4 February 2008. Dopaminergic, glutamatergic, GABAergic, cholinergic. Several hypotheses linking faulty neurotransmission to schizophrenia have emerged over the years. As is often the case with apparently conflicting theories, each may be correct in its own way. So how are they related? That question was addressed by a symposium at last November’s annual meeting of the Society for Neuroscience. As symposium chair John Lisman, Brandeis University, Waltham, Massachusetts, pointed out in his introduction, there is a need to come up with a model of neural networks that can explain not only how these various neurotransmitters tie in with the positive and negative symptoms of the disease, but how they relate to emerging schizophrenia risk genes.
Joseph Coyle, McLean Hospital, Belmont, Massachusetts, started the symposium by reviewing some of the links between glutamatergic hypofunction and novel risk genes for schizophrenia. The glutamatergic hypothesis of schizophrenia is well established (see SRF hypothesis) and stems from the observation that antagonists of the NMDA-type glutamate receptor, such as phencyclidine and ketamine (see SRF related news story), can induce psychosis, as well as some of the physiological signs found in schizophrenia patients, including eye tracking abnormalities (see Avila et al., 2002), frontal cortical hypofunction (see Hertzmann et al., 1990), elevated subcortical dopamine release (see Kegeles et al., 2000), impaired prepulse inhibition (see Braff et al., 2001), and mismatch negativity in auditory event-related potentials (ERPs; see Umbricht et al., 2002). But it is not clear how the glutamatergic hypothesis fits with patterns of inheritance. The explanation may lie not with the genes for NMDA receptors themselves, but in associated genes, suggested Coyle.
Coyle reviewed some of the evidence linking schizophrenia to genetic polymorphisms in genes that regulate NMDA co-agonists such as serine. Researchers reported in 2002 that genetic polymorphisms in two overlapping genes, G72 and G30, associated with schizophrenia (see Chumakov et al., 2002), and it turns out that G72 activates D-amino acid oxidase (DAO), which degrades serine. Thus, genetic links to schizophrenia may be within two degrees of separation from the NMDA receptor, noted Coyle. More recent studies support the link between G72/G30, DAO polymorphisms, and both schizophrenia and bipolar disorder (see Detera-Wadleigh and McMahon, 2006 and Corvin et al., 2007). Though polymorphisms in individual genes may not have large effects, a combination of genetic variations in multiple genes impacting NMDA function, including G72/G30, DAO, and serine racemase, which converts L-serine to the NMDA receptor relevant D-serine, could be significant, Coyle suggested.
Coyle has investigated the serine connection pharmacologically using D-cycloserine, an approved antibiotic which also happens to be a partial NMDA agonist. Though he cautioned that this drug also inhibits serine racemase and therefore might have a complex dose response curve, he has found that at intermediate doses the drug relieves some of the negative symptoms of schizophrenia (see Goff et al., 1995). Other researchers found that glycine, another NMDA co-agonist, also reduces negative symptoms of schizophrenia, Coyle said. His lab also found that there is enhanced activation of some regions of the brain when patients taking D-cycloserine perform a memory task (see Yurgelun-Todd et al., 2005). The data point to a role for NMDA co-agonists in disease pathology and potential treatments.
While several studies show NMDA co-agonists might improve schizophrenia symptoms, it is not clear which neurons in the brain mediate these effects. But a hint came from an earlier study from Robbie Greene’s lab, suggested Coyle. These researchers found that sensitivity to NMDA antagonists is different in pyramidal cells and GABAergic interneurons (see Grunze et al., 1996)—the latter are 10 times more sensitive to receptor antagonists, said Coyle. Coyle is now trying to model serine deficiency in mice to get a better handle on the neurons involved. In collaboration with colleagues at the Ottawa Health Research Institute, Canada, he has created a serine racemase knockout, which is predicted to limit the availability of D-serine. In a poster presented in San Diego, his colleague Chun Ma demonstrated that wild-type mice only get a mild boost in NMDA currents when perfused with serine, but serine racemase knockouts show a dramatic increase in current. In fact, in the knockout animal, long-term potentiation, a key facet of synaptic plasticity, cannot be elicited unless D-serine is applied. Behaviorally, these animals are less active than normal mice, though they seem to perform fairly well in a water maze test of learning and memory.
GABAergic dysfunction has also been linked to the pathology of schizophrenia and related diseases, including bipolar disorder. The GABAergic hypothesis dates back several decades, noted Francine Benes, also from McLean Hospital. In the 1970s it was reported that GABA and glutamate decarboxylase, which catalyzes a key step in GABA synthesis, are lower in brain tissue taken from schizophrenia patients. Those earlier studies were treated with some skepticism, however, because it was thought that the losses may be related to perimortem brain changes rather than the disease itself, Benes said. But further evidence, including decreased GABA uptake, loss of GABAergic interneurons, and compensatory increases in GABA receptors and GABA binding, helped to solidify the idea that GABAergic transmission is dysfunctional in schizophrenia. In addition, the GAD67 isoform of glutamate decarboxylase is reduced in the prefrontal cortex (see Guidotti et al., 2000), and both GAD67 and the smaller GAD65 isoform are reduced in the hippocampus of schizophrenia patients (see Heckers et al., 2002). Genetic polymorphisms in GAD genes have also been linked to schizophrenia (see SRF related news story).
How do GABAergic deficits fit in with other theories of schizophrenia? For one thing, GABAergic dysfunction may be related to loss of NMDA receptor-positive GABAergic cells, which may be a consequence of increased glutamatergic innervation, suggested Benes, thus tying together the glutamatergic and GABAergic components of the disease. Important clues to the role of GABA might be come from studying exactly which GABAergic neurons are affected in schizophrenia.
Benes postulated that amygdalar-hippocampal circuits might play a role in regulating hippocampal GABAergic neurons. To test this, she has developed an in vivo rat model where amygdalar innervation to the hippocampus is disrupted by an intra-amygdalar infusion of picrotoxin, an antagonist for the GABAA receptor. The picrotoxin treatment causes a loss of various GABAergic cells in the CA2 and CA3 region of the hippocampus, including parvalbumin-, calretinin-, and calbindin-positive cells, suggesting innervation by amygdalar neurons is crucial for the well-being of hippocampal interneurons. In addition, inhibitory postsynaptic potentials in the CA3 region of the hippocampus, but not in the CA1, are disrupted by amygdalar picrotoxin treatment (see SRF related news story).
While this picrotoxin treatment may serve as a model for GABAergic dysfunction in schizophrenia, what might underlie amygdalar-hippocampal disruptions in the disease proper? To address this, Benes has used laser dissection microscopy followed by microarray analysis to identify expression changes in the CA2/3 sector in brain samples from schizophrenia patients. Last June she reported that GAD67 expression in the CA2 and CA3 layers of the hippocampus are dramatically lower in both bipolar and schizophrenia patients. She has also found altered expression, either up- or downregulation, of other genes, which can be linked to protein networks that may impinge on GAD expression. GRIK2 and GRIK3, coding for kainate receptor subunits, are upregulated in schizophrenia patients, for example, while genes that may affect GAD more indirectly, such as histone deacetylase 1 and DAXX, are also upregulated in the hippocampus (see SRF related news story). The findings suggest that activity-driven changes in afferent inputs to GABA cells may contribute to unique patterns of gene expression in GABA cells at different points along complex circuits, concluded Benes.
The involvement of the amygdalar-hippocampal circuitry may also help explain the timing of schizophrenia onset, which occurs predominantly during adolescence. Benes showed that amygdalar fibers infiltrate the rat cingulate cortex gradually after birth, reaching maximal innervation during adolescence. At around the same time, these fibers begin to form contacts with GABAergic soma and dendrites in the anterior cingulate cortex. Thus, the ingrowth of amygdalar fibers during adolescence may induce complex molecular changes in GABAergic cells receiving direct inputs from this region. Should this input go awry, perhaps together with altered dopaminergic input, the risk for schizophrenia could be elevated, she concluded.
The involvement of dopamine was addressed by Arvid Carlsson from the University of Gothenburg, Sweden. By one measure, the hypothesis of a dopamine dysfunction in schizophrenia may be considered the most valid—except for the recent clinical trial success of a metabotropic glutamate agonist (see SRF related news story), all antipsychotic drugs so far developed to treat the disease are dopaminergic antagonists. Though that might suggest that pharmacological exploitation of the dopaminergic system is nigh exhausted, Carlsson ventured that as far as dopamine receptor pharmacology is concerned “serious R & D has hardly started yet.” He suggested that the development of new high-throughput screening (HTS) techniques was at least partly to blame for the dearth of new drugs. HTS disregards entirely that the same D2 dopaminergic receptor in one cell may do something entirely different in another cell, he said. Among the shortcomings of HTS, he asserted that the most common version measures the competition of the test molecule with another D2 receptor antagonist (radioactively labeled), based on the assumption that it can predict competition with the physiological agonist dopamine. Also, sometimes the labeled ligand is a D2 receptor agonist other than dopamine, "a fairly hazardous procedure," according to Carlsson.
Almost without exception, Carlsson said, the existence of allosteric binding sites is disregarded. Thus, he concluded, HTS casts too wide a web, and the most interesting drug candidates may well be lost. To illustrate this he described drugs that bind to allosteric sites of the GABAA receptor, such as benzodiazepines, barbiturates, ethanol, and steroids. A similar situation may well exist in the case of dopamine receptors, he noted.
Carlsson reviewed some of the properties of a D2 receptor antagonist called OSU6162. This has low affinity for the D2 receptor, yet it can stabilize behavior by reducing locomotor activity in animals given amphetamines or exposed to a novel stimulating environment, while increasing locomotion in animals that have been habituated to the environment. OSU6162 has a biphasic response curve, enhancing the effect of dopamine at low doses but inhibiting it at higher doses. The drug seems to have complex actions on both orthosteric and allosteric sites. This all speaks to the value of looking at allosteric effects, suggested Carlsson.
Could such dopaminergic modulators be of pharmacological value? Carlsson and colleagues have tested OSU6162 and ACR 16, a similar dopaminergic stabilizer, in schizophrenia patients and found that they reduce both positive and negative symptoms of the disease (in contrast, currently used antipsychotics only ameliorate the positive symptoms). These drugs have also shown promise in Phase 1 and 2 trials for Huntington’s disease, a broad spectrum neurodegenerative disorder with a variety of neurologic and psychiatric symptoms, including psychosis.
Lisman tied the glutamatergic, GABAergic, and dopaminergic components of schizophrenia together in his "halftime" talk. He suggested that the big picture is dominated by two loop circuits. The first loop is between the glutamatergic pyramidal cells of the hippocampus and the dopaminergic neurons they activate in the ventral tegmental area (VTA). Those dopaminergic neurons, in turn, feed back to the pyramidal cells. The second loop is between the pyramidal cells and the GABAergic inhibitory neurons in the hippocampus.
The idea that there is an interaction between the hippocampus, a major site for learning and memory in the brain, and the dopaminergic system came about by the discovery that novelty, such as simply opening a cage door, causes dopamine cells in the VTA to fire. It is not unexpected that the hippocampus would compute novelty, said Lisman, but the interesting thing is the relationship with the VTA. Dopamine release by the VTA has a strong effect in the transition between early and late long-term potentiation (LTP), he explained. Knocking out the D1 dopamine receptor, for example, blocks LTP in the CA1 region of the hippocampus. Lisman said he believes this is the circuit that is affected in schizophrenia, and he outlined growing evidence to support this contention (Lisman and Grace, 2005). The hippocampus becomes chronically activated in the disease, as evident by imaging analysis showing increased cerebral blood volume in the CA1 region of the hippocampus in schizophrenia patients (see Schobel et al., in press) Chemical stimulation of the ventral hippocampus is sufficient to activate dopaminergic neurons in the VTA (see Legault et al., 2000). And in an animal model of the disease, prenatal methylazoxymethanol acetate (MAM) administration, which leads to significantly greater number of spontaneously firing ventral tegmental dopaminergic neurons, disruption of hippocampal activity by tetrodotoxin reverses the elevated dopaminergic activity (see SRF related news story). Taking all the evidence together, the importance of the hippocampus-VTA loop is clear, Lisman said.
The second loop connects the glutamatergic and GABAergic components. What is the relationship between these two sets of neurons? asked Lisman. One key factor could be GAD67. Benes showed that GAD67 is reduced in schizophrenia, and it is well known that NMDA-type glutamate receptor activation stimulates interneurons and that NMDA antagonists lead to a reduction in GAD67. Why is that? Lisman asked. He ventured that there may be some homeostatic control over GAD67 synthesis. For example, he pointed out that Margarita Behrens’s group showed that NMDA antagonist-induced loss of GAD67 can be blocked if calcium levels are elevated (see Kinney et al., 2006). This may be in keeping with the role of fast spiking interneurons, which is to gather information and then process that information to stabilize firing. Calcium entry via NMDA activation may be a key sensor in this process, suggested Lisman. Blocking the NMDA receptor may lead the interneuron to “conclude” that firing is low and that inhibition needs to be relieved. Stopping synthesis of GABA by reducing GAD would be one way to re-balance the system.
One additional level of control over the pyramidal-interneuron loop may be via cholinergic innervation. Lorna Role, State University of New York, Stony Brook, reviewed some of the evidence linking nicotinic acetylcholine receptors with neuregulin, which has turned out to be one of the most studied genes in genetic association studies of schizophrenia. Role and colleagues discovered type 3 neuregulin when searching for regulators of nicotinic receptors. Type 3 neuregulin is essential for viability, since homozygous knockout mice die at birth. Type 3 neuregulin is found in the ventral hippocampus and also in the ventral subiculum, said Role. It is actively transported down axons and seems to colocalize with α7 nicotinic receptors. So what is it doing?
Clues come from heterozygous type 3 neuregulin knockouts. Though these animals do survive past birth, they have altered corticolimbic circuitry and altered behavior, including deficits in short-term memory and in gating sensory stimuli. “The phenotype is reminiscent of some schizophrenia endophenotypes,” said Role. The heterozygous mice perform poorly in a maze test of memory and in tests of prepulse inhibition. The latter is almost fully corrected by administration of nicotine (cigarette smoking is also very prevalent among schizophrenia patients).
What is the underlying cause of these behavioral deficits? Role showed that the animals have fewer parvalbumin-positive interneurons in the mouse equivalent of the human prefrontal cortex. In addition, pyramidal neurons do not seem quite normal. For example, the spines along the axons in the subiculum do not reach the same density as in wild-type mice, particularly in the middle of the axons, where spine levels are almost half wild-type levels. There are also functional consequences. The power of γ band oscillations is weaker than normal and gating-related LTP is lost. Also, unlike wild-type mice where nicotine can elicit sustained changes in the firing pattern in the ventral hippocampus and striatum, those changes are not sustained in type 3 neuregulin heterozygotes.
What are the molecular correlates of these functional changes? The bottom line is that neuregulin is important for targeting of α7-nicotinic receptors to enhance glutamate release in corticolimbic circuits, said Role. Decreased expression of neuregulin and/or α7 nicotinic receptors limits nicotine effects to transient facilitation, she said. In wild-type circuits nicotine effectively lowers the threshold for activity-dependent LTP, whereas in neuregulin heterozygotes LTP cannot be elicited even with tetanic stimulation.
Robert Freedman, from the University of Colorado, Denver, also addressed the involvement of nicotinic receptors in the pathology of schizophrenia. Freedman’s lab found that α-Bungarotoxin, a nicotinic receptor antagonist, affects sensory gating in animals, resulting in behaviors that resemble some endophenotypes found in schizophrenia patients. For example, patients respond almost as strongly to a second auditory stimulus as to one that briefly precedes it, while normal volunteers react much more weakly to the second stimulus. Inhibiting the nicotinic receptor in animals elevates the second sound-induced event-related potential (ERP) to rival the first, much like in patients.
Because the α7-nicotinic receptor gene (CHRNA7) has been cloned, Freedman and colleagues focused on that gene to tease out its relevance to sensory gating. They found that GABA and α-Bungarotoxin colocalize, showing that the nicotinic receptor is present in inhibitory neurons, and they reported that a loss of the receptor in CHRNA7 heterozygous null mice leads to an increase in the ratio of the second to first ERPs. But how do these findings relate to patients?
To test this, Freedman and colleagues have looked at genetic variation in the CHRNA7 gene. He reported that in a large family with two offspring that have schizophrenia, a common allele in the promoter of the gene is always associated with normal sensory gating, whereas a variant allele is associated with abnormal gating, even in individuals who did not have schizophrenia (see also Martin et al., 2007). In humans, there are also decreases in α-Bungarotoxin binding in postmortem brain tissue taken from schizophrenia patients, said Freedman, supporting the idea that nicotinic receptor levels may be linked to pathology.
These correlations between nicotinic receptor deficiency and schizophrenia-related phenotypes and behaviors may explain the prevalence of smoking among patients (by some estimates 90 percent of schizophrenia patients are smokers). But smoking is obviously not the best method of treatment, said Freedman. He has been involved in developing better nicotinic agonists. One of them, DMXB-A (3-[(2,4-dimethoxy)benzylidene]anabaseine), has been tried in a proof-of-concept trial, and Freedman reported that the drug improved scores in several measures including the RBANS (Repeatable Battery for the Assessment of Neuropsychological Status) and measures of auditory sensory gating. Patients also seemed to show improvement in scores of negative symptoms (SANS, Scale for Assessment of Negative Symptoms) and reported feeling better, said Freedman. In summary, there are a number of ways to get inhibitory neurons to function better, and α7-nicotinic receptor agonists may be one of them, he said.
Last but by no means least, Claudia Racca from Newcastle University, England, tied together the GABAergic and glutamatergic theories of schizophrenia. In this disorder there is both disruption of cortical γ oscillations (see Spencer et al., 2004) and reduction of parvalbumin-positive GABAergic interneurons, and the two are related because the γ oscillations depend on these fast-spiking interneurons. Racca is interested in modeling these losses in animals. She reported on two distinct models. One is an in vitro model of acute psychosis induced by ketamine, an NMDA receptor antagonist, while the other is a genetic/developmental animal model based on knocking out the LPA1 receptor gene. LPA1 is a receptor for lysophosphatidic acid, and though its biological role is not entirely clear, loss of the receptor leads to developmental problems that precipitate some phenotypes that are relevant to schizophrenia, including altered sensorimotor gating.
Racca’s findings have been covered in previous SRF news (see SRF related news story). Briefly, she showed that there is a drastic reduction in the power of γ oscillations in layer II of the entorhinal cortex in LPA1 knockout animals. In contrast, γ power in layers V-VI is unaltered. This reduction of γ power correlated with decrease in amplitude and frequency of GABA inhibitory postsynaptic potentials in stellate cells and their disinhibition, leading to increased spiking. This is accompanied by a specific loss of parvalbumin immunoreactivity in layer II of the entorhinal cortex. The reduction in γ power in layer II and the disinhibition of stellate cells are mimicked in the ketamine model. Other studies showed that chronic ketamine treatment also reduced parvalbumin immunoreactivity in neocortex (see Kinney et al., 2006 and SRF related news story). What seems to connect both these models is that GABAergic parvalbumin-positive interneurons in layer II of entorhinal cortex have a large NMDA receptor drive, noted Racca. The findings thus link glutamatergic systems and GABAergic systems and suggest that glutamatergic hypofunction may precipitate a disruption of γ oscillations, which is a proposed endophenotype of schizophrenia.
Lisman concluded the symposium by recounting the importance of the interaction between pyramidal cells and parvalbumin-positive interneurons. NMDAR hypofunction in the parvalbumin interneurons leads to disinhibition and reduced γ oscillations, perhaps contributing to the negative and cognitive symptoms of schizophrenia. The disinhibition in the hippocampus leads to hyperactivation of the hippocampal-VTA loop, leading to increased dopamine release and the positive symptoms of the disease. The beauty of having specific network models is that you can go at it with experimental tests, he suggested. In that vein, knocking out NMDA receptors on parvalbumin interneurons might be an excellent model for schizophrenia, Lisman suggested.—Tom Fagan.