Address correspondence and reprint requests to Pablo A. Gaspar, Clínica Psiquiátrica Universitaria, Hospital Clínico de la Universidad de Chile Av. La Paz 1003-Recoleta, Casilla 70014, Chile. E-mail: email@example.com
Early models for the etiology of schizophrenia focused on dopamine neurotransmission because of the powerful anti-psychotic action of dopamine antagonists. Nevertheless, recent evidence increasingly supports a primarily glutamatergic dysfunction in this condition, where dopaminergic disbalance is a secondary effect. A current model for the pathophysiology of schizophrenia involves a dysfunctional mechanism by which the NMDA receptor (NMDAR) hypofunction leads to a dysregulation of GABA fast- spiking interneurons, consequently disinhibiting pyramidal glutamatergic output and disturbing the signal-to-noise ratio. This mechanism might explain better than other models some cognitive deficits observed in this disease, as well as the dopaminergic alterations and therapeutic effect of anti-psychotics. Although the modulation of glutamate activity has, in principle, great therapeutic potential, a side effect of NMDAR overactivation is neurotoxicity, which accelerates neuropathological alterations in this illness. We propose that metabotropic glutamate receptors can have a modulatory effect over the NMDAR and regulate excitotoxity mechanisms. Therefore, in our view metabotropic glutamate receptors constitute a highly promising target for future drug treatment in this disease.
Schizophrenia is a common chronic mental disease with an unknown etiopathogenic framework. The dominant working model in schizophrenia postulates that genetic (Stefansson et al. 2008; Walsh et al. 2008) and environmental (Fatemi et al. 2000) neurodevelopmental disturbances may lead to dysfunctional neuronal migration, the disorganized cytoarchitecture of cortical layers and synaptic alterations that become associated with schizophrenia (Harrison 1999; Harrison and Weinberger 2005). The structural consequence of these disturbances is an altered brain connectivity, also called neuronal disconnectivity (Friston 1998; Stephan et al. 2006). Disconnectivity may be assessed at different structural and functional levels in the brain, from alterations in tract integrity to functional deficits in neuronal integration (Gaspar et al. 2009). Functional consequences of disconnectivity involve: (i) an altered timing of firing rate at the synaptic level; (ii) delayed coupling of synaptic neurotransmission at neurochemical pathways; and (iii) a loss of rhythm synchronization of brain oscillations at system levels involving brain communication. All these mechanisms have been described in this disease (Gaspar et al. 2009). However, it is important to mention that there is also evidence for an increased connectivity in some domains of the schizophrenic brain. For this reason, we have previously preferred the term ‘aberrant connectivity’ to describe the neuropathological condition underlying schizophrenia, where some functional domains (particularly those related to certain cognitive and perceptual processes) are impaired by disconnectivity, while others (for example, the so-called default network) display an abnormally increased connectivity (Gaspar et al. 2009; Whitfield-Gabrieli et al. 2009).
Although there seems to be general agreement about the disturbances of connectivity in the schizophrenic brain, there is an agitated controversy over the neurochemical mechanisms underlying the pathophysiology of this disease. Two main proposals have been raised in this context: the dopaminergic hypothesis and the glutamatergic hypothesis. In our view, recent evidence strongly points to the latter as the principal mechanism in this condition. In this article, we will review the evidence for the glutamatergic hypothesis of schizophrenia, and will discuss possible therapeutic strategies oriented at modulating glutamate activity via metabotropic glutamate receptors (mGluRs).
The glutamatergic hypothesis in schizophrenia
The dopaminergic hypothesis, implying a hyperactivity of the dopaminergic system, was the principal neurochemical hypothesis of schizophrenia until recently (Snyder et al. 1974; Carlsson et al. 2004). This hypothesis was mainly based on the fact that all typical anti-psychotics exert their effects principally by blocking D2 dopamine receptors (Seeman 2006) and that stimulants enhancing dopaminergic neurotransmission may induce some positive schizophrenia-like symptoms in normal human volunteers, and exacerbated psychotic symptoms in schizophrenic patients (Angrist and Sudilovsky 1978). However, direct support for this proposal has been elusive (Iversen and Iversen 2007).
On the other hand, evidence has accrued over the past years that points to a specific malfunction of glutamatergic receptors in the etiology of schizophrenia (Kim et al. 1980; Javitt 1987). Glutamate is the main excitatory neurotransmitter in the brain and is essential for sensorimotor and cognitive circuitry, which participates in development, synaptic plasticity, neuroprotection and glial-neuronal communication. Glutamate acts on two major classes of receptors: ionotropic glutamatergic receptors: AMPA, NMDA and Kainate, which are ligand-gated ion channels; and mGluRs (Masu et al. 1993). In general terms, the glutamatergic hypothesis of schizophrenia states that hypofunction of this neurotransmitter in cortico-striatal projections provokes a facilitation of thalamo-cortical circuits, producing an augmented sensory input, a decrease in the signal-to-noise ratio and an increase in dopaminergic input because of the disinhibition of the ventral tegmental area in the mesencephalon (Lang et al. 2007). Another possibility, although not strictly an alternative to that above, is that a malfunction of NMDA receptors (NMDARs) in GABAergic interneurons generates a generalized disinhibition in the cerebral cortex (Stahl 2007; Lisman et al. 2008). In fact, alterations in GABAergic interneurons, which receive strong inputs from glutamatergic neurons, are one of the most reproducible neuroanatomic alterations in schizophrenia (Lewis et al. 2005). In the next section, we will explore this second model in some detail.
Part of the evidence that links glutamate receptors (specifically NMDARs) and schizophrenia is that (i) the use of the NMDAR antagonists (MK-801, phencyclidine and ketamine) in rats and humans closely mimics the positive, negative and cognitive symptoms observed in schizophrenia, perhaps better than any other known drug (Krystal et al. 1994; Carlsson et al. 2004; Lisman et al. 2008), and worsens the positive symptoms in chronic- and non-medicated patients (Lahti et al. 1995; Medoff et al. 2001). Furthermore, subanesthetic doses of ketamine correlate with impaired performance on the Wisconsin Card Sorting Test, on spatial and verbal working memory tasks and on verbal declarative memory tasks (Krystal et al. 2000; Rowland et al. 2005), and produce alterations commonly observed in schizophrenics such as a decreased amplitude in event-related potentials (Umbricht et al. 2002). Besides, (ii) drug treatments targeting the glutamatergic receptors improve the clinical state of patients even more efficaciously than drugs that selectively target the monoaminergic system (Malhotra et al. 1997; Conn et al. 2009); and (iii) a strong decrease of the transcripts related to glutamatergic and GABAergic neurotransmission has been consistently observed in schizophrenia by using DNA microarray techniques (Mirnics et al. 2000; Frankle et al. 2003). Furthermore, (iv) functional neuroimaging studies provide evidence of the dysregulation of glutamatergic pathways in schizophrenic patients (van Elst et al. 2005; Gozzi et al. 2008). Finally, (v) at a systems level, neuronal synchronization that is thought to depend partly on glutamatergic regulation over GABAergic neurons is impaired in schizophrenia (Ford et al. 2007) and correlates with perceptual, speech and other cognitive processes (Spencer et al. 2003; Lewis et al. 2005; Mohler 2007; Gaspar et al. 2009). Therefore, the identification of the molecular alterations that generate this glutamatergic dysfunction is crucial for understanding the physiopathology of this disease and planning the future of drug treatment. Considering this evidence, several authors now agree that dopaminergic dysfunction may be secondary to an underlying glutamatergic dysfunction (Tuominen et al. 2005; Lisman et al. 2008).
Therefore, the study of glutamatergic regulation in schizophrenia, particularly of the glutamatergic receptors and their intracellular pathways, might yield important information on the physiopathology of this disease. Furthermore, in order to prevent and stop the progression of disconnectivity in this disease, it may be especially important to discover molecules that regulate glutamatergic activity in these networks.
NMDAR hypofunction and its consequences for schizophrenia
As mentioned, a current model for NMDA participation in schizophrenia implies that, in conditions of NMDAR dysfunction, the glutamatergic pyramidal neurons of the cerebral cortex lose the tonic regulation by fast-spiking GABAergic interneurons (the latter are normally stimulated by NMDAR), leading to a generalized increase of firing rate in pyramidal neurons. Since pyramidal glutamatergic and fast-spiking GABAergic interneurons have been proposed to represent a basic circuit of oscillatory synchronous activity (Bartos et al. 2007), this decrease in GABAergic function leads to a decrease in the functional connectivity or integrative capacity in the cortical networks. High-frequency synchronic activity correlates with several cognitive processes such as working memory and attention (Tallon-Baudry et al. 1998, 2001; Womelsdorf and Fries 2006). Working memory deficits have been considered an essential feature in the physiopathology of this disease and could be a clinical trait to be tested in genetic and neurochemical models in schizophrenia (Goldman-Rakic 1994; Green and Nuechterlein 1999). Such alterations have been consistently observed when assessing spatial and verbal domains in schizophrenics (Huguelet et al. 2000; Conklin et al. 2005), during crisis and compensated states (Park et al. 2002), in first-episode patients and patients with prodromal symptoms (Albus et al. 1996; Eastvold et al. 2007) and in high-risk populations for schizophrenia (Asarnow 1999; Erlenmeyer-Kimling 2000). Therefore, the study of working memory alterations may provide a window to analyze the cognitive deficits in schizophrenia (Fig. 1).
A second consequence of the loss of GABAergic modulation is a direct mechanism of neurotoxicity through non-NMDAR signaling (Lisman et al. 2008). All mechanisms of neurotoxicity described in schizophrenia involve the malfunction of the NMDAR. These mechanisms include (i) direct hypofunction of the NMDAR (Wang et al. 2000), (ii) mitochondrial dysfunction (Ben-Shachar 2002) and (iii) disinhibition of glutamatergic pyramidal neurons from GABAergic modulation (Lewis et al. 2005) (see Fig. 1). This proposal has received support from animal models displaying increasing glutamate levels after the long-term exposition of several antagonists of the NMDAR such as phencyclidine (Wang et al. 2000) and ketamine (Moghaddam et al. 1997). Furthermore, it has been observed that the chronic exposure of NMDAR antagonists induces neurotoxicity (Wang et al. 2001). These observations have led to the proposal of a mechanism of up-regulation of the NMDAR under these conditions. In this way, the direct treatment of schizophrenia based on NMDAR antagonists could trigger a secondary cascade of events, which could lead to toxicity mechanisms and provoke the opposite effect, accentuating the disconnectivity processes. So, the excitotoxicity associated with directly balancing glutamate levels through NMDAR agonist-antagonist modulation limits its therapeutic potential. Thus, other kinds of regulation should be studied to generate safer therapeutic approaches based on NMDAR regulation. With this idea in mind, we will analyze the downstream pathways of the NMDAR and how these may contribute to understanding NMDAR hypofunction in schizophrenia. Then, on the basis of growing evidence we will propose that mGluRs are a possible therapeutic target in schizophrenia.
NMDAR and its altered intracellular pathway in schizophrenia
The NMDARs participate as mediators of excitatory post-synaptic currents, a voltage-dependent mechanism critical for learning, working memory and attention (Coyle et al. 2003). The activity of these receptors is usually regulated by a voltage-dependent and magnesium-dependent mechanism. Furthermore, the expression of the NMDAR genes are modulated by the activity of the receptor itself, which allows calcium entry into the cell, triggering a cascade leading to gene activation. This regulation is the basis for the mechanism of synaptic long-term potentiation and associated morphological changes such as the growth of dendritic buttons that occurs during persistent stimulation of the pyramidal glutamatergic neurons (Lau and Zukin 2007). Many endogenous ligands of the NMDAR regulate the activity of this receptor. Especially important in this context is the glycine modulatory site present in this receptor because of its possible therapeutic implications for schizophrenia (Lane et al. 2008).
Considering that malfunction of the NMDAR could lead to brain excitotoxicity and several behavioral dysfunctions (positive and negative symptoms in schizophrenia), the normal function of the NMDAR must remain under tight cellular regulation. The NMDARs are located in the post-synaptic densities (PSD, cytoskeletal specializations that include the scaffolding protein complex and other signaling proteins). The PSD link the receptor to kinases, phosphatases and other intracellular proteins related to mGluRs, as will be described in the next section. Inside the PSD, the NMDAR is associated with scaffolding proteins such as the PSD protein of 95 KDa (PSD-95) and the synapse-associated protein of 102 KDa (SAP-102). These protein complexes are important in the intracellular trafficking and synaptic delivery of NMDAR (Scannevin and Huganir 2000). The number and subunit composition of the NMDARs are tightly regulated in response of neuronal activity and sensory experience (Lau and Zukin 2007). The NMDAR subunits NR1 and especially NR2 confer most of the biophysical and pharmacological properties to this receptor (Cull-Candy and Leszkiewicz 2004). One of the main mechanisms of regulation of the NMDAR is the balance of phosphorylation in the intracellular C-terminal domain of these subunits. A great number of phosphatases regulate the phosphorylation levels of the NMDAR, through non-receptor tyrosine kinases of the Scr family (Salter and Kalia 2004). While phosphorylation of the NR2b subunit of the NMDAR facilitates the suppression of clathrin-mediated endocytosis of these receptors, dephosphorylation of this subunit triggers NMDAR internalization. Controlling the phosphorylation levels of NMDAR signaling is an important mechanism of glutamatergic receptor-dependent synaptic plasticity (Lau and Zukin 2007).
As in many cellular types, the signal transduction pathway of the NMDAR in glutamatergic neurons depends on the activation of the mitogen-activated protein/NMDAR genes are pathway (Fig. 2). This signaling is finished by a downstream stop signal that involves the activation of the striatal-enriched tyrosine phosphatase that regulates extracellular signal-regulated kinases (ERK) signaling, a mechanism mediated by Ca2+ influx (Paul et al. 2003).
Mechanisms for the rapid internalization of the NMDAR might explain the hypofunction of this receptor in schizophrenia. As previously mentioned, the dephosphorylation of the NMDAR NR2 subunit could produce internalization of the NMDAR through a clathrin-dependent mechanism. It has been recently described that the binding of neuregulin-1 (NRG1) to the ERBB4 receptor produces dephosphorylation of the NR2A subunit, leading to an altered downstream signaling of the NMDAR (Hahn et al. 2006). NRG1 is one of the four proteins of the neuregulin family that act on the family of epidermal growth factor receptors. NRG1 induces proliferation, migration, differentiation and apoptosis in different cell types during neurodevelopment, and participates in synaptic plasticity (Buonanno and Fischbach 2001). Polymorphisms in NRG1 and its epidermal growth factor receptors genes are linked with susceptibility to schizophrenia (Mei and Xiong 2008) and could contribute to the structural and functional disconnectivity in this disease (Gaspar et al. 2009).
Another mechanism for the augmented internalization of the NMDAR in schizophrenia might involve the over-activation of phosphatases in the NMDAR downstream signaling. Serine/threonine Phosphatase PP2B, also known as calcineurin, is a neuron-enriched phosphatase that regulates synaptic plasticity and NMDAR neurotransmission. PP2B dephosphorylates and activates striatal-enriched tyrosine phosphatase, which induces dephosphorylation of the NR2B subunit, promoting internalization of the NMDAR (Braithwaite et al. 2006). Calcineurin knockout mice display an increased locomotor activity, decreased social interaction and impaired attention and working memory function (Zeng et al. 2001; Miyakawa et al. 2003). A potential schizophrenia susceptibility gene, the calcineurin γ catalytic subunit (PPP3CC), has been detected (Gerber et al. 2003). A significant association was reported between some haplotypes of PPP3CC with a Taiwanese sample of schizophrenic patients with deficits in sustained attention and executive processing (Liu et al. 2007). In spite of these promising results, another study failed to reproduce these findings (Kinoshita et al. 2005).
mGluRs: physiology and possible involvement in schizophrenia
The physiology and functional disturbances of mGluRs are relevant topics to schizophrenia since these molecules may have a direct etiopathogenic role on the disorder, but also because as mentioned they may represent useful therapeutical targets to mitigate glutamatergic dysfunction, therefore, alleviating the symptoms of this condition (Moghaddam 2004).
Ligand-gated ion channels (NMDA and AMPA receptors) are responsible for fast excitatory transmission, while mGluRs have a modulatory role (Cartmell and Schoepp 2000; Gasparini et al. 2008). mGluRs are subdivided into three classes according to pharmacological and cell signaling properties.
Group I mGluRs (mGluR1 and mGluR5) are expressed mainly at post-synaptic sites. They activate phospholipase C to generate diacylglycerol and inositol 1,4,5-triphosphate, therefore, increasing the release of calcium from endoplasmic reticulum, which results in protein kinase C activation. These receptors also potentiate L-type calcium channels and inhibit potassium channels, and may also activate other transducing cascades that trigger phosphorylation of ion channels, transcription factors and other target proteins. Through these mechanisms, group I mGluRs increase neuronal excitability and promote long- and short-term plasticity (Benarroch 2008).
Activation of group II mGluRs (mGluR2 and mGluR3) and group III mGluRs (mGluR4, mGluR6, mGluR7 and mGluR8) determines changes on adenylyl cyclase activity and, therefore, on the levels of cAM1P and activity of protein kinase A. Although this coupling seems region-specific, the primary action of group II and group III mGluRs is the decrease of cAMP and protein kinase A activity. (Kim et al. 2008). Aside from their specific mechanisms of action, all families of mGluRs converge on the activation of mitogen-activated protein kinases. This effect is also relevant for long-term plasticity (Kim et al. 2008).
Different glutamatergic pathways interact with each other. At the post-synaptic densities, mGluR5 is physically linked to the NMDAR via homer, shank and PSD-95 (Gray et al. 2009). Group I mGluRs induce the enhancement of NMDAR currents and are involved in the direct phosphorylation of the NMDAR (Pisani et al. 2001; Homayoun et al. 2004). However, other authors have found that activation of these receptors reduces nerve cell death caused by exposure to NMDAR agonists (neuroprotective effect), and could facilitate neurogenesis through a reduction of NMDA-stimulated currents (Baskys et al. 2005). On the other hand, activation of group II and group III mGluRs involved in the regulation of the release of glutamate and other neurotransmitters (Cartmell and Schoepp 2000) may generate neuroprotection. Such an effect has been evident in some neurotoxicity models (Vernon et al. 2008).
Metabotropic glutamate receptors also link the glutamatergic pathway with other neurotransmitter systems, as they modulate GABA and dopaminergic activity (David and Abraini 2002; Durand et al. 2008). Genetic studies suggest that mGluRs may be directly involved in the pathogenesis of schizophrenia. The strongest evidence points to the association of variants of the Glutamate receptor, metabotropic, 3 (GRM3) gene, which codes mGluR3, with the diagnosis of schizophrenia or with some cognitive features of it (Harrison and Weinberger 2005). Association has also been shown with a gene whose product is involved in metabotropic signaling regulation, the regulator of G protein signaling 4 (RGS 4) gene (Levitt et al. 2006). Association studies involving other mGluRs are scant and several of them have produced negative results (Bray et al. 2000; Bolonna et al. 2001; Takaki et al. 2004; Fallin et al. 2005; Ohtsuki et al. 2008). When gene expression has been studied, no changes in mRNA levels of mGluR3 have been found (Harrison et al. 2008), however, one relevant finding is that even when the total amount of the protein mGluR3 is preserved, its dimeric form was found to be decreased (Corti et al. 2007). Other studies have focused on protein expression in several brain regions and found abnormalities involving mGluR1a and mGluR2/3 in the prefrontal cortex of schizophrenic patients (Gupta et al. 2005).
Neuropathological and behavioral disturbances have been observed in knockout animals for different mGluR families (Linden et al. 2002; Brody et al. 2003, 2004; Cryan et al. 2003; Lyon et al. 2008; Gray et al. 2009). From this finding, one may speculate that these molecules are in fact involved in the pathogenesis of schizophrenia. Alternatively, these findings may be an indication that modifications affecting them have consequences that could be useful for clinical practice. We adhere to this second line of thought.
In accordance with this, pharmacological studies at the pre-clinical and clinical levels show that mGluR-modulatory drugs may be useful in alleviating the features of schizophrenia and other neuropsychiatric disorders, such as cognitive disturbances and anxiety (Gray et al. 2009; Kinney et al. 2003, 2005; Krystal et al. 2005; Pietraszek et al. 2005; Smialowska et al. 2007; Lavreysen and Dautzenberg 2008; Palucha-Poniewiera et al. 2008; Paz et al. 2008). Of particular interest is the study by Patil et al. (2007) that involved a group II mGluR agonist administered as the pro-drug LY2140023. Through a randomized, double blind, placebo controlled study, this molecule was shown to be successful in reducing positive and negative symptoms in schizophrenic patients while being well tolerated. In general terms, modulation of mGluRs may contribute to restore regulation of glutamatergic system through the enhancement of NMDAR activity (by enhancing mGluR5) or the reduction of excitatory glutamatergic transmission at key synapses in the prefrontal cortex (by enhancing mGluR2 and mGluR3). Highly selective positive allosteric modulators of these receptors may serve this purpose (Conn et al. 2009).
Neuroleptic treatment focused on monoaminergic targets has enabled the control of the most common symptoms observed in schizophrenics: hallucinations and delusions. Although these kinds of symptoms, also known as positive, are frequently associated with this illness, they are not the most specific. Cognitive dysfunctions, such as attention, working memory and executive functions, have been proposed as core features of this disease (Goldman-Rakic 1994; Green 1996). Cognitive deficits are present in first-episode patients, in a high-risk populations of schizophrenia (Erlenmeyer-Kimling 2000) and are among the best predictors of deficits in daily activities and the long-term functional outcome of patients (Addington et al. 1998; Dickinson and Coursey 2002). Drugs that modulate the glutamatergic system can have an effect on the stabilization of not only positive, but also negative symptoms and cognitive deficits observed in schizophrenia. As we discuss in this article, NMDAR may play a central physiopathological role in this disease.
Hypofunction of the NMDAR underlies the widely distributed domains of disconnectivity in schizophrenia. As commented in the introduction, the ‘disconnectivity’ in schizophrenia involves different domains depending on the level of structural and functional integration in the brain. On one side, we find poor connectivity looking for alterations in the white matter tracts of the postmortem (Crow 1998; Hoffman and McGlashan 1998) and in vivo schizophrenic brain (Hubl et al. 2004; Shergill et al. 2007; Whitford et al. 2007). On the other side, disconnectivity can be understood as an alteration of the temporal correlation of different groups of neurons, named ‘cell assemblies’, which are not directly connected to each other. This mechanism of connectivity has been called neuronal synchronization, and has been found to be strongly decreased in schizophrenia, at least during some cognitive tasks (Spencer et al. 2003; Ford and Mathalon 2008). One possible mechanism for impairing functional connectivity between different brain regions relates to alterations in the NMDAR located in the soma of fast-spiking GABA interneurons of the prefrontal cortex (Lewis et al. 2005). Alterations in this target could explain working memory disturbances through the disruption of neuronal oscillations and synchronization (Gaspar et al. 2009). On the other hand, the hypofunction of the NMDAR affecting GABAergic interneurons in the thalamus, hippocampus and prefrontal cortex would result in an increased glutamatergic, cholinergic and dopaminergic release in the cortex and other localizations (Olney and Farber 1995). These disbalances of neurochemical pathways could explain many of the wide ranging symptoms observed in schizophrenia. Note that as previously mentioned, disconnectivity may be relevant in some functional domains of the schizophrenic brain, but there is also evidence for hyperconnectivity in other domains. Thus, perhaps the term ‘aberrant connectivity’ better fits the general alterations of the schizophrenic brain (Gaspar et al. 2009).
Considering this evidence, many efforts have been made to develop drugs that target NMDAR, although the utility of these agents is limited because of adverse effects that manifest at the cellular and clinical levels. One of these limitations relates to excitotoxic mechanisms triggered by NMDAR stimulation pathways, affecting the GABAergic cortical and hippocampal interneurons (Stone et al. 2007) and related pyramidal glutamatergic neurons among other cell types.
As we have argued, mGluRs could become an important target of new drug treatment in this disease based on their regulatory role over the NMDAR and capacity to prevent excitotoxity mechanisms. Clinical trials are currently testing some NMDAR-glycine site agonists and glycine transporter inhibitors in schizophrenic patients, such as d-cycloserine and sarcosine (N-methylglycine) respectively (Buchanan et al. 2008; Lane et al. 2008). Biomolecules that target mGluRs could help as an adjunctive therapy of direct agonists of the NMDAR, and pre-clinical and clinical evidence has been accumulating in this direction. One of the most promising attempts is the administration of the pro-drug LY2140023, which has been tested on schizophrenic patients.
In conclusion, the aim of this work was to review the glutamatergic hypothesis in schizophrenia, emphasizing an update on the altered intracellular pathways of the NMDAR and mGluRs, and the possibility of using them as targets in the development of new drug treatments for cognitive deficits in schizophrenia. Exhaustive clinical trials will be needed to test this proposal.