ϫ) and mutants E97Q (q), E97D (E), E224Q (f), E224D (Ⅺ), E235Q (OE). E235D could not be measured, since no comparable agonist concentration could be reached. Currents are normalized to the electrical response evoked by the respective 5-HT concentration before the incubation with the antagonist. Representative curves are shown. comparable way, because no complete 5-HT dose-response curve and thus no relative value to I max could be determined.
Binding Experiments Using Radioactive Ligands-To investigate whether the effects due to the different mutations observed by electrophysiology are caused by changes in ligand affinity or channel gating, the ligand-binding properties of the WT and mutant 5-HT 3 R were investigated using the radiolabeled agonist mCPBG and antagonist GR65630. Both the dissociation constants K d of the binding of these ligands to the receptor proteins and the inhibition of radioligand binding by granisetron or mCPBG were investigated. The results are sum-marized in Table I. The expression level of the different mutant proteins as compared with WT inferred from the saturation binding of [ 3 H]GR65630 ranged from 0.3-fold (E97Q and E223Q) to 1.5-fold (E97D), except for E235D, for which less than 0.03-fold was observed which is in the range of the experimental detection limit. The agonist [ 3 H]mCPBG bound with comparable affinity to the mutant receptor protein E97D and E224Q as to WT, whereas for the mutants E97Q and E224D, 2.5-3-fold lower affinities were observed. For E235Q, the dissociation constant was increased more than 5 times, whereas for E235D no significant binding could be observed. The Hill coefficient was close to unity in all cases. Inhibition by the antagonist granisetron of the binding of agonist [ 3 H]mCPBG to the various 5-HT 3 R mutant proteins showed a similar trend.
The affinity of the antagonist [ 3 H]GR65630 for both E97 mutant proteins and E224Q was slightly less (2-3-fold) as compared with WT, whereas for E224D and E235Q a 10-fold or more reduced affinity was observed. As for the agonist In general, large effects on radioligand binding were observed for those mutants, featuring decreased potencies of ligand in electrophysiology.
Localization of 5-HT 3 Rs in Cells-To investigate whether the absence of significant whole cell currents upon 5-HT addition and of radioligand binding to E235D is either due to the mutation in an otherwise intact protein or caused by defect biogenesis of the protein, we applied both binding of GR-flu and immunolocalization, using a monoclonal antibody recognizing the extracellular domain of the receptor (24).
A distinct fluorescence signal due to specific binding of GRflu was observed for all receptor proteins except E235D (Fig. 5,  left column). Staining with the receptor-specific antibody (Fig.  5, middle column) demonstrated the presence of WT and all 5-HT 3 R mutants on the surface of the intact HEK293 cells but was less evident for E97Q and E235D. In permeabilized cells, intracellular localized receptor was detected, especially in the case of E235D. These data suggest that mutation E235D impairs folding, assembly, and/or trafficking of the 5-HT 3 R to the cell surface.
Fluorescence Measurements-Binding of the competitive antagonist GR-flu to WT 5-HT 3 R results in a ϳ70% decrease of its fluorescence intensity, due to a local acidic pH of the binding site; concomitantly, the fluorescence anisotropy increases (9,11). To assess whether one of the Glu residues under investigation is responsible for this fluorescence intensity decrease, fluorescent ligand binding studies were performed on detergent-solubilized membranes of cells expressing the different receptor proteins. Binding of GR-flu to the receptor can be shown unequivocally by an increase of the fluorescence anisotropy, whereas the local environment sensed by the receptorbound GR-flu is indicated by its fluorescence intensity. Fig. 6 summarizes the results obtained for both the fluorescence intensity and anisotropy of receptor-bound GR-flu. The Glu 97 and Glu 224 mutant proteins yielded high fluorescence anisotropies and low fluorescence intensities, comparable with WT, albeit GR-flu bound to E224Q and E224D was about 10 and 15% more fluorescent than GR-flu bound to WT receptor, respectively. GR-flu did not bind to E235D, since the fluorescence anisotropy of GR-flu in the presence of this receptor mutant, with or without an excess of the competing nonfluorescent ligand quipazine, was undistinguishable from that of free GR-flu; also, the fluorescence intensity did not change. Upon binding of GR-flu to E235Q, the fluorescence anisotropy increased from 0.05 up to ϳ0.12. However, the fluorescence intensity remained unchanged. These data demonstrate that the fluorophore of GR-flu when bound to E235Q senses a virtually identical pH as when free in the bulk, suggesting that Glu 235 and the fluorophore of the receptor-bound GR-flu are in close proximity.

DISCUSSION
Negatively charged amino acids have been implicated in ligand binding to ligand-gated ion channels of the Cys loop family. Here, we investigated the role of 3 glutamate residues located in the binding loops C and D, which are conserved in the subunit A of the 5-HT 3 R of different species (Fig. 1). Glu 97 , Glu 224 , and Glu 235 were replaced by glutamine and aspartate. Mutant and WT receptors were transiently expressed in HEK293 cells and characterized using electrophysiology, radioligand binding, immunolocalization, and fluorescent ligand binding. In the following, our results are considered in context with literature data and evaluated by a structural model of the N-terminal domain of the 5-HT 3 R.
Residue Glu 97 -Both the induction of channel currents and inhibition thereof by the antagonist granisetron for both E97Q and E97D are identical to WT. Moreover, the fluorescence properties of receptor-bound GR-flu are identical to those observed in the case of WT. The data on E97D from radioligand binding experiments are virtually identical to WT. However, for E97Q we observed an approximately 2-fold reduced affinity for mCBPG and GR65630 and for granistron in competition with mCPBG; conversely, a 2-fold increased affinity of both granisetron and mCPBG in competition versus the antagonist GR65630 was seen. These data suggest that a negatively charged residue at position 97 has some importance, albeit subtle, for ligand binding to the desensitized state, which is probed by the radioligand binding assays.
Residue Glu 224 -The mutant receptor E224D yielded a 4 -10-fold decrease of both ligand affinity and efficacy, indicating that this mutation affects ligand binding rather than channel gating. In contrast, mutation of Glu 224 to Gln did reduce the efficacy of 5-HT 3-fold without changing the affinity of the agonist mCPBG and its competitor granisetron, suggesting that this mutation might influence channel gating. Curiously, the affinities of GR65630 and its competitor granisetron were 3-4-fold reduced, suggesting that the antagonists bind more strongly to the resting or active state as compared with the desensitized state of E224D. Both mutations of Glu 224 have small but significant effects on the fluorescence intensity of receptor-bound GR-flu, suggesting that Glu 224 is located not too distant from the ligand binding site, either allowing a direct interaction with the chromophore or provoking a protein structural change that affects the fluorescent properties of the chromophore.
Taken together, these data indicate that Glu 224 is involved in both ligand binding and channel gating. In the case of the nAChR, a corresponding Asp in the mouse muscle ␦-subunit (␦Asp 202 ) was mutated to Asn without any effect on agonist affinity or efficacy (18). An explanation of these different results might be that Glu 224 in the 5HT 3 R is located in loop C, whereas the mouse muscle ␦-subunit contributes with loops D-F to the ligand binding site (i.e. the sequence corresponding to loop C in the ␦-subunit is not involved in ligand binding). To our knowledge, no equivalent mutation has been reported for the ␣-subunit of the nAChR.
Residue Glu 235 -Mutation of Glu 235 to Asp had a dramatic effect on both the functional properties of the receptor and its biogenesis. Exceedingly high ligand concentrations were needed to observe a low but significant channel conductivity, indicating that the efficacy of 5-HT was reduced by at least 3 orders of magnitude. The presence of functional receptor protein in the plasma membrane of cells could not be detected by radioligand and fluorescent ligand binding experiments, due to the limited ligand concentrations applicable. Still, immunolocalization showed that small amounts of receptor are present in the cell membrane, whereas the majority of the receptor proteins are located intracellularly.
For E235Q, both the efficacy and affinity of ligands are strongly decreased. Moreover, kinetics of channel opening and desensitization were much slower for E235Q than for WT. This indicates that Glu 235 is involved in ligand binding and channel gating.
Taken together, these observations imply that Glu 235 plays an important role in receptor biogenesis, ligand binding, and channel gating. This residue aligns with ␦Asp 214 of mouse muscle (18) and ␣Asp 200 of Torpedo (19) nAChR. Mutation of these Asp residues to Asn decreased the efficacy of acetylcholine 3-10-fold and caused phenyltrimethylammonium and tetramethylammonium, partial agonists for WT nAChR, to act as antagonists (19). Moreover, also the maximal current upon agonist stimulation was reported to decrease 6-fold (18). These findings indicate that Asp residues in nAChR are involved in receptor biogenesis, ligand binding, and channel gating, like their counterpart Glu 235 in the 5-HT 3 R, whereas the effects observed in this study were more dramatic. This might be caused by the homopentameric nature of the 5-HT 3 R (i.e. each mutant receptor protein carries five mutations, whereas in the case of the nAChR it carries only one or two).
Local Environment and Mobility of Receptor-bound GR-flu-GR-flu, when bound to WT 5-HT 3 R, exhibits a fluorescence anisotropy, r ϭ 0.156; however, when bound to E235Q, a value of only 0.116 is obtained. This apparent discrepancy can be explained by considering the following: (i) the mean fluorescence lifetime of GR-flu free in solution is ϭ 3.5 ns, and bound to WT receptor ϭ 1.56 ns (11); (ii) the limiting anisotropy of GR-flu is r o ϭ 0.21 (9); (iii) the fluorescence lifetime of GR-flu bound to E235Q can be assumed to be similar to that of GR-flu free in solution, since the fluorescence intensities are equal in both cases; (iv) the rotational correlation time of GR-flu bound to WT and E235Q can be assumed to be identical. Using the relation between the fluorescence anisotropy and the rotational correlation time of a fluorophore under restricted motional conditions, r ϭ r o /(1 ϩ /) (26), one notices that a change of r from 0.156 to 0.116 is accompanied by a change of from 1.56 to 3.6 ns.
Distance between GR-flu and Glu 235 -Binding of GR-flu to WT 5-HT 3 R results in a decrease of the fluorescence intensity. This was shown to be due to a local acidic environment of the binding side of the receptor, resulting in an apparent increase of the pK a of the fluorescein moiety of GR-flu of the ⌬pK a ϭ 0.79 Ϯ 0.06 (9). The observed ⌬pK a can be estimated from the following, ⌬pK a ϭ (0.434 q 1 q 2 /4 0 kTd) exp (Ϫd/) (Eq. 6) where q 1 and q 2 represent the respective point charge of the fluorescein and the carboxyl group separated by a distance d, ⑀ 0 is the dielectric constant in vacuum, ⑀ is the dielectric constant of the medium, and k is the Boltzmann constant. The shielding or Debye length 2 ϭ ⑀ 0 ⑀RT/2I(eN A ) 2 , where R represents the gas constant, T the temperature, the density of the solvent, I the ionic strength of the solution, e the elementary charge, and N A the Avogadro number. Since both the fluorescein moiety of receptor-bound GR-flu and the side chain of Glu 235 are located on the outer surface of the receptor protein, one could assume that ⑀ ϭ 78.5, the value for water. However, several studies indicate that on the surface of a protein ⑀ instead ranges from 40 to 10 (27). Taking the observed value ⌬pK a ϭ 0.79 Ϯ 0.06 (9), the distance d between protonatable groups of GR-flu and Glu 235 is estimated to be 0.34 Ϯ 0.03, 0.59 Ϯ 0.04, and 1.11 Ϯ 0.05 nm for ⑀ ϭ 78.5, 40, or 10, respectively. Structural Model of the N-terminal Domain of 5-HT 3 R-A structural model of the N-terminal domain of 5-HT 3 R (Fig. 7) based on the recently resolved structure of the nAChBP predicts the glutamate residues Glu 224 and Glu 235 to be close (within 1-1.5 nm) to the binding pocket, whereas Glu 97 is more remote (about 2.5 nm). It was suggested earlier that the carboxylate side chains of a glutamate might interact electrostatically with the amino group of the ligand and thus help to optimally position the interacting groups in the binding pocket (12,18). Moreover, steric effects due to changes in the length of the amino acid side chains can modify the shape of the binding pocket and alter frequency of the binding events. With respect to our results, the structural model suggests that changes in charge or size of the side chains have more or less strong influence on ligand binding, depending on whether a residue is distant (Glu 97 ) or close (Glu 224 and Glu 235 ) to the binding site. Therefore, it seems likely that Glu 224 and Glu 235 can exert a direct interaction with ligands, whereas Glu 97 is perhaps more involved through small structural changes, affecting subtly the functional properties of the receptor. The distance between GR-flu and Glu 235 experimentally determined above is corroborated by the model, which shows that Glu 235 is located at a distance of 1.3 nm around the center of the ligand binding site. Since Glu 224 and Glu 235 are juxtaposed on neighboring antiparallel ␤-strands forming loop C, and Glu 224 does not seem to be responsible for the ⌬pK a of receptor-bound GR-flu, it is tempting to speculate that, when the pharmacophore of GR-flu is bound in the binding site, the fluorophore protrudes from the binding pocket passing above loop C, distal of the membrane, so that it is located closer to Glu 235 than to Glu 224 .
Conclusions-Our experimental results indicate that Glu 235 is located so close to the ligand-binding site that direct electrostatic interaction with bound ligands occurs. Moreover, this residue is important for receptor biogenesis and is implicated in channel gating. Glu 224 is also involved in ligand binding and channel gating, but mutations of this residue have less pronounced effects as compared with the equivalent mutations on Glu 235 . In general, it turns out that effects on the functional properties of the 5-HT 3 R due to a change in the length of the side chain are much more pronounced than the presence of a negative charge at a particular side chain. This could imply that Glu 224 and Glu 235 are involved in formation of hydrogen bonds implicated in ligand binding or channel gating. Finally, the results show that the structural model is reliable and forms an important base to direct further biochemical and biophysical research.