Despite 50+ years of drug discovery, current antipsychotics have limited efficacy against negative and cognitive symptoms of schizophrenia, and are ineffective—with the exception of clozapine—against any symptom domain for patients who are treatment resistant. Novel therapeutics with diverse non-dopamine D2 receptor targets have been explored extensively in clinical trials, yet often fail due to a lack of efficacy despite showing promise in preclinical development. This lack of translation between preclinical and clinical efficacy suggests a systematic failure in current methods that determine efficacy in preclinical rodent models. In this review, we critically evaluate rodent models and behavioural tests used to determine preclinical efficacy, and look to clinical research to provide a roadmap for developing improved translational measures. We highlight the dependence of preclinical models and tests on dopamine-centric theories of dysfunction and how this has contributed towards a self-reinforcing loop away from clinically meaningful predictions of efficacy. We review recent clinical findings of distinct dopamine-mediated dysfunction of corticostriatal circuits in patients with treatment-resistant vs. non-treatment-resistant schizophrenia and suggest criteria for establishing rodent models to reflect such differences, with a focus on objective, translational measures. Finally, we review current schizophrenia drug discovery and propose a framework where preclinical models are validated against objective, clinically informed measures and preclinical tests of efficacy map onto those used clinically.


Schizophrenia is a syndrome commonly associated with symptoms that are classified into positive (psychosis involving hallucinations and/or delusions), negative (affective flattening, anhedonia, avolition), and cognitive (deficits in memory, attention, learning, executive function) domains [1]. Since its emergence in the 1950s and 60s, the dopamine hypothesis has been the leading theory of schizophrenia pathophysiology [2]. While this hypothesis posited that schizophrenia—without specific reference to symptom domains—occurred as a result of excessive neurotransmission at dopamine receptors, it has since evolved to incorporate new lines of evidence [2]. As it currently stands, the dopamine hypothesis proposes that multiple genetic and environmental factors lead to increased presynaptic dopamine function in the striatum, resulting specifically in psychotic symptoms; while reduced dopaminergic drive to cortical areas leads to negative and cognitive symptoms. While the evidence supporting various tenets of this basic model has not always been consistent, dopamine dysregulation has been the target of all drugs used to treat schizophrenia since the discovery of the first antipsychotic, chlorpromazine, nearly 70 years ago [2,3,4].

Drugs currently approved to treat schizophrenia are generically categorised as either typical or atypical antipsychotics, but all share a common mechanism of action in antagonism of the dopamine D2 receptor and numerous issues relating to efficacy. Drugs are effective against positive symptoms in only ~70% of patients (non-treatment resistant, non-TRS; for review see ref. [5]). The remainder recieve no therapeutic benefit from first-line antipsychotics (treatment-resistant schizophrenia, TRS), leaving clozapine as the sole medicinal option—one associated with significant side effects and lifelong monitoring. Moreover, none produce meaningful improvements in negative nor cognitive symptoms—both of which are associated with poor functional outcomes [6,7,8]. It is evident from both the limited efficacy of current antipsychotics and an array of clinical research that symptoms of schizophrenia occur as a result of diverse brain dysfunction; including changes to a number of neurotransmitter circuitry and immune system function [9,10,11,12]. Despite progress towards a better understanding of the underlying pathophysiology, the landscape of schizophrenia therapeutics has not kept pace [6]. The field still awaits a novel drug without appreciable affinity for D2 receptors that receives widespread clinical use—and this is certainly not due to a lack of effort.

Therapeutics with novel, non-D2 mechanisms have been sought for more than 30 years and include targets as diverse as the receptors, transporters and signalling pathways of dopamine, glutamate, glycine, serotonin, acetylcholine, oxytocin, histamine, opioids and neurosteroids [6]. Compounds for these targets have often shown promising efficacy in preclinical development, mostly in rodents, but consistently fail in clinical trials due to a lack of efficacy. A recent review of 250 clinical trials of drugs with non-D2 mechanisms concluded “…we cannot confidently state that any of the mechanistically novel experimental treatments covered in this review are definitely effective for the treatment of schizophrenia and ready for clinical use.” [6].

While the large number of failed clinical trials may be attributed to shortcomings at multiple stages in the drug discovery process, there is a clear and systematic disconnect between preclinical and clinical efficacy—particularly with respect to negative and cognitive symptoms [13, 14]. This may be unsurprising given the disparate measures by which efficacy is assessed in rodents vs. humans. In clinical trials, various clinical rating scales or cognitive batteries are used to quantify efficacy of schizophrenia therapeutics against positive, negative and cognitive symptoms [15, 16]. Clearly, interview-based clinical rating scales cannot be used in rodents, therefore preclinical efficacy is assessed using behavioural tasks that act as correlates of human symptoms, such as rodent locomotor activity and positive symptoms. While the ultimate goal of assessing efficacy in preclinical models is to provide a prediction of clinical efficacy, there are limited instances in which preclinical efficacy does indeed correlate with clinical efficacy—likely attributed to not only species differences but disparate methods of assessment.

In an industry that increasingly strives to test new drugs in a “fail fast” approach, the inability to reliably predict clinical outcomes poses a significant and costly challenge [17]. Without a critical evaluation and subsequent reimagination of preclinical discovery, it is unlikely such challenges will be overcome. Here, we discuss the past and present landscape of schizophrenia drug development, particularly where the efficacy of novel non-D2 candidates has failed to translate in late-stage clinical trials, then highlight key issues in preclinical development limiting accurate predictions of clinical efficacy. Finally, we consider a revised framework whereby neuroimaging could be used as a translational tool to improve preclinical predictions of clinical outcomes.

The past and the present of schizophrenia drug development

It is clear that current antipsychotics have limited utility in treating the broad spectrum of symptoms in all schizophrenia patients, thus novel non-D2 therapeutics have been sought with the idea that they will treat all symptom domains and be effective in TRS patients. Disappointingly, there has been an overwhelming failure to convert promising preclinical data, and in some cases encouraging early clinical findings, into successes in phase 3 and new drug approvals for the treatment of symptom domains of schizophrenia. Inhibitors of PDE10 (e.g. PF-02545920 and TAK-063), that act downstream of dopamine signalling, have failed to show a significant improvement in PANSS total score in Phase II clinical trials despite showing D2 antagonist-like activity in preclinical models [18, 19]. Similarly, drugs that act to enhance glutamate signalling at either NMDA (e.g. glycine transporter type-1 inhibitor bitopertin and D-amino acid oxidase inhibitor luvadaxistat) or metabotropic glutamate (e.g. mGlu2/3 agonists such as pomaglumetad) receptors have failed to meet primary endpoints in mid-to-late-stage clinical development for negative and cognitive symptoms, despite promising preclinical data [20,21,22,23]. Alpha-7 nicotinic acetylcholine receptor agonists (e.g. encenicline), which have been shown to modulate dopamine, glutamate, and acetylcholine, displayed promising efficacy against cognitive symptoms in a Phase II clinical trial but again failed due to efficacy in two Phase III trials [24, 25].

The aetiology of these failures is unclear; collectively therapeutics targeting solely the glutamatergic system alone appear to have fared poorly in later-stage clinical trials. However, both glycine transporter type-1 inhibitors (e.g., BI-425809) and AMPAkines (e.g., PF-04958242, now BIIB104) remain in active mid-late-stage clinical development for different symptom domains in schizophrenia.

AMPAkines are positive allosteric modulators of AMPA receptors and act to enhance long-term potentiation, a critical process involved in learning and memory formation [26]. The first AMPAkine to be clinically tested for schizophrenia, CX516, was assessed as a possible add-on therapy to standard of care antipsychotic drugs [27]. However, CX516 worsened PANSS scores compared with the placebo group and did not improve patient cognitive outcomes. By virtue of their mechanism of action, AMPAkine clinical trials have largely investigated improvements to cognition—it is unclear whether efficacy may extend to positive or negative symptoms. AMPAkines improve cognition in preclinical cognitive deficit models, thought to be through the enhancement of long-term potentiation and long-term depression [26]. BIIB104 also improved ketamine-induced cognitive deficits in healthy subjects in the Hopkins verbal learning test, however this may be confounded through the activity of ketamine in potentiating AMPA receptor activity [28, 29]. It remains to be seen whether these results will predict efficacy in late-stage clinical development.