|Additional related information may be found at:|
|Neuropsychopharmacology: The Fifth Generation of Progress|
Dopaminergic Mechanisms in Depression and Mania
Dopaminergic Mechanisms in Depression and Mania
Traditional accounts of the biochemical basis of depression have focussed largely on noradrenaline (NA) and serotonin (5-HT), and although most of the evidence that coalesced into the 'catecholamine hypothesis of depression' does not distinguish clearly between NA and dopamine (DA), the potential role of DA was at first overlooked. Following two influential reviews that drew attention to this oversight [89, 114], there has been an upsurge of interest in the possible involvement of DA in affective disorders. In fact, as will be seen below, there is little in the recent clinical evidence to justify this change of fashion; the pressure to reconsider the role of DA in depression arises almost entirely from preclinical developments. One is the now substantial body of work (reviewed below) demonstrating that antidepressant drugs enhance the functioning of mesolimbic DA synapses. However, the major driving force has undoubtedly been the massive research effort around the involvement of DA systems in motivated behaviour (see Mesocorticolimbic Dopaminergic Neurons: Functional and Regulatory Roles, this volume [Le Moal]).
It is important at the outset to recognize that neurotransmitters operate in the brain within well-defined projection systems that may subserve discrete functions. Of the two major forebrain DA projections, the larger nigrostriatal pathway is involved primarily in extrapyramidal motor functions, whereas the smaller mesocorticolimbic system, which innervates limbic structures such as nucleus accumbens, amygdala, ventral hippocampus and prefrontal cortex, supports a variety of behavioural functions related to motivation and reward (118; see Mesocorticolimbic Dopaminergic Neurons: Functional and Regulatory Roles, this volume [Le Moal]); the third major DA pathway, the tuberoinfudibular projection subserves neuroendocrine functions. The behavioural functions of the mesocorticolimbic DA system make this pathway an obvious candidate for investigation in relation to affective disorders. However, this focus on a specific pathway, which contains a relatively small proportion of forebrain DA neurons, creates a very real risk that changes in DA activity relevant to affective psychopathology could be obscured, in studies that employ global measures of brain DA function, or apply anatomically nonspecific interventions.
Two phases can be distinguished in the development of a theoretical perspective on the role of the mesocorticolimbic DA system in motivated behaviour. The first was the formulation by Wise and colleagues of the 'DA hypothesis of reward', which proposed that rewarding events, irrespective of their modality, shared the common property of activating the mesocorticolimbic DA system; conversely, inactivation of DA function would lead to anhedonia, the inability to experience pleasure . This hypothesis, with its obvious implications for a potential role for the mesocorticolimbic DA system as a substrate for affective disorders, has stimulated an extensive body of behavioural pharmacological research aimed at clarifying its precise role(s) in reward, reward-related learning, and affect. More recently, this research perspective has broadened to include consideration of the wider brain circuitry within which the mesocorticolimbic DA system is embedded. In this approach, the major focus shifts to the role of DA in gating the flow of information through the nucleus accumbens, which serves as the major interface through which information in limbic structures gains access to motor output systems. The major non-dopaminergic afferent projections to the nucleus accumbens, which represent the major output pathways of the limbic system, are from amygdala, hippocampus and prefrontal cortex. Each of these three structures is itself innervated by the mesocorticolimbic DA system, and their projections to the nucleus accumbens overlap with one another, and with the mesoaccumbens DA afferents [118; Mesocorticolimbic Dopaminergic Neurons: Functional and Regulatory Roles, this volume [Le Moal]. It is noteworthy that all of these structures afferent to the nucleus accumbens are implicated in affective psychopathology.
While the hypothesis that the mesoaccumbens DA projection functions as a 'reward pathway'  remains controversial, it is now indisputable that this pathway plays a crucial role in the selection and orchestration of goal-directed behaviours, particularly those elicited by incentive stimuli, and in reward-related learning [11,118; see Mesocorticolimbic Dopaminergic Neurons: Functional and Regulatory Roles, this volume [Le Moal]. These properties make a hypofunction of the mesocorticolimbic DA system a prime candidate to mediate the inability to experience pleasure (anhedonia) and loss of motivation (lack of interest) that lie at the heart of major depressive disorder. (Conversely, manic hyperexcitability could readily result from a DA hyperfunction: this possibility has long been recognized ). This hypothesis differs somewhat from earlier 'biochemical theories' in that it not only proposes a relationship between a biochemical entity (DA) and a mental disorder (depression or mania), but also defines explicity the nature of the relationship, in terms of the functional properties of the relevant DA neurons. The hypothesis also defines certain boundary conditions: it involves a limited set of DA projections (the mesocorticolimbic system), and a limited set of depressive symptoms (anhedonia and lack of interest). This chapter reviews critically the recent clinical and preclinical evidence pertaining to the DA hypothesis of affective disorders, with particular reference where possible to the DA innervation of the nucleus accumbens. The three sections of the review deal with: attempts to measure DA function in affective disorder patients; the effects of pharmacological manipulations of DA function in human subjects; and preclinical evidence that antidepressant drugs increase transmission through mesolimbic DA synapses. The emphasis is on recent developments, and in particular, on issues not covered in detail by earlier reviews, which are frequently used in place of references to the older literature [20, 54, 89, 114].
DA FUNCTION IN AFFECTIVE DISORDERS
Numerous studies have attempted to assess forebrain DA function in depressed patients by measuring levels of the DA metabolite homovanillic acid (HVA) in cerebrospinal fluid (CSF). In some studies patients were pretreated with probenecid to block the transport of HVA out of the CSF; this procedure, which measures the accumulation of HVA, is considered to give a better estimate of DA turnover. Most studies have tended to report a decrease in CSF HVA in depressed patients, and this relationship holds strongly in studies using the probenecid technique. Decreases in CSF HVA are particularly pronounced in patients with marked psychomotor retardation. In fact, a 1983 review of this area concluded that "The consistent finding of decreased post-probenecid CSF HVA accumulation in depressed patients, particularly those with psychomotor retardation, is probably the most firmly established observation in the neurochemistry of depression" . More recent studies have not altered this conclusion [20, 54, 90].
DA turnover, measured post-mortem, is also reduced in the caudate nucleus and nucleus accumbens of depressed suicides . As DA uptake is unchanged in depressed suicides [1,18], the decreased turnover apparently reflects decreased DA release. There are also many reports of decreased CSF HVA in depressed suicide attempters . Consistent with these findings, a decrease in 24-hour urinary excretion of HVA and DOPAC has been reported in depressed suicide attempters . As abnormalities of DA metabolism are not observed in nondepressed suicide attempters , these data provide further evidence that decreased DA turnover is a correlate of depression.
Nevertheless, the interpretation of these data is far from straightforward. Although one study has reported that CSF HVA was lower in melancholic than in non-melancholic patients , this relationship is probably explained by the association between low CSF HVA and psychomotor retardation [20, 114], which is a prominent feature of melancholia. In fact, low CSF HVA has been associated with psychomotor slowing (bradyphrenia) not only in depressed patients, but also in Parkinson's disease and Alzheimer's disease . In agitated patients, however, CSF HVA levels are normal or slightly elevated . CSF HVA levels (as well as plasma DA ) are also elevated in delusional patients . Again, this finding may reflect psychomotor change: in a study of psychotic patients, CSF HVA levels were elevated in those with delusions and agitation, but normal in those with delusions but no agitation . CSF HVA levels are usually found to be elevated in mania . These data suggest that CSF HVA levels may reflect motor activity rather than mood, and further raise the problem of whether a reduction in HVA level is the primary cause or a secondary reflection of psychomotor retardation. This latter problem has lain dormant since an early study in which a group of depressed patients were asked to simulate mania: the exercise did increase DA turnover, but also elevated mood .
It is hardly surprising that CSF HVA levels are associated with level of motor activity, since CSF HVA derives largely from the caudate nucleus, on account of its large size and its periventricular location. In schizophrenic patients, decreased CSF HVA concentrations are associated with ventricular enlargement , which is equally common in major depressive disorder . Indeed, PET imaging studies have reported hypometabolism of the head of the caudate nucleus in unipolar and bipolar depressed patients, which may reflect a decreased DA activity in this structure . However, the contribution to CSF HVA of DA release in mesolimbic structures such as the nucleus accumbens and frontal cortex is relatively minor. There is therefore no reason to expect that changes in mesolimbic DA function would be apparent in studies measuring HVA levels in lumbar CSF; it is far more likely that any such changes would be obscured by alterations in nigrostriatal DA function associated with changes in motor output. Thus, although most reviewers have tended to interpret the HVA data as evidence for a DA dysfunction in depression [54, 89, 114], these data are actually silent with respect to the important question of the state of activity in the mesocorticolimbic DA system.
In one series of studies increased levels of DA itself were observed in CSF of melancholic patients, with a tendency towards higher concentrations in patients who were delusional . CSF DA levels have been found to correlate with extraversion in depressed patients . However, the proportion of DA in lumbar CSF originating in the forebrain is unknown.
The possibility that abnormalities of DA turnover may be a feature of depression per se, rather than a consequence of altered motor activity prompts the question of whether there are relevant genetic differences in depressed individuals. Following an initial report of a genetic marker for depression on the short arm of chromosome 11 , many studies have examined the tyrosine hydroxylase gene and other markers located within this region: while some studies have reported positive findings [eg. 100], most have not. Also, no abnormalities of the DA transporter gene have been detected in either unipolar or bipolar depressed patients [44, 66].
In common with virtually all other neuroreceptors, DA receptors are now known to exist in multiple subtypes. The original distinction, between D1 and D2 subtypes, was made on pharmacological grounds: DA interacts differently with these receptors (for example, in the dorsal striatum DA has excitatory effects at D1 receptors while D2 receptors are inhibitory), and the two receptors are coupled to different postsynaptic transduction mechanisms. More recently, five different DA receptors have been cloned, and shown to fall into two families, D1-like (D1 and D5) and D2-like (D2, D3 and D4). D1 and D2 receptors are present in all brain regions that receive a dopaminergic projection, both subtypes are expressed at a high level in the dorsal and ventral striatum, but D1 receptors predominate in prefrontal cortex. DA autoreceptors are of the D2 subtype, with a possible D3 contribution; there are no D1 autoreceptors. D3 and D4 receptors are localized almost exclusively within ‘limbic’ areas, particularly the nucleus accumbens shell, and so are of particular interest in relation to affective disorders (see Molecular Biology of the Dopamine Receptor Subtypes and Dopamine Receptor Expression in the Central Nervous System, this volume [Civelli, Mansour]).
There have as yet been relatively few studies of DA receptors in affective disorder patients. Post-mortem studies have reported no change in D1or D2 receptor binding in depressed suicides  or D4 receptors in patients with major depression , relative to matched controls. One PET imaging study has reported a decrease in D1 receptor binding in the frontal cortex, but not the striatum, of bipolar patients (n=6 euthymic, n=3 manic and n=1 depressed); however, the ligand used in this study (SCH-23390) also binds to 5HT receptors . PET studies of D2 receptors suggest that D2 receptor numbers may be elevated in manic but not in nondelusional depressed patients . SPECT imaging studies have reported either no change , or unilateral  or bilateral  increases in D2 receptor binding in the basal ganglia. The latter findings are compatible with a decrease in DA turnover, but are subject to similar problems of interpretation: the relative contributions of mood and motor activity and the inability of current techniques to image the nucleus accumbens independently of the dorsal striatum. Antidepressant treatment has been associated with an increase in D2 receptor binding, in both post-mortem  and SPECT imaging studies; in the latter study, the extent of clinical recovery was significantly correlated with the size of the increase in D2 receptor binding in striatum and anterior cingulate gyrus .
Although there is strong evidence for a genetic contribution to bipolar affective disorder, molecular genetic studies have so far provided no evidence that bipolar disorder is associated with abnormalities of the genes coding for the D1 [69,74], D2 [22, 50, 66, 69, 74], D3 [44, 66, 91, 102] or D5  receptors. Both positive  and negative  findings have been reported with respect to D4-receptor polymorphisms. An association between a D4-receptor polymorphism and the personality trait of sensation seeking has been reported in several studies [eg. 37], but others have disputed this obserervation [eg. 83].
The tuberoinfundibular DA system has neuroendocrine functions, inhibiting the release of prolactin and stimulating the release of growth hormone (GH). Thus, basal levels of these hormones have been examined as potential markers of DA function in affective disorders, and their reponses to DA agonists have been used to evaluate DA receptor responsiveness. These studies suffer two serious limitations: the inability to generalize any conclusions to the forebrain DA systems, and the involvement of many other neurotransmitters in neuroendocrine regulation; in particular, a stimulatory role of 5HT in prolactin secretion and a stimulatory role of alpha-adrenergic receptors in GH secretion.
Abnormal prolactin levels have frequently been reported in depressed patients, but there is no consistency in the direction of change: low, normal and high values have been reported in different studies [20, 54, 114]. Prolactin levels are reported to be normal in mania [54, 67]. Prolactin responses were also normal in depressed patients following DA agonist [20, 54] or antagonist  challenges. However, two studies have reported a decrease in prolactin levels in seasonal affective disorder (SAD), which was seen in both unipolar and bipolar patients, and was present during both winter depression and summer euthymia [30, 31]. This apparent trait abnormality in SAD patients is consistent either with increased DA function or with decreased 5HT function. The former interpretation is supported by the observation that SAD patients also showed a seasonally-independent increase in spontaneous eye blinking: this behaviour is thought to be under dopaminergic control, being increased by D2 agonists and suppressed by D2 antagonists [30, 31]. Blink rate has been reported to be unaltered in patients with major depression 
Studies of GH are similarly inconclusive. Basal GH levels have been reported to be decreased , normal  or increased  in major depression; no changes were seen in mania . One study reported a blunting of the GH response to apomorphine in major depression, relative to patients with minor depression or normal controls , but no differences were observed in many earlier studies, using either a slightly higher dose of apomorphine (0.75 vs. 0.5 mg), or L-dopa [54, 114]. The group reporting blunted GH responses to apomorphine have reported a difference between major and minor depressives in two further studies [5, 80], and have also reported blunted responses in manic patients  and in suicide attempters . The same group also reported that blunted apomorphine responses in depressed patients were associated with low introversion and anxiety scores on the MMPI, but not with severity of depression ; others have reported a negative correlation between GH response and severity of delusions . Together, these observations suggest that there may be some subsensitivity to apomorphine in a subgroup of depressed patients. If these findings are confirmed, the question remains of whether they reflect DA receptor subsensitivity, or a more general decrease in GH responsivity (it is well established that the GH response to alpha-adrenergic challenges is subsensitive in major depression ). The relevance of GH changes for forebrain DA function also remains to be determined.
MOOD EFFECTS OF DA AGONISTS AND ANTAGONISTS
The psychostimulants amphetamine and methylphenidate cause activation and euphoria in normal volunteers. Although these drugs enhance activity at both DA and noradrenergic synapses, the psychostimulant effects are mediated at DA synapses, since they are antagonized by DA receptor blockers, but not by adrenergic receptor blockers [51, 76, 77]. The euphoric effects of psychostimulants at low doses closely parallel the symptomatology of hypomania, while high doses, particularly when taken repeatedly or chronically, can cause grandiosity, delusions, dysphoria, and all the other symptoms of a full-blown manic episode [51, 85].
Single doses of amphetamine or methylphenidate also cause a transient mood elevation in a high proportion (>50%) of depressed patients ; the response in depressed patients appears similar, in size and in the proportion of subjects responding, to that seen in nondepressed volunteers [23, 76]. Following an initial report by Fawcett & Simonpoulous , a number of studies have used the acute mood response to psychostimulants to predict the clinical response to chronic antidepressant therapy. A review of this literature confirmed that the response to antidepressants was well predicted by the result of an amphetamine challenge (85% improvement in responders vs 43% in nonresponders), but questioned the predictive value of a methylphenidate challenge (66% improvement in responders vs 68% in nonresponders) . However, the amphetamine and methylphenidate studies differ in that the former involved mainly patients treated with imipramine and desipramine, while the latter also included a high proportion of patients treated with 'serotonergic' antidepressants. A reanalysis of the same literature showed that the acute response to methylphenidate does predict antidepressant efficacy, provided that the analysis is restricted to patients treated with 'noradrenergic' antidepressants .
Psychostimulants are not themselves considered to be efficaceous as antidepressants. In early trials, the catecholamine precursor l-DOPA produced a modest global improvement, primarily in retarded patients, but the effect was largely one of psychomotor activation with little effect on mood; in bipolar patients, DOPA frequently caused a switch into hypomania . These data have been interpreted as evidence against a prominent role for DA in depression. However, the effects of DOPA were greatest in patients with the lowest pretreatment CSF HVA levels . This suggests that the effect of DOPA might primarily be to increase DA release in the caudate nucleus, perhaps causing motor side effects that could mask any potentially therapeutic effects of an increase in mesolimbic DA release. It is now known that low doses of amphetamine preferentially release DA within the nucleus accumbens . Despite the absence of clinical trial data, amphetamine continues to find widespread, if little publicized, use in the treatment of depression .
More convincing antidepressant effects have been reported with the directly acting DA agonists piribedil and bromocriptine. These were largely open trials, but there are also controlled studies, including a double-blind trial showing piribedil to be superior to placebo, particularly in patients with low pre-treatment CSF HVA, and two large trials which found no difference in antidepressant efficacy between bromocriptine and imipramine . The antidepressant response to bromocriptine may be greater in bipolar patients , and one study suggests a preferential effect of bromocriptine on emotional blunting . Hypomanic responses during bromocriptine therapy have been reported [55, 105]. In a particularly interesting development, Mouret and colleagues have described striking and rapid therapeutic effects of piribedil in previously non-responsive patients whose sleep EEG showed signs characteristic for Parkinson's disease; in patients not showing these signs, piribedil was ineffective .
Trials of DA agonists in depression are not currently fashionable, but a recent double-blind study found effects superior to placebo and comparable to fluoxetine for pramipexole, a very selective D3-preferring D2/D3 receptor agonist . It is also notable that DA uptake inhibition is a prominent feature of a number of newer antidepressants, including nomifensine, buproprion, and amineptine . The mechanism of action of bupropion, which is widely used both as monotherapy for depression and in combination with SSRIs, appears to involve both dopaminergic and noradrenergic components . Bupropion is also used for smoking cessation, but this effect appears to be independent of its antidepressant properties , and may involve direct nicotinic antagonist actions . Amineptine, which is a relatively selective DA uptake inhibitor, was more efficaceous than clomipramine, and had a faster onset of antidepressant action, in a double-blind trial in retarded patients; another dopaminomimetic agent, minaprine, was also more effective than clomipramine in retarded patients .
Contrary to expectations, given the antidepressant effects of DA agonists, there is also clear evidence that under certain circumstances, neuroleptics, which are DA receptor antagonists, are also active as antidepressants [73, 92]. One potential resolution of this apparent paradox (which will be discussed further below) is that neuroleptics may be antidepressant only at low doses, which act preferentially as DA autoreceptor antagonists and so increase DA turnover. This hypothesis has been advanced in particular in relation to certain atypical antidepressants, such as sulpiride, which are said to have 'activating' properties . Antidepressant effects of sulpiride are seen in a dose range of 50-150mg/day, which is considerably lower than the typical antipsychotic dose of 800-1000mg/day. A DA-activating effect of sulpiride at low doses is supported by the finding that low doses of sulpiride antagonized the sedative actions of apomorphine in human subjects .
Antidepressant effects have also been reported for roxindole, a putatively selective DA autoreceptor agonist. In an open trial, roxindole caused rapid improvements in 8 of 12 patients suffering from a major depressive episode, as well as reducing depression and anergia in schizophrenic patients . Roxindole possesses 5HT uptake-inhibiting and 5HT agonist actions, both of which could contribute to an antidepressant effect, but neuroendocrine data (suppression of prolactin secretion ) suggest that DA agonism is the predominant action of this drug. If, as claimed, roxindole is a selective autoreceptor agonist, the effect should be to decrease DA function. However, it is questionable whether roxindole is antidepressant by virtue of decreasing DA function: the drug also appears to be effective in negative schizophrenia , which is compatible with a DA-activating effect.
Depression is frequently encountered as a side effect of neuroleptic therapy in schizophrenia [89, 106]. This is a complex issue, with debates about whether ‘neuroleptic-induced depression’ is a side effect of treatment, a part of schizophrenia, a secondary effect of having schizophrenia, or the unmasking of a pre-existing depression when psychotic symptoms are brought under control. However, schizophrenic patients on neurolepics are more likely to show full depressive syndromes than those not on neuroleptics, with a strong association between neuroleptic use and anhedonia, and this relationship holds up after controlling for level of psychosis . This suggests that ‘neuroleptic-induced depression’ is genuine, and there are strong grounds for believing that the effect is caused by antagonism of DA receptors. Conversely, neuroleptic drugs also decrease manic symptomatology. Although classical neuroleptics act at a variety of receptor sites, antimanic effects are also observed with drugs that act relatively specifically as DA receptor antagonists . In normal volunteers neuroleptics induce feelings of dysphoria, paralysis of volition and fatigue .
It is still widely believed that the catecholamine depleting drug reserpine causes depression, on the basis of a series of reports in the 1950s, despite the findings of Goodwin et al. , on reanalysis of these data, that the great majority of 'reserpine depression' patients had been incorrectly diagnosed. Patients treated with reserpine tended to display a 'pseudodepression' characterized by psychomotor slowing, fatigue and anhedonia but lacking cognitive features of depression such as hopelessness or guilt. Only a small proportion of patients (5-9%) showed symptoms analogous to major depression, and these patients usually had a prior history of mood disorders . It remains unclear whether the doses of reserpine administered in the 'reserpine depression' studies were sufficient to decrease DA function. However, it may be significant that in the Goodwin et al. reanalysis, major depression was considered to be the correct diagnosis in almost 50% of patients who developed marked psychomotor retardation .
More convincing evidence of an association between DA depletion and depression is seen in the high incidence of depression in Parkinson's disease [72, 89 - but see 110]. At the level of symptomatology, there is substantial overlap between Parkinsonian akinesia and depressive psychomotor retardation [14, 110]. It is difficult to determine whether Parkinsonian depression should be considered a secondary response to loss of motor function, rather than a direct consequence of DA depletion. There is no agreement in the literature as to whether the severity of depression is correlated with the extent of physical impairment. However, Parkinsonian depression is more severe than would be expected from the physical symptoms alone, and the onset of depression can precede the physical disabilities [47, 72].
It is now recognized that Parkinson's disease can not be considered as a pure DA deficiency syndrome: NA, 5HT, ACh, somatostatin and neurotensin are also abnormal . Nevertheless, there are good reasons to relate the symptoms of Parkinsonian depression to DA depletion. In one well-designed study, depressed Parkinsonian patients showed profound attenuation of the euphoric response to methylphenidate, relative to non-depressed Parkinsonian patients, depressed non-Parkinsonian patients, and normal controls . The antidepressant effect of DOPA in Parkinson's disease [3, 46, 89] also points towards a dopaminergic substrate of Parkinsonian depression. In some cases there is clear evidence that mood improvement precedes the improvement in physical symptoms , suggesting that the antidepressant effect cannot be simply explained away as secondary to an improvement in physical symptoms. Antidepressant effects of bupropion  and bromocriptine  have also been reported in Parkinsonian patients.
Neuroleptics as antidepressants
The clinical pharmacology literature reviewed in this section is broadly consistent with the hypothesis that increases in DA function elevate mood and decreases in DA function induce symptoms of depression. However, not all of the data are compatible with this formulation. In particular, the fact that neuroleptics are used to treat depression [73, 92] strikes at the heart of the dopamine/anhedonia/depression hypotheses. This phenomenon therefore requires careful consideration.
One hypothesis, discussed above, is that neuroleptics are administered in depression at low doses that interact selectively with presynaptic autoreceptors. However, while an autoreceptor hypothesis might explain some of the data, particularly those pertaining to sulpiride, it is not necessarily the case that low doses are used when neuroleptics are prescribed as antidepressants. Doses below the antipsychotic range have usually been prescribed in studies of mild, non-endogenous depression, but in delusional depression, neuroleptics are more commonly prescribed at normal antipsychotic doses . However, it is not certain that DA antagonism is the mechanism of antidepressant action. Indeed, in one study, antidepressant effects of cis-flupenthixol were negatively correlated with increase in serum prolactin levels, suggesting that DA blockade might actually antagonize the antidepressant effect . In similar vein, antidepressant effects on withdrawal of neuroleptics are well documented, though the evidence tends to arise from case reports rather than formal studies . In a controlled trial, Del Zompo et al. treated depressed patients with a cocktail of haloperidol and chlorimipramine, and reported marked improvement, relative to a group treated with chlorimipramine alone, when the haloperidol component was withdrawn after three weeks treatment. It was assumed that the improvement resulted from the unmasking of DA receptors rendered supersensitive by chronic neuroleptic treatment . Clearly, more trials of this kind are needed, and the proposed mechanism of action requires confirmation.
It is also questionable whether neuroleptics are truly antidepressant, and examination of the pattern of symptomatic improvement may provide the clearest resolution to the paradox of the antidepressant action of neuroleptics: in brief, there is no evidence that neuroleptics can improve either psychomotor retardation or anhedonia, the core symptom of depression most closely associated with the DA hypothesis. The antidepressant potential of neuroleptics is most firmly established in delusional depression, which responds well to combined therapy with a neuroleptic/tricyclic mixture, but responds poorly if at all to tricylics alone. However, neuroleptics alone are also ineffective in delusional depression: they produce a substantial global improvement, but this arises almost entirely from a decrease in agitation and delusional thinking; motor retardation, lack of energy and anhedonia do not respond to neuroleptic treatment, and indeed, may become worse . In endogenous depressions, while neuroleptics have been claimed to be as effective as tricyclics, or nearly so, this appearance may be spurious, insofar as the studies in question may have seriously underestimated the true effectiveness of tricyclics (owing to a failure to attain adequate plasma drug levels, and other factors) . On the basis of the findings in delusional depression, it seems likely that the global improvement seen in endogenous depressives treated with neuroleptics results from the preponderance in these studies of agitated and delusional patients [73, 92]. This analysis of the place of neuroleptics in the treatment of depression implies that retardation and delusions are mediated by different sets of DA terminals, which may be activated independently . In support of this assumption, it is well established that different components of the mesocorticolimbic DA projection are differentially regulated (see Dopamine Receptor Transcript Localization in Human Brain, this volume [Le Moal])
DOPAMINERGIC MECHANISMS OF ANTIDEPRESSANT ACTION
DA autoreceptor desensitization
Most antidepressant drugs have little effect on DA function following acute administration; in particular, tricyclic antidepressants do not act as potent DA uptake inhibitors , in contrast to their well known effects at adrenergic and serotonergic synapses (though some data suggest that antidepressants may cause significant inhibition of DA uptake within the nucleus accumbens and frontal cortex [24, 27]). Nevertheless, there is now considerable evidence that antidepressants do enhance dopaminergic function following chronic administration.
In one of the earliest studies to demonstrate an antidepressant-induced increase in DA function, Serra et al reported that imipramine, amitriptyline and mianserin all decreased the sedative effect of a low dose of apomorphine. Since this latter effect was assumed to be mediated by stimulation of DA autoreceptors, the results were interpreted as a decrease in autoreceptor sensitivity . However, the evidence that antidepressants desensitize DA autoreceptors is equivocal. There are a number of supportive studies, using a variety of techniques, but equally, there have been failures to replicate all of these data . Some studies have reported that clear evidence of DA autoreceptor subsensitivity was not present until 3-7 days following withdrawal from chronic antidepressant treatment [95, 111]. Another reason to question the relevance of DA autoreceptor desensitization for the clinical action of antidepressants is that these data were obtained in 'normal' rats; rats exposed to chronic mild stress, which has been proposed as an animal model of depression, show evidence of DA autoreceptor desensitization similar to that sometimes seen following chronic antidepressant treatment in 'normal' animals . Finally, changes in apomorphine-induced sedation do not necessarily imply changes in DA autoreceptor function. High doses of apomorphine cause locomotor stimulation, so a decrease in apomorphine-induced sedation might equally well indicate an increase in postsynaptic responsiveness rather than autoreceptor subsensitivity.
Sensitization of D2/D3 receptors
In fact, a substantial body of literature now demonstrates that following chronic treatment, antidepressants do increase the responsiveness of postsynaptic D2/D3 receptors in the mesolimbic system; these effects are seen irrespective of the primary neurochemical action of the drug [62, 115]. The majority of studies have examined the locomotor stimulant response to moderate doses of apomorphine or amphetamine; these responses are consistently elevated following chronic administration of antidepressants. Similar effects were observed using the specific D2/D3 agonist quinpirole . There are well known pharmacokinetic interactions between antidepressants and amphetamine. However, antidepressants also increased the psychomotor stimulant effect when amphetamine, or DA itself, was administered directly to the nucleus accumbens , confirming a true pharmacodynamic interaction. Furthermore, these effects were present within a short time (2h) of the final antidepressant treatment, confirming that, unlike DA autoreceptor desensitization, the increase in responsiveness of postsynaptic D2/D3 receptors is not simply a withdrawal effect. The potentiation of D2/D3 receptor function by chronic antidepressant treatment is confined to mesolimbic terminal regions: antidepressants do not increase the intensity of stereotyped behaviours caused by high doses of amphetamine, which are mediated by DA release within the dorsal striatum . Neither did chronic antidepressant treatment potentiate a DA-mediated neuroendocrine response .
Receptor binding studies have usually failed to detect any alterations in the binding parameters of D2/D3 receptors that would explain the increased functional responses. The majority of these studies are of limited relevance, as they assayed DA receptors in samples of dorsal striatum. Nevertheless, negative findings have also been reported in nucleus accumbens. However, D2/D3 receptors in limbic forebrain (but not dorsal striatum) have an increased affinity for the agonist ligand, quinpirole, following chronic antidepressant administration to rats, and an increase in receptor number has recently been reported, in ventral but not dorsal striatum, using an agonist ligand . Consistent with these observations, a recent study using a conventional antagonist ligand found that a decrease in D2/D3 receptor numbers in limbic forebrain of rats subjected to chronic mild stress was completely reversed by chronic treatment with imipramine . Increased D3-receptor binding in ventral striatal regions, following chronic antidepressant treatment, has also been recently reported .
In addition to increasing the responsiveness of D2/D3 receptors, antidepressants also decrease the number of D1 receptors, following chronic treatment . This effect is associated with a decrease in the ability of DA to stimulate adenyl cyclase , and a decreased behavioural response (grooming) to D1 receptor stimulation , consistent with the binding data. A role for D1 receptor changes in the sensitization of D2/D3 receptors has been proposed , but this seems unlikely, as the downregulation of D1 receptors is species specific: D1 receptors were downregulated by chronic imipramine in rats but not in mice . Furthermore, D1 receptors were not downregulated by chronic imipramine in chronically stressed rats, which did show D2/D3 receptor upregulation . In both of these studies, functionally-relevant behavioural effects of chronic antidepressant treatment were seen in the absence of D1-receptor changes [75, 78]
Role of mesolimbic DA in animal models of depression
Although these data confirm that antidepressants change the functional status of DA receptors in the nucleus accumbens, they give little insight into the role that these changes play in the clinical action of antidepressants. Animal models of depression provide one means of addressing this question, albeit indirectly. The mechanisms by which antidepressants act have been analyzed most extensively in the Porsolt swim test. In this model, rats or mice are required to swim in a confined space, and antidepressants prolong the period in which the animal displays active escape behaviour. Immobility in the swim test may be reversed not only by antidepressants, but also by D2/D3 receptor agonists, applied systemically or to the nucleus accumbens . Conversely, a number of studies have reported that antidepressant effects in the swim test were reversed by DA antagonists ; these include studies in which antidepressants were administered chronically [29, 87]. The effects of chronically administered tricyclic antidepressants were reversed by the administration of sulpiride in the nucleus accumbens, but not in the dorsal striatum . Despite these positive findings, Borsini and Meli urge caution in accepting that the data demonstrate a dopaminergic mechanism of antidepressant action in the swim test, and suggest that the effects of intra-accumbens sulpiride could be related to the presence in the mesolimbic system of non-dopaminergic sulpiride binding sites that also bind antidepressants . The swim test has been criticized on a number of counts, most prominently, that it responds to acute administration of antidepressants, unlike the clinical situation, which requires chronic treatment. This criticism is not entirely justified, since the test only responds acutely to extremely high drug doses, but becomes slowly more sensitive with repeated treatment . However, the validity of the test as a model of depression is extremely weak.
Dopaminergic mechanisms have also been analyzed in animal models of depression more valid and realistic than the swim test. For example, a decrease in D2 receptor binding has been reported in socially subordinate female cynomolgous monkeys, which display many features reminiscent of affective pathology . The most extensive investigations of this type have employed the chronic mild stress procedure, in which rats or mice are exposed chronically (weeks or months) to a variety of mild unpredictable stressors. This causes a generalized decrease in responsiveness to rewards (anhedonia), which can be reversed by chronic administration of tricyclic or atypical antidepressants . These behavioural changes are accompanied by a decrease in D2/D3-receptor binding and D2-mRNA expression in the nucleus accumbens, and a pronounced functional subsensitivity to the rewarding and locomotor stimulant effects of the D2/D3 agonist quinpirole, administered systemically or within the nucleus accumbens. All of these effects are also reversed by chronic antidepressant treatment [34, 117].
The functional role of DA in these antidepressant effects has been examined in studies in which animals successfully treated with antidepressants were treated acutely with D2/D3 receptor antagonists, at low doses that were without effect in non-stressed animals or in untreated stressed animals. This treatment reversed the effects of a wide variety of antidepressants (including tricyclics, specific 5HT or NA uptake inhibitors, or mianserin) . Chronic stress also causes an antidepressant-reversible decrease in aggressive behaviour, and this effect of chronic antidepressant treatment was also reversed by acute administration of DA antagonists . These data argue strongly that an increase in D2/D3 receptor responsiveness may be responsible for the therapeutic action of antidepressants in this model .
The data reviewed in the preceding section present a strong case that elevation of DA transmission in the nucleus accumbens may represent a 'final common pathway' responsible for at least part of the spectrum of behavioural actions of antidepressant drugs. As has been demonstrated for sensitization to psychomotor stimulants, antidepressant sensitization of DA transmission may be initiated by an increase in extracellular DA levels in the vicinity of DA cell bodies in the ventral tegmental area . The mechanisms by which antidepressants bring about these changes are not well understood, but the best guess at present is that the effects are indirectly mediated, via primary actions at NA or 5HT terminals. Whatever the mechanism, the antidepressant-induced increase in mesolimbic DA function is of considerable functional significance. A recent study of eight depressed patients reported that the therapeutic effect of SSRIs was acutely reversed, causing a reinstatement of symptoms in all patients, by a low dose of sulpiride . These results, which are directly comparable to the data from animal studies, imply that an increase in D2/D3 receptor responsiveness may also be responsible for the clinical antidepressant action of SSRIs.
Nevertheless, the evidence supporting a dopaminergic mechanism of antidepressant action is largely preclinical: clinical studies evaluating the role of DA mechanisms in the action of classical antidepressants are sparse. However, the preclinical studies suggest that D2/D3 receptors in the accumbens might represent a potential target for antidepressant action, and it is likely that this receptor population will serve as a focus for novel drug development strategies. The value of this approach remains to be determined, but the initial clinical studies of the antidepressant potential of D2/D3 agonists and DA autoreceptor antagonists are promising.
It remains the case that the preclinical evidence for enhancement of DA transmission following chronic treatment with antidepressant drugs provides the strongest support for the DA hypothesis of affective disorders. The strongest evidence against the DA hypothesis comes from the clinical use of neuroleptics in depression. As discussed above, there are a number of potential resolutions of this troublesome paradox, including the possibility of autoreceptor-selective actions of neuroleptics at low doses, the possibility that neuroleptics control delusions but actually worsen depressive symptoms, and the possibility that DA hypofunction in some terminal fields coexists with DA hyperfunction in other regions [40, 73]. (The latter hypothesis has also been advanced to explain the coexistence of negative and positive symptoms in schizophrenia [26, 40].) There has been little research directed specifically at understanding the place of neuroleptics in the treatment of depression: more is urgently needed.
Setting aside the question of neuroleptics as antidepressants, the effects of pharmacological interventions, in human subjects, lead broadly to the conclusion that inhibiting DA transmission is therapeutic in mania and induces depressive symptomatology in normal volunteers, while stimulation of DA transmission has antidepressant effects and induces manic symptoms. However, the extent of overlap between these pharmacological effects and clinical changes is far from complete. While the effects of psychostimulants provide a good match to the symptoms of mania, the primary effects of neuroleptics or reserpine in normal subjects are fatigue, apathy and dysphoria [10, 45]. Conversely, while l-dopa readily induces hypomania in depressed patients, there is little evidence of mood improvement . Thus, the pharmacological evidence for DA involvement appears rather stronger in mania than in depression. However, this conclusion overlooks the anatomical nonspecificity of these drugs: they are of limited value as research tools for evaluating whether depression is associated with a dysfunction of mesolimbic DA specifically. In contrast to l-dopa, directly acting DA agonists do appear to be effective antidepressants, though the number of controlled trials remains unacceptably low. The clinical efficacy of these agents may reflect a preferential action within the nucleus accumbens, but it is not yet possible to evaluate this hypothesis in human subjects.
Similarly, the inability to measure DA activity within the nucleus accumbens seriously limits the value of virtually all of the correlative studies of DA function in depression and mania. The fact that there are no reliable neuroendocrine changes in affective disorder patients simply tells us that there are no generalized abnormalities of D2/D3 receptors, not that such abnormalities are absent within the nucleus accumbens specifically. Similarly, we have no useful information on the release of DA from mesocorticolimbic terminals in human subjects. The clearest evidence implicating DA in depression, the decrease in CSF HVA concentrations in retarded depression, is intriguing, but appears to relate primarily to changes in motor function. Discovering the direction of causality in this relationship remains an important objective. However, the priority for understanding the role of DA in depression must be to redress the imbalance between the preclinical and the clinical evidence. This requires the development of research tools for human use with sufficient anatomical precision to evaluate DA function within distinct terminal fields.
Syndromes or symptoms?
As noted in the introduction to this chapter, our emerging understanding of the behavioural functions of forebrain DA systems suggests that the involvement of DA in affective disorders might profitably be analyzed at the level of symptoms rather than syndromes. The clinical literature contains a number of findings that support this position. Thus, there is some evidence that emotional blunting responds more rapidly and more completely than other symptoms in depressed patients treated with bromocriptine , and that DA-uptake inhibiting antidepressants may be superior to tricyclics in retarded patients . Conversely, Parkinsonian or pre-Parkinsonian  depressions, which respond to treatment with DA agonists [3, 70, 87], are characterized by decreased motivation and drive, but not by feelings of guilt, self-blame and worthlessness ; these characteristic depressive cognitions are also conspicuously absent from descriptions of neuroleptic- or reserpine-induced depressive states [10, 45].
From a psychobiological standpoint, it seems obvious that the major psychiatric syndromes are likely to involve multiple neurotransmitter systems, which contribute to different syndromes to differing degrees. An obvious research strategy is to investigate, as a first step, the involvement of specific pathways in specific behavioural processes, which need not, on a priori grounds, bear any obvious relationship to nosological boundaries. It is clear that features of what might be termed a DA-deficiency syndrome, involving low CSF HVA, anhedonia, psychomotor slowing, and a good response to DA agonist treatment, are characteristic not only of depression, but also of Parkinson's disease [14, 110, 113, 120], and negative schizophrenia [12, 40, 57, 113]. At the other extreme, there is considerable overlap in symptoms between positive schizophrenia and mania, and these common symptoms are reliably reproduced in psychostimulant-induced psychoses [40, 85]. The extent to which these similar functional outcomes reflect common underlying mechanisms remains to be determined, and represents a major challenge for future research. However, the difficulties of pursuing a research agenda that cuts across DSM-IV diagnostic categories should not be underestimated.