Generation of circadian rhythms and sleep

Sleep and circadian rhythms are not synonymous. Circadian rhythms are endogenous 24-h oscillations in physiology and behaviour that enable an organism to anticipate and adapt to the changing temporal demands of the environment. An internal clock acts to coordinate the 24-h rhythms of multiple cellular and organ systems within an individual so that different aspects of physiology and behaviour are appropriately synchronised to each other. These rhythms arise from a sub-cellular transcriptional–translational feedback loop (Fig. 1a), involving a number of core clock genes (Lowrey and Takahashi 2011; Reppert and Weaver 2002). In mammals, light is the primary time cue (zeitgeber), which entrains the internal clock to the external light environment. Light information is relayed from the eyes to the primary circadian pacemaker located in the suprachiasmatic nuclei (SCN) of the hypothalamus (Moore 1973; Moore and Klein 1974), which in turn regulates physiology and behaviour. Additional oscillators are found in tissues throughout the body, regulating local physiology (Dibner et al. 2010).

Fig. 1
figure 1

Schematic illustration of the mechanisms underlying circadian rhythm generation and sleep regulation. a Light detected by the eye is relayed to the suprachiasmatic nuclei (SCN) in the hypothalamus via the retinohypothalamic tract (RHT), which uses the neurotransmitters glutamate and PACAP. Circadian rhythms are generated by a cell autonomous transcriptional–translational feedback loop (TTFL) involving a set of core clock genes. The molecular clock in the SCN synchronises circadian clocks found in tissues throughout the body, which regulate local physiology. b Sleep is the product of multiple brain regions and neurotransmitters. Abbreviations for brain regions: BF basal forebrain, DR/MR dorsal/medial raphe nucleus, LC locus coeruleus, LDT laterodorsal tegmental nuclei, LH lateral hypothalamus, LPT lateral pontine tegmentum, PPT pedunculopontine tegmental nuclei, SCN suprachiasmatic nuclei, SLD sublaterodorsal nucleus, TMN tuberomammilary nucleus, VLPO ventrolateral preoptic nuclei, vPAG ventral periaqueductal grey, vlPAG ventrolateral periaqueductal grey. Abbreviations for neurotransmitters: 5-HT serotonin, ACh acetylcholine, DA dopamine, GABA γ-aminobutyric acid, Gal galanin, Glut glutamate, His histamine, NA noradrenaline, ORX orexin

Whilst sleep/wake behaviour is perhaps the most obvious output of the circadian system, sleep biology involves much more than the circadian system. In addition to a circadian drive for wakefulness, sleep is under homeostatic regulation, whereby an increased duration of wakefulness leads to an increased need for sleep. The homeostatic drive for sleep is the product of a complex network of brain regions and neurotransmitter pathways (Fig. 1b), none of which are exclusive to the generation of sleep (Tobler 1995). This complexity makes sleep very vulnerable to disruption. Small changes in brain function can have a big impact on sleep, and disrupted sleep leads to multiple health problems (Table 1).

Table 1 Impact of sleep and circadian rhythm disruption on emotional, cognitive and somatic responses

Sleep and circadian rhythm disruption (SCRD) in schizophrenia

The relationship between schizophrenia and abnormal sleep was first described in the late nineteenth century by the German psychiatrist Emil Kraepelin (Manoach and Stickgold 2009). Today, sleep and circadian rhythm disruption (SCRD) is reported in 30–80 % of patients with schizophrenia, and is increasingly recognised as one of the most common features of the disorder (Cohrs 2008). Sleep disturbances in schizophrenia include increases in sleep latency, and reductions in total sleep time, sleep efficiency, REM sleep latency, REM sleep density and slow-wave sleep duration (Cohrs 2008; Manoach and Stickgold 2009).

Schizophrenia is also associated with significant circadian disruption, including the abnormal phasing, instability and fragmentation of rest-activity rhythms (Martin et al. 2001, 2005; Wulff et al. 2006, 2009). Crucially, patients with SCRD score badly on many quality-of-life clinical subscales, highlighting the human cost of SCRD in schizophrenia (Cohrs 2008; Goldman et al. 1996; Hofstetter et al. 2005). To reinforce this, schizophrenia patients often comment that an improvement in sleep is one of their highest priorities during treatment (Auslander and Jeste 2002).

Common brain mechanisms: an explanation for the co-morbidity of SCRD and schizophrenia

Sleep and circadian rhythm disruption in schizophrenia could be viewed as a side effect of antipsychotic medication, given that antipsychotic drugs, particularly typical antipsychotics, have a severe sedative effect when taken at high doses (Miller 2004). This seems unlikely, however, as SCRD affects both medication-naïve (Wulff et al. 2010) and medicated patients (Krystal et al. 2008). Indeed, the emergence of sleep disruption often precedes the diagnosis of schizophrenia, occurring before any drugs have been prescribed (Wulff et al. 2010). Recent data would suggest that, if anything, antipsychotic drug treatment can actually improve sleep quality in schizophrenia (Cohrs 2008; Krystal et al. 2008). More specifically, schizophrenia patients treated with typical antipsychotics show an increase in their sleep efficiency and total sleep time (Cohrs 2008).

Social isolation, and the resulting absence of social constraints, is also routinely suggested as a cause of SCRD in schizophrenia. This hypothesis was recently addressed by comparing the sleep patterns of schizophrenia patients with those of unemployed healthy volunteers (Wulff et al. 2011). Severe SCRD was observed in schizophrenia, but could not be attributed to an absence of routine, as major sleep disruption was observed in schizophrenia patients that followed a fixed routine. Conversely, undisturbed sleep was seen in a number of control participants that did not follow a fixed routine.

A more likely explanation for the co-morbidity of SCRD and schizophrenia is the involvement of common brain mechanisms. As described above, sleep is the product of a complex interaction between multiple brain regions and neurotransmitters. As a consequence, abnormalities in any key neurotransmitter system will impinge upon sleep at multiple levels. Similarly, schizophrenia is a disorder of distributed brain circuits, affecting a range of neurotransmitter systems (Weinberger and Harrison 2011), many of which overlap with those involved in sleep regulation (Wulff et al. 2010). Viewed in this context, it is no surprise that SCRD is common in schizophrenia, or that SCRD will in turn have widespread effects, ranging across many aspects of neural, neuroendocrine and cognitive function (Fig. 2).

Fig. 2
figure 2

The key components in the regulation and maintenance of the sleep/wake cycle and its relationship to mood/cognition and mental health. The sleep/wake cycle is regulated directly by (i) The 24-h body clock/circadian clock located within the suprachiasmatic nuclei (SCN), (ii) wake-dependent homeostatic drivers (e.g. adenosine) that build-up and generate “sleep pressure”, (iii) social behaviours that force a sleep/wake pattern on the individual. The SCN drives wakefulness throughout the day and sleep during the night. This 24-h rhythm interacts with the homeostatic “hourglass oscillator” producing increased sleep pressure during wake and its dissipation during sleep. The circadian system directly regulates multiple neurotransmitter and brain systems that either drive or modulate sleep, including the hypothalamo–pituitary–adrenal axis (HPA) and melatonin from the pineal gland. The SCN also coordinates the 24-h biology of peripheral oscillators, maintaining internal synchrony of 24-h biological timing processes. A mismatch between central and peripheral oscillators (internal desynchrony), as occurs in shift work or jet-lag, is associated with mood imbalance and depression. All of the components within the triangle are modulated by light, which acts to: entrain the circadian pacemaker to the light/dark cycle; alter melatonin production from the pineal gland; modulate the HPA axis; and elevate or suppress levels of mood and cognition. Social behaviours will also change an individual’s exposure to the light/dark cycle (Mistlberger and Skene 2004), and have a major effect on mood/cognition and mental health. The quality of sleep and the stability of the sleep/wake cycle have a direct effect upon mood, cognitive processing and mental health (Table 1)

Sleep and circadian rhythm disruption in schizophrenia could also arise from dysfunction at any point in the circadian axis (i.e. input, oscillator or output). Such defects would impact upon sleep via the circadian drive for wakefulness, as appears to be the case in familial advanced sleep phase syndrome (FASPS), which has previously been linked to mutations in the clock gene Per2 (Toh et al. 2001).

Clinical evidence for common brain mechanisms in SCRD and schizophrenia

Sleep and circadian rhythm disruption is rarely targeted for treatment in schizophrenia, but when it is, patients report improvements in both their sleep quality and psychiatric symptoms (Kantrowitz et al. 2010). In a recent study, insomnia was treated in 15 patients with persistent persecutory delusions and schizophrenia (Myers et al. 2011). Following a cognitive behavioural therapy (CBT) intervention, there were significant reductions in both insomnia and persecutory delusions. At least two-thirds of participants showed a substantial (>25 %) improvement in insomnia, whilst approximately half showed a substantial (>25 %) reduction in persecutory delusions. There were also reductions in levels of hallucinations, anxiety and depression. Although consistent with the existence of common mechanisms in SCRD and schizophrenia, the results of this study should be interpreted with caution, due to a number of methodological limitations; the sample size was small, there was no control group, and 14 of the 15 patients received antipsychotic medication during the CBT intervention.

Significantly, many of the pathologies caused by SCRD are routinely reported as co-morbid with schizophrenia, but are rarely linked to the disruption of sleep. For example, sleep deprivation (Alhola and Polo-Kantola 2007; Chee and Chuah 2008; Horne 1993; Van Dongen et al. 2003) and circadian de-synchronisation (Kyriacou and Hastings 2010) are known to impair cognition in healthy individuals, and cognitive impairment is a core symptom of schizophrenia. Thus, cognitive impairments in schizophrenia could be exacerbated by circadian de-synchronisation and/or disturbed sleep. Consistent with this, associations between sleep and cognitive performance have been reported in medication-naïve (Forest et al. 2007), medicated (Bromundt et al. 2011; Goder et al. 2004, 2008; Wulff and Joyce 2011), and unmedicated schizophrenia patients (Yang and Winkelman 2006). In the latter study, the severity of patients’ cognitive symptoms was inversely related with slow-wave sleep duration and REM sleep density (Yang and Winkelman 2006).

Memory consolidation is just one cognitive function that warrants further investigation. Memory impairment is prevalent in schizophrenia (Aleman et al. 1999), and it has been suggested that sleep makes a crucial contribution to memory consolidation (Stickgold 2005). Reduced overnight consolidation of procedural learning has been demonstrated in schizophrenia patients (Manoach et al. 2004), and more recently, this effect was linked to a reduction in slow-wave sleep duration (Manoach et al. 2010). The negative symptoms of schizophrenia might also be sensitive to SCRD, since circadian de-synchronisation increases negative mood, irritability and affective volatility in healthy volunteers (Kyriacou and Hastings 2010; Murray and Harvey 2010). Consistent with this, improvements in sleep quality are frequently correlated with the amelioration of negative symptoms in schizophrenia patients (Hofstetter et al. 2005; Yamashita et al. 2004).

Using animal models to establish mechanistic links between SCRD and schizophrenia

Animal models provide essential tools to understand the mechanisms underlying neuropathological processes, enabling highly controlled genetic, anatomical, physiological and behavioural studies that are not possible in humans. If, as our central hypothesis proposes, common mechanisms are involved in the pathogenesis of SCRD and schizophrenia, then SCRD should be observed in schizophrenia-relevant animal models. Conversely, animal models of SCRD may display schizophrenia-relevant behavioural and neurobiological abnormalities. There is a rapidly accumulating body of evidence which supports the first of these predictions, while the second prediction has yet to be tested.

SCRD in schizophrenia-relevant mouse models: existing evidence

The hypotheses outlined above apply to genetic, pharmacological and environmental models, and are not species-specific. For the sake of brevity, this paper focuses on schizophrenia-relevant transgenic mouse models. Sleep and circadian function has yet to be investigated in any pharmacological or environmental mouse models, but this represents a promising avenue for future research; it would be particularly interesting to see whether sleep is affected in the PCP, MK-801 and ketamine models of schizophrenia. It should also be noted that SCRD has already been demonstrated in two schizophrenia-relevant transgenic Drosophila models (Sawamura et al. 2008; Zheng and Sehgal 2010).

A number of genes have been linked with schizophrenia, and endophenotypes thereof, through a combination of genetic linkage studies, genome-wide association studies and appraisals of biological plausibility. Those most established in the ‘candidate gene’ literature are Nrg1 (Harrison and Law 2006; Mei and Xiong 2008; Stefansson et al. 2002), Akt1 (Arguello and Gogos 2008), Disc1 (Brandon and Sawa 2011; Johnstone et al. 2011), Grm3 (Harrison et al. 2008), Dao (Chumakov et al. 2002; Verrall et al. 2010), Comt (Tunbridge et al. 2006), Dtnbp1 (Williams et al. 2005) and ErbB4 (Mei and Xiong 2008). More recent additions to the literature include Snap-25 (Corradini et al. 2009), Vipr2 (Vacic et al. 2011), Cckar (Koefoed et al. 2009), Gsk3b (Lipina et al. 2012), Pde4d (Fatemi et al. 2008; Numata et al. 2008; Tomppo et al. 2009), Tcf4 (Brzozka et al. 2010), MIR137 (Ripke et al. 2011) and ZNF804A (O’Donovan et al. 2008). To the best of our knowledge, sleep and circadian function has only been evaluated in knockout or mutant models of four of these genes, namely Snap-25 (Oliver et al. 2012), Vipr2 (Hughes and Piggins 2008), Nrg1 (Johnson et al. 2002) and Cckar (Shimazoe et al. 2008). In all four cases, significant SCRD was observed, as outlined below.

When circadian disruption is observed in a mouse model, it is important to identify the level of the circadian system at which this disturbance arises (Fig. 3). Does the deficit affect inputs to the SCN, the function of the SCN itself, or the physiological outputs of the SCN? Such knowledge is vital, as it will inform the selection and/or development of the most appropriate therapeutic intervention. To date, this approach has only been implemented in the Snap-25 (Oliver et al. 2012) and Vipr2 models (Hughes and Piggins 2008).

Fig. 3
figure 3

Suggested model of the mechanistic links between cognitive and circadian disturbances. We propose the evaluation of both schizophrenia-relevant models and models of circadian disruption, to confirm whether defects at one level are associated with defects at the other. Similarly, therapeutic interventions that correct defects at one level should produce concomitant improvements at the other

How can one determine the level at which circadian disturbances arise? Light input to the SCN can be evaluated by measuring circadian and molecular responses to light (e.g. phase shifting, entrainment, negative masking and clock gene induction) (Albrecht and Foster 2002; Jud et al. 2005), whilst core oscillator function can be assessed at a behavioural level via activity rhythms under constant conditions, and at a cellular level using clock-gene reporter assays (e.g. Per2::Luc) (Savelyev et al. 2011). Clock outputs can be determined from hormonal rhythms, and from gene expression and reporter assays in peripheral tissues. For example, hormones such as corticosterone can be monitored with real-time measurements from faeces (Abraham et al. 2006). Finally, sleep/wake behaviour can be assessed using non-invasive methods such as video tracking (Fisher et al. 2012).

Snap-25 (blind-drunk mutant)

Previous work has shown that the blind-drunk (Bdr) mouse, a model of Snap-25 exocytotic disruption, displays schizophrenia-related endophenotypes that are modulated by environmental stress (Jeans et al. 2007; Oliver and Davies 2009). Bdr mutants also show disturbances in circadian organisation that appear to specifically affect outputs of the SCN (Fig. 4). The rest/activity rhythms of Bdr mice are phase advanced and fragmented under a light/dark cycle. Retinal inputs appear normal in Bdr mice, as light-induced phase shifts, masking, pupil constriction and retinal histology are all unaffected. Similarly, clock gene rhythms within the SCN are normally phased both in vitro and in vivo. However, the 24-h rhythms of arginine vasopressin (Avp) within the SCN and plasma corticosterone are both markedly phase-advanced in Bdr mice. These data suggest that the circadian phenotype of the Bdr mouse arises from a disruption of synaptic connectivity within the SCN that alters critical output signals (Oliver et al. 2012).

Fig. 4
figure 4

Summary of findings in the blind-drunk (Bdr) mouse model of Snap-25 exocytotic disruption. The rest/activity rhythms of Bdr mice are phase advanced and fragmented under a light/dark cycle, reminiscent of the disturbed sleep/wake patterns observed in schizophrenia. Retinal inputs appear normal in mutants, and clock gene rhythms within the suprachiasmatic nuclei (SCN) [e.g. Per2 and vasoactive intestinal peptide (Vip)] are normally phased both in vitro and in vivo. However, the 24-h rhythms of arginine vasopressin (Avp) within the SCN and plasma corticosterone are both markedly phase advanced in Bdr mice. We suggest that the Bdr sleep/wake phenotype arises from a disruption of synaptic connectivity within the SCN that alters critical output signals (Oliver et al. 2012). Abbreviations for brain regions: DMH dorsomedial hypothalamus, sPVZ subparaventricular zone

Vipr2

Recent studies have shown that Vipr2 duplications confer a significant risk for schizophrenia (Vacic et al. 2011). Vasoactive intestinal polypeptide (Vip) and its receptor, Vipr2 (VPAC2), play a critical role within the SCN. Real-time imaging of circadian gene expression in SCN slices has shown that the VPAC2 receptor is required for the maintenance of circadian oscillations within SCN neurons, and for the synchronisation of oscillations between these neurons (Maywood et al. 2006). Vipr2 knockout mice demonstrate circadian abnormalities, including a shortened circadian period of approximately 22 h (Hughes and Piggins 2008). The deficit in these mice must reside in the SCN itself or in its outputs, as light input to the SCN appears relatively normal; exposure to light increases phosphoprotein and immediate early gene expression in the Vipr2 −/− SCN, whilst a subset of Vipr2 −/− mice respond appropriately to nocturnal light pulses, displaying robust phase-shift responses.

Nrg1

Neuregulin 1 (NRG1) is a growth factor involved in neurodevelopment and plasticity, which has been associated with both schizophrenia (Harrison and Law 2006; Mei and Xiong 2008; Stefansson et al. 2002) and schizotypal personality disorder (Lin et al. 2005). There is some evidence that its expression is increased in the brains of schizophrenia patients (Harrison and Law 2006; Hashimoto et al. 2004). Mice heterozygous for a disruption in the Nrg1 gene show disrupted rest/activity rhythms (Johnson et al. 2002), whilst wheel-running activity is inhibited by the long-term infusion of NRG1 into the third ventricle of the hamster brain (Snodgrass-Belt et al. 2005). NRG1 is expressed in the SCN and retinal ganglion cells (Bernstein et al. 2006; Sharif et al. 2009), consistent with its proposed involvement in circadian function.

Cckar

The Cholecystokinin A receptor (CCK-AR) is a G-protein coupled receptor that binds the neuropeptide cholecystokinin (CCK) (Noble et al. 1999). Abnormal levels of CCK mRNA have been observed in the brains of schizophrenia patients (Bachus et al. 1997; Zachrisson et al. 1999), while several studies have reported an association between the Cckar gene and schizophrenia (Koefoed et al. 2009; Tachikawa et al. 2000, 2001; Toirac et al. 2007; Wei and Hemmings 1999). Dopamine may mediate the relationship between CCK and schizophrenia, as CCK-ARs modulate CCK-stimulated dopamine release in the mesolimbic system (Marshall et al. 1991). There is also evidence that CCK plays a role in sleep regulation. The intraperitoneal administration of CCK promotes slow-wave sleep and inhibits locomotor activity in rats (Kapas et al. 1988). CCK may exert its effects through orexin, a neurotransmitter known to influence wakefulness, as CCK activates orexin neurons by binding to CCK-ARs (Tsujino et al. 2005). CCK-ARs are also involved in photoentrainment; light-induced phase shifts are significantly attenuated in Cckar knockout mice, as is light-induced clock gene expression in the SCN (Shimazoe et al. 2008).

SCRD in schizophrenia-relevant mouse models: future studies

The preceding section describes four schizophrenia-relevant mouse models with significant sleep and circadian deficits. The challenge for the future is to identify more of these models and to characterise them fully. Characterisation entails screening for both SCRD and schizophrenia-relevant behaviours; the models mentioned above are only described as ‘schizophrenia-relevant’ by virtue of the fact that the genes in question share a significant association with schizophrenia (or in the case of Snap-25, a plausible biological connection). Which of these mice provides the best model of schizophrenia remains to be seen. The Bdr mutant (Jeans et al. 2007; Oliver and Davies 2009) and NRG1type 1-tg mouse (Deakin et al. 2009) share some features in common with the disorder, while the Vipr2 and Cckar knockout models have not been extensively tested from a behavioural perspective. To redress this balance, a variety of tests could be employed across a range of behavioural domains (Chadman et al. 2009), as described in “Challenges associated with modelling schizophrenia in rodents”. In the paragraphs that follow, we consider four more genes with the potential to link schizophrenia with sleep and circadian function.

ErbB4

NRG1 (see “SCRD in schizophrenia-relevant mouse models: existing evidence”) acts through the same family of tyrosine kinase receptors as TGF-a, a putative inhibitory output signal of the SCN (Kramer et al. 2001). TGF-a acts on ERBB1 receptors, whilst NRG1 acts on ERBB4 receptors. ERBB1 receptors have a known role in locomotor activity and sleep; mice with reduced ERBB1 receptor activity show reduced negative masking by light (Kramer et al. 2001). Future studies should address whether ERBB4 receptors play a similar role. Polymorphisms in ErbB4 have previously been associated with schizophrenia (Mei and Xiong 2008).

Gsk3b

GSK3 is a serine/threonine protein kinase, whose deregulation has been implicated in schizophrenia and bipolar disorder (Lipina et al. 2012). GSK3 activity is inhibited by the putative schizophrenia risk genes Disc1 and Akt1 (Bradshaw and Porteous 2012; Lipina et al. 2012). Significantly, the GSK3 inhibitor TDZD-8 ameliorates hyperactivity and prepulse inhibition (PPI) deficits in the Disc1-L100P mutant, a mouse model of schizophrenia (Lipina et al. 2012). In the Disc1-Q31L mutant mouse, a model of depression, TDZD-8 corrects a PPI deficit, reduces immobility in the forced swim test, and increases social interaction (Lipina et al. 2012). Interestingly, clinicians often prescribe the mood stabiliser lithium (a GSK3 inhibitor) to schizophrenia patients to augment their antipsychotic medication, although the efficacy of this strategy is unclear (Leucht et al. 2004). GSK3b has a known role in circadian function; it phosphorylates clock proteins and regulates several components of the transcriptional–translational feedback loop that generates circadian rhythms (Cross et al. 1995; Iitaka et al. 2005; Martinek et al. 2001; Yin et al. 2006). GSK3b also inhibits CREB DNA-binding activity (Grimes and Jope 2001), which is involved in circadian signal transduction (Lee et al. 2010).

Pde4d

PDE4 is a phosphodiesterase that been associated with schizophrenia in a number of studies (Fatemi et al. 2008; Numata et al. 2008; Tomppo et al. 2009). Like GSK3, PDE4 activity is inhibited by the putative schizophrenia risk gene Disc1 (Bradshaw and Porteous 2012; Lipina et al. 2012). Crucially, the PDE4 inhibitor Rolipram acts as a cognitive enhancer, facilitating long-term potentiation, memory performance and latent inhibition in wildtype rodents (Barad et al. 1998; Davis and Gould 2005; Zhang et al. 2004), and attenuating PPI deficits in the Disc1-L100P mutant model of schizophrenia (Lipina et al. 2012). In addition, Rolipram has antidepressant-like properties, reducing immobility in the forced swim test in wildtype rats (Zhang et al. 2006). From a circadian perspective, PDE4 regulates cAMP signalling, which is involved in circadian signal transduction (O’Neill and Reddy 2012). Intriguingly, a polymorphism in Pde4d has been associated with sleepiness in healthy individuals (Gottlieb et al. 2007).

Tcf4

Several large genome-wide association studies have identified the basic helix-loop-helix (bHLH) transcription factor TCF4 as one of the most significant schizophrenia susceptibility genes (Brzozka et al. 2010). The protein encoded by this gene resembles ID (inhibitor of DNA-binding) proteins, which act as negative regulators of the transcriptional–translational feedback loop that generates intracellular circadian rhythms (Duffield et al. 2009). To date, at least eight different Tcf4 mutants have been produced by ENU mutagenesis, but their circadian profiles have yet to be characterised.

Schizophrenia-relevant abnormalities in mouse models of SCRD

To establish mechanistic links between SCRD and schizophrenia, a parallel approach is to screen mouse models of circadian disruption for schizophrenia-relevant behaviours, such as impaired cognitive function. To our knowledge, the Clock mutant, described below, is the only mouse model of SCRD which has been studied in this way (Roybal et al. 2007). The Clock mutant does not show any schizophrenia-relevant behaviours, but has a striking mania-like phenotype, which can be reversed with the mood stabiliser lithium (Roybal et al. 2007). Future research will reveal whether other mouse models of SCRD have a schizophrenia-relevant phenotype.

Clock (Clock mutant)

The Clock mutant was first identified in 1994 from a circadian screen of ENU mutants (Vitaterna et al. 1994). These animals have an extended circadian period (Vitaterna et al. 1994), and sleep significantly less than wildtypes (Naylor et al. 2000), consistent with the decreased need for sleep observed in patients with mania (Plante and Winkelman 2008). Their behavioural profile is also reminiscent of bipolar patients in the manic state; they show hyperactivity, excessive-reward seeking behaviour [as measured by intracranial self-stimulation (ICSS), cocaine preference and sucrose preference], reduced depression like-behaviour (as measured with the Porsolt forced swim test and learned helplessness test), reduced anxiety-like behaviour (as measured with the open field test and elevated plus maze) (Roybal et al. 2007) and increased exploratory behaviour (Easton et al. 2003).

Therapeutic interventions in mouse models with simultaneous deficits

Once models that display simultaneous circadian and schizophrenia-relevant abnormalities have been identified, therapeutic interventions can be introduced. If schizophrenia and SCRD share a common mechanistic origin, then two further predictions logically follow. The administration of drugs used to treat schizophrenia (e.g. antipsychotics and mood-stabilisers) should produce a concurrent improvement in the animals’ SCRD and schizophrenia-relevant deficits, while therapies that target SCRD should do likewise (Fig. 3). The latter approach could involve either pharmacological (e.g. melatonin administration) or environmental interventions (e.g. modification of the light/dark cycle or scheduled voluntary exercise) (Power et al. 2010). This type of experiment has never been performed in a schizophrenia-relevant animal model, although the pharmacological imposition of sleep has been shown to slow cognitive decline in a transgenic mouse model of Huntington’s disease (Pallier et al. 2007).

Challenges associated with modelling schizophrenia in rodents

Animal studies afford a level of control that is simply not possible in human studies. Human participants differ from each other in all manner of dimensions, including their age, education, medication history, genetic makeup and life experiences. In this context, it is difficult to attribute a specific behavioural observation (e.g. disturbed sleep) to a specific underlying trait (e.g. the possession of a particular risk-conferring gene). In animal studies, there are relatively fewer confounds, so making such attributions is more straightforward. Genetic manipulation is also possible in rodents, enabling the investigator to move beyond correlation and infer causality.

Modelling schizophrenia in animals raises a number of complex theoretical and experimental design issues for the investigator (for reviews, see Arguello and Gogos 2010; Harrison et al. 2012; Nestler and Hyman 2010; Papaleo et al. 2012). The first and perhaps biggest challenge is the process of selecting which gene to manipulate. Despite years of research, no gene has been unequivocally established as conferring increased susceptibility to schizophrenia (Crow 2008). The literature is littered with genes that have yielded positive results in one or two genetic association studies, only for several follow-up studies to draw a blank (Crow 2008). A ‘schizophrenia model’ is only as good as the evidence linking the gene in question with the disorder itself, and from this perspective, the validity of almost any model can be questioned. The likely explanation for these results is that schizophrenia is caused by a large number of genes, each one having a very small effect (Chakravarti 1999). Put simply, no single gene is either necessary or sufficient to cause the disorder. In light of this, the value of single gene models is debatable. Double- or triple-gene models may be more realistic, but these are more complex and costly to produce. A further limitation is that some schizophrenia risk genes (e.g. G72/DAOA) have no known ortholog in rodents (Chumakov et al. 2002).

Besides choosing which gene to manipulate, the investigator must decide how to manipulate it. Knockout models are the simplest and therefore the most common, but again, are unlikely to represent schizophrenia that faithfully. There is no evidence for null mutations in schizophrenia, but there is evidence for the up- or down-regulation of specific genes (Harrison and Weinberger 2005). In this context, transgenic over-expression or heterozygous knockout may prove more suitable tools for modelling schizophrenia. The investigator must also decide whether the genetic manipulation is constitutive or conditional. Most models are constitutive; that is, the manipulation is present throughout the brain and throughout the animal’s lifetime. Given evidence that brain abnormalities in schizophrenia are more pronounced in some brain areas than others, and that they can develop over time (Ellison-Wright et al. 2008), the constitutive approach may not be the most appropriate. Conditional models, in which the timing and location of the manipulation are more tightly regulated, provide a more flexible alternative.

Another complication is the not inconsiderable variation amongst patients labelled with schizophrenia. The disorder is associated with a wide range of symptoms, and two patients with the same diagnosis may present a very different subset of these symptoms (Andreasen 1999). Thus, treating a condition as complex as schizophrenia as a unitary disorder could be a mistake. Perhaps a more appropriate approach would be to model specific symptoms (or endophenotypes) in isolation (Kaffman and Krystal 2012).

In rodents, some schizophrenia-relevant behaviours are easier to model than others. Schizophrenia is characterised by positive symptoms (e.g. hallucinations and delusions), negative symptoms (e.g. avolition and anhedonia) and cognitive symptoms (e.g. impaired working memory, sensorimotor gating and attentional set-shifting) (Andreasen 1995; Elvevag and Goldberg 2000). In humans, hallucinations and delusions are only revealed through patients’ verbal reports, so these symptoms are difficult to investigate in rodents. Nonetheless, hyperlocomotion and stereotypic behaviours are considered by many to reflect the positive symptoms of schizophrenia (Nilsson et al. 1997; Sams-Dodd 1996). Similarly, reduced social interaction in rodents is often presented as a direct analogue of the negative symptoms witnessed in schizophrenia (Lee et al. 2005; Sams-Dodd 1996). Whether these behaviours are conceptually and neurally equivalent seems highly unlikely, however (Bussey et al. 2012; Garner et al. 2006). In contrast, the human and rodent versions of most cognitive tasks are at least superficially similar. PPI, for example, is a measure of sensorimotor gating which can be tested almost identically in humans and rodents (Swerdlow et al. 2008), while the attentional set-shifting task, which measures cognitive flexibility, is modelled directly on its human counterpart, the Wisconsin Card Sorting Test (Bissonette and Powell 2012; Garner et al. 2006). Questions remain as to the suitability of set-shifting tasks for mice, however (Garner et al. 2006). Working memory is also routinely tested in rodents, although again, it is unclear whether the human and rodent versions of these tasks tap the same neural substrates (Sanderson and Bannerman 2010).

Of course, behavioural assessment is not the only way to assess the validity of schizophrenia-relevant models. Investigators can also look for neurophysiological changes reminiscent of those observed in schizophrenia. These include structural changes, such as reduction of cortical thickness and enlargement of the lateral ventricles (Weinberger et al. 1982), and characteristic changes in neurotransmitter systems, such as glutamatergic hypofunction (Konradi and Heckers 2003).

To summarise, it would be a significant overstatement to suggest that schizophrenia can be faithfully modelled in rodents. First, some of the core symptoms of schizophrenia simply cannot be measured in rats or mice, as explained above. Secondly, schizophrenia is a multi-gene disorder, so it seems improbable that a single-gene model could ever demonstrate the full range of symptoms observed in a patient. Finally, there are notable neuroanatomical differences between humans and rodents; the prefrontal cortex, for example, is far more developed in humans than it is in rodents. Despite these limitations, murine models may still yield valuable insights into the roles of schizophrenia risk genes in brain function, such as whether they contribute to sleep and circadian function.

Challenges associated with sleep and circadian phenotyping in rodents

Sleep and circadian rhythms in humans and rodents are not identical, not least as rodents are nocturnal and humans diurnal. In addition, humans typically sleep once every 24 h, whereas rodents tend to alternate between several bouts of sleep and wakefulness (Fisher et al. 2012). As a result, the sleep and circadian phenotype of a rodent model may not translate directly to humans. In addition, environmental light intensity varies continuously in the natural world, whereas most experimental paradigms employ a discrete transition from light to dark and vice versa.

The way we assess the circadian/sleep phenotype of rodents might also cause problems. Crucially, locomotor activity is the dependent variable in most circadian screening paradigms, and locomotor activity is subject to all manner of influences besides circadian function, including basic motor function, anxiety and arousal levels. Therefore, this approach might not be the most appropriate for assessing mouse models with locomotor deficits, such as the schizophrenia-relevant NRG1type 1-tg mouse (Deakin et al. 2009). An obesity phenotype, as seen in Clock mutant mice, could also complicate behavioural phenotyping (Turek et al. 2005). Invasive measures of circadian function (e.g. clock gene rhythms within the SCN) provide a useful alternative, as they are not subject to the same confounds.

Differences in mouse strain must also be taken into consideration. For example, the circadian system of the C57/BL6 mouse is more sensitive to light than that of the C3H mouse. Hence, C57/BL6 mice are able to entrain their circadian behaviour at much lower irradiances (Foster and Helfrich-Forster 2001). There is also evidence that pineal melatonin content varies between mouse strains (Ebihara et al. 1986; Goto et al. 1989). In contrast to wild mice, several inbred strains, including the C57/BL6 mouse, do not have detectable melatonin levels in their pineal glands.

Conclusion and future directions

Sleep and circadian rhythm disruption and schizophrenia are often co-morbid, and this co-morbidity may arise from the involvement of common brain mechanisms. SCRD in schizophrenia is not merely a side effect of antipsychotic medication, nor is it a by-product of an absence of social routine. The treatment of insomnia in schizophrenia patients produces a concomitant improvement in psychiatric symptoms, which provides further support for the hypothesis that common mechanisms are involved. Moreover, many symptoms associated with SCRD, such as impaired cognitive performance, are frequently observed in schizophrenia patients. Animal models provide a means to test our hypothesis of mechanistic overlap. Implicit in our theory are four key predictions that can be tested in rodents:

  1. 1.

    Sleep and circadian rhythm disruption should be observed in schizophrenia-relevant models.

  2. 2.

    Schizophrenia-relevant behavioural abnormalities (e.g. cognitive impairments) may be observed in models of SCRD.

  3. 3.

    Therapies that target SCRD (e.g. melatonin or scheduled voluntary exercise) should ameliorate both SCRD and schizophrenia-relevant behavioural abnormalities in models which display simultaneous deficits.

  4. 4.

    Therapies that target schizophrenia-relevant behavioural abnormalities (e.g. antipsychotic or mood-stabilising drugs) should do likewise.

In this review, we have drawn attention to several schizophrenia-relevant mouse models which show significant sleep and circadian deficits. The challenge for the future is to identify more of these models, characterise them fully, and investigate their responses to therapeutic interventions. Given the continued interest in, and availability of, schizophrenia-relevant mouse models, the evaluation of circadian and sleep physiology and behaviour in these models represents an excellent opportunity to better understand the shared mechanistic basis of SCRD and schizophrenia.