Discussion

Nefazodone, venlafaxine, fluvoxamine and mirtazapine have all recently been introduced in the United States for clinical use. These compounds are inhibitors of monoamine transporters, with the exception of mirtazapine, which appears to be primarily analpha-2 antagonist at auto- and heteroreceptors as well as a potent 5-HT_2A and 5-HT₃antagonist (de Boer and Ruigt, 1995). Unlike venlafaxine and fluvoxamine, which only have high affinity for monoamine transporters, nefazodone possesses potent 5-HT_2A receptor antagonism as well as monoamine transporter antagonist properties (Taylor et al., 1995; Owens et al., 1995). Several years have passed since the comprehensive studies of Richelson and colleagues (Bolden-Watson and Richelson, 1993; Cusack et al., 1994) appeared in which the receptor binding profile of several antidepressants and metabolites were examined. Many earlier studies either focused on a single receptor/transporter or did not include active metabolites for many of the compounds. Moreover, many studies reported IC₅₀ inhibition values which highly depended on both the radioligand used and assay conditions, and were difficult to compare across laboratories. Thus, we examined in detail the binding profile of several antidepressants and their metabolites that are monoamine transporter antagonists with particular attention paid to nefazodone (fig. 1).

Antagonism/inhibition of the SERT was established by competition for [³H]citalopram from either rat frontal cortical membranes or from a HEK-293 cell line stably transfected with the human SERT or antagonism of [³H]5-HT transport into intact HEK-293 cells expressing the human SERT. [³H]Citalopram possesses several advantages compared with other ligands for labeling the SERT including the greatest selectivity, higher specific activity and an affinity which, unlike [³H]paroxetine, provides sufficient signal without depleting free ligand concentration at typical assay volumes and protein amounts (D’Amato et al., 1987; Owenset al., 1996). Although it is not known for certain, citalopram and paroxetine probably label the site near or at the site which serotonin itself occupies for transport (Barker et al., 1994; Barker and Blakely, 1996). Nevertheless, actual inhibition of [³H]monoamine transport may represent the best measure of transporter antagonism.

As shown in table 3 and figure 2, A to C, only the nefazodone metabolite, triazoledione, did not possess any affinity for the SERT. Only moderate affinity for the SERT was observed for nefazodone and its metabolites. Previously, both nefazodone and hydroxynefazodone have demonstrated K_i values between 137 and 181 nmol/l for inhibition of rat [³H]5-HT transport (Bolden-Watson and Richelson, 1993; Taylor et al., 1995). Moreover, serum levels observed in rats which significantly, but not completely, inhibit 5-HT transport are consistent with those obtained with oral nefazodone dosages of 300 to 500 mg/day in humans which demonstrated clinical antidepressant efficacy (Owens et al., 1995; Robinson et al., 1996).

The potency of the various antidepressants at the rat SERT agrees very well with those of other studies which typically included only a few compounds (D’Amato et al., 1987; Plenge and Mellerup, 1991;Cheetham et al., 1993). Unlike, sertraline, amitriptyline and imipramine, the desmethyl metabolites of fluoxetine and venlafaxine had potency similar to their parent compounds. Desmethylsertraline still retains high potency and accumulates to 1.6- to 2.1-fold higher levels than sertraline in plasma, but in vivo data suggest that it may not contribute significantly to inhibition of 5-HT transport clinically (Sprouse et al., 1996). This conclusion may be born out in the finding of significantly reduced potencyversus sertraline in 5-HT transport via the human SERT.

The affinity of the various compounds for displacing [³H]citalopram from the rat and human versions of the SERT were very similar and highly correlated (tables 3 and 4, fig. 4A). The exceptions are the clearly higher potencies of the TCAs amitriptyline, nortriptyline, imipramine, desipramine and chloroimipramine for the human SERT. The increased potency of the tricyclics at the human SERT compared with the rat SERT agrees very well with the findings of Barker and colleagues (Barker et al., 1994; Barker and Blakely, 1996) who used chimeric rat and human SERTs. This property of the tricyclics appears to be attributed to a region near putative transmembrane domain 12 of the human SERT which imparts some species (human) preference for TCAs. Inhibition of [³H]5-HT transport was also highly correlated with the ability of individual compounds to displace [³H]citalopram (tables 3 and 4, fig. 4B). These results agree very well with those of Bolden-Watson and Richelson (1993) and Cheetham et al. (1993).

[³H]Nisoxetine was used to label rat and human NETs. Nisoxetine is 400- and 1000-fold more potent in binding to the NET than the DAT and SERT, respectively. The binding is saturable and Na⁺-dependent to a single class of binding sites (Tejani-Butt, 1992). As shown in table 3 and figure 2, D and E, other than the TCAs, paroxetine was the only other compound possessing moderately high affinity. Even in the face of its high relative SERT selectivity (fig. 3), preliminary studies with serum concentrations of paroxetine similar to those used to treat panic disorder antagonize the NET in rats (M. J. Owens, D. L. Knight and C. B. Nemeroff, unpublished observations). Although nefazodone and trazodone possess similar SERT affinity, trazodone is inactive at the NET suggesting that important structural requirements for NET activity are found in either the phenoxyethyl substituent or the 5-substituent of the triazole moiety, or both (fig. 1).

Our findings for the inhibition of [³H]NE transport in cells expressing the human NET are very similar to those observed for inhibition of [³H]5-HT transport by the human SERT in that the rank order of potency was similar to that observed for radioligand binding (table 3, figs. 2, D and E, and 4D). Of the same drugs tested, our calculated K_ivalues were identical with those reported by Pacholczyk et al. (1991) for the transfected human NET.

As mentioned above, absolute affinity was highly correlated among the compounds for the rat and human NET and between the use of [³H]nisoxetine and [³H]NE (table 3, fig. 4, C and D). Indeed, this correlation was even stronger than that observed for the SERTs because of the similar potencies of the TCAs for rat and human NETs. These data suggest that rat tissue does provide useful data regarding the potency of various compounds for the human NET and that this can be reflected in radioligand binding studies which are easier to perform.

The relative selectivities for the SERT and NET are shown in figure 3. The rank order of selectivity varies slightly depending on the assay method. Desipramine and nortriptyline are clearly the most NET selective of the compounds tested, and sertraline and citalopram are the most SERT selective. Although relatively weak antagonists of the SERT and NET, nefazodone and hydroxynefazodone are the closest to being called “dual uptake inhibitors.” Our data show that venlafaxine andO-desmethylvenlafaxine have, at a minimum, a >15-fold selectivity for the SERT. These selectivities are relative, and with escalating dosages and increased free drug concentrations in serum, even relatively selective drugs can begin to, or fully, bind to other transporters/receptors.

Amitriptyline, nefazodone and hydroxynefazodone were potent at displacing [³H]ketanserin binding from the rat cortical 5-HT_2A receptor (table 3). Trazodone also displayed significant potency (K_i = 20 nmol/l). These data agree closely with data reported previously (Wanderet al., 1986; Seeman, 1993; Cusack et al., 1994). Our results from rat tissue are highly correlated with those of Cusacket al. (1994) who used human cortex (table 5), although our antidepressants displayed somewhat higher affinity in all instances. Once again, we believe this could be related to the method of serial dilution and/or species differences in absolute affinity.

The phenylpiperazine derivatives possessed moderate affinity for the rat hippocampal 5-HT_1A receptor (table 3).mCPP is thought to act as an agonist, whereas it is not known whether the parent drugs are antagonists or agonists, although antagonism is more likely. It has been suggested that combined with the potent 5-HT_2A antagonism, these compounds do augment 5-HT_1A-mediated function (Taylor et al., 1995). These findings agree with those of Richelson and colleagues (Wander et al., 1986; Cusack et al., 1994) and those reported in Seeman (1993). Our data in rat hippocampus was highly correlated with that in human tissue (table 5).

As previously reported by the manufacturer (Paxil, package insert), paroxetine possesses moderately high affinity for muscarinic receptors similar to that observed for the TCA desipramine (table 3). Once again, our data in rats are highly correlated with data observed in humans (Stanton et al., 1993; Cusack et al., 1994) and in Seeman (1993).

As noted under “Methods,” H₁ receptors in guinea pig brain display closer pharmacology to the human H₁ receptor than do rat H₁receptors. Amitriptyline was the most potent drug tested. However, trazodone, nefazodone, hydroxynefazodone and triazoledione all displayed moderately high potency (K_ivalues < 30 nmol/l). These findings are in sharp contrast to those reported by Cusack et al. (1994) who reportedK_i values for nefazodone and trazodone as 24,000 and 1,100 nmol/l, respectively. This discrepancy may be explained by our findings of loss of drug accompanying serial dilution of the drugs (see below) and species differences for the binding affinity of these two drugs (see “Results”). Even with these discrepancies, our data in guinea pig brain was highly correlated with the data reported by Cusack et al. (1994) in human cortex (table 5).

Previous work has provided substantial evidence that rat corticalalpha adrenoceptors have pharmacological characteristics that are different from their human counterparts (Ruffolo et al., 1991). Therefore, we used human cortical tissue for our studies of antidepressant alpha-1 and alpha-2 affinities. For each drug tested, we observed that they possessed a higher affinity for alpha-1 receptors than foralpha-2 receptors (table 3). The high potency displayed by nefazodone and hydroxynefazodone is surprising because the former is reportedly associated with significantly less orthostasis when compared with amitriptyline. Nefazodone and hydroxynefazodone were the only drugs tested that had moderate affinities (<100 nmol/l) for thealpha-2 receptor. As expected, because of the use of the same species tissues, both our alpha-1 andalpha-2 data were highly correlated with the data reported by Cusack et al. (1994). However, they were not any greater than that observed with the 5-HT_2A, 5-HT_1A or muscarinic receptors in which we used rat tissue (table 5).

Although we do not know of any studies of structure-activity relationships between the compounds studied here, the strong positive correlations observed between the affinity of individual compounds for the 5-HT_1A, 5-HT_2A, H₁,alpha-1 and alpha-2 receptors suggests that these individual drugs possess some structural characteristics that determine affinity for several G protein-coupled receptors, the only exception being the muscarinic receptors. Standard hydropathy analyses of the primary sequences for these receptors found that the N-terminal region of the rat m₁ receptor, which contributes most of the muscarinic binding in rat cortex, possesses structural features significantly different from the other receptors (Arnold J. Mandell, personal communication). This portion of the m₁receptor does not appear to be important in binding of “muscarinic” ligands (Brann et al., 1993).

Nefazodone is an effective antidepressant that is marketed as a serotonergic modulating agent that possesses a favorable side effect profile compared with the structurally related antidepressant, trazodone, and compared with the TCAs (Taylor et al., 1995). Unlike the SERT selective antagonists, nefazodone neither produces sexual dysfunction nor alters sleep architecture. Pharmacokinetic analyses find that, at steady-state concentrations, the molar hydroxynefazodone AUC is approximately 35% that of nefazodone. The molar AUC of mCPP is approximately 13% of that of nefazodone and that of the triazoledione tautomer is 160% that of nefazodone (Kaul et al., 1995; Mayol et al., 1994). Although nefazodone is not particularly potent in vitro at antagonizing the SERT and NET, because of the high circulating plasma concentrations of drug, it can produce some transporter inhibition in vivo (Hemrick-Leucke et al., 1994; Owens et al., 1995). Based on metabolic patterns and affinity, the three major metabolites of nefazodone are unlikely to contribute to any transporter inhibition in vivo. Nefazodone and hydroxynefazodone have similar affinity for all the other receptors tested and include potent affinity for the 5-HT_2A site which is thought to represent an important, although not well understood, aspect of its clinical effectiveness (Taylor et al., 1995). These two drugs are also quite potent at the alpha-1 receptor, which is not consistent with the lack of orthostatic side effects and is one instance in which in vitro binding affinity may not accurately predict potential side effects. The relatively potent affinity of nefazodone, hydroxynefazodone and triazoledione at the H₁ receptor are consistent with the sedative properties of nefazodone. Nefazodone and hydroxynefazodone possess moderate affinity for the alpha-2 receptor. One could speculate that this may allow nefazodone to act, in part, in a manner similar to the new antidepressant mirtazapine on alpha-2 auto- and heteroreceptors. Additionally, the moderately high affinity for 5-HT_1A receptors suggests that nefazodone may have some inherent properties that mimic the strategy being utilized by the use of pindolol augmentation as a means to block 5-HT_1A somatodendritic autoreceptors and hasten clinical response and/or convert antidepressant nonresponders to responders. Although trazodone shows a striking similarity to nefazodone in vitro, with the exception of lack of NET antagonism, trazodone rarely produces priapism, whereas nefazodone does not. The mechanism of this difference is not known or discernible from the present binding studies.

The TCAs exhibited a binding profile very similar to that reported previously. Thus, amitriptyline, imipramine, nortriptyline and desipramine showed high affinity for the SERT, particularly the human version, and for the NET in which the secondary amines were more potent. In agreement with previous data, the TCAs had high affinity for the H₁, alpha-1 and muscarinic receptors, which correlates well with their known side effect pattern of sedation, orthostatic hypotension, dry mouth, constipation and tachycardia. One could also speculate that amitriptyline (table 3) and nortriptyline, but not imipramine or desipramine (Cusack et al., 1994), may also act therapeutically via5-HT_2A antagonism.

Venlafaxine, paroxetine, sertraline, citalopram, fluoxetine, fluvoxamine and their various metabolites are all potent antagonists of the SERT. Although marketed as a “dual uptake inhibitor” (Effexor, package insert; Muth et al., 1986), venlafaxine andO-desmethylvenalfaxine are not potent NET antagonistsin vitro, although they do show activity in vivo. This may be explained by the fact that free drug concentrationsin vivo may be relatively high because venlafaxine shows considerably less plasma protein binding (∼20–25% bound) than any of the other compounds and, therefore, likely antagonizes the NET as well as the SERT in vivo. Of the other non-TCAs tested, only paroxetine showed moderately high affinity for the NET, although it was still 2 to 3 orders of magnitude more potent at the SERT. However, as noted earlier, we have preliminary evidence that paroxetine administration in the high therapeutic range may affect the NETin vivo.

In general, the SERT selective antagonists were devoid of any meaningful potency at the various other receptors we examined with two exceptions. Paroxetine has moderately high potency for both muscarinic receptors (K_i = 41 nmol/l) and the NET as noted above. This former finding is undoubtedly responsible for the dry mouth and occasional blurry vision observed with paroxetine use but it may reduce the frequency of diarrhea and loose stools that accompany the use of other SERT antagonists. The second exception is the moderately high potency of sertraline for alpha-1 receptors (K_i = 36 nmol/l). Although one might predict significant orthostasis as a result, this has not been observed clinically, perhaps because concentrations necessary for adequate SERT antagonism are considerably smaller than those needed foralpha-1 blockade.

The curves shown in figure 5 represent the curves we initially observed for several drugs in many different assays. Steep slopes are generally a sign of positive cooperativity which is observed with enzyme kinetics, but not typically in radioligand assays. We considered whether this was a solubility issue; however, an incomplete dissolution of drug would indeed shift the curve to the right but would not change the shape of the curve as observed here. Because these drugs are all weak bases, even at the physiological pH of the buffers, most of the drug would be in an ionized form and likely more soluble in the aqueous buffer than as the free base. Thus, the high degree of ionization produced by the 5 mmol/l HCl dilutions was probably not important for drug dissolution. All the compounds, with the exception of desmethylsertraline and mazindol, were easily solubilized in the 50% ethanol:50% 5 mmol/l HCl solution at 1 mg/ml. Even when using assay buffer alone to perform serial dilutions, the dilution from 1 mg/ml to the most concentrated dilution for the assay was 200 to 250 μg drug/ml assay buffer and resulted in apparently complete solubility. We further confirmed the fact that the drug was fully dissolved and not a particulate suspension by measuring drug levels directly from the serial dilution tubes where we observed a significant loss of drug (table 6 and fig. 6). We further observed that addition to the assay buffer serial dilutions of 25 μl of 1 mol/l HCl directly before HPLC analysis resulted in a significant, but not complete, recovery of drug that was observed to be “missing” in the dilutions of assay buffer alone (data not shown).

The apparent loss of potency and the steep competition curves could plausibly be explained by the presence of a saturable nonspecific binding site not affected by the silanization process on the walls of the glass tubes. Although this is only speculative, this would mean that at lower drug concentrations all the drug is bound to this unidentified site, which results in no competition for radioligand. At the point where this site becomes saturated (i.e.,significantly higher concentrations of serial drug dilution), there is now a large amount of drug available and considerable competition occurs. Because the data are plotted based on assumed concentrations, a very sharp drop in binding (i.e., steep curve) occurs. Whatever the mechanism, we believe that this may represent one explanation why in almost all instances in the present set of studies, the drugs we tested had higher affinity than those previously reported by others for the same compounds. We also believe that, in the future, unless every compound to be tested is examined comparing assay bufferversus dilute acid serial dilution, that serial dilutions for all drugs that are weak bases should be performed in dilute acid rather than assay buffer.