Third generation antipsychotic drugs: partial agonism or receptor functional selectivity?

Abstract

Functional selectivity is the term that describes drugs that cause markedly different signaling through a single receptor (e.g., full agonist at one pathway and antagonist at a second). It has been widely recognized recently that this phenomenon impacts the understanding of mechanism of action of some drugs, and has relevance to drug discovery. One of the clinical areas where this mechanism has particular importance is in the treatment of schizophrenia. Antipsychotic drugs have been grouped according to both pattern of clinical action and mechanism of action. The original antipsychotic drugs such as chlorpromazine and haloperidol have been called typical or first generation. They cause both antipsychotic actions and many side effects (extrapyramidal and endocrine) that are ascribed to their high affinity dopamine D₂ receptor antagonism. Drugs such as clozapine, olanzapine, risperidone and others were then developed that avoided the neurological side effects (atypical or second generation antipsychotics). These compounds are divided mechanistically into those that are high affinity D₂ and 5-HT_2A antagonists, and those that also bind with modest affinity to D₂, 5-HT_2A, and many other neuroreceptors. There is one approved third generation drug, aripiprazole, whose actions have been ascribed alternately to either D₂ partial agonism or D₂ functional selectivity. Although partial agonism has been the more widely accepted mechanism, the available data are inconsistent with this mechanism. Conversely, the D₂ functional selectivity hypothesis can accommodate all current data for aripiprazole, and also impacts on discovery compounds that are not pure D₂ antagonists.

A. The first two generations of antipsychotic drugs

The serendipitous observations about antipsychotic actions of chlorpromazine [1] ultimately led to the finding that these antipsychotic effects were due to antidopaminergic actions [2]. Numerous antipsychotics were subsequently developed that, like chlorpromazine, work primarily as dopamine D₂ receptor antagonists. Drugs of this type (first called typical and now named first generation antipsychotics) include a variety of different chemical classes. These typical drugs can be classified as high, intermediate, or low potency based on their average daily dose range, with their clinical potency often highly correlated with their affinity for the dopamine D₂ receptor [3]. Thus, actions at dopamine systems have been a critical mechanism in all currently approved antipsychotic drugs and many of the compounds in the discovery and development pipeline.

1. Dopamine systems and their receptors

There are three major brain dopamine pathways: the nigrostriatal (from cells in the A9 region); the mesolimbic and mesocortical systems (from cells in the A10 or ventral tegmentum); and the tuberoinfundibular (hypothalamic) system [4]. The first evidence for a biochemical mechanism of dopamine neurotransmission was the observation that dopamine could dosedependently stimulate the synthesis of the second messenger cAMP [5] in a fashion that was antagonized by typical antipsychotic drugs [6]. Both phenothiazine and thioxanthene antipsychotics competitively inhibited the dopamine-stimulated activity of adenylate cyclase in proportion to their clinical potency, leading to the first hypothesis that this was the major functional mechanism of dopamine in the CNS [6, 7]. Studies with antipsychotics in other structural families, however, showed that some behaviorally potent antipsychotics had little potency in inhibiting dopamine-stimulated adenylate cyclase [8]. This discrepancy led to the hypothesis that there were multiple dopamine receptors, one of which stimulated adenylate cyclase [5], whereas the other bound antipsychotics with affinity proportional to clinical antipsychotic potency, but whose functional effect was then unknown [3]. This bifurcation [9, 10] has been codified in the use of the terms D₁-like and D₂-like receptors, respectively.

These two pharmacological families are now known to be the products of five genes, and their biology has been the subject of many chapters and books [11-14]. Both of the “D₁-like” receptor genes are intron-less, and include the D₁ (also called the D_1A in murine species) and the D₅ (also termed D_1B) [15-18]. The “D₂-like” receptors include two major splice variants of the D₂ gene, D_2L (long), and D_2S (short), that together are the most highly expressed of the D₂-like receptors [19-22]. The other D₂-like receptors are D₃ and D₄ receptors, and include D₃ receptor splice variants that are expressed in mice but not humans [20, 23]. The D₄ receptor has variable 16-nucleotide repeats in the region coding for the third intracellular loop that can vary between individuals in periodicity from 4 to 16 repeats [24, 25]. It has been suggested that these D₄ alleles have major effects on behavioral phenotype [26-28], although this idea is controversial [29-31].

In addition to schizophrenia, there has been a great interest of dopamine neurotransmission in the treatment and or etiology of many central disorders such as Parkinson's disease, ADHD, etc. For this reason, there have been major efforts to discover and develop drugs that target dopamine receptors. This has resulted in a rich availability of clinical and research ligands that has facilitated basic research, and has led to many interesting clinical hypotheses.

2. Actions of first generation antipsychotics

After the accidental discovery of chlorpromazine, many other phenothiazine analogs (e.g., fluphenazine, thioridazine, etc., Figure 1) followed it to market. In addition, other first generation drugs were derived from distinct chemical backbones, such as the butyrophenone haloperidol and the thioxanthene flupenthixol (Figure 1).

Figure 1.

Structures of representative first generation (typical) drugs. The top row is representative of the many diverse phenothiazines that have been marketed, whereas the bottom row shows the widely used butyrophenone haloperidol and a representative thioxanthene, flupenthixol.

The widely used first generation antipsychotics like haloperidol are effective in many patients in attenuating the positive symptoms of schizophrenia (e.g., hallucinations and delusions) associated with acute episodes, as well as preventing psychotic relapse [32]. These agents have, however, important limitations. As many as 25-60% of patients are either largely resistant, or only partially responsive, to treatment with conventional antipsychotics. Moreover, these drugs at best have only a modest benefit on negative symptoms (e.g., blunted affect, poverty of thought, emotional and social withdrawal, and lack of drive). The first generation antipsychotics also induce a variety of adverse effects both acutely and with long term exposure. Such side effects reduce compliance and represent major limitations of the drugs for adequate treatment of schizophrenia. Some of the side effects (sedation, autonomic and cardiovascular effects, weight gain) are largely due to off-site actions at other receptors.

These drugs also induced a variety of neurological adverse effects seen both acutely [extrapyramidal side effects (EPS) such as parkinsonism, dystonia, akathisia, as well as neuroendocrine effects] and with long term exposure [e.g., tardive dyskinesia (TD)] [33]. These latter effects seem mediated by the same receptors (dopamine D₂-like) that appear to cause therapeutic effects. Indeed, a majority of patients given recommended therapeutic dosages of typical antipsychotics develop acute EPS [34]. The more worrisome neurological side effect, tardive dyskinesia (TD), can be irreversible, and incidence rates were estimated at about 5% per year [35]. Anticholinergic drugs are effective in attenuating EPS, but have their own unpleasant side effects (e.g. dry mouth, constipation, delirium and memory deficits), further contributing to treatment non-compliance.

Worse, negative symptoms, mood symptoms, and cognitive deficits are only minimally responsive to conventional antipsychotics. The presence of negative symptoms and cognitive impairment often leads to poor social and vocational function [36, 37]. Moreover, approximately 30% of patients with acutely exacerbated psychotic symptoms have minimal response to standard antipsychotics [35, 38]. These factors spurred research for new drugs that would have equal or better efficacy, and markedly decreased side effects. For more than a decade, the side effects that were considered most critical were the neurological (EPS and TD) and neuroendocrine. The first drugs devoid of neurological side effects were originally termed “atypical” antipsychotics, but are now often called second generation.

3. Second generation antipsychotics (the “atypicals”)

For many years there was the common view that effective antipsychotic agents would necessarily produce EPS because these actions were mediated by blockade of the same receptor (i.e., D₂). A great deal of research was devoted to the discovery of drugs that were “atypical” -- although there was no convention about the meaning of the term “atypical.” In its broadest sense, it was used to refer to drugs that had at least equal antipsychotic efficacy to the “typical” drugs, without producing EPS or sustained prolactin elevation [39]. With time and after the development of drugs that could be called “atypical,” the definition was often expanded to include compounds that might have superior antipsychotic efficacy (e.g., in treatment resistant patients) or have beneficial effects against negative symptoms and/or cognitive deficits.

The first drug that was sometimes called atypical was thioridazine (including its metabolite mesoridazine that was also marketed separately). Although once considered the prototypical atypical antipsychotic [40], thioridazine appears to be a moderate affinity antidopaminergic that was different largely because of its intrinsic antimuscarinic affinity [41-43]. There were no major advantages over the concomitant use of a titered dose of an anticholinergic in combination with other first generation drugs. It was the two other types of drugs that were developed that formed the backbone of what became the true atypical antipsychotics, and now make up what are termed second generation drugs. The most remarkable of these was clozapine.

a. “Rich” pharmacology: clozapine and related drugs

Clozapine underwent extensive clinical testing in the 1970's, but its development was halted in the United States and limited in other countries because of the incidence of the potentially fatal side effect, agranulocytosis. Yet clozapine was reborn because it was recognized that it was superior to traditional antipsychotics such as haloperidol and chlorpromazine in treatment of refractory schizophrenia [44]. Clozapine has little liability to induce EPS or TD, and causes far less of an increase in prolactin concentration relative to typical antipsychotics [45]. It has also been suggested that it may ameliorate (or at least not worsen) some of the negative symptoms and cognitive deficits in schizophrenia [46]. Since its re-introduction, the unparalleled clinical efficacy of clozapine has been widely recognized such that it is used despite the need for slow titering and for monitoring for blood dyscrasias. Indeed, this led to broad efforts to both understand the mechanisms that make clozapine unique, and to develop drugs that retain its remarkable efficacy but have fewer side effects.

Clozapine possesses many characteristics that distinguish it from the typical antipsychotics [37, 46]. The drug has modest affinity for some dopamine receptors (DARs) (D₄, as well as D₂ and D₁), but also has affinity for serotonin (5-HT) 5-HT_2A, H₁ histaminergic, muscarinic, and α-adrenergic receptors [47]. Olanzapine has both structural and pharmacological similarity to clozapine, and has affinity for 5-HT_2A, D₂, as well as other dopamine, serotonin, adrenergic, and histamine receptors. Quetiapine and ziprasidone differ somewhat from clozapine in relative affinities for dopamine and a few of these “off-site” targets [48]. The clinical significance of such differences sometimes appears clear (e.g., the 5-HT_1A partial agonist properties of ziprasidone may contribute to low EPS). Yet with such a diverse pharmacological profile, it is not surprising there has been controversy about which pharmacological actions of clozapine are responsible for its clinical efficacy. Because of the needed dose titering, significant occurrence of agranulocytosis and seizures, as well as sedation, hypotension, hypersalivation, and weight gain, there was a broad search for improved drugs that resulted in the marketing of olanzapine, quetiapine, and ziprasidone. There have been numerous reviews [36, 37, 49] of the pharmacology of these agents, all of which share the property of having affinity for numerous receptors that might be related to antipsychotic action and/or side effects. Elucidation of a verifiable single mechanism has escaped the field, and indeed led to the suggestion that the unique actions of clozapine were a result of multi-receptor targeting. Because the exact nature of the multiple receptor actions is critical but not well understood, it has been suggested that the term “dirty” drug should be supplanted by the euphemism “rich” pharmacology [49]. Other drugs with “rich” pharmacology like olanzapine, quetiapine, and ziprasidone are clearly effective and do not have the hematologic side effects of clozapine, but none quite matches the efficacy of clozapine [50]. Thus, the question of mechanism (is it a mixture of receptor actions or a unique effect at one of the target receptors?) is more than of heuristic interest.

b. Dopamine-serotonin antagonists

The second generation drugs include compounds that also work somewhat differently than clozapine. Unlike the relatively low D₂ and 5-HT_2A affinity of clozapine, ziprasidone, and quetiapine, this mechanism involves high affinity blockade of both serotonin 5-HT_2A and dopamine D₂ receptors [51, 52]. Antipsychotics with markedly lower propensity to cause EPS and TD are thought to work by “serotonin-dopamine antagonism”. The prototype of such drugs is risperidone (see Figure 2), although there are several structurally and mechanistically-related compounds approved or in the pipeline. Besides being a high affinity 5-HT_2A and D₂ antagonist, risperidone is relatively devoid of anticholinergic activity and has lower affinity for α-adrenergic and histamine H₁ receptors [53]. At high doses, risperidone causes some dose-related EPS (> 6 mg/day), but the risk of neurological side effects is modest at lower doses [54]. Risperidone causes hyperprolactinemia by virtue of its blockade of D₂ receptors, and is associated with modest weight gain that is a side effect of many first and second generation drugs.

Figure 2.

Structures of representative second generation drugs. The top row shows clozapine and two compounds in which the structural similarity can be easily seen. The bottom row shows the structure of ziprasidone, a fourth compound with “rich” pharmacology, as well as risperidone, the prototype of the “DA-5-HT potent dual antagonist.”

B. Third generation antipsychotics: partial agonists or something more?

1. Auto- and presynaptic dopamine receptors as antipsychotic candidates

One of the major driving forces for antipsychotic drug discovery focusing on the D₂ receptor was the “dopamine hypothesis of schizophrenia” which, in its 1980's rendition, posited that excess dopamine release and/or excess sensitivity of dopaminoreceptive neurons, was the cause of what are now called positive symptoms. Although antagonism of D₂ receptors has been accepted as a way of controlling positive symptoms of schizophrenia [3], a novel hypothesis was developed in the late 1970's. Dopamine autoreceptors (now known to be composed of high densities of D₂ and low densities of D₃ receptors) [11]. When activated, they cause a decrease in both synthesis and release of dopamine, and a decrease in the firing of dopamine neurons [11]. Interestingly, both dopamine and D₂ agonists tend to have higher potency at these autoreceptors versus what are regarded as postsynaptic D₂ functions [55]. This would then explain why the behavioral response to a full D₂ agonist is typically biphasic with respect to dose, with inhibition seen at in low doses (the result of autoreceptor stimulation) and stimulation at higher doses (direct postsynaptic activation). One of the major mechanisms explaining this biphasic effect is greater presynaptic D₂-like receptor reserve [56].

These observations led to the apparently contradictory idea of also using D₂-like agonists to cause antipsychotic effects via one of two alternative mechanisms. One was to use a low dose of a full agonist to achieve selective presynaptic activation, the other to use a partial agonist. According to pharmacological theory, a partial agonist should be far more efficacious at activating the presynaptic receptors where there is high receptor reserve [57]. Indeed, a plethora of in vitro and animal studies suggested that the partial agonist, (-)3-PPP (preclamol), might be an excellent candidate. Although both mechanisms have a sound theoretical basis, the early clinical data were disappointing [58-63]. In retrospect, this may have reflected issues such as getting the “right” presynaptic relative receptor occupancy without temporal fluctuations, or finding a partial agonist with just the “right” intrinsic activity. On the other hand, there is a widely-held belief that a partial agonist with the “just right” degree of partial agonism has now been found, aripiprazole.

2. Aripiprazole as the third generation antipsychotic prototype

Aripiprazole is a relatively new approved antipsychotic drug proffered by its developers as a high affinity, low intrinsic activity partial D₂ agonist. Although the compound has effects on several other receptors, many of the leading figures in schizophrenia biology have labeled aripiprazole as the first “dopamine stabilizer” based on these purported D₂ partial agonist properties [64-66]. According to this view, in situations of high extracellular dopamine concentrations (e.g., in mesolimbic areas involved in positive symptoms), the partial agonist properties of aripiprazole compete with dopamine and cause partial antagonism offering clinical benefit. Conversely, in situations where extracellular dopamine concentrations are low (for example in dopamine circuits involved in working memory), the drug can occupy additional receptors and cause partial activation. A cartoon indicating how this is proposed to occur is shown in Figure 3. On its face, this seems to provide a logical and cogent mechanism that combines classic pharmacological logic about mechanisms of partial agonism with recent information about the biology of schizophrenia.

Figure 3.

Cartoon illustrating how the proposed D₂ dopamine partial agonist mechanism works in third generation antipsychotics. Left column: mesolimbic dopaminergic transmission. Right column: Prefrontal cortical dopaminergic transmission. Broad dotted line: “normal” dopamine activity; solid dark box: Abnormal transmission in schizophrenia; Solid sigmoidal line: actions of partial agonist alone. Dotted sigmoidal line: actions of partial agonist in the presence of endogenous concentrations of dopamine.

In some functional assay systems, aripiprazole is indeed a low-to-moderate intrinsic activity partial agonist [67-69] as required by this prevalent hypothesis. On the other hand, much of the available data with aripiprazole are problematic as it relates to this partial agonist hypothesis. The intrinsic activity and potency of aripiprazole for the D₂-mediated inhibition of cAMP accumulation is cell line-dependent. Thus, the drug has weak partial agonist activity in the CHO-D_2L cell line, but strong partial agonist activity in HEK-D_2L cells [67-69]. In addition, aripiprazole has markedly different potencies at two D_2L-mediated functions within the same cell line [70], or even at the same function in two different cell lines [69]. Moreover, aripiprazole is a pure antagonist at both D₂ agonist-mediated GTPγS binding and GIRK channel activity [69], whereas it is a full agonist in situ for D₂-mediated inhibition of tyrosine hydroxylase [71]. Thus aripiprazole appears to elicit D₂-mediated functional effects that encompass the whole range of classic pharmacological intrinsic activity. This degree of functional discrimination is not seen with other “partial agonists.” These large variations in both intrinsic activity and potency, not explicable by other mechanisms, that have led to the hypothesis that aripiprazole is “functionally selective” [67, 69, 72] at D₂ receptors, not just a simple partial agonist. It is now clear that “functional selectivity” is not a mechanism unique for aripiprazole.

3. Functional selectivity and the notion of intrinsic efficacy

For the last half century, pharmacological theory has posited that ligands could be characterized by the nature of the functional effects elicited by their interaction with their target receptors [73]. These effects are governed by two important properties: affinity, the property of attraction between a ligand and its receptor, and efficacy, the property that allows ligands, once bound, to produce a response [74]. This concept has led to the classification of a drug as a full agonist, a partial agonist, a neutral antagonist, or an inverse agonist. The notion was that each drug had innate “intrinsic efficacy,” a measure of the stimulus per receptor molecule produced by a ligand at its target receptor [75]. A corollary of this idea was that the ability of a ligand to impart (or reduce) stimulus once that ligand is bound to the receptor is an inherent property of the ligand-receptor complex. Intrinsic efficacy is thus differentiated from the more operational term “intrinsic activity” (Ariens, 1954) that refers to the maximal effect (E_max) of a ligand relative to a reference agonist in any single given experimental system. Historically, pharmacologists have viewed intrinsic efficacy as a system-independent parameter that is constant for each ligand at a given receptor, irrespective of where the target receptor is expressed. Differences in the degree of agonism, for example, in different tissue types were assumed to reflect differences in receptor density and/or the strength of stimulus-response coupling. This classification of drug action based on intrinsic efficacy allowed for variations in the quantity of the stimulus that was imparted to the cell, but not the quality. Thus, a full agonist would be expected to activate all of the signaling pathways linked to a receptor to the same degree as the endogenous ligand for that receptor. During the past 15 years, however, a huge body of data has emerged suggesting that the premise of “intrinsic efficacy” may be invalid.

4. Early evidence for functional selectivity of dopaminergic compounds

To our knowledge, the first clear example of dopamine receptor functional selectivity resulted from serendipitous findings with dihydrexidine, a compound originally designed to be a selective D₁ agonist [76]. After dihydrexidine was found to have D₂ affinity [77, 78], both it and a more D₂ selective analog N-n-propyldihydrexidine (propylDHX) were characterized functionally. From both binding and canonical functional assays, both compounds appeared to be full D₂ agonists. They competed for D₂ receptors in heterologous systems or in brain tissue with shallow, guanine-nucleotide-sensitive curves similar to typical agonists [78-80], and were full agonists at D₂ receptor mediated inhibition of adenylate cyclase [79, 80]. Consistent with this, they also inhibited prolactin secretion in vivo.

Unexpectedly however, dihydrexidine and propylDHX bound to D₂–like presynaptic/autoreceptors that mediate inhibition of dopamine neuron firing, dopamine release, or dopamine synthesis [79, 80], but had only antagonist properties. Off-site actions were excluded by both receptor screening and competitive pharmacological studies [78]. These agonist-antagonist actions of a single drug at a single receptor could be replicated in heterologous or single cell systems. Thus, in pituitary lactotrophs, dihydrexidine was a full agonist at D₂-mediated inhibition adenylate cyclase, yet had little intrinsic activity at D₂ receptors coupled to G protein-coupled inwardly-rectifying potassium channels (GIRK), and was an effective antagonist of the functional actions of dopamine [79]. In mesencephalic-derived MN9D cells transfected with the D_2L receptor, both compounds were full agonists at inhibition of adenylate cyclase activity, yet were antagonists at D_2L mediated inhibition of depolarization-induced release of dopamine [80]. Neither dihydrexidine nor propylDHX had effects in untransfected MN9D cells not expressing the D_2L receptor.

In toto, these studies led to the formal idea, to wit, that “functional selectivity” [is] the concept that [some] drugs can cause functional multiplicity even when interacting with a single receptor isoform. …. One mechanism for functional selectivity may be atypical conformation changes [when compared to the endogenous ligand] induced when such drugs bind to the receptor-G protein complex. The particular type(s) of G-protein are dependent on both the type of cell, and the location in the cell, where the receptor of interest is located. Such functional targeting allows drug effects to be refined to a degree not possible just by targeting specific receptor isoforms. This could yield important therapeutic advances, although it introduces a new level of complexity that will require significantly greater understanding of receptor dynamics and the interaction with transduction mechanisms” [81]. This may have been the earliest conceptualization linking a speculative hypothesis [82] to actual data [81], and was soon followed by other reports of similar functional “mismatches” [see 83 for another early example].

5. Is functional selectivity an artifact or a relevant pharmacological mechanism?

For some time, the overall implications of functional selectivity to drug action in vivo was not widely accepted despite some of the available data. For example, classic pharmacology would have predicted that a compound with full intrinsic activity at both D₁ and D₂ receptors (i.e., dihydrexidine at both D₁ and D₂ receptors modulating adenylate cyclase) would have behavioral effects similar to apomorphine or amphetamine, direct or indirect activators of both families of receptor. Clearly, however, dihydrexidine did not have the expected behavioral actions of a mixed D₁:D₂ agonist [84]. This was consistent with the hypothesis that the atypical D₂ signaling properties of the this compound markedly affected its pharmacological actions in vivo.

Equally compelling were studies done with the D₂-selective analog of dihydrexidine, propyldihydrexidine (propylDHX). D₂-like agonists are known to have biphasic behavioral effects, causing locomotor inhibition at low doses and locomotor stimulation and stereotypies at higher doses [85]. PropylDHX, however, only caused modest locomotor inhibition across a wide range of doses, with no competing behaviors that might have interfered with locomotion [86]. This provided additional evidence that functionally selective compounds may well have novel pharmacological actions. As we have summarized recently [72], there is a growing body of data demonstrating functional selectivity of ligands for dozens of receptors. The pharmacological impact of such differential signaling is obviously an important area of future investigation.

6. Does functional selectivity explain more of the actions of aripiprazole?

As discussed earlier, there are clear data demonstrating functionally selective properties of aripiprazole that are inconsistent with simple partial agonism [67, 69, 70]. It may be argued that the clinical effect of aripiprazole can be explained by either mechanism, leaving a clear answer currently in abeyance. Conversely, there are behavioral data that also are inconsistent with the partial agonist mechanism. Aripiprazole has been studied in the unilaterally lesioned 6-hydroxydopamine (6-OHDA) treated rat [87]. This is an extremely useful pharmacodynamic model (sometimes proposed to be a rat model of Parkinson's disease) in which both full and partial dopamine agonists cause the test animal to turn contralaterally in a tight circle towards the side away from the 6-OHDA lesion (i.e., leftward turning if the lesion was to the right-side nigrostriatal pathway). This robust rotation is a result of relative receptor/cellular hypersensitivity of the target receptors on neurons in the lesioned, dopamine-depleted, striatum. If aripiprazole was a partial agonist, it should also cause such contralateral rotation, but it does not [71]. Moreover, not only does it fail to cause rotation on its own, but it is a pure antagonist of the turning caused by known dopamine D₂ agonists [71]. Here then is a situation in which aripiprazole is a pure antagonist in a system with extremely low dopamine tone. These findings are directly contradictory to what is predicted by the partial agonist hypothesis, but consistent with a functional selectivity mechanism.

Interestingly, there is some clinical lore that also may be relevant. Many Parkinson's disease (PD) patients develop psychotic side effects as a result of their use of levodopa and/or dopamine agonists (the latter largely working via D₂ receptors). Clozapine and quetiapine are particularly effective in treating these side effects without worsening Parkinson's symptoms. By similar reasoning as espoused for schizophrenia, aripiprazole (as a partial agonist) also should be very useful in treating these psychotic symptoms when added to the dopaminergic regimen of the PD patient. In fact, aripiprazole not only lacks effectiveness in treating these psychoses, it tends to worsen motor function [88]. These data also are inconsistent with the partial agonist hypothesis.

Thus, the available data better support the hypothesis that aripiprazole works as a functionally selective D₂ ligand where its intrinsic activity varies markedly depending on the signaling environment of the D₂ receptor [67, 69, 72]. It is fascinating to hypothesize how this specifically translates into functional changes in the intact organism. It may be, for example, that the cellular changes that cause post-denervation supersensitivity also change the D₂ signaling environment such that aripiprazole loses intrinsic activity. This would explain the results in the unilateral 6-OHDA rats or in PD patients. Conversely, it might be supposed that, at the critical D₂ receptors in the dopamine mesolimbic terminal fields, aripiprazole is a low intrinsic activity ligand. This may explain the common observation that while well-tolerated for this class of medication, the drug may be slightly less effective than compounds with pure D₂ antagonist effects.

C. Impact of functional selectivity and research issues for next generation drugs

1. Complexity of signaling through D₂ receptors

It is now clear that signaling through modulation of adenylate cyclases is only a fraction of how dopamine receptors modulate cellular function [89, 90]. The first signaling pathway identified for D₂-like receptors was inhibition of cAMP accumulation [91, 92], and D₂ receptors inhibit adenylate cyclase in most clonal and in situ cells [93]. In situ, D₂-like inhibition of adenylate cyclase can lead to presynaptic/autoreceptor decreases in tyrosine hydroxylase activity and in firing of nigrostriatal dopamine neurons. Like other Gα_i/o-coupled receptors, D₂-like receptors modulate many signaling pathways including phospholipases, ion channels, MAP kinases, and the Na⁺/H⁺ exchanger, as well as adenylate cyclases [94]. D₂ activation has major effects on K⁺ currents via dissociation of G_βγ subunits. D₂-like receptors activate a G protein-regulated inwardly rectifying potassium channel (GIRK or Kir3), modulating several potassium currents in various tissues [95-98]. Dopamine release-regulating autoreceptors are coupled to potassium channels [99] rather than to inhibition of adenylate cyclase [100], and there is robust regulation of GIRK currents by D₂ receptors in substantia nigra dopamine neurons [101].

The D₂-like receptors also affect calcium [102-106] and sodium channels [107, 108]. MAP kinases are another set of pathways that can be modulated by D₂ activation [109-116] and stress-activated protein kinase/Jun amino-terminal kinase (SAPK/JNK) [111]. The D₂ receptor can potentiate arachidonic acid release induced by calcium-mobilizing receptors, a response mediated by cytosolic phospholipase A₂ [117-120], and also directly. The D₄, but not D₃, receptor also activates this pathway [120, 121]. Arachidonic acid and its lipoxygenase and cyclooxygenase metabolites have numerous effects on cellular function, including feedback regulation of D₂-like signaling and dopamine transport [122-125]. Finally, the D₂ receptor can also stimulate phospholipase D, catalyzing the hydrolysis of phosphatidylcholine to form choline and phosphatidic acid [126, 127].

A major mediator of D₂-like signaling is the Gα_i/o class of G proteins that are inactivated by pertussis toxin-catalyzed ADP-ribosylation [128, 129], but as noted earlier, non-G protein signaling has become topically important. Most evidence relates to β-arrestin2, a key member of the GPCR desensitization machinery, to mediate downstream signaling through PKB(AKT) and GSK3 [130, 131]. The evidence is based on studies using a variety of genetically altered mice that were deficient in major constituents of GPCR signaling, including GPCR kinases (GRKs), β-arrestins, the dopamine transporter (DAT), and PKB. The data showed simultaneous recruitment of both protein phosphatase 2A (PP2A) and PKB by β-arrestin2 resulting in the dephosphorylation of PKB, and the subsequent loss of the inhibitory effect of PKB on the GSK3 signaling pathway, something indispensable for neuronal survival and differentiation [130-132]. Interestingly, the time course of β-arrestin/PKB signaling at the D₂ receptor takes much longer to occur (hours versus minutes) compared to the G protein-mediated cAMP/PKA response [130, 133, 134]. The slow response of the β-arrestin/PKB pathway can also be indicative of its downstream effects on gene transcription culminating in target protein transcriptional profile changes or even chromatin remodeling [135-138].

In translating findings from in vitro systems, receptor selectivity is a major confound. One complication comes from the fact that there are few, if any, drugs that are highly selective for only one of the D₂-like isoforms [89, 90, 139]. Moreover, there are interesting allelic variants of the D₄ receptor, and two alternative splicing-derived isoforms of the D₂ receptor with different third cytoplasmic loops [i.e., D_2Long (D_2L) and D_2Short (D_2S)]. Coupled with the “off-site” actions that may very well contribute to both useful and side effects, it is not surprising that establishing an unequivocal mechanism(s) of action has been difficult.

2. Consequences of differential effects on signaling pathways mediated by a single receptor

The previous section provides an overview of how some ligands may cause markedly different patterns of signaling via occupation of D₂ receptors. It is this complex matrix of signaling that we have hypothesized is a major mechanism in the unexpected behavioral effects of both experimental drugs and aripiprazole [72, 90, 140]. These manifold, often inter-related signaling patterns underscore the complexity that offers both research challenges and drug discovery opportunities. In terms of the latter, one might consider the D₂ receptor signaling through canonical G protein mechanisms and for non-canonical mechanisms. The potential impact of a functionally selective ligand is illustrated in Figure 4.

Figure 4.

Cartoon illustrating how functional selectivity at a single receptor might lead to an altered balance of therapeutic and side effects [Adapted from 140]. In this example, the “typical” ligand [Left panel] is an agonist that activates two signaling pathways via single receptor. One of these pathways is specifically linked to a therapeutic effect, the other to a side effect. The “functionally selective” ligand [right panel] fully activates only the pathway linked to the therapeutic effect, thereby decreasing side effects mediated by this single receptor.

In theory, it should be possible for a ligand to discriminate among canonical functions mediated by different G protein heterotrimers, but also to cause differential longer-term responses. One of the critical challenges is how to develop high-throughput, physiologically-meaningful screening systems that can distinguish novel drug candidates. The issues inherent in such approaches are not only the assays themselves, but also the cellular background that one chooses to use. It is fair to say that functional selectivity has invigorated the interest in these issues.

3. Dopaminergic mechanisms of candidate compounds

Resolving the alternate hypotheses noted above, as well as determining the true clinical utility of functionally selective drugs, will require other compounds with functionally selective (partial agonist?) properties in vitro that are tested widely in humans. As noted above, our view is that the available data support the hypothesis that functionally selective D₂ properties of aripiprazole (not D₂ partial agonist effects) underlie its unique clinical profile. It is useful to briefly review some of the other compounds whose data may now, or in the future, bear on this question. The following are a few of the purported D₂ partial agonists or otherwise novel D₂ like ligands that may play a role in understanding these issues better (see Figure 5 for key structures).

Figure 5.

Structures of some of the mechanistically interesting compounds.

a. Compounds of specific historical or clinical importance

Preclamol [(-)-3PPP] (Figure 5) is the compound that may be considered the prototype of dopamine agonists with novel actions. The two enantiomers of 3PPP show different, sometimes opposite, effects at dopamine synaptic functions (Hjorth et al 1983). Both isomers are partial D₂ agonists, with the (+) isomer having somewhat lower affinity and lower intrinsic activity partial agonist. Preclamol [(-)-3PPP] preferentially stimulates the dopamine autoreceptor, and has some limbic-selectivity as a partial D₂ agonist. At higher doses it has antagonist effects at the normosensitive postsynaptic D₂ site [141]. Based on these data and the view of schizophrenia illustrated in Figure 3, preclamol was tested for antipsychotic potential, but any immediate effects were not sustained with repeated dosing as occurs with accepted antipsychotic drugs [63]. Although these results did not directly support the partial agonist hypothesis, the short term benefit was taken by many as a reason to pursue this direction. Contemporaneously, Otsuka developed OPC-4392 (Figure 5), a quinolinone derivative that is an aripiprazole predecessor. It was reported to act as an agonist at presynaptic dopamine autoreceptors decreasing synthesis and release of dopamine, while acting as an antagonist at postsynaptic D₂ receptors by inhibiting stereotyped and climbing behavior induced by apomorphine in mice [142-144]. Although OPC-4392 was withdrawn after early clinical trials because it worsened psychoses and increased suicidal ideation, it ultimately led to the discovery of aripiprazole. If the differences in functional selectivity between OPC-4392 and aripiprazole can be elucidated, it very well may offer clues as to the discovery of even better agents.

Substituted benzamides

The benzamides form an interesting family of dopamine receptor antagonists and potential antipsychotics whose pharmacology may well impinge on mechanisms of functional selectivity. In general, the benzamides have preferential affinity for dopamine D₂-like receptors, and many of them have atypical pharmacology in brain tissue that is not completely understood. Some of the early benzamides that have been tested clinically include nemonapride (YM-09151-2) and remoxipride. Radiolabeled nemonapride is a commonly used research tool, and raclopride (FLA 870; A40664) and eticlopride (A38503; FLB131) are high affinity D₂-like ligands that have been used as PET ligands. Of the substituted benzamides, sulpiride is often considered the prototype, and has been widely studied, but only amisulpride (Figure 5) is currently used clinically although many other members of this class have shown varying degrees of clinical efficacy and side effects. Amisulpride has D₂/D₃ selectivity, and low affinity for other dopamine or serotonergic, alpha-adrenergic, histaminergic or cholinergic receptors. It is equieffective to haloperidol and α-flupenthixol for positive symptoms, and at lower does is reported to be significantly more effective in reducing primary negative symptoms than placebo. The clinical properties of amisulpride appear more “atypical” than other members of this family (e.g., sulpiride) despite having similar receptor profiles. The underlying mechanisms for this are unclear, but may well involve functionally selective signaling at the D₂ or other receptors.

Other compounds with partial agonist/functionally selective D₂-like properties

There are many hints in the literature of compounds that might have functionally selective properties at D₂ receptors. Talipexole (B-HT 920) initially was suggested as a partial agonist at dopamine D₂ receptors because of its ability to inhibit both synthesis and utilization of dopamine but not produce any stereotypies on hyperactivity. Studies with stimulation of presynaptic D₂ dopamine receptors by talipexole in comparision with apomorphine inhibited the synthesis of dopamine in the corpus striatum of γ-butyrolactone-treated mice to same extent suggesting that B-HT 920 to be a full agonist at presynaptic D₂ dopamine receptors [145]. Although tested against negative symptoms of schizophrenia [146], talipexole later became a candidate for Parkinson's disease therapy. SDZ HDC 912 is an ergoline derivative dopamine D₂ partial agonist that has intrinsic agonist activity adequate to maintain dopaminergic transmission sufficient to avoid extrapyramidal side-effects as well as hyperprolactinemia [147]. Although the compound had partial agonist properties like the aripiprazole precursor OPC-4392, it also differed in several ways, suggesting it might be functionally selective [148].

Roxindole (EMD 49980) was reported to be a selective agonist of presynaptic dopamine D₂ receptors with little D₁ affinity. It is heuristically less interesting because it has higher affinity for the D₃ and D₄ receptors, and weak partial agonist activity at D₂ receptors [149]. Other compounds of potential interest include UH232 [23, 150], S32504 [151], SDZ 208-912 [152, 153], PD 143188 (CI-1007) [154], SLV-308 (SME-308) [155, 156], F15063 [157, 158], SSR181507 [159], (-)-OSU6162 (PNU-96391A; Figure 5) [160, 161], ACR16 [162, 163], and many others. What is of particular importance is that there has been the view that dopamine stabilization (i.e., the unique clinical effects of aripiprazole) can be explained simply “partial agonist intrinsic activity.” Thus, Tadori et al. [164] compared aripiprazole, bifeprunox, SDZ 208-912, OPC-4392, and ACR16 in terms of degrees of intrinsic activity (i.e., inhibition of forskolinstimulated cAMP accumulation) in clonal CHO cell lines expressing high and low densities of human dopamine D_2L and D_2S receptors. In cells with lower receptor densities, all drugs had partial intrinsic activity (SDZ 208-912 < aripiprazole < bifeprunox=OPC-4392). They concluded that this specific degree of partial agonism was the reason for efficacy in both positive and negative symptoms of schizophrenia, as well as the demonstrated low liability for parkinsonism and hyperprolactinemia. We believe that this explanation fails to account for the clinical failure of preclamol, a compound with intrinsic activity similar to aripiprazole in many systems. Coupled with the fact that aripiprazole has varied intrinsic activity depending on the assay system [67, 69, 70], we feel that it is timely to consider the complexity of D₂-mediated signaling as an alternate mechanism of importance.

b. Examples of recent pipeline drugs

Bifeprunox (DU 127090) (Figure 5) is a novel atypical antipsychotic candidate whose early development was by Solvay-Lundbeck pharmaceuticals. It is a full-spectrum dopamine D₂ receptor partial agonist and also possesses 5-HT_1A receptor agonism. It exhibits low intrinsic activity at the D₂ receptor and high affinity and moderate to high intrinsic activity at the 5-HT_1A receptor. Interpreted as a partial D₂ agonist with 5-HT_1A activity, it is theoretically similar to aripiprazole. Although bifeprunox was shown to be more effective than placebo in the treatment of schizophrenia [165], the US FDA rejected an NDA for it on both efficacy and safety grounds [166]. If functional selectivity rather than partial agonism is indeed the critical mechanism, this leads to the hypothesis that the signaling profile of bifeprunox and aripiprazole differ.

Aplindore (DAB-452) (Figure 5) was another of the partial agonist family that was in development at Wyeth as an atypical antipsychotic. Aplindore has high affinity for D₂/D₃ receptors, but unlike aripiprazole, has low affinity for 5-HT_1A. In a variety of assays (GTPγS binding, MAPK, and calcium flux), the intrinsic activities of aplindore are higher than aripiprazole, and unlike aripiprazole, aplindore induced D₂-dependent contralateral turning in unilaterally 6-OHDA lesioned rats [167]. The schizophrenia indication for aplindore has now been abandoned, and Neurogen is pursuing indications for Parkinson's disease and restless leg syndrome. These data may be interpreted as additional evidence for the importance of functional selectivity, or that the intrinsic activity of a partial agonist must be very low for the drug to be effective in schizophrenia.

4. Challenges for discovery of functionally selective drugs

As the foregoing material makes clear, we believe that the actions of aripiprazole are due to its functionally selective D₂ dopamine receptor properties, as well as possibly involving off-site receptors such as the 5-HT_1A. This hypothesis suggests that even better drugs may be possible by understanding the specific signaling mechanisms that affect the actions of aripiprazole in humans. In this regard, almost every laboratory that played a role in bringing functional selectivity into the pharmacological mainstream also concluded that this mechanism was both a great complication and a great opportunity for drug discovery [72]. As the pharmaceutical industry has become aware of both the benefits and challenges, the obvious question is how to incorporate this mechanism into the discovery of novel compounds.

The lack of a totally predictive animal model for complex psychiatric disorders like schizophrenia is a major research handicap. Indeed, the validity of our models is often dependent on the fact that they respond to current drugs, creating some inherent circular logic. This complicates the selection of novel compounds that might enter preclinical development. Functional selectivity, per se, complicates the issues inherent in setting the parameters for high- or medium throughput screening. Even accepting that aripiprazole may work primarily as a functionally selective D₂ ligand (as opposed to a partial agonist), it is challenging to decide how to assay next-generation aripiprazole-like ligands to select the best of the potential drug candidates. One of the critical issues may be the host cell line to study. The properties of a drug candidate when studied at D₂ receptors expressed heterologously in non-neural cell lines (e.g., HEK293 or CHO) may differ from neurally-derived lines (e.g., C6, MN9D, SK-N-MC, etc.). Of course, even responses in the latter may differ from those of native neurons. These issues are clearly illustrated in the milieu-specific effects that have been shown for aripiprazole [67, 69, 70]. This issue has critical ramifications for drug discovery, especially in the adaptation of systems that while providing very high throughput and high Z-scores, may not always reflect the relevant properties of ligands that are being screened. The impact of cellular milieu cannot be overstated.

As one recent example, Heusler et al. [158] studied the effects of antipsychotic drugs or drug candidates to cause internalization of the human D_2S receptor using hemaglutinin (HA)-tags in HEK293 cells in which there was also increased co-expression of G protein-coupled receptor kinase 2 (GRK2) and β-arrestin2. Dopamine, quinpirole and bromocriptine behaved as full agonists, whereas S(-)-3-(3-hydroxyphenyl)-N-n-propylpiperidine [preclamol (-)-3PPP] and sarizotan were partial agonists. All of the antipsychotics tested (haloperidol, olanzapine, nemonapride, ziprasidone and clozapine) failed to internalize D_2S receptors, whereas aripiprazole had >50% intrinsic activity. Bifeprunox and SSR181507 also were partial agonists whereas SLV313 and F15063) were inactive. Conversely, Urban et al. [72] used FLAG-tagged D_2L receptors ands reported that neither preclamol nor aripiprazole had intrinsic activity. These data underscore how subtle experimental differences may lead to very different conclusions. It is unclear whether the differences were due to the nature of the receptor tag, the use of short- versus long isoform, or the co-expression of GRK2 and β-arrestin2, or some other factor. The task of extrapolating or integrating in vitro functional selectivity findings thus remains a major challenge. We hope that the examples we have provided, however, suggest that demonstrating partial agonism in a single, often non-physiological, assay system is neither mechanistically informative, nor useful in selecting novel drug candidates.

5. Is the “generational” nomenclature right for antipsychotic drugs?

For the last half-century, all approved antipsychotic drugs have had affinity for dopamine D₂-like receptors as an apparently essential aspect of their mechanism of action. There is no fundamental neurobiological reason, however, that requires this to always be the case. Conversely, there is evidence that several non-dopaminergic approaches might lead to a novel antipsychotic drug. Thus, there has been interest in selective ligands for various serotonin receptors (5-HT_2A, 5-HT_2C, 5-HT₇), whether these be agonists, antagonists, or functionally selective ligands. Another research focus has been on glutamatergic interventions, either via modulation of ionotropic or metabotropic glutamate transmission. Moreover, novel therapies may come from treatment of currently resistant facets of schizophrenia (e.g., decreasing cognitive deficits via α₇ nicotinic agonists or D₁ dopamine agonists), or from targets with unexpected actions (e.g., drugs acting at a previously orphan receptor, acting at a known receptor with unexpected effects, or even a more distal aspect of a signaling cascade).

In the interest of effective communication, we would suggest that the field should consider a nomenclature that will accommodate novel directions that are likely to emerge. If indeed several of the therapeutic directions noted above come to pass, the use of “X generation” terminology may quickly lead to a Tower of Babel that, while interpretable to those immersed in the field, will be incomprehensible to practicing physicians, residents, medical students, and even those scientists in closely aligned areas. In other fields (antibiotics, anticancer, etc.), drugs are often categorized according to their mechanism(s) of action. One wonders if communication about antipsychotic drugs might not benefit in a similar way. Thus we could have from the current drugs: “antipsychotics-dopamine antagonists” (1^st generation); “antipsychoticsdopamine-serotonin antagonists” (some 2^nd generation); “antipsychotics-multi-targeted” (other 2^nd generation); and “antipsychotics-dopamine-functionally selective” (3^rd generation). This would then allow growth into possible future classes: “antipsychotics-serotonin 5-HT_2A functionally selective”; “antipsychotics-dopamine D₁ agonists”; “antipsychotics-glutamate metabotropic allosteric modulators”; etc. Although not as simple and elegant as “X Generation antipsychotic,” it might facilitate communication among Generation X scientists.