Autoregulation of Serotonin Neurons: Role in Antidepressant Drug Action

1. The Caudal Linear Nucleus.

The caudal linear nucleus (CLN), described in primates by Törk (1990) and Azmitia (see Azmitia, 1986; Azmitia and Gannon, 1986; Jacobs and Azmitia, 1992; Azmitia and Whitaker-Azmitia, 1995), is the most rostral group of serotonergic neurons in the mesencephalic midline, extending along the rostral boundary of the superior cerebellar decussation. Its ventral limit is defined by the interpeduncular nucleus and dorsally it contacts the dorsal raphe nucleus (DRN) through the gap left between the two medial longitudinal fasciculi. In rats, 5-HT neurons located rostral to the decussation of the superior cerebellar peduncle and above the interpeduncular nucleus have been considered in some cases a rostral extension of the median raphe nucleus (MRN) (Lorez et al., 1978; Parent et al., 1981). However, CLN and MRN neurons do not share common terminal projection fields (Imai et al., 1986) and display dissimilar dendritic morphology (Hornung and Fristchy, 1988). On the other hand, because DRN and CLN 5-HT neurons innervate similar terminal fields (e.g., caudate-putamen in adult rat brain; Imai et al., 1986) and share a common developmental origin (Wallace and Lauder, 1983), an alternative approach proposes that CLN neurons should be considered similar to those in DRN (see Jacobs and Azmitia, 1992).

2. The Dorsal Raphe Nucleus.

The DRN is the largest of the brainstem serotonergic nuclei containing about 50% of 5-HT neurons in the rat CNS (Wiklund and Björklund, 1980; Descarries et al., 1982), 40% in the cat CNS (Wiklund et al., 1981), and 50 to 60% in the human CNS (Baker et al., 1990). Rostrally, the dorsal raphe is bound by the Edingher Westphal nucleus (III) and, caudally, it extends just ventral to the confluence of the fourth ventricle and the cerebral aqueduct (Steinbusch, 1981; Descarries et al., 1982; Imai et al., 1986;Törk, 1990; Jacobs and Azmitia; 1992). In most species, the DRN is composed of several subregions distinguished by their different cell density, morphology, and projections (Azmitia and Gannon, 1986; Baker et al., 1991; Johnson and Ma, 1993): 1) a medial portion, subdivided in turn into dorsomedial and ventromedial components, just below the cerebral aqueduct and surrounding the medial longitudinal fasciculus (MLF), respectively; 2) lateral portions or wings (much more prominent in primates than other mammals due to a lateralization process through phylogeny; Descarries et al., 1982; Azmitia and Gannon, 1986; Baker et al., 1991); and 3) a caudal component. During development (15 days of gestation), 5-HT-immunoreactive cells group themselves into two different clusters, dorsolateral and ventrolateral to the MLF (Wallace and Lauder, 1983). The dorsolateral portion will give rise to the lateral wings of the DRN, whereas the ventrolateral group will split to form the interfascicular portion of DRN and MRN. For this reason, it has been proposed that, in the primate CNS (see Jacobs and Azmitia, 1992; Azmitia and Whitaker-Azmitia, 1995), the interfascicular subregion of the DRN should be best considered an entity along with the MRN. Olzewski and Baxter (1954) had previously defined an anatomical entity, nucleus centralis superior, consisting of three groups of cells: 1) a dorsalis component situated between MLF, 2) a medialis component or MRN, and 3) a lateralis component which includes the 5-HT cells that form the ring around the nucleus reticularis pontis oralis (see Table 1). The absence of anatomical boundaries between the main rostral portion of the DRN (B₇) and the main rostral portion of the MRN (B₈) has also been observed in rats (Descarries et al., 1982) and cats (Wiklund et al., 1981). The main rostral component of the DRN (B₇) also merges, in its rear end, with the caudal component (B₆) of this same nucleus (see Jacobs and Azmitia, 1992; Azmitia and Whitaker-Azmitia, 1995). Furthermore, the latter caudal component is in continuity with the dorsocaudal portion of the MRN (see Jacobs and Azmitia, 1992; Azmitia and Whitaker-Azmitia, 1995) and the lateral wings of the DRN (Lidov and Molliver, 1982). Hence, anatomical proximity between different 5-HT cell clusters provides the ground for a network of homotypic interconnections among 5-HT neurons. Such interconnections constitute in turn the morphological ground for the high degree of cross-talk existing among 5-HT neurons at the somatodendritic level. The major implication underlying this arrangement is that the more closely knit and the larger the number of cells in a functional cluster, the more powerful will be the effect of the group on the target structure toward which it projects (Azmitia, 1986).

3. The Median Raphe Nucleus.

The MRN consists of two distinct parts, i.e., a group of densely packed cells, vertically oriented and concentrated in the midline, and a second group, the paramedian columns, consisting of cells scattered in the periphery of the midline cluster (Köhler and Steinbusch, 1982). The nucleus is situated ventral to the MLF and has a rostrocaudal oblique orientation (Azmitia, 1981; Lidov and Molliver, 1982). In its rostral end, beyond the MLF, it is separated from the DRN in its rostral end by the superior cerebellar peduncle decussation. Its medial portion extends from the mesencephalic interpeduncular nucleus to the trapezoid body in the pons. Laterally, the limits of the nucleus are poorly defined toward the reticular formation (Köhler and Steinbusch, 1982), and in the rat the MRN merges laterally with the 5-HT group in nucleus raphe pontis oralis (Lidov and Molliver, 1982). The MRN forms the second largest cluster of 5-HT neurons in the mammalian CNS (Baker et al., 1990).

4. The Supralemniscal Region.

The last group of 5-HT neurons to be considered corresponds to 5-HT cells in the supralemniscal region. In the rat, 5-HT cells belonging to this group are scattered along the dorsal border of the medial lemniscus and extend their dendrites in between the fiber bundles of the latter (Parent et al., 1981; Steinbusch, 1981). A similarity between supralemniscal 5-HT cells and those of the MRN was first noted by Dahlström and Fuxe (1964)in the rat CNS and has been later confirmed in primates (Hubbard and Di Carlo, 1974). In fact, the cells in the supralemniscal region may be continuous with those of the paramedian columns of the MRN (Azmitia and Gannon, 1986). Unlike rodents in which the supralemniscal cell cluster is predominantly mesencephalic, in humans it is entirely located in the pons (Baker et al., 1991).

1. Sources of Extracellular 5-HT in Rostral Raphe Nuclei

The existence of serotonergic axon terminals in the raphe nuclei of cats and rats has been repeatedly reported (Baraban and Aghajanian, 1981; Chan-Palay, 1982; Descarries et al., 1982; Chazal and Ralston, 1987), but the terminals endowed with synaptic specializations have been found to be consistently in low numbers, ranging from “none” (Baraban and Aghajanian, 1981), “exceedingly small number” (Descarries et al., 1982), to “a few” (Chazal and Ralston, 1987). Furthermore, Baraban and Aghajanian (1981) found 5-HT fibers exclusively in axon bundles rather than in proximity to dendrites or cell bodies, and when 5-HT terminals were observed to make somatodendritic synaptic contacts, these were on non-5-HT neurons (Descarries et al., 1982; Chazal and Ralston, 1987). In the rare cases in which a 5-HT axon terminal was observed in close apposition to a 5-HT cell body, no demonstrable synaptic contact was present (Chazal and Ralston, 1987). Nonsynaptic 5-HT axon terminals are not exclusive to the raphe nuclei; they also exist in projection areas such as the cortex, striatum, and hippocampus (Descarries et al., 1990;Törk, 1990). 5-HT release therefore occurs not only from junctional but also from nonjunctional sites (Descarries et al., 1975,1982, 1990).

The total number of 5-HT axons that reach the rostral raphe nuclei (independent of their synaptic specializations), as well as their origin, is a controversial issue in the literature. In autoradiographic studies of the rat brain, Descarries et al. (1982) reported “only few” [³H]5-HT-labeled axon terminals reaching the DRN; Chan-Palay (1982) found “numerous” such terminals; whereasBaraban and Aghajanian (1981) reported prominent labeling of unmyelinated 5-HT axons. On the other hand, using an immunohistochemical procedure, Brusco et al. (1983) found only few 5-HT axon terminals in rat DRN, whereas Chazal and Ralston (1987) concluded that, in the cat, the latter are “widely distributed and not uncommon”, although “not numerous as compared to non-5-HT terminals”. Furthermore, the fine structure of fibers impinging on the DRN was similar to that of 5-HT fibers found in terminal projection areas (Beaudet and Descarries, 1981), a fact that should be borne in mind when considering whether 5-HT fibers that impinge on the rostral 5-HT nuclei are collaterals or afferents from other 5-HT nuclei. In primates, using immunocytochemistry, Kapadia et al. (1985) have also reported few 5-HT axon terminals in the DRN.

In four of the five above-mentioned studies, the number of 5-HT fibers impinging on the raphe nuclei was found to be scarce. However, the extracellular concentration of 5-HT at the somatodendritic level is twice that observed in projection areas such as the cerebral cortex (e.g., 5.5 and 2.3 nM, respectively; Bel and Artigas, 1992), where the amount of axon terminals is 5.8 × 10⁶varicosities/mm³ (Descarries et al., 1990). It is then difficult to conceive that 5-HT fibers (collaterals or afferents from other nuclei) would be the only source of extracellular 5-HT in the dorsal raphe. An alternative and conceivably important source of extracellular 5-HT in the raphe area would therefore be the soma and dendrites of 5-HT neurons. Indeed, the existence of 5-HT neurons with vesicle-containing dendrites in the cat DRN has been reported (Chazal and Ralston, 1987). These dendrites may be divided into two different types: 1) dendrites in which vesicles were never found to be associated with any membrane specialization, and 2) vesicle-containing dendrites which have synaptic membrane specializations. It was not determined whether the two types of dendrites could belong to the same neuron. It is possible that the different types of dendrites could serve different functions: the ones lacking any junctional specialization could contribute to maintain high extracellular-extrasynaptic somatodendritic concentrations of 5-HT, whereas those bearing specializations could constitute the anatomical basis for dendrodendritic homotypic interactions between 5-HT neurons. In fact, although sometimes vesicle-containing dendrites were found to contact non-5-HT elements, most frequently they were presynaptic to other 5-HT dendrites. Apposition between 5-HT dendrites has also been described in rats (Descarries et al., 1982), rabbits (Felten and Harrigan, 1980), and primates (Kapadia et al., 1985). Synaptic specializations on dendrodendritic contacts were recognized in cats (Chazal and Ralston, 1987), as well as in the primate brain (Kapadia et al., 1985), although 5-HT dendrites containing vesicles were found only in the cat CNS. Despite these vesicle-containing dendrites being present throughout the DRN, they were infrequent as compared to dendrites that did not contain vesicles. Furthermore, 5-HT dendrodendritic synapses constituted a small portion of the total DRN synapses (Chazal and Ralston, 1987). A similar observation has been made in the substantia nigra where only a few of the total dendrodendritic contacts interconnected dopaminergic neurons (Wilson et al., 1977). Still, K⁺-induced depolarization evokes significant dopamine (DA) release from this region (Chéramy et al., 1983). A neurotransmitter-releasing role for 5-HT dendrites was initially proposed by Wang and Aghajanian (1977a, 1978) who observed that inhibition of 5-HT neuron firing following the stimulation of the ascending 5-HT pathway in the ventrotegmental area (VTA) was abolished by the 5-HT synthesis inhibitor p-chlorophenylalanine and restored by 5-hydroxytryptophan. It was also proposed that the latter inhibitory effect could be mediated by recurrent 5-HT axon collaterals. Since their initial reports (Wang and Aghajanian, 1977a, 1978), recurrent inhibition of 5-HT neuron firing activity has been confirmed using intracellular (Park et al., 1982) and extracellular (Piñeyro et al., 1996b) recordings. In the former study, the injection of horseradish peroxidase into a representative 5-HT neuron that showed recurrent inhibition following VTA stimulation revealed a single collateral arising from the neuron’s axon. Although this observation supports the existence of axon collaterals, it is not evident that they constitute the only or even an important anatomical substrate for recurrent inhibition of 5-HT neuron firing activity. Moreover, two major Golgi studies in rat and rabbit (Felten and Cummings, 1979;Dı́az-Cintra et al., 1981) described only a few axon collaterals. Summarizing the above-mentioned evidence, it seems reasonable to conclude that the main source of extracellular 5-HT within the raphe nuclei is of somatodendritic origin.

2. Cell Bodies Found in the 5-HT Nuclei.

Quantitative studies of the total number of 5-HT neurons located in the ascending raphe nuclei indicate that there are about 288,000 in the human brain (Törk, 1990), 33,000 in the cat brain (Wiklund et al., 1981), and 15,200 in the rat brain (see Jacobs and Azmitia, 1992). Moreover, 5-HT neurons represent but a small percentage of the total neuronal population of the raphe nuclei. Using histofluorescence-imaging techniques, Wiklund et al. (1981) reported that in the cat 5-HT neurons constitute 70% of medium-sized cells in the dorsal raphe and 35% of medium-sized neurons of the MRN. This percentage could be lower, serotonergic cells constituting 25 to 50% of the total DRN neuronal population and 20 to 30% of the MRN′s. In the supralemniscal 5-HT cell group (B₉), the percentage of 5-HT neurons appears to be much lower than in the other two nuclei (O’Hearn and Molliver, 1984).

Non-5-HT cells reported in the mesencephalic raphe nuclei include peptidergic and nonpeptidergic neurons. The most numerous non-5-HT perikarya are the peptidergic enkephalin-immunoreactive cells (Uhl et al., 1979; Moss et al., 1981, 1983). Most of enkephalin and 5-HT immunoreactivity in dorsal and median raphe nuclei do not colocalize, thus indicating that enkephalin- and 5-HT-immunopositive cells should be best considered distinct neuronal populations (Tanaka et al., 1993). Other peptides contained in DRN somata include 1) Substance P, abundant in lateral wings (Moss et al., 1983; Magoul et al., 1986) and rostral portion of the nucleus (Ljungdahl et al., 1978). It is worth noting that in bulbospinal neurons Substance P reduces the affinity and increases the density of [³H]5-HT-binding sites (Agnati et al., 1983), indicating that the neuropeptide released from 5-HT neurons may modulate receptor sensitivity to 5-HT; 2) neurotensin, found in cells dorsal to the MLF at mid- and caudal levels of the DRN (Beitz, 1982); 3) neurons positive for vasointestinal polypeptide (VIP), which are few and located just ventral to the aqueduct (Sims et al., 1980; Moss et al., 1983); and 4) somatostatin- and cholecystokinin (CCK)-positive cells, both found within the periaqueductal gray, but few of them lie within the DRN (Vanderhaeghen et al., 1980).

There is also consistent evidence indicating the existence of DA-positive cells in the midbrain raphe nuclei (Hökfelt et al., 1976; Ochi and Shimizu, 1978; Miachon et al., 1984; Trulson et al., 1985; Descarries et al., 1986). Except for Miachon et al. (1984) who reported few tyrosine hydroxylase-positive DA-β-hydroxylase-negative neurons, there is general agreement that DA cell bodies in the midbrain raphe are numerous. Two primary subpopulations of DA neurons have been described: 1) A_10dc, lying on the extreme dorsal border of the rostral half of the DRN, ventral to the cerebral aqueduct and 2) A_10c, occupying the medial aspect of the DRN and extending dorsocaudally from the ventrorostral border of this nucleus, where the cells appear contiguous with those of A₁₀ (Hökfelt et al., 1984). Similar to A₁₀ DA neurons, DA-containing cells in rat DRN project to nucleus accumbens (Stratford and Wirtshafter, 1989) and neostriatum (Descarries et al., 1986) with a low degree of colateralization. This projection pattern differs from that of surrounding 5-HT-immunoreactive neurons which show profuse colateralization and innervate other structures such as septum (de Olmos and Heimer, 1980), prefrontal cortex (O’Hearn and Molliver, 1984; Waterhouse et al., 1986; Imai et al., 1986), and neostriatum (Jacobs et al., 1978; Imai et al., 1986). Stratford and Wirtshafter (1989) have thus suggested that DRN DA cells represent a caudal extension of A₁₀ and hence a completely different population from 5-HT neurons. Furthermore, no colocalization of 5-HT and tyrosine hydroxylase was observed in the raphe nuclei (Vanhatalo et al., 1995). Norepinephrine-positive cells have not been found within the DRN, but are located just caudal to the lateral wing groups at the limits of the locus ceruleus (Grzanna and Molliver, 1980).

The existence of γ-aminobutyric acid (GABA) in midbrain raphe nuclei has been repeatedly reported (Massari et al., 1976; Belin et al., 1979;Gamrani et al., 1979; Nanopoulos et al., 1982; Harandi et al., 1987). In fact, there is as much glutamate decarboxylase activity (a specific marker for GABAergic neurons; Fonnum and Walberg, 1973; Ribak et al., 1976) in DRN as in cerebellar GABA-rich nuclei (Massari et al., 1976). The failure to obtain important changes in γ-aminobutyric acid decarboxylase activity after lesioning afferents to the DRN, as well as the presence in this nucleus of terminals, dendrites, and nerve cell bodies accumulating [³H]GABA (Belin et al., 1979), support the notion that the GABAergic network is predominantly intrinsic to the raphe. GABA-positive cell bodies are less numerous than those for 5-HT (Belin et al., 1979; Harandi et al., 1987), and in one of these studies, 40% of the GABA-positive somata was found to contain 5-HT, whereas 30% of the 5-HT-containing neurons immunostained for GABA (Harandi et al., 1987). Hence, it was proposed that some neurons could be both GABAergic and serotonergic. Coexistence of 5-HT and GABA in midbrain raphe neurons has been repeatedly reported (Nanopoulos et al., 1982; Gamrani et al., 1984; Harandi et al., 1987;Gao et al., 1993); however, the proportion and the nuclei in which colocalization occurs vary among different studies. For example, electron microscopy studies revealed coexistence of both neurotransmitters in the somata of DRN neurons (Harandi et al., 1987). On the other hand, confocal microscopy studies revealed that in DRN and MRN, 5-HT and GABAergic neurons constitute two largely distinct populations, double-labeled neurons being observed only within the raphe magnus, raphe obscurus, and raphe pallidus nuclei (Gao et al., 1993). The latter results have been confirmed by Stamp and Semba (1995)who also observed that in raphe sections only a very small percentage of 5-HT neurons in the medullary raphe nuclei also contain GABA (raphe magnus is the nucleus where the percentage of colocalization was highest and it reached only 3.6%). In the latter study, double-labeled cells were virtually absent in the midbrain raphe nuclei, constituting 0.1 to 0.7% of the total number of cells in DRN, MRN, and the supralemniscal region. Low occurrence of colocalization of 5-HT and GABA has also been found in cultured 5-HT neurons obtained from the ponto-mesencephalic region of neonatal rats (Johnson, 1994a). The use of postnatal rat midbrain-pontine 5-HT-neuron culture has also provided evidence for corelease, and hence colocalization, of 5-HT and glutamate from single 5-HT neurons (Johnson, 1994b). However, the major drawback of the latter technique is that the possibility of glutamatergic function having developed in culture may not be ruled out. Contrary to this last possibility is the fact that coexistence of 5-HT and glutamate in rat and primate brain has been observed in medullo-spinal pathways (Nicholas et al., 1992). 5-HT has also been found to colocalize with the trace amine tryptamine (Dabadie and Geffard, 1993) and with basic fibroblast nerve growth factor (bFGF) in DRN and MRN (Chadi et al., 1993). However, bFGF lacks the signal peptide necessary to be secreted by the classical exocytotic pathway used by neuropeptides (Mignatti et al., 1992) but is released when cells are injured. Rather than as a neurotransmitter, bFGF would then act as an autocrine and paracrine factor which elicits trophic responses. Indeed, a neurotrophic role has been postulated for 5-HT itself (for review, see Jacobs and Azmitia, 1992; Azmitia and Whitaker-Azmitia, 1995). Finally, colocalization of 5-HT and NADPH diaphorase in mesopontine neurons suggest that 5-HT neurons and nitric oxide may be used as neurotransmitters by the same neuron (Johnson and Ma, 1993; Wotherspoon et al., 1994).

1. Efferent Pathways.

The rostral 5-HT nuclei are the main source of 5-HT fibers projecting to telencephalon and diencephalon (Fuxe, 1965; Azmitia and Segal, 1978; Parent et al., 1981; Villar et al., 1987). Although the main contingent of fibers from these nuclei is ascending, they also innervate more sparely numerous brainstem structures (Vertes and Kocsis, 1994), the cerebellar cortex (Waterhouse et al., 1986; Zimny et al., 1988), and the spinal cord (Skagerberg and Björklund, 1985). In the nonhuman primate brain, two main ascending bundles have been described, a dorsal bundle (immediately ventral to the MLF), which receives fibers mainly from lateral wings and the ventromedial portion of the DRN, and a ventral bundle, which receives fibers from the midline DRN and MRN (Schofield and Everitt, 1981, Azmitia and Gannon, 1986). In the human fetus, two ascending axon bundles have also been observed, in the central gray, ventral to the fourth ventricle and the aqueduct ependyma, and between ventromedial and ventrolateral 5-HT groups (Nobin and Bjorklund, 1973). In the rat CNS, two major projection systems to the forebrain have also been described, a transtegmental system, which probably corresponds to the above-mentioned ventral bundle, courses through the midbrain forebrain bundle (MFB) and is the most prominent of the two systems described by Descarries in the rat brain, and a periventricular system, dorsally located along the longitudinal fasciculus of Schütz (Parent et al., 1981). In primates, unlike in rats, the dorsal component (dorsal raphe cortical tract) is much more developed than the MFB system, presumably due to an increase in fibers projecting to the cortex through the dorsal pathway. Moreover, the percentage of myelinated 5-HT fibers in rat MFB is 0.7% of the total immunoreactive 5-HT fibers, whereas it is as much as 25% in primate MFB (Azmitia and Gannon, 1983).

2. Terminal Projection Areas.

Dorsal and median raphe nuclei each innervate specific terminal areas (Bobillier et al., 1975; Azmitia and Segal, 1978; Jacobs et al., 1978), and, in turn, each terminal projection area has its unique topographic representation within the respective nuclei. Labeling studies using wheat germ agglutinin, horseradish peroxidase, or fluorescent dyes have been used to unveil the midbrain raphe projection network. The more rostral portions of the midbrain raphe relate to the basal-ganglia-motor system and caudal areas are more related to the limbic system. Neurons projecting to the striatum occupy the caudal linear nucleus and a rostral portion of the DRN (Jacobs et al., 1978; Imai et al., 1986), whereas those projecting to substantia nigra (Imai et al., 1986) and the motor cortex (O’Hearn and Molliver, 1984; Waterhouse et al., 1986) reside within the rostral portions of the dorsal raphe (Imai et al., 1986). Hippocampus-projecting neurons are situated caudally in the DRN (caudal ventromedial portion and B₆), the MRN, and B₉ (Jacobs et al., 1978; Köhler and Steinbusch, 1982; Imai et al., 1986), similar to neurons projecting to the locus ceruleus (Imai et al., 1986) and to the entorhinal cortex (Köhler and Steinbusch, 1982). The raphe representation of the amygdala, on the other hand, bridges the “basal-ganglia-motor system” and the “limbic system representation” (Jacobs et al., 1978; Imai et al., 1986), possibly as a reflection of the functional diversity of amygdaloid nuclei. The consequence of this topographic arrangement is that the selective activation of a given functional group would simultaneously influence interconnected brain circuits. Neurons projecting to interrelated brain circuits, such as the sensorimotor cortex and cerebellar crus II, the visual cortex and cerebellar paraflocculus (Waterhouse et al., 1986), or to the substantia nigra and caudate-putamen (Imai et al., 1986), the entorhinal cortex and hippocampus (Köhler and Steinbusch, 1982), the trigeminal sensory complex and nucleus accumbens or amygdala (Li et al., 1993), arise from overlapping areas within the different nuclei. Moreover, in these regions, a single neuron may provide a common input to two different but functionally interrelated terminal areas. In the ventromedial portion of the DRN, the same neuron was found to project to at least three different forebrain structures related with the limbic system: septum, medial thalamus, and olfactory cortex (de Olmos and Heimer, 1980). Colateralization provides not only a means for producing concurrent influences on numerous functionally related circuits, but it is also a way of achieving extensive serotonergic innervation from a small number of raphe neurons.

In projection areas, serotonergic axons arising from DRN and MRN raphe have been distinguished on a morphological basis: in the rat brain, dorsal raphe fibers are extremely fine with minute varicosities (less than 1 μm in diameter), whereas those arising from median raphe are distinguished by large spherical varicosities (2–5 μm in diameter;Kosofsky and Molliver, 1987; Mulligan and Törk, 1988). In neocortical areas, dorsal raphe-fine axons have been found to be far more numerous than beaded axons (Kosofsky and Molliver, 1987; Mamounas et al., 1991) and to follow a rostrocaudal pattern of distribution with a greater concentration in more frontal regions. These findings are in agreement with those obtained in retrograde-labeling experiments in which the frontal cortex received twice as many projections from cells in the DRN than those in the parietal and occipital cortex (O’Hearn and Molliver, 1984). Morphological duality has also been reported in the cat cortex (Mulligan and Törk, 1988), but such results were not replicated in the rat brain where immunostained varicosities exhibited similar shape and size irrespective of the cortical region or sector examined (Séguéla et al., 1989).

Over a wide range of doses (2.5–40 mg/kg) and survival times (1 week to 2 months), neurotoxic amphetamine derivatives, such as 3,4-methylenedioxyamphetamine (MDA), 3,4-methylenedioxymetamphetamine (MDMA), and p-chloroamphetamine (PCA), produce degeneration of fine 5-HT axon terminals while sparing beaded 5-HT axons (Mamounas and Molliver, 1988; O’Hearn et al., 1988; Wilson et al., 1989;Mamounas et al., 1991), as well as raphe cell bodies (O’Hearn et al., 1988; Mamounas et al., 1991). The 5-HT-releasing agent fenfluramine produces similar effects (Molliver and Molliver, 1990). These neurotoxins have been used to determine the differential distribution of beaded and fine axons and hence the contribution of DRN and MRN in different projection areas. In the hippocampus and neocortex, different neurotoxins have been reported to produce a regional axon loss/sparing pattern which is coincident with fine and beaded axon distribution.p-chlorophenylalanine (PCPA) or MDA administration caused marked denervation in the parietal and occipital cortices, and a moderate number of axons was spared in the frontal cortex (Mamounas and Molliver, 1988; Mamounas et al., 1991). In the hippocampus, there was a greater density of spared axons, most probably due to the fact that in the hippocampal formation, in comparison to other cortical areas, beaded axons are especially prevalent (Mamounas et al., 1991). An exceptionally large number of axons was spared in stratum radiatum of CA₃, and intact axons were also found in the stratum oriens of CA₃, stratum lacunosum of CA₁, and flanking the granular layer in the dentate gyrus. In contrast, few 5-HT axons remained in the outer molecular layer of the dentate gyrus and stratum oriens and radiatum of CA₁, areas that normally receive DRN innervation (Mamounas et al., 1991). On the other hand, Oleskevich and Descarries (1990) described 5-HT axons projecting to the same layers, but no distinction between fine or beaded fibers was made. This same group has nevertheless reported that 5-HT varicosities with synaptic membrane specializations are slightly larger than their nonjunctional counterparts (Oleskevich et al., 1991). In fact, based on the observation that fine fibers rarely make synaptic contacts while larger varicosities found on beaded axons tend to show distinct synaptic specializations, Törk (1990) proposes that, along with dual morphology and distinct sensitivity to neurotoxins, the existence of axons with and without synaptic contacts supports the idea that the ascending raphe projections form a dual system. It should be noted that the reported percentages of synaptic incidence in terminal areas are very variable: 30 to 80% (of a total of 5.8 × 10⁶ varicosities/mm³) in the rat cortex (Papadopoulos et al., 1987; Séguéla et al., 1989), 3% in the monkey cortex (de Felipe and Jones, 1988), 10 to 15% (of a total of 2.6 × 10⁶varicosities/mm³) in the rat neostriatum (Soghomonian et al., 1987), and 18 to 24% in the hippocampus (of a total of 2.7 × 10⁶varicosities/mm³; Oleskevich et al., 1991).

1. Autoregulation of 5-HT Neuron Firing Activity.

The firing rate of 5-HT neurons is decreased by 5-HT, and this effect is mediated by somatodendritic 5-HT_1A autoreceptors (Aghajanian et al., 1972; Vandermaelen et al., 1986; Blier and de Montigny, 1987). 5-HT and 5-HT_1A agonists inhibit 5-HT firing activity by inducing membrane hyperpolarization which is brought about by a 2-fold mechanism, i.e., by increasing conductance to potassium ions (Aghajanian and Lakoski, 1984; Yoshimura and Higashi, 1985) and by reducing a high-threshold Ca²⁺current (Fig. 1; Penington and Kelly, 1990; Penington and Fox, 1994). In both cases, the response to 5-HT is G protein-mediated via a direct interaction between G proteins and the respective ion channel (Innis and Aghajanian, 1987; Penington et al., 1991; Penington et al., 1993;Penington and Fox, 1994).

Current knowledge of the pharmacological properties of somatodendritic 5-HT_1A autoreceptors is mostly based on electrophysiological studies done in vivo which have assessed the effect of different compounds on the firing activity of 5-HT neurons. Table 3 summarizes these effects. Of the synthetic 5-HT_1A agonists listed therein, lysergic acid diethylamide (LSD) and 8-hydroxy-2-(di-n-propylamino)tetralin (8-OH-DPAT) are among the most effective in producing suppression of firing activity, not only of DRN 5-HT neurons but also that of CA_1/3hippocampus pyramidal neurons. Unlike 8-OH-DPAT which produces maximal activation of postsynaptic 5-HT_1A receptors, most of the drugs that completely inhibit 5-HT neuron firing activity produce partial agonistic effect at postsynaptic 5-HT_1A sites in CA₁ and CA₃ regions of the hippocampus (Yocca and Maayani, 1985; Yocca et al., 1986; Andrade and Nicoll, 1987a). Some of these compounds (e.g., tandospirone and flesinoxan) not only produce a smaller maximal response than 8-OH-DPAT but have also been shown to reduce the inhibitory effect induced by the microiontophoretic application of 5-HT onto CA₃ pyramidal neurons (Godbout et al., 1991; Hadrava et al., 1995). Conversely, several compounds previously thought to act as 5-HT_1Areceptor antagonists at postsynaptic sites have been shown to produce submaximal agonistic effects at presynaptic sites, inhibiting 5-HT neuron firing. These compounds are also best classified as partial agonists. For example, drugs like (+)-WAY 100135, NAN-190, or BMY 7378 block the inhibition induced by 5-HT_1A agonists on forskolin-stimulated adenylyl cyclase activity from hippocampal membranes (Rydelek-Fitzgerald et al., 1990), block the postsynaptic neurochemical, electrophysiological, and behavioral effects of 8-OH-DPAT (Chaput and de Montigny, 1988; Sharp et al., 1990; Routledge et al., 1993; Escandon et al., 1994), but they also decrease 5-HT neuron firing activity (Vandermaelen et al., 1987; Chaput and de Montigny, 1988; Fletcher et al., 1993, Haddjeri and Blier, 1995). The suppression of raphe firing caused by NAN-190 and BMY 7378 is blocked by the β-adrenergic/5-HT_1A antagonist (±)-propranolol (Middlemiss, 1984a; Tricklebank et al., 1985; Greuel and Glaser, 1992). Similarly, the inhibition induced by SDZ 216-235 on 5-HT neuron firing is blocked by the β-adrenergic/5-HT_1A antagonist (−)-tertatolol (Lanfumey et al., 1993; Lejeune et al., 1993), further indicating that these “postsynaptic antagonists” act as partial agonists on presynaptic 5-HT_1A receptors. The suppression of 5-HT neuron firing activity induced by (+)-WAY 100135 may be overcome by pretreatment with the α₁-adrenergic agonist phenylephrine but not by (−)-tertatolol, suggesting that the latter effect is not due to partial agonism on 5-HT_1Areceptors, but rather to the blockade of firing stimulating α₁ adrenoceptors on 5-HT neurons (Lanfumey et al., 1993). Because WAY 100635 and (−)-tertatolol have been consistently shown to block the effect of 5-HT_1Aagonists without affecting 5-HT neuron firing (Jolas et al., 1993;Lanfumey et al., 1993; Lejeune et al., 1993; Prisco et al., 1993;Craven and Grahame-Smith, 1994; Fletcher et al., 1994; Mundey et al., 1994; Table 3), they have been named “silent” antagonists. In spite of its lack of partial agonistic activity on somatodendritic 5-HT_1A receptors, (+)-WAY 100135 does not enter into this category due to its α₁-adrenergic blocking properties (Lanfumey et al., 1993) which induce an observable inhibition of firing in the electrophysiological paradigm.

Two classical hypotheses have been proposed to explain the distinct pharmacological properties of pre- and postsynaptic 5-HT_1A receptors: 1) greater receptor reserve in somatodendritic than in postsynaptic areas (Meller et al., 1990; Yocca, 1990; Greuel and Glaser, 1992; Millan et al., 1992) and 2) different pre- and postsynaptic receptors (see de Montigny and Blier, 1992a,b). The first hypothesis sustains that if no receptor reserve exists, like in the case of the hippocampus (Yocca, 1990; Gozlan et al., 1994), partial agonists will not produce a maximal effect and may block the effect of full agonists. Conversely, if spare receptors are present, at an appropriately high dose, drugs with low intrinsic activity (i.e., partial agonists) may still elicit a maximal response and appear as full agonists. The second hypothesis, suggesting distinct 5-HT_1A receptors at pre- and postsynaptic sites has been proposed to account not only for the differential pre-/postsynaptic activity of partial agonists but also to explain the following observations: 1) different rank order in the effectiveness of 5-HT_1A agonists to inhibit DRN (8-OH-DPAT, gepirone > LSD > 5-HT) versus hippocampus firing activity (5-HT > gepirone > 8-OH-DPAT; Blier and de Montigny, 1987,1990; Chaput and de Montigny, 1988); 2) preferential antagonistic effect of BMY 7378 at postsynaptic sites (Chaput and de Montigny, 1988) and of spiperone at presynaptic receptors (Lum and Piercey, 1988; Blier et al., 1989b, 1993a; Fornal et al., 1994a); and 3) differential effect of long-term administration of 5-HT_1Aagonists which induce desensitization of presynaptic but not of postsynaptic CA₃5-HT_1A-mediated responses (Blier and de Montigny, 1994).

Over the past few years advances in molecular biology, i.e., the expression of G protein-coupled receptors (GPCRs) in surrogate cell systems, have greatly modified the way in which we understand pharmacological properties of GPCRs and consequently the mechanistic drug-receptor models used to simulate drug action (see Kenakin, 1996). The first of such models to consider that R activation by an agonist would result in R-G protein coupling, and subsequent activation of the G protein was the ternary complex model (De Léan et al., 1980). This model has been the standard by which drug-receptor interactions have been explained in the past. More recently, the observed ability of GPCRs for becoming activated independent of the presence of an agonist (Samama et al., 1993; Barker et al., 1994; Chidiac et al., 1994) has guaranteed the appearance of alternative paradigms (extended ternary complex model, Samama et al., 1993; two-state model, Leff, 1995; cubic ternary complex model, Weiss et al., 1996). All models consider that agonist-independent activity is secondary to spontaneous isomerization between the inactive (R) and active (R^*) states of the receptor. Furthermore, according to all of these models, agonists have preferential affinity and stabilize the active conformation of the receptor (R^*). Ligands that stabilize the resting state (R), decreasing the number of spontaneously active receptors, act as inverse agonists, and antagonists would be those drugs that present similar affinities for both, favoring neither state of the receptor. The fundamental consequence of embracing any such paradigms is that affinity, potency, and/or efficacy values should be considered receptor- and system-dependent, the basal R:R^* ratio being a determinant of this dependence. Antagonists, whose effect would remain unmodified by the level of spontaneous activation, would therefore remain as the most reliable tools for receptor classification.

Because none of the drugs producing differential pre- and postsynaptic effects are neutral antagonists (note: spiperone has been recently characterized as an inverse agonist on spontaneously active human 5-HT_1A receptors expressed in Chinese hamster ovary cells; Newman-Tancredi et al., 1997), it cannot be ruled out that differences in efficacy observed in DRN and hippocampus are not due to R:R^* ratios. Furthermore, differential long-term desensitization of presynaptic receptors following long-term administration of 5-HT_1A agonists has been attributed to differential drug efficacy in midbrain and hippocampus (see de Montigny and Blier, 1992a,b). Hence, possible different R:R^* ratios could also account for the latter observation. Finally, differences in the R:R^*ratio may explain a set of drugs being more efficacious in one area than in another, but it does not account for the observation that rank order of agonist efficacy may be different in two regions. The observation of different rank order of efficacy for 5-HT, gepirone, and 8-OH-DPAT in DRN and hippocampus could be due either to the presence of different 5-HT_1A receptor subtypes or to the same 5-HT_1A receptor protein coupling to different G proteins in the two different regions (see Kenakin, 1995 for receptor promiscuity). Paradigms which do not include G proteins in their formulation, such as the two-state model (Leff, 1995), do not account for the latter possibility. A more recent version, the three-state model (Leff et al., 1997), which allows for a single receptor to couple with two different G protein pathways, may be invoked to explain the differences in pharmacology in pre- and postsynaptic 5-HT_1A receptors without the need to propose different receptor subtypes. The coupling of a single receptor to more than one G protein pathway is also addressed by the cubic ternary model. The latter paradigm has the advantage of providing yet another mechanism to explain changes in rank order of efficacy of agonists. The mechanism, known as agonist trafficking of receptor signals (Kenakin, 1995), proposes that different drugs may provoke agonist-specific receptor conformations which would selectively promote G protein coupling in response to activation by different agonists. The possibility of a given conformation being more easily favored in one structure than in another may explain the changes in agonist rank order. Furthermore, the fact that i.c.v. administration of cholera toxin does not alter responsiveness of 5-HT neurons to microiontophoretic applications of 5-HT or 8-OH-DPAT, but reduces that of CA₃ neurons by 90% (Blier et al., 1993b), may also be explained by differential G protein coupling in the two regions.

Extracellular availability of endogenous 5-HT at the cell body level and tonic activation of somatodendritic 5-HT_1Aautoreceptors is yet another factor that may contribute to modify the effects of 5-HT_1A ligands. Thus, Fornal et al. (1994a) have shown that the acute i.v. administration of the 5-HT_1A agonist 8-OH-DPAT or of the partial agonists ipsapirone and buspirone was more effective in inhibiting 5-HT neuron firing activity when cats were inactive (drowsiness) than during active wakefulness (a period of higher neuronal activity, 5-HT neuron depolarization, and hence of 5-HT release). Conversely, the antagonists spiperone and WAY 100635 induced a dose-dependent increase in the firing rate of 5-HT neurons, which was evident during wakefulness but not during sleep (Fornal et al., 1994a,c) when 5-HT neurons are silent, and therefore there is very low depolarization-mediated release of 5-HT to activate somatodendritic 5-HT_1A receptors. Moreover, in anesthetized rats and guinea pigs, in which 5-HT neuron firing activity resembles that of SWS, WAY 100635 had no effect on dorsal raphe neuronal firing (Fletcher et al., 1994; Mundey et al., 1994). It is also worth noting that the 5-HT_1Aantagonistic properties of (+)-WAY 100135 on spontaneous 5-HT neuron firing were not evident in anesthetized rats (Haddjeri and Blier, 1995), in awake freely moving cats (Fornal et al., 1994c), or in an antidromic stimulation paradigm in which 5-HT neuron firing is suppressed by somatodendritic release of 5-HT (Piñeyro et al., 1996b). Yet, at doses similar to the ones used in previous studies (0.5–1 mg/kg i.v.), (+)-WAY 100135 reversed the suppression of 5-HT neuron firing activity induced by the blockade of 5-HT uptake (Arborelius et al., 1995; Hajós et al., 1995). This observation is in keeping with the previous assumption that the action of 5-HT_1A ligands on 5-HT neuron firing activity may differ according to the tonic activation of their somatodendritic autoreceptors by the extracellular concentration of 5-HT. Similarly, S-UH-301, which in the dose range of 0.5 to 4 mg/kg i.v. reduces 5-HT neuron firing, effectively blocked the suppressant effect induced by the systemic administration of selective 5-HT reuptake blocker citalopram (Arborelius et al., 1994, 1995). Moreover, pindolol, which by itself induces no change in 5-HT neuron firing activity at low doses, prevents the inhibitory effect of paroxetine and LSD on this parameter (Romero et al., 1996; Haddjeri et al., 1999).

Pindolol is an antagonist which, in the electrophysiological paradigm, discriminates between presynaptic and postsynaptic hippocampal 5-HT_1A receptors. Like spiperone, pindolol blocks the effect of endogenous 5-HT on presynaptic receptors without interfering with the inhibitory effect induced on CA₃ pyramidal neuron firing activity by the microiontophoretic application of 5-HT (Blier et al., 1993a; Romero et al., 1996). Moreover, pindolol, in contrast with other 5-HT_1A antagonists, was recently shown to be ineffective in blocking the inhibitory effect of 5-HT_1A agonists on the firing activity of CA₁ pyramidal neurons in freely moving rats (Tada et al., 1999). These data do not imply that under certain experimental conditions pindolol cannot act as a 5-HT_1Aantagonist, as shown in the in vitro studies of Corradetti et al. (1998) carried out on hippocampus slices. WAY 100635, on the other hand, has been shown to block the effect of 5-HT on pre- and postsynaptic 5-HT_1A receptors located on CA₃ pyramidal neurons (Haddjeri et al., 1996). Binding studies are consistent with the latter observation, i.e., [³H]WAY 100635 behaves as an antagonist in projection areas such as the hippocampus since guanyl nucleotides fail to influence its K _D andB _max parameters (Khawaja, 1995;Khawaja et al., 1995). The presynaptic selectivity of the other silent antagonist (−)-tertatolol remains to be determined, and, for the time being, pindolol seems the only useful alternative of a presynaptic 5-HT_1A antagonist to be used in clinical trials. Finally, it is also worth mentioning the case of (−)-propranolol, which in spite of its moderate affinity for 5-HT_1A receptors (Pazos and Palacios, 1985; Hoyer and Schoeffter, 1991; Prisco et al., 1993), unlike pindolol, does not block the inhibitory effect of 5-HT on 5-HT neuron firing activity (Sprouse and Aghajanian, 1986; Blier et al., 1989a). Moreover, propranolol does not block the effects of either 5-HT or 8-OH-DPAT on hippocampal neuron firing activity. However, propranolol has been shown to block 5-HT-induced inhibition of forskolin-stimulated adenyl cyclase in hippocampal cells (Dumuis et al., 1988). These differences could be reconciled by assuming that the propranolol-bound-receptor conformation does not interfere with the G_i/K⁺ channel pathway but blocks that of G_i/adenylate cyclase.

The local somatodendritic autoreceptor “short-loop” is only one of the autoregulatory mechanisms of 5-HT neuron firing activity. Several observations suggest that, in fact, the systemic administration of 8-OH-DPAT may regulate 5-HT neuron firing activity by an alternative “long feedback loop”, which also entails 5-HT_1A receptor activation: 1) the effectiveness of microiontophoretic application of 5-HT_1Aagonists to inhibit 5-HT neuron firing activity, but not that of systemic 8-OH-DPAT administration, is reduced following long-term treatment with the 5-HT_1A agonist gepirone and the selective serotonin reuptake inhibitor (SSRI) cericlamine (Blier and de Montigny, 1987; Jolas et al., 1994); 2) the effect of i.v. 8-OH-DPAT on 5-HT neuron firing activity, but not that of its microiontophoretic application onto 5-HT neurons, is increased by short-term lithium treatment (Blier et al., 1987); 3) the inhibitory effect of i.v. administration of 8-OH-DPAT on the firing activity of dorsal raphe serotonergic neurons in rats is attenuated by lesion of the frontal cortex (Ceci et al., 1994); and 4) the local application of pertussis toxin into the dorsal raphe (which inactivates G_i/o-coupled 5-HT_1Aautoreceptors; Innis and Aghajanian, 1987) reduces the effectiveness of intraraphe, but not of systemic 8-OH-DPAT administration, to reduce 5-HT release in terminal projection areas (Romero et al., 1994). Direct evidence for such a long feedback loop being involved in mediating the activity of 5-HT neurons has recently been provided in electrophysiological and microdialysis experiments. The local application of 5-HT_1A agonists in the medial prefrontal cortex produces both a suppression of firing of 5-HT neurons and a decrease in extracellular synaptic level of 5-HT in the dorsal raphe, most likely as a result of the former phenomenon (Hajós and Sharp, 1998; Artigas et al., 1998).

2. Role of Non-5-HT Receptors in the Regulation of 5-HT Neuron Firing Activity.

Autoradiographic and binding studies have documented the presence of different 5-HT-binding sites in the rat raphe nuclei (Waeber et al., 1988; Herrick-Davis and Titeler, 1988;Waeber et al., 1989a; Laporte et al., 1992). However, the role of 5-HT receptors, other than 5-HT_1A in modulating 5-HT neuron firing activity at the cell body level, has not been confirmed. The lack of effect of 1-[3-(trifluoromethyl)phenylpiperazine (TFMPP) and m-chlorophenylpiperazine on the firing activity of 5-HT neurons led Sprouse and Aghajanian (1986, 1987) to conclude that 5-HT_1B receptors were not involved in regulating 5-HT neuron firing activity in the rat brain. The systemic injection and the microiontophoretic application of the preferential 5-HT_2A agonist (±)-2,5-dimethoxy-4-iodoamphetamine (DOI) reduce 5-HT neuron firing frequency. However, this effect could not be blocked by the 5-HT_2A/2C antagonists ketanserin and ritanserin (Wright et al., 1990; Garratt et al., 1991). Moreover, the microiontophoretic application of ketanserin does not block the inhibition of 5-HT neuron firing induced by the neurotransmitter itself. It does, however, reduce the basal firing rates in the majority of 5-HT cells tested (Lakoski and Aghajanian, 1985). These observations, along with the fact that the effectiveness of systemic 8-OH-DPAT but not that of DOI to inhibit the firing activity of 5-HT neurons is decreased following repeated DOI administration (Kidd et al., 1991), suggest that 5-HT₂ receptors are not directly involved in the regulation of 5-HT neuron firing. The role of 5-HT₃ receptors in regulating 5-HT neuron firing activity has also been investigated. The fact that systemic administration of the 5-HT₃ antagonist BRL 46470A does not block the reduction of the dorsal raphe 5-HT neuron firing rate induced by the microiontophoretic application of the 5-HT₃ agonist 2-methyl-5-HT (Haddjeri and Blier, 1995), and that 5-HT neuron firing activity remains unchanged after systemic administration of three different 5-HT₃receptor antagonists, MDL 72222, ICS-205-930, and ondansetron (Adrien et al., 1992), indicate that 5-HT₃ receptors do not contribute to the regulation of 5-HT neuron firing activity. 8-OH-DPAT has often been used as the “gold standard” for defining 5-HT_1A receptors, but there is evidence that it also has reasonable good affinity at the rat (Shen et al., 1993) and guinea pig (To et al., 1995) 5-HT₇ receptor. Moreover, 8-OH-DPAT has also been identified as the most potent agonist at a cyclase-linked receptor not yet fully characterized (Becker et al., 1992), and it also binds to the serotonin transporter (SERT) (Schoemaker and Langer, 1988; Alexander and Wood, 1988; Ieni and Meyerson, 1988). However, the fact that methiothepin blocks 5-HT₇-mediated responses (Terron, 1997; Kitazawa et al., 1998) but not the effect of i.v. 8-OH-DPAT (Blier et al., 1989b) suggests that 5-HT₇ receptors are not involved in 8-OH-DPAT actions on 5-HT neuron firing activity. Furthermore, the dose at which 8-OH-DPAT interferes with 5-HT uptake is two to three orders of magnitude higher than the one currently used in electrophysiological experiments to assess drug effect on neuronal firing activity. Also, without excluding the role of other 5-HT receptors, the extensive list of 5-HT_1A compounds in Table 3 confirms the role of the latter receptor in regulating 5-HT neuron firing.

3. Heteroregulation of 5-HT Neuron Firing Activity.

N-methyl-d-aspartate (NMDA) receptors may elicit excitatory postsynaptic potentials (EPSPs) (Pan and Williams 1989; Pinnock, 1992; Johnson, 1994a) and increase the firing activity of 5-HT neurons (Alojado et al., 1994) in vitro, but they do not seem to maintain a tonic activation of 5-HT neuron firing in vivo (Levine and Jacobs, 1992). However, glutamate does mediate the increase in firing activity observed following presentation of phasic auditory stimuli (Levine and Jacobs, 1992). In turn, excitatory amino acid (EAA) release in the dorsal raphe is negatively regulated by κ-opioid receptors (Pinnock, 1992). Rather unexpectedly, the systemic administration of the NMDA channel blocker (+)-MK-801 has been shown to facilitate the electrical activity of 5-HT neurons in the DRN. Such an observation could be explained by assuming that the channel blocker reduces the facilitation of an inhibitory influence. Microiontophoretic application of GABA onto dorsal raphe 5-HT neurons produces an inhibition of their firing rate (Gallager and Aghajanian, 1976; Levine and Jacobs, 1992). Also, the GABA blocker picrotoxin reduces the suppressant effect on 5-HT neuron activity caused by habenula and pontine reticular formation stimulation (Wang et al., 1976; Wang and Aghajanian, 1977; Stern et al., 1981). In freely moving animals, microiontophoretic application of bicuculline produces a significant increase of 5-HT neuron firing activity during SWS, but not during REM or quiet waking, indicating that GABAergic input is state-dependent and not tonic (Levine and Jacobs, 1992). In anesthetized rats, microiontophoretic and systemic administration of α adrenoceptor antagonists as well as 6-hydroxydopamine pretreatment suppress 5-HT neuron firing activity, suggesting a tonic facilitatory role for noradrenaline on dorsal raphe 5-HT neurons (Baraban and Aghajanian, 1980). Furthermore, activation of α₂-adrenergic autoreceptors by clonidine decreases norepinephrine (NE) output and suppresses the firing activity of 5-HT neurons (Clement et al., 1992;Haddjeri et al., 1996a). In contrast, systemic administration of α adrenoceptor antagonists in the awake freely moving animals produced no change in 5-HT neuron firing activity (Heym et al., 1981). Unlike EAAs, NE does not elicit EPSPs on 5-HT neurons, rather it suppresses the voltage-dependent K⁺ current I_A (Fig. 1; Aghajanian, 1985), leading to a more rapid activation of the low-threshold inward calcium current that triggers the spike at the end of the pacemaker cycle of these neurons. Activation of CCK_A receptors located on 5-HT neurons also stimulates their firing activity (Boden et al., 1991), as does bombesin via the stimulation of neuromedin B receptors (Pinnock et al., 1994).

1. Neurotransmitter Release and Its Regulation: Cellular and Molecular Aspects.

Exocytosis is the main mechanism used by neuronal cells for releasing neurotransmitter molecules. By this process, synaptic vesicles fuse with the plasma membrane and the neurotransmitter(s) contained within them reach(es) the synaptic cleft. Exocytosis is triggered by cell depolarization. Depolarization induces opening of voltage-sensitive calcium channels and subsequent Ca²⁺ entry. Achieving localized concentrations of 10 to 100 μM around the open channel (Smith and Augustine, 1988), the intracellular Ca²⁺ increase constitutes the major coupling signal that links depolarization and exocytotic secretion (reviewed by Burgoyne and Cheek, 1995). One of the main characteristics of neurotransmitter release is its high speed, the complete cycle being achieved in hundreds of milliseconds. This is apparently due to the fact that secretory vesicles are already docked to the plasma membrane so that when Ca²⁺ entry takes place, vesicles in a close vicinity of activated calcium channels will immediately void their content into the synaptic cleft by formation of a fusion pore (see Burgoyne and Cheek, 1995). Docking and fusing of the vesicles to the plasmalemma is achieved by Ca²⁺-sensitive vesicle membrane proteins (Augustine et al., 1985; Smith and Augustine, 1988; Leveque et al., 1992). One of these proteins, synapsin I, in an unphosphorylated state, fixes secretory granules to the cytoskeleton. Once it undergoes Ca²⁺-calmodulin/cAMP-dependent phosphorylation, it releases the vesicles from the cytoskeletal network and allows them to move to the presynaptic membrane (Valtorta et al., 1992) where they will be ultimately docked and voided to the extracellular space. The cytoskeleton is not likely to be involved in a first burst of release, which usually empties already docked vesicles, yet releasing the bound granules from the actin network will facilitate their subsequent recruitment by the plasma membrane in preparation for the arrival of the next axon potential (see Burgoyne and Cheek, 1995).

The amount of neurotransmitter released is subject to receptor-dependent regulation that may theoretically occur at any stage in the release process described above. However, because most of the exocytotic steps are regulated by Ca²⁺ influx in an exponential manner (Augustine and Charlton, 1986), relatively small changes in its influx will be expected to produce profound changes in neurotransmitter release. Serotonin has been shown to enhance release by increasing voltage-activated Ca²⁺ currents in postsynaptic neurons in the substantia nigra (pars compacta) and spinal cord (Nedegaard et al., 1988; Berger and Takahashi, 1990). In contrast, the main autoregulatory effect of 5-HT on its own release is inhibitory. An inhibition in Ca²⁺ influx may be produced either by a direct effect on Ca²⁺channels (shift in the voltage of activation of the channel; Bean, 1989) or by reducing the depolarization time during which Ca²⁺ enters the cell. This reduction in depolarization duration is frequently achieved by speeding the activation of K⁺ or Cl⁻currents that end depolarization linked to the action potential (Berlardetti and Siegelbaum, 1988; Brezina and Erxleben, 1988). Serotonin, acting on 5-HT_1A autoreceptors, has been shown not only to reduce a high threshold Ca²⁺ current (Penington and Kelly, 1990;Penington et al., 1991; Penington and Fox, 1994), but also to increase conductance to potassium ions (Aghajanian and Lakoski, 1984; Yoshimura and Higashi, 1985). It is important to realize that the experiments assessing the cellular and molecular mechanisms of the regulation of neurotransmitter release have been performed on the cell body, and that there is no direct electrophysiological information on the mechanism by which terminal 5-HT autoreceptors modify Ca²⁺influx. Linkage via G protein of neurotransmitter receptors to Ca²⁺ or K⁺ channels is widely used in nature as a means of inhibiting neurotransmitter release (see Miller, 1990; Anwyl, 1991). It could therefore be a possible mechanism by which 5-HT autoreceptors regulate release from 5-HT terminals. On the other hand, agents like cAMP which promote Ca²⁺-independent phosphorylation of synapsin I (Bähler and Greengard, 1987) might induce regulation of neurotransmitter release without modifying Ca²⁺influx to the cell. Indeed, this would be in keeping with not only the observation that adenylate cyclase inhibition by 5-HT_1B/1D receptor activation reduces 5-HT release, but also with the fact that cAMP analogs and stimulation of adenylate cyclase by forskolin increase 5-HT release from brain slices and neuroectodermal cell cultures (Schlicker et al., 1987; Tamir et al., 1990).

Neurotransmitter release from neuronal structures has been classically studied in terms of exocytosis, and it is only recently that carrier-mediated release has been considered as a functionally acceptable mechanism for increasing extracellular concentration of a wide number of neurotransmitters including 5-HT (Levi and Raiteri, 1993). Unlike exocytosis, carrier mediated-release is Na⁺ but not Ca²⁺ dependent, does not rely on a vesicular but on a cytoplasmic pool of neurotransmitter, is not modulated by presynaptic receptors, and is blocked by uptake inhibitors (see Levi and Raiteri, 1993). Indeed, based on the fact that the 5-HT transporter moves 5-HT with Na⁺ and Cl⁻ across the membrane in one step and K⁺ repositions the carrier in a second step (Keyes and Rudnick, 1982; Nelson and Rudnick, 1982; Rudnick, 1986), Rudnick and Wall (1992b,c) have recently shown that purified platelet plasma membrane vesicles, containing the same 5-HT transporter as the one responsible for 5-HT uptake into presynaptic nerve endings (Lesch et al., 1993c), may either accumulate or extrude [³H]5-HT when appropriate transmembrane ionic gradients are imposed. If manipulating the Na⁺ or K⁺ electrochemical gradient changes the direction of neurotransmitter flow, it would then be tempting to assume that 5-HT may be released via this process during depolarization. Such a mechanism could explain observations by McKenna et al. (1991), in which Ca²⁺-independent 5-HT release from rat brain synaptosomes was induced by high extracellular KCl concentrations. However, K⁺-induced, Ca²⁺-independent 5-HT release was not confirmed by Berger et al. (1992) using a similar in vitro preparation.

On the other hand, multiple studies have confirmed that the 5-HT transporter mediates 5-HT release induced by substituted amphetamines such as PCA, the anorexigenic drug fenfluramine, MDA and the drug of abuse “ecstasy” (MDMA; McKenna et al., 1991; Berger et al., 1992;Rudnick and Wall, 1992a,c; Sabol et al., 1992; Gu and Azmitia, 1993;Bonanno et al., 1994). Apart from releasing 5-HT, these drugs also cause a degeneration of fine 5-HT terminals (Mamounas and Molliver, 1988; O’Hearn et al., 1988; Molliver and Molliver, 1990). Therefore, transporter-mediated release of 5-HT has been proposed as a possible neurotoxic mechanism for 5-HT neurons, and different properties of 5-HT carriers in fine and beaded axons could account for the differential effect of these drugs on the two types of 5-HT fibers. More recently, the demonstration that nonneurotoxic amphetamine derivatives also induce transporter-mediated 5-HT release strongly suggests that release through the 5-HT transporter is not sufficient to cause destruction of 5-HT terminals (Rudnick and Wall, 1993). Indeed, several studies indicate that the DA-releasing property of these amines may be a necessary cofactor for 5-HT fiber destruction (Stone et al., 1988;Schmidt et al., 1990; Johnson and Nichols, 1991).

2. Physiological Role of Extracellular 5-HT Bioavailability in Midbrain Raphe Nuclei.

It is important to bear in mind that all of the methods presently used to determine neurotransmitter release, such as synaptosomal or slice superfusion, in vitro or in vivo voltametry and microdialysis, measure not only the release but also the summation of neurotransmitter release, uptake, diffusion, and metabolism, generally referred to as neurotransmitter output. Hence, whenever the term release is used, it is understood that it does not strictly refer to 5-HT release but rather to what is detected by the method used. With this in mind, the extracellular concentration of 5-HT in the DRN has been estimated to be between 3 and 10 nM using in vivo voltametry and microdialysis experiments (Crespi et al., 1988; Bel and Artigas, 1992;Adell et al., 1993). There seems to be general agreement on the fact that extracellular 5-HT levels may vary according to behavioral state changes (Cespuglio et al., 1990; Houdouin et al., 1991; Portas and McCarley; 1994). However, there is still controversy with regard to the direction of these changes. On the one hand, Jouvet’s group has observed the highest 5-hydroxindole peak during sleep (Cespuglio et al., 1990; Houdouin et al., 1991), thus supporting the idea that enhanced 5-HT_1A autoreceptor activation secondary to dendritic 5-HT release triggered by hypnogenic factors such as CLIP or VIP (El Kafi et al., 1994) might be responsible for determining the decrease in 5-HT neuron firing activity observed during SWS and REM (see Jacobs and Azmitia, 1992; Jacobs and Fornal, 1993). On the other hand, Portas and McCarley (1994) argue that extracellular 5-HT in the DRN is highest during wakefulness, lowest during REM, and that extracellular somatodendritic 5-HT availability depends directly on the serotonergic action potential activity. The latter observations are in agreement with results from electrophysiological studies in which 5-HT_1A antagonists were more effective in increasing 5-HT neuron firing activity during wakefulness than during sleep (Fornal et al., 1994a,b). Furthermore, in vitro neurochemical studies indicate that within a range of 5 to 100 Hz, higher stimulation frequencies elicit increasingly higher extracellular concentrations of 5-HT, as measured by fast cyclic voltametry in rat DRN slices (O’Connor and Kruk, 1991).

The role of firing activity on 5-HT release in the raphe nuclei has also been studied using the Na⁺ channel blocker tetrodotoxin (TTX). The local administration of TTX into the DRN or MRN was shown to decrease spontaneous [³H]5-HT release in nonanaesthetized encéphale-isolé cats (Héry et al., 1986) and freely moving rats (Bosker et al., 1994), suggesting that 5-HT release is mainly dependent on firing activity. Other studies did not confirm this view. TTX injection in the immediate vicinity of the DRN (lateral boundaries) did not change the extracellular concentration of 5-HT (Adell et al., 1993). The 5-hydroxyindole acute acid (5-HIAA) peak in the DRN was not altered by intraraphe administration of this Na⁺ channel blocker at a concentration that effectively reduced the voltametric signal in the striatum (Scatton et al., 1985). Although it could be argued that TTX injected in the vicinity and not within the raphe nuclei might have not reached 5-HT neurons, and that extracellular 5-HIAA may not always reflect 5-HT release, results from in vitro experiments have also been variable. When TTX was introduced into the perfusion medium, spontaneous [³H]5-HT release from midbrain raphe slices was increased (Héry et al., 1986), electrically evoked release of 5-HT was reduced (Starkey and Skingle, 1994), and K⁺-induced release of the neurotransmitter was unaffected (El Mansari and Blier, 1996;Piñeyro and Blier, 1996). Interestingly, Pan et al. (1989) have shown that the electrical stimulation of midbrain raphe slices elicits a multicomponent postsynaptic potential in which a fast EPSP precedes the slow inhibitory postsynaptic potential produced by 5-HT_1A receptor activation. This observation was interpreted by the authors as an indication that at least some of the 5-HT is released as a result of synaptically induced excitation of cell bodies, an assumption that seems confirmed by the fact that in this sort of paradigm TTX abolishes not only electrically evoked 5-HT release (Starkey and Skingle, 1994) but also the multiple postsynaptic potential (Pinnock, 1992). On the other hand, TTX resistance of K⁺-induced 5-HT release may be explained by 1) K⁺-induced depolarization, which triggers a direct activation of voltage-sensitive Ca²⁺channels and exocytotic release of 5-HT (experiments to determine the effect of Ca²⁺ channel blockers on K⁺-induced 5-HT release in the presence and absence of TTX may help to assess this possibility) and/or 2) an increase in extracellular K⁺, which may induce carrier-mediated 5-HT release. In favor of the latter mechanism is the observation that 5-HT receptor agonists are much less potent in inhibiting K⁺ than electrically induced 5-HT release from raphe nuclei (Middlemiss, 1987; Starkey and Skingle, 1994;Piñeyro et al., 1995b; Piñeyro and Blier, 1996). Furthermore, in view of the fact that pretreatment with reserpine does not modify 5-HT release in the DRN of freely moving rats, release from a cytoplasmic pool has been proposed (Adell et al., 1993). However, it is worth noting that Ca²⁺ omission greatly reduces or even abolishes electrically and K⁺-induced 5-HT release (Kerwin and Pycock, 1979;Héry et al., 1986; Starkey and Skingle, 1994; Piñeyro et al., 1995b) from midbrain raphe nuclei, thus indicating that exocytosis is the main mechanism involved in neurotransmitter release in this area. Also, in the cat nodose ganglia, which contain cell bodies and dendrites but no 5-HT axons, K⁺-induced 5-HT release is totally abolished in a Ca⁺²-free medium (Fueri et al., 1984). Such an observation is in keeping with the idea not only that soma and dendrites of 5-HT neurons may release 5-HT, but also that somatodendritic release of 5-HT is predominantly exocytotic.

Considerable evidence supports the view that 5-HT release in the cell body area may be regulated by firing-controlling 5-HT_1A autoreceptors. In in vitro superfusion experiments, the 5-HT_1A agonists 8-OH-DPAT, buspirone and ipsapirone have been shown to inhibit electrically evoked [³H]5-HT release from rat midbrain raphe slices, and this effect was blocked by 5-HT_1Aantagonists such as NAN-190, WAY 100135, and S-UH-301 (Starkey and Skingle, 1994; Davidson and Stamford, 1995a; Piñeyro et al., 1995b; Piñeyro and Blier, 1996). It has also been shown that WAY 100135 increases electrically evoked release of [³H]5-HT from midbrain slices in both the absence and presence of a selective 5-HT reuptake blocker (Starkey and Skingle, 1994; Davidson and Stamford, 1995b), an effect that has been attributed to the blockade of tonic activation of 5-HT_1A autoreceptors by the endogenous neurotransmitter. Moreover, local decrease in extracellular 5-HT has also been observed following 8-OH-DPAT injection into the median raphe region (Bosker et al., 1994), further supporting the notion that somatodendritic 5-HT release is subject to local feedback mechanisms through 5-HT_1A autoreceptors. Activation of somatodendritic 5-HT_1A autoreceptors has a major impact on the amount of 5-HT released not only in raphe nuclei but also in forebrain projection areas. For example, infusion of the 5-HT_1A agonist 8-OH-DPAT into the DRN results in a decrease of 5-HT release in the striatum (Bonvento et al., 1992) and the hippocampus (Hutson et al., 1989, Sharp et al., 1989, Adell et al., 1993). Similarly, if the extracellular availability of 5-HT in the biophase of somatodendritic 5-HT_1A receptors is increased by the direct application of the neurotransmitter or intraraphe perfusion of SSRIs or MAOIs, 5-HT release in cortex, striatum, and hippocampus is markedly reduced (Becquet et al., 1990;Adell and Artigas, 1991; Celada and Artigas, 1993). If the injection volume within a specific site is small enough to ensure no diffusion to the corresponding neighboring dorsal or MRN, then reduction of 5-HT release will follow the differential projection pattern of the nucleus into which the agonist was administered: injection of 8-OH-DPAT into the MRN reduces the extracellular 5-HT concentration in the hippocampus but not that of the striatum, and the converse is true for intraDRN administration of 8-OH-DPAT (Bonvento et al., 1992; Kreiss and Lucki, 1994).

As is the case with firing activity, the effect of blocking the somatodendritic 5-HT_1A autoreceptor may be variable on 5-HT release in projection areas and may in part depend on the extracellular concentration of 5-HT at the cell body level. In unanesthetized cats and freely moving rats, systemic administration of drugs that block 5-HT_1A receptors, such as methiothepin, S-UH-301, (+)-WAY 100135, and pindolol, produce no change in striatal or hippocampal 5-HT release (Becquet et al., 1990; Nomikos et al., 1992; Routledge et al., 1993; Romero et al., 1996). In contrast, the systemic administration of S-UH-301 or pindolol before the local infusion of citalopram into the DRN or its systemic administration blocks the reduction caused by the latter drug in hippocampal and striatal 5-HT release (Hjorth, 1993; Romero et al., 1996). Furthermore, infusion of methiothepin into the DRN blocks the reduction in hippocampal 5-HT release caused by systemic administration of the SSRI sertraline (Invernizzi et al., 1991), and systemic administration of the 5-HT_1A/1B antagonist (−)-penbutolol (Hjorth and Sharp, 1993) prevents the decrease in hippocampal 5-HT output caused by systemic citalopram administration (Hjorth, 1993). 5-HT_1A ligands with low intrinsic activity, such as BMY 7378 and NAN-190, have been shown to produce a small decrease in hippocampal 5-HT. This effect has been attributed to 5-HT_1A receptor activation (Hjorth and Sharp, 1990; Sharp et al., 1990). Systemic administration of the partial agonist SDZ 216525 also reduces 5-HT release in terminal projection areas (Sharp et al., 1993b, Gurling et al., 1993). On the other hand, (+)-WAY 100135, which has been shown to inhibit 5-HT neuron firing activity by blocking α₁ adrenoceptors, does not modify hippocampal 5-HT release when given systemically (Routledge et al., 1993).

The idea that, apart from somatodendritic autoreceptors, 5-HT_1A receptors not located on the somata of 5-HT neurons may regulate firing-dependent terminal 5-HT release has now gained considerable support. In fact, inactivation of autoreceptors by intraraphe infusion of pertussis toxin prevents the reducing effect on striatal 5-HT release caused by direct administration of citalopram into the dorsal raphe but does not interfere with the reducing effect caused by the systemic administration of 8-OH-DPAT (Romero et al., 1994a). In another line of evidence, electrophysiological experiments in which 5-HT_1A agonists were microiontophoretically applied onto 5-HT neurons in the DRN and MRN indicate that the former are more responsive than the latter to the local application of such drugs (Blier et al., 1990). However, following systemic administration of 8-OH-DPAT (250 μg/kg), no differential effect on 5-HT released in dorsal and median raphe-innervated areas was observed (Hjorth and Sharp, 1991). Both observations could be reconciled by admitting that post- and not presynaptic 5-HT_1A receptors were being stimulated. On the other hand, also using systemic administration,Sinton and Fallon (1988) have reported a differential sensitivity of dorsal and median raphe neurons to this 5-HT_1Aagonist: only 5 μg/kg of 8-OH-DPAT was needed to abolish 5-HT neuron firing activity in the DRN, whereas 30 μg/kg was not enough to induce cessation of firing of 5-HT neurons in the MRN. Given the differences in doses used, it could be possible that pre- and postsynaptic receptors were differently stimulated in the two studies.

3. Autoregulation of 5-HT Release in the Raphe Nuclei.

The above-mentioned evidence indicates that, by means of a potent feedback control mechanism on 5-HT neuron firing frequency, somatodendritic 5-HT_1A autoreceptors constitute a major presynaptic determinant in the efficacy of 5-HT synaptic transmission. Hence, all auto- or heteroregulatory influences on 5-HT availability in the biophase of 5-HT_1A autoreceptors will contribute to determine the overall efficacy of 5-HT synaptic transmission in the brain.

Apart from being regulated by somatodendritic 5-HT_1A autoreceptors, 5-HT release from 5-HT neurons is controlled by terminal 5-HT_1B/1Dautoreceptors (see Starke et al., 1989; Table4). These receptors, unlike the somatodendritic ones, can modify 5-HT release without altering 5-HT neuron firing activity (e.g., see Crespi et al., 1990). There is now considerable evidence indicating that this firing-independent control of 5-HT release also takes place at the cell body level and that non-5-HT_1A receptors are involved. Using in vivo voltametry, Blier et al. (1990) have shown that the systemic administration of the 5-HT₁ agonist RU 24969 could reduce extracellular availability of 5-hydroxindoles in the DRN of anesthetized rats without altering 5-HT neuron firing frequency. More recently, these results have been confirmed and extended using TFMPP which, following its systemic administration, was also shown to reduce extracellular availability of hydroxindoles in the rat DRN without modifying 5-HT neuron firing activity (Piñeyro et al., 1995, 1996b). Furthermore, the reducing effects of TFMPP and RU 24969 on extracellular availability of 5-hydroxyindoles in the rat DRN are blocked by the non-5-HT_1A agonist mianserin but not by WAY 100135. Evidence indicating that 5-HT release in the DRN may be controlled independently of 5-HT_1Areceptor activation has also been gathered using an electrophysiological paradigm. The latter allowed to infer the amount of extracellular 5-HT available to stimulate somatodendritic 5-HT_1A autoreceptors by measuring changes in 5-HT neuron firing activity, a direct consequence of 5-HT_1A receptor activation, following electrical stimulation of the medial forebrain bundle in the VTA. The 5-HT agonists TFMPP and RU 24969 were shown to reduce the duration of suppression of 5-HT neuron firing evoked by 5-HT pathway stimulation without modifying their basal electrical activity. The effect of these agonists was blocked by (−)-propranolol and mianserin (which do not interfere with 5-HT_1A-mediated electrophysiological responses of 5-HT neurons). This observation was interpreted as further indication that non-5-HT_1Aagonists regulate 5-HT release in the DRN.

Results from in vitro superfusion studies using midbrain raphe slices from guinea pigs also support this idea: 1) the 5-HT_1B/1D antagonist GR 127935 increases electrically evoked release of 5-HT in guinea pig raphe slices (Starkey and Skingle, 1994; El Mansari and Blier, 1996), 2) the 5-HT_1D/1B agonist sumatriptan inhibits 5-HT release and this effect, not altered by (+)-WAY 100135, is blocked by the 5-HT_1B/1D antagonist GR 127935 which also blocks the inhibitory effect of the nonselective agonist 5-methoxytryptamine (5-MeOT). However, neither sumatriptan- nor 5-MeOT-mediated responses are blocked by WAY 100135 (Starkey and Skingle, 1994; El Mansari and Blier, 1996; note: 5-HT_1B/1D nomenclature used in this manuscript conforms to new guidelines adopted by the International Nomenclature Committee3).

According to the new nomenclature, receptors predominantly found in guinea pig brain are of the 5-HT_1B type, with a nonrodent pharmacological profile (Heuring et al., 1986; Bruinvels et al., 1993). Since drugs used to characterize regulation of somatodendritic 5-HT release by non-5-HT_1Areceptors in guinea pig brain do not readily discriminate 5-HT_1B from 5-HT_1Dreceptors in this species, and the former largely outnumber the latter, it seems reasonable to postulate that newly named 5-HT_1B receptors negatively regulate 5-HT release in guinea pig midbrain raphe nuclei. Interestingly, CP93129, a drug that has been characterized as a 5-HT_1B agonist in rodents, produced no effect in electrically evoked release from midbrain guinea pig slices (El Mansari and Blier, 1996). It is not possible for the time being to determine whether the lack of effect of this drug is due to the fact that guinea pig 5-HT_1B receptors do not recognize CP93129 as rodent 5-HT_1B receptors do, or to the fact that 5-HT_1D and not 5-HT_1Breceptors control somatodendritic 5-HT release in the guinea pig brain.

In rodents, where brain 5-HT_1D as well as 5-HT_1B receptors have been detected (Hoyer et al., 1985a; Waeber et al., 1989a; Bruinvels et al., 1993), superfusion studies performed in midbrain raphe slices of rats and mice (Piñeyro et al., 1995a, 1996b) indicated that 5-HT_1B/5-HT_1D-like receptors negatively regulate 5-HT release in this brain area. Evidence accumulated in the above-mentioned studies may be separated into two categories: 1) Observations that support a role for 5-HT_1D-like receptors in the regulation of 5-HT release in anterior raphe nuclei: the 5-HT₁agonist 5-carboxyamidotryptamine (5-CT) induced a concentration-dependent inhibition of the electrically evoked release of [³H]5-HT from preloaded rat raphe slices which was partly blocked by the 5-HT_1B/1Dantagonists GR 127935 and mianserin which in rats has a very high selectivity for 5-HT_1D versus 5-HT_1B receptors; the 5-HT_1B/1D agonist sumatriptan inhibited electrically evoked release of [³H]5-HT from rat and wild-type mice raphe slices, and this effect was blocked by mianserin and GR 127935; and midbrain slices from 5-HT_1B knockout mice maintained a pharmacological response similar to that observed in wild types. 2) Observations that rule out other specific receptor subtypes include: the fact that the effect of sumatriptan was not blocked by S-UH-301 nor (+)-WAY 100135 excludes a possible nonselective activation of 5-HT_1A receptors, and in spite of its inhibitory effect on evoked [³H]5-HT release in terminal regions, the lack of effectiveness of the selective 5-HT_1B agonist CP 93129 to inhibit evoked release from rat and mice midbrain slices strongly suggests that 5-HT_1B receptors are not involved in the sumatriptan-mediated response in midbrain raphe. Rat (as well as human) 5-HT_1E and 5-HT_1F receptors have low affinity for 5-CT, which in our studies effectively mimicked sumatriptan mediated responses. Similarly, 5-HT_1Esites do not bind sumatriptan with high affinity. Studies from human brain indicate that 5-HT_1F sites have low affinity for methiothepin and do not bind propranolol, both of which blocked the effect of sumatriptan on 5-HT release in rodent midbrain slices. Hence, in spite of the fact that sumatriptan binds to 5-HT_1F receptors, it is unlikely that these or the 5-HT_1E subtype may have mediated the sumatriptan-induced inhibition of 5-HT release from rodent brain. Sumatriptan and GR 127935 have very low affinity for murine 5-HT_5A/5B sites. The affinity of sumatriptan for rat 5-HT₇ receptors is also considerably low (although higher than for 5-HT₅). If 5-HT₇ receptors played an important role in inhibiting 5-HT release in midbrain nuclei, the inhibitory effect of 8-OH-DPAT (that binds with higher affinity to 5-HT₇ than to 5-HT_1Dreceptors) on somatodendritic 5-HT release would have been greater, not smaller, than that of sumatriptan. Hence, even if mRNAs for 5-HT_1D, 5-HT₅, and 5-HT₇ receptor subtypes are present in the DRN, the 5-HT_1D receptor, or a subtype bearing a similar pharmacology, was deemed a better choice to account for the effect of different drugs tested on 5-HT release from rat midbrain raphe (see Table 5 for pharmacology of G protein-coupled 5-HT receptors).

In contrast with the above-mentioned observations on the role of 5-HT₁ receptors controlling 5-HT release in the DRN, Davidson and Stamford (1995a) have reported that CP 93129 inhibited in vitro 5-HT release from rat DRN. On the other hand, they have also observed that the effect of CP 93129 was only blocked at scattered time points during continuous superfusion with the 5-HT_1B antagonist isamoltane, an unexpected response since once the blockage is established, if there are no changes in drug concentrations there is no reason for such fluctuation. Furthermore, the stimulation protocol used in the latter study and the one from our group (Piñeyro et al., 1995b) were different; in one, stimulation parameters were set to mimic 5-HT neuron firing frequency during wakefulness (1–5 Hz; Piñeyro et al., 1995b), whereas in the other, pseudosingle pulse stimulations (100 Hz) were used (Davidson and Stamford, 1995a). It has been extensively documented that different stimulation parameters may differ in the concentration of 5-HT induced in the biophase of 5-HT receptors, eliciting different drug actions even if all other experimental variables are kept constant (see below). In these studies, however, not all variables were kept constant, the concentration of CP 93129 used by Davidson and Stamford (1995a) was three times as high as the one used in our study (Piñeyro et al., 1995b). In our paradigm, concentrations of this product higher than 100 nM have been shown to produce significant increases in basal 5-HT release. Because values for this parameter were not reported when a concentration of 300 nM CP 93129 was used, it may not be ruled out that evoked release of 5-HT may have been reduced after treatment with the 5-HT_1B agonist due to unloading of the slices previous to stimulation.

Based on the observations mentioned above, it could be hypothesized that if in rats 5-HT_1B receptors indeed play a regulatory role in midbrain 5-HT release, it would probably involve 5-HT release from collaterals and 5-HT terminals, whereas on the other hand, 5-HT_1D-like receptors would control a larger somatodendritic neurotransmitter release. Because 5-HT terminals are not abundant in DRN and MRN, the previous interpretation would lend support to two observations: 1) lack of effect of 5-HT_1B receptor blockade following systemic administration of an SSRI to rats (Fig.2) and 2) lack of effect of 5-HT_1B receptor blockade during wakefulness, a time when extracellular somatodendric concentrations of 5-HT are high. The former interpretation would also be in keeping with the observation that even if in binding studies the number of 5-HT_1B receptors is much higher than that of 5-HT_1D receptors, the activation of the latter by regulating a greater contingent of releasable 5-HT (somatodendritic) produces a dramatic inhibition of release, whereas the activation of the former has a modest or no effect.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

A, effect of systemic administration of the 5-HT_1A/1B antagonist (−)-penbutolol on spontaneous hippocampal 5-HT release in anesthesized rats. B, effect of hippocamal perfusion of the 5-HT_1A/1B antagonist (−)-penbutolol and systemic administration of the 5-HT_1A antagonist S-UH-301 on spontaneous 5-HT release in rat hippocampus. C, effect of the systemic administration of the 5-HT_1B/1D antagonist GR 127935 on cortical 5-HT release of freely moving guinea pigs. D, effect of the combined administration of a 5-HT_1A (WAY 100135) and the 5-HT_1D antagonist on cortical 5-HT release in freely moving guinea pigs (see text for details).

Comparing the results obtained in microdialysis studies in rats and guinea pigs has helped evaluate the role of 5-HT_1B/1D receptor subtypes in the regulation of 5-HT release in raphe nuclei. In anesthetized rats pretreated with citalopram, the systemic administration of the 5-HT_1A/1B antagonist (−)-penbutolol (8 mg/kg s.c) produces an increase in hippocampal 5-HT release (Hjorth, 1993), most likely by simultaneously blocking activated somatodendritic 5-HT_1A and terminal 5-HT_1Bautoreceptors (Fig. 2A). This effect is similar to that observed following intrahippocampal perfusion of (−)-penbutolol (1 μM in the perfusion medium) in combination with systemic administration of the 5-HT_1A antagonist S-UH-301 (3 mg/kg s.c.; Fig.2B). When administered systemically, there should be no reason why (−)-penbutolol would not block somatodendritic 5-HT_1B receptors in addition to terminal ones. Hence, if 5-HT_1B receptors played a significant inhibitory role in the regulation of midbrain 5-HT release, the logical expectation would be for this type of treatment to induce an additional increase in terminal 5-HT release as compared to intrahippocampal perfusion of (−)-penbutolol, provided that 5-HT_1A autoreceptor function is blocked. Since there was no such observation, the most parsimonious conclusion would be that, in rats, 5-HT_1B receptors do not play a major role in controlling extracellular availability of 5-HT at the somatodendritic level. In freely moving guinea pigs, the results observed are quite different (Price et al., 1994). The systemic administration of the 5-HT_1B/1D antagonist GR 127935 (0.3 mg/kg) produces a marked decrease in extracellular cortical 5-HT (Fig. 2C), an unexpected effect since this drug had been shown to block terminal 5-HT_1B autoreceptors in this species and increase 5-HT release in cortical slices (Price et al., 1994; Table 6). The decrease in release could be due, in theory, to the partial agonistic properties of GR 127935 on guinea pig terminal autoreceptors (Tingley et al., 1994). However, it is improbable that this was the actual cause for reduced in vivo cortical 5-HT release since the systemic administration of WAY 100635 (1 mg/kg) not only blocked, but in fact reversed the inhibitory effect of GR 127935, thus producing a marked elevation in cortical 5-HT levels (Price et al., 1994; Fig. 2D). The reversal by WAY 100635 of the GR 127935 effect can be explained by assuming that GR 127935 blocked 5-HT_1B/1Dreceptors that negatively control 5-HT release at the somatodendritic level. In doing so, such an antagonism would promote a higher extracellular 5-HT level in raphe nuclei and a greater activation of somatodendritic 5-HT_1A autoreceptors. This, in turn, would lead to a reduction in 5-HT neuron firing and thus to a reduction in firing-dependent terminal 5-HT release. The administration of WAY 100635 blocks the 5-HT_1A autoreceptors, the 5-HT neurons recover their firing activity, and the enhancing effect of terminal 5-HT_1B receptor blockade (by GR 127935) on cortical 5-HT release is then unveiled. It would be interesting to test this hypothesis by assessing the effect of systemic administration of GR 127935 in rat brain terminal projection areas. If inhibition resulted, as it did in guinea pigs, this would suggest the blockade of a somatodendritic 5-HT_1D receptor. In such a case, it should be determined whether WAY 100635 could unmask the enhancing effect of GR 127935 produced by terminal 5-HT_1B receptor blockade. Because rat and guinea pig 5-HT_1B and 5-HT_1Dreceptors are species homologs, speculating that the 5-HT_1D subtype may be in control of somatodendritic 5-HT release whereas 5-HT_1Breceptors control terminal release of the neurotransmitter seems plausible. If indeed in humans somatodendritic and terminal autoreceptors are of the 5-HT_1D and 5-HT_1B subtype, respectively, it might be possible to develop selective terminal and somatodendritic ligands. Theoretically, the expected effect of somatodendritic 5-HT_1D activation and of terminal 5-HT_1B receptor blockade would be an enhancement of 5-HT release in projection areas.

Concerning the actual localization of 5-HT_1D or 5-HT_1D-like receptors, the fact that they are functional in midbrain slices indicates that they may be present within the raphe nuclei. Indeed, in slices, the raphe nuclei are separated from their afferent sources. Therefore, the possibility that the receptors in question are located on cell bodies of any of the extra-midbrain afferents to the 5-HT nuclei may be ruled out. Moreover, addition of TTX to the perfusion medium of midbrain raphe slices does not modify the inhibitory effect of sumatriptan on the K⁺-evoked release of [³H]5-HT (El Mansari and Blier, 1996;Piñeyro and Blier, 1996), 5-HT_1D-like receptors have been proposed to be autoreceptors. Finally, GR 127935 has been shown not to block the inhibitory effect of 5-HT on 5-HT neuron firing activity in guinea pig raphe slices (Craven and Grahame-Smith, 1994), and, in rat DRN, TFMPP as well as RU 24969 reduce somatodendritic availability of 5-hydroxyindoles without changing 5-HT neuron firing activity (Blier et al., 1990; Piñeyro et al., 1996b), further suggesting that in anesthetized animals non-5-HT_1A, and possibly 5-HT_1D receptors, contribute to the regulation of somatodendritic release of 5-HT independent of the regulation of 5-HT neuron firing. Such an interpretation is further supported by the observation that in anesthetized guinea pigs, a situation in which extracellular somatodendritic availability is comparatively low, systemic administration of GR 127935 leaves 5-HT neuron firing activity unchanged (Sprouse et al., 1995). On the other hand, it would be expected that in freely moving animals systemic GR 127935 administration would reduce 5-HT neuron if not given in combination with a 5-HT_1A receptor antagonist.

4. Heteroregulation of Neurotransmitter Release in Midbrain Raphe Nuclei.

Numerous neurotransmitters and neuromodulators from afferent terminals or intrinsic non-5-HT neurons influence 5-HT release in the raphe nuclei. For example, low-frequency (1.5 Hz) stimulation of the habenulo-raphe pathway elicits a decrease in 5-HT release in the DRN (Reisine et al., 1982). Similar results are obtained by intraraphe or systemic administration of GABA or GABA agonists (Scatton et al., 1985; Becquet et al., 1990). This pharmacological inhibitory effect on DRN 5-HT release is abolished by transection of the habenulo-raphe pathway (Nishikawa and Scatton, 1985). The need for the integrity of the habenulo-raphe connection has thus been interpreted in two alternative ways: 1) the effect of GABA or GABA agonists is an indirect one, and GABA receptors are located on habenulo-raphe fibers or 2) the inhibitory effect of GABA is made evident only if 5-HT neurons have a certain degree of tonic activation, provided in turn by the habenulo-raphe pathway. It is worth noting that the most consistently described tonic excitatory input originates in locus ceruleus (Baraban and Aghajanian, 1981). The common feature shared by both of the interpretations above is that they assume that the habenulo-raphe pathway must contain an excitatory neurotransmitter, an assumption which is in disagreement with the fact that the main observed effect of habenulo-raphe pathway stimulation is depression of 5-HT neuron firing (72–88% of the 5-HT neurons are inhibited; Stern et al., 1981). However, when a high stimulation frequency (15 Hz) is used, the habenulo-raphe pathway induces a marked increase in 5-HT release in projection areas, an effect that is blocked by injection of kynurenic acid into the DRN (Kalen et al., 1989). Such an observation has ledKalen et al. (1989) to suggest that EAAs are the main neurotransmitters in the habenulo-raphe pathway, and that glutamatergic fibers could have a double effect on 5-HT neuron, i.e., direct phasic stimulation and indirect inhibition by stimulating GABAergic interneurons. Such an observation is in keeping with in vitro findings in which stimulation of midbrain slices with an electrode in the DRN causes fast IPSPs and EPSPs that are respectively blocked by bicuculline or picrotoxin and the NMDA antagonists CNQX and APV (Pan et al., 1989; Pinnock 1992). The EPSPs due to electrically evoked release of EAA from afferent fibers onto 5-HT neurons may be reduced by activation of presynaptic inhibitory κ-opioid receptors located on the glutamatergic fibers (Pinnock, 1992). However, the predominant effect of systemically administered morphine is a stimulation of 5-HT release, an effect contrary to the one expected from activating inhibitory receptors on excitatory fibers impinging onto 5-HT neurons. A possible explanation for such a contradiction could be that the effect of activation of κ-receptors on EAA terminals may be overcome by activation of other opioid receptors, probably also located within the raphe nucleus, as suggested by Tao and Auerbach (1994). Apart from eliciting postsynaptic potentials on 5-HT neurons recorded from midbrain raphe slices, GABA and glutamate have been shown to modulate 5-HT release in rostral rhombencephalic raphe cells in primary cultures (Becquet et al., 1993b). GABA produces its negative modulation predominantly via GABA_A but also GABA_Breceptors, whereas EAA induce 5-HT release by stimulating NMDA receptors (Becquet et al., 1993a). Furthermore, in vivo application of the GABA_A antagonist picrotoxin into the DRN locally increased 5-HT release in unanesthetized rats (Becquet et al., 1990), suggesting that the latter receptors may induce a tonic inhibition of 5-HT release in the dorsal raphe. The release of 5-HT in the DRN is also modulated by tachykinins and catecholamines. Indeed, Substance P has been shown to increase 5-HT release in vitro in midbrain raphe slices (Kerwin and Pycock, 1979) and in vivo following its intraraphe injection (Reisine et al., 1982). The local infusion of amphetamine, apomorphine, and the selective D₂receptor agonist quinpirole also increases 5-HT release in DRN, the effect of apomorphine being blocked by the selective D₂ receptor antagonist raclopride but not the D₁ antagonist SCH 23390 (Ferré and Artigas, 1993; Ferré et al., 1994). The latter observations confirm the role of D₂ receptors in modulating extracellular availability of 5-HT in the DRN, but whether the source of DA are dopaminergic afferents or DA neurons within the nucleus is still unknown. In the case of NA, this catecholamine inhibited K⁺-evoked release of [³H]5-HT from raphe slices. This effect was mimicked by α₂ adrenoceptor agonists clonidine and oxymetazoline, but not by the α₁adrenoceptor agonists phenylephrine and methoxamine. Furthermore, yohimbine and rauwolscine not only blocked the effect of clonidine, but, when administered alone, they both increased the K⁺-induced release of [³H]5-HT. A possible interpretation of the auto- and heteroregulatory connections of 5-HT release in the DRN is given in Fig. 3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Graphic representation of auto- and heteroregulatory connections of midbrain 5-HT neurons: somatodendritic 5-HT_1A as well as 5-HT_1D/1D-like receptors are principally influenced by extracellular 5-HT of somatodendritic origin. They respectively inhibit 5-HT neuron firing or local release of neurotransmitter. Extracellular 5-HT also modulates terminal 5-HT_1B autoreceptors located either on axon collaterals from efferent axons of local 5-HT neurons or those located on axons impinging from neighboring 5-HT nuclei. Activation of α₁adrenoceptors increases 5-HT neuron firing activity and that of α₂ adrenoceptors inhibits 5-HT release. Both receptors are stimulated by NA released from axons originating in locus ceruleus. Pharmacological stimulation of dopaminergic D₂ receptors increases local 5-HT release. It is not clear whether these receptors are physiologically stimulated by DA released from intraraphe DA neurons or dopaminergic projections originating in the VTA. The stimulation of the habenulo-raphe pathway is believed to induce glutamate release, which, via NMDA receptors, evokes a direct EPSP on 5-HT neurons. Glutamate also stimulates interneurons which may contain either Substance P or GABA. Although GABA inhibits 5-HT neuron firing activity (mainly via GABA_A receptors), Substance P may produce an increase in local 5-HT release. Finally, GABAergic and enkephalinergic interneurons both inhibit glutamate release. Dashed lines, afferent nuclei boundaries.

Interestingly, in some cases, heteroregulatory influences may produce their effects via activation of autoregulatory mechanisms, e.g., D₂ receptor stimulation induces an increase in DRN 5-HT release and a decrease in striatal release, the latter effect being blocked by the administration of (+)-WAY 100135 (Ferré et al., 1994). In other cases, heteroregulatory mechanisms may override autoregulation, e.g., local infusion of Substance P enhances 5-HT release in the dorsal raphe (Reisine et al., 1982) and in the hippocampus (Gradin et al., 1992), or the contrary is observed with GABA which, in spite of decreasing somatodendritic availability of 5-HT when injected into the DRN, also decreases striatal 5-HT release (Becquet et al., 1990).

5. Autoregulation of 5-HT Release in Terminal Projection Areas: Cortex and Hippocampus.

Evidence indicating that activation of 5-HT_1B terminal autoreceptors inhibit 5-HT release is extensive and convincing (rat, mouse, guinea pig, pig, rabbit, human brain; see Table 4 for pharmacological profiles of terminal autoreceptors; Hoyer and Middlemiss, 1989; Starke et al., 1989; Göthert, 1990; Middlemiss and Hutson, 1990). Apart from this well established fact, there is now considerable evidence indicating that more than one receptor subtype could be involved in controlling release in the same region, and that some of these receptors might not be of the 5-HT_1B subtype. In rat cortical slices, 8-OH-DPAT has been shown to inhibit electrically or K⁺-evoked 5-HT release (Hamon et al., 1984;Limberger et al., 1991). Because such an effect was not blocked by 300 nM of the 5-HT_1B antagonist isamoltane (Waldmeier et al., 1988; Schoeffter and Hoyer, 1989), it was suggested that 5-HT_1D isamoltane-resistant sites might contribute to regulate 5-HT release in the rat frontal cortex (Limberger et al., 1991). Today we know that 8-OH-DPAT has considerable affinity not only for 5-HT_1D, but also 5-HT₅ and 5-HT₇receptors (Table 5), and also that mRNAs for all three receptor subtypes are present in the DRN (Table 5). This suggests that, similar to the 5-HT_1B receptor subtype, 5-HT_1D, 5-HT₅, as well as 5-HT₇ receptors could be expressed as autoreceptors on 5-HT neurons. The concentrations at which 8-OH-DPAT effectively inhibited evoked 5-HT release in the above-mentioned studies were higher than 100 nM. Thus, it is tempting to speculate that the probability for 5-HT_1D or 5-HT₅ receptors to be involved in this effect is greater than for 5-HT₇ receptors, since the affinity for the latter is very high (Table 5) and 8-OH-DPAT would have been expected to be effective at lower concentrations. Moreover, it should be noted that at concentrations of up to 1 μM 8-OH-DPAT, some groups were unable to demonstrate an inhibition of cortical evoked release of 5-HT (Middlemiss, 1984b; Engel et al., 1986; Maura et al., 1986). Additional evidence supporting the fact that 5-HT receptors controlling 5-HT release in rat hippocampus are heterogeneous has been obtained by our group. In superfusion experiments, sumatriptan (1–1000 nM) and CP 93129 (1–300 nM) induced a dose-dependent inhibition of electrically evoked [³H]5-HT release from rat hippocampal slices. Although the effects of CP 93129 and sumatriptan were blocked by (±)-cyanopindolol (1 μM), only that of sumatriptan was blocked by mianserin (0.3 and 1 μM). On the other hand, only the effect of CP 93129 was blocked by (−)-propranolol (0.3 μM). Moreover, incubation of rat hippocampal slices with the alkylating agent N-ethylmaleimide (NEM), abolished the effect of CP 93129 but not that of sumatriptan (Piñeyro and Blier, 1996). Furthermore, the inhibitory effect of 5-MeOT in rat hippocampus has been shown to remain unaffected by NEM, pertussis, or cholera toxin pretreatment, further indicating that, in the rat, there is an hippocampal 5-HT receptor subpopulation which is indeed resistant to G_i/o protein inactivation (Blier, 1991). Results from superfusion studies in 5-HT_1B knockout mice provide further evidence that a nonhomogeneous terminal autoreceptor population exists: non-5-HT_1B receptors continue to control 5-HT release in terminal projection areas in 5-HT_1B knockout mice (Piñeyro et al., 1995a). Moreover, it is possible that the combination of receptor subtypes in terminal autoreceptor populations may differ among specific projection areas (e.g., cortex and hippocampus). This would not be surprising given that each of the 5-HT nuclei contribute in a different manner to the innervation of distinct terminal fields. In vivo experiments in the frontal cortex and hippocampus of 5-HT_1B knockout mice also support the idea that non-5-HT_1D, non-5-HT_1Breceptors regulate 5-HT release in the mouse frontal cortex and that terminal 5-HT autoreceptor populations in the cortex and hippocampus may be different (Trillat et al., 1996).

In nonrodent species, an inhibitory effect of 8-OH-DPAT on terminal 5-HT release (Feuerstein et al., 1987; Schlicker et al., 1989) is not surprising given the moderate affinity of the drug for the nonrodent 5-HT_1B receptor (Table 5). However, until more selective ligands become available, a role for 5-HT₅, 5-HT₇ receptors, or even other 5-HT receptors not as yet characterized, on terminal 5-HT release may not be ruled out. Methiothepin is the only antagonist that effectively blocks the inhibition induced by 8-OH-DPAT (Feuerstein et al., 1987), but, since methiothepin binds with high affinity to all 5-HT_1D, 5-HT₅, and 5-HT₇ receptor subtypes, no further inferences can be made. In the guinea pig hippocampus, but not in the cortex, methiothepin has been shown to block the inhibitory effect of 5-CT and sumatriptan with less potency than that of 5-HT (Wilkinson and Middlemiss, 1992; Wilkinson et al., 1993), suggesting an heterogeneity in the receptor subtypes regulating 5-HT release in the former but not the latter region, where methiothepin equipotently blocks the effects of both 5-HT and sumatriptan. Differences among 5-HT receptor populations controlling 5-HT release in the cortex and hippocampus in the guinea pig brain has also been reported by El Mansari and Blier (1996). These authors showed that G protein inactivation with NEM attenuates the inhibitory effect of 5-MeOT in the cortex, but not in the hippocampus, although the inhibitory effect of sumatriptan was reduced in NEM-pretreated slices from both regions. The difference between cortical and hippocampal 5-HT autoreceptor populations is further supported by the fact that the inhibitory effect of 5-MeOT on electrically evoked release of [³H]5-HT is attenuated in hippocampus but not frontal cortex slices obtained from guinea pigs treated with paroxetine for 21 days (Blier and Bouchard, 1994; El Mansari et al., 1995).

In vitro as well as in vivo studies have shown that the amount of 5-HT released per electrical impulse increases with decreasing frequencies of stimulation (Göthert 1980; Baumann and Waldmeier, 1981; Chaput et al., 1986a; Blier et al., 1989a), probably due to a lower degree of activation of autoreceptors by prolonging the interval between the stimulation pulses. This interpretation is supported by findings indicating that, at higher stimulation frequencies, the effectiveness of terminal 5-HT receptor agonists (5-HT itself, 5-CT, 5-MeOT) to inhibit 5-HT release is reduced and that of the antagonist methiothepin to enhance this parameter is increased (Baumann and Waldmeier, 1981;Chaput et al., 1986a; Blier et al., 1989a). Another way in which the concentration of endogenous 5-HT in the biophase of 5-HT autoreceptors may be increased is by inhibition of neuronal 5-HT reuptake. Similar to high stimulation frequencies, in the presence of 5-HT reuptake blockers, the effectiveness of 5-HT agonists is reduced (Langer and Moret, 1982; Galzin et al., 1985) and that of antagonists is enhanced (Feuerstein et al., 1987). To explain SSRI-induced changes in efficacy, two alternative interpretations have been proposed: 1) an increase in 5-HT concentration in the biophase of terminal autoreceptors, and 2) an interaction between neuronal 5-HTT and terminal 5-HT receptors (Langer and Moret, 1982; Galzin et al., 1985; Passarelli et al., 1987; Moret and Briley, 1988). Two important arguments against the “molecular link hypothesis” favor the “5-HT biophase hypothesis”: 1) in experiments on synaptosomes in which the released transmitter is washed away rapidly and 5-HT in the biophase remains too low for autoreceptor activation, and 5-HT receptor agonists produce the same inhibition of synaptosomal [³H]5-HT release in the absence or presence of reuptake blockers (Raiteri et al., 1984; Bonanno and Raiteri, 1987), and 2) when very short trains of high frequency pulses are used in brain slices, release of [³H]5-HT is measurable, and yet pulses are too short to generate autoinhibition (Limberger et al., 1991). In the latter circumstances, 5-CT and 5-MeOT generate similar concentration-effect curves for the inhibition of evoked [³H]5-HT release in the absence and presence of different reuptake blockers (Limberger et al., 1990). On the other hand, the fact that 5-HT reuptake blockers continue to reduce the effectiveness of 5-HT autoreceptor agonists following 5-HT depletion by the 5-HT synthesis inhibitor PCPA has been used as an additional argument to support the molecular link hypothesis (Galzin et al., 1985; Passarelli et al., 1987). It is important to notice, however, that even after a 90% depletion of hippocampal 5-HT by PCPA pretreatment, 5-HT pre- and postsynaptic functions were found to be unchanged (Chaput et al., 1990), suggesting that the 5-HT system maintains its adaptability even after a great reduction of endogenous 5-HT.

The effect of changes in 5-HT biophase concentration are particularly evident in the case of partial agonists which can either inhibit or enhance terminal 5-HT release depending on the circumstances. For example, in the absence of the SSRI nitroquipazine or fluoxetine (1 μM), the 5-HT_1B ligand (±)-cyanopindolol inhibited electrically evoked release of 5-HT in the rat hippocampus and enhanced the same parameter when the slices were superfused with the uptake inhibitors (Feuerstein et al., 1987). Similarly, in conditions of negligible autoinhibition (4 pulses at 100 Hz), metergoline inhibited [³H]5-HT release from rabbit frontal cortex slices (Limberger et al., 1991), had no effect when 5-HT biophase was higher (360 pulses at 3 Hz), and enhanced release when 360 pulses at 3 Hz were delivered in the presence of reuptake inhibitors (Feuerstein et al., 1987). Table 6 summarizes the effects on evoked 5-HT release in the cortex and hippocampus, obtained with different terminal autoreceptor antagonists in different species and using different stimulation paradigms. Methiothepin has been shown to consistently enhance evoked release of 5-HT in conditions where autoinhibition exists, suggesting that it might be a pure antagonist. Furthermore, under the same conditions (360 pulses at 3 Hz; rat hypothalamic slices) in which antagonists such as metergoline and alprenolol had no effect by themselves but blocked the effect of LSD, methiothepin induced opposite effects to those of the terminal autoreceptor agonist. This observation has prompted the suggestion that methiothepin could possess inverse agonistic properties at the terminal 5-HT autoreceptor (Moret and Briley, 1993). More recently, this suggestion has been proved, because methiothepin was found to increase [³⁵S]GTPγS binding to human 5-HT_1D and 5-HT_1B receptors (Jones et al., 1995), behaving as an inverse agonist. This observation suggests that methiothepin binding to 5-HT_1B/1Dreceptors favors the uncoupling of the receptor-G protein complex. An inverse agonist at the human terminal autoreceptor that would enhance 5-HT release in terminal projection areas might prove an effective antidepressant with quick onset of action.

5-HT release in guinea pig and rat cortex and hippocampus is also modulated by 5-HT₃ receptors (Galzin et al., 1990; Barnes et al., 1992; Martin et al., 1992; Blier and Bouchard, 1993). Unlike 5-HT_1B/1D, 5-HT₃ receptors are not localized on 5-HT terminals (Blier et al., 1993c); they enhance 5-HT release (Galzin et al., 1990; Barnes et al., 1992; Martin et al., 1992; Blier and Bouchard, 1993) and desensitize within minutes of agonist exposure (seeHoyer 1990; Blier and Bouchard, 1993). They share with terminal autoreceptors their frequency dependence, being more effective at enhancing 5-HT release at lower than at higher frequencies (Blier and Bouchard, 1993).

6. Heteroregulation of Neurotransmitter Release from 5-HT Fibers in Cortex and Hippocampus.

Several in vivo and in vitro studies using brain slices or synaptosomes have provided evidence for the involvement of multiple neurotransmitters in the local regulation of 5-HT release in terminal projection areas. α₂-Adrenergic heteroreceptors on 5-HT terminals in the brain of different species have long been known to inhibit 5-HT release (Gobbi et al., 1990;Raiteri et al., 1990; see Göthert and Schlicker, 1991;Maura et al., 1992b; Blier et al., 1993c). Several findings support the idea that the α₂-adrenergic heteroreceptors have different properties as compared with α₂ autoreceptors regulating NA release in terminal projection areas: 1) they have different pharmacological profiles (Raiteri et al., 1983b; Maura et al., 1992b; Mongeau et al., 1993), 2) α₂ auto- and heteroreceptors are differentially modulated by neuropeptides such as NPY (Martire et al., 1989), and 3) only the former desensitize following long-term treatment with befloxatone, a selective MAO-A inhibitor (see Blier et al., 1993c). The idea of distinct functional properties of adrenoceptors regulating 5-HT and NE release in terminal projection areas has been exploited in the development of new antidepressant drugs. Interestingly, selective α₂ heteroreceptor antagonists, even given acutely, may have the capacity to enhance 5-HT neurotransmission (Haddjeri et al., 1996). Like NE, histamine also exerts a negative regulation of 5-HT release, probably via activation of H₃ heteroreceptors (Fink et al., 1990). On the other hand, acetylcholine (ACh) enhances forebrain 5-HT release by activating nicotinic receptors (Toth et al., 1992; Summers and Giacobini, 1995). The fact that kynurenic acid almost completely prevents the enhancing effect on neurotransmitter release induced by the local administration of nicotine suggests that its effect on 5-HT release is indirect, mediated via glutamic acid release (Toth et al., 1992). Inhibitory amino acids are also involved in the regulation of 5-HT release in cortex and hippocampus. Following systemic administration of benzodiazepine agonists, spontaneous or evoked release of 5-HT in either region has been shown to decrease (Hitchcott et al., 1990; Broderick, 1991; Cheng et al., 1993). Local effects are more ambiguous. In the hippocampal formation, for example, the activation of the GABA_A receptor complex has been shown to produce either a decrease, no effect, or an increase in 5-HT release (Pei et al., 1989; Lista et al., 1990). The fact that the local application of the Cl⁻ channel blocker picrotoxin induced a robust increase in hippocampal 5-HT of freely moving rats supports the idea that GABA exerts a tonic inhibition of 5-HT release in this region (Pei et al., 1989). Numerous polypeptides have also been found to locally modulate 5-HT release in projection areas. For example, PYY and pancreatic polypeptide inhibit cortical 5-HT release via the activation of the same presynaptic G protein-coupled receptor as NPY (Schlicker et al., 1991). On the other hand, Substance P and neurokinin A, two coexisting neuropeptides of the tachykinin family, stimulate 5-HT release in this same area (Iverfeldt et al., 1990). Pharmacological activation of local opioid receptors (δ, κ, and μ) suggests that endogenous hippocampal opiates may also be involved in the negative regulation of 5-HT release in this area (Passarelli and Costa, 1989; Cui et al., 1994).

7. 5-HT_1B versus 5-HT_1D Receptors.

5-HT_1B-binding sites, as opposed to 5-HT_1A sites, were initially pharmacologically described. They represented the sites that bind spiperone and 8-OH-DPAT with low affinity (Pedigo et al., 1981; Middlemiss and Fozard, 1983) but present high affinity for ¹²⁵I-cyanopindolol (Hoyer et al., 1985a,b). They have also been defined by exclusion as [³H]5-HT-binding sites which are neither of the 5-HT_1A nor of the 5-HT_2Csubtype (Blurton and Wood, 1986; Peroutka, 1986; Alexander and Wood, 1987). Originally, these pharmacologically described 5-HT_1B sites were described in rodents (Hoyer et al., 1985a; Waeber et al., 1989a; Waeber and Palacios, 1992) but not in guinea pig, pig, cow, or human brain (Heuring et al., 1986; seeBruinvels et al., 1993 and references within). In nonrodent species, the sites visualized in the presence of saturating concentrations of 8-OHDPAT and mesulergine (Heuring and Peroutka, 1987; Waeber et al., 1988, 1989b) were originally named 5-HT_1D and more recently reclassified as 5-HT_1B. Hence, in spite of their pharmacological differences, sites defined by exclusion in rodent and nonrodent brain would fall under the present 5-HT_1B category. However, it has been established that the sites labeled in this way represent an heterogeneous receptor population composed of at least 5-HT_1B/1D(5-CT-sensitive) and 5-HT_1E (5-CT-insensitive) receptors (Leonhardt et al., 1989; Sumner and Humphrey, 1989; Beer et al., 1992; Lowther et al., 1992; Miller and Teitler, 1992). Furthermore, [³H]5-CT has been shown to label two different subpopulations of 5-HT₁ as high-affinity sites in guinea pig cortex and striatum (Mahle et al., 1991). It was the development of an iodinated radioligand, serotonin-5-O-carboxymethyl-glycyl-¹²⁵I-tyrosinamide (¹²⁵I-GTI; Boulenguez et al., 1991, 1992), that allowed 5-HT_1B binding sites to be directly labeled in human, nonhuman primate, and guinea pig brain, among others (Bruinvels et al., 1991, 1992).

Rodent versus nonrodent 5-HT_1B receptor subtype pharmacology is quite different, whereas that of 5-HT_1D and nonrodent 5-HT_1B-binding sites is very much alike. Two pharmacologically distinct groups may be thus defined, rodent 5-HT_1B and nonrodent 5-HT_1Band all species 5-HT_1D receptors. Compounds that show at least a 1.5 log-unit difference in their affinities for rodent 5-HT_1B versus nonrodent 5-HT_1B/5-HT_1D-binding sites in mammalian brain membranes are summarized in Table7.

Recently, human 5-HT_1B/1D (Hamblin and Metcalf, 1991; Adham et al., 1992; Demchyshyn et al., 1992; Hamblin et al., 1992b; Jin et al., 1992; Levy et al., 1992; Weinshank et al., 1992) and rat 5-HT_1B receptors (Voigt et al., 1991; Adham et al., 1992; Hamblin et al., 1992a; Maroteaux et al., 1992) have been cloned. From these studies, it has become clear that the “old” 5-HT_1D pharmacological phenotype is conferred by two separate genes, i.e., a5-HT_1B gene cloned in nonrodents (Adham et al., 1992; Demchyshyn et al., 1992; Hamblin et al., 1992b;Jin et al., 1992; Levy et al., 1992; Weinshank et al., 1992) and a5-HT_1D gene cloned in rodents and nonrodents (Hamblin et al., 1992a; Weinshank et al., 1992). Although the overall amino acid similarity between 5-HT_1Dand nonrodent 5-HT_1B receptors, which share an almost identical pharmacology, is 61 to 63% (Hamblin et al., 1992b;Weinshank et al., 1992), nonrodent 5-HT_1B and rat 5-HT_1B receptors share a 93% similarity in their deduced amino acid sequence and yet have different pharmacological profiles. When expressed in heterologous systems, nonrodent 5-HT_1B/5-HT_1D and rodent 5-HT_1B receptors maintain the same pharmacological profile as in brain membranes: the former have a higher affinity for sumatriptan, 8-OH-DPAT, and α₂-adrenergic antagonists, and bind CP 93129 and arylalkylamine compounds such as propranolol and pindolol with lower affinity than the rodent 5-HT_1B subtype (Metcalf et al., 1992; Oksenberg et al., 1992; Parker et al., 1993;Adham et al., 1994). Interestingly, a single amino acid difference (asparagine versus threonine at position 355) has been shown to be responsible for rodent versus nonrodent distinct pharmacological profiles. Their pharmacological phenotypes may be interconverted by a single point mutation at amino acid 355 (Metcalf et al., 1992;Oksenberg et al., 1992; Parker et al., 1993; Adham et al., 1994). The relationship between 5-HT_1D, nonrodent 5-HT_1B and rodent 5-HT_1Breceptors has been summarized by Hartig et al. (1992): given the 70 to 80% similarity in their transmembrane domains, human 5-HT_1B/1D and rodent 5-HT_1B/1D receptors should be considered pairs of intraspecies receptor subtypes and members of the same gene product subfamily. Furthermore, given the high (95%) transmembrane homology between human 5-HT_1D and rat 5-HT_1D receptors, on the one hand, and corresponding 5-HT_1B receptors on the other, these two receptor pairs should be considered as species homologs. It was only recently that the 5-HT₂ receptor antagonists ritanserin and ketanserin were found to discriminate between 5-HT_1Dα and 5-HT_1Dβ (1.1–1.8 log-unit difference inK _D values, 5-HT_1D receptors having a higher affinity than nonrodent 5-HT_1B receptors for the 5-HT₂ antagonists (Doménech et al., 1994;Zgombick et al., 1995). Before, these receptors were initially thought to be pharmacologically indistinguishable. Although less selective, GR 127935 and metergoline also distinguish between the two receptors with higher affinity for the nonrodent 5-HT_1B subtype (1 and 0.6 log units, respectively; Skingle et al., 1991;Doménech et al., 1994). Moreover, in transected cells, GR 127935 further distinguishes 5-HT_1D from nonrodent 5-HT_1B receptors by eliciting a full agonistic response for the former and behaving as a silent antagonist for the latter (Pauwels and Colpaert, 1995; Pauwels and Palmier, 1995). However, although transfected systems constitute a useful first approach for studying receptor pharmacology, interpretation of functional data from these studies should be cautious since drug activity may vary according to the level of receptor expression. For example, the arylkylamines propranolol and pindolol, and the ergot derivative metergoline, which have been described as antagonists or partial agonists in in vivo or in vitro studies of terminal 5-HT autoreceptors, were found to behave as full agonists in cells expressing the rat or human 5-HT_1B receptor (Miller et al., 1992; Adham et al., 1993b). Hence, further development of selective 5-HT_1D/nonrodent 5-HT_1B compounds will depend on the availability of naturally occurring systems expressing the different receptors to allow drug evaluation. Hamel and coworkers have demonstrated that nonrodent 5-HT_1B receptors mediate 5-HT-induced contractions in cerebral arteries, establishing this tissue as an appropriate model of nonrodent 5-HT_1B receptor function (Hamel and Bouchard, 1991; Hamel et al., 1993). On the other hand, the low expression of 5-HT_1D relative to 5-HT_1B receptors in mammalian brain (Beer and Middlemiss, 1993; Bruinvels et al., 1993, 1994b; Doménech et al., 1994) has precluded identification of potentially 5-HT_1D-selective compounds using native systems.

The use of autoradiographic and in situ hybridization techniques has allowed us to compare not only the anatomical but also the cellular distribution of 5-HT_1B/1D-binding sites and 5-HT_1B/1D mRNAs (Bruinvels et al., 1994a; Doucet et al., 1995). Autoradiographic studies of¹²⁵I-cyanopindolol (Pazos et al., 1985) and, more recently, ¹²⁵I-GTI (Bruinvels et al., 1993) have revealed a particularly high concentration of 5-HT_1B-binding sites in striatum, substantia nigra, and dorsal subiculum of the rat brain. Yet no mRNA for 5-HT_1B receptors was found in the two latter regions (Voigt et al., 1991; Bruinvels et al., 1993; Doucet et al., 1995). Conversely, the DRN, which exhibited an intense mRNA hybridization signal, displayed low or no 5-HT_1Bbinding (Vergé et al., 1986; Voigt et al., 1991; Bruinvels et al., 1993; Doucet et al., 1995). Mismatches between 5-HT receptor protein and mRNA previously observed in mice have been explained by assuming that 5-HT_1B receptors are transported along fibers far from somas where they are synthesized (Boschert et al., 1994). For example, it has been suggested that 5-HT_1B receptors synthesized in the soma of CA₁ pyramidal neurons (CA₁pyramidal neurons are intensely labeled for mRNA; Doucet et al., 1995) are transported via their glutamatergic projections to reach the dorsal subiculum as heteroreceptors. Similarly, 5-HT_1Bheteroreceptors on striatal projections to the substantia nigra have been proposed to explain the high binding and low hybridization observed in this region (Bruinvels et al., 1994a). Moreover, 5-HT_1B receptors synthesized in the midbrain raphe could be transported via axons to reach terminal areas where they have been described as autoreceptors.

Autoradiographic studies on the human brain indicate that very high densities of 5-HT_1B/1D receptors labeled by¹²⁵I-GTI are present in the basal ganglia, and especially in substantia nigra (Bruinvels et al., 1991; Palacios et al., 1992). A similar population of¹²⁵I-GTI-labeled sites has also been identified in human cortical membranes (Beer and Middlemiss, 1993) and guinea pig basal ganglia (Bruinvels et al., 1994b). In the rat brain,¹²⁵I-GTI also labels 5-HT_1Bas well as 5-HT_1D sites. The latter have been defined as the non-5-HT_1B-¹²⁵I-GTI binding displaced by the 4[2-[4-[3-(trifluoromethyl)phenyl]1-piperazinyl]ethyl]benzeneamine (PAPP, which has more than 2 log units higher affinity for 5-HT_1D than 5-HT_1Breceptors; Schoeffter and Hoyer, 1989). These PAPP displaceable sites represent 15 and 11% of cortical and striatal iodine-labeled sites in the rat brain (Bruinvels et al., 1993). In the rat cortex, the percentage of total ¹²⁵I-GTI sites (5-HT_1B + 5-HT_1D) displaced by PAPP (designated 5-HT_1D sites; 15%) is in agreement with the percentage of total ¹²⁵I-GTI (5-HT_1B + 5-HT_1D) minus¹²⁵I-cytochrome P-450 (CYP; 5-HT_1B) in the same region (19%) and with the percentage of total 5-HT₁ binding represented by non-5-HT_1A/1B/2C (5-HT_1D) cortical sites reported by Herrick-Davis and Titeler (18%; 1988). In contrast, in the striatum, the sites defined by¹²⁵I-GTI (5-HT_1B + 5-HT_1D) minus ¹²⁵I-CYP (5-HT_1B) and non-5-HT_1A/1B/2C (5-HT_1D) sites were 20 and 30%, respectively (Herrick-Davis and Titeler, 1988;Bruinvels et al., 1993), representing more than double the PAPP displacement sites (11%) in this region. In the same study,Herrick-Davis and Titeler (1988) reported more than 40% of non-5-HT_1A/1B/2C sites with 5-HT_1D receptor pharmacology (5-HT > 5-CT ≫ TFMPP > 8-OH-DPAT ≫ (±)-pindolol) in midbrain membranes. On the other hand, Bruinvels et al. (1993) reported an average of 8% PAPP-defined 5-HT_1D sites in the same area, whereas the 5-HT_1B sites labeled by¹²⁵I-CYP in the dorsal raphe and central gray represented approximately 60% of the 5-HT_1B/1Dsites labeled by ¹²⁵I-GTI). These observations may indicate that PAPP does not identify all of the 5-HT_1D sites or that¹²⁵I-GTI labels other non-5-HT_1B/1D sites. Given that the non-5-HT_1A/1B/2C population described byHerrick-Davis and Titeler (1988) showed high affinity for 5-CT, it is improbable for them to be 5-HT_1E/1F sites. However, the presence of multiple receptors within this population cannot be ruled out, since Hill coefficients for sites that recognized 5-CT ranged from 0.62 to 0.86 (Herrick-Davis and Titeler, 1988). Hence, the above-mentioned observations suggest that¹²⁵I-GTI binding displaced by PAPP may not represent all of the 5-HT_1D receptors in the rat raphe area. Rodent midbrain, especially raphe nuclei, is one of the areas expressing the highest levels of 5-HT_1DmRNA (Hamblin et al., 1992a; Bruinvels et al., 1994a). Thus, the high coexpression of 5-HT_1D sites (according toHerrick-Davis and Titeler, 1988) and mRNA could be interpreted as further support for the role of 5-HT_1Dα as somatodendritic autoreceptors in the rat brain. Conversely, results from Bruinvels and coworkers (1993, 1994a,b) would support the idea that most of the 5-HT_1D or its mRNA are transported to terminal areas. On the other hand, there is evidence indicating a matching distribution of 5-HT_1Bsites and mRNA in the rat midbrain; Bruinvels et al., 1993, 1994b;Doucet et al., 1995). However, when tested in functional studies, at concentrations that do not alter basal outflow, the 5-HT_1B agonist CP 93129 did not modify 5-HT release (Piñeyro et al., 1995b). As previously stated, a possible interpretation for this set of observations would be that 5-HT_1B receptors located on afferent or recurrent axons and regulating 5-HT release from these secondary sources (as compared to somatodendritic release) account for the presence of the protein, whereas mRNA would be mainly directed to the synthesis of 5-HT_1B autoreceptors in terminal projection areas.

1. Administration of 5-HT Reuptake Blockers.

The reported effect of acute 5-HT reuptake blockade depends on the dose used and the region examined. At high doses (10 mg/kg i.p or s.c), the acute administration of fluoxetine, citalopram, or sertraline has been shown to induce an increase in extracellular 5-HT in terminal projection areas such as cortex, striatum, or diencephalon (Dailey et al., 1992;Invernizzi et al., 1992a; Perry and Fuller, 1992; Rutter and Auerbach, 1993, see Fuller, 1994). The increase in extracellular 5-HT is dependent on neuronal firing because it is blocked by TTX (Perry and Fuller, 1992) or 8-OH-DPAT (Rutter and Auerbach, 1993). The latter observation is somewhat puzzling given the fact that ED₅₀ of i.v. doses of different SSRIs to inhibit 5-HT neuron firing are within the 0.1 and 0.5 mg/kg range (Blier and de Montigny, 1980; Blier et al., 1984; Chaput et al., 1986b; Gartside et al., 1995; Hajós et al., 1995; Maudhuit et al., 1995; Kasamo et al., 1996), suggesting that at 10 mg/kg, no matter what SSRI is given, the result would be total shut-down of 5-HT neuron firing activity. Even if rats are much more rapid metabolizers than humans, a dose of 10 mg/kg is 15 to 35 times the therapeutic dose. Hence, other studies have been performed using much lower doses of SSRIs. In such cases, the systemic administration of 1 mg/kg citalopram or 32 μmol/kg sertraline (Invernizzi et al., 1991, 1992a) produced no increase in extracellular 5-HT in cortical projection areas, and experiments in which extracellular 5-HT was simultaneously measured in cortex and raphe nuclei reveal that, following systemic administration of these reuptake inhibitors, there is a preferential increase in extracellular 5-HT in the raphe region (Adell and Artigas, 1991; Bel and Artigas, 1992). It is this increase in somatodendritic extracellular 5-HT that activates the powerful 5-HT_1A autoreceptor feedback loop leading 5-HT neurons to establish their own “ceiling” on the extent to which uptake inhibitors increase extracellular 5-HT in terminal projection areas. Reasoning that long-term reuptake blockade induces a desensitization of somatodendritic 5-HT_1A autoreceptors, Bel and Artigas (1993)proposed the possibility to overcome negative feedback and increase extracellular availability of 5-HT in terminal projection areas even when using low doses of SSRIs. Indeed, they treated rats with 1 mg/kg fluvoxamine (s.c.) for 2 weeks, and at the end of this time period the increase in extracellular concentration of 5-HT in frontal cortex of treated rats, still carrying the osmotic minipump, was similar to that observed following a 10-mg/kg acute i.v. dose (Bel and Artigas, 1992,1993). These results have been confirmed in a study by Invernizzi et al. (1994), in which sustained treatment with citalopram (10 mg/kg/day for 14 days) facilitated the enhancing effect on terminal 5-HT release produced by a dose of 1 mg/kg of the SSRI administered 24 h after the end of the treatment. Moreover, this study also showed that the reducing effect of a systemic dose of 25 μg/kg 8-OH-DPAT on terminal 5-HT release was abolished following long-term citalopram administration, thus confirming a desensitization of 5-HT_1A autoreceptors (Invernizzi et al., 1994). On the other hand, Bosker et al. (1995a,b) reported that a 14- or 21-day treatment with oral or s.c. fluvoxamine (3 mg/kg or 6.5 mg/kg at study outset, respectively) did not produce desensitization of 5-HT_1A autoreceptors, facilitation of another dose of fluvoxamine, nor increase in hippocampal 5-HT output. If indeed 5-HT_1A autoreceptor desensitization may in part account for the antidepressant effect of prolonged SSRI administration, an acute reduction of the activation of the feedback mechanism should produce an effect similar to that of sustained SSRI treatment. This assumption has been verified in seven of eight placebo-controlled clinical trials showing that in depressed patients the combination of pindolol and an SSRI produced either a significant acceleration of the antidepressant response or a greater proportion of patients presenting a response at the end of the trial (Maes et al., 1996, 1999; Berman et al., 1997; Pérez et al., 1997; Tome et al., 1997; Zanardi et al., 1997, 1998; Bordet et al., 1998). These results are in agreement with results referred to in a previous section, in which blockade of 5-HT_1A autoreceptors was shown to increase extracellular availability of 5-HT in terminal projection areas after the acute administration of an SSRI.

Apart from inducing desensitization of somatodendritic 5-HT_1A autoreceptors, sustained paroxetine (10 mg/kg/day s.c. for 21 days) administration has been reported to decrease the effectiveness of 5-HT_1B/1D receptor activation in reducing electrically evoked release of 5-HT from rat and guinea pig midbrain raphe slices (El Mansari and Blier, 1996;Piñeyro and Blier, 1996). This same treatment, as well as fluoxetine (5 mg/kg/day i.p. for 21 days), induced an increase in electrically evoked 5-HT release from raphe slices (O’Connor and Kruk, 1994; El Mansari and Blier, 1996; Piñeyro and Blier, 1996), indicating that 5-HT receptors which negatively control 5-HT release in the somatodendritic area are less sensitive to activation by the endogenous neurotransmitter after sustained 5-HT uptake inhibition. In contrast, Bel and Artigas (1993) found no increase in basal extracellular 5-HT levels in the dorsal raphe of rats treated with fluvoxamine (1 mg/kg/day s.c. for 21 days).

Terminal 5-HT_1B/1D autoreceptors in different projection areas including hippocampus, hypothalamus, and orbito-frontal cortex have also been found to desensitize after sustained administration of SSRIs. Desensitization has been demonstrated by a reduced efficacy of 5-HT_1B/1Dagonists to inhibit evoked 5-HT release (Blier and de Montigny, 1983;Blier et al., 1984, 1998a,b; Chaput et al., 1986b; Moret and Briley, 1990; O’Connor and Kruk, 1994; Blier and Bouchard, 1994; El Mansari et al., 1995). Such a desensitization results in a greater release of 5-HT per action potential, as indicated by a greater inhibition of the firing activity of CA₃ pyramidal neurons following 5-HT pathway stimulation (in the absence of changes in postsynaptic receptor sensitivity; Blier and de Montigny, 1983; Blier et al., 1984, 1988a,b; Chaput et al., 1986b), as well as an increase in electrically evoked release of [³H]5-HT from preloaded slices of different projection areas, following a 48-h washout period after SSRI administration (Blier and Bouchard, 1994; El Mansari et al., 1995). Interestingly, not all terminal regions respond in a similar way to the same SSRI treatment, e.g., after a 21-day treatment with paroxetine (10 mg/kg/day s.c.), the inhibitory effect of the agonist 5-methoxytryptamine on the evoked release of 5-HT was attenuated in slices of hippocampus and hypothalamus but not of frontal or orbito-frontal cortex (Blier and Bouchard, 1994, El Mansari et al., 1995). In turn, of the two latter regions, only in the orbito-frontal cortex did desensitization occur after an 8-week treatment with paroxetine (10 mg/kg/day s.c.; El Mansari et al., 1995). It is still unclear whether these different results indicate that 5-HT neurons are endowed with different autoreceptor populations depending on the terminal region to which they project, or whether it is the local influences that determine the different adaptative properties of the same autoreceptor subtype. Sustained administration of a low dose of fluvoxamine was found to induce desensitization of somatodendritic 5-HT_1A receptors but not of 5-HT_1B/1D terminal autoreceptors, the sensitivity of which remained unchanged after fluoxetine (5 mg/kg/day for 21 days) or fluvoxamine (6.7 mg/kg/day for 28 days; Bel and Artigas, 1993;Bosker et al., 1995b; El Mansari et al., 1995) administration. Furthermore, chlomipramine (10 mg/kg/day s.c. for 21 days) did not produce a desensitization of 5-HT_1B terminal autoreceptors in rabbit hypothalamus (Schoups and De Potter, 1988), probably due to the fact that, although in vitro this drug is a potent and highly selective 5-HT reuptake blocker, in vivo it loses its selectivity as soon as it is degraded to chlordesipramine. It is thus possible that a high percentage or complete blockade of 5-HT reuptake sites should be achieved for terminal autoreceptor desensitization to occur.

The importance of terminal autoreceptor plasticity in ensuring an enhanced 5-HT transmission is further supported by the observation that prolonged administration of SSRIs reduces brain 5-HT (Hrdina, 1987;Caccia et al., 1992; Trouvin et al., 1993), indicating that after prolonged reuptake blockade, an increase in extracellular 5-HT availability occurs in the face of a reduction in the total 5-HT tissue content.

2. Prolonged Administration of MAOIs.

In this case the situation is reversed, brain 5-HT being actually increased after sustained blockade of MAO-A (Blier et al., 1986a,b; Celada and Artigas, 1993; Ferrer and Artigas, 1994). On the other hand, similar to SSRIs, the acute systemic administration of selective MAO-A or nonselective MAO-A/MAO-B inhibitors produces an immediate inhibition of 5-HT metabolism and a reduction in 5-HT neuron firing activity (Blier and de Montigny, 1985; Blier et al., 1986a,b). It is not surprising then that, when given acutely, MAOIs and SSRIs, produce a preferential increase in extracellular 5-HT in midbrain raphe nuclei as compared to terminal projection areas. In contrast, prolonged MAOI administration increases extracellular availability of 5-HT to a similar extent in pre- and postsynaptic projection areas (Celada and Artigas, 1993; Ferrer and Artigas, 1994; Bel and Artigas, 1995) with a time course similar to that of the desensitization of somatodendritic 5-HT_1A autoreceptors (Blier and de Montigny, 1985; Blier et al., 1986a,b; Piñeyro and Blier, 1996). In long-term experiments, tranylcypromine given at a dose that had no effect in acute experiments (0.5 mg/kg/day s.c. for 14 days) produced a greater increase in extracellular cortical availability of 5-HT than the acute administration of a dose six times higher, even if the increases in tissue 5-HT concentrations were of 40 and 700%, following the long-term/low-dose and acute/high-dose treatments, respectively. These observations suggest that even if the increase in intracellular 5-HT is almost 20-fold higher following an acute high dose of tranylcypromine, the neurotransmitter is trapped within 5-HT terminals, only a small part of it being available for release (intracellular:extracellular ratio of 5-HT increase in frontal cortex: 11.6 and 5.5 for acute/high-dose and prolonged/low-dose treatments, respectively). In part, increased extracellular availability of 5-HT after long-term MAOI administration is due to desensitization of 5-HT_1A autoreceptors and recovery of 5-HT neuron firing frequency. However, an increase in terminal 5-HT release (hippocampus, cortex, and hypothalamus) is seen not only in vivo in the whole animal, but also in vitro in slices containing only 5-HT terminals (Blier and de Montigny, 1985; Blier et al., 1986a,b; Blier and Bouchard, 1994; Mongeau et al., 1994), indicating a 5-HT_1A-independent enhancement of neurotransmitter release. Unlike long-term treatment with SSRIs, the sensitivity of 5-HT autoreceptors remains unchanged (Blier et al., 1986a,b; Blier and Bouchard, 1994). The question arising then is, what is the mechanism involved in increasing the releasable amount of 5-HT after long-term MAOI administration? It has long been known that the A form of MAO catalyzes the oxidative deamination not only of 5-HT but also of NE (Hall et al., 1969; Yang and Neff, 1973), and that 5-HT terminals are endowed with inhibitory α₂-adrenergic heteroreceptors (Göthert et al., 1980; Göthert et al., 1981; Maura et al., 1982). More recently, our laboratory has shown that enhanced 5-HT release after sustained MAO-A inhibition correlates with the production of α₂-adrenergic heteroreceptor desensitization by these treatments (Blier et al., 1986a,b; Blier and Bouchard, 1994;Mongeau et al., 1994). Moreover, destruction of the NE system impairs the heteroreceptor desensitization caused by the prolonged administration of the reversible MAO-A inhibitor befloxatone (Mongeau et al., 1994).

If acute administration of MAOIs induces a preferential increase of 5-HT release in midbrain raphe nuclei, then, as in the case of SSRIs, the combined administration of a 5-HT_1Aantagonist along with a MAOI should produce a greater enhancement of 5-HT availability in the extracellular space of projection areas than the MAOI by itself. Indeed, depressed patients treated with moclobemide or phenelzine and pindolol showed a reduction in the Hamilton Depression Rate Scale score within the first week of combined treatment (Artigas et al., 1994; Blier and Bergeron, 1995). On the other hand, in microdialysis studies in freely moving rats, the acute administration of supramaximal doses of the nonselective inhibitor tranylcypromine (15 mg/kg i.p. which increases motor activity; Celada and Artigas, 1993; Ferrer and Artigas., 1994) have been shown to produce a considerable increase (500–1100%) in extracellular cortical 5-HT, two to four times higher than the long-term administration of 0.5 mg/kg s.c. for 14 days (224% increase). Furthermore, the ratio of DRN:frontal cortex extracellular 5-HT was six after 15 mg/kg and one after 0.5 mg/kg, indicating that the observed 500 to 1100% increase in extracellular cortical 5-HT following the supramaximal dose of tranylcypromine takes place even with a full activation of the negative somatodendritic autoregulatory feedback. Similar results were observed with the acute administration of the nonselective MAOI pargyline (75 mg/kg i.p.) which elicited 10- to 14-fold increases in caudate-putamen and frontal cortex with the same time course as that seen with tranylcypromine (Kalen et al., 1988b; Carboni and Di Chiara, 1989). In the tranylcypromine series of experiments, tissue concentrations of 5-HT were found to reach a plateau at an acute dose of 3 mg/kg, with no further change at 15 mg/kg, whereas the cortical extracellular concentration of neurotransmitter increased 4-fold with the dose increase from 3 to 15 mg/kg (Ferrer and Artigas, 1994). The latter observation indicates that even if MAO-A activity is completely blocked at the low dose of tranylcypromine, it is not the only factor determining extracellular 5-HT availability. The latter interpretation is supported by the fact that the concurrent administration of either brofaromine (10 mg/kg s.c.) or clorgyline (5 mg/kg i.p.) with deprenyl (2.5 mg/kg i.p; Celada et al., 1994; Bel and Artigas, 1995) also increases the extracellular concentration of cortical 5-HT to a much greater extent than the MAO-A inhibitor by itself. This group of observations may be interpreted as an indication that extracellular changes in 5-HT availability are determined not only by MAO-A activity, but also by 5-HT deamination by MAO-B, as well as the capacity of the brain to store 5-HT.

The previous experiments also indicate that it is possible to induce an acute 250 to 400% increase in extracellular 5-HT concentration in projection areas by the concurrent inhibition of MAO-A and MAO-B. If such a strategy were to be applied to achieve a faster therapeutic response or as a potentiation strategy in patients treated with but not responding to MAOIs (i.e., used to induce a sufficient increase in extracellular 5-HT availability in projection areas without inducing dose-related side effects), various facts should be taken into account: 1) MAO-A (contained in catecholaminergic neurons and terminals; Westlund et al., 1985, 1988) is only 1.4 times more concentrated in DRN than in the cortex or hippocampus, whereas the concentration of MAO-B (contained in 5-HT neurons; Levitt et al., 1982;Westlund et al., 1985, 1988) in the midbrain raphe is 3 times higher than in cortico-hippocampal terminal areas (Saura et al., 1992); 2) MAO-B inhibition has no antidepressant effect by itself (Mann et al., 1984) but contribute to increase terminal 5-HT release when most, if not all, MAO-A activity is blocked; 3) it is possible to produce an almost complete MAO-A blockade without an excessive increase of DRN extracellular concentration of 5-HT (Ferrer and Artigas, 1994); 4) doses of MAO-A/MAO-B inhibitors chosen should favor a terminal versus raphe nuclei blockade of the enzymes; and 5) at doses where nonselective MAOIs produce a significant acute increase in extracellular 5-HT in projection areas, they might also induce collateral side effects. Hence, based on the latter, it could be possible that the minimal dose of selective or nonselective MAO-A inhibitor that completely blocks MAO-A activity (e.g., 3 mg/kg tranylcypromine), administered in combination with one-third of the equivalent dose of 2.5 mg/kg deprenyl used in rats by Artigas’ group (to avoid the disproportionate increase in DRN 5-HT), could induce an immediate and significant increase in extracellular 5-HT in projection areas. Under an equivalent treatment, a quicker therapeutic response could be expected but chances of 5-HT syndrome are higher.

One additional question that immediately arises is, how may the “speeding-up strategies” of SSRI/pindolol or MAOI/pindolol manage to overcome inhibitory terminal autoregulation by 5-HT autoreceptors? Given that supramaximal doses of chlomipramine and tranylcypromine produce a 300 to 1100% increase of the basal 2 nM cortical 5-HT concentration (Bel and Artigas, 1992) and that 5-HT concentrations between 10 and 100 nM inhibit 5-HT release from the rat frontal cortex (Middlemiss, 1986; Limberger et al., 1991), an insufficient increase in extracellular concentration of neurotransmitter does not seem an appropriate explanation. The possibility that terminal autoinhibition could in fact be activated but 5-HT still released is not in agreement with the observation that 100 nM 5-HT has been shown to produce 90% inhibition of electrically evoked release of 5-HT in frontal cortex slices (Limberger et al., 1991). Speculating further, it could also be possible that the amount of releasable 5-HT within the terminal vicinity is of such magnitude that even if autoinhibition is activated by the first bursts of arriving action potentials, the amount of initially released 5-HT is already high enough to increase its extracellular availability. Such an explanation would fit only in the case of MAOIs but not of SSRIs (which do not increase intracellular 5-HT availability).

Moreover, another interesting point to be considered is, why, if both prolonged SSRI and MAOI administration increase extracellular availability of 5-HT only following the former, are terminal autoreceptors desensitized? A possible answer to this question could be that by prolonging reuptake time, SSRIs produce a sustained increase in intrasynaptic 5-HT concentration. On the other hand, the increase produced by MAOIs occurs following the arrival of action potentials to the terminal, and hence extracellular 5-HT concentration increases in an “on-off” manner, probably not suitable for producing desensitization.

3. Antidepressants with α₂ Adrenoceptor Antagonistic Properties.

Antidepressants with α₂antagonists like mianserin and (±)-mirtazapine, whose long-term administration increases the duration of suppression of firing of CA₃ pyramidal neurons produced by 5-HT pathway stimulation, have also been shown to desensitize α₂-adrenergic heteroreceptors on 5-HT neurons (Mongeau et al., 1994; Haddjeri et al., 1997). Other studies of the neurochemical effects of long-term mianserin administration (Raiteri et al., 1983a; Schoups and De Potter, 1988) reported no change in the sensitivity of these receptors. A likely explanation for this discrepancy could be that in the latter studies mianserin was injected i.p., whereas in the studies in which they induced α₂-adrenergic heteroreceptor desensitization, antidepressant drugs were delivered continuously via osmotic minipumps implanted s.c.

4. Activation of 5-HT_1A Receptors by 5-HT_1A Agonists.

Direct activation of presynaptic 5-HT_1A receptors by 5-HT_1A agonists causes inhibition of 5-HT cell firing, synthesis, and release in forebrain areas (Blier and de Montigny, 1987; Hjorth and Magnusson, 1988; Sharp et al., 1989a;Schechter et al., 1990; Godbout et al., 1991). At postsynaptic receptors, biochemical and electrophysiological experiments in the hippocampus (Yocca and Maayani, 1985; Yocca et al., 1986; Andrade and Nicoll, 1987a,b) have shown that clinically available 5-HT_1A agonists such as buspirone (Glitz and Pohl, 1991) act as partial agonists and may inhibit cAMP formation pyramidal neuron firing activity. Since upon acute administration there is net decrease in forebrain extracellular availability of 5-HT and their beneficial clinical antidepressant and anxiolytic effects are usually not observed until a few weeks of administration (see Charney et al., 1990; Glitz and Pohl, 1991), it appears that the presynaptic effects of 5-HT_1A agonists override their postsynaptic actions. Based on electrophysiological results, it has been proposed that tolerance develops to the autoreceptor-mediated effects (Blier and de Montigny, 1987; Schechter et al., 1990; Godbout et al., 1991), and a combination of normal 5-HT firing activity along with simultaneous activation of postsynaptic normosensitive receptors by the drug and endogenous 5-HT may account for their therapeutic actions (see Haddjeri et al., 1998). Neurochemical studies demonstrating 5-HT_1A autoreceptor desensitization after sustained 5-HT_1A agonist administration have been less consistent, with positive (Kreiss and Lucki, 1992) and negative results (Sharp et al., 1993a; Söderpalm et al., 1993). A possible explanation for the discrepancy between electrophysiological and neurochemical studies could be that the sensitivity of 5-HT_1A autoreceptors was tested directly on 5-HT neurons in the former and systemically in the latter. As previously discussed, systemic administration of 5-HT_1A agonists may modify 5-HT neuron electrophysiological and neurochemical activities acting not only at presynaptic but also postsynaptic 5-HT_1Areceptors. Furthermore, the hypothesis generated by electrophysiological data predicts that concurrent administration of a 5-HT_1A agonist with a selective 5-HT_1A autoreceptor blocker should produce rapid enhancement of 5-HT neurotransmission (Blier and de Montigny, 1994), and indeed the administration of buspirone (20 mg/day) with pindolol (2.5 mg/kg thrice daily) produced quick reductions of depressive symptomatology in patients treated with this drug combination (Blier et al., 1997).

1. Molecular Characteristics of the 5-HT Transporter (SERT).

Two main strategies have been used in an attempt to identify the molecular characteristics of the SERT, i.e., biochemical purification and cloning of cDNA coding for the protein. Using digitonin as a detergent, the SERT in rat brain and human platelets has been solubilized and purified in a conformational state that retained a pharmacological profile almost identical with that observed in native membrane preparations (Biessen et al., 1990; Graham et al., 1991;Launay et al., 1992). In a further step, the human placental SERT has been reconstituted after purification, displaying not only the same antidepressant-binding profile as the native carrier but also NaCl-dependent 5-HT transport (Ramamoorthy et al., 1993b). The molecular weight of this functional protein isolated from human placenta is 300,000 (Ramamoorthy et al., 1993b). On the other hand, the other purified proteins have a molecular weight ranging between 55,000 and 78,000 (Biessen et al., 1990; Launay et al., 1992) and that of the cloned SERT itself is about 70,000 (Blakely et al., 1991; Lesch et al., 1993; Ramamoorthy et al., 1993b). Thus, in analogy with the Na⁺/glucose transporter (Stevens et al., 1990), it has been proposed that the SERT may exist as an homotetramer (Ramamoorthy et al., 1993b). However, although dimeric concatenated constructs of the transporter have been shown to possess 5-HT transport activity similar to the monomer’s, concatenated tetramers have substantially lower activity (Chang et al., 1994). The latter observation does not preclude the existence of functional tetramers since simple functional monomers or dimers could associate and efficiently transport 5-HT.

The recent development of site-specific antibodies has allowed further characterization of the SERT (Lawrence et al., 1995a,b; Qian et al., 1995; Ovalle et al., 1995; Wade et al., 1996). The molecular weight of SERT in immunoprecipitates was found to vary according to the structure of origin: rat platelets, 94,000; rat pulmonary membranes, 80,000; rat brain, 71,000 to 76,000; HeLa cells, 90,000 to 200,000 (Ovalle et al., 1995; Wade et al., 1996). Differential N-linked glycosylation has been claimed to account for different molecular weight of CNS and peripheral SERT units. In transfected HeLa cells, the inhibition of glycosylation changes the molecular weight of the transporter, shifting the 90-kDa SERT immunoreactivity band to its presumably unglycosylated state of 56 kDa. In contrast, the mobility of a 200-kDa form of the transporter remained unchanged following glycosylation inhibition. Because unglycosylated monomers have less tendency to aggregate, it has been suggested that the high-weight slow-mobility species may represent a transporter aggregate and that glycosylation may be involved in multimerization (Qian et al., 1995). If this interpretation is correct, the molecular weight of the placental purified transporter (Ramamoorthy et al., 1993b) also suggests the possibility of a tetrameric organization of the transporter (300,000/55,000–78.000 = 5.5–3.8).

Despite the differences in molecular weight, the human CNS and peripheral SERTs are polypeptides encoded by a single gene located on chromosome 17 (Lesch et al., 1993; Ramamoorthy et al., 1993b). The existence of a single hybridizing mRNA, as well as the identity of the cDNAs cloning brain and peripheral rat SERTs, suggest that this is also the case for rodents (Blakely et al., 1991, 1993; Hoffman et al., 1991). Although encoded by a single gene, in humans, unlike rodents, there are different types of SERT mRNAs that have been found in placenta, lung (Ramamoorthy et al., 1993b), and brain (Austin et al., 1994), the abundance of each mRNA species depending on the tissue of origin. The factors determining the expression of different SERT transcripts is presently unknown. In contrast with these results, Lesch et al. (1993) have reported a single hybridizing transcript for human mRNA. These dissenting observations could be explained by differences in probes used.

Analysis of the amino acid sequence of mammalian transporters for 5-HT, NE, and DA shows that 41% of their amino acid residues are identical, homology being highest at the 12 hydrophobic membrane-spanning domain levels, lower in the intracytoplasmic carboxyl- and amino-terminals, and particularly low in the large extracellular loop connecting TM3 and TM4, where the 5-HT carrier has two potential N-linked glycosylation sites (see Amara and Kuhar, 1993; Rudnick and Clark, 1993). Multiple putative phosphorylation sites by protein kinase C (PKC) and protein kinase A, consistent with rapid postranslational regulation of the SERT, are predominantly found in carboxyl and amino termini (Blakely et al., 1991; Hoffman et al., 1991; Lesch et al., 1993, Corey et al., 1994; Demchyshyn et al., 1994). Moreover, recent reports on the organization of the human SERT gene indicate that the latter is endowed with an upstream combination of positive and negative cis-acting elements, including the cAMP response element, that may regulate transcription activity via a promoter unit (Lesch et al., 1994; Heils et al., 1995).

2. The Mechanism of 5-HT Uptake.

The mechanism of 5-HT uptake has been thoroughly studied in platelets (Rudnick, 1977; Nelson and Rudnick, 1979, 1982), mouse brain plasma membrane vesicles (O’Reilly and Reith, 1988; Reith et al., 1989), human placenta brush-border membrane vesicles (Cool et al., 1990; Ramamoorthy et al., 1993b), and, more recently, after stable expression of cloned SERTs, in different expression systems (Blakely et al., 1991; Hoffman et al., 1991;Ramamoorthy et al., 1993a; Corey et al., 1994; Demchyshyn et al., 1994;Gu et al., 1994; Mager et al., 1994). 5-HT is the specific substrate for the transporter, and K _m values reported across different studies are summarized in Table8. Reported values for the turnover number (maximal number of 5-HT molecules carried by one transporter in 1 min) in different systems have also been variable: 500 5-HT molecules/porcine SERT in platelet membrane vesicles/min (Talvenheimo et al., 1979), 110 5-HT molecules/rSERT expressed in parental LLC-PK₁ cells/min (Gu et al., 1994), or 30 5-HT molecules/rSERT expressed in Xenopus oocytes/min (Mager et al., 1994). Tryptamine and its derivatives, as well as 5-HT derivatives and phenylethyamines such as (+)-amphetamine and PCA, are additional substrates for the SERT (Segonzac et al., 1984; Wölfel and Graefe 1992; Mager et al., 1994). On the other hand, tryptophan, 5-hydroxytryptophan, 5-HIAA, histamine, and the catecholamines NE and DA, at concentrations as high as 10 μM, do not significantly bind to this carrier (Ramamoorthy et al., 1993; Hoffman et al., 1991;Wölfel and Graefe, 1992; Corey et al., 1994; Barker and Blakely, 1995). However, higher concentrations of DA, 20 to 40 μM, have been reported not only to bind to the transporter but to exchange with [³H]5-HT (Ramamoorthy et al., 1993;Wölfel and Graefe, 1992; Corey et al., 1994).

For neurotransmitter influx to occur, all type I plasma membrane transporters, by definition, exhibit absolute requirement for Na⁺ in the external medium. The Na⁺ concentration gradient has been demonstrated to be the driving force for 5-HT uptake, and it may not be replaced by other cations. If the gradient is experimentally created, independently of Na⁺ ATPase activity, 5-HT uptake is insensitive to changes in the latter, indicating that 5-HT and Na⁺ fluxes are directly coupled by the transporter (Rudnick, 1977; Kanner and Bendaham, 1985; O’Reilly and Reith, 1988; Ramamoorthy et al., 1993). External Na⁺ increasesV _max and decreasesK _m for 5-HT (Cool et al., 1990). In peripheral SERTs, Na⁺-5-HT stoichiometry has been considered to be 1:1 (Talvenheimo et al., 1983; Cool et al., 1990). In brain membranes, the increase of 5-HT uptake with Na⁺ showed a Hill coefficient of 2, suggesting a requirement of two Na⁺ ions for a transport cycle (O’Reilly and Reith, 1988). More recently, Na⁺stoichiometry has been studied in stably expressed rSERT and an hyperbolic function consistent with a 1:1 stoichiometry was found (Gu et al., 1994). If the sodium gradient is kept, external Cl⁻ increasesV _max and decreasesK _m for 5-HT, fitting a Cl⁻:5-HT stoichiometry of 1:1 (Nelson and Rudnick, 1982; Cool et al., 1990; Gu et al., 1994). Intracellular K⁺ has also been shown to stimulate 5-HT uptake in platelets, placenta brush-border vesicles, and brain vesicles (Nelson and Rudnick, 1979; Reith et al., 1989; Cool et al., 1990). Only in the case of the neuronal transporter, H⁺ could not substitute for K⁺ in enhancing uptake (Reith et al., 1989). The overall stoichiometry of a 5-HT uptake cycle, as proposed by Rudnick and Clark (1993), is 5-HT:Na⁺:Cl⁻:K⁺, 1:1:1:1. In their mechanistic model, the authors assume that the transporter may behave like an ion channel with a gate at each face of the membrane, but with only one gate opening at a time. In a first step, the transporter binds Na⁺, Cl⁻, and 5-HT on the extracellular face of the membrane. To account for cotransport of the three species, it would undergo conformational change and open the intracellular gate only once all three of them are bound to the permeation channel. K⁺ countertransport consists of an inverse conformational change, reorienting the transporter to its “active uptake state” once the K⁺ ion is released to the extracellular space. Given that 5-HT is transported in its cationic form (5-HT⁺; Keyes and Rudnick, 1982; Rudnick et al., 1989), the 1:1:1:1 stoichiometry should give place to an electroneutral process. In agreement with this concept, in most studies, uptake of 5-HT by mammalian transporters was not affected by membrane potential (see Rudnick and Clark, 1993). However, studies of the rSERT present in basophilic leukemia cells (Kanner and Bendaham, 1985), stably expressed hSERT (Laezza et al., 1994), rSERT (Mager et al., 1994), and Drosophila SERT (Corey et al., 1994) indicate that 5-HT uptake may indeed be electrogenic since it depends on membrane potential and/or generates a transport-associated current. These observations would imply not only that voltage gradients across the membrane may regulate uptake, but also that electrogenic transporters may mediate nonvesicular 5-HT release.

3. Anatomical and Cellular Localization of the SERT.

In the periphery, the SERT is expressed in enteric 5-HT neurons (Wade et al., 1996) and non-neuronal cells such as mast cells (Gripenberg, 1976), crypt epithelial cells, and very discretely in enterochromaffin cells (Wade et al., 1996). It is also found in platelets (Rudnick, 1977; Qian et al., 1995), lung membranes (Qian et al., 1995), and maternal brush-border of syncytiotrophoblasts (Cool et al., 1990; Ramamoorthy et al., 1993b).

In the brain, the SERT has been radiolabeled with [³H]imipramine (Langer et al., 1980a,b;Dawson and Wamsley, 1983; Hrdina et al., 1985) and more selective 5-HT uptake inhibitors such as [³H]cyanoimipramine (Wölf et al., 1988; Kovachich et al., 1988; Soucy et al., 1994), [³H]paroxetine (Habert et al., 1985; de Souza and Kuyatt, 1987; Langer et al., 1987; Marcusson et al., 1988), and [³H]citalopram (D’Amato et al., 1987). Although in general the autoradiographic binding pattern of [³H]imipramine was found to be similar to that of [³H]paroxetine (Hrdina et al., 1990) and that of [³H]citalopram (Duncan et al., 1992), there are important regional differences in the density of imipramine- versus SSRI-labeled sites, the former showing a much higher density of binding in forebrain areas such as the cortex and hippocampus. The reason for this discrepancy has been attributed to the fact [³H]imipramine binds to two classes of sites, high and low affinity, but only the high-affinity ones seem related to 5-HT uptake (Moret and Briley, 1986; Marcusson et al., 1986; Hrdina 1987, 1988; see review in D’Amato et al., 1987). For this reason, tritiated SSRIs are the ligands of choice for labeling the brain 5-HT carrier in vitro. Recently, in vitro [³H]cyanoimipramine and [³H]citalopram autoradiograms have been compared to the innervation pattern of 5-HT neurons as marked by [³H]5-HT uptake in rat brain. A similar linear relationship was found for the labeling density of each of these ligands and the density of 5-HT innervation, further indicating the high sensitivity of both [³H]cyanoimipramine and [³H]citalopram to label the SERT (Descarries et al., 1995). [³H]Citalopram binding has also been compared with that of [³H]paroxetine in postmortem human brain tissue, and it was concluded that both of these drugs are highly selective ligands but, because of its higher affinity for the carrier, [³H]paroxetine was suggested as the radioligand of choice for in vitro studies (Arranz and Marcusson, 1994).

Although [³H]paroxetine may be the ligand of choice for in vitro labeling of the SERT, its in vivo distribution of binding resembles that of in vitro imipramine (Biegon and Mathis, 1993), and successful conversion of paroxetine into photon emission tomography (PET) or single photon emission-computerized tomography imaging agents has not been accomplished despite its very high potency for uptake inhibition (see Scheffel et al., 1992). Other drugs with high affinity for the uptake site, such as cyanoimipramine, sertraline, citalopram, or fluoxetine, have been labeled with¹¹C but also displayed relatively low specific to nonspecific binding ratios in vivo (Hashimoto et al., 1987; Lasne et al., 1989; Scheffel and Ricaurte, 1990; Hume et al., 1991). McN-5652-Z (trans-1,2,4,5,6,10β-hexahydro-6-[4-(methylthio)phenyl]pyrrolo[2,1-a]isoquinoline) is another potent blocker of 5-HT uptake that has been¹¹C-tagged and assessed as a PET radiotracer in mouse brain (Shank et al., 1988). This study suggested that McN-5652-Z may label 5-HT uptake sites in vivo with high target:nontarget ratio, holding as a promising radiotracer for human PET studies (Suehiro et al., 1993). In vivo imaging of the 5-HT carrier has also been assessed in rat and nonhuman primate brain with the cocaine analog [(¹²³I]β-CIT ([¹²³I]methyl-3β-(-4-iodophenyl)tropane-2β-carboxylate;Scheffel et al., 1992; Laruelle et al., 1993). Although this cocaine analog binds to the DA transporter as well as the SERT, both sites may be discriminated because of kinetic differences in the way the ligand is taken up or washed out from rich 5-HT and DA innervation areas (Laruelle et al., 1993). Indeed, in a recent PET study in healthy human volunteers, the SERT was distinctly recognized in the medial frontal cortex, brainstem, hypothalamic area, and visual occipital cortex 1 h after injection, and DA transporters were recognized in the basal ganglia 20 h later (Kuikka et al., 1995). In this same study, 5-HT uptake sites were found to be reduced in the frontal cortex of a patient with depression and increased in the occipital cortex of a patient with panic disorder.

Cellular localization of SERTs in the CNS has been accomplished by using site-specific antibodies (Lawrence et al., 1995a,b; Ovalle et al., 1995; Qian et al., 1995). Immunocytochemistry using antibodies directed against sites on the second and third extracellular loops of the 5-HT carrier revealed both neuronal and glial staining in areas of the rat brain containing 5-HT somata and terminals (dorsal raphe and hippocampus) (Lawrence et al., 1995b). In contrast, Qian et al. (1995), using an antibody developed against the intracellular N terminus, found no evidence of glial staining. In this case, SERT-immunoreactive somata and a dense network of SERT-immunoreactive processes were observed in the DRN and on 5-HT terminals in the CA_2–3region of the hippocampus (Qian et al., 1995). It may then be possible that either the glial SERT expression in adult rat brain is not very abundant or the epitope on the N terminus is not expressed (or masked) in the glial SERT. Thus, SERT expression in adult brain astrocytes remains a matter of debate. Using colocalization techniques for glial fibrillary acidic protein and radioactivity for [³H]5-HT, 5-HT uptake activity has been found in primary astrocyte culture (Katz and Kimelberg, 1985; Kimelberg and Katz, 1985) and in 50% (frontal cortex) to 80% (periventricular region) of adult rat brain astrocytes (Anderson et al., 1992). Furthermore, Artigas et al. (1995) found that, after intracortical perfusion of 5-HT, 5,7-dihydroxytryptamine (5,7-DHT)-pretreated rats displayed in vivo SSRI-sensitive 5-HT uptake similar to that of control rats. Our group, on the other hand, using an electrophysiological and in vitro uptake paradigm, found no effect of paroxetine in 5,7-DHT-treated rats (Piñeyro et al., 1994). Our results are in agreement with in situ hybridization studies that have failed to detect any hybridization signal for mRNA in glial cells (Fujita et al., 1993). Like SERT immunoreactivity, SERT mRNA is present in neurons of caudal linear nucleus, DRN, MRN, and caudal 5-HT nuclei, matching the distribution of cell bodies but not that of terminals (Fujita et al., 1993; Austin et al., 1994). Only in one study, in which reverse transcription-polymerase chain reaction was used for amplification, a detectable level of SERT mRNA expression was found in the frontal cortex, hippocampus, and neostriatum, apart from the abundant expression observed in the midbrain raphe complex (Lesch et al., 1993). Such an observation indicates that SERT mRNA may be specifically targeted to different projection areas where it can initiate transporter synthesis, and this opens for consideration the possibility that mRNA transcription could be locally regulated.

4. Pharmacological Properties of the SERT.

The SERT is the pharmacological target for various therapeutic and abused substances. Compounds that block 5-HT reuptake such as tricyclic antidepressants and SSRIs are in the first group, whereas stimulants such as amphetamine and its derivatives, which block 5-HT uptake and promote release, are part of the latter category. Cocaine, although it binds and blocks the 5-HT carrier, is believed to exert most of its behavioral effects by blocking DA rather than 5-HT uptake (Woolverton and Kleven, 1992; Barker and Blakely, 1995; Caron, 1996).

5-HT and NE transporters share their sensitivity for tricyclic antidepressants, tertiary amines such as imipramine and chlomipramine being more potent at the SERT, and secondary amine tricyclics (e.g., desipramine, nortriptyline) at the NE transporter (see Hyttel, 1982;Thomas et al., 1987; Langer and Schoemaker, 1988; Bolden-Watson and Richelson, 1993). [³H]Imipramine-binding sites were among the first antidepressant-binding sites to be described both on 5-HT neurons and platelets (Raisman et al., 1979; Langer et al., 1980a,b). Based on early observations in which the binding of this radioligand was inhibited in a complex manner by 5-HT and SSRIs but competitively by imipramine itself (Langer and Raisman, 1983; Sette et al., 1983), it was initially proposed that imipramine-like antidepressants (imipramine, amitriptyline, or chlomipramine) would act by allosterically regulating the function of the transporter without directly binding to the substrate recognition site (Langer and Raisman, 1983). Conversely, the observation that imipramine and other tricyclics as well as SSRIs inhibit the binding of the selective SERT ligand [³H]paroxetine in a competitive manner (Habert et al., 1985; Marcusson and Eriksson, 1988; Graham et al., 1989;Marcusson et al., 1989, 1992) later suggested that there is a single or at least overlapping binding site for tricyclic and nontricyclic 5-HT uptake inhibitors on the SERT. The idea of an overlapping binding site for the substrate, tricyclic and nontricyclic uptake inhibitors is further supported by the following observations: 1) in membranes from rat and human brain, 5-HT produces competitive displacement not only of [³H]imipramine but also of [³H]paroxetine which in turn may be displaced, fitting a single-site binding model by citalopram, norzimelidine, paroxetine, fluoxetine, indalpine, chlomipramine, and desipramine (Habert et al., 1985; Marcusson et al., 1988, 1989, 1992); 2) [³H]5-HT uptake by dSERT is inhibited in a monophasic manner by paroxetine > fluoxetine > citalopram > cocaine³ 5-HT > desipramine > imipramine (Demchyshyn et al., 1994); 3) Na⁺ ions are needed for paroxetine and imipramine binding, as well as for 5-HT binding and translocation (Wood et al., 1986; Mann and Hrdina, 1992); and 4) preincubation with imipramine protects against the reduction in total [³H]paroxetine binding caused by the sulfydryl group alkylating agent NEM (Graham et al., 1989). It is possible then that the early observations by Langer’s group could be explained by taking into account the existence of two components for [³H]imipramine-specific binding, i.e., high-affinity, Na⁺-dependent binding and low-affinity, Na⁺-independent binding, as defined using desipramine to determine nonspecific binding (Hrdina, 1984, 1987,1988). Indeed, the high-affinity, Na⁺-dependent component of [³H]imipramine binding is completely displaced by 5-HT and nontricyclic reuptake blockers in brain and platelets (Marcusson et al., 1986; Hrdina, 1988; Humphreys et al., 1988).

Even if the above-mentioned findings support the existence of a common recognition site for 5-HT, tricyclics, and nontricyclic antidepressants, there is also compelling evidence indicating that, although overlapping, these sites are distinct, sharing some but not all interacting chemical groups: 1) chemical modification of the platelet SERT withN-ethoxycarbonyl-2-ethoxy-1,2-dihydroquinoline, a reagent that links carbonyl moieties to vicinal amino groups, significantly reduces the total number of [³H]imipramine-binding sites, an effect that is prevented by preincubation with imipramine and 5-HT but not fluoxetine and citalopram (Biessen et al., 1988); 2) oxidation of the SERT in human platelets by the thiol reagent phenylarsine oxide reduces [³H]imipramine binding by 90%, an effect that is prevented by the preincubation with the tricyclic drugs imipramine, cyanoimipramine, chlomipramine, and amitriptyline but not by the nontricyclic reuptake blockers citalopram, fluoxetine, femoxetine, and zimelidine (Biessen et al., 1988). On the other hand, in rat cortical membranes, preincubation with imipramine did protect [³H]paroxetine-binding sites from NEM inactivation (Graham et al., 1989); 3) the sulfydryl-reducing agent dithiothreitol increases the affinity of the human platelet SERT for [³H]imipramine but not for [³H]paroxetine (Tarrant and Williams, 1995); 4) antisera directed against the second extracellular loop of the SERT produce a dose-dependent inhibition of [³H]5-HT uptake but have no effect on [³H]citalopram binding (Lawrence et al., 1995b); and 5) incubation of [³H]citalopram, [³H]imipramine, or [³H]paroxetine in the presence of high micromolar concentrations of 5-HT, citalopram, or paroxetine may induce very different types of changes in the dissociation kinetics of each radioligand, e.g., 200 μM citalopram attenuated the dissociation of [³H]citalopram four times more than that of [³H]paroxetine, whereas paroxetine has an opposite effect, increasing the dissociation rate of [³H]imipramine (Wennogle and Meyerson, 1985;Humphreys et al., 1988; Plenge et al., 1990, 1991; Plenge and Mellerup, 1991). This last set of observations is consistent with the existence of low-affinity sites that may modulate the binding status of the high-affinity site. Further support for allosterism is given by the fact that paroxetine (although in the low nanomolar range) may decrease the affinity of the SERT for [³H]cocaine (Akunne et al., 1992). Cocaine, in turn, binds to a site that may be distinguished from the substrate/antidepressant site because binding of the latter but not that of 5-HT or antidepressants is insensible to Cl⁻ and inhibited by H⁺(Wall et al., 1993). Conversely, the ability of 5-HT to competitively displace cocaine analogs such as 2-β-[³H]carbomethoxy-3-β-[4-fluorophenyl]tropane and β-[¹²⁵I]citalopram argues that cocaine and substrate sites, if not the same, are closely related (Rudnick and Wall, 1991; Wall et al., 1993). Other drugs of abuse such as neurotoxic amphetamine derivatives (MDMA, MDA, PCA) also induce competitive displacement of [³H]imipramine and inhibition of [³H]5-HT transport (Rudnick and Wall, 1992a). Site-directed mutagenesis and chimera construction should help determine which amino acid residues on the transporter interact with all or most of the drugs and which are unique to each ligand. Such studies have suggested that 1) phenylalanine 586 located on the 12th transmembrane domain could be responsible for high-affinity recognition of imipramine (present in human but notDrosophila SERT); 2) (+)-amphetamine interacts with multiple residues in this same transmembrane domain; and 3) contact sites for citalopram have been localized to the second transmembrane domain (Barker et al., 1995).

5. Tianeptine, a Class by Itself?

Tianeptine is a tricyclic agent (dibenzothiazepine nucleus) with a long (aminoheptanoic acid) lateral chain (Labrid et al., 1988). In France, it is indicated for the treatment of “neurotic and reactional depressive conditions” and has been claimed to be a unique type of antidepressant that produces its effect by enhancing 5-HT uptake.

In rat cortical and hippocampal synaptosomes (Mennini et al., 1987;Fattaccini et al., 1990), as well as in rat and human platelets (Kato and Weitsch, 1988; Chambda et al., 1991), the increase in 5-HT uptake induced by tianeptine is secondary to a 20 to 30% increase inV _max. Tianeptine does not displace [³H]paroxetine, [³H]imipramine, nor [³H]d-fenfluramine and does not produce in vitro effects on 5-HT uptake (Kato and Weitsch, 1988). The increase in V _max observed ex vivo at least 1 h after acute administration of tianeptine (Mennini et al., 1987) should then be an indirect effect. This interpretation is supported by the observation that, if given 1 h before sacrifice at a dose similar to the one used in ex vivo experiments (10 mg/kg i.p.) in which V _max is increased, tianeptine does not modify [³H]imipramine binding to rat cortical membranes (Romero et al., 1992). Also, when given acutely (10 mg/kg) or chronically (10 or 20 mg/kg/day for 14 days), tianeptine has no significant effect on 5-HT uptake in mesencephalic synaptosomes where SERTs are most abundant (Mennini et al., 1987). In keeping with this observation, neither acute nor sustained tianeptine administration modify 5-HT neuron firing frequency (Dresse and Scuvée-Moreau, 1988; Piñeyro et al., 1995c). Furthermore, electrophysiological data from our laboratory also indicate that the increase in the firing activity of hippocampal pyramidal neurons after acute administration of tianeptine does not depend on the integrity of 5-HT terminals nor the presence of SERTs, since the effect of the drug is not modified by 5,7-DHT lesions (Piñeyro et al., 1995c). Moreover, analysis of the effect of sustained tianeptine administration also indicates that its effect on 5-HT uptake is not always reproducible. Sustained tianeptine administration increases 5-HT uptake in rat platelets and brain synaptosomes (10–20 mg/kg/day for 14 days; Mennini et al., 1987; Kato and Weitsch, 1988) but not in rat brain slices (20 mg/kg/day for 14 days; Piñeyro et al., 1995d) or human platelets (37.5 mg/kg/day for 10 or 28 days; Chambda et al., 1991). In the brain, the 5-HT uptake-enhancing effect of tianeptine occurs after a 72-h but not a 24-h washout (Mennini et al., 1987; Mocaër et al., 1988). In platelets, on the other hand, a 24-h washout allows the demonstration of a 30% increase in V _max (Kato and Weitsch, 1988). Results from certain binding studies indicate that doses of 10 to 20 mg/kg/day for 14 days (which may enhance cortical and hippocampal 5-HT uptake) produce no change in hippocampal [³H]imipramine- or [³H]paroxetine-binding parameters (Mennini et al., 1987; Mennini and Garattini, 1991; Frankfurt et al., 1993), whereas other studies have shown not only a decrease inB _max for [³H]paroxetine-binding sites in the cortex, hippocampus, and DRN (Watanabe et al., 1993; Kuroda et al., 1994) but also a decrease in SERT mRNA in DRN (Kuroda et al., 1994). Prenatal exposure to tianeptine (20 mg/kg/day for 14 days) has also been shown to reduce B _max for [³H]imipramine in rat cortex (Romero et al., 1992).

Except for the above-mentioned biochemical studies that have assessed the actual effect of tianeptine on 5-HT uptake, evidence supporting the idea that this drug enhances 5-HT reuptake is indirect. One hour after its acute administration (10 mg/kg i.p.), tianeptine was shown to increase brain tissue concentration of 5-HIAA without modifying tissue levels of 5-HT (Fattaccini et al., 1990). At a similar dose, it has also been shown to induce an increase in extracellular 5-HIAA in the hippocampus, hypothalamus, and medullary dorsal horn (de Simoni et al., 1992; Puig et al., 1993). Interestingly, a dose of 20 mg/kg (i.p.) had the opposite effect on extracellular 5-HIAA in the rat hippocampus (Mennini and Garattini, 1991) but increased plasma 5-HIAA levels (Ortiz et al., 1991). The observed increases in 5-HIAA have been interpreted as an increase in intracellular 5-HT turnover secondary to enhanced 5-HT uptake. The opposite effects of 10-mg/kg and 20-mg/kg doses in the hippocampus remain unexplained. Another common approach that has been used to unveil the site of action of tianeptine is its interaction with drugs that are known to modify 5-HT reuptake activity (Fattaccini et al., 1990; Ortiz et al., 1991; de Simoni et al., 1992; Datla and Curzon, 1993). In keeping with its 5-HT uptake-enhancing capacity, tianeptine, given 30 min after different SSRIs or in combination with 5-hydroxytryptophan (5-HTP), respectively reduced or prevented the increase in plasma or extracellular brain 5-HT caused by the aforementioned treatments. In both cases, consistent with the interpretation that tianeptine increases 5-HT uptake and intracellular deamination, its administration potentiated the increase in 5-HIAA, respectively, induced by SSRIs in plasma or 5-HTP in the extracellular cortical fluid (Ortiz et al., 1991; Datla and Curzon, 1993). On the other hand, opposite results have also been observed. In the rat hippocampus, as in many other brain regions, SSRI administration reduced extracellular availability of 5-HIAA, and tianeptine had no effect on this reduction (de Simoni et al., 1992). Furthermore, in the same study, a metabolite of the 5-HT-releasing drug fenfluramine produced the same increase in 5-HIAA independent of whether rats had been pretreated with tianeptine or not (de Simoni et al., 1992). Acute tianeptine administration was also without effect on the 5-HT depletion caused by d-fenfluramine in the cortex and striatum (Fattaccini et al., 1990) and on the increase caused by paroxetine in the time it takes CA₃ pyramidal neurons to recover their firing activity following microiontophoretic application of 5-HT (Piñeyro et al., 1995c). Lack of interaction between tianeptine and drugs that modify SERT activity should be also confronted with yet other findings suggesting that such an interaction may exist: 1) tianeptine no longer induced an increase in tissue 5-HIAA when fenfluramine was previously administered (Fattaccini et al., 1990), 2) in the rat hippocampus the administration of sertraline after tianeptine reverted the enhancing effect of the latter on 5-HIAA production (de Simoni et al., 1992), and 3) after its sustained administration, tianeptine antagonizes the increase in the time of recovery of firing of CA₃ pyramidal from microiontophoretic applications of 5-HT produced by the SSRI paroxetine (Piñeyro et al., 1995d). The effect of tianeptine on basal 5-HT release has also been assessed, and these experiments indicate that neither its acute nor its sustained administration has an effect on this parameter (Mennini et al., 1987; Whitton et al., 1991b). However, both acute and sustained administration reduce K⁺-induced 5-HT release from the rat brain in vitro (IC₅₀ 2 μM in cortex and 0.4 μM in hypothalamus) and in vivo (Mocaër et al., 1988; Whitton et al., 1991b). Most interestingly, as reported by Mocaër et al. (1988), the in vitro effect of tianeptine is partially blocked by methiothepin (1 μM), once again suggesting that a straight-forward interpretation of a decreased 5-HT output due to increased 5-HT uptake may not be the only explanation possible for this effect. Also supporting this view,Bolaños-Jiménez et al. (1993) have recently reported that tianeptine dose-dependently reduced the effect of the 5-HT_1B agonist CGS 12066B on 5-HT release. This group has also shown that without modifying basal outflow, tianeptine (100 μM) may inhibit [³H]ACh release from hippocampal rat synaptosomes (Bolaños-Jiménez et al., 1993). Interestingly, in keeping with the high concentrations used in this study, doses of 30 mg/kg i.p. but not lower ones were found to inhibit in vivo ACh release from the rat hippocampus (Bertorelli et al., 1992).

From the previous analysis, it can be concluded that there is no simple explanation for the effects of tianeptine on the 5-HT system. One of the most consistent observations is that acute as well as prolonged administration of tianeptine may reduce 5-HT neurotransmission when the latter is enhanced by 5-HTP, K⁺-evoked release, or SSRIs. Moreover, in the study by Mennini et al. (1987), in which cortical and hippocampal synaptosomes were shown for the first time to increase V _max, the lowest 5-HT concentration used was 40 nM (at least 15 times higher than the normal extracellular brain concentration of 5-HT), and in that by Fattaccini et al. (1990), the [³H]5-HT concentration was at least 5 times higher than the basal extracellular 5-HT. These neurochemical observations are supported by behavioral studies in which tianeptine has been shown to reduce some of the symptoms of 5-HTP-induced 5-HT syndrome (de Simoni et al., 1992; Datla and Curzon, 1993). Moreover, tianeptine has been shown to attenuate behaviors that have been attributed to stress-induced increases in 5-HT activity, i.e., it attenuates stress-induced open-field behavioral deficits without altering basal locomotion parameters (Broqua et al., 1992; Whitton et al., 1991a; Fontanges et al., 1993). It also abolishes stress-induced decreases in hypothalamic corticotrophin-releasing factor (Delbende et al., 1994), it suppresses decreases in glucocorticoid type I hippocampal receptors induced by isolation rearing without affecting basal levels of the latter nor type II receptors (McEwen, 1991, 1992), and it prevents stress- and corticosterone-induced reduction in CA₃ pyramidal neuron apical tree (Watanabe et al., 1992, 1993). Tianeptine has not been found to alter basal ACTH or corticosteroid plasma levels (Delbende et al., 1993, 1994; Watanabe et al., 1993), nor to modify the efficacy of 5-HT neurotransmission in basal conditions (Piñeyro et al., 1995c). On the other hand, it reduces the recovery time necessary for CA₃ pyramidal neurons to recover their firing activity following microiontophoretic applications of high currents of concentrated 5-HT (Piñeyro et al., 1995c).

The next question that naturally arises is whether the “protective” effect of tianeptine against stress-induced changes is linked to its uptake-enhancing capacity. A recent study by Mennini et al. (1993)directly addressed this question, indicating that tianeptine (10 mg/kg i.p. 1 h before stress) antagonizes the decrease in 5-HT uptake caused by acute noise stress. However, tianeptine has also been shown to induce specific changes in other brain monoamines involved in the response to stress: 1) acute tianeptine administration decreases the firing activity of NA neurons in locus ceruleus without altering the activity of DRN 5-HT neurons (Dresse and Scuvée-Moreau, 1988); 2) without producing marked changes in the 5-HT system, short-term administration of tianeptine (10 mg/kg/day for 4 days) increases NE content and decreases NE turnover in specific nuclei related with mood, i.e., preoptic area, DRN, and sensory cortex (Frankfurt et al., 1994); 3) its sustained administration, like that of desipramine, antagonizes the stress-induced increase in tyrosine hydroxylase mRNA in the NA nuclei (see McEwen, 1991); and 4) acute and prolonged tianeptine treatment may increase extracellular DA concentrations in striatum and nucleus accumbens in a 5-HT-independent manner (Invernizzi et al., 1992), as well as increase DA turnover in the prefrontal cortex (Louilot et al., 1990). Hence, these results further indicate that an interaction with the 5-HT system may not be the only mechanism by which tianeptine attains its stress-protecting effects.

Finally, clinical proof of the antidepressant efficacy of tianeptine has been recently reviewed by Wilde and Benfield (1995). It was concluded that, according to the available information, the antidepressant efficacy of this new drug, administered in the short term, appears similar to that of amitriptyline, imipramine, and fluoxetine, and in patients with coexisting anxiety and depression, tianeptine could be superior to maprotiline. However, it should be noticed that 1) most of the studies lacked a placebo-controlled group; 2) in amitriptyline and maproptyline studies, as well as in some of those in which tianeptine was compared with imipramine, optimal dosages of the standard drugs were not used; 3) in multicenter double-blind studies, in which the efficacy of long-term tianeptine treatment was assessed, efficacy was evaluated taking into account only those patients who completed the treatment, even if 65% of dropouts under tianeptine treatment abandoned the study due to ineffective treatment; and 4) comparative trials of tianeptine and SSRIs other than fluoxetine are needed to further define its role in the treatment of depression. Based on basic and clinical facts, it appears that it may be premature to claim that “tianeptine is a new antidepressant class with a unique 5-HT uptake enhancing profile”. Furthermore, if indeed 5-HT uptake is stimulated by the drug in conditions in which 5-HT neurotransmission is enhanced, then it would seem unlikely that this be the mechanism involved in its antidepressant action since depression is a condition in which the 5-HT system is frequently deficient (see Maes and Meltzer, 1995). It may, however, protect against the deleterious effects that different stressful conditions may impose on the 5-HT system of a depressed patient.