Abstract
It has been suggested that cocaine and mazindol bind to separate sites on the dopamine transporter. In the present study, we address this issue by examining the inhibition by mazindol of the binding of [3H]WIN 35,428 ([3H]2β-carbomethyoxy-3β-(4-fluorophenyl)-tropane), a phenyltropane analog of cocaine, and the inhibition by WIN 35,428 of [3H]mazindol binding to the cloned human dopamine transporter expressed in C6 glioma cells. The design involved the construction of inhibition curves at six widely different radioligand levels, enabling the distinction between the nonlinear hyperbolic competition (i.e., negative allosteric) model and the competitive (i.e., mutually exclusive binding) model. Nonlinear computer curve-fitting analysis indicated no difference in the goodness of fit between the two models; the negative allosteric model indicated an extremely high allosteric constant of ∼ ≥100, which practically equates to the competitive model. The present results suggest that complex interactions reported between cocaine and mazindol in inhibiting dopamine transport are beyond the level of ligand recognition.
The dopamine transporter has been suggested to be the initial target for cocaine in producing its reinforcing and addictive effects (Wise, 1984;Ritz et al., 1987; Carroll et al., 1992). It is widely accepted that blockade of dopamine uptake by cocaine is the result of high-affinity binding of cocaine to the dopamine transporter (Reith et al., 1986; Calligaro and Eldefrawi, 1987; Ritzet al., 1987; Madras et al., 1989). It is not known whether the binding sites for cocaine are unique and whether this accounts for the exceptional abuse potential of cocaine among all compounds that block dopamine uptake (Fahey et al., 1989;Berger et al., 1990; Reith et al., 1992; Carrollet al., 1992; Reith and Selmeci, 1992). Although a competitive model has been advanced for the interaction of cocaine or the close cocaine congener WIN 35,428 and mazindol in equilibrium binding experiments in rodent striatum (Javitch et al., 1984; Calligaro and Eldefrawi, 1987; Reith and Selmeci, 1992; Derschet al., 1994) or rabbit caudate nucleus (Aloyo et al., 1995), more complex interactions have been observed in monkey caudate putamen (Madras et al., 1989) and rat striatum (Berger et al., 1990). In a recent study, Meiergerd and Schenk (1994) reported that cocaine and mazindol interact with separate but interacting sites in inhibiting dopamine uptake in rat striatal suspensions measured by rotating disk electrode voltammetry.
In the present radioligand-binding study, we examined the issue of whether cocaine and mazindol bind to separate or interacting sites on the dopamine transporter by two new experimental approaches. First, the cloned human dopamine transporter is studied rather than the native rodent transporter because it is known that there are a number of differences between the human and rat dopamine transporter (Giroset al., 1992). Although differences have been reported between cloned and native dopamine transporters within the same species, these have been mostly in the area of substrate interaction or translocation, with generally higher Kd values for [3H]dopamine uptake and Ki values for MPP+ (1-methyl-4-phenylpyridinium) or norepinpehrine in inhibiting [3H]dopamine uptake for the clones (Pifl et al., 1993; Giros and Caron, 1993) but comparable inhibitory potencies for a variety of blockers (Piflet al., 1993; Giros and Caron, 1993; Eshleman et al., 1995). Second, rather than relying on the steepness and completeness of inhibition curves obtained at one level of radioligand (Javitch et al., 1984; Calligaro and Eldefrawi, 1987; Madraset al., 1989; Aloyo et al., 1995) or the change in slope but not abscissa intercept in Scatchard plots (Reith and Selmeci, 1992; Dersch et al., 1994), we now report on experiments involving a series of inhibition curves obtained at radioligand concentrations varying over a wide range, as recommended byTomlinson and Hnatowich (1988). This design enables the distinction between linear and nonlinear (“hyperbolic”) competitive interactions; with one level of radioligand, the latter may be mistaken for simple competitive inhibition. The current experimental design involves two pairs, [3H]WIN 35,428 with mazindol and [3H]mazindol with WIN 35,428, studied under conditions yielding one-site binding for either radioligand.
Methods
Models
Hyperbolic competitive model.
First, the present inhibition data were analyzed according to the hyperbolic competitive model as defined by Tomlinson and Hnatowich (1988), which is equivalent to the negative allosterism model of Ehlert (1988). In the model, a receptor (R) bears two classes of interacting binding sites: site 1 is the site to which the ligand (L) binds; site 2 binds inhibitor (I) only. I decreases the affinity of R for L by binding to site 2 through an allosteric relationship:
K
1 and K
3are the equilibrium dissociation constants of the reactions, and β is the allosteric coefficient, which is >1 in the case of negative cooperativity. It can be derived (Tomlinson and Hnatowich, 1988) that:
Competitive model.
It can be easily seen that if β ≫ 1, the hyperbolic model reverts to competitive inhibition:
which is formally indistinguishable from classic competitive inhibition involving one and the same site that binds L and I. With β ≫ 1, it then follows that equation 5 becomes
C6 glioma cells stable expressing the human dopamine transporter.
cDNA encoding the human dopamine transporter was cloned in the laboratory of Dr. Aaron Janowsky (Oregon Health Sciences University, Portland, OR) by screening a human substantia nigra cDNA library with a polymerase chain reaction-amplified probe based on the rat dopamine transporter cDNA sequence (Eshleman et al., 1995). Rat C6 glioma cells were transfected, grown for several passages and frozen in a medium containing 45% Dulbecco’s modified Eagle’s medium, 5% dimethylsulfoxide and 50% fetal bovine serum for storage in liquid nitrogen. For each experiment, one freezing vial was rapidly thawed and seeded into a 75-cm2 flask at a density of ∼100,000 cells/cm2. When the flask reached confluency (after ∼4 days at a density of ∼600,000 cells/cm2), the cells were lysed by trypsinization and seeded into four 75-cm2 flasks. At ∼4 days later, when the flasks reached confluency (∼600,00 cells/cm2), the cells were ready for the binding experiment.
The medium was removed from the cells in the flasks, and the cells were washed with Ca++- and Mg++-free phosphate-buffered saline and lysed with 2 mM HEPES and 1 mM EDTA, pH 7.6, at room temperature. The lysate was centrifuged at 31,000 ×g for 20 min, and the pellet was put into a Brinkmann Polytron (setting 6, 15 sec) in the binding assay buffer (see below). Four flask equivalents were needed for one data set consisting of six levels of radioligand with varying inhibitor.
Inhibition of [3H]WIN 35,428 binding by mazindol and [3H]mazindol binding by WIN 35,428.
Assays were carried out in triplicate in a total volume of 0.2 ml in 1-ml ministrip tubes (Skatron, Sterling, VA). The composition of the assay mixture was 20 μl of radioligand in water, 10 μl of inhibitor in water, 120 μl of assay buffer (33 mM sodium phosphate, pH 7.4, at room temperature) and 50 μl of cell membrane preparations in assay buffer. The incubation was carried out at 0° to 4°C for 2 hr and terminated by filtration with a miniharvesting apparatus (type 11021, Skatron). The detailed procedures of the binding assays were as we previously described (Coffey and Reith, 1994). Nonspecific binding was defined with 100 μM cocaine.
Inhibition curves were constructed by combining varying concentrations of mazindol (0.1, 1, 3, 10, 30, 100 and 1000 nM) with six different levels of [3H]WIN 35,428 (0.725, 1.586, 3.38, 7.14, 14.77 and 30.56 nM or closely related values in follow-up experiments) and,vice versa, varying concentrations of WIN 35,428 (each curve with six of the following points: 0.5, 3, 5, 15, 30, 150, 300, 1500 and 3000 nM) with six different levels of [3H]mazindol (1.1, 2.31, 5.24, 11.11, 23.43 and 49 nM or closely related values in follow-up experiments).
Inhibition of [3H]WIN 35,428 binding by WIN 35,428.
Saturation analysis was carried out by adding increasing concentrations (0.5, 1, 2, 5, 10, 20, 40 and 100 nM) of WIN 35,428 to a constant concetration (0.35 nM) of [3H]WIN 35,428. All other procedures were as above.
Data analysis and statistics.
IC50 values and pseudo-Hill numbers were computed with the equation of the ALLFIT program of DeLean et al. (1978) entered into the Microsoft ORIGIN curve-fitting and plotting software, which was run with total and nonspecific binding entered as constants. IC50 values of inhibitors (mazindol and WIN 35,428) as a function of varying levels of radioligand ([3H]WIN 35,428 and [3H]mazindol, respectively) were plotted as described above. The dissociation constant of the inhibitor (K 3) obtained from the y axis intercept was compared with the dissociation constant of the same compound when used as a radioligand (Kd =K 1 in above models).
Results are expressed as mean ± S.E.M. Data were analyzed by the two-tailed Student’s t test (paired where appropriate) and two-way analysis of variance. Graphic analysis consisted of least-squares linear regression and correlation analysis. The statistical significance of the difference between the goodness of the fit for the apparent competitive and the more complex allosteric model was assessed with the F test [F = [(SSa − SSb)/(dfa − dfb)]/(SSb/dfb), in which SS is sum of squares, df is degrees of freedom, a is the simple model and b is the more complex model] as outline by Munson and Rodbard (DeLeanet al., 1978). The accepted level of significance was .05.
Materials
[3H]WIN 35,428 [lot no. 3141–232; specific activity, 109.3 Ci/mmol; determined by the homologous competition binding method we previously described (Wiener and Reith, 1992)] and [3H]mazindol (lot no. 2824–285; specific activity, 17.0 Ci/mmol) were obtained from DuPont-New England Nuclear (Boston, MA). WIN 35,428 was from Research Biochemicals (Natick, MA). Mazindol was from Sandoz Pharmaceuticals (E. Hanover, NJ). All other chemicals were from Sigma Chemical (St. Louis, MO) or Fisher.
Results
Fitting of K1, K3and β in the hyperbolic competitive model.
The [Lb]/[Lb]max values for a representative experiment of inhibition of [3H]WIN 35,428 binding by mazindol at varying [L] are shown in figure 1A, and the mirror experiment with [3H]mazindol and WIN 35,428 is shown in figure 1B. It can be seen that raising [L] resulted in progressively higher [Lb]/[Lb]max values at a given [I]. Fitting of the data sets of the separate experiments to the hyperbolic competitive model (equation 5) resulted in extremely high values for the allosteric constant β (table1, top). With all parameters unrestricted, best-fit estimates for β ranged from 106 to 240 for all experiments with either [3H]WIN 35,428 or [3H]mazindol, except for one experiment that gave a value of 1.5 × 1013 (table 1, top). There were no statistically significant differences between Kd values that were determined with the compound as the tritiated ligand and Ki values for the same compound as an inhibitor with the alternate compound as the tritiated ligand (table 1, top).
Fitting of K1 andK3 in the competitive model.
Fitting of the data sets to the competitive model (equation 6) yieldedKd and Ki values virtually indistinguishable from those obtained in the hyperbolic competitive model (compare table 2 with table 1). In fact, the goodness of the fit did not improve in going from the simple to the complex model when tested separately by theF test for each data set (table 2). Furthermore, there was a positive correlation between the nonlinear and linearKd values found in separate experiments (r = .999, P < .00001) (fig.2), suggesting that the two models are indistinguishable.
Again, there were no statistically significant differences betweenKd values that were determined with the compound as the tritiated ligand and Ki values for the same compound as an inhibitor with the alternate compound as the tritiated ligand (table 2).
Graphic determination of K1 andK3 in the competitive model.
The inhibition curves for mazindol in the [3H]WIN 35,428 binding assays and for WIN 35,428 in the [3H]mazindol binding assays shifted to the right with increasing [L] (fig. 1, A and B). The Hill values for inhibition of [3H]WIN 35,428 binding were 0.99 ± 0.03 and those for [3H]mazindol binding were 0.98 ± 0.03. The IC50 values plotted as a function of [L] according to the competitive model (equation 8) fell on a straight line for each data set examined (fig. 3, A–F). There were no statistically significant differences betweenKd values that were determined with the compound as the tritiated ligand and Ki values for the same compound as an inhibitor with the alternate compound as the tritiated ligand (table 3). As can be seen in figure 3, there were day-to-day differences in the slopes of the regression lines, which in turn caused variation in theKd estimates, but, on average, the affinities obtained from this analysis were comparable to those from the hyperbolic competitive analysis (table 1, top).
Fitting of β in the hyperbolic competitive model withK1 or K3 as a constant.
It was important to make sure that the estimate for allosterism (β) in the hyperbolic competitive model was not affected in a significant manner by the fitted value ofKd or Ki . Therefore, the data sets were fitted with Ki restrained to the value obtained from the linear regression analysis of the same sets (y intercept in fig. 3, A–F), and again the β values were high, ranging from 132 to 238 except for an extreme outlyer in each group (2.4 × 1010 and 2.5 × 1012) (table 1, bottom). When the Kd value in the hyperbolic competitive model was set to the value arrived at by the competitive graphic analysis, the fitted Ki value for each data set was very close to that deduced from the competitive model (data not shown). Again, β was high, ranging from 94 to 253 with two outlying values (6.2 × 103 and 1.4 × 109) in the total of six experiments (data not shown).
Saturation analysis of [3H]WIN 35,428 binding.
Inhibition of [3H]WIN 35,428 binding by WIN 35,428 was examined under the current conditions (with three independent membrane preparations). Analysis by LIGAND could be performed only with the one-site model in agreement with the linear Scatchard plot (see fig.4 for a representative experiment).
Discussion
One-site binding model for the radioligands used.
There is general consensus that [3H]mazindol binds to only one site on the dopamine transporter (Javitch et al., 1984;Reith and Selmeci, 1992; Dersch et al., 1994). Indeed, membranes prepared from the present C6 glioma cell system expressing the human dopamine transporter displayed only one component of [3H]mazindol binding as determined in a separate set of recent experiments carried out under slightly different assay conditions (Wu et al., 1997, in press). However, [3H]WIN 35,428 binding has been reported to be heterogeneous for the cloned rat (Boja et al., 1992) and human (Pristupa et al., 1994) dopamine transporter expressed in COS cells. In contrast, one-site binding has been reported for the binding of the closely related [125I]RTI-55 ligand to the cloned rat (Gu et al., 1994) and human (Eshleman et al., 1995) dopamine transporter. We examined this issue previously for [3H]WIN 35,428 binding to the current C6 glioma cell preparation with a low concentration of radioligand (0.45 nM) and an extended range of unlabeled WIN 35,428 (0.4–1000 nM) in assays with larger volumes without sucrose as described previously (Reith and Selmeci, 1992); no evidence was found for more than one binding component (Reith et al., 1996). Under the present conditions with sucrose, again only one component was observed (fig. 4). Most likely, the present C6 glioma cell system does not express the low-affinity [3H]WIN 35,428 binding component, which has been suggested to be unrelated to dopamine uptake (Pristupa et al., 1994).
Allosterism in the hyperbolic competitive model vs.competitive binding.
The present results show that if there is an allosteric interrelationship between WIN 35,428 and mazindol binding sites, the allosterism is extreme because of the high β values, practically equating the situation to competitive interaction. Indeed, for any given data set, there was no statistically significant difference between the goodness of the fit of the allosteric (β ∼ ≥100) and competitive (β→ ∞) model. Furthermore, among different data sets, there was a positive correlation between dissociation constants fitted in the competitive model and those fitted in the allosteric model, indicating that day-to-day variations affected the absolute values of the kinetic constants but not the type of applicable model.
When allosterism becomes extreme, as in the present results for the interaction between WIN 35,428 and mazindol binding, it is indistinguishable from competitive interaction. As pointed out previously by Segel (1975), the latter can result from the two compounds binding to one and the same site, to a shared site, to distinct but overlapping sites, to distinct sites involving steric hindrance or to distinct sites linked conformationally. The present modeling approach cannot distinguish between these possibilities but does remove one more obstacle in the attempt to resolve the disparity between previously described complex (Madras et al., 1989;Berger et al., 1990) and simple competitive (Reith and Selmeci, 1992; Aloyo et al., 1995) interactions between cocaine-like compounds and mazindol. As pointed out by Tomlinson and Hnatowich (1988), negative allosteric interactions can masquerade as simple competitive inhibition when varying inhibitor concentrations are examined at only one level of radioligand (classic inhibition curves) or when varying radioligand concentrations are assessed at only one level of inhibitor (classic Scatchard analysis with and without inhibitor). In the present study, inhibitor curves were constructed at six widely different radioligand concentrations with the highest level ∼7-fold ([3H]mazindol) or ∼3-fold ([3H]WIN 35,428) the Kd value close to the factor of 6 used in the computer-simulated examples byTomlinson and Hnatowich (1988) for negative allosterism.
Extrapolation from WIN 35,428 to cocaine binding.
There is an ongoing debate as to whether phenyltropane analogs of cocaine interact with the same site that binds cocaine (Madras et al., 1989;Eshleman et al., 1993; Dersch et al., 1994). This is an important issue because [3H]WIN 35,428 or other iodinated phenyltropane analogs of cocaine are commonly used in binding studies instead of [3H]cocaine because of their higher affinity (Madras et al., 1989; Wall et al., 1993;Reith and Coffey, 1993; Gu et al., 1994; Eshleman et al., 1995; Xu et al., 1995). Classic inhibition and saturation studies, although with the limitations pointed out above, have indicated a competitive interaction between WIN 35,428 and cocaine (Reith and Selmeci, 1992; Reith et al., 1992). In addition, studies on structure-activity relationships within the cocaine and phenyltropane family of compounds have shown a generally similar impact on dopamine transporter activity (Carroll et al., 1992). However, as pointed out by the group of Rothman et al.(Dersch et al., 1994), different ligands may be binding to slightly different domains on the dopamine transporter; for instance, nomifensin inhibits [3H]mazindol binding with aKi value of 24 nM, but [3H]GBR 12935 binding to the dopamine transporter with aKi value of 236 nM, as determined under the same conditions. In contrast, across the four radioligands studied by the group of Rothman under identical conditions, [3H]mazindol, [3H]GBR 12935, [3H]BTCP and [125I]RTI-55, theKi values for WIN 35,428 were in the same range, (108, 44, 36 and 45 nM, respectively), as were theKi values for cocaine (767, 660, 689 and 341 nM) (Dersch et al., 1994; Rothman et al., 1994). This suggests that the binding domains for these radioligands, including [3H]mazindol, overlap in comparable way with the binding domains for WIN 35,428 and cocaine. In a recent study, we observed a similar pH sensitivity of the binding of cocaine methiodide and WIN 35,428 to the dopamine transporter, and we ruled out the implication of varying concentrations of protonated and neutral ligand as a function of pH (Xu and Reith, 1996), which is also consonant with the involvement of a common, pH-sensitive domain in cocaine and WIN 35,428 binding. Furthermore, other approaches, such as protection against N-ethylmaleimide-induced alkylation, have not provided evidence in favor of different binding domains for cocaine and WIN 35,428 (Xuet al., 1997) whereas the same technique suggests differences between blocker and substrate domains (Reith et al., 1996; Xu et al., 1997).
Transporter mutants and chimeras.
If, indeed, different binding domains are involved in WIN 35,428 and mazindol binding, one would expect certain discrete changes in the coding sequence for the dopamine transporter protein to be ineffective on [3H]WIN 35,428 binding but at the same time reduce the potency of mazindol in inhibiting [3H]WIN 35,428 binding. So far, this type of detailed information has not been obtained in the very recent mutagenesis and chimera studies. Evidence for a role of transmembrane domains, as opposed to the large loop between the third and fourth transmembrane domain, or the amino- and carboxyl-terminal tail in the interaction with compounds has been advanced along with different regions involved in substrate and blocker binding in general (Kitayamaet al., 1992; Giros et al., 1994; Buck and Amara, 1994), but further distinctions between certain blockers, such as cocaine and mazindol, must be addressed in future studies.
Effects of cocaine/WIN 35,428 and mazindol on dopamine transport.
Although inhibition of [3H]dopamine translocation into striatal synaptosomes by various uptake blockers, including cocaine and mazindol, has been described to be of a competitive nature (Richelson and Pfenning, 1984; Shank et al., 1987; Krueger, 1990), detailed mechanistic studies are lacking. In recent studies using rotating disk voltammetry for the measurement of dopamine uptake into striatal suspensions, the group of Schenk reported a difference in the mode of interaction between dopamine transport and cocaine, involving an uncompetitive action at Na+ binding sites (McElvain and Schenk, 1992), and that for mazindol, involving competition for dopamine recognition (Meiergerd and Schenk, 1994). These uptake results taken together may not necessarily present a discrepancy from the binding data because binding studies do not address effects of cocaine beyond the recognition step that affect dopamine translocation. However, if we accept that cocaine inhibits dopamine uptake through a mechanism unrelated to interference with dopamine recognition (i.e., through interaction with Na+ binding sites), and if we accept the large body of evidence implicating binding sites for [3H]cocaine and tritiated or iodinated cocaine analog in dopamine uptake inhibition (Ritz et al., 1987; Xu et al., 1995), it follows that the latter binding sites are more closely related to Na+ binding than to dopamine recognition. Meiergerd and Schenk postulate mutually interacting sites at which cocaine and mazindol interact to inhibit uptake (Meiergerd and Schenk, 1994), although the mathematical underpinnings regarding the mutual nature have not been made clear. Thus, the dissociation constant for mazindol in binding to the transporter cocaine complex is estimated to be 345 nM (Meiergerd and Schenk, 1994), but the same constant for cocaine in binding to the transporter-mazindol complex is not presented.
Inspection of the model of Deves (see Meiergerd and Schenk, 1994), used as the basis for the equations by Meiergerd and Schenk (1994), shows that it is implicitly assumed in the derivations that the dissociation constant for cocaine in binding to the transporter mazindol complex is greater than the constant for cocaine binding to the transporter alone by the same factor that increases the dissociation constant for mazindol binding to the transporter cocaine complex compared with mazindol binding by itself. Therefore, the mutual interaction is of the same negative allosteric type as modeled for L and I in our hyperbolic competitive model (equation 1). The observation of Meiergerd and Schenk (1994) that mazindol inhibits dopamine uptake competitively in contrast to the uncompetitive mechanism observed for cocaine is consistent with their suggestion that mazindol interferes with dopamine recognition, whereas cocaine interacts at a distinct site (Na+ binding site), reducing intramembrane transporter turnover. These findings might be related to the short second-time scale of the transport measurements. However, recent voltammetric studies by the group of Wightman, also on a second-time scale, support a competitive instead of uncompetitive mechanism of inhibition for cocaine in the striatum bothin vivo (May et al., 1988) and in vitro (Jones et al., 1995).
Perhaps our current models regarding substrate translocation and blocker action are simplifications, leading to divergent conclusions because important factors (varying between studies) have not yet been taken into account. Be that as it may, the present results suggest that complexities in blocker action, involving separate/interacting sites, are more likely to be found at the level of substrate translocation and transporter reorientation than at the level of recognition.
Acknowledgments
We would like to thank Dr. Aaron Janowsky, Dr. Amy Eshleman and Dr. Kim A. Neve (Oregon Health Sciences University, Portland, OR) for providing the C6 glioma cells expressing the human dopamine transporter.
Footnotes
-
Send reprint requests to: Cen Xu, Ph.D., Department of Biomedical and Therapeutic Sciences, University of Illinois College of Medicine, Box 1649, Peoria IL 61656. E-mail: cen.xu{at}uic.edu
-
↵1 This work was supported by National Institute on Drug Abuse Grant DA-08379 (M. E. A. R.).
- Abbreviations:
- WIN 35
- 428, 2β-carbomethoxy-3β-(4-fluorophenyl)tropane
- Received October 18, 1996.
- Accepted April 22, 1997.
- The American Society for Pharmacology and Experimental Therapeutics