Address correspondence and reprint requests to Dr. M. Popoli at Center of Neuropharmacology, Institute of Pharmacological Sciences, University of Milan, Via Balzaretti 9, 20133 Milano, Italy. E-mail: email@example.com
Abstract: Depression has been treated pharmacologically for over three decades, but the views regarding the mechanism of action of antidepressant drugs have registered recently a major change. It was increasingly appreciated that adaptive changes in postreceptor signaling pathways, rather than primary action of drugs on monoamine transporters, metabolic enzymes, and receptors, are connected to therapeutic effect. For some of the various signaling pathways affected by antidepressant treatment, it was shown that protein phosphorylation, which represents an obligate step for most pathways, is markedly affected by long-term treatment. Changes were reported to be induced in the function of protein kinase C, cyclic AMP-dependent protein kinase, and calcium/calmodulin-dependent protein kinase. For two of these kinases (cyclic AMP- and calcium/calmodulin-dependent), the changes have been studied in isolated neuronal compartments (microtubules and presynaptic terminals). Antidepressant treatment activates the two kinases and increases the endogenous phosphorylation of selected substrates (microtubule-associated protein 2 and synaptotagmin). These modifications may be partly responsible for the changes induced by antidepressants in neurotransmission. The changes in protein phosphorylation induced by long-term antidepressant treatment may contribute to explain the therapeutic action of antidepressants and suggest new strategies of pharmacological intervention.
It may be a matter of surprise for many that the action of antidepressant drugs (ADs), which represent a large and increasing share of the public health expenditure, is not yet (or no longer) explained by a definite mechanism (Sandler, 1998). The initial hypothesis on the action of these drugs was based mainly on the primary effect of tricyclics and monoamine oxidase (MAO) inhibitors (see Table 1) on monoamine transporters and MAOs, respectively (the so-called monoamine hypothesis) (Bunney and Davis, 1965; Schildkraut, 1965). This hypothesis stated that increased availability of monoamine neurotransmitters at synaptic sites and the consequent receptor changes would compensate per se a deficit in monoaminergic transmission and cause therapeutic effect. However, although a useful model, this hypothesis soon was revealed as simplistic. A main reason for this was the temporal discrepancy between the timing of the primary biochemical effect of drugs (minutes, hours) and the onset of therapeutic action (weeks); another reason was the lack of clear correlation between the depressive state and depletion of monoamines. Other reasons in more recent times were the findings of a multitude of effects of AD treatment in postreceptor signaling and the synthesis and use by drug companies of several atypical antidepressants, which do not share a common primary mechanism with typical compounds and yet show therapeutic efficacy. Even the hypothesis that therapeutic efficacy is associated with desensitization of β-adrenoceptors (the most consistent effect found with AD treatment) had to be dismissed lately because of its inconsistency with newer findings (Sanders-Bush et al., 1985; Blier and de Montigny, 1994; Hyman and Nestler, 1996).
Table 1. Primary mechanism of action of principal ADs
MAOI, monoamine oxidase (MAO) inhibitors; SSRI, selective serotonin (5-HT) reuptake inhibitors; SNRI, serotonin and noradrenaline (NA) reuptake inhibitors; NARI, noradrenaline reuptake inhibitors; NaSSA, noradrenergic and specific serotonergic ADs; SARI, serotonin antagonists and reuptake inhibitors. Abbreviations are taken from Stahl (1997).
Irreversible inhibition of MAO A and B, or selective and reversible inhibition of MAO A
Nonselective inhibition of NA, 5-HT, and dopamine reuptake
Selective inhibition of 5-HT reuptake
Selective inhibition of NA and 5-HT reuptake
Selective inhibition of NA reuptake
Enhancement of NA and 5-HT release by blockade of presynaptic α2-adrenergic receptor
Blockade of 5-HT2 receptor and inhibition of 5-HT reuptake
However, there is now a large body of recent preclinical evidence and clinical experience allowing a new systematization of the field and formulation of updated hypotheses (Blier and de Montigny, 1994; Artigas et al., 1996; Hyman and Nestler, 1996; Duman, 1998). A basic assumption is that correlation with therapeutic efficacy should be established only for cellular or molecular events developing over time with the treatment, and not with acute drug effects. Another widely shared assumption is that antidepressant effect cannot be accounted for by a single intraneuronal regulatory event (Blier and de Montigny, 1994; Artigas et al., 1996; Hyman and Nestler, 1996). Rather, there is increasing evidence that several sites in postreceptor signaling cascades are involved in adaptive phenomena related to therapeutic effect. In the present theoretical framework, the different sites may represent various steps of intraneuronal adaptive processes, connected to each other and occurring in a temporal sequence roughly corresponding to the onset of drug action. In the last decade, several such sites have been identified, including (a) receptor-coupled G proteins (changes in expression and activity), (b) neuronal compartment-specific protein kinase activation (treated in detail below), and (c) gene expression [changes in transcription factors, e.g., cyclic AMP (cAMP) response element binding protein (CREB), neurotrophic factors, corticosteroid receptors] (Manji, 1992; Barden et al., 1995; Chen and Rasenick, 1995; Nibuya et al., 1995; Belmaker et al., 1996; Lenox et al., 1996; Manji et al., 1996; Racagni et al., 1996; Avissar et al., 1998). The convergence of these, and several other, modifications in the neuronal macromolecular repertoire and function brings about the long-term (plastic) changes in neurotransmission responsible for the action of ADs on mood, motivation, and cognition (Blier and de Montigny, 1994; Artigas et al., 1996; Hyman and Nestler, 1996; Duman, 1998). For reasons of space, the mechanisms proposed for the antimanic action of lithium will not be addressed in this article (for a recent review on lithium, see Lenox et al., 1998).
PROTEIN PHOSPHORYLATION AND NEURONAL FUNCTION
Protein phosphorylation represents one of the major intracellular regulatory mechanisms (Edelman et al., 1987; Hanks et al., 1988). Protein kinases and phosphatases, which catalyze, respectively, the covalent binding to or the release of phosphate groups from substrate proteins, are with few exceptions primary targets of intracellular second messengers in most signaling cascades. In all cells, these effectors are involved in the regulation of general processes, such as (a) cell cycle, (b) metabolism and intracellular transport, and (c) gene expression (Lawrence, 1992; Taylor et al., 1992). In neurons, in addition to these functions, they regulate several processes more inherent (in some cases specific) to nervous system activity, such as (d) receptor desensitization, (e) ion channel modulation, (f) cytomatrix organization, (g) neurotransmitter biosynthesis, and (h) neurotransmitter release (Nairn et al., 1985; Huganir and Greengard, 1990; Greengard et al., 1993). Many protein kinases are diffused ubiquitously in various cellular districts, yet they are able to regulate different and independent functions by means of selective colocalization with appropriate substrates (Mochly-Rosen, 1995; Faux and Scott, 1996). This implies that if one wants to study the function of protein phosphorylation with regard to different neuronal functions (and to the action of psychotropic drugs), kinase activity in isolated cellular compartments should be analyzed because total cellular changes in activity or expression of protein kinases may not reflect changes occurring at particular loci.
Protein kinase C (PKC), a family of enzymes comprising at least 11 isoforms, contains, as several other kinases, a pseudosubstrate autoinhibitory domain, a catalytic domain interacting with magnesium-ATP and substrate protein, and sites for allosteric interaction with diacylglycerol, phosphatidylserine, and (for some isoforms) calcium (Mochly-Rosen, 1995; Faux and Scott, 1996). Upon activation, the kinase translocates from soluble to particulate compartments (plasma membrane, nucleus, cytoskeleton, neuronal postsynaptic densities). A variegate pattern of regulatory functions is carried out by compartmentalization of selected PKC isoforms together with various protein substrates. Several PKC-binding proteins have been identified, which target PKC to different compartments. Some of them are substrates for the kinase, such as vinculin, annexins, and myristoylated alanine-rich C-kinase substrate (MARCKS) proteins; others are not phosphorylated by PKC and bind to a site different from the substrate-binding pocket and in a phosphatidylserine-independent manner. PKC phosphorylation of MARCKS and neuromodulin decreases their affinity for calmodulin (CaM) and may be important for the subsequent stimulation of CaM-dependent enzymes, including CaM-dependent protein kinases and phosphatases. RACKS (receptors for activated Ckinase) are PKC-binding proteins that are not substrates for the kinase; different RACKs are thought to be important in the translocation of the kinase to specific subcellular locations containing various substrates. Peptides reproducing a short sequence of PKC-binding proteins block the binding of PKC to RACKs and cellular responses linked to PKC activation (Mochly-Rosen, 1995). PKC α, βI-II, γ, ε, δ, and ζ isoforms have been identified in the brain and spinal cord. Most of these isoforms are present in other tissues as well: ε is expressed predominantly in brain and epithelial tissues, and γ is present in brain and spinal cord only (Tanaka and Nishizuka, 1994; Nishizuka, 1995). The various isoforms appear to be differentially distributed in the mammalian CNS, with cell type and cellular compartment specificity. Cellular and intracellular distribution of PKC isoforms suggests isoform-related biological functions, but this specialization has only partly been explored (for this topic, the reader is addressed to the reviews cited above and references therein). Neuronal functions of PKC include the modulation of ion channels and receptor sensitivity (Huganir and Greengard, 1990), of serotonin (5-HT) and noradrenaline (NA) transporters (see below), of use-dependent synaptic plasticity (long-term potentiation) (Pasinelli et al., 1995), and of presynaptic release of neurotransmitter (Nichols et al., 1987).
cAMP-dependent protein kinase (PKA), one of the most thoroughly studied and best biochemically understood kinases, is a central component of the cAMP signaling system, mediating most of the effects due to changes in the intracellular concentration of this second messenger (Taylor et al., 1990). In the absence of cAMP, PKA is a tetramer containing two regulatory (R) subunits and two catalytic (C) subunits. In the holoenzyme, binding of the R subunit, which contains a pseudosubstrate domain, maintains the C subunit in an inactive state by occupying the peptide-binding site and preventing access of other protein substrates. Two classes of R subunits have been identified, R I and R II, with several isoforms cloned for each class. Binding of cAMP by R subunits relieves the inhibitory constraint by dissociating two free C subunits from an R dimer. Once dissociated from R subunits, C subunits are able to phosphorylate a great variety of cellular substrates or to be imported into the nucleus for modulation of gene expression. It is now known that, as for several other protein kinases and phosphatases, different subsets of protein substrates may be affected by PKA activation. The kinase action is restricted to certain compartments by association with a targeting locus, often structural proteins that are components of plasma membranes, cytoskeletal matrix, intracellular organelles, or other cellular districts (Dell’ Acqua and Scott, 1997). Several A-kinase anchoring proteins have been identified, which are responsible for targeting of PKA to plasma membrane, endoplasmic reticulum, Golgi, mitochondria, and peroxisomes. Association with membranes is mediated largely by the R II subunit (R IIb in neurons). In some cases, the targeting locus may coincide with a substrate protein. This appears to be the case for the R II subunit in the neuronal microtubule compartment, where the kinase is anchored to microtubule-associated protein 2 (MAP2), a protein involved in the regulation of microtubule assembly.
The brain contains relatively high levels of PKA. Both I and II isoform classes are present, with a three- to fourfold prevalence of the type II kinase, which seems to be predominantly of neuronal origin (Erlichman et al., 1980). PKA is present in all subcellular fractions, but is particularly enriched at pre- and postsynaptic sites, implying an important role in synaptic transmission and neural plasticity. This is also implied by the high number of receptors for neurotransmitters, hormones, and peptides affecting the cAMP cascade. Among the several identified substrates of PKA in neuronal tissue that are able to affect neurotransmission, one should remember the β-adrenergic receptor, nicotinic acetylcholine receptor, calcium, sodium, and potassium channels, the rate-limiting enzyme in catecholamine biosynthesis tyrosine hydroxylase, neurofilaments, MAP2, and the presynaptic terminal proteins synapsin I and rabphilin (Hausdorff et al., 1990).
Ca2+/CaM-dependent protein kinase II (CaMKII), also referred to as multifunctional CaM kinase, is a ubiquitous enzyme mediating responses to changes in intracellular calcium concentrations. The function of CaMKII has been implicated in processes as diverse as metabolism, cell cycle, gene expression, calcium homeostasis, apoptosis, receptor and ion-channel regulation, cytoskeletal function, synthesis, and release of neurotransmitters (Llinas et al., 1991; McGlade-McCulloh et al., 1993; Braun and Schulman, 1995; Wright et al., 1997; Heist and Schulman, 1998). The autoinhibitory domain and catalytic domain are located on the same molecule. Binding of Ca2+/CaM activates the kinase by releasing the constraint of the autoinhibitory domain on the catalytic domain; autophosphorylation of a single residue (Thr286 in the α isoform) increases by about three orders of magnitude the kinase affinity for CaM. Trapping of CaM makes the kinase calcium-independent (autonomous), a state conserved until it is dephosphorylated by protein phosphatases or phosphorylated on additional (autoinhibitory) residues. CaMKII is able to respond to calcium elevation induced by calcium influx through voltage- or transmitter-activated channels, or by release from intracellular calcium stores.
CaMKII is the most abundant protein kinase in the brain; it is ∼20 times more concentrated compared with nonnervous tissue (∼1% of total protein in cerebral cortex and 2% in the hippocampus) (Braun and Schulman, 1995). The kinase is enriched at synaptic sites, where it is present in presynaptic terminals and in postsynaptic densities located in register with presynaptic active zones. In postsynaptic densities, CaMKII is the most abundant protein and was estimated to be up to 20% of total protein (Kennedy, 1997), although its enrichment may have been overestimated because of postmortem translocation. In view of this synaptic enrichment, functional studies of neuronal CaMKII have focused on its role in synaptic transmission, transmitter release, and synaptic plasticity. At the presynaptic level, CaMKII was shown to increase release when introduced in rat brain synaptosomes or squid synaptic terminals (Nichols et al., 1990; Llinas et al., 1991); stimulation of release was shown to correlate with kinase activation (Gorelick et al., 1988; Tsutsui et al., 1994). Several studies demonstrated that CaMKII activity is essential for the establishment of hippocampal long-term potentiation, a cellular model for synaptic plasticity (Rotenberg et al., 1996).
AD TREATMENT AND PROTEIN PHOSPHORYLATION
It is widely believed that long-term changes in intraneuronal signaling brought about by prolonged AD treatment have a final outcome in neurotransmission modifications occurring mainly in cortical and limbic areas. Accordingly, it is relevant to study protein phosphorylation in such areas in subcellular compartments involved, after AD treatment. Hereafter, the terms acute and long-term (or repeated) treatment are used to indicate, respectively, a single drug administration and a daily administration replicated for several days (usually between 1 and 3 weeks); the length of drug treatment is reported below for the studies performed in the laboratories of the authors of this review.
Long-term administration of fluoxetine or desmethylimipramine (DMI), a selective serotonin reuptake inhibitor (SSRI) and a tricyclic inhibitor of NA and 5-HT reuptake, respectively (Table 1), was found to decrease significantly the basal activity of soluble and particulate PKC in cerebral cortex and hippocampus (Mann et al., 1995). Furthermore, PKC was implicated in the desensitization of 5-HT2A receptors induced by the agonist 1-(2,5-dimethoxy-4-iodophenyl)-2-aminopropane, a finding suggesting the involvement of this kinase in receptor-mediated effects of ADs (Rahimian and Hrdina, 1995). Other reports showed that AD-sensitive hSERT, the human plasma membrane transporter for 5-HT, is a substrate for PKC and that PKC activation or phosphatase inhibition down-regulates 5-HT uptake probably via a reduction of SERT cell surface expression (Ramamoorthy et al., 1998). These results imply that PKC is involved in the control of 5-HT reuptake and that AD-induced 5-HT2A receptor desensitization and the consequent alteration of PKC signaling may affect the transporter’s regulated trafficking. NA transporter activity and surface expression were also shown to be modulated by PKC. However, still little is known as to the action of AD treatment on PKC in subcellular compartments.
The action of AD treatment on the function of PKA and CaMKII has been studied in recent years in two different and complementary neuronal compartments. The two kinases were studied, respectively, in the microtubule fraction (microtubules are localized in cell soma and dendrites, but absent from axons) and in the synaptic vesicle fraction (a presynaptic terminal compartment). It was found (Fig. 1) that long-term (but not acute) AD treatment activates PKA in microtubules and CaMKII in synaptic vesicles in a number of brain areas (depending on the drug used). At the same time, upon stimulation of endogenous phosphorylation, phosphate incorporation in selected substrates located in the two neuronal compartments was increased, linking the long-term action of drugs to selected changes in the cellular response (Perez et al., 1991; Popoli et al., 1995; Racagni et al., 1996).
THE cAMP-DEPENDENT SIGNALING PATHWAY AND THE ACTION OF ADs
Microtubule-associated PKA is activated by AD treatment
Studies in postmortem brain from depressed subjects involve the cAMP-dependent pathway in the pathophysiology of affective disorders. It was found that the binding of cAMP, considered as a measure of the level of PKA regulatory subunits, is reduced significantly in cytosolic fractions of various brain areas from bipolar disorder-affected subjects (Rahman et al., 1997). Another study found the levels of cAMP binding sites unchanged in the brain of AD-free suicides, but decreased in AD-treated suicides (Lowther et al., 1997). It is interesting that β-adrenoreceptor-linked PKA was found to be reduced significantly in skin fibroblasts from depressed patients (Shelton et al., 1996). Several lines of evidence from different laboratories have demonstrated an involvement of the cAMP-dependent signaling pathway in the mechanism of action of AD. Moyer et al. (1986) found, in animals subjected to acute and repeated treatment with DMI, a decrease in the soluble PKA activity in pineal gland. Nestler et al. (1989) found in rat frontal cortex a decrease of PKA activity in the cytosol and an increase in the nuclear fraction, after long-term administration of imipramine, tranylcypromine, or electroconvulsive shock. These effects were produced selectively by ADs and were not found with other psychotropic drugs. More recently, it was found that long-term administration of imipramine decreased cAMP-dependent phosphorylation of tyrosine hydroxylase in the ventral tegmental area (Rosin et al., 1995).
The effect of AD treatment on PKA regulatory subunits was investigated in detail. First, it was found that repeated administration of DMI (20 mg/kg for 10 days) induced a marked increase in the amount of the 52-kDa R II subunit of PKA (Racagni et al., 1992). To study the changes induced in the kinase, the binding of cAMP to PKA regulatory subunits was investigated by using photoaffinity labeling with 8-azido[32P]cAMP (8-N3-[32P]cAMP) (Fig. 2). DMI administration increased in cerebral cortex the binding of cAMP to the R II subunit in the high-speed S1 soluble fraction (containing microtubules), in the isolated crude microtubule fraction, but not in the S2 soluble fraction devoid of microtubules (Racagni et al., 1992). cAMP binding to the R I subunit was not affected, indicating a selective effect on type II PKA. Endogenous phosphorylation experiments showed that DMI induced a marked increase (greater than two-fold) in basal and cAMP-stimulated phosphorylation of the substrate MAP2. This finding was confirmed recently by Miyamoto et al. (1997), who found that MAP2 phosphorylation increased after long-term administration of DMI and that this was coupled with inhibition of microtubule assembly. MAP2 is a major substrate for PKA in microtubules, and it colocalizes with the R II-containing kinase, working at the same time as an anchoring protein and as a substrate (Obar et al., 1989; Rubino et al., 1989). Together with other microtubule-associated proteins, like tau, MAP2 has the function to stimulate the assembly of microtubules by inhibiting tubulin dissociation. This is regulated, in turn, by the affinity of MAP2 for tubulin, which is under the control of phosphorylation by several protein kinases. Phosphorylation of MAP2 on several residues decreases its affinity for tubulin and the stabilization effect on microtubule assembly. Interestingly, it was found that a large fraction (about one third) of type II PKA in the brain is bound to microtubules. This was confirmed by the finding that virtually all of type II PKA in the S1 soluble fraction copurifies with microtubules, suggesting that in this case the effect of the drug is at the same time selective for enzyme isoform and subcellular compartment (Perez et al., 1993). No change in PKA or increase in substrate phosphorylation was observed after acute treatment or incubation in vitro with the drug.
Similar results have been obtained with various monoamine reuptake blockers (Table 1), including the selective 5-HT reuptake blockers fluoxetine, fluvoxamine, and paroxetine, and the selective NA reuptake blocker (+)-oxaprotiline (Perez et al., 1991; Mori et al., 1998a,b). Repeated (10-12 days) treatments with these drugs increased the binding of 8-N3[32P]cAMP to the R II subunit in both S1 and microtubule fraction from cerebral cortex (Fig. 2). Treatment with the isomer (-)-oxaprotiline, devoid of NA transporter blocking action, did not induce any change in cAMP binding, indicating that, at least for these drugs, the action on PKA is linked to monoamine transporter blockade and not to structural properties of the compounds.
Overall, these findings strongly suggested that cAMP-dependent protein phosphorylation in microtubules is a site affected by long-term AD treatment, and that changes in the availability of NA or 5-HT are at least partly necessary to induce these changes. The finding that neither acute treatment nor in vitro addition of drugs was able to induce the same modifications supported the idea that they are part of the adaptive changes induced by ADs in postreceptor signaling mechanisms and could contribute to their therapeutic effect (Racagni et al., 1992).
However, not all ADs induce the same changes in cAMP-dependent phosphorylation. Moclobemide, a reversible inhibitor of MAO A, was found to induce an increase in cAMP binding to R II in S1 supernatant after 21, but not 1, 5, or 12, days of treatment (Mori et al., 1998b). However, when the crude microtubule fraction was isolated, neither cAMP binding to R II nor endogenous phosphorylation of MAP2 was affected by the treatment, suggesting that the PKA component activated by moclobemide is not associated with microtubules and affects substrates different from MAP2. Alternatively, the time course of the effect on PKA might be different, and microtubules could be affected at later times not yet investigated. Nonetheless, this confirmed that PKA is a target for different classes of AD.
Surprisingly, a structurally unrelated compound, S-adenosylmethionine (SAM), showed similarities to ADs. This methyl donor was reported to possess antidepressant properties in a number of clinical trials (Bottiglieri, 1997). After repeated (12 days), but not acute, treatment, SAM induced an increase in cAMP binding to R II and in the endogenous phosphorylation of MAP2 in cerebrocortical microtubules (Zanotti et al., 1998). Although the modifications were less pronounced than with typical ADs, these results strengthened the association between antidepressant effect and involvement of cAMP-dependent phosphorylation, and suggested that a similar outcome at the cellular and molecular levels may be obtained by compounds acting primarily on quite different mechanisms.
A time-dependent translocation of type II PKA
A crucial problem in the study of the mechanism of action of ADs is the correlation between the observed biochemical modifications and the time course of antidepressant effect (usually several weeks). To study the time-related effects of ADs on cAMP binding in cerebral cortex, recently a time course of this molecular change was investigated by treating the animals for various periods with fluvoxamine and paroxetine (Mori et al., 1998a). The animals were treated for 5, 12, and 21 days; photoaffinity labeling of the R II PKA subunit was studied as above with 8-N3[32P]cAMP in the S1 supernatant, in the microtubule fraction obtained from S1, and in the S2 supernatant devoid of microtubules. Five days of treatment did not affect cAMP binding in any of the fractions examined. After 12 days, cAMP binding was increased significantly in S1 and in the microtubule fraction. After 21 days, no more changes were observed in S1 and microtubules, whereas the binding was increased in the S2 supernatant that does not contain microtubules. Although, at present, changes in the affinity of R II for the ligand cannot be ruled out completely, these results suggested a time-dependent translocation of R II to microtubules initially (12 days) and from microtubules to the cytosol afterward (21 days). This time course was consistent with the normal onset of action of the drugs and implied that, in the course of drug-induced adaptive changes, PKA may have undergone translocation between different cellular compartments. The final translocation from microtubules to the cytosol could represent an intermediate step in the transport of the catalytic subunit to the nucleus, as found previously (Nestler et al., 1989). Indeed, it is known that PKA may modulate gene expression by phosphorylating the transcription factor CREB, whose expression is affected by long-term AD treatment (Nibuya et al., 1996). Furthermore, it was found previously that translocation of PKA activity to the nucleus induced by isoproterenol is dependent on microtubules (Schwartz and Costa, 1980).
In summary, a number of studies carried out by several laboratories clearly indicated that the cAMP-dependent signaling system and, in particular, type II PKA are modified by long-term, but not acute, treatment with ADs. Early data and more recent studies converge toward a theoretical framework envisaging activation of type II PKA and translocation of the kinase through different compartments, with consequent phosphorylation of selected substrates. The possible consequences of these phosphorylation events on neuronal function will be discussed in the last paragraph of this review.
LONG-TERM AD TREATMENT MODIFIES NEUROTRANSMISSION
A long-term outcome of AD treatment is an alteration of neurotransmission in certain monoaminergic pathways. Development of this phenomenon during treatment is gradual and is thought to be related to the onset of therapeutic efficacy. This was investigated in detail for 5-HT pathways originating in medial and dorsal raphe nuclei and projecting to limbic and cortical areas (Blier and de Montigny, 1994; Mongeau et al., 1997). A common result of long-term treatment with different classes of ADs is an enhancement of synaptic transmission in these pathways. This result is obtained apparently in different ways by different drug classes. Some drugs (tricyclic ADs) enhance 5-HT neurotransmission mainly by increasing the sensitivity of 5-HT1A postsynaptic receptors. Other drugs (MAO inhibitors and SSRI) mainly affect 5-HT autoreceptors that regulate neuronal firing or transmitter release efficacy, or both, in presynaptic neurons (Blier and de Montigny, 1994; Mongeau et al., 1997). Electrophysiological studies showed that, in the initial phases of treatment, stimulation of 5-HT1A receptors in the raphe by an increased 5-HT level reduces the firing rate of serotonergic neurons. Based on these findings, therapeutic strategies have been devised that combine a classic AD (SSRI) with a presynaptic 5-HT1A receptor antagonist, to overcome initial 5-HT1A receptor stimulation and increase efficacy of therapeutic effect (Artigas et al., 1996). Also, an initially increased level of 5-HT at presynaptic terminals in target areas stimulates 5-HT1B/D receptors, which inhibit stimulation-evoked transmitter release. During treatment, both types of 5-HT receptors become desensitized, leading to a restoration of normal firing and to a net increase in the amount of transmitter released per impulse. The increase in 5-HT release has been demonstrated by electrophysiological recording, by biochemical techniques, and by microdialysis in vivo (Bel and Artigas, 1993; Blier and Bouchard, 1994; Blier and de Montigny, 1994). Therefore, as exemplified for 5-HT, an increase in the efficiency of presynaptic release is an important mechanism by which the efficacy of neurotransmission may be augmented by ADs. However, the actual biochemical mechanisms involved in the AD-induced release increase are mostly unknown at present.
Long-term treatment with ADs increases presynaptic CaMKII activity in the hippocampus
Although the insight into molecular mechanisms regulating presynaptic release of neurotransmitters has greatly deepened recently, still very little is known at the molecular level about the effect of psychotropic drugs on presynaptic exocytotic machinery (Scheller, 1995). Initial findings came from long-term treatment with a number of selective or nonselective 5-HT reuptake blockers (animals treated with paroxetine, fluvoxamine, or venlafaxine for 19, 19, and 12 days, respectively). All the drugs induced a robust and sustained increase in the activity of presynaptic CaMKII (synaptic vesicles and synaptic cytosol), and in the autophosphorylation of the major kinase isoform (α-CaMKII) upon endogenous phosphorylation (Fig. 3A). This change was found in the hippocampus, one of the areas where transmitter release has been shown previously to be increased by AD treatment. No change was found in total cerebral cortex, possibly because an effect in selected cortical areas was masked when tissue from whole cortex was assayed (Popoli et al., 1995). The increase in CaMKII activity was localized selectively to synaptic terminals, because no change was found in homogenate and in synaptosomal membranes. Synaptosomal membranes contain a majority of postsynaptic CaMKII, which is very enriched in postsynaptic densities, associated with portions of postsynaptic membranes copurifying with synaptosomes (Rostas and Dunkley, 1992). The increase in kinase activity and autophosphorylation was not elicited by acute treatment and could not be induced by incubation of synaptic vesicles with drugs in vitro (Popoli et al., 1995).
Long-term treatment with ADs increases phosphorylation of CaMKII presynaptic substrates in the hippocampus
Subsequently, it was shown that the increase in kinase activity induced by AD treatment was accompanied by an increase in the endogenous phosphorylation of selected vesicular substrates (Popoli et al., 1997a). Synapsin I and synaptotagmin, the two major CaMKII substrates in synaptic vesicles, are both part of the molecular machinery regulating transmitter release at nerve terminals. Whereas synapsin I phosphorylation was increased with some, but not all, ADs used, in the case of synaptotagmin all drugs markedly increased the substrate phosphorylation (between two and threefold). This was confirmed when phosphosynaptotagmin was immunoprecipitated following endogenous phosphorylation of synaptic vesicles [synaptotagmin phosphorylation ratio (fluvoxamine versus control) = 2.9 ± 0.7] (Fig. 3B). Western blot analysis showed that both for CaMKII and for synaptotagmin the actual amount of protein associated with synaptic vesicles was unchanged in treated animals, ruling out possible modifications in gene expression or translocation of protein to synaptic terminals. This suggested that the changes in phosphorylation of both kinase and substrate are due solely to the increase in kinase activity.
It is of interest to investigate how an increase in CaMKII substrate phosphorylation may affect the efficacy of transmitter release. Synaptotagmin is the putative calcium sensor in synaptic vesicles and interacts with other proteins of the so-called exocytosis SNARE [N-ethylmaleimide-sensitive fusion protein (NSF) attachment protein (SNAP) receptor (SNARE)] complex, possibly conferring calcium dependence to neuronal exocytosis (Popoli, 1993a; Geppert et al., 1994). Recent findings suggested that interaction in vitro of synaptotagmin with the SNARE protein complex may be strengthened following CaMKII activation (see below) (Popoli et al., 1998).
Presynaptic Ca2+/CaM-dependent protein phosphorylation and the mechanism of action of ADs
The effect of ADs on presynaptic protein phosphorylation was confirmed by various studies. An increase in presynaptic Ca2+/CaM-dependent protein phosphorylation was found after long-term treatment with SSRI, 5-HT and NA reuptake inhibitors (SNRI), selective NA reuptake inhibitors (NARI), and tricyclic ADs (Table 1) (Popoli et al., 1995, 1997a,b; Verona et al., 1998). However, a difference was noted among different classes of drugs, in that selective and nonselective 5-HT reuptake inhibitors increased CaMKII activity and substrate phosphorylation only in the hippocampus, whereas drugs having a higher affinity for NA transporter, such as tricyclics (DMI), induced changes both in the hippocampus and in the frontal cortex. Nonrelated drugs (e.g., benzodiazepines) did not modify CaMKII activity. As for PKA, a nonrelated compound (SAM), shown to possess antidepressant properties (Bottiglieri, 1997), was also tested (12 days). SAM, at variance with typical ADs, increased presynaptic kinase activity in cerebral cortex, but not in the hippocampus (Zanotti et al., 1998); whether this effect of SAM is due to its role in methylation, to interaction with adenosine receptors, or to other reasons remains an open question (Mudd and Cantoni, 1964; Travagli et al., 1994; Zanotti et al., 1998).
The change in presynaptic CaMKII activity is a stable, long-lasting modification, because it is induced by long-term, but not acute, AD treatment, and it is measurable after the time necessary to prepare synaptosomes and subsynaptosomal fractions. It is likely that covalent posttranslational modifications account for the change in activity; this is currently under investigation in a study of kinase kinetic parameters.
Nature of synaptic terminals affected by AD treatment
A major problem posed by AD treatment-induced changes in presynaptic terminals is the nature and number of the pathways involved. With the most selective drugs (SSRI), in view of the described desensitization of presynaptic 5-HT1B/D receptor and consequent facilitatory effects on 5-HT release, one may speculate that in hippocampus mainly 5-HT terminals are involved. However, the magnitude of the phosphorylation change observed is so great that a correlation with the actual number of serotonergic terminals seems difficult. In the hippocampus, treatment with SSRI or nonselective inhibitors induced an increase in vesicular CaMKII activity ranging from 60 to 145%, with a parallel increase in α-CaMKII autophosphorylation of 108-146% and an increase in synaptotagmin phosphorylation of 190-303% (Popoli et al., 1995, 1997a). These figures would suggest that either more than a single terminal type is affected, or the most abundant type of terminal in a single area (i.e., glutamatergic) is also involved (Bouron and Chatton, 1999). Indeed, 5-HT1B/D receptors are present as heteroreceptors in several types of terminals, and it was proposed that 5-HT heteroreceptors may predominate over autoreceptors (Hoyer et al., 1994; Laduron, 1998). The finding that a similar effect was produced by ADs that selectively or nonselectively inhibit NA reuptake added further complexity, suggesting that presynaptic αa-noradrenergic receptor desensitization may also be involved in the activation of CaMKII. Also, it is interesting that DMI, a drug that mainly inhibits NA reuptake, activates presynaptic CaMKII both in hippocampus and in frontal cortex, showing a regional specificity different from SSRI. To establish a clearer correlation between presynaptic machinery activation and type of pathways involved, it will be necessary to use microdialysis studies or measurement of transmitter release from synaptosomes in combination with protein phosphorylation studies.
Changes in efficacy of transmitter release: receptor desensitization or other mechanisms?
A number of microdialysis in vivo studies in recent years have shown clearly that long-term treatment with various types of ADs induces changes in basal and evoked release of monoamine neurotransmitters in several cortical, limbic, and motor areas (including hippocampus, frontal cortex, striatum, and nucleus accumbens) (Kreiss and Lucki, 1995; Mongeau et al., 1997; Page and Abercrombie, 1997). These findings confirmed the measurements of release from cerebral tissue slices from treated animals (Blier and Bouchard, 1994). Therefore, changes found in presynaptic protein phosphorylation and in the release machinery after AD treatment are likely to be a molecular correlate of the changes observed in neurotransmitter release (Racagni et al., 1996). These findings raised the problem as to how the presynaptic protein machinery may be affected by AD treatment.
The first question is whether presynaptic changes are caused by desensitization of presynaptic terminal auto- and heteroreceptors. Whereas the mechanism by which somatodendritic receptors may regulate firing rate is explained at least partly (Williams et al., 1988), less is known about the functional mechanism of presynaptic terminal receptors. Indeed, it was shown that the modulatory action of presynaptic receptors on neurotransmitter release may be accounted for by the G protein-mediated modulation of ion channels (either membrane-delimited or second messenger-mediated), as well as by a direct, although unknown, action on the presynaptic release machinery (Haydon et al., 1991; Scholz and Miller, 1992). Furthermore, it cannot be ruled out at present that desensitization of somatodendritic receptors may also in the long term affect the efficiency of release by other means than just a change in the firing rate. However, it is likely that changes in monoamine synaptic terminal inhibitory receptors, such as 5-HT1B/D, α2, and D2 receptors, may affect release by acting on local (presynaptic) mechanisms. One way by which these receptors exert their action is by affecting adenylate cyclase, the cAMP signaling pathway, and PKA; changes in PKA activity affect the phosphorylation of voltage-activated calcium channels and potassium channels (Nunoki et al., 1989; Levitan, 1994). Therefore, initially the stimulation of receptors by increased synaptic monoamine levels may decrease cAMP production (these receptors are negatively coupled to adenylate cyclase). This will decrease PKA activity and, in turn, increase potassium and decrease calcium conductance, with consequent hyperpolarization of the terminal and decreased probability of release. Successive desensitization of terminal receptors, brought about by AD treatment, may attenuate this action on ion channels and membrane potential and by an unknown mechanism induce a change in CaMKII-dependent phosphorylation of vesicular effectors (synaptotagmin, synapsin I), whereby facilitating transmitter release.
This is not, however, the only hypothesis available. Among the pleiotropic effects exerted by ADs on neuronal cells, a relevant role may be played by the changes induced in calcium homeostasis and signaling. It was shown by various authors that incubation of synaptosomes, primary neuronal cultures, or glioma cultures with ADs decreases the depolarization-induced calcium influx and the calcium rise induced by stimulation of various phospholipase C-coupled receptors, as well as blocking sodium channels (Shimizu et al., 1994; Lavoie et al., 1997). Some of these modifications were observed in conditions in which monoamine reuptake was inactive, suggesting a direct effect of ADs on calcium fluxes. Although the drug concentration used was often higher than comparable drug levels reached in the brain of treated animals, these studies suggested that, in the long term, AD action on ionic conductances and calcium homeostasis may contribute to their effect. If one also takes into account the several reports describing an upregulated calcium mobilization (induced by 5-HT or other extracellular stimuli) in peripheral cells (platelets, lymphocytes) from bipolar or unipolar depressed patients, the hypothesis that modifications of calcium homeostasis may contribute to the AD effect becomes feasible (Dubovsky et al., 1989, 1992; Yamaji et al., 1997; Pancrazio et al., 1998). It is tempting to speculate that a prolonged down-regulation of calcium influx caused by repeated treatment might induce a compensatory change in one or more of the transducers entitled to respond to variations in calcium fluxes (such as CaMKII). Again, functional experiments with synaptosomes and cultured cells should help to investigate this possibility. Obviously, it is possible that several different mechanisms converge on CaMKII, inducing covalent conformational changes that increase its activity. Reports should also be mentioned that show that ADs may inhibit Ca2+/CaM-dependent protein kinases in vitro (Silver et al., 1986; Popoli et al., 1995).
PRE- AND POSTSYNAPTIC CHANGES IN PROTEIN PHOSPHORYLATION AFFECTING NEUROTRANSMISSION: COMPLEMENTARY COMPONENTS IN THE ACTION OF ADs
One of the major consequences of the expanding knowledge about psychotropic drug action is the concept that therapeutic effect cannot be mediated by a single neurobiological adaptation (Blier and de Montigny, 1994; Artigas et al., 1996). This is well illustrated by the present situation in the field of antidepressant treatment. There is a plethora of compounds available (Table 1): several drugs with a more or less understood primary action (mostly monoamine reuptake inhibition); other drugs with different primary actions, sometimes on multiple targets; and others with a totally unrelated or even paradoxical action (e.g., tianeptine). What strikes the pharmacologist is that all these drugs, at variable degrees, possess antidepressant properties. A possible answer to this riddle is that each different drug may act at one or more different transporter, receptor, or intracellular levels; what may be important is the convergence of effects into intracellular signaling and neurotransmission changes beneficial for antidepressant treatment. A way to distinguish beneficial adaptive changes is to detect intraneuronal modifications common to several different drug types.
In most signaling cascades, protein phosphorylation represents an obligatory crossroad; multiple phosphorylation events regulate neuronal function, and an extensive cross talk occurs among different signaling pathways (Selbie and Hill, 1998). It is therefore conceivable that at the neuronal level the modifications described here in serine/threonine kinases are only a portion of the number of modifications induced in protein phosphorylation by AD treatment. However, the phosphorylation systems affected by AD treatment (as shown here) represent, for their location and respective role in neuronal function, complementary components that may be partly responsible for the changes in neurotransmission. What could be the downstream effects of AD-induced PKA and CaMKII activation?
Microtubules are located in neuronal soma and in dendrites, a neuronal district involved in the reception of synaptic signaling. They participate in plastic cytoskeletal changes necessary for synaptic remodeling. Microtubule assembly is regulated by a number of associated proteins, including MAP2, whose function is regulated, in turn, by phosphorylation. When phosphorylated by PKA, MAP2 inhibits microtubule assembly and therefore may favor cytoskeletal changes necessary for longterm (plastic) modifications in neurotransmission (Miyamoto et al., 1997). This would be a somatodendritic (postsynaptic) modification consequent to the treatment with ADs. Furthermore, the time-dependent changes in PKA translocation induced in cerebral cortex by longterm AD treatment are suggestive of sequential steps in the translocation of PKA to the nucleus for regulation of gene expression. It was shown that PKA translocation to the nucleus is microtubule-dependent (Schwartz and Costa, 1980); therefore, the changes in MAP2 phosphorylation and consequent changes in microtubule assembly may also represent a modulatory step in the transport of information to the nucleus for modifications in gene expression.
Equally important for synaptic transmission may be the changes induced by AD treatment in presynaptic protein phosphorylation and in the function of the protein machinery regulating presynaptic release (Greengard et al., 1993). Several lines of evidence showed recently that various phases of the exocytotic—endocytotic cycle of synaptic vesicles are regulated by specific protein—protein interactions. The core of these interactions involves three SNARE proteins forming a tight (sodium dodecyl sulfate-resistant) complex, the putative calcium sensor (synaptotagmin), voltage-gated calcium channels (mostly N and P/Q type), and a number of soluble proteins (NSF and SNAPs) implicated in the disassembly of the SNARE complex (Sollner et al., 1993; Scheller, 1995). The SNARE complex is postulated to carry out most of the molecular interactions connected with docking, maturation, and fusion of vesicles. Most, if not all, of these proteins are substrates of protein kinases present and enriched in synaptic terminals (CaMKII, PKA, PKC); recent evidence showed that phosphorylation—dephosphorylation may affect their interactions and consequently modify the function of the release apparatus (Bennett et al., 1993; Popoli, 1993b; Hirling and Scheller, 1996; Shimazaki et al., 1996; Yokoyama et al., 1997; Stevens and Sullivan, 1998). With regard to CaMKII substrates, it was found that calcium/CaM-dependent phosphorylation of the N-type channel decreases its interaction with the SNARE complex. This is envisaged as a regulatory step that may decrease the inhibitory constraint of SNARE proteins on the channel (Yokoyama et al., 1997), whereby increasing channel responsiveness. Also, it was suggested recently that phosphorylation by CaMKII is required to obtain maximal binding of synaptotagmin to the SNARE complex, a step that may increase calcium sensitivity of the complex (Popoli, 1993a; Popoli et al., 1998). Both these results would be in line with the known facilitatory function of presynaptic CaMKII on the release of neurotransmitters (Nichols et al., 1990; Llinas et al., 1991) and establish a functional link between long-term AD treatment, the sustained increase in presynaptic CaMKII activity, and the facilitation of presynaptic release. A further step in this direction will be the investigation of changes in presynaptic protein—protein interactions after AD treatment.
The last decade brought about a major shift in the theoretical framework addressing the issue of the mechanism of action of psychotropic drugs, thanks in part to the increased knowledge of intracellular signaling. Several lines of evidence suggested that protein phosphorylation a prominent regulatory process used by most signaling pathways, is involved in the long-term action of ADs. Treatment with various drugs was shown to affect activity or translocation of second messenger-regulated protein kinases, notably PKC, PKA, and CaMKII. The effect of ADs on PKA and CaMKII was studied in selected subcellular compartments (microtubules and synaptic vesicles) involved in plastic changes affecting the efficacy of neurotransmission. It was found that activation of kinases by ADs is accompanied by an increase in the phosphorylation of selected substrates in subcellular compartments (MAP2 and synaptotagmin) that may mediate the kinase’s action on cytoskeleton remodeling and transmitter release. We speculate that these changes in protein phosphorylation, occurring with timing compatible with the onset of therapeutic effect in humans, are relevant molecular correlates of the changes in monoaminergic neurotransmission following AD treatment. These findings may be useful in the search for new pharmacological strategies, addressing intracellular signaling in the brain, for the treatment of neuropsychiatric disorders. Furthermore, they may suggest new targets in the search for biological markers in neuropsychiatric diseases. Indeed, it was reported that β-adrenoceptor-linked PKA activity is reduced in fibroblasts of depressed patients (Shelton et al., 1996), suggesting that AD treatment-induced PKA activation observed by various authors may in some way normalize a defect in cAMP-mediated signaling associated with depression. In the absence of satisfactory animal models for mood disorders thus far, the investigation of molecular markers in peripheral cells (fibroblasts, lymphocytes, and platelets) may offer a ready alternative to investigate candidate molecular markers in a still poorly explored area.
The authors wish to acknowledge the skillful scientific collaboration of Silvia Mori, Daniela Tardito, and Marina Verona, who have contributed greatly to the development of the experimental research summarized in this article.