Keywords

Schizophrenia is a disabling psychiatric disorder whose etiopathogenesis is still unclear. Brain research has been mostly focused on neurological diseases but in recent decades, research on the neurobiological basis of mental illnesses has emerged. This lack of information has restrained research from advancing in the understanding of the neurobiological basis of schizophrenia, and has led clinicians to make the diagnosis based on some symptom clusters. Schizophrenia is characterized by positive, negative, and cognitive symptoms. Positive symptoms represent abnormal mental functions, such as hallucinations and delusions. Negative symptoms include social isolation, decreased motivation, and flattened affect. Cognitive symptoms are related to poor executive function, particularly that involving attention and memory. The typical onset of these symptoms starts between late adolescence and early adulthood [1], although neurodevelopmental processes play an important role in schizophrenia. Currently, antipsychotic drugs are effective in reducing positive symptoms but have minimal beneficial effects on cognition. Poor cognitive functions affect the everyday life of schizophrenic patients and contribute most to chronic disability and unemployment [2,3,4]. However, no current treatment or therapy can successfully manage cognitive deficits.

Within the past two decades, numerous efforts in understanding the underlying etiology of cognitive dysfunction of schizophrenia have been made. For that purpose, understanding the neurobiology and circuitry of the forebrain, which support cognitive processes, is of major interest. Several theories regarding the etiology of schizophrenia have been proposed, including, but not limited to, genetic predisposition [5,6,7,8,9], prenatal infection [10, 11], environmental influences [12, 13], or a combination of these [14]. Brain imaging and post-mortem studies have shown anatomical changes in schizophrenic patients, primarily in prefrontal cortex and temporal lobe structures, such as decreased cortical volume [15, 16], altered circuitry and connectivity [17], and changes at the neuronal level [18] that contribute to core cognitive dysfunctions in schizophrenia.

Cognitive dysfunction is pervasive and is independent of other symptoms [3]. The most characteristic finding is the decreased ability in working memory tasks, especially when a high degree of information needs to be processed [19, 20]. The dorsolateral prefrontal cortex (DLPFC) is implicated in working-memory deficits, and brain imaging studies have consistently demonstrated alterations in activation of DLPFC in schizophrenic individuals, particularly during cognitive tasks [21,22,23]. In addition, hippocampal circuitry is altered both regionally and extra-regionally [24]. Within these circuits, GABAergic connections are of particular interest. Local interconnections of GABAergic interneurons onto pyramidal cells show alterations at the synaptic level, resulting in cognitive deficits [25, 26]. Furthermore, afferent and efferent connections between hippocampus and DLPFC also seem to be affected [27, 28]. The aberrant plasticity of the hippocampal–prefrontal cortex pathway may explain the deficits of cognitive processes that require spatial and temporal information [28, 29].

Postmortem studies further support the idea that GABAergic alterations in the prefrontal cortex and hippocampus are implicated in the etiology of schizophrenia. In particular, immunostaining studies reveal a selective decrease of the calcium-binding protein parvalbumin in brains of schizophrenic patients compared with controls, with no concurrent loss of calretinin- or calbindin-immunoreactive interneurons [30–32]. The reason why the decrease of GABAergic interneurons is mostly limited to parvalbumin-containing cells is unclear, but it has been suggested that developmental changes in parvalbumin (PV) expression could make them vulnerable to dysfunction. Recent new studies of postmortem brains indicate that GABAergic interneurons containing the neuropeptide somatostatin are also decreased in orbital regions of the forebrain in schizophrenic patients [33].

Maturation of GABAergic Interneurons

GABAergic interneurons are the main source of cortical inhibition in the mammalian brain [34] and account for 10–25% of total number of neurons, depending on the brain region [35]. The high diversity of GABAergic interneurons in their morphology and functional properties has made the classification challenging. Suffice it to say that GABAergic interneurons play an important role in regulating and orchestrating the activity of pyramidal cells. They also shape cortical plasticity, synaptic wiring, and oscillations during prenatal and postnatal development [35]. An important characteristic of GABAergic circuit development is its long duration. The maturation begins early in embryonic stages and proceeds in several steps before fully-developed cell features are acquired. This involves the maturation of GABA release and reuptake, the ability of neurons to form synapses at defined developmental stages, and the expression of particular proteins that regulate cell signaling to eventually acquire mature electrophysiological properties.

Interneurons containing the calcium-binding protein PV seem to be the most affected GABAergic interneurons in schizophrenia. In particular, the selective downregulation of PV and GAD67 (glutamic acid descarboxilase 67) in the prefrontal cortex is the most consistent finding [36,37,38]. PV+ interneurons can be classified into two morphologically differentiable groups: basket cells and chandelier cells. Basket cells usually synapse in pyramidal cell somas and proximal dendrites, and chandelier cells form axo–axonic connections with the axon initial segment of pyramidal cells. One single basket cell can target large population of pyramidal cells, thereby exerting powerful postsynaptic modulation of excitatory output. In turn, pyramidal cells provide feedback input to parvalbumin positive cells, and this closed loop seems to evoke gamma oscillations, the physiological correlate for proper sensory integration and cognitive functions [39, 40]. Although another mechanism for gamma oscillations generation has also been proposed, the involvement of basket cells in gamma oscillation generation is known to be critical [41]. Schizophrenic patients show reduced power in gamma oscillation in frontal lobe during working memory tasks, in auditory cortex after a train of clicks, and in visual cortex when the scenery needs a perceptual organization [42]. Rhythmic brain activity in gamma-band seems to be necessary to transfer information between brain regions, and a lack of proper communication between brain regions is believed to underlie the pathophysiology of schizophrenia [42]. Given that PV and GAD67 expression is regulated by cortical activity [43, 44], the absence of both in PV+ neurons of schizophrenic patients indicates a lack or decreased activity of PV+ neurons. This would shift brain activity balance towards excitation. The dysregulation of local circuitry could be explained by abnormal neurodevelopmental changes.

Early in neurodevelopment, GABA acts as a depolarizing neurotransmitter, due to chloride accumulation in the cytoplasm. Around the end of the first postnatal week, the expression of potassium-chloride cotransporter KCC2 dramatically increases in the forebrain, and GABA becomes hyperpolarizing [45]. The expression of KCC2 is regulated by local neuronal activity, so it subserves as a stop signal for further migration [46]. Between the second and third postnatal weeks, several changes in GABAergic interneurons take place: (1) maturation of GABAergic immunoreactivity [47], (2) appearance of adult-like electrophysiological properties (increase in firing frequency, high-frequency subthreshold membrane potential oscillations, and changes in membrane resistance) [35, 48], and (3) maturation of axonal plexuses of cortical interneurons [48]. Parvalbumin-expressing cells fully develop neurites at approximately 4–5 weeks [48]. In the first month, there is an increase in GABAergic synapses. The composition of GABAA receptors also changes during development in cortical neurons, and adult form subunits are found at 3–4 weeks [49]. Subunit switch is paralleled with the maturation of GABAergic inhibitory postsynaptic potentials (IPSPs) in the neocortex and hippocampus of parvalbumin-expressing fast-spiking cells [48]. GABAB receptors, a G protein-coupled receptor, also express different subunits during postnatal development [50] (Fig. 25.1). Developmentally regulated subunit expression has important functional implications in physiological properties.

Fig. 25.1
figure 1

Major steps of GABAergic interneuron maturation and neurotransmission

Full size image

Inhibition carried out by GABAergic interneurons is low during neurodevelopment, and acquires adult features in late adolescence or early adulthood [35]. The increase in inhibitory tone is correlated with the development of perineural nets (PNN) [51]. Perineural nets are proteoglycans that wrap certain type of neurons, and are thought to give homeostatic balance to highly active neurons. They are developed in an activity-dependent manner and act as physiological buffers for ions. PNN are particularly present in fast-spiking (FS) neurons for their high activity patterns. The opening and closure of critical periods are regulated by the level of maturation of inhibitory neurons, which in part is determined by the presence of PNN, as they offer synaptic stability [51].

There is evidence of perturbed maturation of GABAergic interneurons in schizophrenia. Impaired interneuron migration during development has been demonstrated through several findings. Various authors have found increased density of interneurons in the superficial white matter of schizophrenic patients [52, 53]. Following knockdown of DISC-1, a gene implicated in schizophrenia susceptibility, tangential migration of MGE-derived interneurons is altered [54]. In addition, neuroregulin-1 (NRG-1) and its receptor ErB4 are also associated with increased risk of schizophrenia, and ErB4 is exclusively expressed in inhibitory interneurons, particularly in PV+ cells [55]. Altered connectivity in local excitation/inhibition circuitry has also been shown: less expression of the α1 subunit at basket cell–pyramidal cell synapses, and overexpression of the α2 subunit at chandelier cell–pyramidal cell synapses [56]. There are decreased axo–axonic synapses and less glutamatergic synaptic input onto parvalbumin-expressing basket cells in mice models [57]. These findings support the hypothesis that parvalbumin-expressing cells get improperly connected during development, resulting in aberrant network activity and plasticity in adulthood. Moreover, the substantial changes that GABAergic system undergoes at late adolescence and the onset of schizophrenic symptoms have the same age-dependency profile.

Animal Models of Schizophrenia

Animal models have been useful for unraveling the pathophysiological mechanism and treatment development in many areas of medicine. The critical obstacle in using animal models for studying psychiatric disorders is rooted in the poor understanding of their neural basis. Moreover, schizophrenia is considered a uniquely human disorder, as it mostly affects perception, thinking, language, and attention. Modeling those features in lower mammals has been controversial, but the high prevalence of schizophrenia (approximately 1% of general population), and the debilitating effects of the disease justify a great effort to study it. In this way, animal models are an indispensable tool.

Several approaches have been taken to model schizophrenia in rodents. Animal models can represent diseases from three different perspectives: (1) reproducing etiopathogenetic factors, (2) simulating signs and symptoms, or (3) predictability of response to treatment. We refer to these approaches as construct validity, face validity, and predictive validity respectively. The faithfulness of each type of validity varies considerably, and the utility of the proposed model depends greatly on the goals of each study. Oftentimes, models with high predictive validity are used for the development of new pharmacological treatments, but animal models with high construct validity provide a better framework for studying pathological processes and outcomes. Given the complexity of the nervous system and the lack of valid pathognomonic biological markers, phenotypes, or genotypes of schizophrenia, a heuristic model that would encompass different aspects of schizophrenia would be desirable [58]. These aspects should include anatomical, neurochemical, behavioral, and cognitive features of schizophrenia, but it is rare that a single model addresses multiple phenomena [58]. The best replicated neurobiological findings are thinning of prefrontal and temporal region cortices [59], and decreased expression of calcium-binding protein parvalbumin and GAD67 enzyme in cortical interneurons [60,61,62,63,64].

Animal models of schizophrenia have been categorized to date in three main groups: neurodevelopmental models, genetic models, and pharmacological models [58, 59, 65,66,67,68,69]. Neurodevelopmental models include obstetrical complications such as gestational malnutrition or prenatal exposure to influenza virus. Early stressors such as maternal separation and social isolation [70,71,72] have also been used, but usually combined with genetic or pharmacological approaches, also named “two-hit models” [73]. Neonatal brain lesions in ventral hippocampus have been widely performed, and present face validity in terms of damaged brain structures, although the disturbance is far more severe than in schizophrenia [74,75,76,77]. In any case, the causative role of any of these approaches is dubious, and thus the construct validity. Schizophrenia is highly heritable, and a genetic component of the disease is beyond discussion. Genes interact with environmental factors, and depending on that interaction the disorder may or may not emerge. Animal genetic models have been perfectly reproduced for some diseases, which give strong construct and face validity. However, given the large number of genes involved in schizophrenia, and their complex interplay [78] with stochastic and environmental factors, it is unlikely that a faithful model can be built based entirely on this approach. Genes that have been involved with increased risk of schizophrenia are dysbindin, neuroregulin-1 (NRG-1), and disrupted-in-schizophrenia 1 (DISC-1), among others [35, 68]. With regard to pharmacological approaches, early studies showed that D2 dopamine receptor antagonists reduced positive symptoms of schizophrenia. Therefore, a dysfunction of dopaminergic neurotransmission has been the most enduring theory as the underlying cause of schizophrenia. Despite the longevity of the dopaminergic hypothesis and its face validity in schizophrenia research, it is now believed that dopaminergic dysfunction is a consequence rather than the cause [79]. Animal models of non-competitive NMDA antagonists are currently the most characterized pharmacological approach. Phencyclidine (PCP) and ketamine have been shown to induce psychosis in healthy humans and to exacerbate positive symptoms of schizophrenic patients [80]. This suggested the involvement of NMDA receptors in the pathophysiology of schizophrenia. The effect of non-competitive NMDA antagonists is not completely understood, but seems to have complex interactions in glutamatergic, dopaminergic, and GABAergic neurotransmission [81]. Altered glutamate transmission and NMDA receptors have been related to negative and cognitive symptoms observed in schizophrenia. NMDA antagonists, unlike dopamine, have strong construct validity for studying cognitive and attention deficits of schizophrenia [65, 67, 71]. Some authors have suggested that NMDA antagonists fail to take into account neurodevelopmental processes, mainly because acute doses of NMDA antagonist have been used in the literature, and short-term consequences measured rather than long-term effects. Nevertheless, in recent years repeated subchronic/chronic administration of NMDA antagonists has been used during the early postnatal period to model cognitive deficits of schizophrenia, and long-term behavioral aspects evaluated (during adolescence and adulthood) [69, 82]. MK-801, also known as dizocilpine, is the most potent and selective drug among non-competitive NMDA antagonists, and therefore widely used to model schizophrenia in rodents [83].

NMDA Hypofunction and Brain Maturation

Glutamate activates intracellular cascades via ionotropic and metabotropic receptors. Glutamate is integral in neurodevelopment. It regulates synaptogenesis, network plasticity, dendritic arborization, neuronal progenitor propagation, and migration [84]. NMDA receptors are the only glutamatergic excitatory receptors postnatally, as functional AMPA receptors are absent at the beginning of the postnatal period [85]. Glutamate can act via ionotropic and metabotropic receptors. Glutamate ionotropic receptors, N-methyl-D-aspartate receptor (NMDAR), are the target in schizophrenia research. Hypofunction of NMDAR plays a role not only in psychiatric diseases like schizophrenia, but also in Alzheimer’s disease or autism [86]. The differences in clinical and neuropathological presentations might account for the timing and cause of NMDAR hypofunction.

NMDARs are made of four subunits, forming a heterotetramer composed by NR1 subunit and the facultative NR2 (A, B, C or D) or NR3 (A or B). NR3 subunits are mainly found in early development. NR2 subunits regulate the channel gating. NR2A subunit is the most abundant throughout the nervous system, but NR2B is predominant in forebrain and hippocampus. Depending on the combination of NMDA subunits, the electrophysiological properties of NMDA receptors vary. NR1-NR2B combinations have longer excitatory postsynaptic potentials in vitro than NR1-NR2A complexes [87]. NMDA receptor subunits are also involved in synaptic plasticity: a shift in subunit expression in a particular receptor potentially changes its functional properties. In fact, NR2B incorporation could increase the time period for synaptic coincidence, thereby enhancing synaptic efficacy and probably memory function. NMDAR subunits also differ in their binding sites: NR1 subunits have glycine binding sites, and NR2 subunits glutamate binding sites. Glycine acts as a co-agonist, meaning that its binding to NMDAR is a prerequisite for the activation of NMDA receptor, together with the removal of the magnesium block. D-serine can also function as a co-agonist when it binds to glycine-B sites of NMDAR. At resting membrane potentials, magnesium ions enter the channel pore and prevent further ion permeation. A membrane depolarization is necessary to dislodge and repel magnesium block, to allow ion flow through the channel. In addition to the heterotetramer, NMDA receptors have postsynaptic densities (PSD), a set of proteins that give structural and functional stability to glutamatergic synapses.

NMDA receptors are tightly related to brain maturation. In PFC, functional NMDARs are expressed in tangentially migrating interneuron precursors [88]. Depending on the electrophysiological properties of interneurons, NMDAR mediated-currents vary. Regular-spiking (RS) interneurons maintain NMDA-mediated currents constant through development, whereas fast-spiking (FS) interneurons have a large decay [89]. This decrease is more prominent from postnatal weeks 2–4 and from weeks 12–15 [89]. NMDAR mediated currents of FS-cells decrease approximately from 75% in juvenile rats to 25% in adult animals. This is probably secondary to changes in NR2 subunits. In fact, brain circuitry maturation usually coincides with NMDAR subunit switch, marking the transition from young to adult neural processing. NR2 subunit switch is cell type-specific in prefrontal cortex, with NR2B levels remaining constant until adulthood in pyramidal cells, but with a gradual replacement from NR2B to NR2A in fast-spiking interneurons, particularly in adolescence [89, 90]. Subunits switch makes NMDARs extremely vulnerable to genetic risk factors and environmental perturbations, and both interact to affect normal brain development [91]. Similarly, NMDAR subunit expression and function in hippocampus is necessary for proper hippocampal development, with NMDAR dysregulation resulting in failures in synaptogenesis and circuit maturation [92,93,94,95].

Recent evidence supports the finding of abnormal glutamatergic transmission and NMDAR hypofunction in schizophrenic individuals [84]. Firstly, multiple genes involved in increased risk for schizophrenia are known to alter NMDAR-mediated signaling [96, 97]. Susceptibility genes for schizophrenia therefore regulate neuronal proliferation, migration, and synaptogenesis. Secondly, dysregulation of NMDAR subunits in postmortem brains of schizophrenic patients, in which NR1 subunits are decreased, further indicates perturbed NMDA function. Thirdly, transgenic mice with low NMDAR expression and animal models of NMDA antagonism present symptoms reminiscent of schizophrenia. NMDA antagonism not only produces behavioral changes, but also patterns of metabolic and neurochemical alterations of the disease [98]. Taken together, it is increasingly recognized that schizophrenia is a neurodevelopmental disorder, in which early brain development is affected [99].

Effects of MK-801 on NMDA Receptors

As stated in a previous section, failure in glutamatergic neurotransmission is known to play a role in the pathophysiology of schizophrenia. MK-801 is a noncompetitive NMDA receptor antagonist that physically blocks the receptor by inserting in the channel pore, binding to several PCP-s binding sites, and preventing the flow of cations through the channel pore. Blocking NMDA receptors results in an excessive release of glutamate that can have an impact on the blocked neuron itself and on downstream brain regions. In fact, early-life MK-801 administration has pro-apoptotic effects shortly after exposure. The ability to activate apoptotic pathways depends on the duration and severity of NMDA blockade. Doses higher than 0.25 mg/kg are necessary to induce irreversible degeneration and cell death [100]. The mechanism for activating apoptotic cell death is not well established, although NMDA receptor coupling to ERK1/2-CREB in early brain development has been proposed to be vital for neurotrophic action of NMDA receptor. MK-801-induced apoptotic injury is believed to result from the dissociation of the NMDA receptor from the ERk1/2-CREB signaling pathway [82]. Ikonomidou et al. [100] reported that only neuronal cells committed apoptosis with no glial cell activation. Studies from the last decade, however, suggest glial impairment in the cerebral cortex of schizophrenics, including reduced glial cell size and density, and glial dysfunction in prefrontal cortex and hippocampus [101, 102]. Astrocytic glutamate metabolism, more specifically glutamate–glutamine–GABA cycle, is also perturbed after MK-801 administration. Glutamate and glutamine levels after repeated injections of high doses of MK-801 (0.5 mg/kg during 6 days) were comparable to those found in first-episode patients of schizophrenia [103]. The susceptibility for MK-801-induced damage correlates with the highest expression of NMDA receptors (around week 2) and growth spurt (peaks at P10). The excessive release of glutamate after NMDA antagonism, mediates excitotoxicity that goes beyond apoptotic pathways, affecting growth cone activity, and neurite extension and branching [104]. Neuronal injury, such as dendritic atrophy, is also seen in postmortem brains [17].

Glutamate-mediated excitotoxicity has considerable functional consequences. GABAergic interneurons are claimed to be 10 times more sensitive to NMDA receptor antagonism than pyramidal cells [105]. In particular, fast-spiking cells that express the calcium-binding protein parvalbumin (PV) seem to be the most vulnerable to damage after NMDA blockade, and their alteration is sufficient to induce behavioral traits that resemble symptoms of schizophrenia. Several studies have demonstrated that repeated MK-801 administration decreases PV immunoreactivity. Chronic NMDAR blockade in juvenile and adult rats diminishes PV+ densities in hippocampus, especially in the dentate gyrus and CA1 region, shortly after exposure [60, 62]. Neurodevelopmental models further support this finding, but usually doses higher than 0.5 mg/kg are needed to induce long-term structural and anatomical changes [106]. Such large doses not only alter hippocampal PV densities, but also PV cells in mPFC, a region involved in higher cognitive functions [61, 63, 64]. Moreover, Nakazawa et al. [41] demonstrated in transgenic mice that the lack of cortical and hippocampal NMDA receptors in GABAergic interneurons was sufficient to evoke schizophrenia-like features. Most of the altered GABAergic interneurons had, in fact, parvalbumin-positive immunoreactivity. The mechanisms by which PV+ FS-cells can be selectively susceptible to damage after MK-801 administration are unclear, although a number of hypotheses have been proposed. FS-cells express Kv1.3 channels that allow a fast repolarization of the membrane, and the ability to fire the next action potential very rapidly. The high frequency of action potential firing means that the open probability of NMDAR in FS-cells is much higher than in any other GABAergic cell or excitatory cells that fire at a slower rate. As MK-801 is an uncompetitive drug and needs the ion channel to be opened for blocking the receptor, the chances of blocking NMDAR of FS-cells is much higher. Another mechanism is the one described by Wang & Gao [107]. They demonstrated that NMDAR of presynaptic glutamatergic terminals targeting pyramidal cells and FS interneurons were distinctly affected after subchronic MK-801 exposure. Presynaptic NMDAR are critical to modulate and facilitate neurotransmitter release. Interestingly, MK-801 completely blocked presynaptic NMDA receptors in glutamatergic terminals that targeted FS interneurons, whereas new NMDA receptors were inserted in presynaptic terminals that made synaptic contact with pyramidal neurons. This mechanistic approach further confirms that synaptic mechanisms of NMDA blockade are cell-type specific. These two hypotheses are not mutually exclusive, and probably MK-801 alters NMDAR by several mechanisms, but all resulting in PV+ FS-cell underactivation, with overall disinhibition of pyramidal cell activity in cerebral cortex.

It is not known whether the NMDAR on astrocytes are blocked, how this could influence glutamate metabolism and transport or glycine release and uptake, and thereby local circuitry [108]. What has been demonstrated so far is that astrocytes affect the glutamatergic system. Notably, the upregulation of glutamate transporter-1 (GLT-1) mRNA [109], protein, and activity [110] in astrocytes has been found in prefrontal cortex of schizophrenic patients. Animal studies revealed that NMDAR antagonist phencyclidine provokes similar findings, although the effects of MK-801 have not been studied yet [111]. Furthermore, a selective deletion of astroglial A2AR, which tightly regulates GLT-1 activity [112], decreases working memory in rodents, as measured in radial arm maze [113]. Glycine transporters (GlyT) are present in both astroglial cells and neurons. It is well established that glycine plays a pivotal role in NMDAR neurotransmission, and GlyT are closely associated with NMDAR. Although it was believed that glycine levels in synaptic cleft were enough to saturate the glycine-B sites of the NMDA receptor, it is now known that these levels are below saturation. Under physiological conditions, glycine is actively removed by the action of presynaptic and postsynaptic glycine transporters (GlyT). Augmenting glycine availability in the synaptic cleft could therefore facilitate NMDAR function, and drugs that inhibit GlyT action have been proved to be effective in improving cognition in animal models of schizophrenia induced by MK-801 (Table 25.1) [114,115,116,117,118].

Table 25.1 Pharmacological compounds and behavioral tasks used to demonstrate that glycine transporter inhibition improves cognitive function in MK-801-induced rodent model of schizophrenia
Full size table

Effects of MK-801 on Brain Circuits and Activity

Acute systemic administration of MK-801 increases mPFC activity, and the decreased signal-to-noise ratio could account for the deficits in mPFC-dependent executive functions. Unraveling the mechanisms by which this occurs could contribute to a better understanding of the dysfunctional brain. As previously suggested, a local disinhibition of mPFC secondary to PV+ FS GABAergic cell underactivation could explain increased excitability of this region. Nevertheless, it has been shown that local infusion of MK-801 in CA1 of hippocampus augments mPFC neural activity in a similar manner to systemic administration [119]. According to these findings, local disinhibition of pyramidal cells in CA1 that send glutamatergic projections to mPFC through hippocampal-PFC pathway could contribute to the overexcitation of mPFC. The hippocampal–PFC pathway supports high-order cognitive functions [120,121,122]. Blot et al. [29] demonstrated an aberrant form of plasticity in this pathway after acute MK-801 administration that correlated with impaired working memory and learning flexibility in rodents. NMDA receptors are involved in brain plasticity, and cellular models of learning and memory formation such as LTP or LTD are NMDAR-dependent. Alterations in plasticity could underlie the pathophysiology of schizophrenic symptoms. In fact, human studies indicate that there is impaired glutamatergic plasticity in schizophrenic brains [123], and the hippocampal–prefrontal pathway is crucial in the pathophysiology [28]. Blot et al. [29] demonstrated that a single dose of MK-801 (0.1 mg/kg) evoked long-lasting response to synaptic input from the ventral hippocampus that was independent of synchronized afferent inputs — necessary for standard LTP. The authors suggested that this was an aberrant form of plasticity. For LTP formation in mPFC of hippocampus–prefrontal pathway synapses, concurrent activation of dopaminergic and NMDA receptors is necessary. As suggested by the authors, an excessive release of glutamate and dopamine by the action of MK-801 in mPFC could be the mechanism by which the aberrant form of plasticity takes place. Furthermore, subchronic administration of MK-801 during 14 days hindered LTP induction. According to the results of Manahan-Vaughan et al. [117], LTP induction and expression was also profoundly impaired in the dentate gyrus 1 week after administering MK-801 acutely (5 mg/kg), and consequently learning deficits were present in MWM. LTP and learning performance were rescued by application of glycine transporter-1 inhibitors discussed before. In a subsequent study of the same group, LTP impairment was also present 4 weeks after drug administration, due to hippocampal hyperactivity and changes in expression of GABA and NMDA receptors in the prefrontal cortex and hippocampus that led to poor inhibitory control of prefrontal cortex [81]. This uncoupling of hippocampal–prefrontal communication could account for learning impairments and memory dysfunction found in MK-801-treated animals.

The nucleus accumbens has been identified as another important brain region that links several findings of schizophrenic disturbances [124,125,126]. The vast majority of neurons in nucleus accumbens are medium spiny neurons, a special type of GABAergic inhibitory neurons that have long-range projections. They receive glutamatergic inputs from hippocampus, prefrontal cortex, and amygdala, and dopaminergic inputs from the ventral tegmental area (VTA) (Fig. 25.2). Medium spiny neurons modulate inhibitory control on thalamocortical glutamatergic neurons. Local blockade of NMDA receptors in nucleus accumbens has been demonstrated to be sufficient to augment PFC activity and impair working memory [127]. Mesolimbic pathway connects dopaminergic neurons from ventral tegmental area with GABAergic neurons of nucleus accumbens. There is a differential dysregulation of mesocortical and mesolimbic dopaminergic pathways in schizophrenia. Mesocortical pathway is hypoactivated, whereas increased firing of dopaminergic neurons in mesolimbic pathway is found. The overexcitation of mesolimbic pathway prevents proper inhibition by accumbal neurons. This results in sensory information overload to cortex and a lack of feedback inhibition to VTA dopaminergic neurons. In turn, this will lead to an excessive release of glutamate in nucleus accumbens through the overactivation of glutamatergic neurons projecting to nucleus accumbens (Fig. 25.2).

Fig. 25.2
figure 2

Schematic representation of the major afferent and efferent connections of the NAc. PFC prefrontal cortex, HPC hippocampus, DM Thal dorsomedial thalamus, VP ventral pallidum, NAc nucleus accumbens, VTA ventral tegmental area

Full size image

Effects of MK-801 on Cognition

Memory function enables storage and retrieval of information over variable periods ranging from seconds to years, and is critical to daily life functioning. In schizophrenic patients, not only memory but all areas of cognition seem to be impaired, suggesting widespread cortical dysfunction. Cognitive symptoms are considered a distinct dimension of the illness, and are relatively independent of positive and negative symptoms. Cognitive deficits include impaired working, spatial, and declarative memory, attentional dysfunction, and poor cognitive flexibility. Working memory [19] and episodic memory [128] are especially sensitive to neuropsychiatric disorders, and appear to be core features of schizophrenia. Valid schizophrenia animal models should mimic at least one of these cognitive deficits.

Behavioral paradigms that evaluate working memory are non-matching to sample of objects or odors, operant tasks, and paradigms that use spatial information, such as maze tasks (delayed alternation, radial-arm maze, Morris water maze) [129]. Using these devices, deficits in working memory have been consistently shown after MK-801 administration [130,131,132,133,134]. Morris water maze (MWM) and radial arm maze (RAM) are the most widely performed tasks for spatial learning and memory assessment. MK-801 impairs spatial learning and memory. Concretely, acquisition, reversal learning, and working memory performance are affected in hidden-platform trials, although reference memory is spared [106, 135,136,137,138,139]. Spatial learning tasks are readily available both in humans and animals, and allow a direct translation of findings. In an attempt to bridge the gap between human and animal research, a variant of Morris water maze has been used in schizophrenic patients, showing decreased navigational abilities in human patients, and further confirming spatial impairment [140]. Deficits in cognitive or behavioral flexibility have been documented in schizophrenic patients [141], a type of executive function carried out by prefrontal cortex. MK-801-induced animal models have also displayed problems in cognitive flexibility measured by reversal learning in MWM, as stated previously, and by active place avoidance tasks in radial arm maze [29, 136]. It is well established that hippocampal neurons are essential for spatial navigation. Nevertheless, the neurocircuitry involved in spatial learning and memory contains different systems that collaborate in serial or parallel fashion. In this context, the role of medial prefrontal cortex is essential. Neurons from cornus ammonis 1 (CA1) send projections to frontal areas, mainly the prelimbic and cingulated cortices [142]. Disconnecting hippocampus from the medial prefrontal cortex impairs spatial memory and spatial working memory in rodents [120, 121]. It seems that spatial information is acquired by the hippocampus and then transferred to mPFC.

Novel object recognition (NOR) is one of the most widely performed preclinical cognitive tests for schizophrenia [143], and it has long been considered the analog of human episodic memory. Multiple schizophrenia-relevant studies have used NOR for exploring cognitive impairment in rodents [144, 145]. NOR is based on a rodent’s natural tendency to explore new stimuli and environments. Following acute administration of MK-801, NOR is severely disrupted [146, 147, 118] but in neurodevelopmental models these results failed to be replicated. In fact, early-life repeated injections of MK-801 have no long-term consequences in NOR, using delays of 1.5 h [148], 2 h [149] and 5 h [150] between acquisition and test trials. Using NOR for evaluating declarative memory has raised some concerns, as it seems that NOR is a familiarity-based test, which is not affected in schizophrenic patients, rather than a recollection-based test [151, 152]. Tests that use associative or relational information are more closely related to human episodic memory. Such tests need temporal and spatial precision of object memory. Li et al. [64] demonstrated that associative recognition memory was impaired in adolescence and adulthood after early-life NMDA blockade. The dissociation of results between NOR and tests that use relational information in MK-801 rodent model could be explained by neural circuitry. Brain wiring for rodent recognition memory involves several structures, but perirhinal (Prh) cortex plays a major role [153]. Although hippocampus (HPC), medial prefrontal cortex (mPFC), mediodorsal thalamus (MD) and postrhinal cortex (PostRh) participate in recognition memory, NOR is particularly sensitive to perirhinal cortex dysfunction, and not to hippocampal alterations. Rather, the role of hippocampus is to integrate object information with spatial or contextual information. Similarly, mPFC integrates spatial information from CA1 subfields of HPC with object information of Prh cortex, using NMDA-dependent synaptic plasticity [154]. Therefore, associative tasks require network interdependency across multiple structures, in which HPC–mPFC-Prh circuits are essential for memory acquisition and retrieval. Furthermore, associative memory depends on NMDA receptor neurotransmission, and hippocampal NMDAR are required for acquisition, but not retrieval, of associative memories [154]. mPFC and HPC are two critical brain regions in the pathophysiology of schizophrenia, so assessing cognitive functions dependent on the interaction of both regions will preferably reveal cognitive deficits in rodents.

Conclusions

Neonatal administration of MK-801 instills a hypofunction of NMDAR that results in a widespread apoptotic injury during synaptogenesis. The GABAergic system is particularly sensitive in the developing brain to transient NMDA blockade. Among subpopulations of GABAergic interneurons, PV+ cells seem to provide the most potent inhibitory input on pyramidal neurons by formation of synaptic contacts in soma, proximal dendrites, and axon initial segment. A decrease in PV+ interneuron population following MK-801 administration therefore has a significant impact on neuronal synchrony and information processing. Disrupted synaptic integration in early brain development results in modified network activity and plasticity in adulthood. This is corroborated by behavioral data that show a constellation of neurobehavioral sequelae that resemble symptoms of schizophrenia.

Mimicking behavioral and cognitive deficits of schizophrenia in animal models is a real challenge, and the validity of cognitive deficits in animal models largely depends on the appropriate behavioral paradigm. Cognitive impairment is best assessed by means of tasks that require mPFC–HPC interactions, a circuit that underlies episodic, working, and spatial memory. Although deficits elicited by repeated NMDA antagonism are unlikely to represent an animal model of schizophrenia per se, a strong body of evidence supports an MK-801-induced neurodevelopmental model of schizophrenia as a valid model for some of the essential deficits occurring in this condition.