Insights into the Mechanism of Action of Antipsychotic Drugs Derived from Animal Models: Standard of Care versus Novel Targets

1. Introduction

Schizophrenia is a devastating developmental disorder that arises from an interaction of genetic susceptibility and environmental factors[1,2,3,4], with an incidence of approximately 1- 1.5% worldwide[5,6,7]. The illness has a devastating impact on individuals and their families/caregivers [8], striking during late adolescence/early adulthood[5,9]. Therefore, there has been enormous effort and resources that have been committed to understanding the pathophysiology of this disorder as a means to derive effective treatments. The first treatment developed was based on dopamine antagonism. This was a serendipitous discovery that was made on the basis of an opportune observation that a modified antihistamine, chlorpromazine, was an effective stabilizer for extended surgical interventions [10,11]. In coining the term “neuroleptic” to describe the neural stabilizing action of the drug, Laborit [12] attracted the attention of clinicians working in a mental asylum as a potential pharmacological intervention for a number of at the time untreatable mental disorders. What they found was that the drug was highly effective for those experiencing schizophrenia [13,14,15,16,17]. It was another 20 years before the drug was proposed to exert its therapeutic action via blockade of dopamine (DA) receptors [18,19], and another 12 years before the DA D2 receptor was identified as the binding site of antipsychotic drugs [20]. This, along with evidence of DA releasing drugs exacerbating schizophrenia psychosis [21,22,23] that gave rise to the DA hypothesis of schizophrenia [24,25], which has dominated drug development in the treatment of schizophrenia for decades.

Because DA is involved in the psychotic symptoms of schizophrenia as well as in regulation of movement, the combined antipsychotic efficacy and associated neurological side effects were considered to be necessary for the actions of these drugs, which was termed the neuroleptic profile [11,26]. However, with the discovery of clozapine and other second-generation antipsychotics, it was discovered that one can separate pharmacologically the antipsychotic properties from the neurological, or “extrapyramidal,” side effects [27,28,29]. Interestingly, this was not due to the development of more selective pharmacological agents, but instead relied on actions off-target from the D2 receptor, most prominently the serotonergic 5HT2 receptor [30] and possibly anticholinergic properties [31]. Thus, although these second generation compounds targeted multiple receptors with various affinities, the drugs still needed to be administered at a dose that caused between 60-80% occupancy of the D2 receptor [32,33,34,35].

How did these drugs achieve their action? This was a quandary, since despite the known efficacy at the D2 receptor, there was little evidence for a dysfunction within the DA system itself; particularly when compared to the psychotomimetic effects of amphetamine, that produce psychosis by increasing DA overflow 20-30 fold over baseline [36,37], which was clearly not present in the schizophrenia patient [21]. Furthermore, the antipsychotic drugs did not have a typical pharmacological profile, in that the drugs typically had to be administered several times to achieve therapeutic actions [38], which was unusual since one would predict greatest efficacy at the first dose with the development of tachyphylaxis with additional doses. Instead, these drugs showed increased efficacy without the need to dramatically increase drug doses to adjust for homeostatic compensation to D2 blockade (e.g., increased D2 receptor number, increased DA synthesis capacity, etc.) [38,39]. Thus, the simple model of abnormally high DA transmission could not explain the pathophysiology of schizophrenia or the therapeutic actions of antipsychotic medications.

2. Mechanism of action of D2 antagonist antipsychotic medications

A breakthrough came from animal studies of the actions of repeated antipsychotic drug administration on the DA system. In recordings from identified DA neurons in the brain of rats, it was found that 3+ weeks of treatment with first-generation antipsychotic drugs led to an inactivation of DA neuron firing – a phenomenon known as depolarization block [40,41,42]. It was found that the first generation drugs caused depolarization block in both the nigrostriatal and mesolimbic DA system [40,41,43,44]. Interestingly, the second-generation drugs, which did not have extrapyramidal side effects but still exhibited therapeutic efficacy, only caused depolarization block in the mesolimbic psychosis-related DA system but not in the extrapyramidal nigrostriatal DA system [45,46,47]. Finally, it was shown that a compound with limited antipsychotic actions but prominent extrapyramidal actions, metoclopramide [48], only produced depolarization block in the nigrostriatal system but not the mesolimbic system [46]. Therefore, it was established that repeated antipsychotic drug-induced depolarization block of the mesolimbic system was associated with antipsychotic efficacy, and depolarization block of the nigrostriatal system with extrapyramidal side effects [43,46,47]. While this was a powerful correlation, there were some caveats: 1) antipsychotic drugs do not need to be administered for weeks to obtain therapeutic efficacy in patients, 2) there was still no evidence for a hyperactive DA system, and 3) DA neuron depolarization block is not the normal state of the DA system.

One potential issue with these preclinical experiments is that they were performed on normal rats. It is well known that the brain has extensive homeostatic mechanisms that can compensate for the continued presence of the drug, and it was likely that the delayed action in the normal animal may have been due to these compensatory mechanisms [49]. What about animal models of schizophrenia, in which the system is already disrupted? We have proposed that in a disrupted system, in which homeostatic processes are either disrupted or ineffective, a therapeutic agent would have significantly different actions as compared to a normal system [49]. For this reason, it is essential to test these agents in an animal model that can approximate at least some of the circuit disruptions that clinical studies have found to also be altered in the schizophrenia brain [49,50].

One issue in this approach is identifying an appropriate animal model. While it is clear that one cannot precisely duplicate a complex uniquely human disorder like schizophrenia in a rodent, one can use clinical studies to guide development of an appropriate model and evaluate its functional validity. Evidence showing that schizophrenia likely has a strong genetic component, with the heritability of schizophrenia a function of number of shared genes in families [1,2]. In addition, there is clear evidence that environmental factors can also engender susceptibility to schizophrenia [1,3,4]. Furthermore, studies by Weinberger and others have shown that disruption of hippocampal function early in life will result in a rodent model that can recapitulate a number of features of schizophrenia [51,52,53,54,55,56]. Using this as a base, we used a developmental disruption model based on administration of the DNA alkylating agent methyl azoxy methanol acetate (MAM) given to pregnant rodents at gestational day 17, and testing them as adults [57,58]. Gestational day 17 was chosen to approximate the human second trimester, which is a period of vulnerability in which infections or trauma will increase the propensity to develop schizophrenia in the offspring [59]. We found that the adult offspring of MAM-treated dams could recapitulate a number of behavioral, pharmacological, neuroanatomical, and neural activity states that have been observed in schizophrenia patients [57,58]. In particular, imaging studies in schizophrenia patients have revealed hyperactivity in the limbic hippocampus [60], which is associated with an increase fluorodopa uptake in the associative striatum [61,62]. Similarly, in the MAM rats we observed hyperactivity in the ventral limbic hippocampus and an increase in ventral tegmental area DA neuron activity [63]. This increase in DA neuron activity would thus be consistent with the increased fluorodopa uptake, since fluorodopa uptake is a metric of the number of active terminals [58,61], and we observed an increase in the number of DA neurons driving these terminals [57,58]. The hyperactivity of the hippocampus appears to be driven by a loss of parvalbumin-containing GABAergic interneurons, which is observed in postmortem schizophrenia brains [64,65] as well as in the MAM model of schizophrenia [66,67]. Furthermore, we showed that activation of the ventral hippocampus would, through a circuit involving the striatum and the ventral pallidum, lead to an increase in the number of active DA neurons in the ventral tegmentum [68] (Figure 1). Thus, we propose that the overactive DA system involved in schizophrenia psychosis is due to loss of parvalbumin inhibition of the limbic hippocampus and consequent overdrive of the DA system. Importantly, while the overdriven DA system is likely the source of the psychotic positive symptoms of schizophrenia [69,70], the limbic hippocampus also exhibits projections to areas involved in affective regulation (e.g., the amygdala) and cognition (i.e., the prefrontal cortex) [71,72,73] (Figure 2). Thus, while targeting the DA system may help to relieve psychosis, it will not impact the negative and cognitive symptoms of this disorder.

Given that the DA neurons are not hyperactive individually, but instead there are more of the neurons active, this would cause a stimulus that increases DA neuron firing would have a substantially greater effect. Specifically, if the DA system is hyper-responsive to stimuli, this would cause an inappropriately high DA response to what may be benign stimuli, a condition known as aberrant salience [74,75,76]. This would lead to inappropriate attribution of threat or salience to benign stimuli (i.e., delusions) (Figure 3).

How would DA D2 antagonists help to alleviate this condition? In the normal animal, administration of antipsychotic drugs, by blocking postsynaptic D2 receptors, would cause a feedback activation of the DA system that, over weeks, results in DA neuron depolarization block. However, in the schizophrenia patient the DA system is ALREADY in a hyperactive state [42]. In the MAM rats, we found that administration of a D2 antagonist, unlike in a normal animal, would add to the present hyper-responsive DA system, with the result that depolarization block develops very soon after drug administration [77]. This is consistent with the clinical literature, with the antipsychotic properties produced rapidly after antipsychotic medication initiation. Moreover, the more psychotic the patient, the more rapid onset of therapeutic action [78,79]. Again, this is consistent with the MAM rat, in that the more overdriven the DA system is initially, the more rapid that addition of a D2 antagonist will produce depolarization block. By producing depolarization block and inactivation of DA neuron firing, this would alleviate the pathological increase in the number of DA neurons active and hyper-responsive to stimuli.

This provides a circuit-based assessment of the mechanism of antipsychotic drug treatment alleviation of psychosis. However, this also exposes several caveats: 1) the “normal” state of the DA system is not depolarization block; this may be why antipsychotic drugs produce negative affective states, which may contribute to low patient compliance [80], and 2) this will impact the DA system not by treating schizophrenia at the site of pathology (i.e., the limbic hippocampus), but instead 5 synapses downstream from the defect [76]. Thus, while the D2 blocking antipsychotic drug may alleviate psychosis, it would be ineffective in treating negative and cognitive deficits of this disorder.

How can some of these issues be circumvented? One more recent development is the use of partial DA agonist drugs, such as aripiprazole or brexpiprazole [81]. These drugs will occupy D2 receptors but, rather than producing a complete blockade, will produce a partial activation of the receptor [82]. Thus, these drugs will act by preventing overstimulation of D2 receptors while providing a baseline level of stimulation. For this reason, these drugs are typically administered at doses that occupy D2 receptors approximately 95% without producing excessive blockade-induced negative affect. We found that, unlike first- and second-generation antipsychotic drugs, the partial agonist aripiprazole appears to directly inhibit DA neuron activity without inducing depolarization block [83]. Therefore, while an improvement over receptor blockade-induced depolarization block, the drugs will nonetheless fail to alleviate the negative and cognitive deficits associated with schizophrenia, and will still impact the normal function of the DA system.

3. Novel target agents

As outlined above, current antipsychotic agents, while effective at treating psychosis, are not well-tolerated by the patient and are not effective at alleviating negative and cognitive symptoms of the disorder, which likely underlies low patient compliance [84,85,86]. A more effective approach would be to target the site of pathology proposed to drive the schizophrenia state. As stated above, the current model suggests that parvalbumin neuron loss in the limbic hippocampus is driving the pathological state. Therefore, one potential mechanism to alleviate this dysfunction would be to either increase GABAergic inhibition at the parvalbumin-hippocampal pyramidal neuron site, or to directly decrease excitability of the hippocampal neurons. There are several compounds that were developed to address this. First, Lilly developed an mGluR2,3 agonist pomaglumetad [87]; a drug to target the hippocampal hyperexcitability. Another compound was developed by Roche; a glycine uptake inhibitor [88] to increase glutamatergic NMDA drive on the parvalbumin interneurons. A third drug developed by Pfizer was the PDE10 inhibitor [89] to diminish the postsynaptic actions of the overactive glutamatergic system. In each case, the compounds showed significant efficacy in animal models of these disorder [88,90,91]. Indeed, in the MAM rats, pomaglumetad was found to be highly effective in normalizing hippocampal activity and DA neuron firing [90]. Furthermore, these compounds showed significant promise in the early phase trials [92,93,94,95]. However, in every case the compounds failed to show separation from placebo in the multicenter clinical trials [96,97,98]. A conclusion drawn from these studies was that one cannot predict the clinical efficacy on the basis of animal models. However, the trial design and logic had a major flaw – whereas in the animal models, the first drug tested was the target compound. However, in the multicenter trials these drugs were tested on long-term schizophrenia patients that had been withdrawn from the drug for only 1-2 weeks, which is the maximum that a therapeutic agent can be ethically withdrawn. The problem is, while this may be sufficient to wash out the compound from the system, it does not return the system to normal. Long-term D2 blockade will result in D2 supersensitivity; therefore once the drug is withdrawn, a normally active DA system would still have a pathologically augmented postsynaptic response. Therefore, once a D2 antagonist is withdrawn and the supersensitive D2 receptors uncovered, the only drug that can act is another D2 antagonist [99].

We had shown that this is the case using another novel compound, a GABA A alpha 5 positive allosteric modulator (PAM). While GABA A synapses are present throughout the brain, the alpha 5 subunit is expressed primarily in the amygdala and the hippocampus; regions in which there is substantial parvalbumin neuron loss [66,67,100,101]. Administration of a novel GABA A alpha 5 PAM had no effect in normal rats; however, in MAM rats it rapidly normalized DA neuron activity, reversed amphetamine-induced hyperlocomotion, and restored hippocampal activity to baseline [102]. However, after administering the D2 antagonist antipsychotic drug haloperidol for only 3 weeks, which was sufficient to induce supersensitivity, this drug was now ineffective at restoring amphetamine responses [77].

These data demonstrate that, in order to adequately evaluate a novel compound, one must test it on a system that is not already perturbed by prior drug administration. So how would one run a clinical trial under such conditions? There are 3 possibilities that may be evaluated: 1) one can test the drug on drug-naïve first episode patients. While this may be problematic with a novel, untested compound, Lilly’s analysis of clinical data on pomaglumetad revealed that the drug was most effective on patients in the early stages of the disorder [94]; a time when supersensitivity may not have fully developed. 2) another possibility is that, although we cannot ethically withdraw patients from an antipsychotic drug for more than 2 weeks, in actuality the patients withdraw themselves all the time, given that there is a nearly 70% noncompliance with their medication [103,104,105,106]. Therefore, targeting patient populations that have demonstrated long-term noncompliance may be an effective strategy. 3) finally, one could use patients that were on medications that do not induce supersensitivity. Data suggest that this may be the partial agonists, given that these drugs stimulate D2 receptors at least partially [81] and also do not induce DA neuron depolarization block [83].

On this basis, to identify effective novel compounds, it would be necessary to alter the manner in which clinical trials are performed. While the prior model may have been effective in identifying DA antagonist medications, they do not appear to be effective in evaluating novel and potentially more effective treatments.

Another approach would be to limit the overactivity of hippocampal pyramidal neurons. One compound that can achieve this effect is evenamide, which will normalize excess glutamate release without affecting baseline levels, which would be ideal for normalizing overactive hippocampal neuron [107]. This drug was found to be effective in reversing selective symptoms produced by amphetamine, PCP, MK-801, or ketamine [108], all of which act via hippocampal hyperactivity [63,109,110]. When tested as an adjunct to standard-of-care to patients that were showing worsening on their current medications, evenamide produced a significant improvement in PANSS [107]. Whether this will also be effective as a monotherapy could yield important insights into how normalizing glutamate overdrive may reverse schizophrenia symptomatology.

4. Conclusion

Based on preclinical data, to identify effective novel compounds, it would be necessary to alter the manner in which clinical trials are performed. While the prior model may have been effective in identifying DA antagonist medications, they do not appear to be effective in evaluating novel and potentially more effective treatments. Current trial designs may be effective at identifying additional D2 antagonist drugs, but not compounds that have a unique site of action. By targeting the site of functional deficits related to hippocampal overdrive, one may be in a position to provide more effective treatments that are better tolerated and can address the broad range of schizophrenia symptomatology.