Abstract
The widely used sodium channel blocker tetrodotoxin (TTX) is a compound that has six hydroxyl residues at the C-4, C-6, C-8, C-9, C-10, and C-11 positions in addition to a guanidinium group, which is positively charged in biological pH range. Thirteen analogs of this toxin with structural modifications involving one or more of these hydroxyls were examined on their affinity to a rat brain membrane preparation, which is known to contain sodium channels abundantly. The equilibrium dissociation constants associated with the binding of TTX and its analogs to the sodium channels were estimated, from their ability to inhibit the binding of [3H]saxitoxin, as follows (in nM): TTX, 1.8; chiriquitoxin, 1.0; 11-oxoTTX, 1.5; 11-norTTX-6,6-diol, 1.6; 11-norTTX-6(S)-ol, 23; 11-norTTX-6(R)-ol, 31; 11-deoxyTTX, 37; 6-epiTTX, 39; 4-epiTTX, 68; 4,9-anhydroTTX, 180; TTX-8-O-hemisuccinate, >380; TTX-11-carboxylic acid, >2300; tetrodonic acid, >3600; 5,6,11-trideoxyTTX, >5000. The reduction of the affinity observed with the analogs involving reduction or translocation of the hydroxyls at C-6 and C-11 is indicative of the contribution of these residues to the binding to sodium channels as hydrogen bond donors. The especially large value of the dissociation constant for TTX-11-carboxylic acid is consistent with the idea that the C-11-hydroxyl forms a hydrogen bond with a carboxylic acid residue of the channel protein. The markedly low affinity of TTX-8-O-hemisuccinate may possibly be ascribable to intramolecular salt-bridge formation, which neutralizes the positive charge of the guanidinium group.
Tetrodotoxin (TTX) and saxitoxin (STX) are highly poisonous natural compounds that are widely used as very useful tools in the study of excitable cells because of their potent and specific blocking action on the voltage-gated sodium channels (Narahashi et al., 1967; Kao, 1986;Hille, 1992). They are believed to bind to the common site, which is present at the external mouth of the channels (Hille, 1992). The structures of TTX (see Mosher, 1986) and STX (Schantz et al., 1975) have long been established. Both toxins are rigid heterocyclic molecules with several hydroxyls on their surfaces in addition to one or two guanidinium groups, which constitute intrinsic parts of the molecules and are positively charged at biological pH (Fig.1).
Previously, we isolated several natural analogs of TTX including 4-epiTTX (Nakamura and Yasumoto, 1985), 6-epiTTX (Yasumoto et al., 1988), 11-oxoTTX (Khora and Yasumoto, 1989), and 11-deoxyTTX (Yasumoto et al., 1988), in which the structure involving one of the six hydroxyls of TTX is specifically changed. Kao and coworkers examined the effects of these and some other TTX analogs on sodium channels by electrophysiological experiments and reached the conclusion that although the hydroxyls at C-9 and C-10 are the most important, those at C-4, C-6, and C-11 also make significant contributions to the binding to the channel (Kao, 1982, 1986; Kao and Yasumoto, 1985; Yang and Kao, 1992; Yang et al., 1992; Wu et al., 1996). In conjunction with studies on the effects of chemical and genetic modifications of the channel, such experiments with toxin derivatives have led to the current view that the high affinity of TTX is mostly attributable to formation of ion pairs and hydrogen bonds between the functional groups of the toxin and the carboxyl residues that are circularly placed around the extracellular mouth of the channel pore (see Catterall, 1995).
However, it should be noted that the view is still only hypothetical. For example, no direct experimental evidence indicating the presence of carboxyl residues in the vicinity of the hydroxyls of TTX has been available to date. Although the guanidinium group of TTX is generally assumed to be essential for the channel blockade, there is no easy way to test the assumption as direct modifications involving this group would also introduce other complex changes in the rest of the molecule. As pointed out by Kao (1986), the current opinion about the role of this moiety is largely dependent on comparison with the 7,8,9-guanidinium group of STX. Also, very little is known about the contribution of the hydroxyl at C-8 because no analog with specific modification at this position has hitherto been available.
During further studies in pursuit of new TTX derivatives, we have recently obtained two novel analogs, TTX-11-carboxylic acid and TTX-8-O-hemisuccinate, by chemical treatment of TTX (seeMethods and Materials). In the latter compound, which is, to our knowledge, the first TTX derivative involving specific modification of the C-8 hydroxyl, the positive charge of the guanidinium group is neutralized by formation of an intramolecular salt-bridge. We have also succeeded in separating two C-6-isomers of 11-norTTX alcohols, 11-norTTX-6(S)-ol (Yotsu et al., 1992) and 11-norTTX-6(R)-ol (Endo et al., 1988), from a mixture of derivatization products of TTX. These new materials gave us opportunities to evaluate further the role of the guanidinium group as well as that of the hydroxyls, especially those in the C-6 end of the molecule.
In this article, we have determined the equilibrium dissociation constants (K0) for the interaction of the novel TTX analogs with rat brain synaptic membrane, which is known to contain sodium channels abundantly (Hartshorne and Catterall, 1984). For this purpose, we measured reduction of [3H]STX binding by addition of the TTX analogs. The validity of the present method depends upon the extent to which the interaction of labeled ligand and unlabeled ligand is exclusive. We applied an analytical procedure that allows us to evaluate quantitatively the degree of exclusiveness as well asK0. There are various factors that make it difficult to directly compareK0 estimated by binding assays with IC50 obtained by electrophysiological experiments (Ritchie and Rogart, 1977). We therefore also carried out the binding assay for some other TTX derivatives whose electrophysiological effects were reported previously (see Table 2 below).
Here, we present the results that suggest the presence of a negative charge near the binding site for the hydroxyl at C-11. We also show that the modification of the C-8 hydroxyl results in a marked reduction of the affinity. Some implications for the structure of the toxin binding site will be discussed.
Materials and Methods
[3H]STX and Rat Brain
[3H]STX was a product (catalogue code TRK877) of Amersham Intl. (Buckinghamshire, UK). The specific radioactivity of [3H]STX was 500 GBq/mmol. Rat brain (catalogue code AT1204) was purchased from Rockland Inc. (Gilbertsville, PA). All other chemical reagents were of analytical grade.
Spectroscopy and Chromatography
1H and 13C NMR spectra were measured using a JEOL GSX-400 spectrometer at 400 MHz and 100 MHz, respectively [in 4% CD3COOD/D2O (v/v)]. The signal of CHD2COOD at 2.06 ppm in1H NMR and the signal of13CD3COOD at 22.4 ppm in13C NMR were used as the references. Fast atom bombardment-mass spectroscopy (FAB-MS; positive mode) in a matrix of glycerol was carried out on a JEOL JMS303HF spectrometer.
Synthetic reactions were monitored by a fluorometric HPLC analyzing system for TTX analogs (Yasumoto and Michishita, 1985; Yotsu et al., 1989). TTX analogs were separated by chromatography on a reversed-phase column. For fluorometric detection, they were derived to fluorophores by heating with alkaline solution (Yotsu et al., 1989). Three different combinations of columns and mobile phases (conditions A, B, and C) were used for HPLC: 1) a Cosmosil 5C18AR column (4.6 × 150 mm; Nacalai Tesque, Kyoto, Japan) with 20 mM ammonium acetate and 7 mM sodium heptanesulfonate buffer (pH 6.6) containing 3% CH3CN (0.7 ml/min); 2) a Develosil ODS-5 column (4.6 × 250 mm; Nomura Chemical Co. Ltd., Seto, Japan) with 50 mM ammonium acetate and 60 mM ammonium heptafluorobutyrate buffer (pH 5.5) containing 3% CH3CN (0.7 ml/min); and 3) a Cosmosil 5C18AR column with 50 mM ammonium acetate and 30 mM ammonium heptafluorobutyrate buffer (pH 5.0) containing 3% CH3CN (0.5 ml/min).
Naturally Occurring TTX Analogs
Chemical structures of TTX analogs tested in the present study are shown in Fig. 1. TTX, 4-epiTTX, and 4,9-anhydroTTX were isolated from pooled eggs of the puffer fishes Fugu poecilonotus and Fugu pardalis (Nakamura and Yasumoto 1985). TTX from these sources was used for the synthesis of 11-oxoTTX, TTX-11-carboxylic acid, 11-norTTX-6,6-diol, 11-norTTX-6(S)-ol, 11-norTTX-6(R)-ol, TTX-8-O-hemisuccinate, and tetrodonic acid (see below). 5,6,11-TrideoxyTTX was isolated from the eggs of F. poecilonotus (Yotsu-Yamashita et al., 1995). 6-EpiTTX and 11-deoxyTTX were isolated from the newt Cynops ensicauda(Yasumoto et al., 1988). Chiriquitoxin was isolated from the frogAtelopus chiriquiensis ( Yotsu et al., 1990 ).
Chemical Modifications of TTX
The scheme for chemical modifications of TTX is shown in Fig.2.
11-oxoTTX, 11-norTTX-6,6-diol, 11-norTTX-6(S)-ol, 11-norTTX-6(R)-ol, and Tetrodonic Acid.
11-OxoTTX (Fig. 2A) was prepared by Pfitzner-Moffatt oxidation of TTX (8.3 μmol) as previously reported (Chicheportiche et al., 1979, Wu et al., 1996). After the reaction, 11-oxoTTX (yield, 5%) was separated from remaining TTX, 4,9-anhydro-11-oxoTTX (major by-product), and 4,9-anhydroTTX by HPLC (condition A) and desalted by HPLC on a GEL 3011-C weak cation exchange column (5 × 500 mm; Hitachi Co., Tokyo, Japan). 11-NorTTX-6,6-diol (yield, 73% from 8.3 μmol of TTX), 11-norTTX-6(S)-ol (yield, 19% from 6.1 μmol of 11-norTTX-6,6-diol), and 11-norTTX-6(R)-ol (yield, 6% from 6.1 μmol of 11-norTTX-6,6-diol) were prepared by oxidation of TTX with H5IO6 and subsequent reduction with NaBH3CN (Fig. 2B) as reported by Pavelka et al. (1982). Separation of 11-norTTX-6(S)-ol (RF, 0.75) and 11-norTTX-6(R)-ol (RF, 0.58) was achieved by TLC with Silica-gel 60 (Merck Co., Darmstadt, Germany) using pyridine/ethyl acetate/acetic acid/H2O, 15:7:3:6 (volume ratio) as the mobile phase. Tetrodonic acid (Fig. 2C) was prepared by the method of Goto et al. (1965) and purified by HPLC on a G1000PW TSK gel filtration column (10 × 500 mm; Toso Co., Tokyo, Japan) with 0.05 M acetic acid.
TTX-11-Carboxylic Acid.
TTX-11-carboxylic acid was prepared by oxidation of 11-oxoTTX (Fig. 2A). The partially purified mixture of 11-oxoTTX and TTX (molecular ratio, 1:1) from the reaction mixture of Pfitzner-Moffatt oxidation of TTX (8.3 μmol) (Wu et al., 1996) was lyophilized, dissolved in 0.05 M acetic acid (1.5 ml), and then mixed with a solution of NaClO2 (80%, 11 μmol) in 0.05 M acetic acid (400 μl) and 2-methyl-2-butene (211 μl, 2 mmol) to start oxidation. After stirring for 30 min at 20°C, the peak corresponding to 11-oxoTTX at 11.2 min disappeared, and a peak corresponding to TTX-11-carboxylic acid appeared at 7.2 min on the fluorometric HPLC chromatogram (condition B). The solvent was removed by evaporation, and the residue was dissolved in H2O (1 ml) and then applied to a SEP-PAK CM weak cation exchange cartridge column (Waters Co., Milford, MA) equilibrated with H2O. TTX was completely trapped by the column, whereas TTX-11-carboxylic acid appeared in the flow-through, which was then applied to an HPLC on a G1000PW TSK gel filtration column (see above) using 0.05 M acetic acid at a flow rate of 1.0 ml/min. TTX-11-carboxylic acid (yield, 7% from TTX) was eluted together with 4,9-anhydroTTX-11-carboxylic acid in the molecular ratio of 10:1 (estimated by 1H NMR) in 20 to 22 ml. FAB-MS of the sample obtained revealed a peak corresponding to a protonated ion at m/z 334. The 1H NMR signals were assigned by1H-1H correlation spectroscopy. δ (ppm): 2.31[1H, d, J 10.3 Hz, H-4a], 4.04[1H, s, H-9], 4.28[1H, br s, H-8], 4.51[1H, br s, H-7], 4.68[1H, br s, H-5], 5.53[1H, d, J 10.3 Hz, H-4]. The oxidation of the aldehyde group at C-6 of 11-oxoTTX to the carboxylic acid was supported by the downfield shifts of the H-5 (0.31 ppm) and H-7 (0.30 ppm) signals compared with those of 11-oxoTTX.
TTX-8-O-Hemisuccinate.
To synthesize TTX-8-O-hemisuccinate (Fig. 2D), TTX was first converted to a mixture containing 4,9-anhydro-8-O-hemisuccinate as the major component (40%), essentially by the method of Strong and Keana (1976). Briefly, a suspension of TTX (6.26 μmol) in H2O (5 ml) was magnetically stirred, and excess succinic anhydride (1.8 mmol) was added portionwise at room temperature over 10 min. During the addition of succinic anhydride, the pH was kept in a range from 5 to 6 with saturated Ba(OH)2 in H2O. One hour later, the reaction was stopped by addition of H2O (3 ml), and the precipitate of barium succinate was removed by centrifugation. The supernatant was then applied to an activated charcoal column (10 × 50 mm) equilibrated with H2O. After washing the column with H2O (12 ml), the reaction products were eluted with 20 ml of acetic acid/ethanol/H2O (5:50:45, v/v/v), and the solvent was removed by evaporation. FAB-MS and 1H NMR spectroscopy identified the following components in the sample: 4,9 anhydroTTX-8-O-hemisuccinate (protonated ion, m/z 402; yield, 40%) and 4,9-anhydroTTX-8-O,11-O-dihemisuccinate (m/z 502; yield, 4%) (see Fig. 2D). This mixture of reaction products was then dissolved in 5% (v/v) trifluoroacetic acid/H2O (100 ml) and allowed to stand at 37°C for 22 h. After this incubation time, the fluorometric HPLC (condition C) revealed a new peak corresponding to TTX-8-O-hemisuccinate at 5.2 min. Because of partial hydrolysis of the ester bonds, the peaks corresponding to TTX, 4,9-anhydroTTX also appeared. The molecular ratio of TTX, 4,9-anhydroTTX, TTX-8-O-hemisuccinate, and 4,9-anhydroTTX-8-O-hemisuccinate in the reaction mixture was 4:1:10:10 based on HPLC. After removal of the volatile components, the residue was again dissolved in H2O (1 ml) and applied to a GEL 3011-C column (see above) equilibrated with H2O. The column was then washed with H2O (40 ml) and developed with 0.05 M acetic acid. TTX, 4,9-anhydroTTX, TTX-8-O-hemisuccinate (yield, 34% from TTX), and 4,9-anhydroTTX-8-O-hemisuccinate were eluted successively at 21 to 26 ml, 27 to 29 ml, 30 to 32 ml, and 33 to 42 ml. FAB-MS of the preparation of TTX-8-O-hemisuccinate revealed a protonated ion at m/z 420. The signals on1H and 13C NMR spectra were assigned by 1H-1H correlation spectroscopy and13C-1H correlation spectroscopy. 1H NMR, δ (ppm): 2.51[1H, d,J 9.0 Hz, H-4a], 2.75[2H, dd, J 4.5, 8.1 Hz, 2′ or 3′-CH2], 2.83[2H, dd, J 4.5, 8.1 Hz, 2′ or 3′-CH2], 4.03[1H, s, H-9], 4.03[1H, d, J 14.1 Hz, H-11], 4.05[1H, d, J 14.1 Hz, H-11], 4.24[1H, br s, H 7], 4.32[1H, br s, H-5], 5.48[1H, br s, H-8], 5.53[1H, d, J 9.0 Hz, H-4].13C NMR 32.1, 32.8[C-2′, C-3′], 40.1[C-4a], 58.2[C-8a], 64.9[C-11], 70.8[C-9], 71.0[C-6], 73.3[C-5], 74.6[C-4], 75.3[C-8], 76.6[C-7], 110.8[C-10], 156.6[C-2], 176.2, 181.5[C-1′, C-4′]. The position of the hemisuccinated hydroxyl group of TTX-8-O-hemisuccinate was determined on the basis of the downfield shifts of H-8 (1.28 ppm) and C-8 (2.5 ppm) and the upfield shifts of C-7 (3.1 ppm) and C-8a (1.5 ppm) compared with those of TTX. Notably, 11-O-monohemisuccinyl derivatives (such as 4,9-anhydroTTX-11-O-hemisuccinate or TTX-11-O-hemisuccinate) were not detected throughout the reaction steps described above. Furthermore, hydrolysis of 4,9-ether bond in 4,9-anhydro-11-O-hemisuccinate, which we obtained by a different succinylation method with succinic anhydride andp-toluenesulphonic acid in CH3CN, gave only TTX and 4,9-anhydroTTX (M.Y.-Y., unpublished data). These observations suggest that the 11-O-ester bond is much more liable to hydrolysis than the 8-O-ester bond. The higher stability of the TTX-8-O-hemisuccinate than the TTX-11-O-hemisuccinate may be ascribable to intramolecular salt-bridge formation between the hemisuccinyl carboxyl acid and the guanidinium group of TTX-8-O-hemisuccinate (see below).
Purity and Quantification of the TTX Analogs
The degrees of TTX contamination in the preparation of the analogs used except for 4-epiTTX and 4,9-anhydroTTX were less than 1% (mol/mol), as judged by 1H NMR spectroscopy and fluorometric HPLC. The preparation of 4-epiTTX and 4,9-anhydroTTX contained up to 1% (mol/mol) of TTX as a result of spontaneous equilibration with TTX (Mosher, 1986). The preparation of TTX-8-O-hemisuccinate contained about 0.5% of TTX produced by hydrolysis of the ester bond under the condition of binding assay. The preparation of TTX-11-carboxylic acid contained 10% (mol/mol) of its 4,9-anhydrated form. Because of the difficulty of weighing the small amount of hygroscopic TTX analogs, quantification was made by 1H NMR spectroscopy using TTX as the standard. The 1H NMR spectra of four different amounts of TTX (173–690 nmol, quantified by weighing) were obtained in 3.98% CD3COOD/D2O (525 μl, v/v) containing 0.00038% t-butanol (v/v) as the internal standard, and the ratios (Y) of the signal intensity of H-4 of TTX (δ, 5.54 ppm) to that of t-butanol (δ, 1.24 ppm) were plotted against the amount of TTX (X; in nmol). Linear regression of the (X, Y) data gave a line,Y = 0.00309X-0.0216 (correlation coefficient, r = 0.997), which was used to convert the NMR signal intensity ratios similarly measured for the TTX analogs to their amounts.
Preparation of Rat Brain Synaptic Membrane
Rat brain synaptic membrane was prepared essentially by the procedure described by Hartshorne and Catterall (1984). Briefly, the rat brain tissue (20 g) was homogenized with a glass homogenizer in 200 ml of an ice-cold buffer containing 5 mM Tris, 0.32 M sucrose and four protease inhibitors: 0.1 mM phenylmethylsulfonyl fluoride, 1 mM iodoacetamide, 1 mM 1,10-phenanthroline monohydrate, and 1 mM pepstatin A (pH 7.4, adjusted with HCl), and sedimented at 700g for 10 min. The supernatant was saved, and the pellet was resuspended in 200 ml of buffer of the same composition was centrifuged as before. The supernatants obtained by this and the previous centrifugation steps were combined and again centrifuged at 11,500g for 20 min. The precipitate was resuspended with a buffer containing 50 mM Tris, 1 mM EDTA, and the four protease inhibitors (pH 7.4, adjusted with HCl) and centrifuged as before. The membrane suspension thus prepared was kept frozen at −80°C for up to 2 months. The protein content was measured according to the method of Lowry et al. (1951).
Binding Assay
Rat brain synapse membrane fraction (60 μg/ml protein) was incubated with [3H]STX in 2 ml of the incubation medium of the following composition: 5 mM HEPES/Tris, 130 mM choline chloride, 5.5 mM glucose, 0.8 mM MgSO4, and 5.4 mM KCl (pH 7.4). After 15 min of incubation at 0°C, the mixture was decanted into the inlets of vacuum chamber equipped with GC/C glass fiber filters (24-mm diameter; Whatman International, Ltd., Maidstone, UK) (see Catterall et al., 1979, and Takai et al., 1995). To remove unbound [3H]STX, the filters were washed thrice with 1 ml of the washing medium containing 5 mM HEPES/Tris, 163 mM choline chloride, 1.8 mM CaCl2, and 0.8 mM MgSO4 (pH 7.4), and the bound radioactivity was determined with 8 ml of an EX-H scintillation cocktail (Dojindo Labs, Kumamoto, Japan). To measure the nonspecific binding, excess (3 μM) TTX was added to the medium.
Molecular Modeling
Molecular modeling of TTX-8-O-hemisuccinate was performed with Sybyl v6.1a software (Tripos Associates, St. Louis, MO) using the Tripos force field (Clark et al., 1989) with the Powell method and the molecular orbital packaging charges (Stewart, 1990).
Analysis of Data
The symbols used in the following mathematical descriptions are defined in Table 1.
Estimation of B andK1.
When an isotope-labeled ligand, L1, interacts with a single species of binding sites, B, in the absence of unlabeled ligand, the steady-state concentration of the BL1
complex, φ, is given as an explicit function of the total concentration of L1,L1
(Takai et al., 1995) as follows:
A more conventional way to estimate B andK1 and K1 from (L1, φ) data is to use a linearizing transformation of eq. 1:
Estimation of K0.
Binding of a labeled ligand, L1, to a single species of binding sites in the presence of an unlabeled ligand, L0, is represented by a general interaction model as shown in Scheme FS1. For simplicity, we now assume that the reduction of free concentrations of L0 and L1 as a result of their binding to B is negligible; i.e.,L0 = [L0] andL1 = [L1]. This conventional assumption is reasonable ifL0 > 10B andL1 > 10B. The dependence of φ on L0 is then described by the following equation, which can readily be derived from the conservation equations and the equations describing the mass-action low:
Competitive binding of two ligands can be defined as a limiting case of the above-mentioned model, where k → ∞ and therefore [BL0L1] → 0 (see SchemeFS1). In this case, equation 3 reduces to the following:
A normalized form of eq. 4 is as follows:
Fitting and Statistics.
The model functions were fitted to data by nonlinear least-squares regression using the Origin software, version 5.0 (Microcal Software, Northampton, MA), which utilizes the Levenberg-Marquardt algorithm (Sen and Srivastava, 1990). Fitting calculations were performed using as weight the reciprocal of the square of the standard error at each value of the independent variable, and the values obtained thereby with the standard errors were compared by a modified t test (Carroll and Ruppert, 1988). Differences were taken as statistically significant if a two-tailed probability of less than 0.05 was obtained.
Results
Binding Assay.
Figure 3 shows the relationship between the total concentration of [3H]STX (L1) and its saturable binding to the rat brain membrane (60 μg of protein/ml). Equation 1 was fitted well to the data by nonlinear least-squares regression, indicating that the binding is attributable to a class of high-affinity receptors with the concentration (B) of 0.093 ± 0.004 nM and the equilibrium dissociation constant (K1) of 0.43 ± 0.04 nM (total number of experiments, n= 21).
As shown in Fig. 4, the specific binding of [3H]STX to the receptor sites was inhibited by addition of the unlabeled TTX derivatives in a dose-dependent manner. In these experiments, the total molar concentration of [3H]STX (L1) was kept constant at 2.5 nM and that of the binding sites (B) was set so that it would not exceed 0.2 nM; thus,L1 > 12B. It seems therefore reasonable to use equation 3 as a regression model for the (L0, φ) data (seeMaterials and Methods). Nonlinear least-squares fitting of this model function to the data gave the regression curves that appeared to fit the data points satisfactorily (Fig. 4). The fitting also gave estimates of k (equation 4), which fell in the range 1015 to 1046. The very large values of k indicate that the binding of the labeled and unlabeled ligands to the rat brain membrane is extremely exclusive. Indeed, the competitive binding model (equation 4) gave indistinguishably similar regression curves (see Fig. 4) and estimates of K0 to those given by the general interaction model. The K0 values enlisted in Table 2 are those obtained by the fitting with the competitive binding model.
Table 2 also gives γ, the ratio of theK0 value of each derivative to that of TTX, from which the difference of standard free-energy change of binding δΔG was estimated by the conversion equationδΔG = RTlnγ. For comparison, the differences of the free-energy change were similarly calculated using the ratio of the IC50 values obtained previously by electrophysiological experiments (Table 2;δΔG′). (See Ritchie and Rogart, 1977, and Kao, 1986, for the merits of using ratios of IC50 rather than IC50 values themselves in evaluating electrophysiologically obtained dose-inhibition data.) Considering the differences of the methods, preparations, and conditions used, there seems to be a reasonable agreement between the values ofδΔG and δΔG′ except for those of 4-epiTTX and 11-norTTX-6,6-diol (see below).
Generally, the K0 value for a TTX analog that has lower affinity compared with TTX tends to be underestimated in the presence of contamination by TTX. This tendency becomes more marked in proportion to the magnitude of γ. Theoretically, the molar ratio of TTX to its analog should be sufficiently lower than the reciprocal of the γ value for reliable estimation of the K0 value for the analog. Because it is usually difficult to keep TTX contamination of TTX analogs much lower than 1%, theK0 values obtained by the present binding assay experiments for the analogs of low affinity (γ > 100) should be regarded as lower limits of the actual values. The preparation of TTX-8-O-hemisuccinate, for example, caused a measurable decrease in the binding of [3H]STX in the concentration range 0.1 to 30 μM (Fig. 4). From this result, theK0 value for this analog was estimated at 380 nM, a value 210-fold larger (i.e., γ = 210) than that for TTX (Table 2). However, the fluorometric HPLC detected in this preparation about 0.5% of TTX, which was possibly generated by hydrolysis of the ester bond (see Materials and Methods). Note that the percentage of TTX contamination is very close to the reciprocal of the γ value. This means that the observed reduction of the [3H]STX binding was mostly related to the contamination by TTX rather than to the analog itself. Thus, the actual K0 value for TTX-8-O-hemisuccinate is probably much larger than 380 nM.
The present binding assay experiments with the rat synaptic membrane detected a considerably larger reduction of the affinity with 4-epiTTX (δΔG ≅ 9.0 kJ/mol) than did the previous voltage-clamp experiments on the squid giant axon (δΔG′ ≅ 2.7 kJ/mol). The preparation of 4-epiTTX inevitably contains up to 1% (mol/mol) of TTX as a result of spontaneous equilibration with TTX (seeMaterials and Methods). As the tendency ofK0 to be underestimated because of TTX contamination is more marked with analogs of lower affinity (see above), the real difference of δΔG andδΔG′ for 4-epiTTX can be even larger. The contribution of the C-4 hydroxyl to the binding of TTX to the sodium channels may be different in different tissues (seeDiscussion).
In the previous electrophysiological experiments, Kao (1982) showed that 11-norTTX-6,6-diol (often simply referred to as 11-norTTX) was 4 times less active (δΔG′ = 3 kJ/mol) on the squid giant axon and 13 times less active (δΔG′ = 6 kJ/mol) on frog muscle fibers (Table 2) compared with TTX. In the present experiments with the rat synaptic membrane, 11-norTTX-6,6-diol exhibited almost the same affinity (δΔG ≅ 0 kJ/mol) as did TTX (Table 2). We cannot decide from the present experiments to what extent the apparently large difference between δΔG andδΔG′ for 11-norTTX-6,6-diol can be ascribed to the difference of the methods and/or tissues used. As argued by Kao (1982), such variability may well be attributable to the fact that 11-norTTX-6,6-diol exists in a facile equilibrium between the active hemilactal form and the possibly less active 10,7-lactone form, which is affected by various factors including the pH of the medium and the duration for which the derivative is left in solution (see also Mosher, 1986).
Molecular Modeling.
Figure 5 shows the minimal-energy conformation of TTX-8-O-hemisuccinate predicted by the computer-assisted modeling, which estimated the distance between the guanidinium carbon (C-2) and the two carboxylate oxygens of the hemisuccinyl group at 2.9 Å and 4.9 Å. The molecular orbital packaging charge calculation indicated that the center of the positive charge (+0.94) was localized on this C-2 because of the resonance within the protonated guanidinium group. The modeling also generated several other structures corresponding to the local minima of the energy function. Although they were slightly different in the direction of the C-11-hydroxyl, the relative position of the hemisuccinate group with respect to the guanidinium group was essentially the same in these alternative structures as in the one shown in Fig. 5. Thus, the guanidinium group is very likely to form an intramolecular ionic bond with the carboxylate in the 8-O-hemisuccinyl group.
Discussion
The roles of the hydroxyls at C-6 and C-11 of the TTX molecule in the binding of the toxin to the sodium channel have long escaped proper recognition, although various derivatives with modifications involving these hydroxyls have been obtained since the days of the earlier studies conducted in connection with the proof of structure of TTX (seeMosher, 1986). In the previous electrophysiological study (Yang et al., 1992) and the present binding assay experiments, we have shown that 6-epiTTX and 11-deoxyTTX, in which the hydroxyls are translocated or reduced, exhibit markedly lower affinities than TTX. In contrast, high affinity is retained in the derivatives like chiriquitoxin and 11-oxoTTX, which have all the six hydroxyls including those at C-6 and C-11 in the same position as does TTX (Table 2). [Note that in aqueous solution, the aldehyde group of 11-oxoTTX is present as the hydrated form, −CH(OH)2.] These observations clearly demonstrate the involvement of the two hydroxyls in the toxin binding to the sodium channel.
The standard free-energy change for binding, ΔG, of TTX to the sodium channel is estimated by the equation, ΔG =RTlnK0, at −50 kJ/mol. We have shown that removal of the C-11 hydroxyl (11-deoxyTTX) results in a decrease in the absolute value of ΔG by 7.5 kJ/mol (Table2, δΔG). The change of ΔG of this magnitude is consistent with the idea that the hydroxyl at C-11 acts as a donor for a hydrogen bond, which is possibly formed with a carboxyl residue of the channel protein (see below). Although the epimeric translocation at C-6 (6-epiTTX) changes the position of the two hydroxyls at C-6 and C-11, the δΔG (7.7 kJ/mol) corresponding to this structural modification is comparable with that for removal of only one hydroxyl at C-11 (11-deoxyTTX). This apparent discrepancy was left unexplained in the previous electrophysiological experiments (Yang and Kao, 1992), in which 6-epiTTX exhibited even higher affinity than did 11-deoxyTTX. In the present experiments, we have shown that 11-norTTX-6(S)-ol and 11-norTTX-6(R)-ol exhibit almost the same affinity as 6-epiTTX. This suggests that the hydroxyl at the 6(R)-position is as effective as the hydroxyl at the 6(S)-position or the C-11-hydroxyl in forming a hydrogen bond. Indeed we see that the affinity of 11-norTTX-6,6-diol to the rat brain membrane is similar to that of TTX. The nature of effects of the epimeric translocation on the effectiveness of the hydroxyls in the C-6 end would be further clarified if we could obtain a derivative in which both the hydroxyls are specifically removed (e.g., 6,11-dideoxyTTX). We have shown that 5,6,11-trideoxyTTX retains almost no affinity. This derivative, however, involves other complex changes including dehydrogenation of the C-10-hydroxyl, which is believed to be essential for the affinity.
Specific replacement of the hydroxymethyl group at C-6 to a carboxylic acid (TTX-11-carboxylic acid) has been shown to cause a large (γ > 1300) increase in theK0 (Table 2). This marked change in the affinity may have a special implication for the property of the possible binding site for the C-11-hydroxyl. According to the current model, α subunits of the sodium channel assume the tertiary structure in which each of the four repeat domains (I–IV) with its six transmembrane helix segments (S1–S6) is circularly disposed to form a central ion pore (see, for example, Catterall, 1995). The short segments (named SS1–SS2 regions) of the extended interhelical loop connecting S5 and S6 in the extracellular side of the membrane are thought to fold back into the pore and contribute to its ion selectivity and conductance properties. It has long been known that the affinity of TTX and STX to the sodium channel is reduced or abolished by procedures that protonate or covalently modify carboxylic acid residues of the extracellular side of the membrane (see Hille, 1992). Mutational experiments have now identified several acidic amino acid residues of the SS2 regions as being crucially important for the binding of TTX and STX (Kontis and Goldin, 1993; Noda et al., 1989;Terlau et al., 1991). It seems therefore reasonable to speculate that the C-11-hydroxyl contributes to the binding by forming a hydrogen bond with one of the carboxylic acid residues. Electrostatic repulsion between the negative charges of the carboxyl groups would explain the remarkable reduction of affinity observed with TTX-11-carboxylic acid.Nakayama et al. (1992) showed in an eel sodium channel preparation that covalent labeling using a TTX derivative, in which a photoreactive label was attached to the C-11 position, resulted in incorporation of the label into domains III and IV, whereas no incorporation was detected in domain I. On the basis of such observations, they proposed a binding model in which TTX is fitted into a pocket constituted by all four SS2 regions by orienting its guanidinium residue to domains I and II and its C-6 end to domain IV. Lipkind and Fozzard (1994) have discussed that it is physically reasonable to dock the hemisphere of the TTX molecule including the guanidinium and the hydroxyls at C-9 and C-10 of TTX into the pocket of negative charges constituted by the carboxyls belonging to the SS2 of domains I and II. If we were to accept such an orientation of TTX in its binding site, the most plausible candidate for the acidic residue as the acceptor for the C-11-hydroxyl, which is located on the opposite side of that of the guanidinium group, should be Asp-1717, the only negatively charged residue in the SS2 region of domain IV of the rat brain sodium channel II. Terlau et al. (1991) reported that neutralization of the charge by replacement of this residue with asparagine resulted in an increase in the IC50 of TTX from 18 nM to 350 nM. They observed only a minimal change in the TTX affinity with similar mutational neutralization of Asp-1426, the only acidic residue in the SS2 of domain III. It is worth noting that the ratio (= 19) of the IC50 value for the Asn-1717 mutant to that for the wild-type sodium channel (Terlau et al., 1991) is in close agreement with the ratio (= 21) of theK0 value for 11-deoxyTTX to that for TTX (Table 2, γ).
The hydroxyl at C-9 as well as that at C-10 is generally thought to be the most important of the six hydroxyls of TTX in its binding to the sodium channel (see Kao, 1986). This current opinion about the importance of the C-9-hydroxyl is mainly based on the observation ofKao and Yasumoto (1985), who examined the effects of 4-epiTTX and 4,9-anhydroTTX of the sodium channel current in the squid axon by means of the voltage-clamp. Because the IC50 was 4 nM for TTX, 13 nM for 4-epiTTX, and 300 nM for 4,9-anhydroTTX, they concluded that the contribution of the C-9-hydroxyl to the toxin binding is much larger than that of the C-4-hydroxyl. The present binding assay experiments with the rat brain membranes detected a considerably larger change in the affinity with 4-epiTTX than did the previous electrophysiological experiments (see Results). However, theδΔG value obtained in the present experiments for 4,9-anhydroTTX is comparable with the values estimated from the results of previous electrophysiological experiments. In the rat brain sodium channel, the C-4-hydroxyl may contribute to the toxin binding to the same extent as the C-9-hydroxyl.
TTX-8-O-hemisuccinate, which is, to our knowledge, the first TTX derivative with specific modification of the C-8-hydroxyl, gave a much larger K0 value than did TTX (γ > 210; see Table 2 and Results). There are several equally plausible explanations for the marked reduction of the affinity. 1) Molecular modeling estimates the distance between the guanidinium carbon and the carboxylate oxygen of the hemisuccinyl group at 2.9 Å for a minimal-energy conformer of TTX-8-O-hemisuccinate (Fig. 5). This strongly suggests that the guanidinium group forms an intramolecular ionic bond with the carboxylate of the 8-O-hemisuccinyl group. (If this is really the case, the present result is also the first demonstration of modifying the guanidinium group of TTX without completely destroying the binding activity.) Such an intramolecular ionic bond would greatly reduce the intermolecular ionic interaction between the acidic residues of the channel protein and the guanidinium group, which is believed to be the most essential for the action of TTX on the sodium channels. 2) The 8-O-hemisuccinate moiety is relatively bulky. It is therefore possible that some part of the reduction of the affinity is related to a steric interaction of this moiety with the wall of the binding pocket for TTX. 3) It is also possible that there is a receptor site in the binding pocket that forms a hydrogen bond with the C-8-hydroxyl; if so, the loss of this hydroxyl by the esterification would result in a reduction of the affinity. At the moment, it is impossible to decide which of these possible factors is primarily responsible for the observed phenomenon. Two or even all of the three factors might well be jointly involved. To address this question, derivatives with simpler modification at C-8 of TTX (e.g., 8-epiTTX or 8-deoxyTTX) would be very useful.
Footnotes
-
Send reprint requests to: Mari Yotsu-Yamashita, Graduate School of Agriculture, 1-1 Tsutsumidori-Amamiyamachi, Aoba-ku, Sendai 981-8555, Japan. E-mail:myama{at}biochem.tohoku.ac.jp
-
↵1 This work was supported by Grants-in-Aid 07102002 and 10760043 from the Ministry of Education, Science, Sports and Culture of Japan, a Suntory Institute for Bioorganic Research grant, and grants from the Naito Foundation and the Hayashi Memorial Foundation for Female Natural Scientists. A.T. is a member of a Research for the Future Program of the Japan Society for the Promotion of Science (project number: 96L00504).
-
↵2 Present address: Japan Food Research Laboratories, 6-11-10 Nagayama, Tama-shi, Tokyo 206-0025, Japan.
- Abbreviations:
- TTX
- tetrodotoxin
- STX
- saxitoxin
- FAB-MS
- fast atom bombardment-mass spectroscopy
- Received December 9, 1998.
- Accepted February 16, 1999.
- The American Society for Pharmacology and Experimental Therapeutics