Introduction

Schizophrenia remains a disabling illness characterized in most patients by residual symptoms and recurrent relapse despite recent advances in pharmacological treatment. All antipsychotic agents share in common antagonism of dopamine D2 receptors; the more recent “second-generation” agents additionally share serotonin (5-HT2) receptor antagonism but differ from each other in possessing diverse patterns of activity at other receptors (Schotte et al. 1996; Zhang and Bymaster 1999). Among the newer agents, only clozapine has clearly demonstrated enhanced efficacy in patients refractory to D2 antagonists, although roughly half of patients still fail to respond to clozapine (Kane et al. 1988). No treatments have been established for patients who fail clozapine or are only partially responsive (Barnes et al. 1996).

New pharmacological models of schizophrenia have provided novel pharmacological strategies for the treatment of the disorder. Of particular interest is a model that posits hypofunction of the glutamatergic N-methyl-d-aspartate (NMDA) receptor, resulting in dysregulation of cortical neuronal activity and glutamate transmission (Olney and Farber 1995; Goff and Coyle 2001; Krystal et al. 2003). For decades it has been recognized that the potent NMDA channel blocker, phencyclidine (PCP), produces psychotic symptoms in abusers that are remarkably similar to symptoms of schizophrenia (Allen and Young 1978; Bakker and Amini 1961; Javitt and Zukin 1991; Adler et al. 1999); in chronic users, an irreversible psychosis with cognitive impairment has also been described (Rainey and Crowder 1975; Jentsch and Roth 1999). Infusion of ketamine, which also blocks the NMDA channel, also simulates the full range of symptoms of schizophrenia in healthy subjects (Krystal et al. 1994). These symptoms were not attenuated by haloperidol (Krystal et al. 1999a), though haloperidol has been used to treat psychotic symptoms induced by phencyclidine in an emergency-room situation (Giannini et al. 1987). In patients with schizophrenia stabilized on haloperidol, ketamine infusion produces a florid relapse of symptoms (Lahti et al. 1995; Malhotra et al. 1997a). Clozapine is able to blunt these effects (Malhotra et al. 1997a,b).

Glycine agonists and glycine transport inhibitors have been shown to reduce negative symptoms of schizophrenia when added to antipsychotic drugs (Tuominen et al. 2005). These agents enhance NMDA receptor function (Bergeron et al. 1998) through activation of the NMDA glycine allosteric modulatory site (Dannhardt and Kohl 1998). In preclinical studies, glycinergic agents suppress phencyclidine-induced behavioural disruption and glutamate release in rats (Javitt et al. 1999). In clinical trials, the glycine site partial agonist, d-cycloserine, the full agonists, glycine and d-serine, and the glycine transporter I inhibitor, sarcosine, have all demonstrated therapeutic benefit for negative symptoms and, in some studies, benefit for psychotic and cognitive symptoms when added to antipsychotic agents other than clozapine (Heresco-Levy 2000; Tsai et al. 1998,2004). d-Cycloserine monotherapy significantly improved negative symptoms in a single-blind trial (van Berckel et al. 1996). Thus, these results suggest that hypofunction of NMDA receptors might also occur in schizophrenia and might be involved in the generation of negative symptoms. However, clinical results for the partial agonist d-cycloserine have been mixed, since in other studies d-cyloserine added to antipsychotic drugs was found to worsen positive symptoms (van Berckel et al. 1999), or else have no therapeutic benefit (Goff et al. 2004). Furthermore, the combination of d-cycloserine with clozapine resulted in worsening of negative symptoms (Goff et al. 1999; Heresco-Levy and Javitt 2004) and addition of other glycine site agonists to clozapine did not produce clinical benefit (Evins et al. 2000; Potkin et al. 1999; Tsai et al. 1999). Unlike most other atypical antipsychotic drugs, clozapine may facilitate NMDA receptor activation through inhibition of type 1 glycine transporters (Javitt et al. 2004) or through a dopamine D1-dependent mechanism (Chen and Yang 2002; Ninan and Wang 2003). Changes in serum or brain glutamate concentrations after switching from conventional agents to clozapine or olanzapine have tended to correlate with enhanced efficacy for negative symptoms (Goff et al. 2002; Evins et al. 1997). It is possible that these effects interfere in some way with the modulation of NMDA receptor function by glycinergic drugs. In conclusion, studies with glycinergic agents support the hypothesis for a hypofunction of NMDA receptors in schizophrenia. Furthermore, a number of susceptibility genes for schizophrenia are associated with NMDA receptor signalling (Harrison and Weinberger 2005). However, activation of the glycine modulatory site on the NMDA receptor may not be sufficient to treat the range of symptoms of schizophrenia and may be complicated by adverse interactions with clozapine. Other strategies to address the downstream consequences of NMDA receptor hypofunction should be considered.

NMDA antagonists facilitate the release of dopamine (Miller and Abercrombie 1996; Svensson 2000), both in the limbic system (Carboni et al. 1989; Steinpreis and Salamone 1993; Hertel et al. 1995) and frontal cortex (Deutch et al. 1987; Hondo et al. 1994; Adams and Moghaddam 1998). However, dopamine antagonists do not consistently block the psychotic symptoms induced by NMDA antagonists in humans (Krystal et al. 1999a), and dopamine depletion is unable to prevent NMDA-antagonist-induced behaviour in rats (Carlsson and Carlsson 1989). These observations argue against a primary role for dopamine in the generation of psychotic symptoms by NMDA antagonists. NMDA antagonists also enhance the release of glutamate (Moghaddam et al. 1997) and cause alterations in the firing pattern of cortical pyramidal neurons (Shi and Zhang 2003; Jackson et al. 2004). Thus, it is hypothesized that dysregulation of glutamate transmission may be central to the psychotomimetic effects of NMDA antagonists.

The alteration of neuronal activity and enhanced release of glutamate is suggested to arise through two mechanisms. First, attention has focused on a subgroup of NMDA receptors localized on inhibitory GABAergic interneurons. This follows in part from the finding that phencyclidine and ketamine bind with highest affinity to this subgroup (Grunze et al. 1996). It is postulated that diminished NMDA receptor activity on inhibitory interneurons in the schizophrenic brain results in reduced inhibitory drive at GABA-A receptors on glutamatergic pyramidal neurons (Fig. 1). The disinhibition of pyramidal neurons, leading to a disruption of their normal pattern of firing and synchronization (Shi and Zhang 2003; Jackson et al. 2004), results in increased release of glutamate and activation of non-NMDA receptors (Olney and Farber 1995). Drugs that modulate glutamate release, such as the selective type 2/type 3 metabotropic glutamate receptor (mGluR2/3) agonist, LY354740, can block the effects of PCP in rodents (Adams and Moghaddam 1998; Schoepp and Marek 2002), and produce a dose-dependent suppression of ketamine-induced impairment of working memory by ketamine in human volunteers (Krystal et al. 2005).

Fig. 1
figure 1

Hypothesis for glutamate neuron dysregulation induced by NMDA receptor antagonists. Selective block of NMDA receptors on interneurons in the cortex leads to disinhibition of pyramidal neurons leading to altered firing and increased glutamate transmission via non-NMDA glutamate receptors (Olney and Farber 1995). NMDA antagonists may also increase 5-HT release in the cortex. This is proposed to lead to activation of 5-HT2a receptors that further enhance glutamate release (Aghajanian and Marek 1999). Pyr Cortical pyramidal neuron, Inter inhibitory interneuron

Full size image

Secondly, NMDA antagonists elevate serotonin levels in the prefrontal cortex and hippocampus (Martin et al. 1998). Serotonin, acting at 5-HT2A receptors, in turn can enhance the release of glutamate (Aghajanian and Marek 1999) (Fig. 1). In support of this mechanism, 5-HT2 antagonists have been shown to reduce behavioural effects of NMDA antagonists in rats (Bakshi and Geyer 1997; Carlsson et al. 1999; Gleason and Shannon 1997). The selective 5-HT2A antagonist, M100907, has been shown to suppress NMDA-antagonist-induced c-fos expression (Habara et al. 2001), disruption of paired-pulse inhibition (Varty et al. 1999), impairment in attentional performance (Mirjana et al. 2004), and increased glutamate release (Ceglia et al. 2004).

The results with M100907 and with the mGluR2/3 agonist, LY354740, support the hypothesis that disruption of glutamate transmission is an important downstream consequence of NMDA receptor antagonism that might underlie the psychotomimetic effects of ketamine or phencyclidine. However, the situation is likely to be more complex, since LY354740 did not significantly attenuate ketamine-induced expression of psychosis or negative symptoms in healthy volunteers (Krystal et al. 2005), and M100907 monotherapy was not beneficial in patients with schizophrenia (Talvik-Lotfi et al. 2000; de Paulis 2001). In contrast, combination of 5-HT2a receptor antagonism with D2 receptor blockade appears to be an effective treatment strategy, since most atypical antipsychotic drugs have significant activity at both receptors (Breese et al. 2002; Seeman 2002). At this point, it would seem prudent to hypothesise that schizophrenia involves dysregulation of both glutamate and dopamine transmission.

There is increasing evidence that schizophrenia is associated with a gradual neurodegeneration (Konradi and Heckers 2003; Krystal et al. 2003). We may hypothesise that dysregulation of glutamate release might lead to excitotoxicity and neuronal injury. Chronic administration of NMDA antagonists has been shown to cause neuronal degeneration in posterior cingulate and retrosplenial cortex of rats (Olney and Farber 1995; Farber et al. 2002; Konradi and Heckers 2003). However, it remains to be seen whether the pattern of damage induced by NMDA antagonists reflects that observed in the post-mortem schizophrenic brain.

The glutamate neuron dysregulation hypothesis provides a basis from which to propose new pharmacological approaches to the treatment of schizophrenia. In this review, we examine the evidence for the efficacy of lamotrigine in combination with antipsychotic drugs for the treatment of schizophrenia. We consider how this efficacy might relate to glutamate neuron dysregulation, the mechanism by which lamotrigine influences neuronal activity and glutamate transmission, and the possible ways in which the drug might interact with antipsychotic medications. We then consider whether lamotrigine monotherapy would be effective in the treatment of schizophrenia within a framework of dysregulation of glutamate and dopamine transmission in schizophrenia. Based on this, we argue that the treatment of schizophrenia may require a combination of anti-dopaminergic and anti-glutamatergic drugs, at least during the active psychotic phase of the illness.

Clinical and experimental medicine findings with lamotrigine

Lamotrigine is a broad spectrum anticonvulsant (Messenheimer 1995) used also in the treatment of bipolar disorder (Table 1). Its downstream effects on neuronal function include inhibition of glutamate release. Thus, consistent with the glutamate neuron dysregulation hypothesis for schizophrenia, pre-administration of 300 mg of lamotrigine 2 h before ketamine challenge attenuated psychotic, negative, and cognitive effects in a placebo-controlled trial in healthy volunteers (Anand et al. 2000a). Clinical experience also supports a potential therapeutic effect of lamotrigine in schizophrenia (Table 1). Dursun and McIntosh (1999) added lamotrigine (125–250 mg/day) to clozapine in six treatment-resistant schizophrenic patients in a 24-week open-label trial. Lamotrigine was well tolerated and produced a mean 75% reduction in Brief Psychiatric Rating Scale (BPRS) total scores, with maximal improvement occurring in all but one subject between weeks 8 and 16. In subsequent open trials, Dursun and Deakin (2001) found that lamotrigine augmentation of olanzapine, risperidone, and conventional antipsychotics was of considerably less benefit compared with augmentation of clozapine, although the number of patients in this study was small. Tiihonen et al. (2003) conducted a 10-week, double-blind, placebo-controlled, parallel-group trial adding 200 mg of lamotrigine daily to clozapine in 36 treatment-resistant schizophrenic patients. Compared with placebo, lamotrigine significantly reduced ratings of psychosis, but not negative symptoms. The investigators noted that auditory hallucinations were particularly responsive, although only 20% of subjects accounted for most of the therapeutic benefit. Kremer et al. (2004) recently reported improvement of psychotic symptoms with the addition of up to 400 mg of lamotrigine daily to typical or atypical antipsychotics in 38 treatment-resistant schizophrenia patients. Improvement was significant compared with placebo in completers but did not achieve significance in the primary LOCF analysis. Kremer et al. also reported no significant improvement in negative symptoms. Whereas the study by Tiihonen et al. added lamotrigine to clozapine only, Kremer et al. found that psychosis improved equally in patients treated with typical and atypical agents.

Table 1 Clinical evidence for the efficacy of lamotrigine in bipolar disorder and schizophrenia
Full size table

Since lamotrigine and other anticonvulsant drugs with different mechanisms of action are commonly used in the treatment of psychiatric disorders (Rogawski and Loscher 2004), we briefly consider two other anticonvulsants that, like lamotrigine, are used in the treatment of bipolar disorder. Valproate is the anticonvulsant most frequently prescribed to patients with schizophrenia (Citrome et al. 2000). In a recent double-blind, randomized study of 249 schizophrenic patients admitted for acute exacerbation of symptoms, the combination of valproate added to risperidone or olanzapine produced a more rapid response than antipsychotic monotherapy; however, the difference in response between groups diminished over time and was no longer significant by week 4 of treatment (Citrome et al. 2004; Casey et al. 2003). In contrast, three controlled studies found no therapeutic benefit of valproate adjunctive therapy in chronic, treatment-refractory patients (Ko et al. 1985; Dose et al. 1998; Hesslinger et al. 1999; Basan and Leucht 2004). Another anticonvulsant, carbamazepine, was found to improve symptoms when added to antipsychotic medication in patients with schizophrenia who had EEG abnormalities (Neppe 1983). The combination of carbamazepine with antipsychotic agents is complicated by potential induction by carbamazepine of antipsychotic metabolism (Arana et al. 1986), which would reduce exposure to the antipsychotic agent. The fact that despite this symptoms improved in the study by Neppe (1983) should be taken as further support for the efficacy of carbamazepine in this setting. However, in a placebo-controlled, crossover trial of carbamazepine monotherapy, no improvement in the symptoms of chronic outpatients was observed (Carpenter et al. 1991).

Summary of clinical studies

Thus, clinical data with lamotrigine support the use of drugs that reduce glutamate release as adjunctive treatments for patients with schizophrenia. However, it remains to be seen whether this will be confirmed by larger double-blind placebo-controlled trials. There are insufficient data to conclude whether or not a similar anticonvulsant drug, carbamazepine, may also have adjunctive efficacy in schizophrenia, whereas studies suggest that valproate, an anticonvulsant with a different mechanism of action, does not produce a sustained effect. It is notable that lamotrigine and valproate also have different clinical profiles in the treatment of bipolar disorder (Ketter and Calabrese 2002), whereas they have similar broad-spectrum efficacy in the treatment of epilepsy (Loscher 1999). The glutamate neuron dysregulation model of schizophrenia predicts that facilitation of NMDA receptor function on inhibitory interneurons or inhibition of glutamate release should produce similar therapeutic effects. Whereas evidence now supports the therapeutic benefit of both approaches, the pattern of clinical efficacy has not been consistent. Glycine site agonists predominantly treat negative symptoms (Tuominen et al. 2005), whereas lamotrigine seems to improve only positive symptoms (Tiihonen et al. 2003; Kremer et al. 2004). Evidence for efficacy with glycine site agonists has been strongest when added to conventional antipsychotics and weakest when added to clozapine, whereas the opposite may be true with lamotrigine. The difference between these two approaches is discussed further later on. However, more data are needed from large clinical trials to verify the preliminary observations.

Efficacy of lamotrigine in NMDA antagonist models of psychosis

With the assumption that NMDA antagonists can be used to model schizophrenia in the laboratory, at least at a superficial level, they are increasingly used to investigate the potential efficacy of novel antipsychotic drugs, particularly those that do not rely on dopamine D2 receptor blockade (Krystal et al. 1999b). As discussed earlier, drugs that negatively modulate glutamate transmission, such as 5-HT2a antagonists (Mirjana et al. 2004; Ceglia et al. 2004) and mGluR2/3 agonists (Schoepp and Marek 2002; Carter et al. 2004; Krystal et al. 2005), have already been shown to reduce some NMDA-antagonist-induced symptoms in humans or behavioural or neurochemical effects in rodents. In the following section we review the few studies that have examined the effects of lamotrigine and two other anticonvulsant drugs, carbamazepine and valproate, in NMDA antagonist models (Table 2). The first model, disruption of prepulse inhibition (PPI) by psychotomimetic agents, is used frequently to assess the antipsychotic potential of novel drugs (Geyer et al. 2001). The second model, in which psychotomimetic agents were used to disrupt an operant reversal-learning task, was chosen because it may allow assessment of the effects of the novel agents on cognitive function, which is a symptom domain of particular importance in schizophrenia. The final model, in which NMDA antagonists induce neurodegeneration, intends to provide information about the neuroprotective potential of novel drugs, since this is likely to be an important long-term aim of antipsychotic treatment.

Table 2 Preclinical evidence for the efficacy of lamotrigine in psychiatry models
Full size table

Lamotrigine prevents the disruption of prepulse inhibition by ketamine in mice

Prepulse inhibition can be used as a measure of sensorimotor gating, a process by which organisms filter information from the environment (reviewed by Braff and Geyer 1990; Braff et al. 2001). PPI is deficient in certain psychiatric disorders, notably schizophrenia (Braff et al. 1992), Huntington's disease (Braff et al. 2001), and mania (Perry et al. 2001a). Disruption of PPI can be achieved pharmacologically in rodents using NMDA receptor antagonists (Mansbach and Geyer 1989) or drugs that increase dopamine D2 receptor activation, such as d-amphetamine (Ralph et al. 1999). These pharmacological disruptions can be prevented by certain antipsychotic drugs and, as a result, the model is frequently used to study or predict the efficacy of novel agents (Bakshi et al. 1994; Varty and Higgins 1995; Swerdlow and Geyer 1998; Geyer et al. 2001).

The ability of lamotrigine to prevent d-amphetamine- and ketamine-induced disruption of PPI in 129SvPasIco mice was recently investigated by Brody et al. (2003). Lamotrigine at a dose of 27 mg/kg i.p. prevented disruption of PPI by ketamine, but the drug had no effect on d-amphetamine-induced disruption. Lamotrigine alone had no effect on startle magnitude or on PPI itself. In a separate study by the same group, carbamazepine (50 mg/kg i.p.) also prevented disruption of PPI by ketamine, but not amphetamine (Ong et al. personal communication). However, carbamazepine also significantly reduced startle magnitude, which may confound interpretation of the PPI data. In contrast, sodium valproate (400 mg/kg i.p.) did not prevent ketamine or amphetamine-induced disruption of PPI. Valproate did significantly reduce baseline startle magnitude, although it did not affect baseline PPI. These results suggest that anticonvulsant efficacy per se is insufficient to prevent the disruptive effects of ketamine on PPI, but that sodium channel inhibition may be important.

To what extent do these results support the efficacy of lamotrigine in schizophrenia? PPI is robust and ubiquitous across species, including rodents and humans, suggesting that it is a fundamental trait of sensory processing in mammals. PPI deficits have been correlated with positive and negative symptoms of schizophrenia (Perry and Braff 1994; Perry et al. 1999; Weike et al. 2000), in particular thought disorder (Perry et al. 1999). It is hypothesized that impairment in sensorimotor gating may leave patients unable to screen out distracting stimuli (Perry and Braff 1994). There is evidence that the treatment of patients with antipsychotic drugs improves and, in some reports, normalizes PPI (Weike et al. 2000; Kumari et al. 2002), although this has not been confirmed by all studies (Perry et al. 2001b). However, in preclinical studies that used ketamine to disrupt PPI, the efficacy of lamotrigine or carbamazepine confirms only that these drugs are able to inhibit some mechanism downstream of NMDA receptor antagonism, presumably the enhanced release of glutamate. Thus, the degree of similarity between the pathology underlying schizophrenia and the mechanism underlying disruption of rodent behaviour by ketamine is of critical interest, in particular since there is evidence that the effects of NMDA antagonists on PPI in rodents may be qualitatively different in humans. An early study of healthy human volunteers by Karper et al. (1994) did observe a small reduction in PPI with ketamine, but startle magnitude was also altered, which tends to confound analysis of PPI data. Subsequent studies with ketamine in healthy subjects found that the drug either did not alter (van Berckel et al. 1998) or else increased PPI (Duncan et al. 2001; Abel et al. 2003), despite the co-appearance of psychotic symptoms. A possible explanation for the apparent discrepancy may be the difficulty in testing a range of doses and pretreatment times in human volunteer studies; thus it can be difficult to compare results between human and animal studies (discussed by Krystal et al. 2003). Thus, although the PPI deficit induced by ketamine in mice was prevented by pretreatment with lamotrigine or carbamazepine (Brody et al. 2003), it is not necessarily the case that these drugs would reverse an established PPI deficit in patients with schizophrenia. It would be valuable to conduct a clinical study in patients with schizophrenia to investigate the ability of lamotrigine to reverse the deficits in PPI.

Lamotrigine prevents the disruption of a reversal-learning task by PCP in rats

Reversal-learning tasks probe the ability of an individual to adapt to changing stimulus-reward contingencies and depend on the integrity of the ventral prefrontal cortex (Clark et al. 2004). Performance of these tasks is often disrupted in patients with schizophrenia (Mackintosh and Little 1969) and is thought to be a consequence of hypofrontality (Weinberger and Berman 1996). For example, in the Wisconsin card-sorting task, patients with schizophrenia tend to persevere with the initial contingency and find it difficult to adapt when a new contingency is introduced (Deicken et al. 1995). A similar pattern of disruption can be achieved with ketamine in healthy volunteers (Krystal et al. 2000). Reversal learning can also be demonstrated in non-human primates (Smith et al. 1999) and rodents (Jentsch and Taylor 2001; Abdul-Monim et al. 2003). In the latter, performance could be disrupted by administration of PCP or d-amphetamine. The ability of lamotrigine to prevent PCP or d-amphetamine-induced disruption of reversal learning in rodents was studied by Idris et al. (2005). Lamotrigine (5–20 mg/kg i.p.) dose-dependently reduced the deficit in reversal learning induced by PCP, but did not prevent the deficit induced by d-amphetamine. In the same study, the atypical antipsychotic clozapine partially reduced the deficit in reversal learning induced by PCP, but also had no effect against d-amphetamine-induced deficits. In a previous study by the same group, the atypical antipsychotic, ziprasidone, prevented PCP-induced disruption of reversal learning, whereas the classical antipsychotic, haloperidol was ineffective (Abdul-Monim et al. 2003). However, haloperidol could prevent disruption induced by d-amphetamine (Idris et al. 2005). Sodium valproate at doses up to 200 mg/kg i.p. was unable to reduce the deficit in reversal learning induced by PCP or d-amphetamine (Idris et al. personal communication). Carbamazepine has yet to be studied in this model.

These results are similar to those from the PPI model and similar arguments can be applied to their interpretation—specifically, that interaction of lamotrigine with the mechanism by which PCP disrupts reversal learning in the rat does not necessarily predict efficacy of lamotrigine against an established reversal-learning deficit in patients with schizophrenia. However, an advantage of the reversal-learning paradigm over PPI is that disruption by NMDA antagonists does appear to translate from animals to humans (Krystal et al. 2000). Despite this, it remains to be determined whether NMDA-induced disruption of reversal learning and the disruption observed in patients with schizophrenia share a common substrate. Given that reversal-learning tasks have been shown to depend on functioning of the orbitofrontal cortex (reviewed by Clark et al. 2004), an informative way forward might be to use functional imaging techniques in healthy human subjects to examine the effects of ketamine on this brain region during performance of a reversal-learning task and compare these results with studies of reversal learning in patients (Volz et al. 1997; Riehemann et al. 2001). The orbitofrontal cortex may be of particular interest for studying the mechanism of action of lamotrigine because lamotrigine has been shown to increase the BOLD response selectively in this region during application of transcranial magnetic stimulation over the frontal cortex (Li et al. 2004). Finally, as in the case of PPI, a study of the ability of lamotrigine to improve a deficit in reversal learning in patients with schizophrenia would add a great deal to our understanding of the relationship between reversal learning deficits in animal models, human volunteer studies, and patients.

Lamotrigine reduces the number of cortical neurons injured by MK-801 in rats

The predominant view of schizophrenia is that it is caused by a neurodevelopmental dysfunction (Weinberger 1996; Bunney et al. 1997; Miyamoto et al. 2003). However, there is increasing evidence that the disorder is associated with a gradual neurodegenerative decline (Konradi and Heckers 2003; Krystal et al. 2003; Lewis et al. 2003). The neurodegenerative component may be operative at disease onset or may develop only later in the course of the disease. Neuroimaging studies have provided evidence for both. Brain volume reductions have been reported in regions such as the prefrontal cortex (Lim et al. 1996; reviewed in Bogerts 1999) and hippocampus (reviewed in Weinberger 1999), and similar reductions have been observed in first-onset patients (Vita et al. 1997; Davies et al. 1998; Cahn et al. 2002). Furthermore, longitudinal imaging studies of patients with schizophrenia document gradual morphometric changes with disease progression (DeLisi et al. 1997; Gur et al. 1998; Lieberman et al. 2001, 2005; Mathalon et al. 2001; Wood et al. 2001; Ho et al. 2003). Post-mortem studies of gliosis or apoptotic cell death suggest that gross cell loss does not occur in schizophrenia (Heckers 1997; Falke et al. 2000); however, gliosis may not be apparent if atrophy of the neuropil and synaptic disconnection are the major deficits. Indeed, immunocytochemical and ultrastructural studies found reductions in neuropil, dendritic arborization, synaptic spine density, and white matter density in post-mortem schizophrenia brains (reviewed in Konradi and Heckers 2003; Krystal et al. 2003; Lewis et al. 2003; Jones 2004). Region-specific increases and decreases in glutamate receptor expression (reviewed in Deakin and Simpson 1997) and alterations of white matter connectivity (Buchsbaum et al. 1998; Lim et al. 1999) may also occur. Deficits in cortical connectivity may be close to the root cause of schizophrenia. They may give rise to the aberrant thought processes characteristic of schizophrenia and may also fuel other changes, including further alterations in the neuropil, and altered function of the dopamine system (see later).

Progressive changes in cortical connectivity in schizophrenia would also be consistent with dysregulation of glutamate transmission. We can speculate that low doses of NMDA antagonists mimic acutely the cortical disconnectivity of schizophrenia and hence closely reproduce schizophrenic thought processes. Over a longer time course, NMDA antagonists may also mimic in rats some of the neuropathology observed in patients (Farber et al. 2002). In this study, which used the NMDA antagonist MK-801, Farber et al. (2002) found that sodium channel blocking agents, lamotrigine, carbamazepine, phenytoin, riluzole, and TTX all reduced the number of injured cortical neurons. They also found that valproate and other non-sodium channel blocking anticonvulsant drugs, such as gabapentin and felbamate, could also protect neurons. A similar effect has been observed with the mGluR2/3 agonist, LY379268 (Carter et al. 2004). These data suggest that unlike the behavioural disruption seen with NMDA antagonists, a wider range of pharmacological approaches might be effective. One explanation for this may be that the neurotoxicity induced by NMDA antagonists may not follow from the disruption of pyramidal neuron activity, as proposed in the model shown in Fig. 1, but instead may arise through an alternative mechanism unrelated to the psychotomimetic effects of the drugs (cf. Nakki et al. 1996). This suggests that NMDA antagonists may not necessarily model the neurodegenerative process in schizophrenia, which in any case occurs over a much longer time course. There is a need for further study of the nature and time course of the pathology caused by NMDA antagonist drugs, the similarity of the pathology to that observed in patients with schizophrenia, and the reliance of this pathology on altered glutamate transmission. However, even though NMDA antagonists may not accurately model the neurodegeneration associated with schizophrenia, they provide further evidence, in support of other models, for a broad neuroprotective effect of lamotrigine (Ketter et al. 2003; Rogawski and Loscher 2004; and see Table 3), valproate (Li et al. 2002; Gould et al. 2004; Hao et al. 2004; Vajda 2002; Rogawski and Loscher 2004), and possibly carbamazepine (Mai et al. 2002; but also see Nonaka et al. 1998). Thus, these drugs could offer some benefit as neuroprotectants for patients with schizophrenia.

Table 3 Preclinical evidence for the efficacy of lamotrigine in neurology models
Full size table

Finally, in relation to augmentation strategies, severity of tardive dyskinesia has been found to correlate with the concentration of excitatory amino acids in CSF of schizophrenia patients, suggesting that glutamatergic neurotoxicity may contribute to irreversible neurological injury from conventional antipsychotics as well (Goff et al. 1995). Thus, add-on treatment with neuroprotective drugs may be of long-term benefit for patients treated chronically with antipsychotic drugs.

Summary of preclinical studies

The preclinical data reviewed here demonstrate that acute treatment with lamotrigine and carbamazepine, but not sodium valproate, can prevent the disruption of rodent behaviour by NMDA antagonists. Because both of these drugs, but not valproate, are sodium channel inhibitors that can modulate the release of glutamate (see below), these studies add to a growing body of evidence that implicates glutamate transmission and the activity of cortical pyramidal neurons in the behavioural effects of NMDA antagonists. These results are consistent with the hypothesis that glutamate neuron dysregulation generates the psychotic symptoms observed in human subjects given ketamine. In addition to the observable similarity between symptoms induced by ketamine and symptoms of schizophrenia, there is increasing evidence that a disruption of glutamate transmission also occurs in schizophrenia.

The lack of efficacy of valproate in the two behavioural models reviewed above is surprising given that the effects of NMDA antagonists are proposed to arise through reduced activity of GABAergic interneurons (Olney et al. 1999; Krystal et al. 2003) (Fig. 1). Krystal et al. (1998) previously found that the benzodiazepine, lorazepam, was unable to reduce many of the effects of ketamine in healthy volunteers, although the drug did reduce the emotional distress caused by the psychotomimetic. A difficulty with studies involving GABAergic drugs is sedation that can confound interpretation of the results. However, it is recognised that GABA-enhancing strategies may have a place in the acute treatment of psychosis to reduce anxiety (Parepally et al. 2002). The use of valproate in this setting is supported by the clinical data reviewed earlier (Citrome et al. 2004; Casey et al. 2003). In addition, given the positive effects of valproate and other GABA-enhancing drugs vs NMDA-antagonist-induced neurodegeneration (Farber et al. 2002), these drugs may also confer some neuroprotective benefit.

Finally, an important area for further preclinical research will be the study of the effects of lamotrigine and other novel antipsychotic agents given chronically, preferably in models in which behavioural disruption has already been established, e.g. after chronic treatment with NMDA antagonists (e.g. Jentsch et al. 1997; Schwabe et al. 2005). Lamotrigine, given chronically, has been studied in models of relevance to epilepsy with mixed results (Postma et al. 2000; Stratton et al. 2003); other studies, reviewed below, have examined the effects of chronic lamotrigine on amino acid and monoamine levels in rodents (Hassel et al. 2001; Ahmad et al. 2004a,b) and humans (Kuzniecky et al. 2002).

Possible mechanisms of action of lamotrigine in schizophrenia

Sodium channel inhibition

Lamotrigine is thought to bind and stabilize the inactivated state of voltage-gated sodium channels (Kuo 1998; Xie et al. 1995; Liu et al. 2003), thus reducing the ability of a neuron to generate high-frequency trains of action potentials. The affinity of lamotrigine for the inactivated state of the human brain type II sodium channel, based on recombinant cell line studies, is estimated to be in the range 10–30 μM (Xie et al. 1995; Kuo 1998). This concentration range is similar to that in the brains of patients with epilepsy during treatment with lamotrigine (Walker et al. 2000). Use-dependent inhibition of brain sodium channels is thought to be central to the anticonvulsant efficacy of the drug. Similar results have been obtained with carbamazepine (Willow et al. 1995; Kuo 1998); furthermore, in a recent review, Ambrosio et al. (2002) concluded that inhibition of sodium channels is most likely responsible for the anticonvulsant efficacy of carbamazepine. However, other molecular interactions of carbamazepine such as inhibition of NMDA-mediated activity (Cunha et al. 2002 and see below), might contribute to poorer tolerability, limiting its apparent efficacy and use. Valproate, in contrast to lamotrigine and carbamazepine, has only weak effects on fast sodium currents at concentrations up to 1 mM (Johannessen 2000), although it has been shown to inhibit burst firing of neurons at therapeutically relevant concentrations (6–30 μM; McLean and Macdonald 1986). The mechanism in this latter case is suggested to involve inhibition of so-called persistent sodium currents (Taverna et al. 1998).

Lamotrigine does not distinguish between the different subtypes of brain sodium channel (Whitaker et al. 2001) and so will stabilize inactivation of each, thus regulating high-frequency action potential generation at the soma and propagation of these potentials along the axon, and perhaps affecting dendritic integration of high-frequency synaptic input. This is likely to be a highly effective mechanism to limit neuronal activity in some seizure disorders, but it is unclear how this mechanism might confer antipsychotic efficacy without understanding how neurons behave and interact during psychotic episodes in patients with schizophrenia. A reduction in burst firing of cortical neurons (Shi and Zhang 2003; Jackson et al. 2004) but an increase in overall firing (Jackson et al. 2004) was observed in rats after exposure to NMDA antagonists. In the latter study, the firing of individual neurons was less organized, and there was reduced synchronization between neurons. It will be important to investigate how lamotrigine might affect neuronal network activity in the prefrontal cortex in the presence of PCP or other psychotomimetic drugs.

Activation of many G-protein-coupled receptors can lead to phosphorylation of sodium channels, which then alters their function (Cantrell and Catterall 2001). More importantly, in the context of the role of dopamine in schizophrenia, activation of cAMP-dependent protein kinase A can phosphorylate sodium channel alpha subunits, leading to diminished channel conductance in hippocampal pyramidal neurons and striatal medium spiny neurons (reviewed by Cantrell and Catterall 2001). Activation of dopamine D2 receptors, which inhibit adenylate cyclase and so reduce PKA activation, has been shown to enhance sodium currents in medium spiny neurons projecting to the substantia nigra (Surmeier et al. 1992). Furthermore, dopamine D2 receptor stimulation may enhance the transition of neurons in the striatum from a hyperpolarised “downstate” to a depolarised “upstate” by influencing sodium channel function (Surmeier and Kitai 1997). Thus, the influence of lamotrigine on sodium channel function under these conditions will be an interesting subject for research. The influence of atypical antipsychotic drugs on sodium channel function also merits investigation. In the prefrontal cortex, olanzapine enhanced the firing and excitability of pyramidal neurons after 21 days of treatment (Gronier and Rasmussen 2003). If sodium channels are involved, then compounds like lamotrigine that interact with the channels in a use-dependent manner may be well placed to augment the therapeutic effect of the antipsychotic drug, as suggested by the emerging clinical data.

In conclusion, sodium channels are critical for neuronal activity and transmitter release. In this review, we argue that dysregulation of cortical neuronal activity and glutamate release is central to the effects of psychotomimetic drugs and may contribute to the pathology of schizophrenia. Thus, modulation of neuronal activity and glutamate release via sodium channel inhibition is likely to be central to the ability of lamotrigine and carbamazepine to prevent NMDA-antagonist-induced behavioural disruption in the PPI and reversal-learning models, and may be responsible for the add-on efficacy of these drugs in schizophrenia.

Calcium channel inhibition

Some in vitro effects of lamotrigine persist in the presence of the sodium channel blocker, tetrodotoxin (TTX). Cunningham and Jones (2000) found that inhibition of spontaneous excitatory post-synaptic potentials (EPSPs) in the entorhinal cortex persisted in the presence of tetrodotoxin. Indirect evidence suggests that lamotrigine may inhibit N-type voltage-activated calcium channels (Stefani et al. 1996; Wang et al. 1996a,b; Hainsworth et al. 2001), since the inhibitory effects of the drug on evoked EPSPs were occluded by the selective blocker, ω-conotoxin (Wang et al. 1996a,b). However, lamotrigine did not inhibit N-type calcium channels expressed in a recombinant cell line, which suggests that the drug does not interact directly with the channels (T. Dale, personal communication). In contrast, Hainsworth et al. (2003) found that lamotrigine (10 μM) could inhibit human R-type, but not T-type calcium channels in recombinant cell lines. This may be important to follow up since R-type channels are thought to be involved in the dendritic release of dopamine from neurons in the ventral tegmentum or substantia nigra (Bergquist and Nissbrandt 2003).

High concentrations of carbamazepine (up to 1 mM) have been shown to interact with presynaptic calcium channels (Zhu et al. 2002). Lower concentrations (1–50 μM) appeared to affect postsynaptic calcium entry, either through NMDA receptor channels (Hough et al. 1996) or through inhibition of L-type calcium channels (Ambrosio et al. 2002; but see also Lingamaneni and Hemmings 2003). Furthermore, in a study by Cunha et al. (2002), carbamazepine potently inhibited the NMDA component of the EPSP, an effect thought to be unrelated to its presynaptic inhibitory effects. To our knowledge, valproate has not been shown to interact with voltage-gated calcium channels in the central nervous system (Loscher 2002).

These data suggest that lamotrigine interaction with presynaptic voltage-gated calcium channels may allow control over spontaneous glutamate release and may differentiate the drug from carbamazepine and valproate. Whether this interaction confers adjunctive efficacy in schizophrenia remains to be proven. Calcium channel blockers, such as nifedipine or nilvadipine, which inhibit L-type calcium channels, have been trialed in patients with schizophrenia with mixed success (Suddath et al. 1991; Yamada et al. 1996). However, where positive effects have been seen, it is possible that enzyme induction leading to higher plasma levels of the co-administered antipsychotic might have played a part (Stedman et al. 1991). There is also some evidence that the L-type calcium channel inhibitor, nimodipine, can attenuate ketamine-induced psychotic symptoms in recovering alcoholics (Krupitsky et al. 2001). Investigation of drugs that specifically block N-type calcium channels, such as cilnidipine (Takahara et al. 2004) or ziconotide (Miljanich 2004), would be useful.

Other ion channels

A recent study by Poolos et al. (2002) found that lamotrigine (50–100 μM), but not carbamazepine (50 μM), had a greater inhibitory effect on dendritically evoked neuronal firing compared to firing evoked at the soma of hippocampal pyramidal neurons. They attributed the effect to activation of a hyperpolarisation-activated cation current (Ih), which is predominant in the dendrites of these neurons (Magee et al. 1998). Berger and Luscher (2004) also found an increase in Ih in cortical layer V pyramidal neurons after application of lamotrigine (50–100 μM), although in this case dendritic input to the soma was not attenuated; instead an increase in somatic excitability was observed in the presence of lamotrigine, which the authors attributed to depolarization mediated by the activation of Ih. These finding are intriguing and merit further study. However, close attention to therapeutically relevant concentrations of the sodium channel blockers will be important to assess the clinical relevance of the effects. Indeed, Berger and Luscher conclude that effects of lamotrigine on Ih are likely to be of less significance than effects on sodium channels at therapeutic drug concentrations. In addition, study of the effect of lamotrigine on recombinant hyperpolarisation-activated cyclic nucleotide-gated (HCN) ion channels, which are thought to contribute to Ih (Robinson and Siegelbaum 2003), would be useful.

There are also published data suggesting an interaction between lamotrigine and voltage-gated potassium channels. Grunze et al. (1998) and Zona et al. (2002) reported that lamotrigine (100–500 μM) could increase a transient potassium current in hippocampus and cortex, respectively, although effects were small and required high drug concentrations. In a recent study, lamotrigine, again at relatively high concentrations (100–300 μM), was shown to inhibit the A-type potassium current in a hippocampal cell line, H19-7 (Huang et al. 2004). Zona et al. (1990) previously showed that carbamazepine can also augment a transient potassium current in cortical neurons, although valproate does not (Zona and Avoli 1990). In contrast, Walden et al. (1993) reported an enhancement of late potassium currents by valproate in neurons of the snail. Studies using recombinant systems with mammalian channels will be required to investigate these interactions further, but given the high concentrations of lamotrigine and carbamazepine required to obtain effects, it seems unlikely that they contribute to the therapeutic efficacy of the drug.

Glutamate release

Lamotrigine has been shown to reduce glutamate transmission evoked by electrical stimulation in vitro. Wang et al. (1996a,b) showed that low concentrations of lamotrigine (10 μM) could inhibit EPSPs evoked in the basolateral amygdala. Calabresi et al. (1996, 1999) reported inhibition of EPSPs in cortex (IC50 ∼60 μM) and striatum (IC50 27 μM). However, higher concentrations of lamotrigine were required to produce robust inhibition in the hippocampus (32% inhibition with 100 μM; Langosch et al. 2000). In contrast, Xie et al. (1995) saw no effect of lamotrigine (50 μM) in hippocampal slices. Where inhibition of the EPSP has been observed, a presynaptic site of action was assumed, since lamotrigine had no effect on responses produced by direct application of glutamate receptor agonists. Furthermore, paired-pulse facilitation of the electrically evoked EPSP was increased by lamotrigine, a classic sign of presynaptic inhibition (Manabe et al. 1993; Sanchez-Prieto et al. 1996). The discrepancy between findings may be in part due to the different brain regions studied, but may also arise from differences in stimulation parameters. Waldmeier et al. (1995) studied [3H]glutamate release from cortical slices evoked either by electrical stimulation or the sodium channel opener, veratridine. The sodium channel blocker, tetrodotoxin, could inhibit both, but was at least tenfold less potent against electrical than veratridine-induced release. Similarly, lamotrigine was more potent at inhibiting glutamate release evoked by veratridine (IC50 ∼31 μM; Waldmeier et al. 1995) but was unable to inhibit release from cortical synaptosomes evoked by depolarization with potassium chloride (Lingamaneni and Hemmings 1999). In a recent in vivo study, Ahmad et al. (2004a) reported that lamotrigine, at therapeutically relevant concentrations, could inhibit an increase in extracellular glutamate induced by veratridine applied via a microdialysis probe to the hippocampus. The inhibitory effect of lamotrigine increased with chronic dosing (21 days), although this could be explained by increases in plasma and brain concentrations of the drug over the first few days of dosing. Perhaps of more relevance to schizophrenia, Idris et al. (2004) found that lamotrigine could prevent the increase in tissue levels of glutamate in the frontal cortex induced by acute treatment of rats with PCP. Thus, the method used to stimulate release can affect modulation by sodium channels blockers, and this may be relevant when considering the efficacy of sodium channel blockers in relation to the neuronal activity that might underlie symptoms of schizophrenia.

In addition to evoked glutamate release, lamotrigine has also been shown to inhibit spontaneous release through a mechanism that may not depend on sodium channels. Cunningham and Jones (2000) recorded spontaneous excitatory postsynaptic currents from neurons of the entorhinal cortex and showed that lamotrigine (50 μM) could inhibit the frequency of these events, even in the presence of tetrodotoxin. Von Wegerer et al. (1997) also examined spontaneous events in hippocampal slices bathed in a low-magnesium solution. Under these conditions, spontaneous, presumed calcium-mediated extracellular field potentials were recorded and shown to be inhibited by low concentrations of lamotrigine (minimum effective concentration, 2 μM). The relevance of spontaneous glutamate release in vitro to in vivo neuronal function is not known, although spontaneous excitatory transmission could be important in setting the signal-to-noise level in a circuit or in setting inhibitory tone through background activation of interneurons. Thus, the observation that lamotrigine may modulate spontaneous glutamatergic events merits further investigation.

Carbamazepine can also inhibit glutamate transmission in vitro via a presynaptic mechanism, although high concentrations are required. Cunha et al. (2002) found that the drug inhibited EPSPs and attenuated the presynaptic fibre volley in the hippocampus (EC50 263 μM). Carbamazepine has also been shown to inhibit glutamate release evoked by electrical stimulation of striatal slices (IC50 >100 μM; Waldmeier et al. 1995). In contrast to the two sodium channel blockers, sodium valproate has been shown to increase basal glutamate release in cortical slices (Dixon and Hokin 1997), although the drug concentration required (0.5–2 mM) was above that required for clinical efficacy. Chronic treatment with valproate has also been shown to up-regulate expression of the glutamate transporter, GLT-1 (Ueda and Willmore 2000), although this effect is unlikely to contribute to the acute effects of the drug in the animal models reviewed earlier.

Thus, lamotrigine and, to some extent, carbamazepine can inhibit glutamate transmission with potency dependent on stimulation parameters. Interaction with brain sodium channels is likely to contribute to inhibition of stimulated, although perhaps not spontaneous, glutamate transmission.

GABA release

Recent preclinical, post-mortem, and genetic data point to abnormalities in GABA neurotransmission, as well as glutamate, in schizophrenia (Benes 2000; Lewis 2000). The increase in glutamate release after treatment with PCP and ketamine is thought to be secondary to a reduction in tonic GABAergic inhibition of pyramidal neurons (Olney et al. 1999). Paulson et al. (2003) have shown that 18-day treatment of rats with MK-801 increases the cortical expression of the GABA transporter, GAT-3. Idris et al. (2004) found that acute treatment with PCP resulted in a significant decrease in tissue levels of GABA in the frontal and cingulate cortex, the same areas where increases in glutamate levels were observed. Therefore, enhancement of tonic GABA transmission might be an alternative mechanism by which drugs can prevent the psychotomimetic effects of NMDA antagonists. There is mixed evidence that lamotrigine can enhance GABA levels. Cunningham and Jones (2000) found that lamotrigine increased spontaneous GABA-mediated synaptic events in the entorhinal cortex. Whereas Braga et al. (2002) showed that lamotrigine could inhibit spontaneous inhibitory postsynaptic currents recorded from neurons of the basolateral amygdala (significant inhibition at 10 μM). Chronic treatment with lamotrigine increased GABA levels in the hippocampus (Hassel et al. 2001), although no change in GABA was noted in cortical areas. Four weeks of treatment with lamotrigine resulted in an increase in GABA in the occipital cortex of healthy human volunteers measured by magnetic resonance spectroscopy (Kuzniecky et al. 2002). In contrast to these studies of spontaneous GABA release or measurement of tissue levels, Ahmad et al. (2004a) found that lamotrigine could inhibit veratridine-induced GABA release in the hippocampus in vivo; they also showed that this effect persisted after 21 days of treatment with lamotrigine. Further studies should compare the temporal and region-specific changes induced by lamotrigine on both tonic and phasic GABA release.

Carbamazepine has been shown to inhibit GABA release evoked by veratridine or electrical stimulation (Waldmeier et al. 1995), a result that mirrors the drug's effect on glutamate release. To our knowledge, there are no reports of an enhancement of GABA release by carbamazepine, although there are some data to suggest that carbamazepine might enhance GABA-A mediated responses via interaction with an allosteric site on the receptor (Granger et al. 1995), and changes in GABA-B receptor density have been observed after treatment with both carbamazepine and valproate (Motohashi et al. 1989; Motohashi 1992). Effects of valproate on GABAergic transmission have been extensively explored (Loscher 1999; Johannessen 2000). Significant increases in whole-brain GABA levels have been observed in rats after valproate at 200 mg/kg (Loscher et al. 1985). The mechanism through which valproate enhances GABA levels remains to be confirmed, although inhibition of the enzyme succinate semialdehyde dehydrogenase is considered a likely candidate (Johannessen 2000).

Whatever the molecular mechanism by which valproate and lamotrigine enhance GABA levels, the lack of efficacy of valproate in either the NMDA models or in patients with schizophrenia would suggest that enhancing GABA is not sufficient to confer therapeutic benefit.

Monoamine release

Lamotrigine has been shown to interact with 5-HT, noradrenaline, and dopamine uptake transporters, but only at high, non-therapeutic concentrations (Table 4). In vitro studies also found an inhibitory action of lamotrigine on monoamine oxidase MAO-A and MAO-B activities in rat brain homogenates (E. Southam, personal communication). The inhibition was concentration dependent and reversible. K i values were determined as 15 μM (MAO-A) and 18 μM (MAO-B). Similar effects of lamotrigine on MAO-A activity were seen with human liver preparations. However, these in vitro observations were not matched by evidence of ex-vivo MAO inhibition or altered monoamine disposition in vivo. Thus, after the administration of lamotrigine to rats, there was no (MAO-A) or minimal (MAO-B) reduction in brain MAO activities when assayed ex vivo. Furthermore, in vivo brain microdialysis failed to detect alterations in extracellular hippocampal or frontal cortex monoamine concentrations (reviewed in Xie and Hagan 1998). In the same study, lamotrigine had no effect on tyramine-induced hypertension in rats or 5-hydroxytryptophan (5-HTP)-induced head shaking in mice, providing strong evidence that the drug does not perturb monoamine metabolism in vivo. The absence of observable effects of lamotrigine on monoamine disposition in vivo may be explained by the competitive and highly reversible nature of the interaction of lamotrigine with MAO isoforms, such that it is competed out by endogenous ligand.

Table 4 The pharmacology of lamotrigine
Full size table

In a recent study by Ahmad et al. (2004b), acute treatment with lamotrigine (10–20 mg/kg) decreased extracellular dopamine and 5-HT levels in dialysate from the ventral hippocampus of freely moving rats, but without affecting metabolite levels. Chronic treatment produced an increase in dopamine and 5-HT levels after 2 days, but these had returned to baseline by day 21. In conclusion, lamotrigine does not appear to have a direct effect on monoamine levels but could influence these indirectly through other mechanisms.

Interactions with antipsychotic drugs

The clinical data currently support the add-on efficacy of lamotrigine in combination with antipsychotic drugs. There is some very preliminary indication that efficacy may be apparent only in combination with clozapine (Dursun and Deakin 2001), although a larger study found equal efficacy with other atypical antipsychotic drugs (Kremer et al. 2004). Notwithstanding the possibility, based on the preclinical evidence, that lamotrigine may be an effective antipsychotic drug when given as monotherapy (discussed later), it is important to consider how the drug might interact with typical and atypical antipsychotic drugs. There are three scenarios: First, lamotrigine might enhance exposure to the antipsychotic medication. Second, lamotrigine may reduce some of the side effects associated with antipsychotic drugs, and so improve tolerability, compliance, and apparent efficacy of the antipsychotic. Finally, lamotrigine might augment the efficacy of the antipsychotic medication through a synergistic or additive mechanism.

Considering the first scenario, lamotrigine is not known to inhibit any of the cytochrome P-450 enzymes (GlaxoSmithKline, data on file) responsible for metabolism of the major atypical antipsychotic drugs. Thus, lamotrigine would not be predicted to increase exposure to any of these drugs. A formal drug interaction study in healthy subjects has been carried out with lamotrigine and olanzapine (Ascher et al. 2004). The study found that plasma levels of olanzapine at steady state were comparable when the drug was administered with lamotrigine or with placebo. In contrast, plasma levels of lamotrigine were lower when the drug was administered with olanzapine compared with placebo. There has been one case report of increased exposure to clozapine in the presence of lamotrigine (Kossen et al. 2001); however, no change was observed in a larger study carried out by Tiihonen et al. (2003). Although the effects of lamotrigine on the pharmacokinetics of clozapine was not a main objective of this clinical study, trough concentrations of clozapine obtained after 14 weeks of treatment with lamotrigine (200 mg for the last 4 weeks) do not indicate a significant drug interaction. Evaluation of antipsychotic exposures should be included in all future studies of the efficacy of add-on lamotrigine, but for now, pharmacokinetic interaction would not seem to provide an explanation for efficacy.

The second scenario, that lamotrigine might reduce the side-effect burden of the antipsychotic treatment, is harder to address with the limited clinical date available so far. An examination of the incidence of extrapyramidal side effects in patients given lamotrigine in addition to their antipsychotic drug may be worthwhile, since there are hints that lamotrigine has some benefit in the treatment of Parkinson's disease (Zipp et al. 1995), although significant improvements in motor function have not been observed (Zipp et al. 1995; Shinotoh et al. 1997). An open-label study in patients diagnosed with Alzheimer's disease showed some treatment improvement in depressed mood, word recognition, and naming with lamotrigine treatment (Tekin et al. 1998), and lamotrigine may have a beneficial effect on cognition in patients with epilepsy (Aldenkamp and Baker 2001). Furthermore, healthy volunteers on low doses of lamotrigine showed improvement on measures of cognitive activation and alertness (Aldenkamp et al. 2002).

In the final scenario, lamotrigine and antipsychotic drugs may act through the same biochemical mechanism or pathway. Dopamine is central to the mechanism of action of antipsychotic drugs, since antipsychotic efficacy closely correlates with antagonist activity at the dopamine D2 receptor (Creese et al. 1976). This also holds for the atypical antipsychotic drugs, and the correlation between D2 receptor occupancy and antipsychotic efficacy has been verified using non-invasive imaging (Nordstrom et al. 1993; Kapur et al. 1999), with an occupancy in excess of 65% necessary for efficacy (Seeman 2002). However, lamotrigine does not bind to dopamine receptors, nor does it affect the availability or release of dopamine in vivo (Table 4). It remains to be seen whether interaction with R-type calcium channels, observed in a recombinant cell line (Hainsworth et al. 2003), might confer some effect on dopamine neuron activity (Bergquist and Nissbrandt 2003). Furthermore, it would be useful to study the impact of use-dependent sodium channel inhibition on the firing of dopamine neurons after treatment with NMDA antagonists. Thus, there is reason to suspect that lamotrigine might interact directly with the dopamine system, although experimental data are lacking. Given that D2 receptor occupancy appears to be a critical parameter in the efficacy and side-effect potential of antipsychotic drugs, it might be valuable to examine this parameter in the presence and absence of lamotrigine.

Clozapine and other atypical antipsychotic drugs also have significant affinity for other receptors, notably dopamine D1, D4, noradrenergic alpha-1, 5-HT2a, 5-HT2c, and muscarinic M1, which may be responsible for increased efficacy (Coward 1992; Seeman 1992) or lower incidence of extrapyramidal symptoms (EPS) (Duncan et al. 1999; Seeman 2002). Lamotrigine does not bind to any of these receptors nor does it influence the availability or release of noradrenaline or 5-HT (Table 4).

Since we have argued that the efficacy of lamotrigine arises through inhibition of glutamate release, it is appropriate to consider whether antipsychotic drugs also affect glutamatergic transmission (reviewed in Heresco-Levy 2003). As mentioned earlier, most atypical antipsychotic drugs are antagonists of 5-HT2a receptors (Seeman 2002). Activation of these receptors within the cortex can enhance the release of glutamate (Aghajanian and Marek 1999). Unfortunately there is no direct evidence that the combination of lamotrigine with a selective 5-HT2a receptor antagonist produces an additive or synergistic inhibition of glutamate release, but it would be valuable to carry out such a study. Lamotrigine could in theory augment inhibition of glutamate release mediated by blockade of 5-HT2a receptors by atypical antipsychotic drugs. However, clozapine has actually been shown to increase levels of glutamate in the cortex (Daly and Moghaddam 1993; Yamamoto and Cooperman 1994). This observation is difficult to reconcile with the efficacy of lamotrigine in combination with Clozapine. Further study of the effects of these two drugs on glutamate release is warranted.

Clozapine may differentiate from other atypical antipsychotic drugs in other ways and perhaps should be considered separately from the other drugs in add-on trials with lamotrigine. Clozapine has been shown to enhance the excitatory response of cortical pyramidal neurons to application of NMDA (Arvanov et al. 1997). This enhancement may involve activation of dopamine D1 receptors, since it was absent in dopamine-depleted slices and could be blocked by the D1-selective antagonist, SCH23390 (Chen and Yang 2002; Ninan and Wang 2003). D1 receptor activation by other drugs has also been shown to promote NMDA receptor function (Morari et al. 1994; Dunah and Standaert 2001; Flores-Hernandez et al. 2002). Clozapine has significant affinity for dopamine D1 receptors, although it is still unclear whether the drug acts as an agonist or antagonist (Tauscher et al. 2004). A downstream consequence of D1 receptor activation leading to NMDA receptor facilitation appears to be inhibition of sodium and N-type calcium currents (Fienberg et al. 1998), so it is conceivable that lamotrigine might augment the impact of D1 receptor activation through such a mechanism. Reduced activation of cortical D1 receptors has been linked to negative and cognitive symptoms of schizophrenia (Davis et al. 1991; reviewed in Laruelle et al. 2003). Alternatively or in addition, clozapine may facilitate NMDA receptor activation through inhibition of type 1 glycine transporters (GlyT-1) (Javitt et al. 2004). However, this is unlikely to be a site of synergism between clozapine and lamotrigine, since the latter does not interact with this transporter (H. Herdon, personal communication). In contrast, these effects of clozapine on NMDA receptor function may be relevant to the worsening of symptoms when in combination with d-cycloserine (Goff et al. 1999; Heresco-Levy and Javitt 2004). Clinical data for clozapine and glycinergic agents suggest that negative symptoms of schizophrenia may be closely linked to NMDA receptor function; the absence of a compelling direct effect of lamotrigine on NMDA receptor function might be consistent with the absence of clinical effect of the drug on negative symptoms. However, this should be revisited as soon as further clinical data for lamotrigine are available.

Interactions between lamotrigine and antipsychotic drugs at a systems level

In the previous section we discussed how lamotrigine and antipsychotic drugs might interact within specific neurochemical systems. However, it is equally possible that the drugs work on different systems and their effects combine at a higher level to influence symptoms of schizophrenia. We can propose that lamotrigine regulates the firing of pyramidal neurons and the release of glutamate, whereas the antipsychotic drugs have their principal effect on dopamine transmission. The interaction between glutamate and dopamine in the schizophrenic brain has been a subject of significant discussion (Laruelle et al. 2003; Kapur 2003; Lisman and Otmakhova 2001). Dopamine is proposed to have a facilitatory rather than causal role in the generation of the symptoms of schizophrenia (Laruelle and Abi-Dargham 1999; Lisman and Otmakhova 2001; Kapur 2003). In support of this, d-amphetamine induces a form of psychosis in volunteers (Griffith et al. 1972; Angrist and Gershon 1970), but the symptoms are not considered to be fully similar to those of the schizophrenic patient. On the other hand, the drug can exacerbate positive symptoms in patients with schizophrenia (Lieberman et al. 1987). Furthermore, unlike NMDA receptor antagonists, treatment with d-amphetamine does not seem to induce negative symptoms, but may improve these symptoms in some patients (Angrist et al. 1982; van Kammen and Boronow 1988). Laruelle and Abi-Dargham (1999) and Kapur (2003) have described dopamine as the “wind of the psychotic fire”, increasing the salience of abnormal thoughts or hallucinations. Lisman and Otmakhova (2001) suggest that dopaminergic activity signals novelty, perhaps again increasing the impact of aberrant cortical activity on conscious thought. A consequence of this view is that psychosis is only temporarily subdued by dopamine-blocking drugs (Kapur 2003), which seems to be borne out by the frequent observation that identical symptoms remerge if the antipsychotic medication is withdrawn. Glutamate neuron dysregulation, on the other hand, may lie closer to the root cause of schizophrenia and may be responsible for the aberrant thought processes characteristic of the disorder. This view is supported by the acute psychotomimetic effects of ketamine in humans. The recent studies by Shi and Zhang (2003) or Jackson et al. (2004) show how altered glutamate transmission induced by PCP may be associated with disorganized firing of pyramidal neurons in the cortex. We postulate that this disorganization may mimic that induced by aberrant cortical connectivity in the schizophrenic brain. Thus, glutamatergic dysregulation (or disorganized neuronal activity) in combination with altered dopamine transmission may generate the recognizable symptoms of schizophrenia. The model may explain some of the observations from human volunteer studies with lamotrigine and ketamine. Lamotrigine, acting on the glutamate system, prevents the aberrant thoughts and hallucinations experienced by healthy subjects given ketamine (Anand et al. 2000a), but leaves intact or perhaps even enhances the feelings of euphoria that are likely associated with increased dopamine release (Anand et al. 2000b). In contrast, antipsychotic drugs, acting mainly on the dopaminergic component of ketamine's effect, produce only a modest reduction of the symptoms induced in human volunteers (Malhotra et al. 1997b; Lahti et al. 2001) or else are ineffective (Krystal et al. 1999b). Similarly, antipsychotic drugs are relatively ineffective against behavioural disruption induced by NMDA antagonists in preclinical species (Duncan et al. 2000; Abdul-Monim et al. 2003; Idris et al. 2005).

Although NMDA antagonists may recreate acutely the neurophysiological and neurochemical conditions necessary for the emergence of psychotic symptoms, the situation in schizophrenia is likely to be more complex. It will be important to study the efficacy of drugs like lamotrigine in monotherapy to determine whether an anti-glutamatergic treatment strategy would be effective alone. However, for the following reasons it is likely that efficacy would not be seen with lamotrigine monotherapy, but combination with anti-dopaminergic medications may be essential, at least during the active psychotic phase of the illness. Very briefly, schizophrenia is widely considered to be a developmental disorder that appears during adolescence (Keshavan and Hogarty 1999) and that is preceded by a prodromal phase (Yung and McGorry 1996). This phase is proposed to include or else cause dysregulation of cortical neural activity and glutamate transmission. For example, neurodegeneration may already have taken place by first onset of psychotic symptoms (Vita et al. 1997; Davies et al. 1998; Thompson et al. 2001; Cahn et al. 2002). Multiple lines of evidence suggest that alterations in glutamate transmission can impact the dopamine system. Cortical glutamatergic efferents exert significant control over the firing pattern of mesolimbic and mesocortical dopamine neurons (Tong et al. 1995; Overton and Clark 1997; Takahata and Moghaddam 2000). Furthermore, PCP has been shown to increase the firing of dopamine neurons in the ventral tegmental area (VTA) of anaesthetized rats (French 1994; reviewed by Svensson 2000). Takahata and Moghaddam (2003) also recently showed that the motoric effects of PCP in rats could be prevented by local inhibition of AMPA glutamate receptors in the VTA. Thus, the impact of glutamate on dopamine neuron activity may be sufficient to facilitate the emergence of psychotic symptoms (Grace 1991). Furthermore, since dopamine exerts significant control over glutamate transmission in cortical and limbic brain areas, a disrupted dopamine system may well cause further dysregulation of glutamate transmission and alterations in cortical function, creating a vicious cycle of glutamate–dopamine interactions (Laruelle et al. 2003) (Fig. 2).

Fig. 2
figure 2

A combination of dysregulation of glutamate and dopamine systems may contribute to the expression of psychosis and the symptoms of schizophrenia. Aberrant cortical connectivity is postulated to cause alterations of glutamatergic transmission; this in turn may lead to alterations in the dopamine system. A vicious cycle of altered glutamate and dopamine transmission drives further degenerative alterations in cortical connectivity. The principal site of action of lamotrigine is postulated to be on cortical neuronal activity and glutamate transmission; the drug may also influence the dopamine system directly, but there is currently no strong evidence for this. In contrast, atypical antipsychotic drugs are presumed to have their major effect through control of dopamine receptor activation; however, they may also influence the glutamate system through other receptor interactions

Full size image

Although speculative, this framework does allow predictions to made about how schizophrenia might be treated. First, as soon as a patient presents with symptoms of schizophrenia, treatment with anti-dopaminergic drugs will be necessary. However, since antipsychotic treatment does not address the underlying glutamatergic dysfunction, except perhaps through some mild modulatory effects mediated by D1 or 5-HT2a receptors, cessation of treatment would inevitably lead to the re-emergence of symptoms. Secondly, drugs designed to reduce glutamate release may not be sufficient to correct the glutamatergic dysregulation arising from the combination of aberrant cortical connectivity and secondary alterations of dopaminergic transmission. Reducing glutamate transmission may also not be sufficient to reset the dopamine system back to its normal state, given the complex homeostatic mechanisms involved. Consequently, the model suggests that both anti-dopaminergic and anti-glutamatergic drugs should be combined in the early treatment of schizophrenia. However, as soon as symptoms have been controlled, it may be possible to withdraw the patient from anti-dopaminergic agents and maintain only the glutamate-reducing drug, to control the underlying dysfunction and to provide prophylaxis against the re-emergence of a permissive dopaminergic state.

Conclusions

There is increasing evidence from clinical studies that lamotrigine can augment the efficacy of antipsychotic drugs in the treatment of schizophrenia. The efficacy of lamotrigine may be explained within a glutamate neuron dysregulation hypothesis of schizophrenia and may arise primarily through the drug's ability to influence glutamate transmission and neural activity in the cortex. However, clinical data so far contrasts with other glutamate-related approaches, such as activation of the glycine site on NMDA receptors, since the improvement with lamotrigine appears to be mainly in positive symptoms. At a molecular level, lamotrigine most likely acts through inhibition of voltage-dependent sodium channels, although other molecular interactions, e.g. with N-type or R-type calcium channels, cannot be ruled out. These effects of lamotrigine may add to or synergize with effects of some atypical antipsychotic drugs acting on glutamate transmission via inhibition or 5-HT2a receptors, activation of dopamine D1 receptors, or inhibition of glycine uptake. However, we also propose that lamotrigine (modulating glutamate transmission) and antipsychotic drugs (modulating dopamine transmission) may act at two different points on a cycle of pathophysiology that underlies the symptoms of schizophrenia (Fig. 2). Based on this model, we propose new strategies for the treatment of schizophrenia using a combination of anti-dopaminergic and anti-glutamatergic drugs.